Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Research ArticleAveraging Principles for Nonautonomous Two-Time-ScaleStochastic Reaction-Diffusion Equations with Jump
Yong Xu 12 and Ruifang Wang1
1Department of Applied Mathematics Northwestern Polytechnical University Xirsquoan 710072 China2MIIT Key Laboratory of Dynamics and Control of Complex Systems Northwestern Polytechnical UniversityXirsquoan 710072 China
Correspondence should be addressed to Yong Xu hsux3nwpueducn
Received 9 January 2020 Revised 7 July 2020 Accepted 14 July 2020 Published 19 September 2020
Academic Editor Eulalia Martınez
Copyright copy 2020 Yong Xu and Ruifang Wang is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
In this paper we aim to develop the averaging principle for a slow-fast system of stochastic reaction-diffusion equations driven byPoisson randommeasures e coefficients of the equation are assumed to be functions of time and some of them are periodic oralmost periodic erefore the Poisson term needs to be processed and a new averaged equation needs to be given For thisreason the existence of time-dependent evolution family of measures associated with the fast equation is studied and proved thatit is almost periodic Next according to the characteristics of almost periodic functions the averaged coefficient is defined by theevolution family of measures and the averaged equation is given Finally the validity of the averaging principle is verified by usingthe Khasminskii method
1 Introduction
e slow-fast systems are widely encountered in biologyecology and other application areas In this paper we are
concerned with the following nonautonomous slow-fastsystems of stochastic partial differential equations (SPDEs)on a bounded domain O of Rd(dge 1)
zuϵzt
(t ξ) A1(t)uϵ(t ξ) + b1 t ξ uϵ(t ξ) vϵ(t ξ)( 1113857 + f1 t ξ uϵ(t ξ)( 1113857zωQ1
zt(t ξ) + 1113946
Zg1 t ξ uϵ(t ξ) z( 1113857
z 1113957N1
zt(t ξ dz)
zvϵzt
(t ξ) 1ϵ
A2(t) minus α( 1113857vϵ(t ξ) + b2 t ξ uϵ(t ξ) vϵ(t ξ)( 11138571113858 1113859 +1ϵ
radic f2 t ξ uϵ(t ξ) vϵ(t ξ)( 1113857zωQ2
zt(t ξ) + 1113946
Zg2 t ξ uϵ(t ξ) vϵ(t ξ) z( 1113857
z 1113957Nϵ2
zt(t ξ dz)
uϵ(0 ξ) x(ξ)
vϵ(0 ξ) y(ξ)
ξ isin O
N1uϵ(t ξ) N2vϵ(t ξ) 0 tge 0 ξ isin zO
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(1)
HindawiComplexityVolume 2020 Article ID 9864352 22 pageshttpsdoiorg10115520209864352
where ϵ≪ 1 is a positive parameter and α is a sufficientlylarge fixed constante operatorsN1 andN2 are boundaryoperatorse stochastic perturbations ωQ1 ωQ2 and 1113957N1
1113957Nϵ2
are mutually independent Wiener processes and Poissonrandom measures on the same complete stochastic basis(ΩF Ft1113864 1113865tge0P) which will be described in Section 2 Fori 1 2 the operator Ai(t) and the functions bi fi andgi
depend on time and we assume that the operator A2(t) isperiodic and the functions b1 b2 f2 andg2 are almostperiodic
e goal of this paper is to establish an effective ap-proximation for the slow equation of original system (1) byusing the averaging principle e averaged equation isobtained as follows
zu
zt(t ξ) A1(t)u(t ξ) + B1(u(t))(ξ) + f1(t ξ u(t ξ))
zωQ1
zt(t ξ) + 1113946
Zg1(t ξ u(t ξ) z)
z 1113957N1
zt(t ξ dz)
u(0 ξ) x(ξ) ξ isin O
N1u(t ξ) 0 tge 0 ξ isin zO
⎧⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
(2)
where B1 is the averaged coefficient which will be given inequation (6) To demonstrate the validity of the averagingprinciple we prove that for any Tgt 0 and ηgt 0 it yields
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
L2(O)gt η1113888 1113889 0 (3)
where u is the solution of the averaged equation (2)e theory of the averaging principle originated by Laplace
and Lagrange has been applied in celestial mechanics oscil-lation theory radiophysics and other fieldse firstly rigorousresults for the deterministic case were given by Bogolyubov andMitropolskii [1] Moreover Volosov [2] and Besjes [3] alsopromoted the development of the averaging principle engreat interest has appeared in its application to dynamicalsystems under random perturbations An important contri-bution was that in 1968 Khasminskii [4] originally proposedthe averaging principle for stochastic differential equations(SDEs) driven by Brownian motion Since then the averagingprinciple has been an active area of research Many studies onthe averaging principle of SDEs have been presented eg Givon[5] Freidlin andWentzell [6] Duan [7]ompson [8] andXuand his coworkers [9ndash11] Recently effective approximationsfor slow-fast SPDEs have been received extensive attentionCerrai [12 13] investigated the validity of the averagingprinciple for a class of stochastic reaction-diffusion equationswith multiplicative noise In addition Wang and Roberts [14]Pei and Xu [15ndash17] and Li [18] and Gao [19ndash21] also studiedthe averaging principles for the slow-fast SPDEs
e abovementioned works mainly considered autono-mous systems For the autonomous systems as long as theinitial value is given the solution of which only depends on the
duration of time not on the selection of the initial timeHowever if the initial time is different the solution of non-autonomous equations with the same initial data will also bedifferent erefore compared with the autonomous systemsthe dynamic behaviors of the nonautonomous systems aremore complex which can portray more actual models Che-pyzhov and Vishik [22] studied long-time dynamic behaviorsof nonautonomous dissipative system Carvalho [23] dealt withthe theory of attractors for the nonautonomous dynamicalsystems Bunder and Roberts [24] considered the discretemodelling of nonautonomous PDEs
In 2017 the averaging principle has been presented fornonautonomous slow-fast system of stochastic reaction-diffusion equations by Cerrai [25] But the system of thispaper was driven by Gaussian noises which is considered asan ideal noise source and only can simulate fluctuations nearthe mean value Actually due to the complexity of the ex-ternal environment random noise sources encountered inpractical fields usually exhibit non-Gaussian propertieswhich may cause sharp fluctuations It should be pointed outthat Poisson noise one of the most ubiquitous noise sourcesin many fields [26ndash28] can provide an accurate mathe-matical model to describe discontinuous random processessome large moves and unpredictable events [29ndash31] So inthis paper we are devoted to developing the averagingprinciple for nonautonomous systems of reaction-diffusionequations driven by Wiener processes and Poisson randommeasures
e key to using the averaging principle to analyzesystem (1) is the fast equation with a frozen slow componentx isin L2(O)
zvxy
zt(t ξ) A2(t) minus α( 1113857v
xy(t ξ) + b2 t ξ x(ξ) v
xy(t ξ)( 11138571113858 1113859 + f2 t ξ x(ξ) v
xy(t ξ)( 1113857
zωQ2
zt(t ξ) + 1113946
Zg2 t ξ x(ξ) v
xy(t ξ) z( 1113857
z 1113957Nϵ2
zt(t ξ dz)
vxy(s ξ) y(ξ) ξ isin O
N2vxy(t ξ) 0 tge s ξ isin zO
⎧⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
(4)
2 Complexity
By dealing with the Poisson terms we prove that anevolution family of measures (μx
t t isin R) on L2(O) for fastequation (4) exists en assuming thatA2(t) is periodic andb2 f2 andg2 are almost periodic we prove that the evolutionfamily ofmeasures is almost periodicMoreover with the aid ofeorem 210 in [32] we prove that the family of functions
t isin R⟼1113946L2(O)
B1(t x y)μxt (dy)1113896 1113897 (5)
is uniformly almost periodic for any x in a compact setK sub L2(O) where B1(t x y)(ξ) b1(t ξ x(ξ) y(ξ)) forany x y isin L2(O) and ξ isin O
According to the characteristics of almost periodicfunction ([25] eorem 34) we define the averaged co-efficient B1 as follows
B1(x) ≔ limT⟶infin
1T
1113946T
01113946
L2(O)B1(t x y)μx
t (dy)dt x isin L2(O)
(6)
Finally the averaged equation is obtained through theaveraged coefficient B1 Using the classical Khasminskiimethod to the present situation the averaging principle iseffective
e abovementioned notations will be given in Section 2and in this paper cgt 0 below with or without subscripts willrepresent a universal constant whose value may vary indifferent occasions
2 Notations Assumptions and Preliminaries
Let O be a bounded domain of Rd(dge 1) having a smoothboundary In this paper we denote H the separable Hilbertspace L2(O) endowed with the usual scalar product
langx yrangH 1113946O
x(ξ)y(ξ)dξ (7)
and with the corresponding norm middotH e norm in Linfin(O)
will be denoted by middotinfinFurthermore the subspace D((minus A)θ) [33ndash35] of the
generator A is dense in H and endowed with the norm
Λθ (minus A)θΛ
H Λ isin D (minus A)
θ1113872 1113873 (8)
for 0le θlt 1 0lt tleT According to eorem 613 in [36]there exists cθ gt 0 such that
(minus A)θe
At
Hle cθt
minus θ (9)
We denote by Bb(H) the Banach space of the boundedBorel functions φ H⟶ R endowed with the sup-norm
φ0 ≔ supxisinH
|φ(x)| (10)
and Cb(H) is the subspace of the uniformly continuousmappings Moreover D([s T]H) denotes the space of allcadlag path from [s T] into H
We shall denote that L(H) is the space of the boundedlinear operators in H and denote L2(H) the subspace ofHilbertndashSchmidt operators endowed with the norm
Q2 Tr QlowastQ[ ]
1113968 (11)
In slow-fast system (1) the Gaussian noises zωQ1 zt(t ξ)
and zωQ2 zt(t ξ) are assumed to be white in time andcolored in space in the case of space dimension dgt 1 fortge 0 and ξ isin O And ωQi (t ξ)(i 1 2) is the cylindricalWiener processes defined as
ωQi (t ξ) 1113944infin
k1Qiek(ξ)βk(t) i 1 2 (12)
where ek1113864 1113865kisinN is a complete orthonormal basis in Hβk(t)1113864 1113865kisinN is a sequence of mutually independent standardBrownian motion defined on the same complete stochasticbasis (ΩF Ft1113864 1113865tge0P) and Qi is a bounded linear operatoron H
Next we give the definitions of Poisson random mea-sures 1113957N1(dt dz) and 1113957N
ϵ2(dt dz) Let (ZB(Z)) be a given
measurable space and v(dz) be a σ-finite measure on itDpi
t i 1 2 represents two countable subsets of R+
Moreover let p1t t isin Dp1
tbe a stationaryFt-adapted Poisson
point process on Z with the characteristic v and p2t t isin Dp2
t
be the other stationaryFt-adapted Poisson point process onZ with the characteristic vϵ Denote by Ni(dt dz) i 1 2the Poisson counting measure associated with pi
t ie
Ni(tΛ) ≔ 1113944sisinD
pitslet
IΛ pit1113872 1113873 i 1 2
(13)
Let us denote the two independent compensated Poissonmeasures
1113957N1(dt dz) ≔ N1(dt dz) minus v1(dz)dt
1113957Nϵ2(dt dz) ≔ N2(dt dz) minus
1ϵv2(dz)dt
(14)
where v1(dz)dt and 1ϵv2(dz)dt are the compensatorsRefer to [28 37] for a more detailed description of the
stochastic integral with respect to a cylindrical Wienerprocess and Poisson random measure
For any t isin R the operators A1(t) and A2(t) are sec-ond-order uniformly elliptic operators having continuouscoefficients on O e operators N1 and N2 are theboundary operators which can be either the identity op-erator (Dirichlet boundary condition) or a first-order op-erator (coefficients satisfying a uniform nontangentialitycondition) We shall assume that the operatorAi(t) has thefollowing form
Ai(t) ci(t)Ai + Li(t) t isin R i 1 2 (15)
where Ai is a second order uniformly elliptic operator[38 39] with continuous coefficients on O which is inde-pendent of t In addition Li(t) is a first-order differentialoperator which has the form
Li(t ξ)u(ξ) langli(t ξ)nablau(ξ)rangRd t isin R ξ isin O (16)
e realizations of the differential operators Ai and Li
in H are Ai and Li Moreover A1 and A2 generate twoanalytic semigroups etA1 and etA2 respectively
Complexity 3
Now we give the following assumptions
(A1)
(a) For i 1 2 the function ci R⟶ R is con-tinuous and there exist c0 cgt 0 such that
c0 le ci(t)le c t isin R (17)
(b) For i 1 2 the function li R times O⟶ Rd iscontinuous and bounded
(A2) For i 1 2 there exists a complete orthonormalsystem eik1113966 1113967
kisinN in H and two sequences of nonneg-ative real numbers αik1113966 1113967
kisinN and λik1113966 1113967kisinN such that
Aieik minus αikeik
Qieik λikeik kge 1
κi ≔ 1113944infin
k1λρi
ik eik
2infin ltinfin
ζ i ≔ 1113944infin
k1αminus βi
ik eik
2infin ltinfin
(18)
for some constants ρi isin (2 +infin] and βi isin (0 +infin)
such that
βi ρi minus 2( 11138571113858 1113859
ρi
lt 1 (19)
Remark 1 For more comments and examples about theassumption (A2) of the operators Ai and Qi the reader canread [12]
(A3) e mappings b1 R times O times R2⟶ R f1 R times O times
R⟶ R g1 R times O times R times Z⟶ R are measur-able and the mappings b1(t ξ middot) R2⟶ R
f1(t ξ middot) R⟶ R g1(t ξ middot z) R⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z More-over for all pge 1 there exist positive constantsc1 and c2 such that for all x1 x2 isin R we have
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c1 1 + x11113868111386811138681113868
1113868111386811138681113868p
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 1113857 minus g1 t ξ x2 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c2 x1 minus x21113868111386811138681113868
