73
Pad´ e approximations of generalized hypergeometric series Tapani Matala-aho Tokyo 2009 March 02

Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations of generalized

hypergeometric series

Tapani Matala-aho

Tokyo 2009 March 02

Page 2: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Abstract

We shall present short proofs for type II Pade approximations of

the generalized hypergeometric and q-hypergeometric series

F (t) =∞∑

n=0

∏n−1k=0 P(k)∏n−1k=0 Q(k)

tn, Fq(t) =∞∑

n=0

∏n−1k=0 P(qk)∏n−1k=0 Q(qk)

tn. (1)

In a q-exponential case we will discuss how certain modified

approximations give sharp linear independence results. Further, a

comparison is done between the remainder series approximations of

the exponential series (Prevost and Rivoal) and our modified

approximations for a q-analogue of the exponential series.

Page 3: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Arithmetic Motivation

An interesting part of Number Theory is involved with a question

of arithmetic nature of explicitly defined numbers.

-Irrationality

-Linear independence over a field

-Transcendency

Even more interesting with a quantitative setting.

Page 4: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Generalized Hypergeometric series

Let P(y) and Q(y) be polynomials and define generalized

(classical) hypergeometric and (basic) q-hypergeometric series

F (t) =∞∑

n=0

[P]n[Q]n

tn, Fq(t) =∞∑

n=0

[P; q]n[Q; q]n

tn, (2)

where

[P]n = [P(y)]n =n−1∏k=0

P(k), (3)

[P; q]n = [P(y); q]n =n−1∏k=0

P(qk). (4)

Page 5: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Classical hypergeometric series

Pochhammer symbol (generalized factorial)

(a)0 = 1, (a)n = a(a + 1) · · · (a + n − 1) n ∈ Z+. (5)

(1)n = n!

Hypergeometric series

AFB

(a1, ..., aA

b1, ..., bB

∣∣∣ t) =∞∑

n=0

(a1)n · · · (aA)n

n!(b1)n · · · (bB)ntn (6)

Page 6: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Classical case

First we will study the classical series F (t) with it’s derivativies

∆bF (t), where ∆ = t ddt .

Denote d = max{deg P(y), deg Q(y)} and let d ,m ∈ Z+ and the

numbers α1, ..., αm be given.

We start by giving explicit type II Pade approximations for the

series

∆bF (tαj), b = 0, 1, ..., d − 1; j = 1, ...,m. (7)

Our construction is based on a product expansion a la Maier

[Potenzreihen irrationalen Grenzwertes. J. Reine Angew. Math.

156, 93–148 (1927)]

Page 7: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Maier’s product formula

Let l ,m ∈ Z+ and α = t(α1, ..., αm) be given and define

σi = σi (l , α) bym∏

t=1

(αt − w)l =ml∑i=0

σiwi . (8)

Thenml∑i=0

σi ikαi

t = 0 (9)

for all t ∈ {1, ...,m}; k ∈ {0, ..., l − 1}.

Page 8: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Maier’s product formula

Moreover

σi = (−1)i∑

i1+...+im=i

(l

i1

)· · ·(

l

im

)· αl−i1

1 · · ·αl−imm . (10)

Denote

Σh =∑

h1+...+hm=h

(l

h1

)· · ·(

l

hm

)· αh1

1 · · ·αhmm (11)

Page 9: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/Classical case

Let b, d , l ,m, λ ∈ N, b < d and choose m numbers α1, ..., αm. Put

Bl ,λ(t) =ml∑i=0

tml−iσi (l , α)[Q]i+λ+bl/dc−1

[P]i+λ. (12)

Then

Bl ,λ(t)∆bF (αj t)− Al ,λ,b,j(t) = Rl ,λ,b,j(t), (13)

where

degt Bl ,λ(t) = ml , degt Al ,λ,b,j(t) ≤ ml + λ− 1 (14)

ordt=0

Rl ,λ,b,j(t) ≥ ml + bl/dc+ λ. (15)

Page 10: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/Classical case

Thus we have a gap of lenght bl/dc in the power series expansion

Bl ,λ(t)∆bF (αj t) = Al ,λ,b,j(t) + Rl ,λ,b,j(t). (16)

The polynomials Bl ,λ(t) are Pade approximant denominators in

variable t for the functions Fb,j(t) = ∆bF (tαj),

b = 0, 1, ..., d − 1; j = 1, ...,m.

