Click here to load reader
Upload
xiang-gui
View
214
Download
2
Embed Size (px)
Citation preview
Applied Mathematics and Computation 234 (2014) 214–222
Contents lists available at ScienceDirect
Applied Mathematics and Computation
journal homepage: www.elsevier .com/ locate /amc
The finite element method for computing the ground statesof the dipolar Bose–Einstein condensates
http://dx.doi.org/10.1016/j.amc.2014.01.0850096-3003/� 2014 Elsevier Inc. All rights reserved.
⇑ Corresponding author.E-mail address: [email protected] (X.-G. Li).
Dong-Ying Hua, Xiang-Gui Li ⇑School of Applied Science, Beijing Information Science and Technology University, Beijing 100192, PR China
a r t i c l e i n f o
Keywords:Finite element methodBose–Einstein condensatesGeneralized solution
a b s t r a c t
A finite element approximation for computing the ground states of the dipolar Bose–Ein-stein condensates with a nonlocal nonlinear convolution term is presented in one dimen-sion. Following the idea of the imaginary time method, we compute the ground state finitemethod solution of the Bose–Einstein condensates by solving a nonlinear parabolic differ-ential–integral equation. Theoretical analysis is given to show the existence and conver-gence of this finite method solution. Numerical results are given to verify efficiency ofour numerical method.
� 2014 Elsevier Inc. All rights reserved.
1. Introduction
Since 1995, experimental realization of Bose–Einstein condensation (BEC) of trapped atomic gases (mainly alkali atoms)has spurred a great deal of experimental, theoretical and numerical activities, all aimed at gaining a deeper insight into thefascinating behavior of this novel state of matter [1]. In 2005, the first realization of a dipolar BEC of 52Cr [2,3] was observedat Stuttgart University. Needless to say, this dramatic breakthrough on the experimental front has stimulated a correspond-ing wave of studies on both theoretical and numerical fronts. Computational methods have been proposed [4–11] to analyzethe ground states and dynamical behaviors of Bose–Einstein condensates, which can be described by the stationary andtime-dependent Gross–Pitaevskii equation (GPE), respectively. As we know, degenerate quantum gases are usually domi-nated by s-wave contact interactions which is isotropic and short range, while dipolar quantum gases are governed bythe d-wave symmetry of dipole–dipole interactions which is anisotropic and long range [12,13]. Though it has becomeapparent that the contact interactions between the condensed atoms govern most of the observed phenomena, this long-range difference gives rise to novel properties [14,15]. In fact, the long-range nature of dipolar interactions means thatthe GPE which governs the BEC is not only nonlinear but also nonlocal. At temperature T much smaller than the critical tem-perature, a dipolar BEC in the mean field theory is well described by the macroscopic wave function wðx; tÞ whose evolutionis governed by the three-dimensional (3D) GPE [16–18]
i�hwtðx; tÞ ¼ ��h2
2mr2 þ VðxÞ þ U0jwj2 þ Vdip � jwðx; tÞj2
" #wðx; tÞ; x 2 R3; t > 0; ð1:1Þ
with vanishing boundary conditions, where �h is the Planck constant, m the mass of the atom and VðxÞ is the external trappingpotential which is generally harmonic, that is, VðxÞ ¼ m
2 ðx21x2 þx2
2y2 þx23z2Þ with xi ði ¼ 1;2;3Þ representing the trap fre-
quency in x; y; z directions respectively. U0 ¼ 4p�h2as=m is the local interaction between dipoles in the condensate with as
D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222 215