1113868111386811138681113868p
(20)
(A4) e mappings b2 R times O times R2⟶ R f2 R times O times
R2⟶ R g2 R times O times R2 times Z⟶ R are mea-surable and the mappings b2(t ξ middot) R2⟶ R
f2(t ξ middot) R2⟶ R g2(t ξ middot z) R2⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z
Moreover for all qge 1 there exist positive con-stants c3 and c4 such that for all(xi yi) isin R2 i 1 2 we have
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
1113868111386811138681113868qυ2(dz)
le c3 1 + x11113868111386811138681113868
1113868111386811138681113868q
+ y11113868111386811138681113868
1113868111386811138681113868q
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
minus g2 t ξ x2 y2 z( 1113857|qυ2(dz)
le c4 x1 minus x21113868111386811138681113868
1113868111386811138681113868q
+ y1 minus y21113868111386811138681113868
1113868111386811138681113868q11138731113872 (21)
Remark 2 For any (t ξ) isin R times O and x y h isin H z isin Z weshall set
B1(t x y)(ξ) ≔ b1(t ξ x(ξ) y(ξ))
B2(t x y)(ξ) ≔ b2(t ξ x(ξ) y(ξ))
F1(t x)h1113858 1113859(ξ) ≔ f1(t ξ x(ξ))h(ξ)
F2(t x y)h1113858 1113859(ξ) ≔ f2(t ξ x(ξ) y(ξ))h(ξ)
G1(t x z)h1113858 1113859(ξ) ≔ g1(t ξ x(ξ) z)h(ξ)
G2(t x y z)h1113858 1113859(ξ) ≔ g2(t ξ x(ξ) y(ξ) z)h(ξ)
(22)
Due to (A3) and (A4) for any fixed (t z) isin (RZ) themappings
B1(t middot) H times H⟶ H B2(t middot) H times H⟶ H
F1(t middot) H⟶L(H) F2(t middot) H times H⟶L(H)
G1(t middot z) H⟶L(H) G2(t middot z) H times H⟶L(H)
(23)
are Lipschitz continuous and have linear growth conditionsNow for i 1 2 we define
ci(t s) ≔ 1113946t
sci(r)dr slt t (24)
and for any ϵgt 0 and βge 0 set
Uβϵi(t s) e1ϵci(ts)Ai minus βϵ(tminus s)
slt t (25)
For ϵ 1 we write Uβi(t s) and for ϵ 1 and β 0 wewrite Ui(t s)
Next for any ϵgt 0 βge 0 and for anyu isin C([s t] W
1p0 (O)) we define
ψβϵi(u s)(r) 1ϵ
1113946r
sUβϵi(r ρ)Li(ρ)u(ρ)dρ slt rlt t
(26)
In the case ϵ 1 we write ψβi(u s)(r) and in the caseϵ 1 and β 0 we write ψi(u s)(r)
Remark 3 By using the same argument as Chapter 5 of thebook [40] we can get that there exists a unique evolution
system U(t s) eA 1113938
t
sc(r)dr
(0le sle tleT) for c(t)A More-over we also can get that ψβϵi(u s)(t) is the solution of
4 Complexity
du(t) 1ϵ
Ai(t) minus β( 1113857u(t)dt tgt s u(s) 0 (27)
3 A Priori Bounds for the Solution
With all notations introduced above system (1) can berewritten in the following abstract form
duϵ(t) A1(t)uϵ(t) + B1 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+ F1 t uϵ(t)( 1113857dωQ1(t)
+ 1113946Z
G1 t uϵ(t) z( 1113857 1113957N1(dt dz)
dvϵ(t) 1ϵ
A2(t) minus α( 1113857vϵ(t) + B2 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ(t) vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ(t) vϵ(t) z( 1113857 1113957Nϵ2(dt dz)
uϵ(0) x vϵ(0) y
(28)
According to Remark 2 we know that the coefficients ofsystem (28) satisfy global Lipschitz and linear growthconditions and the assumptions (A1)ndash(A4) are uniformwith respect to t isin R So using the same argument as[28 41 42] it is easy to prove that for any ϵgt 0 Tgt 0 andx y isin H there exists a unique adapted mild solution (uϵ vϵ)
to system (28) in Lp(Ω D([0 T]H times H)) is means thatthere exist two unique adapted processes uϵ and vϵ inLp(Ω D([0 T]H)) such that
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
vϵ(t) Uαϵ2(t 0)y + ψαϵ2 vϵ 0( 1113857(t)
+1ϵ
1113946t
0Uαϵ2(t r)B2 r uϵ(r) vϵ(r)( 1113857dr
+1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
+ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(29)
Lemma 1 Under (A1)ndash(A4) for any pge 1 and Tgt 0 thereexists a positive constant cpT such that for any x y isin H andϵ isin (0 1] we have
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 (30)
1113946T
0E vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (31)
Proof For fixed ϵ isin (0 1] and x y isin H for any t isin [0 T]we denote
Γ1ϵ(t) ≔ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
Ψ1ϵ(t) ≔ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(32)
Setting Λ1ϵ(t) ≔ uϵ(t) minus Γ1ϵ(t) minus Ψ1ϵ(t) we have
ddtΛ1ϵ(t) c1(t)A1Λ1ϵ(t) + L1(t) Λ1ϵ(t) + Γ1ϵ(t)1113872
+ Ψ1ϵ(t)1113873B1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
Λ1ϵ(0) x
(33)
For any pge 2 because B1(middot) is Lipschitz continuoususing Youngrsquos inequality we have1p
ddtΛ1ϵ(t)
p
Hlangc1(t)A1Λ1ϵ(t)Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langL1(t) Λ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
le c Λ1ϵ(t)
p
H+ c Γ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c Ψ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873
HΛ1ϵ(t)
pminus 1H
le cp Λ1ϵ(t)
p
H+ cp 1 + Γ1ϵ(t)
p
H+ Ψ1ϵ(t)
p
H+ vϵ(t)
p
H1113872 1113873
(34)
is implies that
Λ1ϵ(t)
p
Hle e
cptx
p
H + cp 1113946t
0e
cp(tminus r) 1 + Γ1ϵ(r)
p
H1113872
+ Ψ1ϵ(r)
p
H+ vϵ(r)
p
H1113873dr
(35)
Complexity 5
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
where ϵ≪ 1 is a positive parameter and α is a sufficientlylarge fixed constante operatorsN1 andN2 are boundaryoperatorse stochastic perturbations ωQ1 ωQ2 and 1113957N1
1113957Nϵ2
are mutually independent Wiener processes and Poissonrandom measures on the same complete stochastic basis(ΩF Ft1113864 1113865tge0P) which will be described in Section 2 Fori 1 2 the operator Ai(t) and the functions bi fi andgi
depend on time and we assume that the operator A2(t) isperiodic and the functions b1 b2 f2 andg2 are almostperiodic
e goal of this paper is to establish an effective ap-proximation for the slow equation of original system (1) byusing the averaging principle e averaged equation isobtained as follows
zu
zt(t ξ) A1(t)u(t ξ) + B1(u(t))(ξ) + f1(t ξ u(t ξ))
zωQ1
zt(t ξ) + 1113946
Zg1(t ξ u(t ξ) z)
z 1113957N1
zt(t ξ dz)
u(0 ξ) x(ξ) ξ isin O
N1u(t ξ) 0 tge 0 ξ isin zO
⎧⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
(2)
where B1 is the averaged coefficient which will be given inequation (6) To demonstrate the validity of the averagingprinciple we prove that for any Tgt 0 and ηgt 0 it yields
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
L2(O)gt η1113888 1113889 0 (3)
where u is the solution of the averaged equation (2)e theory of the averaging principle originated by Laplace
and Lagrange has been applied in celestial mechanics oscil-lation theory radiophysics and other fieldse firstly rigorousresults for the deterministic case were given by Bogolyubov andMitropolskii [1] Moreover Volosov [2] and Besjes [3] alsopromoted the development of the averaging principle engreat interest has appeared in its application to dynamicalsystems under random perturbations An important contri-bution was that in 1968 Khasminskii [4] originally proposedthe averaging principle for stochastic differential equations(SDEs) driven by Brownian motion Since then the averagingprinciple has been an active area of research Many studies onthe averaging principle of SDEs have been presented eg Givon[5] Freidlin andWentzell [6] Duan [7]ompson [8] andXuand his coworkers [9ndash11] Recently effective approximationsfor slow-fast SPDEs have been received extensive attentionCerrai [12 13] investigated the validity of the averagingprinciple for a class of stochastic reaction-diffusion equationswith multiplicative noise In addition Wang and Roberts [14]Pei and Xu [15ndash17] and Li [18] and Gao [19ndash21] also studiedthe averaging principles for the slow-fast SPDEs
e abovementioned works mainly considered autono-mous systems For the autonomous systems as long as theinitial value is given the solution of which only depends on the
duration of time not on the selection of the initial timeHowever if the initial time is different the solution of non-autonomous equations with the same initial data will also bedifferent erefore compared with the autonomous systemsthe dynamic behaviors of the nonautonomous systems aremore complex which can portray more actual models Che-pyzhov and Vishik [22] studied long-time dynamic behaviorsof nonautonomous dissipative system Carvalho [23] dealt withthe theory of attractors for the nonautonomous dynamicalsystems Bunder and Roberts [24] considered the discretemodelling of nonautonomous PDEs
In 2017 the averaging principle has been presented fornonautonomous slow-fast system of stochastic reaction-diffusion equations by Cerrai [25] But the system of thispaper was driven by Gaussian noises which is considered asan ideal noise source and only can simulate fluctuations nearthe mean value Actually due to the complexity of the ex-ternal environment random noise sources encountered inpractical fields usually exhibit non-Gaussian propertieswhich may cause sharp fluctuations It should be pointed outthat Poisson noise one of the most ubiquitous noise sourcesin many fields [26ndash28] can provide an accurate mathe-matical model to describe discontinuous random processessome large moves and unpredictable events [29ndash31] So inthis paper we are devoted to developing the averagingprinciple for nonautonomous systems of reaction-diffusionequations driven by Wiener processes and Poisson randommeasures
e key to using the averaging principle to analyzesystem (1) is the fast equation with a frozen slow componentx isin L2(O)
zvxy
zt(t ξ) A2(t) minus α( 1113857v
xy(t ξ) + b2 t ξ x(ξ) v
xy(t ξ)( 11138571113858 1113859 + f2 t ξ x(ξ) v
xy(t ξ)( 1113857
zωQ2
zt(t ξ) + 1113946
Zg2 t ξ x(ξ) v
xy(t ξ) z( 1113857
z 1113957Nϵ2
zt(t ξ dz)
vxy(s ξ) y(ξ) ξ isin O
N2vxy(t ξ) 0 tge s ξ isin zO
⎧⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎩
(4)
2 Complexity
By dealing with the Poisson terms we prove that anevolution family of measures (μx
t t isin R) on L2(O) for fastequation (4) exists en assuming thatA2(t) is periodic andb2 f2 andg2 are almost periodic we prove that the evolutionfamily ofmeasures is almost periodicMoreover with the aid ofeorem 210 in [32] we prove that the family of functions
t isin R⟼1113946L2(O)
B1(t x y)μxt (dy)1113896 1113897 (5)
is uniformly almost periodic for any x in a compact setK sub L2(O) where B1(t x y)(ξ) b1(t ξ x(ξ) y(ξ)) forany x y isin L2(O) and ξ isin O
According to the characteristics of almost periodicfunction ([25] eorem 34) we define the averaged co-efficient B1 as follows
B1(x) ≔ limT⟶infin
1T
1113946T
01113946
L2(O)B1(t x y)μx
t (dy)dt x isin L2(O)
(6)
Finally the averaged equation is obtained through theaveraged coefficient B1 Using the classical Khasminskiimethod to the present situation the averaging principle iseffective
e abovementioned notations will be given in Section 2and in this paper cgt 0 below with or without subscripts willrepresent a universal constant whose value may vary indifferent occasions
2 Notations Assumptions and Preliminaries
Let O be a bounded domain of Rd(dge 1) having a smoothboundary In this paper we denote H the separable Hilbertspace L2(O) endowed with the usual scalar product
langx yrangH 1113946O
x(ξ)y(ξ)dξ (7)
and with the corresponding norm middotH e norm in Linfin(O)
will be denoted by middotinfinFurthermore the subspace D((minus A)θ) [33ndash35] of the
generator A is dense in H and endowed with the norm
Λθ (minus A)θΛ
H Λ isin D (minus A)
θ1113872 1113873 (8)
for 0le θlt 1 0lt tleT According to eorem 613 in [36]there exists cθ gt 0 such that
(minus A)θe
At
Hle cθt
minus θ (9)
We denote by Bb(H) the Banach space of the boundedBorel functions φ H⟶ R endowed with the sup-norm
φ0 ≔ supxisinH
|φ(x)| (10)
and Cb(H) is the subspace of the uniformly continuousmappings Moreover D([s T]H) denotes the space of allcadlag path from [s T] into H
We shall denote that L(H) is the space of the boundedlinear operators in H and denote L2(H) the subspace ofHilbertndashSchmidt operators endowed with the norm
Q2 Tr QlowastQ[ ]
1113968 (11)
In slow-fast system (1) the Gaussian noises zωQ1 zt(t ξ)
and zωQ2 zt(t ξ) are assumed to be white in time andcolored in space in the case of space dimension dgt 1 fortge 0 and ξ isin O And ωQi (t ξ)(i 1 2) is the cylindricalWiener processes defined as
ωQi (t ξ) 1113944infin
k1Qiek(ξ)βk(t) i 1 2 (12)
where ek1113864 1113865kisinN is a complete orthonormal basis in Hβk(t)1113864 1113865kisinN is a sequence of mutually independent standardBrownian motion defined on the same complete stochasticbasis (ΩF Ft1113864 1113865tge0P) and Qi is a bounded linear operatoron H
Next we give the definitions of Poisson random mea-sures 1113957N1(dt dz) and 1113957N
ϵ2(dt dz) Let (ZB(Z)) be a given
measurable space and v(dz) be a σ-finite measure on itDpi
t i 1 2 represents two countable subsets of R+
Moreover let p1t t isin Dp1
tbe a stationaryFt-adapted Poisson
point process on Z with the characteristic v and p2t t isin Dp2
t
be the other stationaryFt-adapted Poisson point process onZ with the characteristic vϵ Denote by Ni(dt dz) i 1 2the Poisson counting measure associated with pi
t ie
Ni(tΛ) ≔ 1113944sisinD
pitslet
IΛ pit1113872 1113873 i 1 2
(13)
Let us denote the two independent compensated Poissonmeasures
1113957N1(dt dz) ≔ N1(dt dz) minus v1(dz)dt