Also we say that (13–15) define a Pade approximation with the

degree and order parameters

[degt B, degt A ≤, ordt=0

R ≥] = [ml ,ml + λ− 1,ml + bl/dc+ λ]

(17)

Page 11: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

Write

Bl ,λ(t) =ml∑

h=0

thbl ,λ,h =ml∑i=0

tml−iσi

[Q]i+λ+bl/dc−1

[P]i+λ(18)

and study the expansion of the product

Bl ,λ(t)∆bF (tαj) =∞∑

N=0

rNtN , (19)

where

rN =∑

h+n=N

bl ,λ,hnbfnα

nj (20)

Page 12: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

Put

N = ml + λ+ a (21)

Then

rN =ml∑i=0

σi

[Q]i+λ+bl/dc−1

[P]i+λ

[P]i+λ+a

[Q]i+λ+a(i + λ+ a)bαi+λ+a

j (22)

or

Page 13: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

rN =ml∑i=0

σi[P]i+λ+a

[P]i+λ

[Q]i+λ+bl/dc−1

[Q]i+λ+a(i + λ+ a)bαi+λ+a

j (23)

If now

0 ≤ a ≤ bl/dc − 1 (24)

then

rN = αa+λj

ml∑i=0

σi (i + λ+ a)bαij ·

[P(i + λ)]a[Q(i + λ+ a)]bl/dc−a−1 (25)

Page 14: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

By binomial theorem we may denote

(i + a + λ)b =b∑

k=0

sk ik , (26)

P(i + b) =d∑

kb=0

pkbik , Q(i + b) =

d∑kb=0

qkbik ∀b ∈ Z,

where pb, qb, sk ’s are independent of i .

Page 15: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

Thus

[P(i + λ)]a =

ad∑K=0

∑k0+...+ka−1=K

pk0 · · · pka−1 ik0 · · · ika−1

=ad∑

K=0

PK iK , (27)

and similarly

[Q(qa+i+λ)]bl/dc−a−1 =

(bl/dc−a−1)d∑K=0

QK iK , (28)

where PK ,QK ’s are independent of i .

Page 16: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

So,

rN = αa+λj

(a+bl/dc−a−1)d+b∑K=0

∑k+K1+K2=K

skPK1QK2

ml∑i=0

σiαij i

k+K1+K2

(29)

and finally

rN = αa+λj

(bl/dc−1)d+b∑K=0

∑k+K1+K2=K

skPK1QK2

ml∑i=0

σi iKαi

j , (30)

where 0 ≤ K ≤ (bl/dc − 1)d + b ≤ l − 1.

Page 17: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Proof

Hence, by using the property

ml∑i=0

σi ikαi

j = 0 (31)

for j ∈ {1, ...,m}; k ∈ {0, ..., l − 1} we have

rN = rml+λ+a = 0 ∀ 0 ≤ a ≤ bl/dc − 1. (32)

NOTE: The crucial point in the proof is formula (23) which in fact

shows how to find the right denominator polynomials.