the s-wave scattering length, and it can be tuned between positive for repulsive interaction and negative for attractive inter-action by using a Feshbach resonance. VdipðxÞ is the long-range anisotropic dipolar interaction potential between two dipolesgiven by
VdipðxÞ ¼Cdd
4p1� 3 cos2 h
jxj3; x 2 R3; ð1:2Þ
where h is the angle between the polarization axis n ¼ ðn1;n2;n3Þ and the vector x, i.e., cos h ¼ n�xjxj . For magnetic dipoles we
have Cdd ¼ l0l2d , where l0 is the magnetic vacuum permeability and ld the dipole moment; and for electric dipoles we have
Cdd ¼ d2=l0, where l0 is the vacuum permittivity and d the electric dipole moment. Both the amplitude and sign of Cdd can be
tuned by rotation of the polarization axis. The wave function wðx; tÞ is normalized according to
kwk2 ¼Z
R3jwðx; tÞj2 dx ¼ N ð1:3Þ
where N is the number of the atoms in the dipolar BEC. After the dimensionless transformation [7], that is, by rescaling
lengths with x ¼ a0~x; t ¼ ~tx0
, w ¼ffiffiffiNp
~w
a3=20
, where x0 ¼minfx1;x2;x3g; a0 ¼ffiffiffiffiffiffiffiffi
�hmx0
q, we can get
i~w~t ¼ �12
~r2 ~wþ 12
c21~x2 þ c2
2~y2 þ c23~z2� �
~wþ bj~wj2 ~wþ k Udip � j~wj2� �
~w
with b ¼ U0N�hx0a3
0¼ 4pasN
a0, k ¼ mN
a0�h2, ci ¼xix0ði ¼ 1;2;3Þ and Udip ¼ Cdd
4p1�3 cos2 hj~xj3
. It’s easy to see k ¼ b=U0. To simplify the notation, we
remove the tilde from the above equation, and get the following dimensionless nonlocal GPE:
iwtðx; tÞ ¼ �12r2 þ VðxÞ þ bjwðx; tÞj2 þ kUdip � jwðx; tÞj2
� �wðx; tÞ ð1:4Þ
under the constraint kwðx; tÞk ¼ 1, where VðxÞ ¼ 12 c2
1x2 þ c22y2 þ c2
3z2� �
and Udip ¼ Cdd4p
1�3 cos2 hjxj3
. It is well known that the aboveGPE has two conservation quantities, i.e, the number of the particles and the total energy. That is to say,
ZR3jwðx; tÞj2 dx �
ZR3jwðx; 0Þj2 dx ¼
ZR3jw0ðxÞj
2 dx ¼ 1 ð1:5Þ
and
Eðwðx; tÞÞ ¼Z
R3
12jrwj2 þ VðxÞjwj2 þ b
2jwj4 þ k
2ðUdip � jwj2Þjwj2
� �dx � Eðwðx;0ÞÞ ð1:6Þ
In this paper, we shall be interested in effectively one-dimensional BECs. Despite their mathematical simplicity, lowerdimensional dipolar BECs [19–21] bear nonetheless a significant physical interest as well as a clear advantage for numericalcomputations. For example, in view of experiments on the transport behavior of condensates in elongated optical trapswhich are periodic in only one direction of space [22]; for strong trap anisotropic the time scales along the compressedand elongated axes are very different, which makes an accurate numerical treatment hard. Instead of solving the full 3Dproblem, it is hence desirable to find governing equations for lower dimensional dipolar BECs which are suitable for efficientnumerical methods. We focus on the computation of the ground state. Besides the intrinsic interest of computing the groundstate of BEC, the availability of ground state numerical solvers is also important for BEC dynamics because the dynamicbehavior of BEC can be very sensitive to initial conditions. In this paper we adopt a finite element method rather than apseudospectral method [7] successfully used in dipolar Bose–Einstein condensates in 3D because that technic used in 3Dcan’t be extended to 1D. In addition, the convergence analysis of the pseudospectral is not an easy task. So now considerthe following 1D nonlinear differential-integral GPE
iwtðx; tÞ ¼ �12r2wðx; tÞ þ VðxÞwþ bjwðx; tÞj2wðx; tÞ þ kðUdip � jwðx; tÞj2Þwðx; tÞ; x 2 R; t > 0; ð1:7Þ
where VðxÞ ¼ x2
2 , Udip ¼ jxj�a with 0 < a < 1. And the normalized conservation constraint of the number of atoms isRRjwj2 dx ¼ 1.This paper is organized as follows. In Section 2, we show the theoretical detail of computing the ground state using a finite
element method. In Section 3, numerical examples are shown to verify our results. In Section 4, conclusion is made.
