• Aucun résultat trouvé

International Conference in honour of Srinivasa Ramanujan Centre for Advanced Study in Mathematics

N/A
N/A
Protected

Academic year: 2022

Partager "International Conference in honour of Srinivasa Ramanujan Centre for Advanced Study in Mathematics"

Copied!
18
0
0

Texte intégral

(1)

“NumberTheory & Discrete Mathematics”

International Conference in honour of Srinivasa Ramanujan Centre for Advanced Study in Mathematics

Panjab University, Chandigarh (October 2/6, 2000.)

Multiple Zeta Values

and

Euler-Zagier Numbers

by

Michel WALDSCHMIDT

http : //www.math.jussieu.fr/ miw/articles/ps/MZV.ps

(2)

2

1. Introduction

For s 1 , . . . , s k in Z with s 1 ≥ 2, ζ (s 1 , . . . , s k ) = X

n

1

> ··· >n

k

≥ 1

n 1 s

1

· · · n k s

k

.

k = 1 integer values of Riemann zeta function ζ (s).

Euler: ζ (s)π s ∈ Q for s even ≥ 2.

Fact: No known other algebraic relations between values of Riemann zeta function at positive integers.

Expected: there is no further relation : Are the numbers

π, ζ (3), ζ (5), . . . , ζ (2n + 1), . . . algebraically independent?

Means:

For n ≥ 0 and P ∈ Q [X 0 , X 1 , . . . , X n ] \ { 0 } , P ¡

π, ζ (3), ζ (5), . . . , ζ (2n + 1) ¢

6

= 0 ?

(3)

3

F. Lindemann (1882): π is transcendental.

R. Ap´ery (1978): ζ (3) is irrational.

T. Rivoal (2000): infinitely many irrational numbers among ζ (3), ζ (5), . . . , ζ (2n + 1), . . .

Theorem (T. Rivoal). Let ² > 0. For any suffi- ciently large n, the Q -vector space spanned by the n numbers

ζ (3), ζ (5), . . . , ζ (2n + 1) has dimension

≥ 1 − ²

1 + log 2 · log n.

The proof also yields:

There exists on odd integer j with 5 ≤ j ≤ 169 such that the three numbers

1, ζ (3), ζ (j )

are linearly independent over Q .

(4)

4

2. Sketch of Proof of Rivoal’s Theorem

Goal: Given a sufficiently large odd integer a, con- struct a sequence of linear forms in (a + 1)/2 vari- ables, with integer coefficients, such that the numbers

` n = p 0n +

(a − X 1)/2

i=1

p in ζ (2i + 1) satisfy, for n → ∞ ,

| ` n | = α n+o(n) and

| p in | ≤ β n+o(n) with

α ' a 2a and β ' (2e) 2a . It will follow that the (a + 1)/2 numbers

1, ζ (3), ζ (5), . . . , ζ (a)

span a Q -vector space of dimension at least 1 + log α

log β ' log a

1 + log 2

(Nesterenko’s Criterion).

(5)

5

Explicit construction of the linear forms Previous works of R. Ap´ery, F. Beukers,

E. Nikishin, K. Ball, D. Vasilyev, . . .

Pochammer symbol: (m) 0 = 1 and, for k ≥ 1, (m) k = m(m + 1) . . . (m + k − 1).

Set r = £

a(log a) 2 ¤

. Define

d m = l.c.m. of { 1, 2, . . . , m } ,

R n (t) = n! a 2r (t − rn + 1) rn (t + n + 2) rn (t + 1) a n+1 , S n =

X ∞ k=0

R n (k), ` n = d a 2n S 2n . Write the partial fraction expansion

R n (t) =

X a

i=1

X n

j =0

c ijn (t + j + 1) i with

c ijn = 1 (a − i)!

µ d dt

¶ a − i ¡

R n (t)(t + j + 1) a ¢

| t= − j − 1 .

(6)

6

Set p in = d 2n q i,2n where q 0,n = −

X a

i=1

X n

j =1

c ijn

j − 1

X

k=0

1 (k + 1) i and

q in =

X n

j=0

c ijn (1 ≤ i ≤ a).

Estimate for | p in | : c ijn = 1

2πi Z

| t+j +1 | =1/2

R n (t)(t + j + 1) i 1 dt.