Page 18: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world

q-series factorials (q-Pochhammer symbols):

(b, a)0 = (b, a; q)0 = 1

(b, a)n = (b, a; q)n = (b − a)(b − aq)...(b − aqn−1), n ∈ Z+

(a)n = (a; q)n = (1, a)n = (1− a)(1− aq) · · · (1− aqn−1) (33)

(q)n = (q; q)n = (1− q)...(1− qn)

q-hypergeometric (basic) series

AΦB

(a1, ..., aA

b1, ..., bB

∣∣∣ t) =∞∑

n=0

(a1)n...(aA)n

(q)n(b1)n...(bB)ntn. (34)

Page 19: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world

The q-binomial coefficients are defined by[nk

]=

(q; q)n

(q; q)k(q; q)n−k(35)

q-binomial theorem [2, p. 490, Corollary 10.2.2(c)]:

(b, a; q)n =n∑

k=0

[nk

]q(k

2)bn−k(−a)k ∀n ∈ N (36)

Page 20: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/q-world

Stihl [Arithmetische Eigenschaften spezieller Heinescher Reihen.

Math. Ann. 268, 21–41 (1984)]

Let l ,m ∈ Z+ and α = t(α1, ..., αm) be given and define

σq,i = σq,i (l , α) by

m∏t=1

(αt ,w ; q−1)l =ml∑i=0

σq,iwi (37)

Thenml∑i=0

σq,i (αjqk)i = 0 (38)

for all j ∈ {1, ...,m}; k ∈ {0, ..., l − 1}.

Page 21: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/q-world

Moreover, by q-binomial theorem,

σq,i = (−1)iq−m( l2)Σq,ml−i (39)

holds with

Σq,h = Σq,h(l , α) =∑

i1+...+im=h

[l

i1

]· · ·[

l

im

q(i12)+...+(im

2 )αi11 · · ·α

imm . (40)

Page 22: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/q-world

Define a operator J = Jt by Jf (t) = f (qt) and denote again

d = max{deg P(y), deg Q(y)}.

Stihl constructed the following explicit type II Pade approximations

in variable t for the d series JbF (t), 0 ≤ b ≤ d − 1 at m points.

Page 23: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/q-world

Let b, l ,m, λ ∈ N, 0 ≤ b < d = max{deg P(y), deg Q(y)} and

choose m numbers α1, ..., αm. Put σq,i = σq,i (l , α) and

Bl ,λ(t) =ml∑

h=0

thbl ,λ,h =ml∑i=0

tml−iσq,i

[Q; q]i+λ+bl/dc−1

[P; q]i+λ(41)

Then

Bl ,λ(t)JbF (αj t)− Al ,λ,b,j(t) = Rl ,λ,b,j(t), (42)

holds with the parameters

[ml ,ml + λ− 1,ml + bl/dc+ λ]. (43)

Page 24: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Pade approximations/q-world

This means that the polynomials Bl ,λ(t) are Pade approximant

denominators in variable t for the functions Fb,j(t) = JbF (tαj),

b = 0, 1, ..., d − 1; j = 1, ...,m.

The linear independence results by applying the above explicit

Pade approximations of q-hypergeometric series restrict usually to

the following class of functions

F (t) =∞∑

n=0

qm(n2)

(q)n(a1)n...(al)ntn, m ≥ 2, l ≤ m − 2 (44)

with |q| < 1.

Page 25: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world/analogues of exponential series

Hence, for example, the linear independence question of

q-analogues of exponential function and Euler’s divergent series

have resisted the attacts from the explicit Pade approximations.

q-analogues of exponential series:

Eq(z) =∞∑

n=0

1

(q; q)nzn, Eq(z) =

∞∑n=0

q(n2)

(q; q)nzn (45)

A q-analogue of Euler’s divergent series:

Dq(t) =∞∑

n=0

(q; q)ntn. (46)

Page 26: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Arithmetic of q-series

Amou M., Andre Y., Bertrand D., Bezivin, Borwein P., Bundschuh

P., Duverney D., Katsurada M., Nesterenko Yu., Nishioka K.,

Prevost M., Rivoal T., Stihl Th., Shiokawa I., Waldscmidt M.,

Wallisser R., Vaananen K., Zudilin W.