2. The computation of the ground states
The existence and uniqueness of the ground states was discussed in [19]. To obtain the ground state from the numericalpoint of view, we use the imaginary time method [4], that is,
wtðx; tÞ ¼12r2w� VðxÞwðx; tÞ � bjwðx; tÞj2wðx; tÞ � kðUdip � jwðx; tÞj2Þwðx; tÞ ð2:1Þ
216 D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222
withR
Rjwðx; tÞj2 dx ¼ 1. The resulting equation is now a nonlinear parabolic equation. The generalized solution or the weak
form of (2.1) is defined as follows: to find wðx; tÞ 2 L1ð0; T;H10Þ, which satisfies
ZR
wt/dx ¼Z
R
12r2 � VðxÞ � bjwj2 � kðUdip � jwj2
� �w/dx;Z
R
wðx;0Þ/dx ¼Z
R
w0ðxÞ/dx;8/ðxÞ 2 H1
0:
8>><>>: ð2:2Þ
The numerical solution of the corresponding problem (2.2) is: to find the real function whðx; tÞ 2 VhðXÞ, which satisfies
ZXwht/dx ¼Z
X
12r2 � VðxÞ � bjwhj
2 � kðUdip � jwhj2Þ
� �wh/dx;Z
Xwhðx;0Þ/dx ¼
ZR
w0ðxÞ/dx;
8>><>>: 8/ðxÞ 2 VhðXÞ; ð2:3Þ
where w0ðxÞ is the initial condition. Here the finite element method is used to get a numerical solution. Supposewhðx; tÞ ¼
Pmi¼0 aiðtÞ/iðxÞ, where /iðxÞ is the shape function at the mesh grid xi in the continuous linear space VhðXÞ with X
being a bounded area ½�a; a� approximating the whole space R and aiðtÞ is continuous with respect to the time variable t.Eq. (2.3) results in
Xm
i¼0
a0iðtÞZ
X/iðxÞ/jðxÞdx ¼ � 1
2
Xm
i¼0
aiðtÞZ
Xr/iðxÞ � r/jðxÞdx�
Xm
i¼0
aiðtÞZ
XVðxÞ/iðxÞ/jðxÞdx
� bXm
i¼0
aiðtÞZ
X
Xm
k¼0
akðtÞ/kðxÞ
2
/iðxÞ/jðxÞdx
� kXm
i¼0
aiðtÞZ
XUdip �
Xm
k¼0
akðtÞ/kðxÞ
2
0@
1A/iðxÞ/jðxÞdx; 0 6 j 6 m:
For simplicity, we denote Aij ¼R
X /iðxÞ/jðxÞdx, Bij ¼R
X r/iðxÞ � r/jðxÞdx, Cij ¼R
X VðxÞ/iðxÞ/jðxÞdx, DðaÞij ¼R
X
Pmk¼0 akðtÞ
/kðxÞj
2/iðxÞ/jðxÞdx and KðaÞij ¼R
X Udip �Pm
k¼0 akðtÞ/kðxÞ 2� �
/iðxÞ/jðxÞdx. If the element Aij is looked upon as an element
of the matrix A, it’s easy to see that A is symmetric, as well as the corresponding matrix B;C;DðaÞ and KðaÞ. By simple cal-culation, for all 1 6 j 6 m, we get
Aj�1j ¼16
h; Bj�1j ¼ �1h; Cj�1j ¼
Z xj
xj�1
VðxÞ/j�1ðxÞ/jðxÞdx;
DðaÞj�1j ¼1
20ða2
j�1 þ a2j Þ þ
115
aj�1aj
�h;
and for all 1 6 j 6 m� 1,
Ajj ¼23
h; Bjj ¼2h; Cjj ¼
Z xjþ1
xj�1
VðxÞ/2j ðxÞdx;
DðaÞjj ¼1
30ða2
j�1 þ a2jþ1Þ þ
25a2
j þ1
10ajðaj�1 þ ajþ1Þ
�h:
Moreover, we have
A00 ¼ Amm ¼h3; B00 ¼ Bmm ¼
1h; C00 ¼
Z x1
x0
VðxÞ/20ðxÞdx; Cmm ¼
Z xm
xm�1
