25.03.2013 Views

Gauss hypergeometric equation as a representation of the formal KZ ...

Gauss hypergeometric equation as a representation of the formal KZ ...

Gauss hypergeometric equation as a representation of the formal KZ ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Acknowledgment<br />

The author expresses his deep gratitude to Pr<strong>of</strong>essor Kimio Ueno. Pr<strong>of</strong>essor<br />

Ueno is his academic supervisor and undertook <strong>the</strong> chief examiner for this<br />

<strong>the</strong>sis. Without his warm and careful mentorship, this <strong>the</strong>sis would not attain<br />

completion.<br />

He is also grateful to Pr<strong>of</strong>essor Kiichiro H<strong>as</strong>himoto, Pr<strong>of</strong>essor Yumiko Hironaka<br />

and Pr<strong>of</strong>essor Jun Murakami for <strong>as</strong>suming second readers on examining<br />

this <strong>the</strong>sis and for giving many useful advices. Fur<strong>the</strong>rmore, Pr<strong>of</strong>essor<br />

Hironaka gave <strong>the</strong> chances <strong>of</strong> speaking on <strong>the</strong> main topic <strong>of</strong> this <strong>the</strong>sis to <strong>the</strong><br />

author in twice.<br />

He thanks Doctor Jun-ichi Okuda and members <strong>of</strong> Ueno’s laboratory for<br />

valuable advice and discussions.<br />

Shu Oi<br />

Major in Ma<strong>the</strong>matical Science<br />

Graduate School <strong>of</strong> Science and Engineering<br />

W<strong>as</strong>eda University<br />

3-4-1, Okubo Shinjuku-ku<br />

Tokyo 169-8555, Japan<br />

e-mail: shu oi@toki.w<strong>as</strong>eda.jp


Contents<br />

1 Multiple zeta values, <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> 7<br />

1.1 Multiple zeta values and multiple polylogarithms <strong>of</strong> one variable 7<br />

1.1.1 Multiple zeta values . . . . . . . . . . . . . . . . . . . 7<br />

1.1.2 Multiple polylogarithms <strong>of</strong> one variable . . . . . . . . . 8<br />

1.1.3 The free shuffle algebra h generated by letters x, y . . . 9<br />

1.1.4 The double shuffle relation <strong>of</strong> MZVs . . . . . . . . . . . 11<br />

1.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> . . . . . . . . . . . . . . . . . . . . . 12<br />

1.2.1 The moduli space M0,n and <strong>the</strong> cubic coordinate . . . 12<br />

1.2.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on M0,n . . . . . . . . . . . . 13<br />

1.2.3 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable . . . . . . . . . 14<br />

1.2.4 The normalized fundamental solution <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong><br />

<strong>equation</strong> <strong>of</strong> one variable and Drinfel’d <strong>as</strong>sociator . . . . 15<br />

1.3 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> . . . . . . . . . . . . . . . . . 17<br />

1.3.1 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and its solution . . . . 18<br />

1.3.2 Existing researches about <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

and MZVs . . . . . . . . . . . . . . . . . . . . . . 19<br />

2 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> 21<br />

2.1 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> . . . . . . . . . . . 21<br />

2.2 Analytic properties <strong>of</strong> MPLs . . . . . . . . . . . . . . . . . . . 22<br />

2.3 One dimensional <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> . 25<br />

2.4 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong><br />

1<strong>KZ</strong> <strong>equation</strong> . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

2.5 The expression <strong>of</strong> <strong>the</strong> <strong>hypergeometric</strong> function by MPLs . . . 27<br />

2.5.1 Definitions and Notations . . . . . . . . . . . . . . . . 27<br />

2.5.2 Main Result . . . . . . . . . . . . . . . . . . . . . . . . 28<br />

2.5.3 The image <strong>of</strong> word in H by <strong>the</strong> <strong>representation</strong> ρ0 . . . . 29<br />

2.5.4 Asymptotic properties <strong>of</strong> ρ0(H0) and Φ0 . . . . . . . . 32<br />

2.5.5 Pro<strong>of</strong> <strong>of</strong> Theorem 5 . . . . . . . . . . . . . . . . . . . . 32<br />

iii


iv<br />

2.6 The connection formula between <strong>the</strong> regular solutions to <strong>Gauss</strong><br />

<strong>hypergeometric</strong> <strong>equation</strong> at z = 0 and z = 1 . . . . . . . . . . 34<br />

2.6.1 The inverse <strong>of</strong> <strong>the</strong> fundamental solution matrix on <strong>the</strong><br />

neighborhood <strong>of</strong> z = 1 . . . . . . . . . . . . . . . . . . 34<br />

2.6.2 The expansion <strong>of</strong> <strong>the</strong> connection matrix <strong>as</strong> a series <strong>of</strong><br />

<strong>the</strong> zeta values . . . . . . . . . . . . . . . . . . . . . . 36<br />

2.6.3 Functional relations obtained from <strong>the</strong> (1, 1)-element<br />

<strong>of</strong> <strong>the</strong> connection formula (2.21) . . . . . . . . . . . . . 39<br />

2.6.4 Various examples <strong>of</strong> functional relations <strong>of</strong> MPLs . . . 43<br />

2.7 Functional relations derived from <strong>the</strong> connection formula between<br />

irregular solutionsChapter. 22.7. FUNCTIONAL RE-<br />

LATIONS DERIVED FROM THE CONNECTION FORMULA<br />

BETWEEN IRREGULAR SOLUTIONSCHAPTER. 22.7 FUNC-<br />

TIONAL RELATIONS DERIVED FROM Z = 0, ∞CHAPTER.<br />

22.7 FUNCTIONAL RELATIONS DERIVED FROM Z =<br />

0, ∞2.7 Functional relations derived from z = 0, ∞ . . . . . . 50<br />

2.7.1 The fundamental solution matrix on <strong>the</strong> neighborhood<br />

<strong>of</strong> z = ∞ . . . . . . . . . . . . . . . . . . . . . . . . . . 51<br />

2.7.2 The functional relations derived from <strong>the</strong> (1, 1)-element<br />

<strong>of</strong> <strong>the</strong> connection formula between z = 0 and ∞ . . . . 56<br />

3 General <strong>representation</strong>s and many variable <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>s<br />

63<br />

3.1 Generalized <strong>hypergeometric</strong> <strong>equation</strong>s and multiple zeta values 63<br />

3.2 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong> and<br />

MZVs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66<br />

3.2.1 The <strong>formal</strong> generalized <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variables . . 66<br />

3.2.2 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong> 67<br />

3.3 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables . . . . . . . . . . . . 68<br />

3.3.1 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables . . . . . . . . 68<br />

3.3.2 The reduced bar algebra and iterated integral <strong>of</strong> two<br />

variables . . . . . . . . . . . . . . . . . . . . . . . . . . 69<br />

3.3.3 The normalized fundamental solution to <strong>the</strong> <strong>formal</strong><br />

2<strong>KZ</strong> <strong>equation</strong> . . . . . . . . . . . . . . . . . . . . . . . 71<br />

3.3.4 Decomposition <strong>the</strong>orem <strong>of</strong> <strong>the</strong> normalized fundamental<br />

solution and <strong>the</strong> generalized harmonic product relation 73<br />

3.3.5 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> . . . . . . 75


Introduction<br />

In this <strong>the</strong>sis, we discuss relationship between <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

and multiple zeta values through <strong>the</strong> viewpoint <strong>of</strong> <strong>representation</strong>s <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable b<strong>as</strong>ed on [O]. We regard <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and apply<br />

algebraic <strong>the</strong>ory <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and multiple zeta values. And<br />

<strong>the</strong>n, comparing <strong>the</strong> results with <strong>the</strong> connection formul<strong>as</strong> <strong>of</strong> <strong>the</strong> solutions to<br />

<strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>, we obtain various relations <strong>of</strong> <strong>the</strong> multiple<br />

polylogarithms and multiple zeta values.<br />

Multiple zeta values are real numbers defined <strong>as</strong> <strong>the</strong> series<br />

ζ(k1, . . . , kr) =<br />

1<br />

n k1<br />

1 · · · nkr r<br />

n1>n2>···>nr>0<br />

for index (k1, . . . , kr) <strong>of</strong> positive integers (k1 > 1). They are expansions <strong>of</strong><br />

Riemann zeta values (that is, <strong>the</strong> c<strong>as</strong>e <strong>of</strong> r = 1). Multiple zeta values <strong>of</strong> <strong>the</strong><br />

c<strong>as</strong>e <strong>of</strong> r = 2 are introduced and researched by Euler, and in 1990s, <strong>the</strong>y<br />

are re-discovered by H<strong>of</strong>fman and Zagier. Since <strong>the</strong>n, algebraic properties<br />

<strong>of</strong> <strong>the</strong>m have been studied actively. There exists various relations among<br />

multiple zeta values, and obtaining all <strong>the</strong> relations among multiple zeta<br />

values and finding <strong>the</strong> structure <strong>of</strong> Q-vector space spanned by all multiple<br />

zeta values are great aims on <strong>the</strong> number <strong>the</strong>ory. For <strong>the</strong>se purposes, Ihara-<br />

Kaneko-Zagier([I<strong>KZ</strong>]) and Racinet([R]) conjectured that all <strong>the</strong> Q-linearly<br />

relations <strong>of</strong> multiple zeta values can be obtained from <strong>the</strong> regularized double<br />

shuffle relation <strong>of</strong> multiple zeta values.<br />

Multiple zeta values are regarded <strong>as</strong> <strong>the</strong> limit <strong>of</strong> multiple polylogarithms<br />

zn1 Lik1,...,kr(z) =<br />

n1>n2>···>nr>0<br />

n k1<br />

1 · · · n kr<br />

r<br />

<strong>as</strong> z tends to 1−0 and considering differential <strong>equation</strong>s satisfied by multiple<br />

polylogarithms is very useful to study multiple zeta values. Moreover, multiple<br />

polylogarithms are very interesting subject in <strong>the</strong>mselves. For instance,<br />

1


2 Introduction<br />

<strong>the</strong>y have various functional relations such <strong>as</strong> Euler’s inversion formula, Landen’s<br />

formula and five term relations <strong>of</strong> di-logarithms (that is, <strong>the</strong> c<strong>as</strong>e <strong>of</strong><br />

r = 1, k1 = 2).<br />

The <strong>formal</strong> Knizhnik-Zamolodchikov <strong>equation</strong> (<strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>,<br />

for short) , introduced by Drinfel’d([D]), is a differential <strong>equation</strong> on <strong>the</strong><br />

moduli space M0,n <strong>of</strong> <strong>the</strong> configuration space <strong>of</strong> n points on P 1 over <strong>the</strong><br />

action <strong>of</strong> PGL(2, C). The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> is derived from <strong>the</strong> <strong>equation</strong><br />

satisfied by correlation functions on <strong>the</strong> con<strong>formal</strong> field <strong>the</strong>ory and appears<br />

on <strong>the</strong> knot <strong>the</strong>ory because <strong>the</strong>ir monodromy <strong>representation</strong> yields <strong>the</strong> braid<br />

group([Kon],[LM]). For n = 4, <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on M0,4 is regarded<br />

<strong>as</strong> an <strong>equation</strong> <strong>of</strong> one variable on P 1 ;<br />

dG<br />

dz =<br />

<br />

X<br />

z<br />

<br />

Y<br />

+ G,<br />

1 − z<br />

where X, Y are non-commutative <strong>formal</strong> elements. Recently algebraic properties<br />

<strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable and its solutions are explored<br />

([D],[G2],[I<strong>KZ</strong>],[OkU] and so on). Especially, by means <strong>of</strong> <strong>the</strong> <strong>the</strong>ory <strong>of</strong> iterated<br />

integrals and free shuffle algebr<strong>as</strong>, <strong>the</strong> generating function <strong>of</strong> all multiple<br />

polylogarithms with indeterminate elements X, Y becomes <strong>the</strong> fundamental<br />

solution to <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable normalized at z = 0, and<br />

<strong>the</strong> connection coefficient <strong>of</strong> certain solutions, referred to <strong>as</strong> Drinfel’d <strong>as</strong>sociator,<br />

is expressed <strong>as</strong> <strong>the</strong> generating function <strong>of</strong> all <strong>the</strong> multiple zeta values.<br />

Drinfel’d <strong>as</strong>sociator plays a very important role on <strong>the</strong> arithmetic geometry.<br />

In this context, Deligne-Ter<strong>as</strong>oma([DT]) and Furusho([F]) showed that <strong>the</strong><br />

pentagon and hexagon relations <strong>of</strong> Drinfel’d <strong>as</strong>sociator leads <strong>the</strong> double shuffle<br />

relation <strong>of</strong> multiple zeta values. Hence it is conjectured that <strong>the</strong> relations<br />

<strong>of</strong> Drinfel’d <strong>as</strong>sociator yields all <strong>the</strong> Q-linearly relations <strong>of</strong> multiple zeta values.<br />

This conjecture suggests that all <strong>the</strong> relations <strong>of</strong> multiple zeta values<br />

can be interpreted <strong>as</strong> a connection problem <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>.<br />

Fur<strong>the</strong>rmore, one can define <strong>the</strong> iterated integral <strong>of</strong> many variables by<br />

using <strong>the</strong> Orlik-Solomon algebra <strong>of</strong> hyperplane arrangements and Chen’s reduced<br />

bar construction([C2],[Koh],[OT],[B]) and can discuss <strong>the</strong> algebraic<br />

structure <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> many variables and <strong>the</strong> relationship<br />

to multiple zeta values([B],[OU]). For multiple zeta values, it suffice to consider<br />

<strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>s <strong>of</strong> one and two variables, however, <strong>the</strong> <strong>formal</strong><br />

<strong>KZ</strong> <strong>equation</strong> <strong>of</strong> three or more variables is also important in <strong>the</strong> sight <strong>of</strong> a<br />

study <strong>of</strong> differential <strong>equation</strong>s. It is thought <strong>of</strong> <strong>as</strong> corresponding bigger cl<strong>as</strong>s<br />

<strong>of</strong> relations containing multiple L values and so on.


On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable is <strong>the</strong> universal<br />

Fuchsian <strong>equation</strong> which h<strong>as</strong> three regular singular points at 0, 1, ∞ on<br />

P 1 . Thus we can obtain specific Fuchsian <strong>equation</strong> by replacing <strong>the</strong> <strong>formal</strong><br />

elements X, Y with certain square matrix (that is, taking a <strong>representation</strong>).<br />

On <strong>the</strong>se specific <strong>equation</strong>s, algebraic properties <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> is<br />

inherited restrictively via <strong>representation</strong>s. And <strong>the</strong>n, under <strong>representation</strong>s,<br />

one can consider analytic manipulations (taking limit, infinite sum, specializing<br />

indeterminate elements, and so on) to relations derived from <strong>the</strong> <strong>formal</strong><br />

1<strong>KZ</strong> <strong>equation</strong>. Moreover if <strong>the</strong> specific Fuchsian <strong>equation</strong> h<strong>as</strong> certain analytic<br />

properties, such <strong>as</strong> an integral expression <strong>of</strong> <strong>the</strong> solution, an connection<br />

formula and so on, we can obtain various non-trivial functional relations <strong>of</strong><br />

multiple polylogarithms and relations <strong>of</strong> multiple zeta values by combining<br />

<strong>the</strong>se analytic properties and <strong>the</strong> algebraic <strong>the</strong>ory on <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>.<br />

However, previously syn<strong>the</strong>sizing studies on Fuchsian <strong>equation</strong>s <strong>as</strong> a<br />

<strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> are not made.<br />

The most important and elemental <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

is <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

z(1 − z) d2f df<br />

+ (γ − (α + β + 1)z) − αβf = 0,<br />

dz2 dz<br />

where α, β, γ are complex parameters. It is a general form <strong>of</strong> second order<br />

Fuchsian <strong>equation</strong>s which have three regular singular points on P 1 . Fur<strong>the</strong>rmore<br />

<strong>the</strong> integral expression <strong>of</strong> <strong>the</strong> solution, which is known <strong>as</strong> <strong>Gauss</strong><br />

<strong>hypergeometric</strong> function, and <strong>the</strong> connection formul<strong>as</strong> are well known explicitly<br />

and completely ([WW],[Ha] and so on). Especially, <strong>the</strong> connection<br />

coefficients can be expressed <strong>as</strong> a factor <strong>of</strong> Gamma functions.<br />

In previous time, Ter<strong>as</strong>oma([T1]) said that values <strong>of</strong> Selberg type integrals<br />

(<strong>the</strong>y contain <strong>the</strong> limit <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> function <strong>as</strong> z tends to 1) can<br />

be written <strong>as</strong> series <strong>of</strong> multiple zeta values, and Ohno-Zagier([OZ]) showed<br />

that <strong>the</strong> generating function <strong>of</strong> sum <strong>of</strong> multiple zeta values fixed weight,<br />

depth and height satisfies <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and derived various<br />

relations <strong>of</strong> multiple zeta values. However, <strong>the</strong>se existing research did not<br />

refer to <strong>the</strong> viewpoint <strong>of</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and <strong>of</strong> a<br />

connection problem <strong>of</strong> a differential <strong>equation</strong>. Additionally, even though it is<br />

well known that <strong>Gauss</strong> <strong>hypergeometric</strong> function can be expressed <strong>as</strong> a series<br />

<strong>of</strong> multiple polylogarithms by solving <strong>hypergeometric</strong> <strong>equation</strong> by successive<br />

integration([AoK]), but <strong>the</strong> explicit expression is not given yet.<br />

Compared with <strong>the</strong>se existing researches, on [O], <strong>the</strong> author suggested <strong>the</strong><br />

framework <strong>of</strong> obtaining relations <strong>of</strong> multiple polylogarithms by <strong>the</strong> methods<br />

above via <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and establish <strong>the</strong> way to<br />

3


4 Introduction<br />

obtain relations systematically. Moreover by regarding <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and by calculating<br />

<strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution to <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>,<br />

<strong>the</strong> author succeeded in obtaining <strong>the</strong> expansion <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong><br />

function <strong>as</strong> a series <strong>of</strong> multiple polylogarithms. This expansion is effected<br />

on <strong>the</strong> universal covering space <strong>of</strong> P 1 − {0, 1, ∞} and can apply to <strong>the</strong> connection<br />

problem <strong>of</strong> solutions not only at z = 0, 1 but also at z = ∞. The<br />

author derived new functional relations <strong>of</strong> multiple polylogarithms from <strong>the</strong><br />

connection formula <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> between solutions at<br />

z = 0 and z = 1, and solutions at z = 0 and z = ∞. Since relations <strong>of</strong><br />

multiple zeta values derived <strong>the</strong> limit <strong>as</strong> z tends to 1 − 0 <strong>of</strong> <strong>the</strong>m restore<br />