Page 27: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world numbers

p-adic, p ∈ P:∞∑

n=1

pn

1− pn, (47)

∞∑n=1

pn

n∏i=1

1± pi

, (48)

∞∑l=0

pl2∏lj=1(1± pj)2

(49)

Page 28: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world numbers

1 +p

1 +

p2

1 +

p3

1 + . . .(50)

∞∏n=1

(1 + kpn), k = 1, ..., p − 1, (51)

∞∑n=1

pnn∏

i=1

(1 + kpi ), k = 1, ..., p − 1, (52)

Page 29: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world numbers

Real, p ∈ Z \ {0,±1}:∞∑

n=1

1

1− pn, (53)

∞∑n=1

1n∏

i=11± pi

, (54)

∞∑l=0

1∏lj=1(1± pj)2

, (55)

Page 30: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world numbers

1 +p−1

1 +

p−2

1 +

p−3

1 + . . ., (56)

∞∏n=1

(1 + kp−n), k = 0, 1, ..., p − 1, (57)

∞∑n=1

p−nn∏

i=1

(1 + kp−i ), k = 0, 1, ..., p − 1. (58)

Page 31: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world numbers

1

F1 +

1

F3 +

1

F5 + . . ., (59)

1

F2 +

1

F4 +

1

F6 + . . ., (60)

∞∑n=0

1

Fan+b, (61)

∞∑n=0

1

Lan+b, (62)

where a, b,∈ Z+, Fn and Ln are the Fibonacci and Lucas numbers,

respectively; F0 = 0,F1 = 1, L0 = 2, L1 = 1.

Page 32: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world/An analogue of Euler’s divergent series

Euler’s divergent series:∞∑

n=0

n!tn. (63)

A q-analogue of Euler’s divergent series:

Dq(t) =∞∑

n=0

(q; q)ntn. (64)

a special case of

Db(t) =∞∑

n=0

(b; q)ntn (65)

Page 33: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

In the following we will present explicit Pade type approximations

in variable t for the q-exponential series

Fx(t) =∞∑

k=0

q(k2)(−tx)k

(t; q)k+1. (66)

First, we have

Fx(t) = Dx(t) =∞∑

n=0

(x ; q)ntn (67)

Page 34: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

Define q-factorial polynomials

(x ; q)n =n−1∏h=0

(1− xqh) =n∑

k=0

s(n, k)xk (68)

and set s(n, k) = 0, when k < 0 or n < k .

Thenn∑

k=0

s(n, k)(−x)n−k = q(n2)xn ∀ n ∈ N (69)

which plays an essential role in the following modification.

Page 35: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

Define the following polynomials

Bl ,ν(t) =ml∑

h=0

bl ,ν,h(t)th, (70)

with

bl ,ν,h(t) = (−1)ml−hq(ml+ν2 )−(ml+ν−h

2 )(t; q)ml+ν−hΣq,h (71)

Page 36: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

Then by (68)

Bl ,ν(t) =ml+ν∑H=0

bl ,ν,HtH (72)

with

bl ,ν,H = qm( l2)

∑ml−i+f =H

0≤f≤i+ν≤ml+ν

q(ml+ν2 )−(i+ν

2 )s(i + ν, f )σq,i (73)

Page 37: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Pade approximations

Now we get the following modified Pade type approximations of

the second kind for the m functions D−αj (t), (j = 1, ...,m) in

variable t. Let l , ν ∈ N and j = 1, ...,m, then

Bl ,ν(t)D−αj (t) + Al ,ν,j(t) = Ll ,ν,j(t) (74)

with

[ml + ν,ml + ν − 1, (m + 1)l + ν] (75)

By (75) we have a diagonal type Pade approximation with the free

parameter ν.