VðxÞ/2mðxÞdx;
D00 ¼a2
0
5þ a2
1
30þ a0a1
10
� �h; Dmm ¼
a2m�1
30þ a2
m
5þ am�1am
10
� �h:
And it’s easy to see Aij ¼ Bij ¼ Cij ¼ DðaÞij ¼ 0 for all ji� jjP 2. Now we need to handle with KðaÞ. First, the convolution termFðx; tÞ ¼ Udip � j
Pmi¼0 aiðtÞ/iðxÞj
2 is considered. Notice
Fðx; tÞ :¼Z xm
x0
Pmi¼0 aiðtÞ/iðyÞ
2jx� yja
dy ¼ a20ðtÞ
Z x1
x0
/20ðyÞ
jx� yjadyþ a2
mðtÞZ xm
xm�1
/2mðyÞjx� yja
dy
þXm�1
i¼1
a2i ðtÞ
Z xiþ1
xi�1
/2i ðyÞ
jx� yjadyþ 2
X06i<j6m
aiðtÞajðtÞZ xm
x0
/iðyÞ/jðyÞjx� yja
dy: ð2:4Þ
D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222 217
Now we only need to handle the following three kinds of terms. That is,
LiðxÞ :¼Z xi
xi�1
/2i ðyÞ
jx� yjady ¼ 1
h2
Z xi
xi�1
ðy� xþ x� xi�1Þ2
jx� yjady;
RiðxÞ :¼Z xiþ1
xi
/2i ðyÞ
jx� yjady ¼ 1
h2
Z xiþ1
xi
ðxiþ1 � xþ x� yÞ2
jx� yjady;
and
MiðxÞ :¼Z xiþ1
xi
/iðyÞ/iþ1ðyÞjx� yja
dy ¼ 1
h2
Z xiþ1
xi
ðxiþ1 � xþ x� yÞðy� xþ x� xiÞjx� yja
dy:
Through careful computation, and let C ¼ 13�a� 2
2�aþ 11�a, then we have
LiðxÞ ¼1
h2
�Cðxi�1 � xÞ3�a þ ðxi � xÞ3�a
3� a� 2ðxi�1 � xÞðxi � xÞ2�a
2� aþ ðxi�1 � xÞ2ðxi � xÞ1�a
1� a; x 6 xi�1;
Cðx� xi�1Þ3�a þ ðxi � xÞ3�a
3� aþ 2ðx� xi�1Þðxi � xÞ2�a
2� aþ ðx� xi�1Þ2ðxi � xÞ1�a
1� a; xi�1 < x < xi;
Cðx� xi�1Þ3�a � ðx� xiÞ3�a
3� aþ 2ðx� xi�1Þðx� xiÞ2�a
2� a� ðx� xi�1Þ2ðx� xiÞ1�a
1� a; x P xi:
8>>>>>>>><>>>>>>>>:
RiðxÞ ¼1
h2
Cðxiþ1 � xÞ3�a � ðxi � xÞ3�a
3� aþ 2ðxiþ1 � xÞðxi � xÞ2�a
2� a� ðxiþ1 � xÞ2ðxi � xÞ1�a
1� a; x 6 xi;
Cðxiþ1 � xÞ3�a þ ðx� xiÞ3�a
3� aþ 2ðxiþ1 � xÞðx� xiÞ2�a
2� aþ ðxiþ1 � xÞ2ðx� xiÞ1�a
1� a; xi < x < xiþ1;
�Cðx� xiþ1Þ3�a þ ðx� xiÞ3�a
3� a� 2ðx� xiþ1Þðx� xiÞ2�a
2� aþ ðx� xiþ1Þ2ðx� xiÞ1�a
1� a; x P xiþ1:
8>>>>>>>><>>>>>>>>:
MiðxÞ ¼1
h2
�ðxiþ1 � xÞðxi � xÞ1� a
ðxiþ1 � xÞ1�a � ðxi � xÞ1�a� �
þ xiþ1 þ xi � 2x2� a
ðxiþ1 � xÞ2�a � ðxi � xÞ2�a� �
� 13� a
ðxiþ1 � xÞ3�a � ðxi � xÞ3�a� �
; x 6 xi;
ðxiþ1 � xÞðx� xiÞ1� a
ðxiþ1 � xÞ1�a þ ðx� xiÞ1�a� �
þ xiþ1 þ xi � 2x2� a
ðxiþ1 � xÞ2�a � ðx� xiÞ2�a� �
� 13� a
ðxiþ1 � xÞ3�a þ ðx� xiÞ3�a� �
; xi < x < xiþ1;
ðx� xiþ1Þðx� xiÞ1� a
ðx� xiþ1Þ1�a � ðx� xiÞ1�a� �
xiþ1 þ xi � 2x2� a
ðx� xiþ1Þ2�a � ðx� xiÞ2�a� �
þ 13� a
ðx� xiþ1Þ3�a � ðx� xiÞ3�a� �
; x P xiþ1:
8>>>>>>>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>>>>>>>:
So we have
Fðx; tÞ ¼ a20ðtÞR0ðxÞ þ
Xm�1
i¼1
a2i ðtÞ LiðxÞ þ RiðxÞð Þ þ a2
mðtÞLmðxÞ þ 2Xm�1
i¼0
aiðtÞaiþ1ðtÞMiðxÞ ð2:5Þ