1113957Nϵ2(dt dz) ≔ N2(dt dz) minus
1ϵv2(dz)dt
(14)
where v1(dz)dt and 1ϵv2(dz)dt are the compensatorsRefer to [28 37] for a more detailed description of the
stochastic integral with respect to a cylindrical Wienerprocess and Poisson random measure
For any t isin R the operators A1(t) and A2(t) are sec-ond-order uniformly elliptic operators having continuouscoefficients on O e operators N1 and N2 are theboundary operators which can be either the identity op-erator (Dirichlet boundary condition) or a first-order op-erator (coefficients satisfying a uniform nontangentialitycondition) We shall assume that the operatorAi(t) has thefollowing form
Ai(t) ci(t)Ai + Li(t) t isin R i 1 2 (15)
where Ai is a second order uniformly elliptic operator[38 39] with continuous coefficients on O which is inde-pendent of t In addition Li(t) is a first-order differentialoperator which has the form
Li(t ξ)u(ξ) langli(t ξ)nablau(ξ)rangRd t isin R ξ isin O (16)
e realizations of the differential operators Ai and Li
in H are Ai and Li Moreover A1 and A2 generate twoanalytic semigroups etA1 and etA2 respectively
Complexity 3
Now we give the following assumptions
(A1)
(a) For i 1 2 the function ci R⟶ R is con-tinuous and there exist c0 cgt 0 such that
c0 le ci(t)le c t isin R (17)
(b) For i 1 2 the function li R times O⟶ Rd iscontinuous and bounded
(A2) For i 1 2 there exists a complete orthonormalsystem eik1113966 1113967
kisinN in H and two sequences of nonneg-ative real numbers αik1113966 1113967
kisinN and λik1113966 1113967kisinN such that
Aieik minus αikeik
Qieik λikeik kge 1
κi ≔ 1113944infin
k1λρi
ik eik
2infin ltinfin
ζ i ≔ 1113944infin
k1αminus βi
ik eik
2infin ltinfin
(18)
for some constants ρi isin (2 +infin] and βi isin (0 +infin)
such that
βi ρi minus 2( 11138571113858 1113859
ρi
lt 1 (19)
Remark 1 For more comments and examples about theassumption (A2) of the operators Ai and Qi the reader canread [12]
(A3) e mappings b1 R times O times R2⟶ R f1 R times O times
R⟶ R g1 R times O times R times Z⟶ R are measur-able and the mappings b1(t ξ middot) R2⟶ R
f1(t ξ middot) R⟶ R g1(t ξ middot z) R⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z More-over for all pge 1 there exist positive constantsc1 and c2 such that for all x1 x2 isin R we have
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c1 1 + x11113868111386811138681113868
1113868111386811138681113868p
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 1113857 minus g1 t ξ x2 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c2 x1 minus x21113868111386811138681113868
1113868111386811138681113868p
(20)
(A4) e mappings b2 R times O times R2⟶ R f2 R times O times
R2⟶ R g2 R times O times R2 times Z⟶ R are mea-surable and the mappings b2(t ξ middot) R2⟶ R
f2(t ξ middot) R2⟶ R g2(t ξ middot z) R2⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z
Moreover for all qge 1 there exist positive con-stants c3 and c4 such that for all(xi yi) isin R2 i 1 2 we have
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
1113868111386811138681113868qυ2(dz)
le c3 1 + x11113868111386811138681113868
1113868111386811138681113868q
+ y11113868111386811138681113868
1113868111386811138681113868q
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
minus g2 t ξ x2 y2 z( 1113857|qυ2(dz)
le c4 x1 minus x21113868111386811138681113868
1113868111386811138681113868q
+ y1 minus y21113868111386811138681113868
1113868111386811138681113868q11138731113872 (21)
Remark 2 For any (t ξ) isin R times O and x y h isin H z isin Z weshall set
B1(t x y)(ξ) ≔ b1(t ξ x(ξ) y(ξ))
B2(t x y)(ξ) ≔ b2(t ξ x(ξ) y(ξ))
F1(t x)h1113858 1113859(ξ) ≔ f1(t ξ x(ξ))h(ξ)
F2(t x y)h1113858 1113859(ξ) ≔ f2(t ξ x(ξ) y(ξ))h(ξ)
G1(t x z)h1113858 1113859(ξ) ≔ g1(t ξ x(ξ) z)h(ξ)
G2(t x y z)h1113858 1113859(ξ) ≔ g2(t ξ x(ξ) y(ξ) z)h(ξ)
(22)
Due to (A3) and (A4) for any fixed (t z) isin (RZ) themappings
B1(t middot) H times H⟶ H B2(t middot) H times H⟶ H
F1(t middot) H⟶L(H) F2(t middot) H times H⟶L(H)
G1(t middot z) H⟶L(H) G2(t middot z) H times H⟶L(H)
(23)
are Lipschitz continuous and have linear growth conditionsNow for i 1 2 we define
ci(t s) ≔ 1113946t
sci(r)dr slt t (24)
and for any ϵgt 0 and βge 0 set
Uβϵi(t s) e1ϵci(ts)Ai minus βϵ(tminus s)
slt t (25)
For ϵ 1 we write Uβi(t s) and for ϵ 1 and β 0 wewrite Ui(t s)
Next for any ϵgt 0 βge 0 and for anyu isin C([s t] W
1p0 (O)) we define
ψβϵi(u s)(r) 1ϵ
1113946r
sUβϵi(r ρ)Li(ρ)u(ρ)dρ slt rlt t
(26)
In the case ϵ 1 we write ψβi(u s)(r) and in the caseϵ 1 and β 0 we write ψi(u s)(r)
Remark 3 By using the same argument as Chapter 5 of thebook [40] we can get that there exists a unique evolution
system U(t s) eA 1113938
t
sc(r)dr
(0le sle tleT) for c(t)A More-over we also can get that ψβϵi(u s)(t) is the solution of
4 Complexity
du(t) 1ϵ
Ai(t) minus β( 1113857u(t)dt tgt s u(s) 0 (27)
3 A Priori Bounds for the Solution
With all notations introduced above system (1) can berewritten in the following abstract form
duϵ(t) A1(t)uϵ(t) + B1 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+ F1 t uϵ(t)( 1113857dωQ1(t)
+ 1113946Z
G1 t uϵ(t) z( 1113857 1113957N1(dt dz)
dvϵ(t) 1ϵ
A2(t) minus α( 1113857vϵ(t) + B2 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ(t) vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ(t) vϵ(t) z( 1113857 1113957Nϵ2(dt dz)
uϵ(0) x vϵ(0) y
(28)
According to Remark 2 we know that the coefficients ofsystem (28) satisfy global Lipschitz and linear growthconditions and the assumptions (A1)ndash(A4) are uniformwith respect to t isin R So using the same argument as[28 41 42] it is easy to prove that for any ϵgt 0 Tgt 0 andx y isin H there exists a unique adapted mild solution (uϵ vϵ)
to system (28) in Lp(Ω D([0 T]H times H)) is means thatthere exist two unique adapted processes uϵ and vϵ inLp(Ω D([0 T]H)) such that
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
vϵ(t) Uαϵ2(t 0)y + ψαϵ2 vϵ 0( 1113857(t)
+1ϵ
1113946t
0Uαϵ2(t r)B2 r uϵ(r) vϵ(r)( 1113857dr
+1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
+ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(29)
Lemma 1 Under (A1)ndash(A4) for any pge 1 and Tgt 0 thereexists a positive constant cpT such that for any x y isin H andϵ isin (0 1] we have
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 (30)
1113946T
0E vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (31)
Proof For fixed ϵ isin (0 1] and x y isin H for any t isin [0 T]we denote
Γ1ϵ(t) ≔ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
Ψ1ϵ(t) ≔ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(32)
Setting Λ1ϵ(t) ≔ uϵ(t) minus Γ1ϵ(t) minus Ψ1ϵ(t) we have
ddtΛ1ϵ(t) c1(t)A1Λ1ϵ(t) + L1(t) Λ1ϵ(t) + Γ1ϵ(t)1113872
+ Ψ1ϵ(t)1113873B1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
Λ1ϵ(0) x
(33)
For any pge 2 because B1(middot) is Lipschitz continuoususing Youngrsquos inequality we have1p
ddtΛ1ϵ(t)
p
Hlangc1(t)A1Λ1ϵ(t)Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langL1(t) Λ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
le c Λ1ϵ(t)
p
H+ c Γ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c Ψ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873
HΛ1ϵ(t)
pminus 1H
le cp Λ1ϵ(t)
p
H+ cp 1 + Γ1ϵ(t)
p
H+ Ψ1ϵ(t)
p
H+ vϵ(t)
p
H1113872 1113873
(34)
is implies that
Λ1ϵ(t)
p
Hle e
cptx
p
H + cp 1113946t
0e
cp(tminus r) 1 + Γ1ϵ(r)
p
H1113872
+ Ψ1ϵ(r)
p
H+ vϵ(r)
p
H1113873dr
(35)
Complexity 5
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
By dealing with the Poisson terms we prove that anevolution family of measures (μx
t t isin R) on L2(O) for fastequation (4) exists en assuming thatA2(t) is periodic andb2 f2 andg2 are almost periodic we prove that the evolutionfamily ofmeasures is almost periodicMoreover with the aid ofeorem 210 in [32] we prove that the family of functions
t isin R⟼1113946L2(O)
B1(t x y)μxt (dy)1113896 1113897 (5)
is uniformly almost periodic for any x in a compact setK sub L2(O) where B1(t x y)(ξ) b1(t ξ x(ξ) y(ξ)) forany x y isin L2(O) and ξ isin O
According to the characteristics of almost periodicfunction ([25] eorem 34) we define the averaged co-efficient B1 as follows
B1(x) ≔ limT⟶infin
1T
1113946T
01113946
L2(O)B1(t x y)μx
t (dy)dt x isin L2(O)
(6)
Finally the averaged equation is obtained through theaveraged coefficient B1 Using the classical Khasminskiimethod to the present situation the averaging principle iseffective
e abovementioned notations will be given in Section 2and in this paper cgt 0 below with or without subscripts willrepresent a universal constant whose value may vary indifferent occasions
2 Notations Assumptions and Preliminaries
Let O be a bounded domain of Rd(dge 1) having a smoothboundary In this paper we denote H the separable Hilbertspace L2(O) endowed with the usual scalar product
langx yrangH 1113946O
x(ξ)y(ξ)dξ (7)
and with the corresponding norm middotH e norm in Linfin(O)
will be denoted by middotinfinFurthermore the subspace D((minus A)θ) [33ndash35] of the
generator A is dense in H and endowed with the norm
Λθ (minus A)θΛ
H Λ isin D (minus A)
θ1113872 1113873 (8)
for 0le θlt 1 0lt tleT According to eorem 613 in [36]there exists cθ gt 0 such that
(minus A)θe
At
Hle cθt
minus θ (9)
We denote by Bb(H) the Banach space of the boundedBorel functions φ H⟶ R endowed with the sup-norm
φ0 ≔ supxisinH
|φ(x)| (10)
and Cb(H) is the subspace of the uniformly continuousmappings Moreover D([s T]H) denotes the space of allcadlag path from [s T] into H
We shall denote that L(H) is the space of the boundedlinear operators in H and denote L2(H) the subspace ofHilbertndashSchmidt operators endowed with the norm
Q2 Tr QlowastQ[ ]
1113968 (11)
In slow-fast system (1) the Gaussian noises zωQ1 zt(t ξ)
and zωQ2 zt(t ξ) are assumed to be white in time andcolored in space in the case of space dimension dgt 1 fortge 0 and ξ isin O And ωQi (t ξ)(i 1 2) is the cylindricalWiener processes defined as
ωQi (t ξ) 1113944infin
k1Qiek(ξ)βk(t) i 1 2 (12)
where ek1113864 1113865kisinN is a complete orthonormal basis in Hβk(t)1113864 1113865kisinN is a sequence of mutually independent standardBrownian motion defined on the same complete stochasticbasis (ΩF Ft1113864 1113865tge0P) and Qi is a bounded linear operatoron H
Next we give the definitions of Poisson random mea-sures 1113957N1(dt dz) and 1113957N
ϵ2(dt dz) Let (ZB(Z)) be a given
measurable space and v(dz) be a σ-finite measure on itDpi
t i 1 2 represents two countable subsets of R+
Moreover let p1t t isin Dp1
tbe a stationaryFt-adapted Poisson
point process on Z with the characteristic v and p2t t isin Dp2
t
be the other stationaryFt-adapted Poisson point process onZ with the characteristic vϵ Denote by Ni(dt dz) i 1 2the Poisson counting measure associated with pi
t ie
Ni(tΛ) ≔ 1113944sisinD
pitslet
IΛ pit1113872 1113873 i 1 2
(13)
Let us denote the two independent compensated Poissonmeasures
1113957N1(dt dz) ≔ N1(dt dz) minus v1(dz)dt
1113957Nϵ2(dt dz) ≔ N2(dt dz) minus
1ϵv2(dz)dt
(14)
where v1(dz)dt and 1ϵv2(dz)dt are the compensatorsRefer to [28 37] for a more detailed description of the
stochastic integral with respect to a cylindrical Wienerprocess and Poisson random measure
For any t isin R the operators A1(t) and A2(t) are sec-ond-order uniformly elliptic operators having continuouscoefficients on O e operators N1 and N2 are theboundary operators which can be either the identity op-erator (Dirichlet boundary condition) or a first-order op-erator (coefficients satisfying a uniform nontangentialitycondition) We shall assume that the operatorAi(t) has thefollowing form
Ai(t) ci(t)Ai + Li(t) t isin R i 1 2 (15)
where Ai is a second order uniformly elliptic operator[38 39] with continuous coefficients on O which is inde-pendent of t In addition Li(t) is a first-order differentialoperator which has the form
Li(t ξ)u(ξ) langli(t ξ)nablau(ξ)rangRd t isin R ξ isin O (16)
e realizations of the differential operators Ai and Li
in H are Ai and Li Moreover A1 and A2 generate twoanalytic semigroups etA1 and etA2 respectively
Complexity 3
Now we give the following assumptions
(A1)
(a) For i 1 2 the function ci R⟶ R is con-tinuous and there exist c0 cgt 0 such that
c0 le ci(t)le c t isin R (17)
(b) For i 1 2 the function li R times O⟶ Rd iscontinuous and bounded
(A2) For i 1 2 there exists a complete orthonormalsystem eik1113966 1113967
kisinN in H and two sequences of nonneg-ative real numbers αik1113966 1113967
kisinN and λik1113966 1113967kisinN such that
Aieik minus αikeik
Qieik λikeik kge 1
κi ≔ 1113944infin
k1λρi
ik eik
2infin ltinfin
ζ i ≔ 1113944infin
k1αminus βi
ik eik
2infin ltinfin
(18)
for some constants ρi isin (2 +infin] and βi isin (0 +infin)
such that
βi ρi minus 2( 11138571113858 1113859
ρi
lt 1 (19)
Remark 1 For more comments and examples about theassumption (A2) of the operators Ai and Qi the reader canread [12]
(A3) e mappings b1 R times O times R2⟶ R f1 R times O times
R⟶ R g1 R times O times R times Z⟶ R are measur-able and the mappings b1(t ξ middot) R2⟶ R