Ohno-Zagier’s result, <strong>the</strong>se results gives <strong>the</strong> expansion <strong>of</strong> Ohno-Zagier’s result<br />

to functional relations <strong>of</strong> multiple polylogarithms and <strong>the</strong> interpretation<br />

<strong>of</strong> <strong>the</strong>m on <strong>the</strong> <strong>the</strong>ory <strong>of</strong> differential <strong>equation</strong>s. Fur<strong>the</strong>rmore since connection<br />

formul<strong>as</strong> between z = 0, 1 and between z = 0, ∞ are clearly independent algebraically,<br />

<strong>the</strong> relations <strong>of</strong> multiple zeta values derived from z = 0, ∞ might<br />

differ essentially from Ohno-Zagier relation.<br />

This <strong>the</strong>sis is organized <strong>as</strong> follows. This <strong>the</strong>sis consists <strong>of</strong> three chapters.<br />

Chapter 1 is an introduction <strong>of</strong> b<strong>as</strong>ic knowledges and existing results. In<br />

Section 1.1, we define multiple zeta values and multiple polylogarithms, and<br />

show some famous relations <strong>of</strong> multiple zeta values. We also give <strong>the</strong> iterated<br />

integral expression <strong>of</strong> multiple polylogarithms and multiple zeta values by using<br />

<strong>the</strong> free shuffle algebra. In Section 1.2, we review <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

(especially <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable), its normalized fundamental<br />

solutions and Drinfel’d <strong>as</strong>sociator. In Section 1.3, we outline <strong>Gauss</strong><br />

<strong>hypergeometric</strong> <strong>equation</strong> and introduce existing researches on <strong>the</strong> <strong>equation</strong><br />

and multiple zeta values, in particular Ohno-Zagier’s results.<br />

Chapter 2 is <strong>the</strong> main part <strong>of</strong> this <strong>the</strong>sis. In this chapter, we discuss a<br />

relationship between <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and multiple polylogarithms<br />

from a viewpoint <strong>of</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> b<strong>as</strong>ed<br />

on [O]. In Section 2.1, we define a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

and its fundamental solutions, and in Section 2.2 we discuss analytic properties<br />

<strong>of</strong> multiple polylogarithms (Proposition 1) in order to show convergence<br />

<strong>of</strong> <strong>representation</strong>s <strong>of</strong> <strong>the</strong> fundamental solution. In Section 2.3, we express<br />

one dimensional <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>. They are <strong>the</strong><br />

simplest examples <strong>of</strong> <strong>representation</strong>s, but <strong>the</strong>se <strong>representation</strong>s are trivial<br />

and <strong>the</strong>re is no non-trivial relation <strong>of</strong> multiple zeta values.<br />

In Section 2.4, we redefine <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> <strong>as</strong> a <strong>representation</strong><br />

<strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>, and in Section 2.5, we calculate <strong>the</strong><br />

<strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution normalized at z = 0 concretely


and derive <strong>the</strong> expansion <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> function <strong>as</strong> a series <strong>of</strong> multiple<br />

polylogarithms (Theorem 5). This is <strong>the</strong> first main result <strong>of</strong> this <strong>the</strong>sis.<br />

This expansion is regarded <strong>as</strong> <strong>the</strong> iterated integral expression <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong><br />

function.<br />

Next, in Section 2.6, we apply <strong>the</strong> same method <strong>of</strong> Section 2.5 to <strong>the</strong><br />

<strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution normalized at z = 1 (Proposition<br />

12), and consider <strong>the</strong> connection formula between <strong>the</strong> regular solution<br />

at z = 0 and z = 1. Then we obtain various functional relations <strong>of</strong> multiple<br />

polylogarithms (Theorem 15) <strong>as</strong> <strong>the</strong> second main result <strong>of</strong> this <strong>the</strong>sis.<br />

Fur<strong>the</strong>rmore, we have various relations <strong>of</strong> multiple zeta values, <strong>as</strong> known <strong>as</strong><br />

Ohno-Zagier relation, <strong>as</strong> <strong>the</strong> limit <strong>of</strong> <strong>the</strong>se functional relations. Specializing<br />

<strong>the</strong>se relations, we show also some interesting relations, for instance <strong>the</strong> sum<br />

formula <strong>of</strong> multiple polylogarithms (Proposition 17), which is <strong>the</strong> functional<br />

relation <strong>of</strong> multiple polylogarithms extending <strong>the</strong> sum formula <strong>of</strong> multiple<br />

zeta values.<br />

For <strong>the</strong> connection formula between singular solutions or solutions on<br />

z = ∞ are generally complicated. However, we describe partially <strong>the</strong> relations<br />

derived from <strong>the</strong> connection formul<strong>as</strong> between <strong>the</strong> solutions on z = 0<br />

and z = ∞ (Proposition 22 and Corollary 23) in Section 2.7. These relations<br />

contains <strong>the</strong> expression <strong>of</strong> <strong>the</strong> values <strong>of</strong> Riemann zeta function on even integers<br />

by using Bernoulli numbers and expressions <strong>of</strong> multiple zeta values such<br />

<strong>as</strong> ζ(m, 1, . . . , 1) <strong>as</strong> a polynomial <strong>of</strong> Riemann zeta values, and are thought <strong>of</strong><br />

<strong>as</strong> different relations from Ohno-Zagier relation.<br />

The main <strong>the</strong>me <strong>of</strong> this <strong>the</strong>sis, a relationship between <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> and multiple zeta values via a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong><br />

<strong>KZ</strong> <strong>equation</strong>, is <strong>the</strong> first step <strong>of</strong> <strong>the</strong> investigation on <strong>the</strong> <strong>representation</strong> <strong>the</strong>ory<br />

<strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>. We mention general <strong>representation</strong>s <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable and <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two<br />

variables in Chapter 3. In Section 3.1, we consider <strong>the</strong> <strong>representation</strong>s <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable with respect to <strong>the</strong> generalized <strong>hypergeometric</strong><br />

<strong>equation</strong>s and review existing works on <strong>the</strong>se <strong>equation</strong>s and multiple<br />

zeta values. In Section 3.2, we introduce <strong>the</strong> <strong>formal</strong> generalized <strong>KZ</strong> <strong>equation</strong><br />

<strong>of</strong> one variable, which is <strong>the</strong> universal Fuchsian <strong>equation</strong> with many regular<br />

singular points on P 1 , and <strong>the</strong>ir <strong>representation</strong>s. In Section 3.3, we discuss<br />

algebraic properties <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables according to<br />

[OU], which is a subsequent work <strong>of</strong> <strong>the</strong> author after [O], and explain its<br />

<strong>representation</strong>s such <strong>as</strong> Appell <strong>hypergeometric</strong> <strong>equation</strong>s. These topics are<br />

under investigation, but very important subjects to apply our research.<br />

5


Chapter 1<br />

Multiple zeta values, <strong>the</strong> <strong>formal</strong><br />

<strong>KZ</strong> <strong>equation</strong> and <strong>Gauss</strong><br />

<strong>hypergeometric</strong> <strong>equation</strong><br />

In this chapter, we preliminary recall <strong>the</strong> fundamental properties and known<br />

results on multiple zeta values, multiple polylogarithms, shuffle algebr<strong>as</strong>, <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>. Algebraic <strong>as</strong>pects<br />

<strong>of</strong> multiple polylogarithms and <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable play<br />

essential role in this <strong>the</strong>sis.<br />

1.1 Multiple zeta values and multiple polylogarithms<br />

<strong>of</strong> one variable<br />

1.1.1 Multiple zeta values<br />

Multiple zeta values (MZVs, for short) are real numbers <strong>as</strong>sociated with <strong>the</strong><br />

index <strong>of</strong> positive integers (k1, k2, . . . , kr), k1 ≥ 2 defined <strong>as</strong><br />

ζ(k1, k2, . . . , kr) =<br />

<br />

n1>···>nr>0<br />

1<br />

n k1<br />

1 · · · n kr<br />

r<br />

. (1.1)<br />

We call k1 + · · · + kr <strong>the</strong> weight and r <strong>the</strong> depth <strong>of</strong> index (k1, . . . , kr). If<br />

r = 1, <strong>the</strong> multiple zeta values ζ(k1) are nothing but <strong>the</strong> values <strong>of</strong> Riemann<br />

zeta function at positive integers. There are many relations among multiple<br />

zeta values, for instance;<br />

(i) The sum formula (Granville[Gr], Zagier[Z2])<br />

7


8 Chapter. 1<br />

For all k > r ≥ 1,<br />

(ii) The duality formula<br />

<br />

k1+···+kr=k<br />

k1≥2,ki≥1<br />

ζ(k1, . . . , kr) = ζ(k). (1.2)<br />

For an index k = (k1, . . . , kr) = (a1 + 1, 1, . . . , 1 , . . . , <strong>as</strong> + 1, 1, . . . , 1 )<br />

<br />

<br />

b1−1 times<br />

bs−1 times<br />

(ai, bi ≥ 1), we define <strong>the</strong> dual index k ′ <strong>as</strong><br />

Then we have<br />

(iii) Ohno relation ([Oh])<br />

k ′ = (bs + 1, 1, . . . , 1 , . . . , b1 + 1, 1, . . . , 1 ).<br />

<br />

<br />

<strong>as</strong>−1 times<br />

a1−1 times<br />

ζ(k) = ζ(k ′ ). (1.3)<br />

For all index k = (k1, . . . , kr), k1 ≥ 2, we denote <strong>the</strong> dual index by<br />

k ′ = (k ′ 1, . . . , k ′ r ′). Then<br />

<br />

ζ(k1+c1, . . . , kr+cr) =<br />

<br />

ζ(k ′ 1+c ′ 1, . . . , k ′ r ′ +c′ r ′) (1.4)<br />

c1+···+cr=l<br />

c ′ 1 +···+c′<br />

r ′=l<br />

holds. These relations include <strong>the</strong> sum formula and duality formula.<br />

We denote by Zk (k > 1) <strong>the</strong> Q vector space spanned by MZVs whose<br />

weight is k, and define Z0 = Q, Z1 = 0, Z = ∞ k=0 Zk. It is conjectured<br />

by Zagier and o<strong>the</strong>rs that Z = ∞ k=0 Zk and dim Zk = dk where <strong>the</strong> sequence<br />

{dk} is defined by <strong>the</strong> recurring formula d0 = d2 = 1, d1 = 0, dk =<br />

dk−2 + dk−3. Goncharov([G1]) and Ter<strong>as</strong>oma([T2]) showed dim Zk ≤ dk in<br />

<strong>the</strong> context <strong>of</strong> mixed Tate motives.<br />

1.1.2 Multiple polylogarithms <strong>of</strong> one variable<br />

We define <strong>the</strong> multiple polylogarithms <strong>of</strong> one variable (MPLs, for short) <strong>as</strong><br />

Lik1,...,kr(z) =<br />

<br />

n1>···>nr>0<br />

z n1<br />

n k1<br />

1 · · · n kr<br />

r<br />

. (1.5)<br />

These series converge absolutely on |z| < 1. Moreover, if k1 > 1, <strong>the</strong>y<br />

converge also at z = 1 and <strong>the</strong> limit values become MZVs,<br />

lim<br />

z→1−0 Lik1,...,kr(z) = ζ(k1, . . . , kr). (1.6)


1.1 MVZs,MPLs 9<br />

Multiple polylogarithms have <strong>the</strong> iterated integral expressions<br />

z<br />

Lik1,...,kr(z) =<br />

0<br />

where <strong>the</strong> iterated integral<br />

by<br />

z<br />

z0<br />

dz dz dz dz dz dz<br />

◦ · · · ◦ ◦ · · · ◦ ◦ · · · ◦ ,<br />

z z<br />

1 − z z z<br />

1 − z<br />

(1.7)<br />

k1−1 times<br />

kr−1 times<br />

z<br />

ω1 ◦ · · · ◦ ωr =<br />

z0<br />

ω1 ◦ · · · ◦ ωr, (ωi’s are 1-forms <strong>of</strong> dz) is defined<br />

z<br />

z0<br />

ω1(z ′ z ′<br />

)<br />

z0<br />

ω2 ◦ · · · ◦ ωr (r > 1)<br />

inductively. By using <strong>the</strong> iterated integral expression, multiple polylogarithms<br />

Lik1,...,kr(z) can be continued <strong>as</strong> many-valued analytic function on<br />

P 1 − {0, 1, ∞} (for detail, see Section 2.2).<br />

We can also obtain <strong>the</strong> iterated integral expression <strong>of</strong> MZVs:<br />

ζ(k1, . . . , kr) =<br />

1<br />

0<br />

dz dz dz dz dz dz<br />

◦ · · · ◦ ◦ · · · ◦ ◦ · · · ◦ ,<br />

z z<br />

1 − z z z<br />

1 − z<br />

(1.8)<br />

k1−1 times<br />

kr−1 times<br />

where k1 ≥ 2 and <strong>the</strong> duality formula <strong>of</strong> MZVs can be viewed <strong>as</strong> <strong>the</strong> transformation<br />

<strong>of</strong> <strong>the</strong> variable t = 1 − z by <strong>the</strong> iterated integral expression.<br />

1.1.3 The free shuffle algebra h generated by letters<br />

x, y<br />

Let A be a set <strong>of</strong> letters and C〈A〉 a non-commutative polynomial algebra<br />

<strong>of</strong> letters A over C. We denote by 1 <strong>the</strong> unit <strong>of</strong> C〈A〉 (empty word) and ◦<br />

<strong>the</strong> product <strong>of</strong> C〈A〉 by concatenation (we will omit it if <strong>the</strong>re is no worry<br />

<strong>of</strong> confusing). We define <strong>the</strong> shuffle product x on C〈A〉 inductively <strong>as</strong><br />

w x 1 = 1 x w = w,<br />

(a1 ◦ w1) x (a2 ◦ w2) = a1 ◦ (w1 x (a2 ◦ w2)) + a2 ◦ ((a1 ◦ w1) x w2)<br />

where w, w1, w2 are words <strong>of</strong> C〈A〉 and a1, a2 are letters in A. Then S(A) =<br />

(C〈A〉, x, 1) becomes an <strong>as</strong>sociative and commutative algebra due to Reutenauer([Re]).<br />

This is referred to <strong>as</strong> <strong>the</strong> free shuffle algebra generated by A. The free shuffle<br />

algebra S(A) h<strong>as</strong> a grading S(A) = ∞ s=0 Ss(A) with respect to <strong>the</strong> length<br />

<strong>of</strong> words and h<strong>as</strong> a Hopf algebra structure with respect to <strong>the</strong> coproduct ¯ ∆,<br />

<strong>the</strong> counit ¯ε and <strong>the</strong> antipode ¯ S defined <strong>as</strong>


10 Chapter. 1<br />

¯∆(a1 ◦ · · · ◦ ar) =<br />

r<br />

(a1 ◦ · · · ◦ ai) ⊗ (ai+1 ◦ · · · ◦ ar), (1.9)<br />

i=0<br />

¯ε(a1 ◦ · · · ◦ ar) = 0, (1.10)<br />

¯S(a1 ◦ · · · ◦ ar) = (−1) r ar ◦ · · · ◦ a1<br />

(1.11)<br />

for a1, . . . , ar ∈ A, where we <strong>as</strong>sume that a1 · · · a0 = ar+1 · · · ar = 0. For<br />

details <strong>of</strong> Hopf algebr<strong>as</strong>, see [K].<br />

For A = {x, y}, we denote by h = S({x, y}) <strong>the</strong> free shuffle algebra<br />

generated by letters x, y and h0 = C1 + hy (resp. h10 = C1 + xhy) <strong>the</strong><br />

subspace <strong>of</strong> h spanned by elements ended with y (resp. started with x and<br />

ended with y). Clearly h0 and h10 are both x-subalgebra <strong>of</strong> h. Fur<strong>the</strong>rmore h<br />

is regarded <strong>as</strong> a polynomial algebra <strong>of</strong> x over h0 (resp. a polynomial algebra<br />

<strong>of</strong> x, y over h10 ) under x multiplication, namely<br />

∞<br />

h = h 0 x x n (= h 0 [x]) (1.12)<br />

=<br />

n=0<br />

∞<br />

m,n=0<br />

h 10 x x m x y n (= h 10 [x, y]), (1.13)<br />

where an stands for a x · · · x a<br />

= n!a<br />

n times<br />

n . By this isomorphism, we define <strong>the</strong><br />

regularization map regi : h → hi (i = 0, 10) <strong>as</strong> follows;<br />

reg 0 (w) = <strong>the</strong> constant term <strong>of</strong> w in <strong>the</strong> decomposition (1.12), (1.14)<br />

reg 10 (w) = <strong>the</strong> constant term <strong>of</strong> w in <strong>the</strong> decomposition (1.13). (1.15)<br />

The regularization map reg0 satisfies <strong>the</strong> following properties([I<strong>KZ</strong>]);<br />

wx n n<br />

= reg 0 (wx n−j ) x x j<br />

for w ∈ h 0 , (1.16)<br />

j=0<br />

reg 0 (wyx n ) = (−1) n (w x x n )y for n ≥ 0, w ∈ h. (1.17)<br />

Under <strong>the</strong>se notations, we identify <strong>the</strong> letters x and y <strong>as</strong> 1-forms dz<br />

z and<br />

dz , and <strong>the</strong> iterated integral<br />

1−z<br />

z<br />

0<br />

<strong>as</strong> a linear map from h 0 to <strong>the</strong> algebra <strong>of</strong><br />

analytic functions (we define 1 = 1 <strong>the</strong> constant function). From properties<br />

<strong>of</strong> iterated integrals, <strong>the</strong> map<br />

z<br />

0<br />

z<br />

0<br />

w1 x w2 =<br />

is a x-homomorphism, namely<br />

z<br />

0<br />

w1<br />

z<br />

0<br />

w2


1.1 MVZs,MPLs 11<br />

for all words w1, w2 ∈ h 0 . Thus multiple polylogarithms Li can be regarded<br />

<strong>as</strong> a homomorphism from h 0 to <strong>the</strong> algebra <strong>of</strong> analytic functions on P 1 −<br />