and therefore KðaÞij ¼R
X Fðx; tÞ/iðxÞ/jðxÞdx. Then we have the following ordinary differential equations
A~a0ðtÞ ¼ � 12
Bþ C þ bDð~aÞ þ kKð~aÞ� �
~aðtÞ ð2:6Þ
with ~aðtÞ ¼ ða0ðtÞ;a1ðtÞ; . . . ;am�1ðtÞ;amðtÞÞT . The boundary condition wðx; tÞ ¼ 0 on @X gives a0ðtÞ ¼ amðtÞ ¼ 0 for all t > 0.The continuous initial condition wðx;0Þ ¼ w0ðxÞ shows that
ZXwhðx;0Þ/jðxÞdx ¼
ZX
w0ðxÞ/jðxÞdx; 0 6 j 6 m; ð2:7Þ
in the discrete level. That is,
A~að0Þ ¼~b ð2:8Þ
with the jth component of~b beingR
X w0ðxÞ/jðxÞdx. These equations can be solved by Runge–Kutta method or the finite dif-ference method. We solve the ODE system (2.6) with the semi-implicit backward Euler scheme
218 D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222
Aþ s 12
Bþ C þ bDð~akÞ þ kKð~akÞ� � �
~akþ1 ¼ A~ak ð2:9Þ
with the initial condition~a0 ¼ ~að0Þ in (2.8). As all the /jðxÞ ðj ¼ 0;1; . . . ;mÞ is independent, the above equations has a uniquelocal solution on ½0; t0�. Furthermore, it’s easy to see
ZXjwhðx; tÞj
2dx ¼ ~aT A~a
where ~a :¼ ~aðtÞ and after each discrete time level, whðx; tÞ should be normalized to satisfyR
X jwhðx; tÞj2 dx ¼ 1, that is,
whðx; tÞ :¼ whðx;tÞkwhðx;tÞk
. In addition, the total energy is
Eðwhðx; tÞÞ ¼12~aT B~aþ~aT C~aþ b
2~aT Dð~aÞ~aþ k
2~aT Kð~aÞ~a ¼ ~aT 1
2Bþ C þ b
2Dð~aÞ þ k
2Kð~aÞ
� �~a ð2:10Þ
Here, 12~aT B~a, ~aT C~a, b
2~aT Dð~aÞ~a and k
2~aT Kð~aÞ~a are the kinetic energy, the potential energy, the internal energy and the dipolar
energy, respectively.Next, we will prove that the numerical solution exists globally and is convergent to the generalized solution of (2.1) if
b P 4kð2aÞ1�a
1�a .
Proposition 1. Suppose the initial condition w0ðxÞ is bounded uniformly and b P 0, then we have
kwhðx; tÞkL2 6 C; ð2:11Þ
where the constant C is independent of h and t.
Proof. Set /ðxÞ ¼ wh in (2.3) and it is easy to see
12
ddt
ZXjwhj
2 dx ¼ �12
ZXjrwhj
2 dx�Z
XVðxÞjwhj
2 dx� bZ
Xjwhj
4 dx� kZ
XðUdip � jwhj
2Þjwhj2 dx;
which shows
ddtkwhk
2L2 6 0; ð2:12Þ
i.e.,
kwhðx; tÞkL2 6 kwhðx;0ÞkL2 :
Moreover, by (2.7) we have
Zwhðx;0Þwhðx;0Þdx ¼Zw0ðxÞwhðx;0Þdx;
which results in
kwhðx;0ÞkL2 6 kw0ðxÞkL2
and the proof is completed. h
Proposition 2. Suppose kwhtðx;0Þk2L2 is bounded with respect to h uniformly and b P 4kð2aÞ1�a
1�a , then we have
kwhtðx; tÞkL2 6 C; ð2:13Þ
where C is independent of h and t.