Page 38: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

Further

Al ,ν,j(t) =ml+ν−1∑

N=0

al ,ν,j ,NtN , (76)

where

al ,ν,j ,N = −∑

H+n=N

bl ,ν,H(−αj ; q)n, (77)

and

Page 39: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/q-world

Ll ,ν,j(t) = t(m+1)l+νq(ml+ν2 )+m( l

2)+νl(−αj ; q)lανj Sl ,ν,j(t). (78)

where

Sl ,ν,j(t) =∞∑

k=0

sl ,ν,j ,ktk (79)

with

sl ,ν,j ,k = qkν(−αjql ; q)k

m∏t=1

(αt , αjqk+1; q)l . (80)

Page 40: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Proof

The expansion

Bl ,ν(t)D−αj (t) =∞∑

N=0

rNtN , (81)

holds with

rN =∑

H+n=N

bl ,ν,H(−αj ; q)n. (82)

Page 41: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Proof

Set

N = ml + ν + a 0 ≤ a ≤ l − 1

then

rN = qm( l2)

ml∑i=0

i+ν∑f =0

q(ml+ν2 )−(i+ν

2 )s(i + ν, f )σq,i (−αj ; q)i+ν−f +a =

qm( l2)

ml∑i=0

σq,i (−αj ; q)aq(ml+ν

2 )−(i+ν2 )·

i+ν∑f =0

s(i + ν, f )(−αjqa; q)i+ν−f (83)

Page 42: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Proof

Here the inner f -sum is evaluated by (69) and so

rml+ν+a = qm( l2)+(ml+ν

2 )(−αj ; q)a

ml∑i=0

σq,i (αjqa)i+ν =

qm( l2)+(ml+ν

2 )+aν(−αj ; q)aανj

m∏t=1

(αt , αjqa; q−1)l = 0 (84)

for any 0 ≤ a ≤ l − 1 .

Page 43: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Proof

Next we consider the case a = l + k , k ∈ N. Then

rN = r(m+1)l+ν+k =

qm( l2)+(ml+ν

2 )+(l+k)ν(−αj ; q)l+kανj

m∏t=1

(αt , αjqk+1; q)l . (85)

Page 44: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl

Note that by Stihl

C (x , t) =ml∑

h=0

(xt)h(t; q)(m+1)l+λ−h· (86)

(−1)ml−hq(ml+λ+12 )−(ml+λ−h+1

2 )Σq,h

is a denominator polynomial in variable x and

Bl ,ν(t) =ml∑

h=0

(tx)h(t; q)ml+ν−h (87)

(−1)ml−hq(ml+ν2 )−(ml+ν−h

2 )Σq,h

is a denominator polynomial in variable t for

Fx(t) =∞∑

k=0

q(k2)(−tx)k

(t; q)k+1. (88)

Page 45: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl

By using the above modified approximations we may prove linear

independence measures for certain values of the functions Da(z)

and Ea(z), which can be regarded as q-analogues of Euler’s

divergent series and the usual exponential series.

For the q-exponential function Eq(z), our result asserts the linear

independence (over any number field) of the values 1 and Eq(αj)

(j = 1, ...,m) together with its measure having the exponent

ω = O(m), which sharpens the known exponent ω = O(m2)

obtained by a certain refined version of Siegel’s lemma Amou,

Matala-aho, Vaananen [On Siegel-Shidlovskii’s theory for

q-difference equations. Acta Arith. 127, 309–335 (2007)]

Page 46: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl

Let p be a prime number. Then we have the linear independence

of the p-adic numbers

∞∏n=1

(1 + kpn), k = 0, 1, ..., p − 1 (89)

over Q with a measure having the exponent

ω < 2p (90)

Page 47: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Valuations, heights

Let K be a fixed number field of degree κ = [K : Q], v a place of

K and | |v the associated absolute value on the completion Kv

with a local degree κv = [Kv : Qv ].

If the finite place v of K lies over the prime p, we write v |p, for an

infinite place v of K we write v |∞. We normalize the absolute

value | |v of K so that

|p|v = p−1, if v |p, (91)

|x |v = |x |, if v |∞, (92)

where | | denotes the ordinary absolute value in Q.