{0, 1, ∞}. We denote by<br />

Li(x k1−1 y · · · x kr−1 y; z) = Lik1,...,kr(z) (1.18)<br />

via <strong>the</strong> iterated integral. We also express ζ(k1, . . . , kr) = ζ(x k1−1 y · · · x kr−1 y)<br />

by MZVs and regard ζ <strong>as</strong> a homomorphism from h 10 to R.<br />

We extend <strong>the</strong> domain <strong>of</strong> Li(•; z) from h 0 to h via Li(x; z) = log z and<br />

x-homomorphism. That is, by virtue <strong>of</strong> (1.16),<br />

Li(wx n ; z) =<br />

n<br />

j=0<br />

Li(reg 0 (wx n−j ); z) logj z<br />

j!<br />

(1.19)<br />

for w ∈ h 0 . These extended MPLs are also many-valued analytic functions on<br />

P 1 − {0, 1, ∞} and satisfy <strong>the</strong> following differential recursive relations([Ok]);<br />

d Li(xw; z)<br />

dz<br />

d Li(yw; z)<br />

dz<br />

= 1<br />

Li(w; z),<br />

z<br />

(1.20)<br />

= 1<br />

Li(w; z).<br />

1 − z<br />

(1.21)<br />

1.1.4 The double shuffle relation <strong>of</strong> MZVs<br />

The iterated integral formulation <strong>of</strong> MPLs induce <strong>the</strong> (integral) shuffle product<br />

<strong>of</strong> MZVs;<br />

ζ(w1 x w2) = ζ(w1) x ζ(w2) (1.22)<br />

for any words w1, w2 ∈ h 10 . Fur<strong>the</strong>rmore MZVs have ano<strong>the</strong>r product-sum<br />

relation referred to <strong>as</strong> <strong>the</strong> harmonic product (or series shuffle product) defined<br />

by <strong>the</strong> following due to H<strong>of</strong>fman([Ho]).<br />

We denote by χk = x k−1 y and w = χk1 · · · χkr <strong>the</strong> word <strong>of</strong> h 0 . The<br />

harmonic product ∗ <strong>of</strong> h 0 is defined by<br />

w ∗ 1 = 1 ∗ w = w, (1.23)<br />

(χk1w1) ∗ (χk2w2) = χk1(w1 ∗ (χk2w2)) + χk2((χk1w1) ∗ w2)<br />

+ (χk1+k2)(w1 ∗ w2) (1.24)<br />

inductively for any words w, w1, w2 ∈ h 0 . Then (h 0 , ∗, 1) becomes an <strong>as</strong>sociative<br />

and commutative algebra and <strong>the</strong> relation


12 Chapter. 1<br />

ζ(w1 ∗ w2) = ζ(w1)ζ(w2) (1.25)<br />

holds. This is a generalization <strong>of</strong> <strong>the</strong> relations via <strong>the</strong> series expression<br />

ζ(k1)ζ(l1) = <br />

=<br />

1<br />

n n1>0<br />

k1<br />

1<br />

<br />

<br />

n1>m1>0<br />

<br />

m1>0<br />

1<br />

m l1<br />

1<br />

+ <br />

n1=m1>0<br />

+ <br />

m1>n1>0<br />

<br />

= ζ(k1, l1) + ζ(k1 + l1) + ζ(l1, k1).<br />

1<br />

n k1<br />

1 m l1<br />

1<br />

We provide an interpretation <strong>of</strong> <strong>the</strong> harmonic product <strong>as</strong> iterated integrals<br />

in Section 3.3.4.<br />

Combining <strong>the</strong> shuffle and harmonic product <strong>of</strong> MZVs, one can obtain<br />

<strong>the</strong> non-trivial Q-linear relations among MZVs referred to <strong>as</strong> <strong>the</strong> regularized<br />

double shuffle relation([I<strong>KZ</strong>]);<br />

ζ(reg 10 (w1 x w2 − w1 ∗ w2)) = 0 (1.26)<br />

for w1 ∈ h 0 , w2 ∈ h 10 . Ihara-Kaneko-Zagier([I<strong>KZ</strong>]) and Racinet([R]) conjectured<br />

that <strong>the</strong> regularized double shuffle relations contain <strong>the</strong> all Q-linear<br />

relations <strong>of</strong> MZVs.<br />

1.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

In this section, we introduce <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong>. First we discuss <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on <strong>the</strong> moduli space M0,n. Next, we define <strong>the</strong> <strong>formal</strong> <strong>KZ</strong><br />

<strong>equation</strong> <strong>of</strong> one variable <strong>as</strong> <strong>the</strong> c<strong>as</strong>e <strong>of</strong> M0,4. Since <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

<strong>of</strong> one variable plays essential roles in Chapter 2, we review <strong>the</strong> algebraic<br />

properties <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable carefully according to<br />

[OkU]. We note that <strong>the</strong> c<strong>as</strong>e <strong>of</strong> M0,5 (<strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two<br />

variables) appears in Section 3.3.<br />

1.2.1 The moduli space M0,n and <strong>the</strong> cubic coordinate<br />

Let (P 1 ) n ∗ be a configuration space <strong>of</strong> n points on P 1<br />

(P 1 ) n ∗ = {(x1, . . . , xn) ∈ (P 1 ) n | xi = xj (i = j)}. (1.27)<br />

We denote by M0,n = PGL(2, C)\(P 1 ) n ∗ <strong>the</strong> moduli space <strong>of</strong> (P 1 ) n ∗ over <strong>the</strong><br />

actions


1.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> 13<br />

xi ↦→ axi<br />

<br />

a<br />

<strong>of</strong><br />

c<br />

<br />

b<br />

d<br />

∈ PGL(2, C).<br />

+ b<br />

(1.28)<br />

cxi + d<br />

We introduce <strong>the</strong> simplicial coordinate system<br />

(y1, . . . , yn−3) <strong>of</strong> M0,n via<br />

yi = xi − xn−2<br />

xi − xn<br />

xn−1 − xn<br />

xn−1 − xn−2<br />

and <strong>the</strong> cubic coordinate system (z1, . . . , zn−3) via<br />

z1 = y1, zi = yi<br />

yi−1<br />

(1 ≤ i ≤ n − 3) (1.29)<br />

(2 ≤ i ≤ n − 3) (1.30)<br />

due to Brown([B]). Since <strong>the</strong> cross ratio yi is an invariant under <strong>the</strong> action<br />

<strong>of</strong> linear fractional transformation, <strong>the</strong> simplicial and <strong>the</strong> cubic coordinate<br />

define <strong>the</strong> coordinate system <strong>of</strong> M0,n. Roughly speaking, <strong>the</strong> simplicial coordinate<br />

means fixing <strong>the</strong> three points yn−2 = 0, yn−1 = 1 and yn = ∞, and<br />

<strong>the</strong> cubic coordinate means blowing up on <strong>the</strong> origin to be a normal crossing<br />

point <strong>of</strong> <strong>the</strong> divisors.<br />

1.2.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on M0,n<br />

We consider <strong>the</strong> Pfaffian system on (P 1 ) n ∗<br />

dG = ΩG, Ω = <br />

1≤i


14 Chapter. 1<br />

due to Arnold([A]). By using <strong>the</strong> IPBR and <strong>the</strong> AR, one can show that <strong>the</strong><br />

<strong>equation</strong> (1.31) is integrable and h<strong>as</strong> PGL(2, C) invariance ([B], [OU]). Thus<br />

it is an integrable system on M0,n.<br />

Since <strong>the</strong> IPBR is homogeneous, X h<strong>as</strong> a grading with respect to <strong>the</strong><br />

degree <strong>of</strong> Lie polynomials and <strong>the</strong> universal enveloping algebra U(X) =<br />

⊕ ∞ s=0 Us(X) h<strong>as</strong> also a grading with respect to <strong>the</strong> length <strong>of</strong> words. We denote<br />

by I <strong>the</strong> unit <strong>of</strong> U(X) and U(X) <strong>the</strong> completion <strong>of</strong> U(X) with respect to this<br />

grading. U(X) (resp. U(X)) h<strong>as</strong> a Hopf algebra structure (resp. topological<br />

Hopf algebra structure) by <strong>the</strong> coproduct ∆, <strong>the</strong> counit ε and <strong>the</strong> antipode<br />

S defined <strong>as</strong><br />

for A ∈ X.<br />

∆(A) = 1 ⊗ A + A ⊗ 1 algebra morphism, (1.34)<br />

ε(A) = 0, algebra morphism (1.35)<br />

S(A) = −A anti-algebra morphism (1.36)<br />

1.2.3 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable<br />

For n = 4, <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> (1.31) is represented via <strong>the</strong> cubic coordinate<br />

system <strong>as</strong><br />

dG = ΩG, Ω = ζ1Z1 + ζ11Z11, (1.37)<br />

where z = z1, Z1 = X12, Z11 = −X13, ζ1 = dz<br />

z and ζ11 = dz . We call this<br />

1−z<br />

<strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable (<strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong>, for short).<br />

The IPBR (1.32) for {Z1, Z11} is trivial, thus <strong>the</strong> infinitesimal pure braid<br />

Lie algebra X is a free Lie algebra generated by Z1, Z11. The universal enveloping<br />

algebra U(X) is a ring <strong>of</strong> non-commutative polynomials <strong>of</strong> <strong>the</strong> variables<br />

Z1, Z11 namely U(X) = C〈Z1, Z11〉, and U(X) = C〈〈Z1, Z11〉〉 is an<br />

algebra <strong>of</strong> <strong>the</strong> non-commutative <strong>formal</strong> power series.<br />

We denote by B <strong>the</strong> dual Hopf algebra <strong>of</strong> U(X). This is a free shuffle<br />

algebra generated by ζ1, ζ11. We remark that <strong>the</strong> AR (1.33) for {ζ1, ζ11} is<br />

only ζ1 ∧ ζ11 = 0 and B is a 0-th cohomology <strong>of</strong> <strong>the</strong> reduced bar complex <strong>of</strong><br />

<strong>the</strong> exterior algebra generated by ζ1, ζ11 (see also Section 3.3.2).<br />

Through <strong>the</strong> identification ζ1 = x, ζ11 = y, B is nothing but <strong>the</strong> free<br />

shuffle algebra h introduced in Section 1.1.3. In what follows, we use <strong>the</strong><br />

notation h, x = dz<br />

z<br />

and y = dz<br />

1−z instead <strong>of</strong> B, ζ1 and ζ11. We also denote by<br />

X = Z1, Y = Z11 and H = U(X). In this notation, we can express <strong>the</strong> <strong>formal</strong><br />

<strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable (1.37) <strong>as</strong> a H-valued Fuchsian <strong>equation</strong><br />

dG<br />

dz =<br />

<br />

X Y<br />

+ G. (1.38)<br />

z 1 − z


1.2 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> 15<br />

We note that <strong>the</strong> Hopf algebra structure <strong>of</strong> H is given by (1.34),(1.35)<br />

and (1.36) and <strong>the</strong> Hopf algebra structure <strong>of</strong> h is given by (1.9),(1.10) and<br />

(1.11).<br />

1.2.4 The normalized fundamental solution <strong>of</strong> <strong>the</strong> <strong>formal</strong><br />

<strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable and Drinfel’d<br />

<strong>as</strong>sociator<br />

We consider a solution L(z) to <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> (1.38) which have a<br />

decomposition<br />

L(z) = ˆ L(z)z X<br />

(1.39)<br />

where ˆ L(z) is holomorphic at z = 0 and ˆ L(0) = I. This is referred to <strong>as</strong><br />

<strong>the</strong> fundamental solution to (1.38) normalized at z = 0. Since <strong>the</strong> <strong>equation</strong><br />

(1.38) is Fuchsian, if <strong>the</strong>re exists, <strong>the</strong> normalized fundamental solution is <strong>the</strong><br />

unique solution <strong>of</strong> (1.38) which h<strong>as</strong> a <strong>as</strong>ymptotic behavior L(z)z −X → I <strong>as</strong><br />

z tends to 0.<br />

According to Okuda([Ok]), we define H-valued analytic function H0(z) <strong>as</strong><br />

H0(z) = <br />

w:word <strong>of</strong> h<br />

Li(w; z)W, (1.40)<br />

where W is <strong>the</strong> word <strong>of</strong> H determined <strong>as</strong> replacing x, y in w by X, Y . (1.20)<br />

and (1.21) imply that H0(z) satisfies <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> (1.38). Fur<strong>the</strong>rmore,<br />

by virtue <strong>of</strong> (1.16), one can calculate <strong>as</strong><br />

H0(z) = <br />

=<br />

=<br />

=<br />

=<br />

w:word <strong>of</strong> h<br />

<br />

Li(w; z)W =<br />

∞<br />

s<br />

w:word <strong>of</strong> h0 s=0 j=0<br />

<br />

∞<br />

∞<br />

w:word <strong>of</strong> h0 i=0 j=0<br />

<br />

w:word <strong>of</strong> h<br />

<br />

w:word <strong>of</strong> h<br />

<br />

s∈Z≥0<br />

w:word <strong>of</strong> h 0<br />

Li(wx s ; z)W X s<br />

Li(reg 0 (wx s−j ); z) logj (z)<br />

j! W Xs−j X j<br />

Li(reg 0 (wx i ); z) logj (z)<br />

j! W Xi X j<br />

Li(reg 0 (w); z)W<br />

Li(reg 0 (w); z)W<br />

∞<br />

<br />

j=0<br />

z X .<br />

log j (z)<br />

j! Xj


16 Chapter. 1<br />

Since Li(reg 0 (w); z) is regular at z = 0 for all word w ∈ h and<br />

Li(reg 0 (w); 0) =<br />

<br />

1 (w = 1)<br />

0 (w = 1)<br />

(1.41)<br />

holds, w:word <strong>of</strong> h Li(reg0 <br />

(w); z)W is regular at z = 0 and ˆ L(0) = I.<br />

Therefore H0(z) is <strong>the</strong> unique fundamental solution to (1.38) normalized<br />

at z = 0. The inverse <strong>of</strong> H0(z) is given <strong>as</strong><br />

(H0(z)) −1 = <br />

w:word <strong>of</strong> h<br />

Li( ¯ S(w); z)W (1.42)<br />

by using <strong>the</strong> antipode ¯ S <strong>of</strong> h. Indeed one can show e<strong>as</strong>ily that<br />

<br />

w1:word <strong>of</strong> h<br />

=<br />

<br />

w1,w2:word <strong>of</strong> h<br />

= <br />

w:word <strong>of</strong> h (w)<br />

= <br />

= I,<br />

w:word <strong>of</strong> h (w)<br />

Li(w1; z)W1<br />

<br />

w2:word <strong>of</strong> h<br />

Li(w1 x ¯ S(w2); z)W1W2<br />

<br />

Li(w ′ x ¯ S(w ′′ ); z)W<br />

<br />

Li(¯ε(w)1; z)W<br />

Li( ¯ S(w2); z)W2<br />

where <strong>the</strong> sum in <strong>the</strong> third expression stands for <strong>the</strong> Sweedler’s notation<br />

¯∆(a) = <br />

(a) a′ ⊗ a ′′ .<br />

We remark that <strong>the</strong> fundamental solution normalized at z = 0 can be<br />

constructed not to use <strong>the</strong> regularization map reg 0 <strong>as</strong> follows (for detail, see<br />

[OU]). We define ˆ Ls(z) recursively <strong>as</strong><br />

ˆL0(z) = I, (1.43)<br />

z <br />

ˆLs+1(z)<br />

1<br />

=<br />

z [X, ˆ Ls(z)] + 1<br />

1 − z Y ˆ <br />

Ls(z) dz. (1.44)<br />

0<br />

Thus we can show 1<br />

z [X, ˆ Ls(z)] is holomorphic at z = 0 for all s by induction<br />

on s, <strong>the</strong>n ˆ Ls(z) is well-defined for all s and ˆ L(z) = ∞<br />

s=0 ˆ Ls(z) is <strong>the</strong><br />

fundamental solution normalized at z = 0. In term <strong>of</strong> iterated integral,


1.2 The <strong>formal</strong> <strong>KZ</strong> eqaution 17<br />

it is expressed <strong>as</strong><br />

ˆL(z) = <br />

z<br />

(k1,...,kr)<br />

0<br />

x k1−1 kr−1<br />

◦ y · · · ◦ x ◦ y<br />

× ad(X) k1−1 kr−1<br />

µ(Y ) · · · ad(X) µ(Y )(I) (1.45)<br />

= <br />

(k1,...,kr)<br />

Li(x k1−1 ◦ y · · · ◦ x kr−1 ◦ y; z)<br />

× ad(X) k1−1 µ(Y ) · · · ad(X) kr−1 µ(Y )(I), (1.46)<br />

where (k1, . . . , kr) runs over all indexes <strong>of</strong> positive integers and ad (resp.<br />

µ) stands for an adjoint ad(A)(W ) = [A, W ] (resp. a left multiplication<br />

µ(A)(W ) = AW ) for A ∈ {X, Y }, W ∈ H.<br />

We also consider <strong>the</strong> fundamental solution normalized at z = 1, which is<br />

<strong>the</strong> unique solution with an expression<br />

L (1) (z) = ˆ L (1) (z)(1 − z) Y , (1.47)<br />

where ˆ L (1) (z) is holomorphic at z = 1 and ˆ L (1) (1) = I. One can show <strong>the</strong><br />

function H1(z) below is <strong>the</strong> fundamental solution normalized at z = 1:<br />

H1(z) = <br />

w:word <strong>of</strong> h<br />

Li(σ(w); 1 − z)W, (1.48)<br />

where σ is an ◦-involution <strong>of</strong> h defined by σ(x) = −y, σ(y) = −x.<br />

The connection coefficient between H0(z) and H1(z) is referred to <strong>as</strong><br />

Drinfel’d <strong>as</strong>sociator and denoted by Φ<strong>KZ</strong>(X, Y ) = (H1(z)) −1 H0(z). Drinfel’d<br />

<strong>as</strong>sociator Φ<strong>KZ</strong>(X, Y ) is expressed <strong>as</strong> a generating function <strong>of</strong> MZVs;<br />

Φ<strong>KZ</strong>(X, Y ) = <br />

w:word <strong>of</strong> h<br />

due to Le-Murakami([LM]) and Okuda-Ueno([OkU]).<br />

ζ(reg 10 (w)) (1.49)<br />

1.3 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

In Section 1.3.1, we briefly review <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and <strong>Gauss</strong><br />

<strong>hypergeometric</strong> functions. Details for <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> can<br />

be found on [Ha], [Ki], [WW] and o<strong>the</strong>r texts. Next in Section 1.3.2, we<br />

introduce some existing researches about <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and<br />

MZVs.