Proof. Differentiate (2.3) with respect to t, then set / ¼ wht in the resulting equation, we obtain
ZXwhttwht dx ¼ 12
ZXr2whtwht dx�
ZX
VðxÞjwhtj2 dx� b
ZXðjwhj
2whÞtwht dx� kZ
XðUdip � jwhj
2Þwh
� �twht dx;
That is,
12
ddtkwhtk
2L2 ¼ � 1
2krwhtk
2L2 �
ZX
VðxÞjwhtj2 dx� b
ZX
12
@
@tjwhj
2� �2
þ jwhj2jwhtj
2
" #dx� k
ZXðUdip � jwhj
2Þtwhwht dx
� kZ
XðUdip � jwhj
2Þjwhtj2 dx 6 �b
ZXjwhj
2jwht j2 dx� k
ZXðUdip � jwhj
2Þtwhwht dx
¼ � bZ
Xjwhj
2jwht j2 dx� k
2
ZZX2
1jx� yja
@
@tjwhðy; tÞj
2 @
@tjwhðx; tÞj
2 dxdy;
D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222 219
Notice
ZZX21jx� yja
@
@tjwhðy; tÞj
2 @
@tjwhðx; tÞj
2 dxdy
6 4ZZ
X2
jwhðx; tÞwhtðx; tÞjjx� yj
a2
� jwhðy; tÞwhtðy; tÞjjx� yj
a2
dxdy
6 4ZZ
X2
jwhðx; tÞj2jwhtðx; tÞj
2
jx� yjadxdy
!12 ZZ
X2
jwhðy; tÞj2jwhtðy; tÞj
2
jx� yjadxdy
!12
¼ 4ZZ
X2
jwhðx; tÞj2jwhtðx; tÞj
2
jx� yjadxdy
and
ZZX2jwhðx; tÞj2jwhtðx; tÞj
2
jx� yjadxdy ¼
Z a
�ajwhðx; tÞj
2jwhtðx; tÞj2 dx
Z a
�a
1jx� yja
dy
¼Z a
�ajwhðx; tÞj
2jwhtðx; tÞj2Z x
�a
1jx� yja
dyþZ a
x
1jx� yja
dy� �
dx
¼Z a
�ajwhðx; tÞj
2jwhtðx; tÞj2 1
1� aðaþ xÞ1�a þ ða� xÞ1�a� �
dx
62 � ð2aÞ1�a
1� a
ZXjwhðx; tÞj
2jwhtðx; tÞj2 dx;
we have
12
ddtkwhtk
2L2 6 �bþ 4kð2aÞ1�a
1� a
! ZXjwhðx; tÞj
2jwhtðx; tÞj2 dx:
Choosing b sufficiently large such that b P 4kð2aÞ1�a
1�a , we have
12
ddtkwhtk
2L2 6 0;
therefore,
kwhtðx; tÞk2L2 6 kwhtðx; 0Þk
2L2 6 C
which completes the proof. h
Proposition 3. Under the assumptions of Propositions 1 and 2, we have
krwhkL2 6 C; kVðxÞjwhj2kL1 6 C;
kwhkL4 6 C; kðUdip � jwhj2Þjwhj
2kL1 6 C:ð2:14Þ
Proof. By (2.3), we have
ZX
whtwh dx ¼ �12
ZXjrwhj
2 dx�Z
XVðxÞjwhj
2 dx� bZ
Xjwhj
4 dx� kZ
XðUdip � jwhj
2Þjwhj2 dx; ð2:15Þ
which shows
12
ZXjrwhj
2 dxþZ
XVðxÞjwhj
2 dxþ bZ
Xjwhj
4 dxþ kZ
XðUdip � jwhj
2Þjwhj2 dx 6 kwhtkL2kwhðx; tÞkL2 ;
Therefore,
12
ZXjrwhj
2 dxþZ
XVðxÞjwhj
2 dxþ bZ
Xjwhj
4 dxþ kZ
XðUdip � jwhj
2Þjwhj2 dx 6 C: ð2:16Þ
As the terms in the left hand side of (2.16) are all nonnegative, the proof is complete. h
Theorem 1. Under the assumptions of Propositions 1–3, there exists any T > 0 such that the finite element solution whðx; tÞ isconvergent to the generalized solution (2.2) on ½0; T�.