Estimate for | ` n | : S n =

¡ (2r + 1)n + 1 ¢

!

n! 2r+1 · I n , I n =

Z

[0,1]

a+1

F (x) · dx 1 dx 2 . . . dx a+1 (1 − x 1 x 2 · · · x a+1 ) 2 , F (x 1 , x 2 , . . . , x a+1 ) =

à Q a+1

i=1 x r i (1 − x i ) (1 − x 1 x 2 · · · x a+1 ) 2r+1

! n

.

(7)

7

3. Shuffle Product for Series Reflexion Formula:

ζ (s)ζ (s 0 ) = X

n ≥ 1

n s · X

n

0

≥ 1

(n 0 ) s

0

= X

n>n

0

≥ 1

n s (n 0 ) s

0

+ X

n

0

>n ≥ 1

n s (n 0 ) s

0

+ X

n ≥ 1

n s s

0

= ζ (s, s 0 ) + ζ (s 0 , s) + ζ (s + s 0 ).

Example:

ζ (s) 2 = 2ζ (s, s) + ζ(2s).

For s = 2: ζ (2) = π 2 /6, ζ (4) = π 4 /90, ζ (2, 2) = X

m>n ≥ 1

(mn) 2 = π 4 120 · Other example:

ζ (2)ζ (3) = ζ (2, 3) + ζ (3, 2) + ζ (5).

(8)

8

Shuffle relations arising from the series representation.

ζ (s)ζ (s 0 ) = X

σ

ζ (σ),

where σ = (σ 1 , . . . , σ h ) ranges over the tuples ob- tained as follows:

s → ( s 1 0 s 2 · · · s k )

s 0 → ( 0 s 0 1 s 0 2 · · · 0 ) σ = ( s 1 s 0 1 s 2 + s 0 2 · · · s k ) Hence max { k, k 0 } ≤ h ≤ k + k 0 .

Example: k = k 0 = 1, s = s, s 0 = s 0 , then

s → (s 0) (0 s) s

s 0 → (0 s) (s 0 0) s 0

σ = (s s 0 ) (s 0 s) s + s 0

so that

{ σ 1 , σ 2 , σ 3 } = { (s, s 0 ), (s 0 , s), s + s 0 } .

(9)

9

Other Description:

Alphabet with two letters X = { x 0 , x 1 } . Words: X = { x a 0

1

x b 1

1

· · · x a 0

h

x b 1

h

} . Non-commutative polynomials: Qh X i .

For s ≥ 1 set y s = x s 0 1 x 1 .

For s = (s 1 , . . . , s k ) with s i ≥ 1, set x s = y s

1

· · · y s

k

= x s 0

1

1 x 1 x s 0

2

1 x 1 · · · x s 0

k

1 x 1 .

The number k of factors x 1 is the depth of the word x s and of the tuple s.

The number p = s 1 + · · · + s k of letters is the weight . The set of such x s ’s is X x 1 togeether with the null word e (corresponds to ∅ with k = 0).

Convergent words: x 0 X x 1 S

{ e } .

Set ζ (w) = ζ (s) for w = x s with ζ ( ∅ ) = ζ (e) = 1.

Convergent polynomials: Qh X i conv ⊂ Qh X i .

Extend ζ by linearity to Qh X i conv .

(10)

10

Law ∗ on Qh X i conv :

e ∗ w = w for w ∈ X x 1

and, for s ≥ 1 and t ≥ 1, w and w 0 in X x 1 , (y s w ) ∗ (y t w 0 ) =

y s (w ∗ y t w 0 ) + y t (y s w ∗ w 0 ) + y s+t (w ∗ w 0 ).

Then

x s ∗ x s

0

= X

σ

x σ .

Proposition. For w and w 0 in x 0 X x 1 , ζ (w)ζ (w 0 ) = ζ (w ∗ w 0 ).

Connection with quasi-symmetric functions Commutative infinite alphabet: t = { t 1 , t 2 , . . . } Formal power series: Q [[t]].

To w = x s ∈ X x 1 associate F w (t) = X

n

1

> ··· >n

k

≥ 1

t s n

11

· · · t s n

kk

. Then for w and w 0 in X x 1 we have

F w (t)F w

0

(t) = F w w

0

(t).

For w ∈ x 0 X x 1 , ζ (w) is the value of F w (t) with

t n = 1/n, (n ≥ 1).