18 Chapter. 1<br />

1.3.1 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and its solution<br />

<strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> is a second order Fuchsian differential <strong>equation</strong><br />

expressed <strong>as</strong><br />

z(1 − z) d2f df<br />

+ (γ − (α + β + 1)z) − αβf = 0, (1.50)<br />

dz2 dz<br />

where α, β, γ are complex parameters. The <strong>equation</strong> h<strong>as</strong> three regular singular<br />

points 0, 1, ∞ on P 1 .<br />

In what follows, we <strong>as</strong>sume that <strong>the</strong> parameter α, β, γ, α−β and γ −α−β<br />

are not integers. Under this <strong>as</strong>sumption, <strong>the</strong> linearly independent solutions<br />

ϕ (i)<br />

0 (z), ϕ (i)<br />

1 (z) (i = 0, 1, ∞) on <strong>the</strong> neighborhoods <strong>of</strong> z = 0, 1, ∞ are given<br />

<strong>as</strong> follows;<br />

ϕ (0)<br />

0 (z) = F (α, β, γ; z) =<br />

∞<br />

n=0<br />

(α)n(β)n<br />

(γ)nn! zn , (1.51)<br />

ϕ (0)<br />

1 (z) = z 1−γ F (α + 1 − γ, β + 1 − γ, 2 − γ; z), (1.52)<br />

ϕ (1)<br />

0 (z) = F (α, β, α + β + 1 − γ; 1 − z), (1.53)<br />

ϕ (1)<br />

1 (z) = (1 − z) γ−α−β F (γ − α, γ − β, γ − α − β + 1; 1 − z), (1.54)<br />

ϕ (∞)<br />

0 (z) = z −α F (α, α + 1 − γ, α − β + 1; 1/z), (1.55)<br />

ϕ (∞)<br />

1 (z) = z −β F (β, β + 1 − γ, β − α + 1; 1/z), (1.56)<br />

where we define <strong>the</strong> branch <strong>of</strong> <strong>the</strong>se complex power by <strong>the</strong> principal values<br />

and (α)n stands for <strong>the</strong> Pochhammer symbol (α)n = (α + n − 1)(α + n −<br />

2) · · · (α + 1)α. The function F (α, β, γ; z) is <strong>Gauss</strong> <strong>hypergeometric</strong> function.<br />

It h<strong>as</strong> Euler’s integral expression<br />

F (α, β, γ; z) =<br />

Γ(γ)<br />

Γ(α)Γ(γ − α)<br />

1<br />

t<br />

0<br />

α−1 (1 − t) γ−α−1 (1 − zt) −β dt (1.57)<br />

and is continued analytically to P1 − {0, 1, ∞} <strong>as</strong> a many-valued function.<br />

We also regard to all ϕ (i)<br />

j ’s (i = 0, 1, ∞, j = 0, 1) <strong>as</strong> many-valued functions<br />

on P1 − {0, 1, ∞}.<br />

The connection matrices <strong>of</strong> <strong>the</strong>se solutions are given by <strong>the</strong> following


1.3 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> 19<br />

formula via <strong>the</strong> Euler’s integral expression <strong>of</strong> <strong>hypergeometric</strong> function;<br />

(ϕ (0)<br />

0 , ϕ (0)<br />

1 ) = (ϕ (1)<br />

0 , ϕ (1)<br />

1 )C 01 ,<br />

C 01 ⎛<br />

Γ(γ)Γ(γ − α − β)<br />

⎜<br />

=<br />

Γ(γ − α)Γ(γ − β)<br />

⎝<br />

Γ(γ)Γ(α + β − γ)<br />

Γ(α)Γ(β)<br />

(ϕ (0)<br />

0 , ϕ (0)<br />

1 ) = (ϕ (∞)<br />

0 , ϕ (∞)<br />

1 )C 0∞ ,<br />

C 0∞ ⎛<br />

e<br />

⎜<br />

= ⎝<br />

−πiα Γ(γ)Γ(β − α)<br />

Γ(β)Γ(γ − α)<br />

e−πiβ Γ(γ)Γ(α − β)<br />

Γ(α)Γ(γ − β)<br />

Γ(2 − γ)Γ(γ − α − β)<br />

⎞<br />

Γ(1 − α)Γ(1 − β)<br />

Γ(2 − γ)Γ(α + β − γ)<br />

Γ(α + 1 − γ)Γ(β + 1 − γ)<br />

⎟<br />

⎠ , (1.58)<br />

eπi(γ−α−1) Γ(2 − γ)Γ(β − α)<br />

Γ(β + 1 − γ)Γ(1 − α)<br />

eπi(γ−β−1) ⎞<br />

⎟<br />

Γ(2 − γ)Γ(α − β)<br />

⎠ .<br />

Γ(α + 1 − γ)Γ(1 − β)<br />

(1.59)<br />

1.3.2 Existing researches about <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> and MZVs<br />

The first result for <strong>the</strong> relation between <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and<br />

MZVs is due to Ohno-Zagier([OZ]). For index k = (k1, . . . , kr), k1 > 1, we<br />

call <strong>the</strong> number #{i|ki > 1} <strong>the</strong> height <strong>of</strong> k and denote by G(k, n, s; z) <strong>the</strong><br />

sum <strong>of</strong> MPLs which have fixed weight k, depth n and height s. Ohno-Zagier<br />

showed that <strong>the</strong> generating function<br />

Φ(z) =<br />

<br />

n≥s≥1,k≥s+n<br />

G(k, n, s; z)u k−n−s v n−s w s−1<br />

satisfies <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and can express <strong>as</strong><br />

Φ(z) =<br />

(1.60)<br />

1<br />

(1 − F (α − u, β − u, 1 − u, z)), (1.61)<br />

uv − w<br />

where α + β = u + v, αβ = w.<br />

Taking <strong>the</strong> limit <strong>of</strong> (1.61) <strong>as</strong> z tends to 1 by using<br />

Γ(γ)Γ(γ − α − β)<br />

lim F (α, β, γ; z) =<br />

z→1 Γ(γ − α)Γ(γ − β)<br />

(1.62)<br />

and <strong>the</strong> expansion <strong>of</strong> Gamma function by Riemann zeta values (see Section<br />

2.6.2), <strong>the</strong>y showed <strong>the</strong> relations known <strong>as</strong> Ohno-Zagier relation<br />

Φ(1) =<br />

1<br />

(1 − exp(<br />

uv − w<br />

∞<br />

n=2<br />

ζ(n)<br />

n Sn(u, v, w))), (1.63)


20 Chapter. 1<br />

where Sn(u, v, w) = u n +v n −α n −β n , α+β = u+v, αβ = w. They obtained<br />

also many relations <strong>of</strong> MZVs by specializing <strong>the</strong> Ohno-Zagier relation,for<br />

instance <strong>the</strong> sum formula, ζ(2, . . . , 2)<br />

=<br />

<br />

l times<br />

π2s and Le-Murakami formula.<br />

(2s+1)!<br />

Ter<strong>as</strong>oma([T1]) said that <strong>the</strong> values <strong>of</strong> Selberg integral, which contains<br />

<strong>the</strong> special values <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> function at z = 1, can be written<br />

<strong>as</strong> a linear combination <strong>of</strong> MZVs. Ohno-Zagier’s result gives an example <strong>of</strong><br />

Ter<strong>as</strong>oma’s <strong>as</strong>sertion. In chapter 2, we show again <strong>the</strong> expansion <strong>of</strong> Φ(z) and<br />

interpret <strong>as</strong> an iterated integral expression <strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>.<br />

Moreover we also interpret Ohno-Zagier’s result <strong>as</strong> a connection problem <strong>of</strong><br />

<strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> and extend to functional relations <strong>of</strong> MPLs.<br />

The second result is <strong>the</strong> formula<br />

ζ(3, 1, . . . , 3, 1)<br />

=<br />

<br />

2n times<br />

2π4n<br />

. (1.64)<br />

(4n + 2)!<br />

This formula is conjectured by Zagier and proved by Borwein-Bradley-Broadhurst-<br />

Lisoněk ([BBBL]). They show that <strong>the</strong> generating function <strong>of</strong> Li3,1,...,3,1(z) is<br />

expressed <strong>as</strong><br />

∞<br />

Li3,1,...,3,1<br />

<br />

n=0 2n times<br />

(z)t 4n t −t t −t<br />

= F ( , , 1; z)F ( , , 1; z). (1.65)<br />

1 + i 1 + i 1 − i 1 − i<br />

Indeed one can show that <strong>the</strong> both sides <strong>of</strong> <strong>the</strong> <strong>equation</strong> above are eliminated<br />

by <strong>the</strong> action <strong>of</strong><br />

<br />

(1 − z) d<br />

2 <br />

z<br />

dz<br />

d<br />

2 − t<br />

dz<br />

4 . (1.66)<br />

Taking <strong>the</strong> limit <strong>of</strong> (1.65) <strong>as</strong> z tends to 1 and using<br />

we obtain (1.64).<br />

lim F (a, −a, 1, z) =<br />

z→1<br />

1<br />

Γ(1 − a)Γ(1 + a)<br />

sin πa<br />

= , (1.67)<br />

πa


Chapter 2<br />

<strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

<strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

This chapter is <strong>the</strong> main part <strong>of</strong> this <strong>the</strong>sis. We discuss <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong>, according<br />

to [O]. We will obtain two main results. The first one is <strong>the</strong> expansion <strong>of</strong><br />

<strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> to a series <strong>of</strong> parameters whose coefficients<br />

are multiple polylogarithms on <strong>the</strong> universal covering space <strong>of</strong> P 1 −{0, 1, ∞}.<br />

The second result is system <strong>of</strong> functional relations <strong>of</strong> MPLs, which restore<br />

<strong>the</strong> Ohno-Zagier relations <strong>of</strong> MZVs <strong>as</strong> <strong>the</strong> limit <strong>as</strong> z tends to 1. These relations<br />

are followed from <strong>the</strong> connection formula between <strong>the</strong> regular solutions<br />

<strong>of</strong> <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> at z = 0 and z = 1. Fur<strong>the</strong>rmore, we will<br />

lead various relations <strong>of</strong> MPLs with respect to <strong>the</strong> connection formula <strong>of</strong> <strong>the</strong><br />

solutions between z = 0 and z = ∞.<br />

2.1 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong><br />

Let X = C{X, Y } be a free Lie algebra <strong>of</strong> letters X, Y and ρ : X → M(n, C)<br />

be a <strong>representation</strong> <strong>of</strong> X. We denote a corresponding <strong>representation</strong> <strong>of</strong> U(X)<br />

by <strong>the</strong> same symbol ρ. We call <strong>the</strong> M(n, C)-valued differential <strong>equation</strong><br />

dG<br />

dz =<br />

<br />

ρ(X) ρ(Y )<br />

+ G (2.1)<br />

z 1 − z<br />

21


22 Chapter. 2<br />

a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> by ρ and <strong>the</strong> <strong>formal</strong> sum<br />

ρ(H0(z)) = <br />

Li(w; z)ρ(W ) (2.2)<br />

w<br />

a <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution H0(z) <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong>.<br />

In general, a <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution does not converge,<br />

but if <strong>the</strong> <strong>formal</strong> sum (2.2) converges absolutely and uniformly, it<br />

gives <strong>the</strong> fundamental solution to <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong><br />

(2.1). Fur<strong>the</strong>rmore a <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution can be<br />

viewed <strong>as</strong> a solution expressed by <strong>the</strong> iterated integral.<br />

2.2 Analytic properties <strong>of</strong> MPLs<br />

In preparation for proving convergence <strong>of</strong> <strong>representation</strong>s <strong>of</strong> <strong>the</strong> fundamental<br />

solution, we discuss an analytic continuation <strong>of</strong> <strong>the</strong> multiple polylogarithms<br />

<strong>of</strong> one variable to <strong>the</strong> universal covering space <strong>of</strong> P 1 − {0, 1, ∞}.<br />

Proposition 1. Let U be <strong>the</strong> universal covering space <strong>of</strong> P 1 − {0, 1, ∞} and<br />

K be a compact subset <strong>of</strong> U. There exists a constant MK depending only on<br />

K such that, for any word w ∈ h,<br />

| Li(w; z)| < MK ∀z ∈ K. (2.3)<br />

A b<strong>as</strong>ic idea to prove this proposition is due to Lappo-Danilevsky on [L]<br />

p.159-163.<br />

Let π : U → P1 − {0, 1, ∞} be <strong>the</strong> canonical projection and <strong>the</strong> real<br />

interval I be a simply-connected subset <strong>of</strong> U such that arg(z) = arg(1−z) = 0<br />

on I. That is, on z ∈ I, Li(w; z) h<strong>as</strong> an expansion (1.5) for all word w ∈ h0 and log z h<strong>as</strong> an expansion log z = − ∞ (1−z)<br />

n=1<br />

n<br />

. We note that, on z ∈ I,<br />

n<br />

Li(w; z) converges to 0 <strong>as</strong> z tends to 0 for any word w = 1 and log z converges<br />

to 0 <strong>as</strong> z tends to 1.<br />

Let z, p be points <strong>of</strong> U and Cz p be a path on U from p to z. For any word<br />

w ∈ h, we define MPLs with prescribing an initial point and a path by an<br />

inductive way such <strong>as</strong><br />

Lip,Cz (xw; z) =<br />

p<br />

Lip,Cz (yw; z) =<br />

p<br />

<br />

<br />

C z p<br />

C z p<br />

Lip,C z p<br />

Lip,C z p<br />

(w; z)<br />

dz, (2.4)<br />

z<br />

(w; z)<br />

dz, (2.5)<br />

1 − z<br />

Lip,Cz (1; z) = 1. (2.6)<br />

p<br />

These MPLs satisfy <strong>the</strong> following properties.


2.2 Analytic properties <strong>of</strong> MPLs 23<br />

Lemma 2. ([L])<br />

(i) For any word w ∈ h and points p, z ∈ U, <strong>the</strong> value <strong>of</strong> Lip,Cz (w; z) does<br />

p<br />

not depend on choice <strong>of</strong> a path Cz p on U (in o<strong>the</strong>r words, <strong>the</strong> value <strong>of</strong><br />

Lip,Cz (w; z) depends only a homotopy cl<strong>as</strong>s <strong>of</strong> <strong>the</strong> integral contour on<br />

p<br />

P1 − {0, 1, ∞}).<br />

(ii) Let δ be <strong>the</strong> distance between π(C p z ) and {0, 1};<br />

δ = dist(π(C p z ), {0, 1}) = inf<br />

z1∈π(C p z )<br />

z2∈{0,1}<br />

and σ be <strong>the</strong> length <strong>of</strong> <strong>the</strong> path π(C z p). Then we have<br />

| Lip,Cz 1<br />

(w; z)| < p |w|!<br />

|z1 − z2|, (2.7)<br />

<br />

σ<br />

|w|<br />

, (2.8)<br />

δ<br />

where |w| stands for <strong>the</strong> length <strong>of</strong> <strong>the</strong> word w (that is, <strong>the</strong> number <strong>of</strong><br />

letters in w).<br />

The next lemma, which is also due to [L], follows from <strong>the</strong> coproduct<br />

structure <strong>of</strong> h (1.9) <strong>as</strong> a Hopf algebra.<br />

Lemma 3. Let w = a1a2 · · · ar be a word <strong>of</strong> h 0 , where each ai denotes <strong>the</strong><br />

letter x or y (i = 1, . . . , r, ar = y), and C z p be a path from p to z on U.<br />

Choosing a point q on Cz p, we divide <strong>the</strong> path Cz p <strong>as</strong> Cz p = Cz q ◦ Cq p. Then<br />

(w; z) satisfies<br />

Lip,C z p<br />

Lip,Cz (w; z) =<br />

p<br />

r<br />

i=0<br />

Liq,C z q (a1 · · · ai; z) Li p,C q p (ai+1 · · · ar; q). (2.9)<br />

Here we use a convention such <strong>as</strong> Liq,C z q (a1 · · · ai; z) = 1 for i = 0 and<br />

Li p,C q p (ai+1 · · · ar; q) = 1 for i = r.<br />

Clearly, for a word w ∈ h 0 and p ∈ I, <strong>the</strong> usual MPL Li(w; z) defined in<br />

Section 1.1.2 is characterized <strong>as</strong><br />

Li(w; z) = lim<br />

ε→0<br />

ε∈I<br />

Liε,C z p ◦[ε,p](w; z), (2.10)<br />

where [ε, p] stands for <strong>the</strong> path from ε to p on <strong>the</strong> interval I.<br />

One can prove <strong>the</strong> following lemma e<strong>as</strong>ily.<br />

Lemma 4. For any z ∈ (0, 1<br />

2 ] ⊂ R and any word w ∈ h0 , we have<br />

| Li(w; z)| ≤ 1. (2.11)


24 Chapter. 2<br />

Pro<strong>of</strong>. We prove this lemma by induction on <strong>the</strong> length <strong>of</strong> <strong>the</strong> word w. If<br />

|w| = 0, Li(w; z) = Li(1; z) = 1 clearly satisfies <strong>the</strong> claim. If |w| = 1, we<br />

have | Li(y; z)| = | − log(1 − z)| ≤ | log( 1<br />

1<br />

)| ≤ 1 on z ∈ (0, 2 2 ].<br />

We <strong>as</strong>sume | Li(w; z)| ≤ 1 for a word w = xk1−1 kr−1 0 y · · · x y ∈ h . Thus<br />

we obtain<br />

<br />

<br />

<br />

| Li(xw; z)| = <br />

<br />

≤<br />

≤<br />

and<br />

<br />

<br />

| Li(yw; z)| = <br />

<br />

on z ∈ (0, 1<br />

2 ].<br />

≤<br />

<br />

n1>···>nr>1<br />

<br />

n1>···>nr>1<br />

<br />

n1>···>nr>1<br />

z<br />

0<br />

z<br />

0<br />

z n1<br />

n k1+1<br />

1 n k2<br />

2 · · · nkr r<br />

r<br />

<br />

<br />

<br />

<br />

<br />

|z| n1<br />

n k1+1<br />

1 n k2<br />

2 · · · nkr ≤<br />

<br />

1 n1<br />

2<br />

n k1<br />

1 n k2<br />

2 · · · n kr<br />

r<br />

<br />

n1>···>nr>1<br />

= Li(w; 1<br />

) ≤ 1<br />

2<br />

<br />

dz <br />

Li(w; z) <br />

1 − z ≤<br />

z<br />

dz<br />

|Li(w; z)|<br />

0 1 − z<br />

dz<br />

= − log(1 − z) ≤ 1<br />

1 − z<br />

Using <strong>the</strong>se lemm<strong>as</strong>, we now prove <strong>the</strong> proposition.<br />

n k1+1<br />

1<br />

<br />

1 n1<br />

2<br />

n k2<br />

2 · · · n kr<br />

r<br />

Pro<strong>of</strong> <strong>of</strong> Proposition 1. Let p(K) and l(K) be constants depending only on<br />