220 D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222
Proof. By Propositions 1–3, kwhkL2 , krwhkL2 , kwhtkL2 , kwhkL4 , and kðUdip � jwhj2Þjwhj
2kL1 are uniformly bounded, and therefore,the finite element solution is not only a local solution on ½0; t0�, but can be extended to any time interval ½0; T�. In addition, theboundedness of the above norms also implies there exists a subseries (still denoted by wh) for any fixed t, and wh has thefollowing properties: wh is strongly convergent to wðx; tÞ in L2,rwh;wht are weakly convergent torw;wt in L2, respectively, wh
is weakly convergent to w in L4 and ðUdip � jwhj2Þjwhj
2 is weakly convergent to ðUdip � jwj2Þjwj2 in L1. Let h! 0 in (2.3), and wehave
ZR
wt/dx ¼ 12
ZR
r2w/dx�Z
R
VðxÞw/dx� bZ
R
jwj2w/dx� kZ
R
ðUdip � jwj2Þw/dx; 8/ðxÞ 2 VhðXÞ: ð2:17Þ
As the space Vh composed by the basis /iðxÞ is condense in H10, which shows (2.17) is valid for all /ðxÞ 2 H1
0. That is, the lim-itation wðx; tÞ of whðx; tÞ as h! 0 is the generalized solution (2.2). That is to say, the generalized solution (2.2) existsglobally. h
3. Numerical examples
In this section, the ground states of a dipolar BEC with different parameters are given by the numerical scheme above. Theinitial condition is chosen to be w0ðxÞ ¼ 1
p1=4 e�x2=2. The ground state wg :¼ wkþ1h ¼
Pmj¼0 akþ1
j /jðxÞ is reached numerically in (2.6)when
k~akþ1 �~akk1 :¼ max06j6M
jakþ1j � ak
j j 6 e :¼ 10�6:
0.35
−20 −15 −10 −5 0 5 10 15 200
0.05
0.1
0.15
0.2
0.25
0.3
x
ψg
Fig. 1. Ground state solution wg of a dipolar BEC with k ¼ 10 and b ¼ 500.
0 0.5 1 1.5 2 2.5 3 3.50
1
2
3
4
5
6
7
8
9
t
Energy
EkinEpotEdip
0 0.5 1 1.5 2 2.5 3 3.510
20
30
40
50
60
70
80
90
100
110
t
Energy
EintEtotal
Fig. 2. The energy evolution of a dipolar BEC with k ¼ 10 and b ¼ 500.
−8 −6 −4 −2 0 2 4 6 80
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
x
ψg
Fig. 3. Ground state solution wg of a dipolar BEC with k ¼ 0:8 and b ¼ 30.
0 0.5 1 1.5 2 2.5 30
1
2
3
4
5
6
7
8
t
Energy
EkinEpotEintEdipEtotal
Fig. 4. The energy evolution of a dipolar BEC with k ¼ 0:8 and b ¼ 30.
−8 −6 −4 −2 0 2 4 6 80
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
x
ψg
Fig. 5. Ground state solution wg of a dipolar BEC with k ¼ 0:8 and b ¼ 80.
D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222 221
Example 1. We solve this problem on ½�16;16� with h ¼ 1=8 and s ¼ 0:01. Fig. 1 shows the ground state wgðxÞ of a dipolarBEC with a ¼ 0:5; b ¼ 500; k ¼ 10. And in Fig. 2, the energy evolution is given.
222 D.-Y. Hua, X.-G. Li / Applied Mathematics and Computation 234 (2014) 214–222
Example 2. We solve this problem on ½�8;8� with h ¼ 1=32 and s ¼ 0:005. Fig. 3 shows the ground state wgðxÞ of a dipolarBEC with a ¼ 0:5; b ¼ 30; k ¼ 0:8. And in Fig. 4, the energy evolution is given.
Example 3. In this example, we keep all the parameters the same as in Example 2 except b ¼ 80. The ground state is shownin Fig. 5, and it is short and stout than that in Fig. 3, which is consistent with the fact that the stronger the repulsive inter-action between the particles is, the bigger the cloud size will be.
In the above numerical examples, the kinetic energy, the internal energy, the dipolar energy and the total energy all de-crease, while the potential energy increases.
4. Conclusion
In this paper, a finite element method is presented to compute the ground states of the dipolar Bose–Einstein conden-sates. As the dipole–dipole interactions are anisotropic and long range, the governing equation described by a Gross–Pitaev-skii equation is nonlocal and nonlinear which brings difficulty in numerical simulation as well as in theoretical analysis. Thefinite element method rather than other methods is adopted in 1D to compute the ground states because in 3D, a dipolarBose–Einstein condensates can be rewritten as a coupled Gross–Pitaevskii–Poisson equations [7], where the time indepen-dent Poisson equation can be solve by the finite element method efficiently. Numerical examples shows our method is reli-able and efficient. And the extension of this method to 3D is in progress and will be presented in our future work.
