0% found this document useful (0 votes)
23 views9 pages

Geometry of Multiple Zeta Values

This document discusses the geometry of multiple zeta values. It begins by introducing multiple zeta values and some known relations between them. It then discusses how multiple zeta values can be expressed as iterated integrals, and how the Drinfeld associator relates multiple zeta values to the fundamental group of P1 − {0, 1, ∞}. Finally, it discusses how the Grothendieck–Teichmüller group is related to mixed Tate motives over Q and Z.

Uploaded by

qwerty
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
23 views9 pages

Geometry of Multiple Zeta Values

This document discusses the geometry of multiple zeta values. It begins by introducing multiple zeta values and some known relations between them. It then discusses how multiple zeta values can be expressed as iterated integrals, and how the Drinfeld associator relates multiple zeta values to the fundamental group of P1 − {0, 1, ∞}. Finally, it discusses how the Grothendieck–Teichmüller group is related to mixed Tate motives over Q and Z.

Uploaded by

qwerty
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Geometry of multiple zeta values

Tomohide Terasoma

Abstract. Many relations are known between multiple zeta values ζ (k1 , . . . , kn ). A relation
coming from the associator condition for the Drinfeld associator, the generating function of
multiple zeta values, is a geometric relation. By the theory of mixed motives, we can control the
dimension of the rational linear hull of multiple zeta values. The harmonic shuffle relation also
comes from geometry, and more strongly, this is implied by the associator relation.

Mathematics Subject Classification (2000). Primary 14C30; Secondary 14F42.

Keywords. Multiple zeta value, mixed motif.

1. Introduction

Let k1 , . . . , kn ≥ 1 be integers such that kn ≥ 2. We define a multiple zeta value


ζ (k1 , . . . , kn ) by
 1
ζ (k1 , . . . , kn ) = .
k1
0<m1 <m2 <···<mn m1 · · · mknn

We define the weight w of the index (k1 , . . . , kn ) by w = k1 + · · · + kn . Many


relations between multiple zeta values are known for a long time; for example,

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

The first one is a quadratic relation and the second and the third ones are linear
relations between multiple zeta values. Several systematic methods are known to
produce a series of relations for multiple zeta values: iterated integral shuffle relation,
duality relation, harmonic shuffle relation, and so on. The iterated integral relation
and the duality relation are a part of the associator relation, which is closely related to
the Grothendieck–Teichmüller group. These relations produce many linear relations
between multiple zeta values. What is very interesting is that all the known rational
relations preserve the weights introduced above. So it is natural to expect that all
Q-relations come from geometry.

Proceedings of the International Congress


of Mathematicians, Madrid, Spain, 2006
© 2006 European Mathematical Society
628 Tomohide Terasoma

2. Iterated integral expression


A multiple zeta value has an iterated integral expression, which enables us to study
multiple zeta values from a geometric point of view. Let ω1 , . . . , ωn be one-forms
on a manifold X and let γ : [0, 1] → X be a path starting from a point a and ending
with a point b. An iterated integral is defined by
 
ω1 ω2 · · · ωn = pr ∗1 ω1 ∧ pr ∗2 ω2 ∧ · · · ∧ pr ∗n ωn ,
γ 0<tn <tn−1 <···<t1 <1

where pr i : [0, 1]n → [0, 1] is the i-th projection. A multiple zeta value is expressed
as   kn −1  k1 −1
dx dx dx dx
ζ (k1 , . . . , kn ) = ··· .
[0,1] x 1−x x 1−x
To control many relations it is convenient to consider the “generating function” of
multiple zeta values which is called the Drinfeld associator ([Dr]). Let A = Ce0 , e1 
be the non-commutative formal power series ring generated by e0 and e1 over C and
ω = e0xdx + ex−1
1 dx
an A-valued one-form. The Drinfeld associator  is defined by
  1−t 
 = (e0 , e1 ) = lim t −e1 exp ω t e0 ,
t→0 t
b ∞ b
where exp a ω = 1+ i=1 a ω · · · ω. The multiple zeta value (−1)n ζ (k1 , . . . , kn )
i-times
appears as the coefficient of e0kn −1 e1 · · · e0k1 −1 e1 .