K such <strong>as</strong><br />

<br />

p(K) = min dist(π(K), {0, 1}), 1<br />

<br />

, (2.12)<br />

2<br />

⎛<br />

⎞<br />

l(K) = max<br />

z∈K<br />

⎝ inf<br />

C : path from p(K) to z on U<br />

dist(π(C),{0,1})≥p(K)<br />

length(π(C)) ⎠ . (2.13)<br />

In <strong>the</strong> right hand side <strong>of</strong> (2.13), we identify <strong>the</strong> number p(K) ∈ (0, 1]<br />

with a<br />

2<br />

point in I ⊂ U.<br />

By virtue <strong>of</strong> Lemma 3 and (2.10), for any word w = a1a2 · · · ar ∈ h0 , we<br />

obtain<br />

Li(w; z) =<br />

r<br />

i=0<br />

Lip(K),C z p(K) (a1 · · · ai; z) Li(ai+1 · · · ar; p(K)). (2.14)


2.2 Analytic properties <strong>of</strong> MPLs 25<br />

Consequently, by Lemma 2 and Lemma 4, we have<br />

| Li(w; z)| ≤<br />

<<br />

r<br />

i=0<br />

r<br />

i=0<br />

| Lip(K),C z p(K) (a1 · · · ai; z)|| Li(ai+1 · · · ar; p(K))| (2.15)<br />

1<br />

i!<br />

i l(K)<br />

< exp<br />

p(K)<br />

Fur<strong>the</strong>rmore, for any word w ∈ h 0 ,<br />

| Li(wx s ; z)| ≤<br />

s<br />

j=0<br />

< exp<br />

<br />

l(K)<br />

.<br />

p(K)<br />

| Li(reg 0 (wx s−j | log z|j<br />

); z)|<br />

<br />

l(K)<br />

exp<br />

p(K)<br />

j!<br />

max | log z|<br />

z∈K<br />

<br />

l(K)<br />

Thus <strong>the</strong> estimation (2.3) holds for MK = exp<br />

p(K)<br />

<br />

.<br />

<br />

+ max | log z| .<br />

z∈K<br />

(2.16)<br />

2.3 One dimensional <strong>representation</strong>s <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong><br />

Let ρ be a one dimensional <strong>representation</strong> <strong>of</strong> X and we denote by ρ(X) =<br />

α, ρ(Y ) = β ∈ C. The <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> reads<br />

dG<br />

dz =<br />

<br />

α β<br />

+ G. (2.17)<br />

z 1 − z<br />

Clearly <strong>the</strong> <strong>equation</strong> h<strong>as</strong> <strong>the</strong> solution z α (1 − z) −β on U uniquely up to scalar<br />

multiplication.<br />

On <strong>the</strong> o<strong>the</strong>r hand, we try to calculate <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental<br />

solution <strong>as</strong> follows:<br />

ρ(H0(z)) = <br />

Li(w; z)ρ(W )<br />

=<br />

=<br />

w<br />

∞<br />

<br />

p,q=0 w: word <strong>of</strong> h<br />

w consists p x’s and q y’s<br />

∞<br />

Li(x p x y q ; z)α p β q<br />

p,q=0<br />

Li(w; z)α p β q


26 Chapter. 2<br />

=<br />

=<br />

where we use x p = xp<br />

∞ log<br />

p,q=0<br />

p (z)<br />

Li 1,...,1 (z)α<br />

p! <br />

q times<br />

p β q<br />

<br />

∞<br />

log p (z)<br />

p! αp<br />

<br />

∞<br />

(− log(1 − z)) q<br />

β<br />

q!<br />

q<br />

<br />

p=0<br />

= z α (1 − z) −β ,<br />

<br />

q times<br />

p! and Li 1,...,1<br />

q=0<br />

(z) = (− log(1 − z)) q . The <strong>representation</strong><br />

ρ(H0(z)) clearly converge for |α|, |β| < 1 by virtue <strong>of</strong> Proposition 1. Then a<br />

one dimensional <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> is trivial.<br />

2.4 <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> <strong>as</strong> a <strong>representation</strong><br />

<strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong><br />

Next, we consider two dimensional <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong>.<br />

Let α, β and γ be complex parameters and ρ0 : X → M(2, C) be a<br />

<strong>representation</strong> defined by<br />

ρ0(X) =<br />

<br />

0 β<br />

0 0<br />

, ρ0(Y ) =<br />

. (2.18)<br />

0 1 − γ<br />

α α + β + 1 − γ<br />

Then <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> by ρ0 is <strong>the</strong> <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> (1.50). Indeed one can rewrite <strong>the</strong> <strong>equation</strong> (1.50) to<br />

<strong>the</strong> system;<br />

<br />

d v1 1 0 β<br />

=<br />

+<br />

dz v2 z 0 1 − γ<br />

1<br />

<br />

0 0<br />

v1<br />

, (2.19)<br />

1 − z α α + β + 1 − γ v2<br />

where v1 = f and v2 = 1 df<br />

z . By Section 1.3, we can write <strong>the</strong> fundamental<br />

β dz<br />

solution matrix <strong>of</strong> (2.19) on a neighborhood <strong>of</strong> z = i (i = 0, 1, ∞) <strong>as</strong><br />

<br />

<br />

where ϕ (i)<br />

j<br />

Φi =<br />

1<br />

β<br />

ϕ (i)<br />

0<br />

d z dz ϕ(i) 0<br />

1<br />

β<br />

ϕ (i)<br />

1<br />

d z dz ϕ(i) 1<br />

, (2.20)<br />

are linearly independent solutions defined at (1.51) ∼ (1.56). The<br />

connection formul<strong>as</strong> <strong>of</strong> each fundamental solution matrices are given by<br />

Φ −1<br />

1 Φ0 = C 01 , (2.21)<br />

Φ −1<br />

∞ Φ0 = C 0∞ , (2.22)


2.4 <strong>Gauss</strong> <strong>hypergeometric</strong> eq. <strong>as</strong> rep. 27<br />

where C 01 , C 0∞ are connection coefficients defined by (1.58) and (1.59).<br />

It is well known fact that any second order Fuchsian <strong>equation</strong> which h<strong>as</strong><br />

three regular singular points on P 1 can be transformed to <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong>.<br />

2.5 The expression <strong>of</strong> <strong>the</strong> <strong>hypergeometric</strong> function<br />

by MPLs<br />

In this section, we give a concrete expression <strong>of</strong> ρ0(H0(z)) <strong>the</strong> <strong>representation</strong><br />

<strong>of</strong> <strong>the</strong> fundamental solution to <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> and expand <strong>Gauss</strong><br />

<strong>hypergeometric</strong> function F (α, β, γ; z) <strong>as</strong> a series <strong>of</strong> <strong>the</strong> multiple polylogarithms<br />

<strong>of</strong> one variable.<br />

2.5.1 Definitions and Notations<br />

We now redefine <strong>the</strong> weight, depth and height <strong>of</strong> indexes by using <strong>the</strong> term<br />

<strong>of</strong> <strong>the</strong> shuffle algebra h. For a word w in h, we define <strong>the</strong> weight |w|, <strong>the</strong><br />

depth d(w) and <strong>the</strong> height h(w) <strong>of</strong> w <strong>as</strong> follows;<br />

|w| := <strong>the</strong> number <strong>of</strong> letters in w, (2.23)<br />

d(w) := <strong>the</strong> number <strong>of</strong> y which appears in w, (2.24)<br />

h(w) := (<strong>the</strong> number <strong>of</strong> yx which appears in w) + 1. (2.25)<br />

Denote by gi(k, n, s) (i = 0, 10) <strong>the</strong> sum <strong>of</strong> all words in h i with fixed<br />

weight k, depth n and height s, namely<br />

gi(k, n, s) =<br />

<br />

w∈h i<br />

|w|=k, d(w)=n<br />

h(w)=s<br />

w. (2.26)<br />

Set g10(k, n, s; z) = 0 if k < n + s, n < s or k, n, s ≤ 0, and g0(k, n, s; z) = 0<br />

if k < n + s − 1, n < s or k, n, s ≤ 0. We note that, if a word w =<br />

x k1−1 y · · · x kr−1 y belongs to h 10 , we have h(w) = #{i|ki ≥ 2}. This is <strong>the</strong><br />

original definition <strong>of</strong> <strong>the</strong> height given by [OZ] appeared in Section 1.3.2.<br />

Hence one obtains <strong>the</strong> expression<br />

g10(k, n, s) =<br />

<br />

k1+···+kn=k<br />

k1≥2, k2,...,kn≥1<br />

#{i|ki≥2}=s<br />

x k1−1 yx k2−1 y · · · x kn−1 y. (2.27)


28 Chapter. 2<br />

We denote by Gi(k, n, s; z) <strong>the</strong> sum <strong>of</strong> MPLs <strong>as</strong>sociated to gi(k, n, s);<br />

Gi(k, n, s; z) := Li(gi(k, n, s); z) =<br />

<br />

Li(w; z). (2.28)<br />

Especially, we have <strong>the</strong> following formula;<br />

G10(k, n, s; z) =<br />

<br />

k1+···+kn=k<br />

k1≥2, k2,...,kn≥1<br />

#{i|ki≥2}=s<br />

w∈h i<br />

|w|=k, d(w)=n<br />

h(w)=s<br />

Lik1,...,kn(z). (2.29)<br />

Now <strong>the</strong> <strong>equation</strong> (1.20) implies <strong>the</strong> following differential relation;<br />

z d<br />

dz G10(k, n, s; z) = G0(k − 1, n, s; z). (2.30)<br />

For any word W in H, we also define <strong>the</strong> weight |W |, <strong>the</strong> depth d(W ) and<br />

<strong>the</strong> height h(W ) <strong>of</strong> W in a similar f<strong>as</strong>hion;<br />

|W | := <strong>the</strong> number <strong>of</strong> letters in W,<br />

d(W ) := <strong>the</strong> number <strong>of</strong> Y which appears in W ,<br />

h(W ) := (<strong>the</strong> number <strong>of</strong> Y X which appears in W ) + 1.<br />

2.5.2 Main Result<br />

Theorem 5. Assume that |1 − γ|, |α + 1 − γ|, |β + 1 − γ| and<br />

|α + β + 1 − γ| < 1.<br />

Then we have <strong>the</strong> expression <strong>of</strong> F (α, β, γ; z), <strong>the</strong><br />

2<br />

<strong>hypergeometric</strong> function, <strong>as</strong> follows;<br />

F (α, β, γ; z) = 1 + αβ <br />

G10(k, n, s; z)<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

× (1 − γ) k−n−s (α + β + 1 − γ) n−s ((α + 1 − γ)(β + 1 − γ)) s−1 . (2.31)<br />

The series in <strong>the</strong> right hand side converges absolutely and uniformly on any<br />

compact subset K <strong>of</strong> <strong>the</strong> universal covering space U <strong>of</strong> P 1 − {0, 1, ∞}.<br />

This expression is equal to Ohno-Zagier’s result (1.61) in Section 1.3.2<br />

through u = 1 − γ, v = α + β + 1 − γ, w = (α + 1 − γ)(β + 1 − γ) in |z| < 1<br />

and |α|, |β|, |1 − γ| are sufficiently small. Thus this <strong>the</strong>orem can be regarded<br />

<strong>as</strong> an expansion <strong>of</strong> Ohno-Zagier’s result to <strong>the</strong> universal covering space <strong>of</strong><br />

P 1 − {0, 1, ∞} and an interpretation <strong>as</strong> an iterated integral expression <strong>of</strong><br />

<strong>Gauss</strong> <strong>hypergeometric</strong> function.


2.5 The expression <strong>of</strong> <strong>hypergeometric</strong> function 29<br />

We prove this <strong>the</strong>orem in <strong>the</strong> following. Put p = 1 − γ, q = α + β + 1 − γ<br />

for convenience.<br />

2.5.3 The image <strong>of</strong> word in H by <strong>the</strong> <strong>representation</strong> ρ0<br />

We consider <strong>the</strong> <strong>representation</strong> ρ0 : X → M(2, C) given by (2.18)<br />

ρ0(X) =<br />

0 β<br />

0 p<br />

<br />

, ρ0(Y ) =<br />

0 0<br />

α q<br />

<br />

. (2.32)<br />

As previously mentioned, <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong><br />

(1.38) by ρ0<br />

<br />

d ρ0(X)<br />

G =<br />

dz z + ρ0(Y<br />

<br />

)<br />

G<br />

1 − z<br />

(2.33)<br />

is <strong>the</strong> <strong>hypergeometric</strong> <strong>equation</strong> (1.50) and <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental<br />

solution H0(z)<br />

ρ0(H0(z)) = <br />

Li(w; z)ρ0(W ) (2.34)<br />

w<br />

gives , if it converges absolutely, <strong>the</strong> fundamental solution matrix which h<strong>as</strong><br />

<strong>the</strong> <strong>as</strong>ymptotic property ρ0(H0)z −ρ0(X) → I (z → 0). In order to compute<br />

<strong>the</strong> series <strong>of</strong> <strong>the</strong> right hand side <strong>of</strong> (2.34) and to show its convergence, we<br />

prepare <strong>the</strong> following lemma.<br />

Lemma 6. For any non-empty word W ∈ H, <strong>the</strong> <strong>representation</strong> ρ0(W ) is<br />

given by<br />

where<br />

ρ0(W ) = p |W |−d(W )−h(W ) q d(W )−h(W ) (αβ + pq) h(W )−1 M, (2.35)<br />

⎧<br />

<br />

αβ βq<br />

αp pq<br />

<br />

0 0<br />

⎪⎨<br />

αp pq<br />

M = <br />

0 βq<br />

0 pq<br />

<br />

0 0<br />

⎪⎩<br />

0 pq<br />

(if W ∈ XHY ),<br />

(if W ∈ Y HY or W = Y ),<br />

(if W ∈ XHXor W = X),<br />

(if W ∈ Y HX).<br />

(2.36)


30 Chapter. 2<br />

Pro<strong>of</strong>. This is proved by straightforward computation <strong>as</strong> follows.<br />

First, by e<strong>as</strong>ily induction, we have<br />

ρ0(X n ) = p n−1 ρ0(X),<br />

ρ0(Y n ) = q n−1 ρ0(Y ),<br />

ρ0((XY ) n ) = (αβ + pq) n−1<br />

For W = X a1 Y b1 · · · X <strong>as</strong> Y bs (ai, bi ≥ 1), we have<br />

<br />

αβ βq<br />

.<br />

αp pq<br />

ρ0(W ) = (p a1−1 q b1−1 ρ0(XY )) · · · (p <strong>as</strong>−1 q bs−1 ρ0(XY ))<br />

= p P s<br />

i=1 (ai−1) q P s<br />

i=1 (bi−1) ρ0(XY ) s<br />

= p P s<br />

i=1 (ai−1) q P s<br />

i=1 (bi−1) (αβ + pq) s−1<br />

<br />

αβ βq<br />

.<br />

αp pq<br />

Under <strong>the</strong> rewriting W = X a1 Y b1 · · · X <strong>as</strong> Y bs = X k1−1 Y · · · X kr−1 Y (k1 ≥<br />

2, ki ≥ 1), s = #{i|ki > 1}, s i=1 (bi − 1) = #{i|ki = 1}, s i=1 (ai <br />

− 1) =<br />

ki=1 (ki − 2) holds. Thus we obtain<br />

ρ0(X k1−1 Y · · · X kr−1 Y )<br />

= p ¯P (ki−2) q #{i|ki=1} (αβ + pq) #{i|ki>1}−1<br />

for k1 > 1, k2 . . . , kr > 0, where ¯ (ki − 2) stands for<br />

Now we have also<br />

ρ0(X k1−1 Y · · · X kr−1 Y )<br />

¯<br />

(ki − 2) := <br />

(ki − 2).<br />

ki=1<br />

= p ¯P (ki−2) q #{i|ki=1}−1 (αβ + pq) #{i|ki>1}<br />

ρ0(X k1−1 Y · · · X kr−1 Y X s )<br />

⎧<br />

⎪⎨<br />

=<br />

⎪⎩<br />

p ¯P (ki−2)+s−1 q #{i|ki=1} (αβ + pq) #{i|ki>1}<br />

<br />

0 0<br />

α q<br />

p ¯P (ki−2)+s−1 q #{i|ki=1}−1 (αβ + pq) #{i|ki>1}+1<br />

<br />

0 β<br />

0 p<br />

<br />

0 0<br />

0 1<br />

<br />

αβ βq<br />

,<br />

αp pq<br />

(k1 = 1),<br />

(s > 0, k1 > 1),<br />

(s > 0, k1 = 1).