f1(t ξ middot) R⟶ R g1(t ξ middot z) R⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z More-over for all pge 1 there exist positive constantsc1 and c2 such that for all x1 x2 isin R we have
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c1 1 + x11113868111386811138681113868
1113868111386811138681113868p
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 1113857 minus g1 t ξ x2 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c2 x1 minus x21113868111386811138681113868
1113868111386811138681113868p
(20)
(A4) e mappings b2 R times O times R2⟶ R f2 R times O times
R2⟶ R g2 R times O times R2 times Z⟶ R are mea-surable and the mappings b2(t ξ middot) R2⟶ R
f2(t ξ middot) R2⟶ R g2(t ξ middot z) R2⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z
Moreover for all qge 1 there exist positive con-stants c3 and c4 such that for all(xi yi) isin R2 i 1 2 we have
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
1113868111386811138681113868qυ2(dz)
le c3 1 + x11113868111386811138681113868
1113868111386811138681113868q
+ y11113868111386811138681113868
1113868111386811138681113868q
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
minus g2 t ξ x2 y2 z( 1113857|qυ2(dz)
le c4 x1 minus x21113868111386811138681113868
1113868111386811138681113868q
+ y1 minus y21113868111386811138681113868
1113868111386811138681113868q11138731113872 (21)
Remark 2 For any (t ξ) isin R times O and x y h isin H z isin Z weshall set
B1(t x y)(ξ) ≔ b1(t ξ x(ξ) y(ξ))
B2(t x y)(ξ) ≔ b2(t ξ x(ξ) y(ξ))
F1(t x)h1113858 1113859(ξ) ≔ f1(t ξ x(ξ))h(ξ)
F2(t x y)h1113858 1113859(ξ) ≔ f2(t ξ x(ξ) y(ξ))h(ξ)
G1(t x z)h1113858 1113859(ξ) ≔ g1(t ξ x(ξ) z)h(ξ)
G2(t x y z)h1113858 1113859(ξ) ≔ g2(t ξ x(ξ) y(ξ) z)h(ξ)
(22)
Due to (A3) and (A4) for any fixed (t z) isin (RZ) themappings
B1(t middot) H times H⟶ H B2(t middot) H times H⟶ H
F1(t middot) H⟶L(H) F2(t middot) H times H⟶L(H)
G1(t middot z) H⟶L(H) G2(t middot z) H times H⟶L(H)
(23)
are Lipschitz continuous and have linear growth conditionsNow for i 1 2 we define
ci(t s) ≔ 1113946t
sci(r)dr slt t (24)
and for any ϵgt 0 and βge 0 set
Uβϵi(t s) e1ϵci(ts)Ai minus βϵ(tminus s)
slt t (25)
For ϵ 1 we write Uβi(t s) and for ϵ 1 and β 0 wewrite Ui(t s)
Next for any ϵgt 0 βge 0 and for anyu isin C([s t] W
1p0 (O)) we define
ψβϵi(u s)(r) 1ϵ
1113946r
sUβϵi(r ρ)Li(ρ)u(ρ)dρ slt rlt t
(26)
In the case ϵ 1 we write ψβi(u s)(r) and in the caseϵ 1 and β 0 we write ψi(u s)(r)
Remark 3 By using the same argument as Chapter 5 of thebook [40] we can get that there exists a unique evolution
system U(t s) eA 1113938
t
sc(r)dr
(0le sle tleT) for c(t)A More-over we also can get that ψβϵi(u s)(t) is the solution of
4 Complexity
du(t) 1ϵ
Ai(t) minus β( 1113857u(t)dt tgt s u(s) 0 (27)
3 A Priori Bounds for the Solution
With all notations introduced above system (1) can berewritten in the following abstract form
duϵ(t) A1(t)uϵ(t) + B1 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+ F1 t uϵ(t)( 1113857dωQ1(t)
+ 1113946Z
G1 t uϵ(t) z( 1113857 1113957N1(dt dz)
dvϵ(t) 1ϵ
A2(t) minus α( 1113857vϵ(t) + B2 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ(t) vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ(t) vϵ(t) z( 1113857 1113957Nϵ2(dt dz)
uϵ(0) x vϵ(0) y
(28)
According to Remark 2 we know that the coefficients ofsystem (28) satisfy global Lipschitz and linear growthconditions and the assumptions (A1)ndash(A4) are uniformwith respect to t isin R So using the same argument as[28 41 42] it is easy to prove that for any ϵgt 0 Tgt 0 andx y isin H there exists a unique adapted mild solution (uϵ vϵ)
to system (28) in Lp(Ω D([0 T]H times H)) is means thatthere exist two unique adapted processes uϵ and vϵ inLp(Ω D([0 T]H)) such that
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
vϵ(t) Uαϵ2(t 0)y + ψαϵ2 vϵ 0( 1113857(t)
+1ϵ
1113946t
0Uαϵ2(t r)B2 r uϵ(r) vϵ(r)( 1113857dr
+1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
+ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(29)
Lemma 1 Under (A1)ndash(A4) for any pge 1 and Tgt 0 thereexists a positive constant cpT such that for any x y isin H andϵ isin (0 1] we have
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 (30)
1113946T
0E vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (31)
Proof For fixed ϵ isin (0 1] and x y isin H for any t isin [0 T]we denote
Γ1ϵ(t) ≔ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
Ψ1ϵ(t) ≔ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(32)
Setting Λ1ϵ(t) ≔ uϵ(t) minus Γ1ϵ(t) minus Ψ1ϵ(t) we have
ddtΛ1ϵ(t) c1(t)A1Λ1ϵ(t) + L1(t) Λ1ϵ(t) + Γ1ϵ(t)1113872
+ Ψ1ϵ(t)1113873B1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
Λ1ϵ(0) x
(33)
For any pge 2 because B1(middot) is Lipschitz continuoususing Youngrsquos inequality we have1p
ddtΛ1ϵ(t)
p
Hlangc1(t)A1Λ1ϵ(t)Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langL1(t) Λ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
le c Λ1ϵ(t)
p
H+ c Γ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c Ψ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873
HΛ1ϵ(t)
pminus 1H
le cp Λ1ϵ(t)
p
H+ cp 1 + Γ1ϵ(t)
p
H+ Ψ1ϵ(t)
p
H+ vϵ(t)
p
H1113872 1113873
(34)
is implies that
Λ1ϵ(t)
p
Hle e
cptx
p
H + cp 1113946t
0e
cp(tminus r) 1 + Γ1ϵ(r)
p
H1113872
+ Ψ1ϵ(r)
p
H+ vϵ(r)
p
H1113873dr
(35)
Complexity 5
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
Now we give the following assumptions
(A1)
(a) For i 1 2 the function ci R⟶ R is con-tinuous and there exist c0 cgt 0 such that
c0 le ci(t)le c t isin R (17)
(b) For i 1 2 the function li R times O⟶ Rd iscontinuous and bounded
(A2) For i 1 2 there exists a complete orthonormalsystem eik1113966 1113967
kisinN in H and two sequences of nonneg-ative real numbers αik1113966 1113967
kisinN and λik1113966 1113967kisinN such that
Aieik minus αikeik
Qieik λikeik kge 1
κi ≔ 1113944infin
k1λρi
ik eik
2infin ltinfin
ζ i ≔ 1113944infin
k1αminus βi
ik eik
2infin ltinfin
(18)
for some constants ρi isin (2 +infin] and βi isin (0 +infin)
such that
βi ρi minus 2( 11138571113858 1113859
ρi
lt 1 (19)
Remark 1 For more comments and examples about theassumption (A2) of the operators Ai and Qi the reader canread [12]
(A3) e mappings b1 R times O times R2⟶ R f1 R times O times
R⟶ R g1 R times O times R times Z⟶ R are measur-able and the mappings b1(t ξ middot) R2⟶ R
f1(t ξ middot) R⟶ R g1(t ξ middot z) R⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z More-over for all pge 1 there exist positive constantsc1 and c2 such that for all x1 x2 isin R we have
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c1 1 + x11113868111386811138681113868
1113868111386811138681113868p
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g1 t ξ x1 z( 1113857 minus g1 t ξ x2 z( 11138571113868111386811138681113868
1113868111386811138681113868pυ1(dz)
le c2 x1 minus x21113868111386811138681113868
1113868111386811138681113868p
(20)
(A4) e mappings b2 R times O times R2⟶ R f2 R times O times
R2⟶ R g2 R times O times R2 times Z⟶ R are mea-surable and the mappings b2(t ξ middot) R2⟶ R
f2(t ξ middot) R2⟶ R g2(t ξ middot z) R2⟶ R areLipschitz continuous and linearly growing uni-formly with respect to (t ξ z) isin R times O times Z
Moreover for all qge 1 there exist positive con-stants c3 and c4 such that for all(xi yi) isin R2 i 1 2 we have
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
1113868111386811138681113868qυ2(dz)
le c3 1 + x11113868111386811138681113868
1113868111386811138681113868q
+ y11113868111386811138681113868
1113868111386811138681113868q
1113872 1113873
sup(tξ)isinRtimesO
1113946Z
g2 t ξ x1 y1 z( 11138571113868111386811138681113868
minus g2 t ξ x2 y2 z( 1113857|qυ2(dz)
le c4 x1 minus x21113868111386811138681113868
1113868111386811138681113868q
+ y1 minus y21113868111386811138681113868
1113868111386811138681113868q11138731113872 (21)
Remark 2 For any (t ξ) isin R times O and x y h isin H z isin Z weshall set
B1(t x y)(ξ) ≔ b1(t ξ x(ξ) y(ξ))
B2(t x y)(ξ) ≔ b2(t ξ x(ξ) y(ξ))
F1(t x)h1113858 1113859(ξ) ≔ f1(t ξ x(ξ))h(ξ)
F2(t x y)h1113858 1113859(ξ) ≔ f2(t ξ x(ξ) y(ξ))h(ξ)
G1(t x z)h1113858 1113859(ξ) ≔ g1(t ξ x(ξ) z)h(ξ)
G2(t x y z)h1113858 1113859(ξ) ≔ g2(t ξ x(ξ) y(ξ) z)h(ξ)
(22)
Due to (A3) and (A4) for any fixed (t z) isin (RZ) themappings
B1(t middot) H times H⟶ H B2(t middot) H times H⟶ H
F1(t middot) H⟶L(H) F2(t middot) H times H⟶L(H)
G1(t middot z) H⟶L(H) G2(t middot z) H times H⟶L(H)
(23)
are Lipschitz continuous and have linear growth conditionsNow for i 1 2 we define
ci(t s) ≔ 1113946t
sci(r)dr slt t (24)
and for any ϵgt 0 and βge 0 set
Uβϵi(t s) e1ϵci(ts)Ai minus βϵ(tminus s)
slt t (25)
For ϵ 1 we write Uβi(t s) and for ϵ 1 and β 0 wewrite Ui(t s)
Next for any ϵgt 0 βge 0 and for anyu isin C([s t] W
1p0 (O)) we define
ψβϵi(u s)(r) 1ϵ
1113946r
sUβϵi(r ρ)Li(ρ)u(ρ)dρ slt rlt t
(26)
In the case ϵ 1 we write ψβi(u s)(r) and in the caseϵ 1 and β 0 we write ψi(u s)(r)
Remark 3 By using the same argument as Chapter 5 of thebook [40] we can get that there exists a unique evolution
system U(t s) eA 1113938
t
sc(r)dr
(0le sle tleT) for c(t)A More-over we also can get that ψβϵi(u s)(t) is the solution of
4 Complexity
du(t) 1ϵ
Ai(t) minus β( 1113857u(t)dt tgt s u(s) 0 (27)
3 A Priori Bounds for the Solution
With all notations introduced above system (1) can berewritten in the following abstract form
duϵ(t) A1(t)uϵ(t) + B1 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+ F1 t uϵ(t)( 1113857dωQ1(t)
+ 1113946Z
G1 t uϵ(t) z( 1113857 1113957N1(dt dz)
dvϵ(t) 1ϵ
A2(t) minus α( 1113857vϵ(t) + B2 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ(t) vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ(t) vϵ(t) z( 1113857 1113957Nϵ2(dt dz)
uϵ(0) x vϵ(0) y
(28)
According to Remark 2 we know that the coefficients ofsystem (28) satisfy global Lipschitz and linear growthconditions and the assumptions (A1)ndash(A4) are uniformwith respect to t isin R So using the same argument as[28 41 42] it is easy to prove that for any ϵgt 0 Tgt 0 andx y isin H there exists a unique adapted mild solution (uϵ vϵ)
to system (28) in Lp(Ω D([0 T]H times H)) is means thatthere exist two unique adapted processes uϵ and vϵ inLp(Ω D([0 T]H)) such that
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
vϵ(t) Uαϵ2(t 0)y + ψαϵ2 vϵ 0( 1113857(t)
+1ϵ
1113946t
0Uαϵ2(t r)B2 r uϵ(r) vϵ(r)( 1113857dr
+1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
+ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(29)
Lemma 1 Under (A1)ndash(A4) for any pge 1 and Tgt 0 thereexists a positive constant cpT such that for any x y isin H andϵ isin (0 1] we have
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 (30)
1113946T
0E vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (31)
Proof For fixed ϵ isin (0 1] and x y isin H for any t isin [0 T]we denote
Γ1ϵ(t) ≔ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
Ψ1ϵ(t) ≔ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(32)
Setting Λ1ϵ(t) ≔ uϵ(t) minus Γ1ϵ(t) minus Ψ1ϵ(t) we have
ddtΛ1ϵ(t) c1(t)A1Λ1ϵ(t) + L1(t) Λ1ϵ(t) + Γ1ϵ(t)1113872
+ Ψ1ϵ(t)1113873B1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
Λ1ϵ(0) x
(33)
For any pge 2 because B1(middot) is Lipschitz continuoususing Youngrsquos inequality we have1p
ddtΛ1ϵ(t)
p
Hlangc1(t)A1Λ1ϵ(t)Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langL1(t) Λ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
le c Λ1ϵ(t)
p
H+ c Γ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c Ψ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873
HΛ1ϵ(t)
pminus 1H
le cp Λ1ϵ(t)
p
H+ cp 1 + Γ1ϵ(t)
p
H+ Ψ1ϵ(t)
p
H+ vϵ(t)
p
H1113872 1113873
(34)
is implies that
Λ1ϵ(t)
p
Hle e
cptx
p
H + cp 1113946t
0e
cp(tminus r) 1 + Γ1ϵ(r)
p
H1113872
+ Ψ1ϵ(r)
p
H+ vϵ(r)
p
H1113873dr
(35)
Complexity 5
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
du(t) 1ϵ
Ai(t) minus β( 1113857u(t)dt tgt s u(s) 0 (27)
3 A Priori Bounds for the Solution
With all notations introduced above system (1) can berewritten in the following abstract form
duϵ(t) A1(t)uϵ(t) + B1 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+ F1 t uϵ(t)( 1113857dωQ1(t)
+ 1113946Z
G1 t uϵ(t) z( 1113857 1113957N1(dt dz)
dvϵ(t) 1ϵ
A2(t) minus α( 1113857vϵ(t) + B2 t uϵ(t) vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ(t) vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ(t) vϵ(t) z( 1113857 1113957Nϵ2(dt dz)
uϵ(0) x vϵ(0) y
(28)
According to Remark 2 we know that the coefficients ofsystem (28) satisfy global Lipschitz and linear growthconditions and the assumptions (A1)ndash(A4) are uniformwith respect to t isin R So using the same argument as[28 41 42] it is easy to prove that for any ϵgt 0 Tgt 0 andx y isin H there exists a unique adapted mild solution (uϵ vϵ)
to system (28) in Lp(Ω D([0 T]H times H)) is means thatthere exist two unique adapted processes uϵ and vϵ inLp(Ω D([0 T]H)) such that
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
vϵ(t) Uαϵ2(t 0)y + ψαϵ2 vϵ 0( 1113857(t)
+1ϵ
1113946t
0Uαϵ2(t r)B2 r uϵ(r) vϵ(r)( 1113857dr
+1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
+ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
(29)
Lemma 1 Under (A1)ndash(A4) for any pge 1 and Tgt 0 thereexists a positive constant cpT such that for any x y isin H andϵ isin (0 1] we have
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 (30)
1113946T
0E vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (31)
Proof For fixed ϵ isin (0 1] and x y isin H for any t isin [0 T]we denote
Γ1ϵ(t) ≔ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
Ψ1ϵ(t) ≔ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(32)
Setting Λ1ϵ(t) ≔ uϵ(t) minus Γ1ϵ(t) minus Ψ1ϵ(t) we have
ddtΛ1ϵ(t) c1(t)A1Λ1ϵ(t) + L1(t) Λ1ϵ(t) + Γ1ϵ(t)1113872
+ Ψ1ϵ(t)1113873B1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
Λ1ϵ(0) x
(33)
For any pge 2 because B1(middot) is Lipschitz continuoususing Youngrsquos inequality we have1p
ddtΛ1ϵ(t)
p
Hlangc1(t)A1Λ1ϵ(t)Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langL1(t) Λ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 tΛ1ϵ(t) + Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Γ1ϵ(t) + Ψ1ϵ(t) vϵ(t)1113872 1113873
minus B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
+langB1 t Ψ1ϵ(t) vϵ(t)1113872 1113873Λ1ϵ(t)rangH Λ1ϵ(t)
pminus 2H
le c Λ1ϵ(t)
p
H+ c Γ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c Ψ1ϵ(t)
H Λ1ϵ(t)
pminus 1H
+ c B1 t Ψ1ϵ(t) vϵ(t)1113872 1113873
HΛ1ϵ(t)
pminus 1H
le cp Λ1ϵ(t)
p
H+ cp 1 + Γ1ϵ(t)
p
H+ Ψ1ϵ(t)
p
H+ vϵ(t)
p
H1113872 1113873
(34)
is implies that
Λ1ϵ(t)
p
Hle e
cptx
p
H + cp 1113946t
0e
cp(tminus r) 1 + Γ1ϵ(r)
p
H1113872
+ Ψ1ϵ(r)
p
H+ vϵ(r)
p
H1113873dr
(35)
Complexity 5
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
According to the definition of Λ1ϵ(t) for any t isin [0 T]we have
uϵ(t)
p
Hle cpT 1 +x
p
H + suprisin[0T]
Γ1ϵ(r)
p
H+ sup
risin[0T]
Ψ1ϵ(r)
p
H1113888 1113889
+ cpT 1113946T
0vϵ(r)
p
Hdr
(36)
So
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + E suptisin[0T]
Γ1ϵ(t)
p
H1113888
+ E suptisin[0T]
Ψ1ϵ(t)
p
H1113889 + cpT 1113946
T
0E vϵ(r)
p
Hdr
(37)
According to Lemma 41 in [12] with θ 0 it is easy toprove that
E suptisin[0T]
Γ1ϵ(t)p
H le cpT 1113946T
01 + Euϵ(r)
p
H1113872 1113873dr (38)
Due to (A3) using Kunitarsquos first inequality we get
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpE 1113946
T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
2
Hv1(dz)dr1113888 1113889
p2
+ cpE1113946T
01113946Z
ec1(tr)A1G1 r uϵ(r) z( 1113857
p
Hv1(dz)dr
le cpE 1113946T
01 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946T
01 + uϵ(r)
p
H1113872 1113873drle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr
(39)
E suptisin[0T]
Ψ1ϵ(t)
p
Hle cpT 1113946
T
01 + E uϵ(r)
p
H1113872 1113873dr (40)
Substituting (38) and (39) into (37) we obtain
E suptisin[0T]
uϵ(t)
p
Hle cpT 1 +x
p
H + 1113946T
0E vϵ(r)
p
Hdr1113888 1113889
+ cpT 1113946T
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(41)
Now we have to estimate
1113946T
0E vϵ(r)
p
Hdr (42)
For any t isin [0 T] we set
Γ2ϵ(t) ≔1ϵ
radic 1113946t
0Uαϵ2(t r)F2 r uϵ(r) vϵ(r)( 1113857dw
Q2(r)
Ψ2ϵ(t) ≔ 1113946t
01113946Z
Uαϵ2(t r)G2 r uϵ(r) vϵ(r) z( 1113857 1113957Nϵ2(dr dz)
(43)
Let Λ2ϵ(t) ≔ vϵ(t) minus Γ2ϵ(t) minus Ψ2ϵ(t) we have
ddtΛ2ϵ(t)
1ϵ
c2(t)A2 minus α( 1113857Λ2ϵ(t) +1ϵL2(t)
middot Λ2ϵ(t) + Γ2ϵ(t) + Ψ2ϵ(t)1113872 1113873
+1ϵB2 t uϵ(t)Λ2ϵ(t) + Γ2ϵ(t)1113872
+ Ψ2ϵ(t)1113873 Λ2ϵ(0) y
(44)
For any pge 2 because αgt 0 is large enough by pro-ceeding as in (34) we can get
1p
ddtΛ2ϵ(t)
p
Hle minus
α2ϵΛ2ϵ(t)
p
H
+cp
ϵ1 + uϵ(t)
p
H+ Γ2ϵ(t)
p
H+ Ψ2ϵ(t)
p
H1113872 1113873
(45)
According to the Gronwall inequality we have
Λ2ϵ(t)
p
Hle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r)
middot 1 + uϵ(r)
p
H+ Γ2ϵ(r)
p
H+ Ψ2ϵ(r)
p
H1113872 1113873dr + e
minus (αp2ϵ)ty
p
H
(46)
6 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
According to the definition of Λ2ϵ(t) for any t isin [0 T]we have
E vϵ(t)
p
Hle cpE Γ2ϵ(t)
p
H+ cpE Ψ2ϵ(t)
p
H+ cpe
minus (αp2ϵ)ty
p
H
+cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H1113872
+ E Ψ2ϵ(r)
p
H1113873dr
(47)
erefore by integrating with respect to t using Youngrsquosinequality we obtain
1113946t
0E vϵ(r)
p
Hdrle cp(t)y
p
H + cp 1113946t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889 +
cp
ϵ1113946
t
01 + E uϵ(r)
p
H+ E Γ2ϵ(r)
p
H+ E Ψ2ϵ(r)
p
H1113872 1113873dr
times 1113946t
0e
minus (αp2ϵ)rdrle cp(t) 1 +yp
H1113872 1113873 + cp 1113946t
0E uϵ(r)
p
Hdr + 1113946
t
0E Γ2ϵ(r)
p
Hdr + 1113946
t
0E Ψ2ϵ(r)
p
Hdr1113888 1113889
(48)
According to the BurkholderndashDavisndashGundy inequalityby proceeding as Proposition 42 in [12] we can easily get
1113946t
0E Γ2ϵ(r)
p
Hdrle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(49)
Concerning the stochastic term Ψ2ϵ(t) using Kunitarsquosfirst inequality we have
E Ψ2ϵ(t)
p
Hle cpE
1ϵ
1113946t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
01113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2G2 r uϵ(r) vϵ(r) z( 1113857
p
Hv2(dz)dr
lecp
ϵp2E1113946
t
0e
minus (α(pminus 2)ϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873dr 1113946
t
0e
minus (αpϵ)(pminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵE1113946
t
0e
minus (αpϵ)(tminus r) 1 + uϵ(r)
p
H+ vϵ(r)
p
H1113872 1113873drle
cp
ϵ1113946
t
0e
minus (αp2ϵ)(tminus r) 1 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(50)
By integrating with respect to t both sides and usingYoungrsquos inequality we have
1113946t
0E Ψ2ϵ(r)
p
Hdrle
cp
ϵ1113946
t
0e
minus (αp2ϵ)rdr middot 1113946t
01 + E uϵ(r)
p
H1113872
+ E vϵ(r)
p
H1113873drle cp(t) 1113946
t
01 + E uϵ(r)
p
H+ E vϵ(r)
p
H1113872 1113873dr
(51)
Substituting (49) and (51) into (48) we get
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + 1113946t
0E uϵ(r)
p
Hdr1113888 1113889
+ cpT(t) 1113946t
0E vϵ(r)
p
Hdr
(52)
As cp(0) 0 and cp(t) is a continuous increasingfunction we can fix t0 gt 0 such that for any tle t0 we havecp(t)le 12 so
Complexity 7
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
1113946t
0E vϵ(r)
p
Hdrle cp(t) 1 +y
p
H + E suprisin[0t]
uϵ(r)
p
H1113888 1113889 t isin 0 t01113858 1113859
(53)
Due to (41) and (53) for any t isin [0 t0] we can get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT(t)E suprisin[0t]
uϵ(r)
p
H+ cpT 1113946
t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr
(54)
Similarly we also can fix 0lt t1 le t0 such that for anytle t1 we have cpT(t)le 12 so
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873
+ cpT 1113946t
0E sup
σisin[0r]
uϵ(σ)
p
Hdr t isin 0 t11113858 1113859
(55)
According to the Gronwall inequality we get
E suprisin[0t]
uϵ(r)
p
Hle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(56)
Substituting (56) into (53) it yields
1113946t
0E vϵ(r)
p
Hdrle cpT 1 +x
p
H +yp
H1113872 1113873 t isin 0 t11113858 1113859
(57)
For any pge 2 by repeating this in the intervals[t1 2t1] [2t1 3t1] etc we can easily get (31) Substituting(31) into (41) and using the Gronwall inequality again we get(30) Using the Holder inequality we can estimate (30) and(31) for p 1
Lemma 2 Under (A1)ndash(A4) there exists θgt 0 such that forany Tgt 0 pge 1 x isin D((minus A1)
θ) with θ isin [0 θ) and y isin Hthere exist a positive constant cpθT gt 0 such that
supϵisin(01]
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (58)
Proof Assuming that x isin D((minus A1)θ)(θ ge 0) for any
t isin [0 T] we have
uϵ(t) U1(t 0)x + ψ1 uϵ 0( 1113857(t)
+ 1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t
0U1(t r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t
01113946Z
U1(t r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz)
(59)
Concerning the second term ψ1(uϵ 0)(t) we get
ψ1 uϵ 0( 1113857(t)
p
θ le cp 1113946t
0minus A1( 1113857
θe
c1(tr)A1L1(r)uϵ(r)dr
p
H
le cpθ 1113946t
0(t minus r)
minus θL1(r)uϵ(r)
Hdr1113888 1113889
p
le cpθ suprisin[0T]
uϵ(r)
p
H1113946
t
0(t minus r)
minus θdr1113888 1113889
p
le cpθT 1 +xp
H +yp
H1113872 1113873
(60)
For any pge 2 according to the proof of Proposition 43in [12] and thanks to (30) and (31) it is possible to show thatthere exists a 1113957θ ge 0 such that for any θle 1113957θand12 we have
E suptisin[0T]
1113946t
0U1(t r)B1 r uϵ(r) vϵ(r)( 1113857dr
p
θle cpθT 1 +x
p
H +yp
H1113872 1113873
(61)
E suptisin[0T]
Γ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (62)
Concerning the stochastic term Ψ1ϵ(t) using the fac-torization argument we have
Ψ1ϵ(t) cθ 1113946t
0(t minus r)
θminus 1e
c1(tr)A1ϕϵθ(r)dr (63)
where
ϕϵθ(r) 1113946r
01113946Z
(r minus σ)minus θ
ec1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
(64)
Next for any pge 2 let 1113954θ ((14) minus (12p))and12p forany θle 1113954θ according to (A3) and Lemma 1 using Kunitarsquosfirst inequality and the Holder inequality we get
8 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
Ψ1ϵ(t)
p
θ le cθ 1113946t
0(t minus r)
θminus 1 ϕϵθ(r)
θdr1113888 1113889
p
le cθ suprisin[0t]
ϕϵθ(r)
p
θ 1113946t
0(t minus r)
θminus 1dr1113888 1113889
p
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus θ
minus A1( 1113857θe
c1(rσ)A1G1 σ uϵ(σ) z( 1113857 1113957N1(dσ dz)
p
H
le cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus 2θ
(r minus σ)minus 2θ
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)dσ1113874 1113875p2
+ cpθT suprisin[0t]
1113946r
01113946Z
(r minus σ)minus pθ
(r minus σ)minus pθ
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)dσ le cpθT sup
risin[0t]
1113946r
01113946Z
G1 σ uϵ(σ) z( 1113857
2H
v1(dz)1113874 1113875p2
dσ1113890
times 1113946r
0(r minus σ)
minus 4pθ(pminus 2)dσ1113874 1113875(pminus 2)2
1113891 + cpθT suprisin[0t]
1113946r
0(r minus σ)
minus 2pθdσ middot supσisin[0r]
1113946Z
G1 σ uϵ(σ) z( 1113857
p
Hv1(dz)1113888 11138891113890 1113891
le cpθT 1 + supσisin[0T]
uϵ(σ)
p
H1113888 1113889
(65)
So due to (30) we get
E suptisin[0T]
Ψ1ϵ(t)
p
θ le cpθT 1 +xp
H +yp
H1113872 1113873 (66)
Hence if we choose θ ≔ 18and1113957θand1113954θ thanks to (60)ndash(62)and (66) for any pge 2 and θlt θ we get
E suptisin[0T]
uϵ(t)
p
θ le cpθT 1 +xp
θ +yp
H1113872 1113873 (67)
Using the Holder inequality we can estimate (58) forp 1
Lemma 3 Under (A1)ndash(A4) for any θ isin [0 θ) and 0le hle 1there exists β(θ)gt 0 such that for anyTgt 0 pge 1 x isin D((minus A1)
θ) y isin H and t isin [0 T] it holds
supϵisin(01]
E uϵ(t) minus uϵ(t + h)
p
H
le cpθT hβ(θ)p
+ h1113872 1113873 1 +xp
θ +yp
H1113872 1113873
(68)
Proof For any tge 0 with t t + h isin [0 T] we have
uϵ(t + h) minus uϵ(t) U1(t + h t) minus I( 1113857uϵ(t) + ψ1 uϵ t( 1113857(t + h) + 1113946t+h
tU1(t + h r)B1 r uϵ(r) vϵ(r)( 1113857dr
+ 1113946t+h
tU1(t + h r)F1 r uϵ(r)( 1113857dw
Q1(r)
+ 1113946t+h
t1113946Z
U1(t + h r)G1 r uϵ(r) z( 1113857 1113957N1(dr dz) ≔ 11139445
i1I
it
(69)
By proceeding as the proof of Proposition 44 in [12] and(60) fix θ isin [0 θ) for any pge 1 and it is possible to showthat
E I1t
p
Hle cpθTh
θp 1 +xp
θ +yp
H1113872 1113873
E I2t
p
Hle cpTh
pminus 1 1 +xp
θ +yp
H1113872 1113873
E I3t
p
Hle cpTh
p 1 +xp
θ +yp
H1113872 1113873
E I4t
p
Hle cpTh
(pminus 2)2minus β1 ρ1minus 2( )ρ1( )(p2) 1 +xp
θ +yp
H1113872 1113873
(70)
According to the proof of (39) using the Holder in-equality and (30) we have
E I5t
p
Hle cpE 1113946
t+h
t1 + uϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+ cpE1113946t+h
t1 + uϵ(r)
p
H1113872 1113873dr le cp h
(pminus 2)2+ 11113872 1113873
middot 1113946t+h
t1 + E uϵ(r)
p
H1113872 1113873dr
le cpT hp
2 + h1113874 1113875 1 +xp
H +yp
H1113872 1113873
(71)
en if we take pgt 1 such that
β1 ρ1 minus 2( 1113857
ρ1
p
p minus 2lt 1 (72)
We can get
Complexity 9
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
E uϵ(t + h) minus uϵ(t)
p
H
le cpT h(1minus 1p )p
+ hp
+ h(12)minus (1p)minus β1 ρ1minus 2( )2ρ1( )p
+ hp2
+ h1113874 1113875
times 1 +xp
H +yp
H1113872 1113873 + cpθThθp 1 +x
p
θ +yp
H1113872 1113873
(73)
As we are assuming |h|le 1 (68) follows for any pgep bytaking
β(θ) ≔ min θ 1 minus1p
112
minus1p
minusβ1 ρ1 minus 2( 1113857
2ρ112
1113896 1113897 (74)
From the Holder inequality we can estimate (68) forpltp
E uϵ(t + h) minus uϵ(t)
p
Hle E uϵ(t + h) minus uϵ(t)
p
H1113876 1113877
pp
(75)
So we have (68)According to the above lemma we get that for everyϵ isin (0 1] the function uϵ(t) is uniformly bounded aboutt isin [0 T] and it is also equicontinuous at every point oft isin [0 T] In view of eorem 123 in [43] we can infer thatthe set uϵ1113864 1113865ϵisin(01] is relatively compact in D([0 T]H) Inaddition according to eorem 132 in [43] and the abovelemma by using Chebyshevrsquos inequality we also can get thatthe family of probability measures L(uϵ)1113864 1113865ϵisin(01] is tight inP(D([0 T]H))
4 An Evolution Family of Measures for theFast Equation
For any frozen slow component x isin H any initial conditiony isin H and any s isin R we introduce the following problem
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) v(s) y
(76)
where
wQ2(t)
wQ21 (t)
wQ22 (minus t)
⎧⎨
⎩
if tge 0
if tlt 0
1113957N2prime(t z)
1113957N1prime(t z)
1113957N3prime(minus t z)
⎧⎪⎪⎨
⎪⎪⎩
if tge 0
if tlt 0
(77)
for two independent Q2-Wiener processes wQ21 (t) and