Page 48: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Valuations, heights

Further, the notation

||α||v = |α|κv/κv , κv = [Kv : Qv ], (93)

will be used in the sequel. The height H(α) of α is defined by the

formula

H(α) =∏v

||α||∗v , ||α||∗v = max{1, ||α||v} (94)

and the height H(α) of the vector α = t(α1, ..., αm) ∈ Km is given

by

H(α) =∏v

||α||∗v , ||α||∗v = maxi=1,...,m

{1, ||αi ||v}. (95)

Page 49: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Valuations, heights

Further, for any place v of K, and q ∈ K∗, ||q||v 6= 1, we define

the characteristic λ by

λ = λq =log H(q)

log ||q||v. (96)

Page 50: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

To state our results we denote

ω = ωq =u0

u0 + λqs0, (97)

where

s0 = m2 + m + m√

m2 + m, (98)

u0 = m2 + m + (m + 1)√

m2 + m. (99)

Now we fix a place v of K throughout the following.

Page 51: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

Let m ∈ Z+ be arbitrary, a, q, α1, ..., αm ∈ K∗, and |q|v < 1.

Denote by f (t) each of the functions

Dt(a), Eq(t),∞∏

n=0

(1− tqn) (100)

and assume

a /∈ q−N, αi /∈ q−N, αi /∈ αjqZ for all i 6= j , (101)

−(

1 +1

m +√

m2 + m

)< λq ≤ −1. (102)

Page 52: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

Then the numbers 1, f (α1), ..., f (αm) belonging to Kv are linearly

independent over K. Further, there exist positive constants

c , d ,H0 depending on a and αi such that

|k0 + k1f (α1) + ...+ kmf (αm)|v >c

Hωκ/κv+d(log H)−1/2(103)

for all

k = t(k0, k1, ..., km) ∈ Km+1 \ {0}

with H = max(H(k),H0).

Page 53: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

Here we note that λq ≤ −1 always holds for |q|v < 1, and the

following cases in particular assert λq = −1:

1. K = I, v is the infinite place of K, and 1/q ∈ ZK;

2. K = Q, v = p ∈ P, and q = pl , l ∈ Z+;

3. K = Q(√

5), q = ((1−√

5)/2)l , l ∈ Z+.

Now, if we take the value λ = −1, then we have

ω = m + 1 +√

m2 + m < 2m + 2 (104)

and in general we have ω = O(m),too.

Page 54: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

In the cases 1–3 we have

|L0 + L1f (α1) + ...+ Lmf (αm)|p >c

H2m+2+d(log H)−1/2(105)

for all

L = t(L0, L1, ..., Lm) ∈ Zm+1 \ {0}

with H = maxi=0,...,m

(|Li |,H0).

Page 55: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Modified Maier-Stihl/Applications

Let p ∈ P. Then in each of the four sets∞∏

n=1

(1 + kpn), k = 0, 1, ..., p − 1, (106)

∞∑n=1

pnn∏

i=1

(1 + kpi ), k = 0, 1, ..., p − 1, (107)

∞∏n=1

(1 + kp−n), k = 0, 1, ..., p − 1, (108)

∞∑n=1

p−nn∏

i=1

(1 + kp−i ), k = 0, 1, ..., p − 1. (109)

of p numbers we have the linear independence of the p numbers

over Q with a measure having an exponent

ω < 2p. (110)

Page 56: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-world

q-difference equations

Eq(qz) = (1− z)Eq(z), Eq(z) = (1 + z)Eq(qz). (111)

By using (111) one gets the well-known Euler formulae

Eq(z) =1

(z)∞, Eq(z) = (−z)∞. (112)

Page 57: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

q-identities

From (112) we get

Eq(t)Eq(−t) = 1 (113)

which implies

E1/q,1/q(t)Eq,q(qt) = 1 (114)

The following connects q-exponentials and q-divergent

∞∑n=0

q(n2)αn

(qz)nzn = 1 + αzD−α,q(qz) (115)

Page 58: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique

Prevost and Rivoal [Remainder Pade Approximants for the

Exponential Function. Constr. Approx. 25, 109–123 (2007)]

presented new Pade type approximations for the classical

exponential series using the remainder series technique, invented in

Prevost [A new proof of the irrationality of ζ(2) and ζ(3) using

Pade approximants. J. Comput. Appl. Math. 67, 219–235 (1996)].