2.5 The expression <strong>of</strong> <strong>hypergeometric</strong> function 31<br />

On <strong>the</strong> o<strong>the</strong>r hand, by definition <strong>of</strong> height, we have<br />

⎧<br />

#{i|ki > 1} (k1 > 1, s = 0),<br />

⎪⎨<br />

#{i|ki > 1} + 1 (k1 = 1, s = 0),<br />

h(X k1−1 Y · · · X kr−1 Y X s ) =<br />

Thus <strong>the</strong>se results yield <strong>the</strong> lemma.<br />

#{i|ki ⎪⎩<br />

> 1} + 1 (k1 > 1, s > 0),<br />

#{i|ki > 1} + 2 (k1 = 1, s > 0).<br />

By putting δ = max(|αβ|, |αp|, |βq|, |pq|, 1), we obtain <strong>the</strong> following corollaries.<br />

Corollary 7. For any word W ∈ H, <strong>the</strong>re exists a constant δ depending only<br />

on α, β, p, q such that<br />

||ρ0(W )|| ≤ δ|p| |W |−d(W )−h(W ) |q| d(W )−h(W ) |αβ + pq| h(W )−1 , (2.37)<br />

where ||A|| denotes <strong>the</strong> maximal norm <strong>of</strong> a matrix A = (aij)1≤i,j≤2, namely<br />

||A|| = maxi,j |aij|.<br />

Corollary 8. If |1 − γ|, |α + 1 − γ|, |β + 1 − γ|, |α + β + 1 − γ| < 1,<br />

<strong>the</strong> 2<br />

<strong>representation</strong> ρ0(H0(z)) converges absolutely and uniformly on any compact<br />

subset K <strong>of</strong> <strong>the</strong> universal covering space U <strong>of</strong> P1 − {0, 1, ∞}.<br />

Pro<strong>of</strong>. Let K be a compact subset <strong>of</strong> U. By using Proposition 1 and <strong>the</strong> fact<br />

that <strong>the</strong> number <strong>of</strong> words with weight k in h is 2k , we can show that <strong>the</strong>re<br />

exists a constant MK depending only on K such that<br />

<br />

Li(w; z) k<br />

< 2 MK ∀z ∈ K. (2.38)<br />

w∈h: word, |w|=k<br />

d(w)=n, h(w)=s<br />

By Corollary 7 we have<br />

||ρ0(H0(z))|| = || <br />

Li(w; z)ρ0(W )|| (2.39)<br />

≤ 1 + δ <br />

w<br />

<br />

k,n,s w<br />

|w|=k, d(w)=n<br />

h(w)=s<br />

<br />

≤ 1 + δMK<br />

< 1 + 4δMK<br />

k,n,s<br />

<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

<br />

Li(w; z) |p| |W |−d(W )−h(W ) |q| d(W )−h(W ) |αβ + pq| h(W )−1<br />

2 k |p| k−n−s |q| n−s |αβ + pq| s−1<br />

(2|1 − γ|) k−n−s (2|α + β + 1 − γ|) n−s<br />

× (2|α + 1 − γ|)(2|β + 1 − γ|) s−1 .


32 Chapter. 2<br />

Hence <strong>the</strong> series ρ0(H0(z)) converges absolutely and uniformly on K if<br />

|1 − γ|, |α + 1 − γ|, |β + 1 − γ|, |α + β + 1 − γ| < 1<br />

2 .<br />

2.5.4 Asymptotic properties <strong>of</strong> ρ0(H0) and Φ0<br />

From <strong>the</strong> discussion above, it follows that <strong>the</strong> <strong>representation</strong> ρ0(H0(z)) is <strong>the</strong><br />

fundamental solution matrix <strong>of</strong> <strong>the</strong> <strong>hypergeometric</strong> <strong>equation</strong> on U. Hence<br />

<strong>the</strong>re exists a linear relation between ρ0(H0(z)) and Φ0, which is also a fundamental<br />

solution matrix in a neighborhood <strong>of</strong> z = 0 defined by (2.20), <strong>as</strong><br />

follows.<br />

Lemma 9.<br />

<br />

1 1<br />

Φ0(z) = ρ0(H0(z)) . (2.40)<br />

0<br />

Pro<strong>of</strong>. Let D be a domain in C defined by D = {z ∈ C | |z| < 1} − {ℜz ≤<br />

0, ℑz = 0} and specify branches <strong>of</strong> all MPLs, which appears in ρ0(H0(z)), by<br />

<strong>the</strong> expansion (1.5) and a branch <strong>of</strong> log z by <strong>the</strong> principal value on D (that<br />

is, D is a domain in U which includes <strong>the</strong> interval I and π(D) is simplyconnected).<br />

It is enough to prove that both sides <strong>of</strong> (2.40) have <strong>the</strong> same<br />

<strong>as</strong>ymptotic property <strong>as</strong> z tends to 0 in D.<br />

By definition, ρ0(H0(z))z −ρ0(X) → I <strong>as</strong> z → 0 holds. On <strong>the</strong> o<strong>the</strong>r hand,<br />

since <strong>the</strong> formula<br />

holds, we have<br />

<br />

1 1<br />

Φ0(z)<br />

0<br />

−1<br />

p z<br />

β<br />

−ρ0(X)<br />

=<br />

z −ρ0(X)<br />

<br />

1<br />

=<br />

<br />

F00(z)<br />

1<br />

β zF ′ 00(z)<br />

p<br />

β<br />

β<br />

p (z−p − 1)<br />

0 z −p<br />

<br />

β<br />

p (F01(z) − F00(z))<br />

1<br />

p (pF01(z) + zF ′ 01(z) − zF ′ 00(z))<br />

(2.41)<br />

<strong>as</strong> z → 0, where F00 and F01 are regular functions on |z| < 1 introduced<br />

through ϕ (0)<br />

0 = F00, ϕ (0)<br />

1 = z1−γ −1<br />

1 1<br />

F01. Hence Φ0(z) p and ρ0(H0(z))<br />

0 β<br />

are <strong>the</strong> same fundamental solution matrix to <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>.<br />

2.5.5 Pro<strong>of</strong> <strong>of</strong> Theorem 5<br />

Now we prove Theorem 5 by using <strong>the</strong> preceding lemm<strong>as</strong>.<br />

<br />

→ I


2.5 The expression <strong>of</strong> <strong>hypergeometric</strong> function 33<br />

Pro<strong>of</strong> <strong>of</strong> Theorem 5. From Corollary 8, if |1 − γ|, |α + 1 − γ|, |β + 1 − γ|, |α +<br />

β + 1 − γ| < 1<br />

2 , any matrix element <strong>of</strong> ρ0(H0(z)) converges absolutely and<br />

uniformly on any compact subset K <strong>of</strong> U. Hence we see that ϕ (0)<br />

<br />

0 (z) =<br />

1 1<br />

ρ(H0(z)) p satisfies <strong>the</strong> same property, here Aij stands for <strong>the</strong><br />

0 β 11<br />

(i, j) element <strong>of</strong> a matrix A. This is computed <strong>as</strong> follows;<br />

ϕ (0)<br />

0 (z) =<br />

<br />

I + <br />

<br />

Li(w; z)ρ0(W )<br />

w=1<br />

= 1 + <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

= 1 + αβ <br />

<br />

|w|=k,d(w)=n<br />

h(w)=s<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

= 1 + αβ <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

11<br />

Li(w; z)p k−n−s q n−s (αβ + pq) s−1 αβ<br />

G10(k, n, s; z)p k−n−s q n−s (αβ + pq) s−1<br />

G10(k, n, s; z)(1 − γ) k−n−s (α + β + 1 − γ) n−s<br />

× ((α + 1 − γ)(β + 1 − γ)) s−1 . (2.42)<br />

In <strong>the</strong> discussion above, we have <strong>as</strong>sumed that 1 − γ, α, β = 0. However,<br />

<strong>the</strong> formula (2.42) makes sense even if 1 − γ → 0, α → 0 and β → 0.<br />

Corollary 10. The solution ϕ (0)<br />

1 (z) to <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>, which<br />

h<strong>as</strong> <strong>the</strong> exponent 1 − γ at z = 0, is given <strong>as</strong> follows;<br />

ϕ (0)<br />

1 (z) = z 1−γ<br />

<br />

1 + (α + 1 − γ)(β + 1 − γ) <br />

k,n,s<br />

G10(k, n, s; z)<br />

× (γ − 1) k−n−s (α + β + 1 − γ) n−s (αβ) s−1<br />

<br />

. (2.43)<br />

Remark. This corollary is proved immediately by using <strong>of</strong> Theorem 5 and<br />

<strong>the</strong> formula ϕ (0)<br />

1 (z) = z1−γF (α+1−γ, β+1−γ, 2−γ; z). However one can also<br />

prove this <strong>as</strong> an algebraic way to compute ϕ (0)<br />

<br />

1 1<br />

1 (z) = ρ0(H0(z))<br />

.<br />

0<br />

p<br />

β<br />

12


34 Chapter. 2<br />

2.6 The connection formula between <strong>the</strong> regular<br />

solutions to <strong>Gauss</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> at z = 0 and z = 1<br />

In this section, we investigate various functional relations <strong>of</strong> <strong>the</strong> multiple<br />

polylogarithms <strong>of</strong> one variable by considering <strong>the</strong> (1, 1)-element <strong>of</strong> <strong>the</strong> connection<br />

formula (2.21), and obtain <strong>the</strong> relations <strong>of</strong> <strong>the</strong> multiple zeta values<br />

<strong>as</strong> known <strong>as</strong> Ohno-Zagier relation([OZ]) by taking <strong>the</strong> limit <strong>as</strong> z → 1.<br />

For <strong>the</strong> purpose, we first express <strong>the</strong> inverse element <strong>of</strong> Φ1 (2.20) <strong>as</strong> a series<br />

<strong>of</strong> MPLs, and expand <strong>the</strong> gamma functions which appear in <strong>the</strong> connection<br />

matrix C 01 (2.21) <strong>as</strong> a series <strong>of</strong> Riemann zeta values.<br />

2.6.1 The inverse <strong>of</strong> <strong>the</strong> fundamental solution matrix<br />

on <strong>the</strong> neighborhood <strong>of</strong> z = 1<br />

Let G be a solution to <strong>the</strong> <strong>hypergeometric</strong> <strong>equation</strong> (2.33). Thus <strong>the</strong> trans-<br />

posed inverse matrix t G −1 satisfies<br />

d t −1<br />

G =<br />

dt<br />

tρ0(Y )<br />

t<br />

+<br />

tρ0(X) tG−1 , (2.44)<br />

1 − t<br />

where t = 1 − z. So we define a <strong>representation</strong> ρ1 : X → M(2, C) such <strong>as</strong><br />

ρ1(X) = t <br />

0<br />

ρ0(Y ) =<br />

0<br />

<br />

α<br />

, ρ1(Y ) =<br />

q<br />

t <br />

0<br />

ρ0(X) =<br />

β<br />

<br />

0<br />

.<br />

p<br />

(2.45)<br />

The matrix-valued function t Φ −1<br />

1<br />

<strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> (1.38) by ρ1<br />

d<br />

G =<br />

dt<br />

is a fundamental solution matrix <strong>of</strong> <strong>the</strong><br />

<br />

ρ1(X)<br />

t + ρ1(Y )<br />

1 − t<br />

<br />

G. (2.46)<br />

On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> <strong>representation</strong> ρ1(H0(t)) is also a fundamental solution<br />

matrix <strong>of</strong> (2.46). It is nothing but ρ0(H0(z)) up to changing <strong>the</strong> variable<br />

z → t and <strong>the</strong> parameters (α, β, p, q) → (β, α, q, p). Similarly <strong>as</strong> Lemma 9,<br />

we obtain<br />

Lemma 11. There exists a linear relation<br />

<br />

t −1<br />

1<br />

Φ1 = ρ1(H0(t))<br />

0<br />

αβ<br />

(α+β−γ)q<br />

β<br />

α+β−γ<br />

<br />

. (2.47)


2.6 Connection formula between z = 0, 1 35<br />

Pro<strong>of</strong>. It is suffice to show<br />

t Φ −1<br />

1<br />

<br />

1<br />

0<br />

αβ<br />

(α+β−γ)q<br />

β<br />

α+β−γ<br />

−1<br />

t −ρ1(X) → I<br />

<strong>as</strong> t tends to 0. We express solutions to <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong><br />

ϕ (1)<br />

0 , ϕ (1)<br />

1 <strong>as</strong> ϕ (1)<br />

0 = F10(t), ϕ (1)<br />

1 = t 1−q F11, where F10(t), F11(t) are holomorphic<br />

at t = 0. By e<strong>as</strong>y calculation, we have<br />

t −1<br />

Φ1 = 1<br />

1 −<br />

det Φ1<br />

det Φ1 = 1<br />

<br />

αβ<br />

1<br />

0<br />

t −ρ1(X) =<br />

β (1 − t)((1 − q)t−qF11 + t1−qF ′ 1<br />

11)<br />

β (1 − t)F ′ 10<br />

−t1−qF11 F10<br />

β (1 − t)t−q ((1 − q)F10F11 + tF10F ′ 11 − tF ′ 10F11),<br />

−1 <br />

α 1 − q =<br />

0 − 1−q<br />

<br />

,<br />

β<br />

<br />

,<br />

(α+β−γ)q<br />

β<br />

α+β−γ<br />

α 1 q (t−q − 1)<br />

0 t−q where F ′ stands for d F . Thus we obtain<br />

dt<br />

t −1<br />

lim Φ1 t→0<br />

<br />

1<br />

0<br />

αβ<br />

(α+β−γ)q<br />

β<br />

α+β−γ<br />

(1 − t)t<br />

= lim<br />

t→0<br />

−q<br />

det Φ1<br />

−1<br />

t −ρ1(X)<br />

− 1<br />

β ((1 − q)F11 + tF ′ 11)<br />

<br />

,<br />

α<br />

βq ((1 − q)F11 + tF11) − 1−q<br />

α<br />

q tF11 − 1−q<br />

β F10<br />

−tF11<br />

1<br />

= lim<br />

t→0 − 1<br />

<br />

1<br />

α<br />

− (1 − q)F11<br />

β βq<br />

((1 − q)F10F11<br />

β (1 − q)F11 − 1−q<br />

β2 F ′ 10<br />

0 − 1−q<br />

β F10<br />

<br />

= β<br />

<br />

1−q α(1−q) 1−q<br />

− + β βq β<br />

1 − q<br />

2<br />

<br />

αβ<br />

q = I.<br />

0<br />

1−q<br />

β<br />

β 2 F ′ 10<br />

By this lemma, <strong>the</strong> (1, 1) and (2, 1)-elements <strong>of</strong> t Φ −1<br />

1 lead to <strong>the</strong> following<br />

proposition immediately.<br />

Proposition 12. Assume that |1 − γ|, |α + 1 − γ|, |β + 1 − γ| and<br />

|α + β + 1 − γ| < 1.<br />

The (1, 1) and (1, 2)-elements <strong>of</strong> Φ−1<br />

2 1 are expressed <strong>as</strong>


36 Chapter. 2<br />

follows;<br />

(Φ −1<br />

1 )11 = 1 + αβ <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

G10(k, k − n, s; 1 − z) (2.48)<br />

× (1 − γ) k−n−s (α + β + 1 − γ) n−s ((α + 1 − γ)(β + 1 − γ)) s−1 ,<br />

(Φ −1<br />

1 )12 = β <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

G0(k − 1, k − n, s; 1 − z) (2.49)<br />

× (1 − γ) k−n−s (α + β + 1 − γ) n−s ((α + 1 − γ)(β + 1 − γ)) s−1 .<br />

The series <strong>of</strong> <strong>the</strong> right hand sides converge absolutely and uniformly on any<br />

compact subset K <strong>of</strong> U.<br />

Pro<strong>of</strong>. Lemma 11 says that (Φ −1<br />

1 )11 = ρ1(H0)11, (Φ −1<br />

1 )12 = ρ1(H0)12 and<br />

ρ1(H0(t)) is equal to ρ0(H0(z)) up to changing <strong>the</strong> variable z → t and <strong>the</strong><br />

parameters (α, β, p, q) → (β, α, q, p). Thus we obtain<br />

(Φ −1<br />

1 )11 = 1 + αβ <br />

G10(k, n, s; t)q k−n−s p n−s (αβ + pq) s−1<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

= 1 + αβ <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

= 1 + αβ <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

G10(k, k − n, s; t)q n−s p k−n−s (αβ + pq) s−1<br />

G10(k, k − n, s; 1 − z)<br />

× (1 − γ) k−n−s (α + β + 1 − γ) n−s ((α + 1 − γ)(β + 1 − γ)) s−1 .<br />

The second <strong>equation</strong> for (Φ −1<br />

1 )12 is followed by<br />

(Φ −1<br />

1 )12 = ρ1(H0)12 = 1<br />

d<br />

t<br />

α dt ρ1(H0)11 = 1<br />

α<br />

t d<br />

dt (Φ−1<br />

1 )11.<br />

2.6.2 The expansion <strong>of</strong> <strong>the</strong> connection matrix <strong>as</strong> a series<br />

<strong>of</strong> <strong>the</strong> zeta values<br />

To compare both sides <strong>of</strong> <strong>the</strong> connection formula (2.21), we expand <strong>the</strong><br />

<strong>as</strong> a series<br />

(1, 1)-element <strong>of</strong> <strong>the</strong> connection matrix (C 01 )11 = Γ(γ)Γ(γ−α−β)<br />

Γ(γ−α)Γ(γ−β)


2.6 Connection formula between z = 0, 1 37<br />

<strong>of</strong> Riemann zeta values. We set p = 1 − γ, q = α + β + 1 − γ and r =<br />

(α + 1 − γ)(β + 1 − γ).<br />

The following formula for <strong>the</strong> gamma function is well known cl<strong>as</strong>sically<br />

([WW]);<br />

1<br />

Γ(1 − z)<br />

where c := limn→∞( n 1<br />

k=1 k<br />

= exp(−cz −<br />

∞<br />

n=2<br />

− log n) is Euler constant.<br />

ζ(n)<br />

n zn ), (2.50)<br />

For a sequence <strong>of</strong> complex numbers a = (a1, a2, . . .) and a non-negative<br />

integer n, we introduce Schur polynomial Pn(a) through <strong>the</strong> following generating<br />

function;<br />

∞<br />

exp( anz n ) =<br />

n=1<br />

that is ⎧<br />

⎪⎨ P0(a) = 1,<br />

⎪⎩ Pn(a) = <br />

k1+2k2+3k3+···=n<br />

We also define <strong>the</strong> integers N (n)<br />

i,j <strong>as</strong><br />

∞<br />

Pn(a)z n , (2.51)<br />

n=0<br />

a k1<br />

1<br />

k1!<br />

a k2<br />

2<br />

k2!<br />

a k3<br />

3<br />

· · · (n ≥ 1).<br />

k3!<br />

⎧<br />

⎪⎨<br />

N<br />

⎪⎩<br />

(n)<br />

i,j = 0 (i < 0 or j < 0),<br />

N (0)<br />

0,0 = 1 (i = j = 0),<br />

a n + b n = <br />

N (n)<br />

i,j (a + b)i (ab) j<br />

(o<strong>the</strong>rwise).<br />

i,j<br />

(2.52)<br />

(2.53)<br />

Since an + bn is a symmetric polynomial <strong>of</strong> a and b, this definition is welldefined<br />

and we have N (n)<br />

i,j = 0 if i + 2j = n. We denote by Ni,j = N (i+2j)<br />

i,j .<br />

Under <strong>the</strong>se notations, one can show <strong>the</strong> following lemma and proposition.<br />