wQ22 (t) and two independent compensated Poisson mea-
sures 1113957N1prime(dt dz) and 1113957N3prime(dt dz) with the same Levymeasure are both defined as in Section 2
According to the definition of the operator ψα2(middot s) weknow that the mapping ψα2(middot s) C([s T] H)⟶C([s T]H) is a linearly bounded operator and it is Lip-schitz continuous Hence we have that for anyx y isin H pge 1 and sltT there exists a unique mild solution
vx(middot s y) isin Lp(Ω D([0 T]H)) (see [28]) in the followingform
vx(t s y) Uα2(t s)y + ψα2 v
x(middot s y) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s y)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s y)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s y) z( 1113857 1113957N2prime(dr dz)
(78)
Moreover if the space D(RH) endowed with the to-pology of uniform convergence on bounded intervals anFt1113864 1113865tisinR-adapted process vx isin Lp(Ω D([0 T]H)) is a mildsolution of the equation
dv(t) A2(t) minus α( 1113857v(t) + B2(t x v(t))1113858 1113859dt
+ F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz)
(79)
where t isin R en for every slt t we have
vx(t) Uα2(t s)v
x(s) + ψα2 v
x s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r) z( 1113857 1113957N2prime
(dr dz)
(80)
In the following for any x isin H and any adapted processv we set
Γα(v s)(t) ≔ 1113946t
sUα2(t r)F2(r x v(r))dw
Q2(r) tgt s
Ψα(v s)(t) ≔ 1113946t
s1113946Z
Uα2(t r)G2(r x v(r) z) 1113957N2prime(dr dz) tgt s
(81)
For any 0lt δ lt α and any v1 and v2 with slt t by pro-ceeding as in the proof of Lemma 71 in [44] it is possible toshow that there exists pgt 1 such that for any pgep we have
suprisin[st]
eδp(rminus s)
E Γα v1 s( 1113857(r) minus Γα v2 s( 1113857(r)
p
H
le cp1L
p
f2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(82)
where Lf2is the Lipschitz constant of f2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 and slt tFor the stochastic term Ψα(v s)(t) using Kunitarsquos first
inequality ([26] eorem 4423) we get
10 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
E Ψα v1 s( 1113857(t) minus Ψα v2 s( 1113857(t)
p
Hle cpE 1113946
t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus α(tminus r)
ec2(tr)A2 G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 11138571113858 1113859
p
Hv2prime(dz)dr
le cpE 1113946t
s1113946Z
eminus 2α(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
2H
v2prime(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
eminus αp(tminus r)
G2 r x v1(r) z( 1113857 minus G2 r x v2(r) z( 1113857
p
Hv2prime(dz)dr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)e
minus 2δ(tminus s)e2δ(rminus s)
E v1(r) minus v2(r)
2Hdr1113888 1113889
p2
+ cpLpg2
1113946t
se
minus p(αminus δ)(tminus r)e
minus δp(tminus s)eδp(rminus s)
E v1(r) minus v2(r)
p
Hdr
le cpLpg2
1113946t
se
minus 2(αminus δ)(tminus r)dr1113888 1113889
p2
+ 1113946t
se
minus p(αminus δ)(tminus r)dr⎡⎣ ⎤⎦ times eminus δp(tminus s) sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
le cp1L
pg2
(α minus δ)cp2e
minus δp(tminus s) suprisin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(83)
so
suprisin[st]
eδp(rminus s)
E Ψα v1 s( 1113857(r) minus Ψα v2 s( 1113857(r)
p
Hle cp1
Lpg2
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H (84)
where Lg2is the Lipschitz constant of g2 and cp1 and cp2 are
two suitable positive constants independent of αgt 0 andslt t
Moreover using (A4) we can show that
suprisin[st]
eδp(rminus s)
E Γα(v s)(r)
p
Hle cp1
Mp
f2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(85)
suprisin[st]
eδp(rminus s)
E Ψα(v s)(r)
p
Hle cp1
Mpg2
(α minus δ)cp2
suprisin[st]
eδp(rminus s) 1 + Ev(r)
p
H1113872 1113873
(86)
where Mf2andMg2
are the linear growth constants off2 andg2 and cp1 and cp2 are two suitable positive constantsindependent of αgt 0 and slt t
For any fixed adapted process v let us introduce theproblem
dρ(t) A2(t) minus α( 1113857ρ(t)dt + F2(t x v(t))dωQ2(t)
+ 1113946Z
G2(t x v(t) z) 1113957N2prime(dt dz) ρ(s) 0
(87)
We denote that its unique mild solution is ρα(v s) ismeans that ρα(v s) solves the equation
ρα(v s)(t) ψα2 ρα(v s) s( 1113857(t) + Γα(v s)(t)
+ Ψα(v s)(t) slt tltT(88)
Due to (82) and (84) using the same argument as (58) in[25] it is easy to prove that for any process v1 v2 and0lt δ lt α we have
suprisin[st]
eδp(rminus s)
E ρα v1 s( 1113857(r) minus ρα v2 s( 1113857(r)
p
H
le cp1Lp
(α minus δ)cp2times sup
risin[st]
eδp(rminus s)
E v1(r) minus v2(r)
p
H
(89)
where L max Lb2 Lf2
Lg21113966 1113967
Similarly thanks to (85) and (86) for any process v and0lt δ lt α we can prove that
suprisin[st]
eδp(rminus s)
E ρα(v s)(r)
p
Hle cp1
Mp
(α minus δ)cp2
suprisin[st]
eδp(rminus s)
E 1 +v(r)p
H1113872 1113873
(90)
where M max Mb2 Mf2
Mg21113966 1113967
Complexity 11
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
Lemma 4 Under (A1)ndash(A4) there exists δ gt 0 such that forany x y isin H and pge 1
E vx(t s y)
p
Hle cp 1 +x
p
H + eminus δp(tminus s)
yp
H1113872 1113873 slt t
(91)
Proof We set Λα(t) ≔ vx(t s y) minus ρα(t) whereρα(t) ρα(vx(middot s y) s)(t) is the solution of problem (87)with v vx(middot s y) Using Youngrsquos inequality we have1p
ddtΛα(t)
p
Hlelang A2(t) minus α( 1113857Λα(t)Λα(t)rangH Λα(t)
pminus 2H
+langB2 t xΛα(t) + ρα(t)( 1113857 minus B2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2H
+langB2 t x ρα(t)( 1113857Λα(t)rangH Λα(t)
pminus 2Hle minus α Λα(t)
p
H
+ c Λα(t)
p
H+ c 1 +xH + ρα(t)
H1113872 1113873 Λα(t)
pminus 1H
le minus α Λα(t)
p
H+ cp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(92)
Because α is large enough we can find η α minus cp gt 0such that
ddtΛα(t)
p
Hle minus ηp Λα(t)
p
H+ cp 1 +x
p
H + ρα(t)
p
H1113872 1113873
(93)
According to the Gronwall inequality we have
Λα(t)
p
Hle e
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(94)
So for any pge 1
vx(t s y)
p
Hle cp ρα(t)
p
H+ cpe
minus ηp(tminus s)y
p
H + cp 1 +xp
H1113872 1113873
+ cp 1113946t
se
minus ηp(tminus r) ρα(r)
p
Hdr
(95)
Fixing 0lt δ lt η according to (90) we get
eδp(tminus s)
E vx(t s y)
p
Hle cpe
δp(tminus s)E ρα(t)
p
H+ cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873 + cp 1113946t
seδp(rminus s)
E ρα(r)
p
Hdr
le cp1Mp
(α minus δ)cp2sup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873 + cpe
p(δminus η)(tminus s)y
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ cp1Mp
(α minus δ)cp21113946
t
ssup
risin[st]
eδp(rminus s) 1 + E v
x(r s y)
p
H1113872 1113873dr
(96)
Taking α1 (2cp1Mp)1cp2 + δ when αge α1 we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
+ 1113946t
ssup
risin[st]
eδp(rminus s)
E vx(r s y)
p
Hdr
(97)
Due to the Gronwall lemma we have
suprisin[st]
eδp(rminus s)
E vx(r s y)
p
Hle cpy
p
H + cpeδp(tminus s) 1 +x
p
H1113872 1113873
(98)
Hence we get (91)
Lemma 5 Under (A1)ndash(A4) for any t isin R and x y isin H forall pge 1 there exists ηx(t) isin Lp(ΩH) such that
lims⟶minus infin
E vx(t s y) minus ηx
(t)
p
H 0 (99)
Moreover for any pge 1 there exists some δp gt 0 suchthat
E vx(t s y) minus ηx
(t)
p
Hle cpe
minus δp(tminus s) 1 +xp
H +yp
H1113872 1113873
(100)
Finally ηx is a mild solution in R of (79)
Proof Fix hgt 0 and define
ρ(t) vx(t s y) minus v
x(t s minus h y) slt t (101)
We know that ρ(t) is the unique mild solution of theproblem
12 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
dρ(t) A2(t) minus α( 1113857ρ(t) + B2 t x vx(t s y)( 1113857 minus B2 t x vx(t s minus h y)( 11138571113858 1113859dt
+ F2 t x vx(t s y)( 1113857 minus F2 t x vx(t s minus h y)( 11138571113858 1113859dwQ2(t)
+1113946Z
G2 t x vx(t s y)( 1113857 minus G2 t x vx(t s minus h y)( 11138571113858 1113859 1113957N2prime(dt dz)
ρ(s) y minus vx(s s minus h y)
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩
ρ(t) Uα2(t s) y minus vx(s s minus h y)( 1113857 + ψα2(ρ s)(t)
+ 1113946t
sUα2(t r) B2 r x v
x(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
+ 1113946t
sUα2(t r) F2 r x v
x(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r) G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime
(dr dz)
(102)
Multiply both sides of the above equation by eδp(tminus s)Because α large enough according to Lemma 24 in [25] wehave
eδp(tminus s)
Eρ(t)p
H le cpe(δminus α)p(tminus s)
E ec2(ts)A2 y minus v
x(s s minus h y)( 1113857
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
B2 r x vx(r s y)( 1113857 minus B2 r x v
x(r s minus h y)( 11138571113858 1113859dr
p
H
+ cpE 1113946t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859dw
Q2(r)
p
H
+ cpE 1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859 1113957N2prime(dr dz)
p
H
≔ 11139444
i1I
it
(103)
According to Lemma 31 in [12] we know that for anyJ isinL(Linfin(D)H)capL(H L1(D)) with J Jlowast and sge 0 wehave
esAi JQi
22 leKis
minus βi ρi minus 2( )ρi eminus α ρi+2( )ρi( )s
J2L Linfin(D)H( )
(104)
where
Ki βi
e1113888 1113889
βi ρi minus 2( )ρi
ζ ρi minus 2( )ρi
i κ2ρi
i (105)
Taking pgt 1 such that pβ2(ρ2 minus 2)[ρ2(p minus 2)]lt 1en using the BurkholderndashDavisndashGundy inequality andKunitarsquos first inequality we can get that for any pgep and0lt δ lt α it yields
I2t le cpL
p
b2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se
(δminus α)(tminus r)dr1113888 1113889
p
le cp
Lp
b2
(α minus δ)p suprisin[st]
eδp(rminus s)
Eρ(r)p
H (106)
Complexity 13
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
I3t le cpE 1113946
t
se
c2(tr)A2e(δminus α)(tminus r)
eδ(rminus s)
F2 r x vx(r s y)( 1113857 minus F2 r x v
x(r s minus h y)( 11138571113858 1113859Q2
2
2dr1113888 1113889
p2
le cpLp
f2K
p22 1113946
t
se2(δminus α)(tminus r)
c2(t r)minus β2 ρ2minus 2( )ρ2( )e
minus α ρ2+2( )ρ2c2(tr)dr1113888 1113889
p2
times suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946t
se2(δminus α)(tminus r)
c0(t minus r)1113858 1113859minus β2 ρ2minus 2( )ρ2( )dr1113888 1113889
p2
le cpLp
f2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0r
minus β2 ρ2minus 2( )ρ2( )(p(pminus 2))dr1113888 1113889
(pminus 2)2
1113946tminus s
0e
minus p(αminus δ)rdr1113888 1113889
le cp
Lp
f2
α minus δsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
I4t le cpE 1113946
t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
2
Hv2prime
(dz)dr1113888 1113889
p2
+ cpE1113946t
s1113946Z
ec2(tr)A2e
(δminus α)(tminus r)eδ(rminus s)
G2 r x vx(r s y)( 1113857 minus G2 r x v
x(r s minus h y)( 11138571113858 1113859
p
Hv2prime
(dz)dr
le cpLpg2
suprisin[st]
eδp(rminus s)
Eρ(r)p
H middot 1113946tminus s
0e
minus 2(αminus δ)rdr1113888 1113889
p2
+ 1113946tminus s
0e
minus p(αminus δ)rdr⎡⎣ ⎤⎦le cp
Lpg2
(α minus δ)cpsup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(107)
Hence we have
suprisin[st]
eδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H
+ cp1Lp
(α minus δ)cp2sup
risin[st]
eδp(rminus s)
Eρ(r)p
H
(108)
erefore for αgt 0 large enough we can find 0lt δp lt αsuch that
cp1L
α minus δp1113872 1113873cp2lt 1 (109)
is implies that
suprisin[st]
epδp(rminus s)
Eρ(r)p
H le cp y minus vx(s s minus h y)
p
H (110)
Let δp pδp thanks to Lemma 4 we have
E vx(t s y) minus v
x(t s minus h y)
p
H
le cpeminus δp(tminus s)
y minus vx(s s minus h y)
p
H
le cpeminus δp(tminus s) 1 +x
p
H +yp
H + eminus δph
yp
H1113872 1113873
(111)
Because Lp(ΩH) has completeness for any pgep lets⟶ minus infin in (111) there exists ηx(t) isin Lp(ΩH) such that(99) holds en if we let h⟶infin in (111) we obtain (100)Using the Holder inequality we can get (99) and (100) holdsfor any pltp
If we take y1 y2 isin H use the same arguments forvx(t s y1) minus vx(t s y2) slt t we have
E vx
t s y1( 1113857 minus vx
t s y2( 1113857
p
Hle cpe
minus δp(tminus s)y1 minus y2
p
H slt t
(112)
Let s⟶ minus infin this means that the limit ηx(t) does notdepend on the initial condition y isin H
Finally we prove that ηx(t) is a mild solution of (79)Due to the limit ηx(t) does not depend on the initialcondition we can let initial condition y 0 For any slt t
and hgt 0 we have
vx(t s minus h 0) Uα2(t s)v
x(s s minus h 0) + ψα2 v
x(middot s minus h 0) s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x v
x(r s minus h 0)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x v
x(r s minus h 0)( 1113857dw
Q2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x vx(r s minus h 0) z( 1113857 1113957N2prime
(dr dz)
(113)
Let h⟶infin on both sides due to (99) we can get forany slt t have
ηx(t) Uα2(t s)ηx
(s) + ψα2 ηx s( 1113857(t)
+ 1113946t
sUα2(t r)B2 r x ηx
(r)( 1113857dr
+ 1113946t
sUα2(t r)F2 r x ηx
(r)( 1113857dwQ2(r)
+ 1113946t
s1113946Z
Uα2(t r)G2 r x ηx(r) z( 1113857 1113957N2prime
(dr dz)
(114)
is means that ηx(t) is a mild solution of (79)For any t isin R and x isin H we denote that the law of the
random variable ηx(t) is μxt and we introduce the transition
evolution operator
Pxstφ(y) Eφ v
x(t s y)( 1113857 slt t y isin H (115)
14 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
where φ isinBb(H)Due to (91) and (99) for any pge 1 we have
suptisinR
E ηx(t)
p
Hle cp 1 +x
p
H1113872 1113873 x isin H (116)
so that
suptisinR
1113946H
yp