Acknowledgement
This work was supported by National Natural Science Foundation of China (No. 11171032) and Beijing Municipal Educa-tion Commission (No. KM201110772017).
References
[1] L. Pitaevskii, S. Stringari, Bose–Einstein Condensation, Oxford University, New York, 2003.[2] A. Griesmaier, J. Werner, S. Hensler, J. Stuhler, T. Pfau, Bose–Einstein condensation of chromium, Phys. Rev. Lett. 94 (2005). article 160401.[3] J. Stuhler, A. Griesmaier, T. Koch, et al, Observation of dipole–dipole interaction in a degenerate quantum gas, Phys. Rev. Lett. 95 (2005). article 150406.[4] M.L. Chiofalo, S. Scci, M.P. Tosi, Ground state of trapped interacting Bose–Einstein condensates by an explicit imaginary-time algorithm, Phys. Rev. E 62
(2000) 7438–7444.[5] W. Bao, Q. Du, Computing the ground state solution of Bose–Einstein condensates by a normalized gradient flow, SIAM J. Sci. Comput. 25 (2004) 1674–
1697.[6] W. Bao, W. Tang, Ground-state solution of Bose–Einstein condensate by directly minimizing the energy functional, J. Comput. Phys. 187 (2003) 230–
254.[7] W. Bao, Y. Cai, H. Wang, Efficient numerical methods for computing ground states and dynamics of dipolar Bose–Einstein condensates, J. Comput. Phys.
229 (2010) 7874–7892.[8] D.Y. Hua, X.G. Li, J. Zhu, A mass conserved time splitting method for the nonlinear Schrodinger equation, Adv. Differ. Equ. 85 (2012).[9] M. Caliari, A. Ostermann, S. Rainer, M. Thalhammer, A minimisation approach for computing the ground state of Gross–Pitaevskii systems, J. Comput.
Phys. 228 (2009) 349–360.[10] Z. Huang, P.A. Markowich, C. Sparber, Numerical simulation of trapped dipolar quantum gases: collapse studies and vortex dynamics, Kinet. Relat.
Models 3 (2010) 181–194.[11] K. Góral, L. Santos, Ground state and elementary excitations of single and binary Bose–Einstein condensates of trapped dipolar gases, Phys. Rev. A 66
(2002). article 023613.[12] T. Lahaye, C. Menotti, L. Santos, M. Lewenstein, T. Pfau, The physics of dipolar bosonic quantum gases, Rep. Prog. Phys. 72 (2009). article 126401.[13] L. Santos, G.V. Shlyapnikov, P. Zoller, M. Lewenstein, Bose–Einstein condensation in trapped dipolar gases, Phys. Rev. Lett. 85 (2000) 1791–1794.[14] T. Lahaye, J. Metz, B. Fröhlich, et al, d-wave collapse and explosion of a dipolar Bose–Einstein condensate, Phys. Rev. Lett. 101 (2008). article 080401.[15] K. Góral, K. Rza� _zewski, T. Pfau, Bose–Einstein condensation with magnetic dipole–dipole forces, Phys. Rev. A 61 (2000). article 051601(R).[16] S. Yi, L. You, Trapped atomic condensates with anisotropic interactions, Phys. Rev. A 61 (2000). article 041604(R).[17] S. Yi, L. You, Trapped condensates of atoms with dipole interactions, Phys. Rev. A 63 (2001). article 053607.[18] S. Yi, L. You, Calibrating dipolar interaction in an atomic condensate, Phys. Rev. Lett. 92 (2004). article 193201.[19] R. Carles, P.A. Markowich, C. Sparber, On the Gross–Pitaevskii equation for trapped dipolar quantum gases, Nonlinearity 21 (2008) 2569–2590.[20] Y. Cai, M. Rosenkranz, Z. Lei, W. Bao, Mean-field regime of trapped dipolar Bose–Einstein condensates in one and two dimensions, Phys. Rev. A 82
(2010). article 043623.[21] S. Ronen, D. Bortolotti, J. Bohn, Bogoliubov modes of a dipolar condensate in acylindrical trap, Phys. Rev. A 74 (2006). article 013623.[22] B.P. Anderson, M.A. Kasevich, Macroscopic quantum interference from atomic tunnel arrays, Science 282 (1998) 1686–1689.