3. Period of fundamental group of P1 − {0, 1, ∞}


The Drinfeld associator  describes the period of the fundamental group of
M0,4 = P1 − {0, 1, ∞} for tangential base points introduced by Deligne [De]. Let
→ →
|M0,4 | = {01, 10, . . . } be a set of tangential points of M0,4 and p, q be elements
of |M0,4 |. The comparison isomorphism
comp : Q[π1 (P1 − {0, 1, ∞}, p, q)]ˆ ⊗ C Qe0 , e1  ⊗ C (1)
defines a mixed Hodge structure on Q[π1 (P1 − {0, 1, ∞}, p, q)]ˆ. The Drinfeld as-
sociator is equal to the image of [0, 1] under the comparison homomorphism. The
Drinfeld associator satisfies the associator relations arising from the compatibility
condition for homomorphisms of mixed Hodge structures. In general, the associator
relations can be formulated via the compatibility condition for isomorphisms between
topological and de Rham fundamental groups. To interpret an associator as a func-
torial isomorphism we introduce the fundamental category (Fund) and the notion of
algebroids.
Geometry of multiple zeta values 629

Definition 3.1. Let K be a field. A K-algebroid U = {Uab }ab over a set B consists of
(1) a family of K-vector spaces {Uab }ab indexed by a, b ∈ B, and
(2) a family of homomorphisms
Ubc ⊗ Uab → Uac ,
which is associative.
We assume the following properties:
(1) For a ∈ B the associative ring Uaa has a unit.
(2) The vector space Uab is a free left Ubb -module of rank one and a free right
Uaa -module of rank one.
(3) The natural homomorphism
Ubc ⊗Ubb Uab → Uac
is an isomorphism.
We can define the notion of K-Hopf algebroids similarly. For a Q-algebraic variety,
by attaching a completion of the Q-linear hull of Betti and de Rham fundamental
groupoids, we get two functors
UB , UDR : (Var/Q) → (Hopf-alg/Q)
from the category (Var/Q) of Q-algebraic varieties to the category (Hopf-alg/Q) of
Q-Hopf algebroids. The comparison map obtained by Hodge theory gives a functorial
isomorphism of these two functors over C.
The object of the fundamental category (Fund) consists of four spaces M0,4 , M0,5 ,
M0,4 − {0, ∞} and the punctured disc ∗ , with tangential points. Morphisms are
generated by
(1) inclusions ∗ → M0,4 around points 0, 1 or ∞,
(2) “infinitesimal inclusions” M0,4 → M0,5 , and
(3) natural inclusion M0,4 → M0,4 − {0, ∞}.
We can define two functors of Betti and de Rham fundamental algebroids over a set
of tangential points UB (?) = Q[π1B (?)]ˆ and UDR (?) = Q[π1DR (?)]ˆ: (Fund) →
(Hopf-alg/Q). The set of functorial isomorphisms from UB ⊗ C to UDR ⊗ C is
denoted by Isom(UB ⊗ C, UDR ⊗ C).
Definition 3.2 (Associator). (1) Let ρ and e be the canonical generators of π1 (∗ , +)
∗ ∗
x in U ( ). Here + denotes a tangential point of  defined
and the dual of dx DR

by the local coordinate x. For an element ϕ ∈ Isom(U ⊗ C, U ⊗ C) we define


B DR

λ(ϕ) ∈ C× by λ(ϕ) = ϕ(log(ρ))/e. Thus we have a map


λ : Isom(UB ⊗ C, UDR ⊗ C) → C× .
630 Tomohide Terasoma

(2) We define the set of associator Ass as the inverse image


λ−1 (2πi) ⊂ Isom(UB ⊗ C, UB ⊗ C)
of 2πi under the map λ.
Remark 3.3. The functorial isomorphism ϕ is determined by the element  =
ϕ([0, 1]) ∈ UDR (M0,4 ) = Ce0 , e1 . The condition for an element  to be able
to be continued to a functorial isomorphism is nothing but the classical condition
for associators (see [Dr]). By this correspondence, the Drinfeld associator corre-
sponds to the classical comparison map. This has been communicated to the author
by M. Matsumoto.
Definition 3.4 (Grothendieck–Teichmüller group). (1) Let X be an algebraic variety
over Q. The Ql completion of the etale fundamental groupoid of a variety X ⊗ Q is
denoted by π1l (X ⊗ Q). The completion of the Ql -linear hull of π1l (X ⊗ Q) is denoted
by Ul (X). If X is one of ∗ = Spec(Q[[x]]), M0,4 or M0,5 , we can similarly define
similar tangential base points (see [I]). The Hopf algebroid Ul (X) gives a functor
Ul : (Fund) → (Hopf-alg/Ql ).
(2) Let ∗ = B, DR or l. The group GT∗ = Aut(U∗ ) of functorial automorphisms
of U∗ is called the ∗-Grothendieck–Teichmüller group. The groups GTB , GTDR and
GTl are pro-algebraic groups over Q, Q and Ql , respectively. For an element of ϕ ∈
Aut(U∗ ), by attaching ϕ(∗ ) ∈ Aut Hopf -alg (U∗ (∗ )), we have a homomorphism of
pro-algebraic groups:
λ : GT∗ = Aut(U∗ ) → Gm ,
where Gm is the multiplicative group. The kernel of λ : GT∗ → Gm is denoted by
(1)
GT∗ . It is a pro-nilpotent algebraic group.
(3) By the action of Gal(Q/Q) we have a natural homomorphism Gal(Q/Q) →
GTl .