(11)

11

4. Shuffle Product for Integrals

Let s = (s 1 , . . . , s k ) with s i ≥ 1; set p = s 1 + · · · + s k . Define ² i ∈ { 0, 1 } for 1 ≤ i ≤ p by

x s = x ²

1

· · · x ²

p

.

For instance for s = (2, 3) with p = 5:

x (2,3) = x 0 x 1 x 2 0 x 1 , (² 1 , . . . , ² 5 ) = (0, 1, 0, 0, 1).

Define differential forms:

ω 0 (t) = dt

t et ω 1 (t) = dt 1 − t · Let ∆ p be the simplex in R p :

p = { t ∈ R p ; 1 > t 1 > · · · > t p > 0 } . Proposition. For s 1 ≥ 2,

ζ (s) = Z

p

ω ²

1

(t 1 ) · · · ω ²

p

(t p ).

Example:

ζ (2, 3) = Z

5

dt 1

t 1 · dt 2

1 − t 2 · dt 3

t 3 · dt 4

t 4 · dt 5 1 − t 5 · Proof. Expand 1/(1 − t) = P

s ≥ 0 t s .

(12)

12

Shuffle on X :

e tt w = w tt e = w,

and, for i and j in { 0, 1 } , u and v in X ,

(x i u) tt (x j v) = x i (u tt x j v) + x j (x i u tt v )

Example. Computation of y 2 tt y 3 = x 0 x 1 tt x 2 0 x 1 : get x 0 x 1 x 2 0 x 1 once, x 2 0 x 1 x 0 x 1 three times and x 3 0 x 2 1 six times. Hence

y 2 tt y 3 = y 2 y 3 + 3y 3 y 2 + 6y 4 y 1 . Corollary.

ζ (w)ζ (w 0 ) = ζ (w tt w 0 ) for w and w 0 in x 0 X x 1 .

Proof. The Cartesian product ∆ p × ∆ p

0

is the union

of (p + p 0 )!/p!p 0 ! simplices.

(13)

13

Example. From

y 2 tt y 3 = y 2 y 3 + 3y 3 y 2 + 6y 4 y 1 we deduce

ζ (2)ζ(3) = ζ (2, 3) + 3ζ (3, 2) + 6ζ(4, 1).

On the other hand the shuffle relation for series gives ζ (2)ζ (3) = ζ (2, 3) + ζ (3, 2) + ζ (5),

hence

ζ (5) = 2ζ (3, 2) + 6ζ(4, 1).

There are further relations.

Example:

x 1 tt x 0 x 1 = x 1 x 0 x 1 + 2x 0 x 2 1 and

x 1 ∗ x 0 x 1 = x 1 x 0 x 1 + x 0 x 2 1 + x 2 0 x 1 , hence

x 1tt x 0 x 1 − x 1 ∗ x 0 x 1 = x 0 x 2 1 − x 2 0 x 1 . Fact: ζ (x 0 x 2 1 ) = ζ (x 2 0 x 1 ).

Euler : ζ (2, 1) = ζ (3).

(14)

14

Proposition. For w and w 0 in x 0 X x 1 , ζ (w)ζ (w 0 ) = ζ (w ∗ w 0 ),

ζ (w)ζ (w 0 ) = ζ (w tt w 0 ) and

ζ(x 1 tt w − x 1 ∗ w) = 0.

5. Symbolic Multizeta

Define Ze(s) for each s = (s 1 , . . . , s k ), with k ≥ 0 and s i ≥ 1. Next define Ze(w) for w in X x 1 by Ze(x s ) = Ze(s). Convergent symbols: Ze(s) with s 1 ≥ 2 or k = 0; these are the Ze(w) with w in x 0 X x 1 together with Ze(e) = Ze( ∅ ).

Algebra of Convergent MZV:

MZV conv is the commutative algebra over Q gener- ated by the convergent symbols Ze(s) with the rela- tions

Ze(w)Ze(w 0 ) = Ze(w ∗ w 0 ), Ze(w)Ze(w 0 ) = Ze(w tt w 0 ) and

Ze(x 1 tt w − x 1 ∗ w) = 0

for w and w 0 in x 0 X x 1 .

(15)

15

Main Diophantine Conjecture. The specializa- tion morphism from MZV conv into C which maps Ze(s) onto ζ (s) is injective.