To be more precise, they constructed approximations in variable t

for the exponential remainder series

Φxαj (t) =∞∑

k=0

(xαj)k

(1− 1/t)k, j = 1, ...,m (116)

Page 59: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

It is interesting that the above series Φx(t) and Dx(t) share a

formal similarity. Namely

Φx(t) =∞∑

k=0

(−tx)k

[1− ty ]k+1=∞∑

k=0

xk

(1− 1/t)k; (117)

[1− ty ]k+1 = (1− t · 0)(1− t · 1)(1− t · 2) · · · (1− t · k) (118)

and

Dx(t) =∞∑

k=0

q(k2)(−tx)k

[1− ty ; q]k+1= −1

t

∞∑k=0

(x/q)k

(1/t; 1/q)k+1; (119)

[1− ty ; q]k+1 = (1− tq0)(1− tq1)(1− tq2) · · · (1− tqk) (120)

Page 60: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

By applying the above modified Maier’s method we may give

another proof for the new Pade approximations for the exponential

remaider series

Φx(t) =∞∑

k=0

(−tx)k

[1− ty ]k+1=∞∑

k=0

xk

(1− 1/t)k. (121)

given by Prevost and Rivoal. Prevost and Rivoal showed that

Φx(t) =∞∑

n=0

Tn(−x)tn, (122)

where Tn(x) are Touchard polynomials

Page 61: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

Tn(x) =n∑

k=0

S2(n, k)xk (123)

defined with Stirling numbers of the second kind S2(n, k). Further,

In the following s1(n, k) denote the Stirling numbers of first kind.

So our aim is to construct explicit simultaneous Pade type

approximations in variable t for the series

Φ−αj (t) =∞∑

n=0

Tn(αj)tn, j = 1, ...,m. (124)

Page 62: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

We now define

Bl ,ν(t) =ml∑

h=0

bl ,ν,h(t)th, (125)

with

bl ,ν,h(t) = (−1)ml−h[1− ty ]ml+ν−hΣh(α), (126)

Then

Bl ,ν(t) =ml+ν∑H=0

bl ,ν,HtH , (127)

with

bl ,ν,H =∑

ml−i+f =H0≤f≤i+ν≤ml+ν

s1(i + ν, i + ν − f )σi (α) (128)

Page 63: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

Let l , ν,m ∈ N and choose m numbers α1, ..., αm. Then

Bl ,ν(t)Φ−αj (t) + Al ,ν,j(t) = Ll ,ν,j(t) (129)

with

[ml + ν,ml + ν − 1, (m + 1)l + ν] (130)

give a diagonal type Pade approximation of the second kind with a

free parameter ν.

Page 64: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier/Proof

Now we note that

[1− ay ]k = (1−a ·0)(1−a ·1)(1−a ·2) · · · (1−a ·(k−1)) = (131)

ak(1/a)k =k∑

i=0

s1(k , i)ak−i =k∑

i=0

s1(k , k − i)ai

Thus

Bl ,ν(t) =ml∑i=0

tml−i [1− ty ]i+νσi =

ml+ν∑H=0

tH∑

ml−i+f =H0≤f≤i+ν≤ml+ν

s1(i + ν, i + ν − f )σi =

ml+ν∑H=0

bl ,ν,HtH , (132)

Page 65: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

We next study the expansion of the product

Bl ,ν(t)Φ−αj (t) =∞∑

N=0

rNtN , (133)

where

rN =∑

H+n=N

bl ,ν,HTn(αj) (134)

Set

N = ml + ν + a, 0 ≤ a ≤ l − 1, a ∈ N. (135)