4. Mixed Tate motives and Grothendieck–Teichmüller group


Following Levine, Voevodsky, Hanamura ([L]) we can define an abelian category
MTMQ of mixed Tate motives over Q. Goncharov ([G]) defined a full subcate-
gory MTMZ of mixed Tate motives over Z. By the classical comparison map (1),
UB+DR (P1 − {0, 1, ∞}) → → becomes a mixed Hodge structure of mixed Tate type.
01,10
This mixed Hodge structure is obtained by the nearby fiber (= the limit of mixed Hodge
→ →
structure) at 01, 10 of the variation of mixed Hodge structure U(P1 − {0, 1, ∞})ab
of two variables a, b. It is natural to expect that the near by fiber of a family of
mixed motives is also a mixed motif, which is not well formulated up to now. In fact,
for Hodge realization, a limit of a mixed Hodge structure depends on “the tangential
structure”, i.e. the choice of a branch of “log(t)”. But in our setting, we are very
happy to have the following.
Geometry of multiple zeta values 631

Theorem 4.1 (Deligne–Goncharov [DG], Terasoma [T]). There exists an object


UM (P1 − {0, 1, ∞}) → → in MTMZ whose Hodge realization is isomorphic to
01,10
UB+DR (P1 − {0, 1, ∞}) → → .
01,10
We state a consequence of the above theorem in the language of Tannakian
category. Let H DR : MTMZ → (VecQ ) be the de Rham realization functor, let
π1 (MTMZ , H DR ) be the Tannaka fundamental group (see [DM]) of MTMZ for the
fiber functor H DR and let λ : π1 (MTMZ , H DR ) → Gm be the homomorphism ob-
tained by the functor attaching the associated graded module for weight filtration.
The kernel of λ is denoted by π1 (MTMZ , H DR )(1) . Let U(π1 (MTMZ , H DR )) be
the completion of the Q-linear hull of π1 (MTMZ , H DR ). An object in MTMZ
corresponding to the representation U(π1 (MTMZ , H DR )) of π1 (MTMZ , H DR ) is
denoted by UM (π1 (MTMZ )). By the above theorem and the definition of Tan-
naka fundamental group, H DR UM (P1 − {0, 1, ∞}) → → ) becomes a (homogeneous)
01,10
U(π1 (MTMZ , H DR ))-module. Therefore the periods of UM (P1 − {0, 1, ∞}) are
controlled by those of UM (π1 (MTMZ )). Taking into account the real structures we
have the following corollary.
Corollary 4.2 (Conjectured by Zagier, ([G], [T])). Let Ln be the Q-vector space
generated by multiple zeta values of weight n. Then we have dimQ Ln ≤ dn , where dn
is defined by the generating function ∞ i=0 dn t = 1−t 2 −t 3 .
n 1

Either
(1) by comparing Gal(Q/Q) and π1 (MTMZ , H l ) via the homomorphism
Gal(Q/Q) → π1 (MTMZ , H l ) by using Hain–Matsumoto’s result, or
(2) by using the infinitesimal embedding UM (M0,4 ) → UM (M0,5 ) of mixed Tate
motives according to Deligne–Goncharov,
we have a homomorphism of pro-algebraic groups
Rep : π1 (MTMZ , H B ) → GTB
which is compatible with the weight homomorphisms λ. The Deligne–Ihara conjec-
ture asserts
Conjecture 4.3 (Deligne–Ihara). The homomorphism Rep is an isomorphism.
Remark 4.4. The injectivity of Rep is equivalent to the following statement:
Any mixed Tate motif over Z is generated by UM (P1 − {0, 1, ∞}) → → .
01,10
As a consequence the injectivity of Rep implies that periods of any mixed Tate motif
over Z are Q-linear combinations of multiple zeta values.
Note that U(π1 (MTMZ , H B )(1) ) is known to be a free Lie algebra generated by
c3 , c5 , c7 , . . . ([DG]), and the description of GTB is combinatorial. Therefore the
above conjecture is (in principle) a combinatorial question. It is answered in the
positive up to a quite high degree.
632 Tomohide Terasoma