Lemma 13. In <strong>the</strong> algebra <strong>of</strong> <strong>formal</strong> power series C[[a, b]], we have<br />

<br />

∞<br />

Aia i<br />

<br />

∞<br />

Aib i<br />

<br />

∞<br />

<br />

l<br />

<br />

=<br />

i=0<br />

i=0<br />

i=0<br />

i=0<br />

k,l=0<br />

i=j≥0<br />

i=0<br />

AiA2l+k−iNk,l−i<br />

Pro<strong>of</strong>.<br />

<br />

∞<br />

Aia i<br />

<br />

∞<br />

Aib i<br />

<br />

<br />

= + <br />

+ <br />

<br />

i>j≥0<br />

j>i≥0<br />

= <br />

A 2 l (ab) l + <br />

l≥0<br />

l≥0<br />

n>1<br />

(a + b) k (ab) l . (2.54)<br />

AiAja i b j<br />

AlAl+n(ab) l (a n + b n )


38 Chapter. 2<br />

= A 2 0 + <br />

+ <br />

l≥0<br />

l≥1<br />

= A 2 0 + <br />

A 2 l (ab) l<br />

<br />

i,j<br />

(i,j)=(0,0)<br />

l≥1<br />

A 2 l (ab) l<br />

+ <br />

AlAl+2jN0,j(ab) l+j<br />

l≥0<br />

j>0<br />

+ <br />

l≥0<br />

i>0<br />

= A 2 0 + <br />

l≥1<br />

+ l−1<br />

l≥1<br />

j=0<br />

+ <br />

i>0<br />

l≥0<br />

= A 2 0 + <br />

=<br />

l≥1<br />

+ <br />

i>0<br />

l≥0<br />

∞<br />

<br />

l<br />

i,l=0<br />

j=0<br />

j≥0<br />

A 2 l (ab) l<br />

AlAl+i+2j(ab) l Ni,j(a + b) i (ab) j<br />

AlAl+i+2jNi,j(a + b) i (ab) l+j<br />

AjA2l−jN0,l−j(ab) l<br />

l<br />

j=0<br />

l<br />

j=0<br />

l<br />

j=0<br />

AjA2l+i−jNi,l−j(a + b) i (ab) l<br />

AjA2l−jN0,l−j(ab) l<br />

AjA2l+i−jNi,l−j(a + b) i (ab) l<br />

AjA2l+i−jNi,l−j<br />

<br />

(a + b) i (ab) l<br />

Proposition 14. For <strong>the</strong> sequence ζ = (0, ζ(2) ζ(3) ζ(4)<br />

, , , · · · ) and <strong>the</strong> com-<br />

2 3 4<br />

plex numbers p = 1 − γ, q = α + β + 1 − γ and r = (α + 1 − γ)(β + 1 − γ),<br />

we obtain <strong>the</strong> following expansion;<br />

Γ(γ)Γ(γ − α − β)<br />

Γ(γ − α)Γ(γ − β)<br />

∞<br />

<br />

k l m<br />

<br />

i + j<br />

=<br />

i<br />

k,l,m=0<br />

i=0 j=0 µ=0<br />

Pk−i(ζ)Pl−j(ζ)Pµ(−ζ)Pi+j+2m−µ(−ζ)Ni+j,m−µ<br />

<br />

× p k q l r m . (2.55)


2.6 Connection formula between z = 0, 1 39<br />

Pro<strong>of</strong>.<br />

Γ(γ)Γ(γ − α − β) Γ(1 − p)<br />

=<br />

Γ(γ − α)Γ(γ − β) ec(1−p) Γ(1 − q)<br />

ec(1−q) ec(1−(α+1−γ)) e<br />

Γ(1 − (α + 1 − γ))<br />

c(1−(β+1−γ))<br />

Γ(1 − (β + 1 − γ)<br />

∞ ζ(n)<br />

= exp(<br />

n<br />

n=2<br />

pn ∞ ζ(n)<br />

) exp(<br />

n<br />

n=2<br />

qn )<br />

∞ ζ(n)<br />

× exp(−<br />

n<br />

n=2<br />

(α + 1 − γ)n ∞ ζ(n)<br />

) exp(−<br />

n<br />

n=2<br />

(β + 1 − γ)n )<br />

∞<br />

= Pn(ζ)p n<br />

∞<br />

Pn(ζ)q n<br />

∞<br />

Pn(−ζ)(α + 1 − γ) n<br />

∞<br />

Pn(−ζ)(β + 1 − γ) n .<br />

n=0<br />

n=0<br />

n=0<br />

Here by making use <strong>of</strong> Lemma 13, we have<br />

Γ(γ)Γ(γ − α − β)<br />

Γ(γ − α)Γ(γ − β)<br />

∞<br />

= Pn(ζ)p n<br />

∞<br />

Pn(ζ)q n<br />

n=0<br />

n=0<br />

× <br />

<br />

k + l<br />

k<br />

m<br />

k,l,m<br />

µ=0<br />

= <strong>the</strong> right hand side <strong>of</strong> (2.55).<br />

n=0<br />

Pµ(−ζ)Pk+l+2m−µ(−ζ)Nk+l,m−µ<br />

<br />

p k q l r m<br />

2.6.3 Functional relations obtained from <strong>the</strong> (1, 1)-element<br />

<strong>of</strong> <strong>the</strong> connection formula (2.21)<br />

Theorem 15. For |1 − γ|, |α + β + 1 − γ|, |(α + 1 − γ)(β + 1 − γ)| < 1,<br />

<strong>the</strong> 2<br />

<strong>equation</strong><br />

ϕ (0)<br />

0 (z)ϕ (1)<br />

0 (z) − 1 d<br />

z<br />

β dz ϕ(0) 0 (z) 1<br />

α<br />

(1 − z) d<br />

dz ϕ(1)<br />

0 (z) =<br />

Γ(γ)Γ(γ − α − β)<br />

Γ(γ − α)Γ(γ − β) (2.56)<br />

yields functional relations <strong>of</strong> multiple polylogarithms. Especially by expanding<br />

(2.56) <strong>as</strong> a series <strong>of</strong> (1 − γ), (α + β + 1 − γ) and ((α + 1 − γ)(β + 1 − γ)),


40 Chapter. 2<br />

we have<br />

<br />

<br />

¯G10(k + l + 2m, l + m, m; z) + ¯ G10(k + l + 2m, k + m, m; 1 − z)<br />

k,l,m<br />

+ <br />

k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ =m<br />

= <br />

<br />

k<br />

k,l,m<br />

¯G10(k ′ + l ′ + 2m ′ , l ′ + m ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ ; 1 − z)<br />

+ ¯ G0(k ′ +l ′ +2m ′ −1, l ′ +m ′ , m ′ ; z)G0(k ′′ +l ′′ +2m ′′ +1, k ′′ +m ′′ +1, m ′′ +1; 1 − z)<br />

i=0<br />

l<br />

j=0 µ=0<br />

m<br />

<br />

i + j<br />

i<br />

× (1 − γ) k (α + β + 1 − γ) l ((α + 1 − γ)(β + 1 − γ)) m<br />

Pk−i(ζ)Pl−j(ζ)Pµ(−ζ)Pi+j+2m−µ(−ζ)Ni+j,m−µ<br />

where ¯ Gi(k, n, s; z) = Gi(k, n, s; z) − Gi(k, n, s + 1; z)<br />

(i = 0, 10).<br />

Pro<strong>of</strong>. We try to expand <strong>the</strong> (1, 1)-element <strong>of</strong> (2.21)<br />

× (1 − γ) k (α + β + 1 − γ) l ((α + 1 − γ)(β + 1 − γ)) m ,<br />

(2.57)<br />

(Φ −1<br />

1 )11(Φ0)11 + (Φ −1<br />

1 )12(Φ0)21 = (C 01 )11 (2.58)<br />

<strong>as</strong> series in p, q, r. According to Proposition 14, (C01 )11 is <strong>the</strong> right hand side<br />

<strong>of</strong> (2.57). On <strong>the</strong> o<strong>the</strong>r hand, from Theorem 5 and Proposition 12, we had<br />

already<br />

(Φ0)11 = ϕ (0)<br />

0 (z) = 1 + αβ <br />

G10(k, n, s; z)p k−n−s q n−s r s−1 ,<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

(Φ0)21 = 1 d<br />

z<br />

β dz ϕ(0) 0 (z) = α <br />

G0(k − 1, n, s; z)p<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

k−n−s q n−s r s−1 ,<br />

(Φ −1<br />

1 )11 = ϕ (1)<br />

0 (z) = 1 + αβ <br />

G10(k, k − n, s; 1 − z)p k−n−s q n−s r s−1 ,<br />

(Φ −1<br />

1 )12 = − 1<br />

α<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

d<br />

(1 − z)<br />

dz ϕ(1) 0 (z) = β <br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

G0(k − 1, k − n, s; 1 − z)p k−n−s q n−s r s−1 .


2.6 Connection formula between z = 0, 1 41<br />

By αβ = r − pq, we have<br />

(Φ0)11 = 1 + (r − pq) <br />

= 1 + (r − pq)<br />

= 1 +<br />

= 1 +<br />

= 1 +<br />

−<br />

−<br />

∞<br />

k,l,m=0<br />

∞<br />

k,l,m=0<br />

∞<br />

k,l,m=0<br />

m=0<br />

∞<br />

k,l,m=0<br />

k,l=0<br />

∞<br />

k,l,m=0<br />

k,n,s>0<br />

k≥n+s<br />

n≥s<br />

∞<br />

k,l,m=0<br />

G10(k, n, s; z)p k−n−s q n−s r s−1<br />

G10(k + l + 2m + 2, l + m + 1, m + 1; z)p k q l r m<br />

G10(k + l + 2m + 2, l + m + 1, m + 1; z)p k q l r m+1<br />

G10(k + l + 2m + 2, l + m + 1, m + 1; z)p k+1 q l+1 r m<br />

G10(k + l + 2m, l + m, m; z)p k q l r m<br />

G10(k + l + 2m, l + m, m + 1; z)p k q l r m<br />

¯G10(k + l + 2m, l + m, m; z)p k q l r m ,<br />

where G10(k + l, l, 0; z) = G10(l + 2m, l + m, m + 1; z) = G10(k + 2m, k +<br />

m, m + 1; z) = 0. In <strong>the</strong> same way, we have also<br />

Thus we obtain<br />

(Φ −1<br />

1 )11 = 1 +<br />

(Φ −1<br />

1 )11(Φ0)11 =<br />

<br />

×<br />

1 +<br />

<br />

1 +<br />

∞<br />

k,l,m=0<br />

∞<br />

k,l,m=0<br />

∞<br />

k,l,m=0<br />

¯G10(k + l + 2m, k + m, m; z)p k q l r m .<br />

¯G10(k + l + 2m, l + m, m; z)p k q l r m<br />

¯G10(k + l + 2m, k + m, m; 1 − z)p k q l r m


42 Chapter. 2<br />

and<br />

+ <br />

= 1 + <br />

¯G10(k + l + 2m, l + m, m; z)<br />

k,l,m<br />

+ ¯ <br />

G10(k + l + 2m, k + m, m; 1 − z p k q l r m<br />

<br />

k,l,m k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ <br />

¯G10(k<br />

=m<br />

′ + l ′ + 2m ′ , k ′ + m ′ , m ′ ; z)<br />

× ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ <br />

; 1 − z) p k q l r m<br />

(Φ −1<br />

1 )12(Φ0)21 = αβ<br />

<br />

<br />

G10(k + l + 2m + 1, l + m + 1, m + 1; z)<br />

k,l,m<br />

<br />

<br />

× G10(k + l + 2m + 1, k + m + 1, m + 1; 1 − z)<br />

= <br />

⎜<br />

⎝<br />

k,l,m<br />

= <br />

<br />

⎛<br />

k,l,m<br />

<br />

− <br />

⎞<br />

⎟<br />

k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ k<br />

=m−1<br />

′ +k ′′ =k−1<br />

l ′ +l ′′ =l−1<br />

m ′ +m ′′ ⎟<br />

⎠<br />

=m<br />

G10(k ′ + l ′ + 2m ′ + 1, l ′ + m ′ + 1, m ′ + 1; z)<br />

× G10(k ′ + l ′ + 2m ′ + 1, k ′ + m ′ + 1, m ′ + 1; 1 − z)p k q l r m<br />

<br />

k,l,m k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ =m<br />

= <br />

<br />

k,l,m k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ =m<br />

¯G0(k ′ + l ′ + 2m ′ − 1, l ′ + m ′ , m ′ ; z)<br />

G0(k ′′ + l ′′ + 2m ′′ + 1, k ′′ + m ′′ , m ′′ ; 1 − z)p k q l r m<br />

G0(k ′ + l ′ + 2m ′ + 1, l ′ + m ′ , m ′ ; z)<br />

¯G0(k ′′ + l ′′ + 2m ′′ − 1, k ′′ + m ′′ , m ′′ ; 1 − z)p k q l r m<br />

Therefore (Φ −1<br />

1 )11(Φ0)11 + (Φ −1<br />

1 )12(Φ0)21 is calculated to <strong>the</strong> left hand side <strong>of</strong><br />

(2.57).<br />

Since limz→1 G10(k, n, s; z) = G10(k, n, s; 1), limz→1 G10(k, n, s; 1−z) = 0,<br />

and limz→1 G0(k, n, s; z)G0(k ′ , n ′ , s ′ ; 1 − z) = 0, <strong>the</strong> limit <strong>of</strong> <strong>the</strong> <strong>equation</strong><br />

(2.57) <strong>as</strong> z → 1 and 0 implies <strong>the</strong> following corollary.<br />

<br />

.


2.6 Connection formula between z = 0, 1 43<br />

Corollary 16.<br />

<br />

k,l,m<br />

= <br />

k,l,m<br />

= <br />

<br />

k<br />

k,l,m<br />

¯G10(k + l + 2m, l + m, m; 1)(1 − γ) k (α + β + 1 − γ) l ((α + 1 − γ)(β + 1 − γ)) m<br />

(2.59)<br />

¯G10(k + l + 2m, k + m, m; 1)(1 − γ) k (α + β + 1 − γ) l ((α + 1 − γ)(β + 1 − γ)) m<br />

i=0<br />

l<br />

j=0 µ=0<br />

m<br />

<br />

i + j<br />

i<br />

Pk−i(ζ)Pl−j(ζ)Pµ(−ζ)Pi+j+2m−µ(−ζ)Ni+j,m−µ<br />

× (1 − γ) k (α + β + 1 − γ) l ((α + 1 − γ)(β + 1 − γ)) m .<br />

The <strong>equation</strong> between <strong>the</strong> first and <strong>the</strong> second lines is <strong>the</strong> duality formula<br />

(1.3) for fixed weight, depth and height. The <strong>equation</strong> between <strong>the</strong> first (or<br />

<strong>the</strong> second) and <strong>the</strong> third lines is Ohno-Zagier relation originally shown by<br />

[OZ].<br />

2.6.4 Various examples <strong>of</strong> functional relations <strong>of</strong> MPLs<br />

In what follows, computing <strong>the</strong> coefficient <strong>of</strong> p k q l r m <strong>of</strong> <strong>the</strong> formula (2.57)<br />

for some lower l and m, or by specializing <strong>the</strong> parameters, we show various<br />

concrete relations <strong>of</strong> <strong>the</strong> multiple polylogarithms <strong>of</strong> one variable.<br />

The c<strong>as</strong>e <strong>of</strong> m = 0 and l = 1<br />

By e<strong>as</strong>y calculation we have<br />

(<strong>the</strong> coefficient <strong>of</strong> p k q 1 r 0 <strong>of</strong> <strong>the</strong> left hand side <strong>of</strong> (2.57))<br />

= ¯ G10(k + 1, 1, 0; z) + ¯ G10(k + 1, k, 0; 1 − z)<br />

+ <br />

k ′ +k ′′ <br />

¯G10(k<br />

=k<br />

′ + 1, 1, 0; z) ¯ G10(k ′′ , k ′′ , 0; 1 − z)<br />

+ ¯ G0(k ′ , 1, 0; z)G0(k ′′ + 1, k ′′ <br />

+ 1, 1; 1 − z)<br />

k ′ +k ′′ <br />

¯G10(k<br />

=k<br />

′ , 0, 0; z) ¯ G10(k ′′ + 1, k ′′ , 0; 1 − z)<br />

+ ¯ G0(k ′ − 1, 0, 0; z)G0(k ′′ <br />

+ 2, 2, 1; 1 − z)<br />

+ <br />

= −G10(k + 1, 1, 1; z) − G10(k + 1, k, 1; 1 − z)<br />

− <br />

k ′ +k ′′ =k<br />

<br />

G0(k ′ , 1, 1; z)G0(k ′′ + 1, k ′′ + 1, 1; 1 − z)<br />

= − Lik+1(z) − Li2, 1,...,1<br />

<br />

k−1 times<br />

(1 − z) −<br />

k<br />

Lii(z) Li1,...,1<br />

<br />

i=1<br />

k−i+1 times<br />

<br />

(1 − z).