Then

H = ml − i + f , H + n = N. (136)

Page 66: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

It follows that n = i + ν − f + a and thus

rN =ml∑i=0

σi

i+ν∑f =0

s1(i + ν, i + ν − f )Ti+ν−f +a(αj) =

ml∑i=0

σi

i+ν∑f =0

a+ν+i−f∑e=0

s1(i + ν, i + ν − f )S2(a + ν + i − f , e)αej (137)

Denote for shortly I = i + ν, K = I − f . Then we are led to study

the following inner sums

Page 67: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

Ha =I∑

K=0

a+K∑e=0

s1(I ,K )S2(a + K , e)αej =

a+I∑e=0

αej

I∑K=e−a

s1(I ,K )S2(a + K , e) =

a+I∑e=0

αej

I∑K=0

s1(I ,K )S2(a + K , e) (138)

Page 68: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Generalized Stirling orthogonality

Let a ∈ N. For all I , e ∈ N we have

I∑K=0

s1(I ,K )S2(a + K , e) =∑

0≤b,d≤a

Cb,d(a)edδI ,e−b (139)

where the numbers Cb,d(a) do not depend on I and e.

Page 69: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

First

H0 =I∑

e=0

αej

I∑K=0

s1(I ,K )S2(K , e) =

I∑e=0

αej δIe = αI

j = αi+νj (140)

which by (9) shows

rml+ν =ml∑i=0

σiαi+νj = ανj

ml∑i=0

σiαij = 0. (141)

Page 70: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

Further

H1 =I∑

e=0

αej

I∑K=0

s1(I ,K )S2(K + 1, e) =

I∑e=0

αej (δI ,e−1 + eδIe) = αI+1

j + IαIj = αi+ν+1

j + (i + ν)αi+νj (142)

which by (9) shows

rml+ν+1 =ml∑i=0

σiαi+ν+1j + (i + ν)αi+ν

j =

(αν+1j + νανj )

ml∑i=0

σiαij + ανj

ml∑i=0

σi iαij = 0. (143)

Page 71: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

And in full generality by (139) we have

Ha =a+I∑e=0

αej

∑0≤b,d≤a

Cb,dedδI ,e−b =

∑0≤b,d≤a

Cb,d

a+I∑e=0

αej e

dδI ,e−b =

∑0≤b,d≤a

Cb,dαI+bj (I + b)d =

∑0≤b,d≤a

Cb,dαi+ν+bj (i + ν + b)d =

∑0≤b,d≤a

Cb,dαi+ν+bj

d∑g=o

(d

g

)ig (b + ν)d−g (144)

Page 72: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

Remaider series technique/Modified Maier

which gives

rml+ν+a =ml∑i=0

σiHa =

∑0≤b,d≤a

d∑g=0

Cb,dαi+ν+bj

(d

g

)(b + ν)d−g

ml∑i=0

σiαij i

g (145)

where 0 ≤ g ≤ a ≤ l − 1. Hence, again by using (9) we may

deduce

rml+ν+a = 0 (146)

for any 0 ≤ a ≤ l − 1 .

Page 73: Padé approximations of generalized …cc.oulu.fi/~tma/TOKYOSLIDES.pdfAbstract We shall present short proofs for type II Pad e approximations of the generalized hypergeometric and

References

I [1]. Amou M., Matala-aho T., Vaananen K.: On

Siegel-Shidlovskii’s theory for q-difference equations. Acta

Arith. 127, 309–335 (2007)

I [2]. Andrews G., Askey R., Roy R.: Special Functions,

Encyclopedia of Mathematics and its Applications, 71.

Cambridge University Press, Cambridge, 1999.

I [3]. Chudnovsky G. V.: Pade approximations to the

generalized hypergeometric functions. I. J. Math. pures et

appl. 58, 445–476 (1979)

I [4]. Hata M.: Remarks on Mahler’s Transcendence Measure

for e. J. Number Theory 54, 81–92 (1995)