5. Harmonic shuffle relation


In this section we introduce the harmonic shuffle relation according to Hoffman [H]
and its regularized version by Ihara–Kaneko–Zagier and Racinet ([IKZ], [R]). The
rearrangement of the product of two multiple zeta values expressed by infinite series
leads to a Z-linear combination of multiple zeta values: for example, we have
 1  1  1  1
ζ (2)ζ (2) = 2
· 2
= 4
+ 2 = ζ (4) + 2ζ (2, 2).
n m n n m2
2
0<n 0<m 0<n 0<n<m

A relation of this type is called harmonic shuffle relation. Using the technique of
regularization we also consider a relation of this type for non-convergent sums. As a
consequence Racinet and Ihara–Kaneko–Zagier obtained a wider class of relations,
the so-called “regularized harmonic shuffle relation”. We briefly recall the formula-
tion by Racinet of the regularized harmonic shuffle relation. We have the following
approximation of the partial summation
 1
ζN (k1 , . . . , kn ) = k1
0<m1 <m2 <···<mn <N m1 . . . mknn

= Pk1 ,...,kn (log N + γ ) + O(N − ) (2)

by a real coefficient polynomial Pk1 ,...,kn (T ). Using Boutet de Monvel–Zagier’s the-


orem, the generating series of Pk1 ,...,kn (T ) can be computed from the Drinfeld as-
sociator. By rearranging the partial summation up to N and using the asymptotic
expression (2), the product Pk1 ,...,kn (T )Pk1 ,...,k   (T ) becomes a Z-linear combination
n
of polynomials of the form Pk1 ,...,k  (T ). In order to formulate the result it is conve-
n
nient to introduce a new coproduct, the harmonic coproduct. Let W be a subalgebra
C ⊕ Ce0 , e1 e1 of Ce0 , e1 . This algebra is isomorphic to a weighted comple-
tion Cy1 , y2 . . .  of the free algebra generated by y1 = −e1 , y2 = −e0 e1 , . . . ,
yi = −e0i−1 e1 , . . . . We define a harmonic coproduct ∗ : W → W ⊗ W by an
algebra homomorphism given by


n
∗ (yn ) = yi ⊗ yn−i ,
i=0

with y0 = 1. Let DR = 1 + ϕ0 e0 + ϕ1 e1 be the Drinfeld associator and set DR,Y =


1 + ϕ1 e1 ∈ W . Then there is a unique formal power series DR (s) ∈ 1 + s 2 C[[s]]
such that
DR (−e0 ) DR (−e1 )
abDR,Y (e0 , e1 ) = ,
DR (−e0 − e1 )

DR,Y is the image of DR,Y in the abelianization C[[e0 , e1 ]]. We define the
where ab
−1
DR,Y by DR,Y = DR (y1 ) DR,Y ∈ W .
modified Y -series mod mod
Geometry of multiple zeta values 633

Theorem 5.1 (Racinet [R], Ihara–Kaneko–Zagier [IKZ] in a different formalism).


With the above notation, we have

DR,Y ) = DR,Y ⊗ DR,Y .


∗ (mod mod mod

In other words, mod


DR,Y is a group-like element under the harmonic coproduct.

6. Fake Hodge realization and harmonic shuffle relation


As is shown in Sections 3 and 5, the origin of the harmonic shuffle relation comes
from the rearrangement of partial series, and that of the associator relation comes
essentially from the functoriality for infinitesimal inclusions from M0,4 into M0,5 .
Though two origins are quite different, we can show that the associator relation implies
the harmonic shuffle relation. The contents of this section is a result of collaboration
with P. Deligne.
For an associator  = 1+ϕ0 e0 +ϕ1 e1 set Y = 1+ϕ1 e1 = Y (y1 , y2 , . . . ) ∈ W .
Theorem 6.1 (Deligne–Terasoma). (1) Let ab Y be the image of Y in the abelian-
ization C[[e0 , e1 ]]. Then there exists a unique element  (s) in 1 + s 2 C[[s]] such
that
 (−e0 )  (−e1 )
Y (e0 , e1 ) =
ab .
 (−e0 − e1 )
(2) Using  obtained in (1) we define
−1
Y (y1 , y2 , . . . ) :=
mod  (y1 ) Y (y1 , y2 , . . . ) ∈ W.

Then we have
Y ) = Y
∗ (mod ⊗ mod
mod
Y .

The proof of the above theorem is divided into two parts.

6.1. Fake Hodge realization attached to an associator. By Definition 3.2 an asso-


ciator defines Hopf algebroid objects U(M0,4 ) and U(M0,5 ) in the category

M = VecQ ×VecC VecQ


= {(V B , V DR , ϕ) | V B and V DR are Q-vector spaces, ϕ : V B ⊗ C −→ V DR ⊗ C
is an isomorphism of C vector spaces}.