Hμxt (dy)le cp 1 +x
p
H1113872 1113873 x isin H (117)
According to the above conclusion by using the samearguments as ([25] Proposition 53) we know that thefamily μx
t1113864 1113865tisinR defines an evolution system of probabilitymeasures on H for equation (76) is means that for anyt isin R μx
t is a probability measure on H and it holds that
1113946H
Pxstφ(y)μx
s (dy) 1113946Hφ(y)μx
t (dy) slt t (118)
for every φ isin Cb(H) Moreover we also have
Pxstφ(y) minus 1113946
Hφ(y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868le ce
minus δ1(tminus s) 1 +xH( 1113857 (119)
Lemma 6 Under (A1)ndash(A4) the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (120)
is tight in P(D([0 T]H))
Proof According to the definition of vx(middot s 0) for any tgt s
and hgt 0 we have
vx(t + h s 0) minus v
x(t s 0) Uα2(t + h t) minus I1113872 1113873v
x(t s 0)
+ ψα2 vx(middot s 0) t( 1113857(t + h) + 1113946
t+h
tUα2(t + h r)
middot B2 r x vx(r s 0)( 1113857dr + 1113946
t+h
tUα2(t + h r)
middot F2 r x vx(r s 0)( 1113857dw
Q2(r) + 1113946t+h
t
middot 1113946Z
Uα2(t + h r)G2 r x vx(r s 0) z( 1113857 1113957N2prime
(dr dz)
(121)Due to the assumption (A4) we know that the mappings
B2 F2 andG2 are linearly growing For any tgt s 0lt hlt 1and x isin H analogous to the proof of Lemma 3 and using theestimate (91) we can get
E vx(t + h s 0) minus v
x(t s 0)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873
(122)
where κ(p) is a function of p and satisfies κ(p)gt 0
From Lemma 5 we know that the limit ηx(t) does notdepend on the initial condition and we can get
E ηx(t + h) minus ηx
(t)
p
HleE ηx
(t + h) minus vx(t + h s 0)
p
H
+ E vx(t + h s 0) minus v
x(t s 0)
p
H+ E v
x(t s 0) minus ηx
(t)
p
H
(123)
en thanks to (100) and (122) let s⟶infin we have
E ηx(t + h) minus ηx
(t)
p
Hle cp h
κ(p)+ h1113872 1113873 1 +x
p
H1113872 1113873 (124)
Moreover we have proved that the family μxt1113864 1113865tisinR defines
an evolution system of probability measures for (76) Byusing Chebyshevrsquos inequality and (116) it yields
limn⟶infin
μxt ηx(t) ηx(t)
Hge n1113872 1113873le lim
n⟶infin
E ηx(t)
2H
n2
le limn⟶infin
c21 +x2H
n2 0
(125)
In addition for any tgt s hgt 0 and fixed ϵgt 0 accordingto the (124) we have
limh⟶0
μxt ηx(t) ηx(t + h) minus ηx(t)
Hge ϵ1113872 1113873
le limh⟶0
E ηx(t + h) minus ηx(t)
2H
ϵ2
le limh⟶0
c2hκ(2) + h( 1113857 1 +x2H1113872 1113873
ϵ2 0
(126)
According to eorem 132 in [43] (125) and (126)imply that the family of measures
Λ ≔ μxt t isin R x isin H1113864 1113865 (127)
is tight in P(D([0 T]H))
In order to get the averaged equation wemust ensure theexistence of the averaged coefficient B1 So we need theevolution family of measures satisfying some nice propertiesWe give the following assumption
(A5)
(a) e functions c2 R⟶ (0infin) andl2 R times O⟶ Rd are periodic with the sameperiod
(b) e families of functions
B1R ≔ b1(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
B2R ≔ b2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
FR ≔ f2(middot ξ σ) ξ isin O σ isin BR2(R)1113864 1113865
GR ≔ g2(middot ξ σ z) ξ isin O σ isin BR2(R) z isin Z1113864 1113865
(128)
are uniformly almost periodic for any Rgt 0
Remark 4 Similar with the proof of Lemma 62 in [25] weget that under (A5) for any Rgt 0 the families of functions
B1(middot x y) (x y) isin BHtimesH(R)1113864 1113865
B2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
F2(middot x y) (x y) isin BHtimesH(R)1113864 1113865
G2(middot x y z) (x y z) isin BHtimesH(R) times Z1113864 1113865
(129)
Complexity 15
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
are uniformly almost periodicAs we know above A2(middot) is periodic Remark 4 holds
and the family of measures Λ is tight By proceeding as [45]we can prove that the mapping
t⟼μxt t isin R x isin H (130)
is almost periodic
5 The Averaged Equation
Lemma 7 Under (A1)ndash(A5) for any compact set K sub H thefamily of functions
t isin R⟼1113946H
B1(t x y)μxt (dy) x isin K1113882 1113883 (131)
is uniformly almost periodic
Proof AsK is a compact set inH so it is bounded and thereexist some Rgt 0 such thatK sub BH(R) at is for any x isin K
we have xleR
Now let us define
Φ(t x) 1113946H
B1(t x y)μxt (dy) (t x) isin R times H (132)
en for any t τ isin R n isin N and x isin K according toassumption (A3) and equation (117) we can get
|Φ(t + τ x) minus Φ(t x)|le 1113946yH le n
B1(t + τ x y)μxt+τ(dy) minus 1113946
yH le nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946yH gtn
B1(t + τ x y)μxt+τ(dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868+ 1113946
yH gt nB1(t + τ x y)μx
t (dy)
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ 1113946H
B1(t + τ x y)μxt (dy) minus 1113946
HB1(t x y)μx
t (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le supxisinKyH le n
B1(t + τ x y)
H 1113946H
μxt+τ minus μx
t( 1113857(dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
+c(1 +x)
n+ 1113946
HB1(t + τ x y) minus B1(t x y)
Hμ
xt (dy)
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
(133)
Fixed some n such that c(1 + x)nle ϵ3 en for anyt isin R due to the mapping t⟼μx
t and the families offunctions B1(middot x y) are almost periodic we can find someτ isin R such that
|Φ(t + τ x) minus Φ(t x)|lt ϵ (134)
So the function Φ(middot x) is almost periodic for any x isin KBy preceding as in Lemma 5 we can get that under
(A1)ndash(A4) for any fixed x1 x2 isin K and pge 1 there existscp gt 0 such that
supsltt
E vx1(t s 0) minus v
x2(t s 0)
p
Hle cp x1 minus x2
p
H (135)
According to (99) and (100) let s⟶ 0 it yields
suptisinR
E ηx1(t) minus ηx2(t)
p
Hle cp x1 minus x2
p
H (136)
Hence thanks to (A3) and (136) we can conclude thatfor any x1 x2 isin K we have
Φ tx1( 1113857 minus Φ tx2( 1113857
H le E B1 tx1ηx1(t)( 1113857 minus B1 tx2η
x2(t)( 1113857
H
le c x1 minus x2
H+ E ηx1(t) minus ηx2(t)
2H
1113874 111387512
1113888 1113889
le c x1 minus x2
H
(137)
is means that the family of functions Φ(t middot) t isin R
is equicontinuous about x
In view of eorem 210 in [32] the above conclusionsindicate that Φ(middot x) x isin K is uniformly almost periodicHence the proof is complete
According to eorem 34 in [25] we define
B1(x) ≔ limT⟶infin
1T
1113946T
01113946H
B1(t x y)μxt (dy)dt x isin H
(138)
And thanks to (A3) and (117) we have that
B1(x)
Hle c 1 +xH( 1113857 (139)
Lemma 8 Under (A1)ndash(A5) for any Tgt 0 s σ isin R andx y isin H
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857dt minus B1(x)
H
lec
T1 +xH +yH( 1113857 + α(T x)
(140)
for some mapping α [0infin) times H⟶ [0infin) such that
16 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
supTgt0
α(T x)le c 1 +xH( 1113857 x isin H (141)
And for any compact set K sub H we have
limT⟶infin
supxisinK
α(T x) 0 (142)
Proof We denote
ψxB1(t y) ≔ B1(t x y) minus 1113946
HB1(t x w)μx
t (dw) (143)
So
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt1113888 1113889
2
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857ψx
B1 t vx(t s y)( 11138571113858 1113859dtdr
2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857P
xrtψ
xB1 r v
x(r s y)( 11138571113960 1113961dtdr
le2
T2 1113946s+T
s1113946
s+T
rE ψx
B1 r vx(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
1113872 111387312dtdr
(144)
Due to (A3) (91) and (117) we have
E ψxB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682 le cE B1 r x v
x(r s y)( 1113857
2H
+ cE 1113946H
B1(r x w)
Hμxr (dw)1113874 1113875
2
le c 1 +x2H + E v
x(r s y)
2H
1113874 1113875
le c 1 +x2H + e
minus 2δ(rminus s)y
2H1113872 1113873
(145)
and according to (100) we get
E Pxrtψ
xB1 r v
x(r s y)( 1113857
111386811138681113868111386811138681113868111386811138682
E Pxrt B1 r x v
x(r s y)( 1113857 minus 1113946
HB1(r x w)μx
r (dw)1113876 1113877
2
H
E B1 t x vx(t s y) minus B1 t x ηx
(t)( 1113857( 1113857
2H
le cE vx(t s y) minus ηx
(t)
2Hle ce
minus δ2(tminus s) 1 +x2H +y
2H1113872 1113873
(146)
Let δ δ22 it follows
E1T
1113946s+T
sB1 t x v
x(t s y)( 1113857 minus 1113946
HB1(t x w)μx
t (dw)1113876 1113877dt
H
lec
T1 +xH +yH( 1113857 1113946
s+T
s1113946
s+T
re
minus δ(tminus s)dtdr1113888 1113889
12
lec
T1 +xH +yH( 1113857
(147)
Since the family of functions (131) is uniformly almostperiodic according to eorem 34 in [25] we can get thatthe limit
limT⟶infin
1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt (148)
converges to B1(x) uniformly with respect to s isin R andx isin K (K is a compact set in H) erefore if we define
α(T x) 1T
1113946s+T
s1113946H
B1(t x w)μxt (dw)dt minus B1(x)
H
(149)
we get the conclusion (142) Moreover thanks to the as-sumption (A3) equations (117) and (139) we also can get
α(T x)le1T
1113946s+T
s1113946H
B1(t x w)
Hμxt (dw)dt
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ B1(x)
Hle c 1 +xH( 1113857
(150)
Hence the proof is complete
Complexity 17
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
Now we introduce the averaged equation
du(t) A1(t)u(t) + B1(u(t))1113858 1113859dt + F1(t u(t))dwQ1(t)
+ 1113946Z
G1(t u(t) z) 1113957N1(dt dz) u(0) x isin H
(151)
Due to assumption (A3) we can easily get that themapping B1 is Lipschitz continuous So for any x isin H Tgt 0and pge 1 equation (151) admits a unique mild solutionu isin Lp(Ω D([0 T]H))
6 Averaging Principles
In this section we prove that the slow motion uϵ convergesto the averaged motion u as ϵ⟶ 0
Theorem 1 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isin [0 θ)) and y isin H for any Tgt 0 and ηgt 0 we have
limϵ⟶0
P suptisin[0T]
uϵ(t) minus u(t)
Hgt η1113888 1113889 0 (152)
where u is the solution of the averaged (151)
Proof For any h isin D(A1)capLinfin(O) we have
languϵ(t) hrangH langx hrangH + 1113946t
0langA1(r)uϵ(r) hrangHdr
+ 1113946t
0langB1 uϵ(r)( 1113857 hrangHdr
+ lang1113946t
0F1 r uϵ(r)( 1113857dw
Q1(r) hrangH
+ lang1113946t
01113946Z
G1 r uϵ(r) z( 1113857 1113957N1(dr dz) hrangH
+ Rϵ(t)
(153)
where
Rϵ(t) ≔ 1113946t
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
(154)
According to the proof in Section 3 we know that thefamily L(uϵ)1113864 1113865ϵisin(01] is tight in P(D([0 T]H)) Hence inorder to prove eorem 1 it is sufficient to provelimϵ⟶0Esuptisin[0T]|Rϵ(t)| 0
For any ϵgt 0 and some deterministic constant δϵ gt 0 wedivide the interval [0 T] in subintervals of the size δϵ In eachtime interval [kδϵ (k + 1)δϵ] k 0 1 lfloorTδϵrfloor we definethe following auxiliary fast motion 1113954vϵ
d1113954vϵ(t) 1ϵ
A2(t) minus α( 11138571113954vϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858 1113859dt
+1ϵ
radic F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857dωQ2(t)
+ 1113946Z
G2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 1113957Nϵ2(dt dz)
(155)
According to the definition of 1113954vϵ we know thatan analogous estimate to (31) holds So for any pge 1 wehave
1113946T
0E 1113954vϵ(t)
p
Hdtle cpT 1 +x
p
H +yp
H1113872 1113873 (156)
Lemma 9 Under (A1)ndash(A5) fixing x isin D((minus A1)θ)
(θ isint[n0q hθ)) and y isin H there exists a constant κgt 0 suchthat if
δϵ ϵlnϵminus κ
(157)
we have
limϵ⟶0
suptisin[0T]
E 1113954vϵ(t) minus vϵ(t)
p
H 0 (158)
Proof Let ϵgt 0 be fixed For k 0 1 lfloorTδϵrfloorand t isin [kδϵ (k + 1)δϵ] let ρϵ(t) be the solution of theproblem
dρϵ(t) 1ϵ
A2(t) minus α( 1113857ρϵ(t)dt +1ϵ
radic Kϵ(t)dωQ2(t)
+ 1113946Z
Hϵ(t z) 1113957Nϵ2(dt dz) ρϵ kδϵ( 1113857 0
(159)
where
Kϵ(t) ≔ F2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus F2 t uϵ(t) vϵ(t)( 1113857
Hϵ(t z) ≔ G2 t uϵ kδϵ( 1113857 1113954vϵ(t) z( 1113857 minus G2 t uϵ(t) vϵ(t) z( 1113857
(160)
We get
ρϵ(t) ψαϵ2 ρϵ kδϵ( 1113857(t) + Γϵ(t) + Ψϵ(t) t isin kδϵ (k + 1)δϵ1113858 1113859
(161)
where
Γϵ(t) 1ϵ
radic 1113946t
kδϵUαϵ2(t r)Kϵ(r)dw
Q2(r)
Ψϵ(t) 1113946t
kδϵ1113946Z
Uαϵ2(t r)Hϵ(r z) 1113957Nϵ2(dr dz)
(162)
If we denote Λϵ(t) ≔ 1113954vϵ(t) minus vϵ(t) and ϑϵ(t) ≔Λϵ(t) minus ρϵ(t) we have
dϑϵ(t) 1ϵ
A2(t) minus α( 1113857ϑϵ(t) + B2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 11138571113858
minus B2 t uϵ(t) vϵ(t)( 11138571113859dt
(163)
Due to Lemma 3 for αgt 0 large enough using Youngrsquosinequality we have
18 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
1p
ddt
ϑϵ(t)
p
Hle1ϵ lang c2(t)A2 + L2(t) minus α( 1113857ϑϵ(t) ϑϵ(t)rangH ϑϵ(t)
pminus 2H
+1ϵlangB2 t uϵ kδϵ( 1113857 1113954vϵ(t)( 1113857 minus B2 t uϵ(t) vϵ(t)( 1113857 ϑϵ(t)rangH ϑϵ(t)
pminus 2H
lec
ϵϑϵ(t)
p
Hminusαϵϑϵ(t)
p
H+
c
ϵuϵ kδϵ( 1113857 minus uϵ(t)
H ϑϵ(t)
pminus 1H
+c
ϵ1113954vϵ(t) minus vϵ(t)
H ϑϵ(t)
pminus 1H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵuϵ kδϵ( 1113857 minus uϵ(t)
p
H+
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
le minusα2ϵ
ϑϵ(t)
p
H+
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)pϵ + δϵ1113872 1113873 +
cp
ϵ1113954vϵ(t) minus vϵ(t)
p
H
(164)
Using the Gronwall inequality we get
ϑϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵ1113954vϵ(r) minus vϵ(r)
p
Hdr
(165)
By proceeding as Lemma 63 in [13] we prove that
E Γϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(166)
For the other stochastic term Ψϵ(t) using Kunitarsquos firstinequality and the Holder inequality we get