Algebras MZV and MZV

tt

: generators Ze(s) with s = (s 1 , . . . , s k ), k ≥ 0, s j ≥ 1, and ∗ (resp. tt ) defined by

Ze(w) ∗ Ze(w 0 ) = Ze(w ∗ w 0 ) resp.

Ze(w) tt Ze(w 0 ) = Ze(w tt w 0 ) for w ∈ X x 1 .

Remark.

x 1 ∗ x 1 = 2x 2 1 + x 0 x 1 and x 1 tt x 1 = 2x 2 1 , hence

Ze(x 1 ) ∗ Ze(x 1 ) = 2Ze(x 2 1 ) + Ze(x 0 x 1 ) while

Ze(x 1 ) tt Ze(x 1 ) = 2Ze(x 2 1 )

and ζ (x 0 x 1 ) = ζ (2) 6 = 0.

(16)

16

Conjecture of

Zagier, Drinfeld, Kontsevich and Goncharov.

For p ≥ 2 let d p denote the dimension of the Z - module in MZV conv spanned by the 2 p 2 elements Ze(s) for s = (s 1 , . . . , s k ) of length p and s 1 ≥ 2.

Conjecture. We have

d 1 = 0, d 2 = d 3 = d 4 = 1 and

d p = d p 2 + d p 3 for p ≥ 4.

For each p ≥ 1, define Z p as the Q -vector space spanned by the Ze(s) with s convergent of weight p;

set Z 0 = Q . Then the sum of Z p (p ≥ 0) is direct, and the conjecture means

X

p ≥ 0

q p dim Q Z p = 1

1 − q 2 − q 3 ·

Remark: d p → ∞ by Rivoal’s result.

(17)

17

6. Further Results

Ecalle: for weight ´ ≤ 10, independent generators are Ze(2), Ze(3), Ze(5), Ze(7), Ze(9),

Ze(6, 2), Ze(8, 2).

Polylogarithms

Classical: for s ≥ 1 and | z | < 1, Li s (z) = X

n ≥ 1

z n n s .

Higher dimension: for s = (s 1 , . . . , s k ) with s i ≥ 1, Li s (z ) = X

n

1

> ··· >n

k

≥ 1

z n

1

n s 1

1

· · · n s k

k

Then

Li u

tt

v (z ) = Li u (z )Li v (z ) If s 1 ≥ 2, then Li s (1) = ζ (s).

Work of Petitot (Lille): the functions Li w , with w

in X , are linearly independent over C .

(18)

18

7. Related Topics

Further connections with:

Combinatoric (theory of quasisymmetric functions, Radford’s Theorem and Lyndon words)

Lie and Hopf algebras

Resurgent series (´ Ecalle’s theory)

Mixed Tate motives on Spec Z (Goncharov’s work) Monodromy of differential equations

Fundamental group of the projective line minus three points and Belyi’s Theorem

Absolute Galois group of Q

Group of Grothendieck-Teichm¨ uller Knots theory and Vassiliev invariants K -theory

Feynman diagrams and quantum field theory Quasi-triangular quasi-Hopf algebras

Drinfeld’s associator Φ KZ (connexion of Knizhnik-

Zamolodchikov).

Références

Documents relatifs

We consider Galois group Inf-gal (L/K) of the differential field extension L/K in general Galois theory of Umemura, and compare it to Galois group Gal(L P V /K) of the

In this process the double points lying in the line that we have removed create isolated points and we must allow to blow them up as well in order to cover the case when

Wojtkowiak, On ℓ -adi iterated integrals, III Galois ations on fundamental groups, Nagoya Math. Wojtkowiak, A Note on Funtional Equations of ℓ -adi Polylogarithms, Journal of

In the same way that PRC fields generalize PAC fields, Artin-Schreier structures enrich Galois groups by taking into account the orderings.. The corresponding Artin-Schreier

We should note here that in Rousseau’s thesis, this result has been proven also when the residue field of k is not perfect, and under some additional hypothesis on the reductive group

k-scheme X. the subgroup of the Galois group generated by the inertia groups over all components of D is abelian)... Then fi 1 D is an effective divisor on

In characteristic zero, any finite group scheme is reduced, and its representations certainly form a semisimple category. GROTHENDIECK: Éléments de Géométrie

rithm is the collection of data regarding individual elements of the Galois group, for which the standard method is to factor the polynomial moduloD. various &#34;good&#34;