The i-th projection induces a homomorphism pr i : U(M0,5 ) → U(M0,4 ) of Hopf


algebroids in M. For an algebroid U in M, we can introduce a notion of “U-modules”
in the category M. For a U(M0,5 )-module M we can define a U(M0,4 )-module
R j pr i M by using cohomological technology. Note that an isomorphism

R j pr i M B ⊗ C −→ R j pr i M DR ⊗ C
634 Tomohide Terasoma

is encoded in the object R j pr i M, which can be computed from the associator chosen
first. This isomorphism is called a fake comparison map associated to an associator.
We can also define the notion of “perverse” sheaf, and so on. These modules are
equipped with a mixed Hodge structure which is different from the natural one. They
are called a fake Hodge realization.

6.2. Multiplicative convolution. Following Deligne we introduce a geometric inter-


pretation of harmonic shuffle relation. Let A be the category of topological perverse
sheaves on M0,4 − {0, ∞} Gm smooth outside of {1} with a nilpotent monodromy.
We introduce an equivalence relation on A as follows. A morphism F1 → F2 in A is
equivalent if the induced homomorphism of vanishing cycles is an isomorphism. The
quotient category under this equivalence relation is denoted by A. Then A is again
an abelian category and we can show that this abelian category is equivalent to the
category of W B modules, where

W B = UB→ → log ρ1 ⊕ Q,
10,10

ρ1 being the canonical generator of local monodromy around {1}. For two perverse
sheaves F1 and F2 , we can define a biadditive functor ∗ : A × A → A by the
multiplicative convolution

F1 ∗ F2 = pH 1 R pr 5∗ (pr ∗1 (F1 ) ⊗ pr ∗2 (F2 )).

The equivalence class of the above convolution F1 ∗ F2 depends only on the equiva-
lence class of F1 and F2 and, as a consequence, the convolution induces a biadditive
functor ∗ : A × A → A. The corresponding ring homomorphism W B → W B ⊗ W B
is denoted by ∗ . This picture can be filled into the M = VecQ ×VecC VecQ
world with a comparison isomorphism attached to an associator . By the com-
putation of the cohomology of algebroids we can show that the de Rham realization
∗ : W DR → W DR ⊗ W DR is equal to the harmonic coproduct. By using the fake
comparison isomorphism we have the theorem.

References
[De] Deligne, P., Le groupe fondamental de la droite projective moins trois points. In Galois
groups over Q (Berkeley, CA, 1987), Math. Sci. Res. Inst. Publ. 16, Springer-Verlag,
New York 1989, 79–297.
[DG] Deligne, P., Goncharov, A., Groupe fondamentaux motivique de Tate mixte. Ann. Sci.
Ecole Norm. Sup. (4) 38 (2005), 1–56.
[DM] Deligne, P., Milne, J. S., Tannakian Categories. In Hodge Cycles, Motives, and Shimura
Varieties, Lecture Notes in Math. 900, Springer-Verlag, Berlin 1982, 101–228.
[Dr] Drinfeld, V. G., On quasitriangular quasi-Hopf algebras and on a group that is closely
connected with Gal(Q/Q). Leningrad Math. J. 2 (1991), 829–860.
Geometry of multiple zeta values 635

[G] Goncharov, A., Multiple polylogarithm and Mixed Tate motives. math.AG.0103059.
[H] Hoffman, M., Multiple harmonic series. Pacific J. Math. 152 (1992), 275–290.
[I] Ihara, Y., On the embedding of Gal(Q/Q) into GT. In Grothendieck theory of dessins
d’enfants, London Math. Soc. Lecture Note Ser. 200, Cambridge University Press, Cam-
bridge 1994, 289–321.
[IKZ] Ihara, K., Kaneko, M., Zagier, D., Derivation and double shuffle relations for multiple
zeta values. MPIM2004-100.
[L] Levine, M., Mixed Motives. Math. Surveys Monographs 57, Amer. Math. Soc., Provi-
dence, RI, 1998.
[R] Racinet, G., Doubles mélanges des polylogarithmes multiples aux racines de l’unité.
Publ. Math. Inst. Hautes Études Sci. 95 (2002), 185–231.
[T] Terasoma, T., Mixed Tate motives and multiple zeta values. Invent. Math. 149 (2002),
339–369.

Graduate School of Mathematical Sciences, University of Tokyo, 3-8-1 Komaba, Tokyo,


153-8914 Japan
E-mail: [email protected]

You might also like