E Ψϵ(t)
p
Hle cpE
1ϵ
1113946t
kδϵ1113946Z
eminus (αϵ)(tminus r)
ec2(tr)ϵ( )A2Hϵ(r z)
2
Hv2(dz)dr1113888 1113889
p2
+cp
ϵE1113946
t
kδϵ1113946Z
eminus αϵ(tminus r)
eA2c2(tr)ϵ
Hϵ(r z)
p
Hv2(dz)dr
lecp
ϵp2E 1113946
t
kδϵe
minus (2αϵ)(tminus r)uϵ kδϵ( 1113857 minus uϵ(r)
2H
+ 1113954vϵ(r) minus vϵ(r)
2H
1113874 1113875dr1113888 1113889
p2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
Hdr
lecp
ϵp21113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr 1113946
t
kδϵe
minus (2αϵ)(ppminus 2)(tminus r)dr1113888 1113889
(pminus 2)2
+cp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1113946
t
kδϵE uϵ kδϵ( 1113857 minus uϵ(r)
p
H+ E 1113954vϵ(r) minus vϵ(r)
p
H1113872 1113873dr
lecp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873 +
cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(167)
According to Lemma 24 in [25] and equations(161)ndash(167) we obtain
E 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵ1 +x
p
θ +yp
H1113872 1113873 δβ(θ)p+1ϵ + δ2ϵ1113872 1113873
+cp
ϵ1113946
t
kδϵE 1113954vϵ(r) minus vϵ(r)
p
Hdr
(168)From the Gronwall lemma this meansE 1113954vϵ(t) minus vϵ(t)
p
Hle
cp
ϵδβ(θ)p+1ϵ + δ2ϵ1113872 1113873e
cpϵ( 1113857δϵ 1 +xp
θ +yp
H1113872 1113873
(169)
For t isin [0 T] selecting δϵ ϵlnϵminus κ then if we take
κlt β(θ)p(β(θ)p + 1 + cp)and1(2 + cp) we have (158)
Lemma 10 Under the same assumptions as inGeorem 1 forany Tgt 0 we have
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868 0 (170)
Proof According to the definition of B1 we get that themapping B1 H⟶ H is Lipschitz continuous Using as-sumption (A3) and Lemmas 3 and 9 we have
Complexity 19
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
limϵ⟶0
E suptisin[0T]
Rϵ(t)1113868111386811138681113868
1113868111386811138681113868le limϵ⟶0
E1113946T
0langB1 r uϵ(r) vϵ(r)( 1113857 minus B1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 hrangH
11138681113868111386811138681113868111386811138681113868dr
+ limϵ⟶0
E suptisin[0T]
1113946t
0langB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ(r)( 1113857 hrangHdr
1113868111386811138681113868111386811138681113868
1113868111386811138681113868111386811138681113868
le limϵ⟶0
cThH suprisin[0T]
E uϵ(r) minus uϵ lfloorrδϵrfloorδϵ( 1113857
H + suprisin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ lfloorrδϵrfloorδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
hH 1113944
lfloorTδϵrfloor
k01113946
(k+1)δϵ
kδϵE B1 uϵ kδϵ( 1113857( 1113857 minus B1 uϵ(r)( 1113857
Hdr
le limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)ϵ + δϵ1113872 1113873 + sup
risin[0T]
E 1113954vϵ(r) minus vϵ(r)
H1113890 1113891
+ limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
+ limϵ⟶0
cThH 1 +xθ +yH( 1113857 δβ(θ)+1ϵ + δ2ϵ1113872 1113873 lfloorTδϵrfloor + 1( 1113857
(171)
So we have to show that
limϵ⟶0
1113944
lfloorTδϵrfloor
k0E 1113946
(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 0
(172)
If we set ζϵ δϵϵ we have
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868 E 1113946
δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113954vϵ kδϵ + r( 1113857( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
E 1113946δϵ
0langB1 kδϵ + r uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(rϵ)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
δϵE1ζϵ
1113946ζϵ
0langB1 kδϵ + ϵr uϵ kδϵ( 1113857 1113957v
uϵ kδϵ( )vϵ kδϵ( )(r)1113874 1113875 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
(173)
where 1113957vuϵ(kδϵ)vϵ(kδϵ)(r) is the solution of the fast motion (76)with the initial datum given by vϵ(kδϵ) and the frozen slowcomponent given by uϵ(kδϵ) In addition the noises in (76)are independent of uϵ(kδϵ) and vϵ(kδϵ) According to theproof of Lemma 8 we get
E 1113946(k+1)δϵ
kδϵlangB1 r uϵ kδϵ( 1113857 1113954vϵ(r)( 1113857 minus B1 uϵ kδϵ( 1113857( 1113857 hrangHdr
11138681113868111386811138681113868111386811138681113868
11138681113868111386811138681113868111386811138681113868
le δϵc
ζϵ1 + E uϵ kδϵ( 1113857
H + E vϵ kδϵ( 1113857
H1113872 1113873hH
+ δϵhHEα ζϵ uϵ kδϵ( 1113857( 1113857
(174)
In Section 3 we proved that the family
uϵ kδϵ( 1113857 ϵ gt 0 k 0 lfloorTδϵrfloor1113864 1113865 (175)is tight en by proceeding as the proof of equation (821)in [25] we also can get that for any ηgt 0 there exists acompact set Kη sub H such that
Eα ζϵ uϵ kδϵ( 1113857( 1113857le supxisinKη
α ζϵ x( 1113857 +η
radicc 1 +xH +yH( 1113857
(176)
Due to the arbitrariness of η and according to Lemma 8we can get (172) Furthermore (170) holds
rough the above proof eorem 1 is established
Data Availability
No data were used to support this study
20 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
Conflicts of Interest
e authors declare that they have no conflicts of interest
Acknowledgments
is work was partly supported by the National NaturalScience Foundation of China under Grant no 11772255 theFundamental Research Funds for the Central Universitiesthe Research Funds for Interdisciplinary Subject ofNorthwestern Polytechnical University the Shaanxi Projectfor Distinguished Young Scholars the Shaanxi ProvincialKey RampD Program (nos 2020KW-013 and 2019TD-010)and the Seed Foundation of Innovation and Creation forGraduate Students in Northwestern Polytechnical Univer-sity (ZZ2018027)
References
[1] N Bogolyubov and Y Mitropolskii Asymptotic Methods inthe Geory of Nonlinear Oscillations Gordon and BreachScience Publishers New York NY USA 1961
[2] V M Volosov ldquoAveraging in systems of ordinary differentialequationsrdquo Russian Mathematical Surveys vol 17 no 6pp 1ndash126 1962
[3] J G Besjes ldquoOn asymptotic methods for non-linear differ-ential equationsrdquo Journal de Mecanique vol 8 p 357 1969
[4] R Khasminskii ldquoOn the averaging principle for stochasticdifferential Ito equationsrdquo Kybernetika vol 4 pp 260ndash2791968
[5] D Givon ldquoStrong convergence rate for two-time-scale jump-diffusion stochastic differential systemsrdquoMultiscale Modelingamp Simulation vol 6 no 2 pp 577ndash594 2007
[6] M Freidlin and A Wentzell Random Perturbations of Dy-namical Systems Springer Science and BusinessMedia BerlinHeidelberg 2012
[7] J Duan andWWang Effective Dynamics of Stochastic PartialDifferential Equations Elsevier Berlin Heidelberg 2014
[8] W F ompson R A Kuske and A H Monahan ldquoSto-chastic averaging of dynamical systems with multiple timescales forced with α-Stable noiserdquo Multiscale Modeling ampSimulation vol 13 no 4 pp 1194ndash1223 2015
[9] Y Xu J Duan and W Xu ldquoAn averaging principle forstochastic dynamical systems with Levy noiserdquo Physica DNonlinear Phenomena vol 240 no 17 pp 1395ndash1401 2011
[10] Y Xu B Pei B Pei and R Guo ldquoStochastic averaging forslow-fast dynamical systems with fractional Brownian mo-tionrdquo Discrete amp Continuous Dynamical Systems-B vol 20no 7 pp 2257ndash2267 2015
[11] Y Xu B Pei and J-L Wu ldquoStochastic averaging principle fordifferential equations with non-Lipschitz coefficients drivenby fractional Brownian motionrdquo Stochastics and Dynamicsvol 17 no 02 p 1750013 2017
[12] S Cerrai ldquoA Khasminskii type averaging principle for sto-chastic reaction-diffusion equationsrdquo Ge Annals of AppliedProbability vol 19 no 3 pp 899ndash948 2009
[13] S Cerrai ldquoAveraging principle for systems of reaction-dif-fusion equations with polynomial nonlinearities perturbed bymultiplicative noiserdquo SIAM Journal on Mathematical Anal-ysis vol 43 no 6 pp 2482ndash2518 2011
[14] W Wang and A J Roberts ldquoAverage and deviation for slow-fast stochastic partial differential equationsrdquo Journal of Dif-ferential Equations vol 253 no 5 pp 1265ndash1286 2012
[15] B Pei Y Xu and G Yin ldquoAveraging principles for SPDEsdriven by fractional Brownian motions with random delaysmodulated by two-time-scale markov switching processesrdquoStochastics and Dynamics vol 18 no 4 p 1850023 2018
[16] B Pei Y Xu and J-L Wu ldquoStochastic averaging for sto-chastic differential equations driven by fractional Brownianmotion and standard Brownian motionrdquo Applied Mathe-matics Letters vol 100 p 106006 2020
[17] B Pei Y Xu Y Xu and Y Bai ldquoConvergence of p-th mean inan averaging principle for stochastic partial differentialequations driven by fractional Brownian motionrdquo Discrete ampContinuous Dynamical Systems-B vol 25 no 3 pp 1141ndash1158 2020
[18] Z Li and L Yan ldquoStochastic averaging for two-time-scalestochastic partial differential equations with fractionalbrownian motionrdquo Nonlinear Analysis Hybrid Systemsvol 31 pp 317ndash333 2019
[19] P Gao ldquoAveraging principle for the higher order nonlinearschrodinger equation with a random fast oscillationrdquo Journalof Statistical Physics vol 171 no 5 pp 897ndash926 2018
[20] P Gao ldquoAveraging principle for stochastic Kuramoto-Siva-shinsky equation with a fast oscillationrdquo Discrete amp Contin-uous Dynamical Systems-A vol 38 no 11 pp 5649ndash56842018
[21] P Gao ldquoAveraging principle for multiscale stochastic Klein-Gordon-Heat systemrdquo Journal of Nonlinear Science vol 29no 4 pp 1701ndash1759 2019
[22] V Chepyzhov and M Vishik Attractors for Equations ofMathematical Physics American Mathematical SocietyProvidence Rhode Island USA 2002
[23] A Carvalho J Langa and J Robinson Attractors for Infinite-Dimensional Non-autonomous Dynamical Systems SpringerScience and Business Media Berlin Germany 2012
[24] J E Bunder and A J Roberts ldquoResolution of subgridmicroscale interactions enhances the discretisation ofnonautonomous partial differential equationsrdquo AppliedMathematics and Computation vol 304 pp 164ndash1792017
[25] S Cerrai and A Lunardi ldquoAveraging principle for nonau-tonomous slow-fast systems of stochastic reaction-diffusionequations the almost periodic caserdquo SIAM Journal onMathematical Analysis vol 49 no 4 pp 2843ndash2884 2017
[26] D Applebaum Levy Processes and Stochastic CalculusCambridge University Press Cambridge UK 2009
[27] J Duan An Introduction to Stochastic Dynamics CambridgeUniversity Press Cambridge UK 2015
[28] S Peszat and J Zabczyk Stochastic Partial DifferentialEquations with Levy Noise An Evolution Equation ApproachCambridge University Press Cambridge UK 2007
[29] J Bertoin Levy Processes Cambridge University PressCambridge UK 1998
[30] X Zhang Y Xu B Schmalfuszlig and B Pei ldquoRandomattractors for stochastic differential equations driven by two-sided Levy processesrdquo Stochastic Analysis and Applicationsvol 37 no 6 pp 1028ndash1041 2019
[31] Z Li L Yan and L Xu ldquoStepanov-like almost automorphicsolutions for stochastic differential equations with Levynoiserdquo Communications in Statistics - Geory and Methodsvol 47 no 6 pp 1350ndash1371 2018
[32] A Fink ldquoAlmost periodic differential equationsrdquo LectureNotes in Mathematics Vol 377 Springer-Verlag BerlinGermany 1974
[33] B Boufoussi and S Hajji ldquoSuccessive approximation ofneutral functional stochastic differential equations with
Complexity 21
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity
jumpsrdquo Statistics amp Probability Letters vol 80 no 5-6pp 324ndash332 2010
[34] J Luo ldquoFixed points and exponential stability of mild solu-tions of stochastic partial differential equations with delaysrdquoJournal of Mathematical Analysis and Applications vol 342no 2 pp 753ndash760 2008
[35] J Luo and T Taniguchi ldquoe existence and uniqueness fornon-Lipschitz stochastic neutral delay evolution equationsdriven by Poisson jumpsrdquo Stochastics and Dynamics vol 9no 1 pp 135ndash152 2009
[36] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer-Verlag BerlinGermany 1983
[37] G D Prato and J Zabczyk Stochastic Equations in InfiniteDimensions Cambridge University Press Cambridge UK2014
[38] L C Evans Partial Differential Equations American Math-ematical Society Cambridge UK 2nd edition 2010
[39] H Amann Linear and Quasilinear Parabolic Problems Bir-khaeuser Basel Switzerland 1995
[40] A Pazy Semigroups of Linear Operators and Applications toPartial Differential Equations Springer Science amp BusinessMedia Berlin Germany 2010
[41] S Cerrai ldquoStochastic reaction-diffusion systems with multi-plicative noise and non-Lipschitz reaction termrdquo ProbabilityGeory and Related Fields vol 125 no 2 pp 271ndash304 2003
[42] B Pei Y Xu and J-L Wu ldquoTwo-time-scales hyperbolic-parabolic equations driven by Poisson random measuresexistence uniqueness and averaging principlesrdquo Journal ofMathematical Analysis and Applications vol 447 no 1pp 243ndash268 2017
[43] P Billingsley Convergence of Probability Measures JohnWiley amp Son Berlin Germany 1968
[44] S Cerrai ldquoAsymptotic behavior of systems of stochasticpartial differential equations with multiplicative noise sto-chastic partial differential equations and applications-viirdquo2006
[45] G D Prato and C Tudor ldquoPeriodic and almost periodicsolutions for semilinear stochastic equationsrdquo StochasticAnalysis and Applications vol 13 pp 13ndash33 1995
22 Complexity