44 Chapter. 2<br />

On <strong>the</strong> o<strong>the</strong>r hand,<br />

(<strong>the</strong> coefficient <strong>of</strong> p k q 1 r 0 <strong>of</strong> <strong>the</strong> right hand side <strong>of</strong> (2.57)) (2.60)<br />

=<br />

k<br />

i=0<br />

1<br />

<br />

i + j<br />

Pk−i(ζ)P1−j(ζ)Pi+j(−ζ) =<br />

i<br />

j=0<br />

= <strong>the</strong> coefficient <strong>of</strong> z k ∞ ζ(n)<br />

in exp(<br />

n<br />

n=2<br />

zn ) d<br />

dz exp(−<br />

∞<br />

n=2<br />

= <strong>the</strong> coefficient <strong>of</strong> z k ∞<br />

in − ζ(i)z i−1<br />

= −ζ(k + 1).<br />

Consequently we obtain<br />

Lik+1(z) + Li2, 1,...,1<br />

<br />

k−1 times<br />

(1 − z) +<br />

i=2<br />

k<br />

k<br />

(i + 1)Pk−i(ζ)Pi+1(−ζ)<br />

i=0<br />

Lii(z) Li1,...,1<br />

<br />

i=1<br />

k−i+1 times<br />

ζ(n)<br />

n zn )<br />

(1 − z) = ζ(k + 1). (2.61)<br />

This is known <strong>as</strong> Euler’s inversion formula for polylogarithms.<br />

The c<strong>as</strong>e <strong>of</strong> m = 0 and l = 2<br />

Similarly we have<br />

(<strong>the</strong> coefficient <strong>of</strong> p k q 2 r 0 <strong>of</strong> <strong>the</strong> left hand side <strong>of</strong> (2.57)) (2.62)<br />

= ¯ G10(k + 2, 2, 0; z) + ¯ G10(k + 2, k, 0; 1 − z)<br />

+ <br />

k ′ +k ′′ <br />

¯G10(k<br />

=k<br />

′ + 2, 2, 0; z) ¯ G10(k ′′ , k ′′ , 0; 1 − z)<br />

+ ¯ G0(k ′ + 1, 2, 0; z)G0(k ′′ + 1, k ′′ <br />

+ 1, 1; 1 − z)<br />

k ′ +k ′′ <br />

¯G10(k<br />

=k<br />

′ + 1, 1, 0; z) ¯ G10(k ′′ + 1, k ′′ , 0; 1 − z)<br />

+ ¯ G0(k ′ , 1, 0; z)G0(k ′′ + 2, k ′′ <br />

+ 1, 1; 1 − z)<br />

k ′ +k ′′ <br />

¯G10(k<br />

=k<br />

′ , 0, 0; z) ¯ G10(k ′′ + 2, k ′′ , 0; 1 − z)<br />

+ ¯ G0(k ′ − 1, 0, 0; z)G0(k ′′ + 3, k ′′ <br />

+ 1, 1; 1 − z)<br />

+ <br />

+ <br />

= −G10(k + 2, 2, 1; z) − G10(k + 2, k, 1; 1 − z)<br />

− <br />

k ′ +k ′′ =k<br />

+ <br />

k ′ +k ′′ =k<br />

<br />

G0(k ′ + 1, 2, 1; z)G0(k ′′ + 1, k ′′ + 1, 1; 1 − z)<br />

<br />

G10(k ′ + 1, 1, 1; z)G10(k ′′ + 1, k ′′ , 1; 1 − z)<br />

− G0(k ′ , 1, 1; z)G0(k ′′ + 2, k ′′ + 1, 1; 1 − z)


2.6 Connection formula between z = 0, 1 45<br />

and<br />

= − Lik+1,1(z) − Li3, 1,...,1 (1 − z)<br />

<br />

k−1 times<br />

− <br />

k ′ +k ′′ =k<br />

+ <br />

Lik ′ ,1(z) Li 1,...,1(1<br />

− z)<br />

<br />

k ′′ +1 times<br />

k ′ +k ′′ Lik<br />

=k<br />

′ +1(z) Li2, 1,...,1<br />

<br />

k ′′ −1 times<br />

− <br />

(1 − z)<br />

k ′ +k ′′ Lik ′(z) Li2,1,...,1<br />

<br />

=k<br />

k ′′ (1 − z)<br />

times<br />

= − Lik+1,1(z) − Li3, 1,...,1(1<br />

− z) − Li1(z) Li2, 1,...,1<br />

<br />

<br />

k−1 times<br />

k−1 times<br />

−<br />

k<br />

(1 − z)<br />

Lii,1(z) Li1,...,1<br />

<br />

i=1<br />

k−i+1 times<br />

(1 − z),<br />

(<strong>the</strong> coefficient <strong>of</strong> p k q 2 r 0 <strong>of</strong> <strong>the</strong> right hand side <strong>of</strong> (2.57)) (2.63)<br />

k 2<br />

<br />

i + j<br />

=<br />

Pk−i(ζ)Pl−j(ζ)Pi+j(−ζ)<br />

i<br />

i=0 j=0<br />

= <strong>the</strong> coefficient <strong>of</strong> z k in 1<br />

2 exp(<br />

∞ ζ(n)<br />

n zn ) d2<br />

∞<br />

exp(−<br />

dz2 i=1<br />

n=2<br />

k + 1 1 k−1<br />

= − ζ(k + 2) + ζ(i + 1)ζ(k − i + 1).<br />

2 2<br />

Therefore, we have<br />

Lik+1,1(z) + Li3, 1,...,1(1<br />

− z) + Li1(z) Li2, 1,...,1<br />

<br />

<br />

k−1 times<br />

k−1 times<br />

+<br />

k<br />

Lii,1(z) Li1,...,1<br />

<br />

i=1<br />

k−i+1 times<br />

= k + 1 1 k−1<br />

ζ(k + 2) − ζ(i + 1)ζ(k − i + 1).<br />

2 2<br />

i=1<br />

n=2<br />

ζ(n)<br />

n zn )<br />

(1 − z) (2.64)<br />

(1 − z)<br />

Especially taking <strong>the</strong> limit <strong>as</strong> z tends to 1, we get <strong>the</strong> formula shown by<br />

Euler([Z1]);<br />

ζ(k + 1, 1) −<br />

k + 1 1 k−1<br />

ζ(k + 2) + ζ(i + 1)ζ(k − i + 1) = 0. (2.65)<br />

2 2<br />

i=1


46 Chapter. 2<br />

The sum formula for MPLs<br />

By comparing <strong>the</strong> coefficients <strong>of</strong> α 1 in both sides <strong>of</strong> (2.57) <strong>as</strong> a series in<br />

p = (1 − γ), q ′ = (β + 1 − γ), we obtain <strong>the</strong> following proposition;<br />

Proposition 17. For any positive integers k > n > 0,<br />

<br />

G10(k, n, s; z) + <br />

G10(k, k − n, s; 1 − z)<br />

s<br />

+ <br />

k ′ +k ′′ =k<br />

n ′ +n ′′ =n<br />

<br />

s ′<br />

s<br />

G0(k ′ , n ′ , s ′ ; z) <br />

G0(k ′′ , k ′′ − n ′′ , s ′′ ; 1 − z) = ζ(k). (2.66)<br />

s ′′<br />

Particularly, <strong>the</strong> limit <strong>of</strong> <strong>the</strong> formula (2.66) for z → 1 is <strong>the</strong> sum formula<br />

<strong>of</strong> <strong>the</strong> multiple zeta values (1.2)<br />

Pro<strong>of</strong>. Put q0 = (β + 1 − γ). We have<br />

<br />

G10(k, n, s; 1) = ζ(k). (2.67)<br />

s<br />

d<br />

<br />

<br />

(<strong>the</strong> right hand side <strong>of</strong> (2.57)) <br />

dα α→0<br />

= d<br />

<br />

exp(<br />

dα<br />

ζ(n)<br />

n (α + β + 1 − γ)n ) exp(− ζ(n)<br />

n (α + 1 − γ)n )<br />

× exp( ζ(n)<br />

n (1 − γ)n ) exp(− ζ(n)<br />

n (β + 1 − γ)n )<br />

<br />

n−1 n−1<br />

= ζ(n)(α + β + 1 − γ) − ζ(n)(α + 1 − γ)<br />

× exp( ζ(n)<br />

n (α + β + 1 − γ)n ) exp(− ζ(n)<br />

n (α + 1 − γ)n )<br />

× exp( ζ(n)<br />

n (1 − γ)n ) exp(− ζ(n)<br />

n (β + 1 − γ)n <br />

<br />

)<br />

<br />

n−1 n−1<br />

= ζ(n)(β + 1 − γ) − ζ(n)(1 − γ)<br />

= − <br />

ζ(k + 1)p k + <br />

ζ(l + 1)q l 0.<br />

k<br />

l<br />

On <strong>the</strong> o<strong>the</strong>r hand, we have<br />

<br />

¯Gi(k, n, m; z) = 0, (2.68)<br />

m<br />

<br />

m ¯ Gi(k, n, m; z) = <br />

Gi(k, n, m; z) (2.69)<br />

m<br />

m<br />

α→0<br />

α→0


2.6 Connection formula between z = 0, 1 47<br />

by definition for i = 0, 10 and for all k, n. By using <strong>the</strong>m, one can calculate<br />

<strong>the</strong> left hand side <strong>of</strong> (2.57) <strong>as</strong> follows.<br />

d<br />

<br />

<br />

(<strong>the</strong> left hand side <strong>of</strong> (2.57)) <br />

dα α→0<br />

= <br />

<br />

¯G10(k + l + 2m, l + m, m; z) + ¯ G10(k + l + 2m, k + m, m; 1 − z)<br />

k,l,m<br />

+ <br />

k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ =m<br />

¯G10(k ′ + l ′ + 2m ′ , l ′ + m ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ ; 1 − z)<br />

+ ¯ G0(k ′ +l ′ +2m ′ −1, l ′ +m ′ , m ′ ; z)G0(k ′′ +l ′′ +2m ′′ +1, k ′′ +m ′′ +1, m ′′ +1; 1 − z)<br />

The first term <strong>of</strong> (2.70) is calculated <strong>as</strong><br />

<br />

k,l,m<br />

= <br />

k,l,m<br />

× (lp k+m q l+m−1<br />

0<br />

¯G10(k + l + 2m, l + m, m; z)(lp k+m q l+m−1<br />

0<br />

l ¯ G10(k + l + 2m, l + m, m; z)p k+m q l+m−1<br />

0<br />

+ <br />

k,l,m<br />

m ¯ G10(k + l + 2m, l + m, m; z)p k+m−1 q l+m<br />

0<br />

= <br />

(l + 1 − m) ¯ G10(k + l + 1, l + 1, m; z)p k q l 0 + <br />

k,l,m<br />

+ mp k+m−1 q l+m<br />

0 ). (2.70)<br />

+ mp k+m−1 q l+m<br />

0 )<br />

k,l,m<br />

<br />

m ¯ G10(k + l + 1, l, m; z)p k q l 0<br />

= <br />

<br />

− <br />

G10(k + l + 1, l + 1, m; z) + <br />

<br />

G10(k + l + 1, l, m; z)<br />

k,l<br />

m<br />

In <strong>the</strong> same way, we have<br />

<br />

k,l,m<br />

¯G10(k + l + 2m, k + m, m; 1 − z)(lp k+m q l+m−1<br />

0<br />

m<br />

+ mp k+m−1 q l+m<br />

0 )<br />

p k q l 0.<br />

= <br />

<br />

− <br />

G10(k + l + 1, k, m; 1 − z) + <br />

<br />

G10(k + l + 1, k + 1, m; 1 − z)<br />

k,l<br />

m<br />

<strong>as</strong> <strong>the</strong> second term <strong>of</strong> (2.70). For <strong>the</strong> third term <strong>of</strong> (2.70), we obtain<br />

<br />

k,l,m k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ =m<br />

m<br />

¯G10(k ′ + l ′ + 2m ′ , l ′ + m ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ ; 1 − z)<br />

× (lp k+m q l+m−1<br />

0<br />

+ mp k+m−1 q l+m<br />

0 )<br />

p k q l 0


48 Chapter. 2<br />

= <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′ ,m ′′<br />

= <br />

+ <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

¯G10(k ′ + l ′ + 2m ′ , l ′ + m ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ ; 1 − z)<br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′ ,m ′′<br />

+ <br />

= <br />

m ′ ,m ′′<br />

m ′<br />

<br />

× (l ′ + l ′′ )p k′ +m ′ +k ′′ +m ′′<br />

q l′ +m ′ +l ′′ +m ′′ −1<br />

0<br />

¯G10(k ′ + l ′ + 2m ′ , l ′ + m ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 2m ′′ , k ′′ + m ′′ , m ′′ ; 1 − z)<br />

m ′ ,m ′′<br />

× (m ′ + m ′′ )p k′ +m ′ +k ′′ +m ′′ −1 q l ′ +m ′ +l ′′ +m ′′<br />

0<br />

(l ′ + l ′′ + 1 − m ′ − m ′′ ) ¯ G10(k ′ + l ′ , l ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 1, k ′′ , m ′′ ; 1 − z)<br />

(m ′ + m ′′ ) ¯ G10(k ′ + l ′ , l ′ , m ′ ; z) ¯ G10(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

<br />

(l ′ + l ′′ + 1) <br />

× p k′ +k ′′<br />

q l′ +l ′′<br />

0<br />

¯G10(k ′ + l ′ , l ′ , m ′ ; z) <br />

<br />

¯G10(k ′′ + l ′′ + 1, k ′′ , m ′′ ; 1 − z)<br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′<br />

m ′′<br />

+ <br />

(−m ′ ) ¯ G10(k ′ + l ′ , l ′ , m ′ ; z) <br />

¯G10(k ′′ + l ′′ + 1, k ′′ , m ′′ ; 1 − z)<br />

= 0.<br />

+ <br />

m ′′<br />

¯G10(k ′ + l ′ , l ′ , m ′ ; z) <br />

(−m ′′ ) ¯ G10(k ′′ + l ′′ + 1, k ′′ , m ′′ ; 1 − z)<br />

m ′<br />

m ′′<br />

+ <br />

m ′<br />

m ′ G10(k ¯ ′ + l ′ , l ′ , m ′ ; z) <br />

m ′′<br />

¯G10(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

+ <br />

¯G10(k ′ + l ′ , l ′ , m ′ ; z) <br />

m ′′ G10(k ¯ ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

m ′<br />

m ′′<br />

L<strong>as</strong>tly for <strong>the</strong> fourth term <strong>of</strong> (2.70),<br />

<br />

× p k′ +k ′′<br />

q l′ +l ′′<br />

0<br />

¯G0(k ′ +l ′ +2m ′ −1, l ′ +m ′ , m ′ ; z)G0(k ′′ +l ′′ +2m ′′ +1, k ′′ +m ′′ +1, m ′′ +1; 1 − z)<br />

k,l,m k ′ +k ′′ =k<br />

l ′ +l ′′ =l<br />

m ′ +m ′′ × (lp =m<br />

k+m q l+m−1<br />

0 + mp k+m−1 q l+m<br />

0 )<br />

¯G0(k ′ +l ′ +2m ′ −1, l ′ +m ′ , m ′ ; z)G0(k ′′ +l ′′ +2m ′′ +1, k ′′ +m ′′ +1, m ′′ +1; 1 − z)<br />

= <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′ ,m ′′<br />

= <br />

+ <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′ ,m ′′<br />

× (l ′ + l ′′ )p k′ +m ′ +k ′′ +m ′′<br />

q l′ +m ′ +l ′′ +m ′′ −1<br />

0<br />

¯G0(k ′ +l ′ +2m ′ −1, l ′ +m ′ , m ′ ; z)G0(k ′′ +l ′′ +2m ′′ +1, k ′′ +m ′′ +1, m ′′ +1; 1 − z)<br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′ ,m ′′<br />

× (m ′ + m ′′ )p k′ +m ′ +k ′′ +m ′′ −1 l<br />

q ′ +m ′ +l ′′ +m ′′<br />

0<br />

<br />

(l ′ + l ′′ + 1 −m ′ −m ′′ ) ¯ G0(k ′ + l ′ − 1, l ′ , m ′ ; z)G0(k ′′ + l ′′ + 2, k ′′ + 1, m ′′ + 1; 1 − z)<br />

+ (m ′ + m ′′ ) ¯ G0(k ′ + l ′ − 1, l ′ , m ′ ; z)G0(k ′′ + l ′′ + 2, k ′′ + 2, m ′′ <br />

+ 1; 1 − z)<br />

× p k′ +k ′′<br />

q l′ +l ′′<br />

0


2.6 Connection formula between z = 0, 1 49<br />

= <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

= <br />

+ <br />

m ′<br />

+ <br />

m ′<br />

+ <br />

m ′<br />

+ <br />

k ′ ,k ′′ ,l ′ ,l ′′<br />

m ′<br />

+ <br />

m ′<br />

<br />

(l ′ + l ′′ + 1) <br />

¯G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 2, k ′′ + 1, m ′′ + 1; 1 − z)<br />

m ′<br />

m ′′<br />

(−m ′ ) ¯ G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 2, k ′′ + 1, m ′′ + 1; 1 − z)<br />

m ′<br />

¯G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

(−m ′′ )G0(k ′′ + l ′′ + 2, k ′′ + 1, m ′′ + 1; 1 − z)<br />

m ′′<br />

m ′ G0(k ¯ ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 2, k ′′ + 2, m ′′ + 1; 1 − z)<br />

m ′′<br />

¯G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

m ′′ G0(k ′′ + l ′′ + 2, k ′′ + 2, m ′′ + 1; 1 − z)<br />

<br />

− <br />

= <br />

k,l<br />

Thus<br />

m ′′<br />

× p k′ +k ′′<br />

q l′ +l ′′<br />

0<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 2, k ′′ + 1, m ′′ + 1; 1 − z)<br />

m ′<br />

m ′<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 2, k ′′ + 2, m ′′ + 1; 1 − z)<br />

<br />

k ′ +k ′′ =k+1<br />

l ′ +l ′′ m<br />

=l<br />

′<br />

− <br />

k ′ +k ′′ =k<br />

l ′ +l ′′ =l+1<br />

<br />

m ′<br />

m ′′<br />

× p k′ +k ′′<br />

q l′ +l ′′<br />

0<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

m ′′<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z) <br />

G0(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

× p k q l 0.<br />

<strong>the</strong> coefficient <strong>of</strong> p k q l 0 in <strong>the</strong> left hand side <strong>of</strong> (2.70)<br />

= <br />

G10(k + l + 1, l, m; z) − <br />

G10(k + l + 1, l + 1, m; z)<br />

m<br />

m<br />

+ <br />

G10(k + l + 1, k + 1, m; 1 − z) − <br />

G10(k + l + 1, k, m; 1 − z)<br />

m<br />

+ <br />

<br />

k ′ +k ′′ =k+1<br />

l ′ +l ′′ m<br />

=l<br />

′<br />

− <br />

k ′ +k ′′ =k<br />

l ′ +l ′′ =l+1<br />

<br />

m ′<br />

m ′<br />

m<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z)G0(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)<br />

G0(k ′ + l ′ − 1, l ′ , m ′ ; z)G0(k ′′ + l ′′ + 1, k ′′ + 1, m ′′ ; 1 − z)


50 Chapter. 2<br />

holds. Therefore we have<br />

<br />

G10(k, n, s; z) + <br />

G10(k, k − n, s; 1 − z) (2.71)<br />

s<br />

+ <br />

k ′ +k ′′ =k<br />

n ′ +n ′′ =n<br />

<br />

s ′<br />

s<br />

G0(k ′ , n ′ , s ′ ; z) <br />

G0(k ′′ , k ′′ − n ′′ , s ′′ ; 1 − z)<br />

= <br />

G10(k, n − 1, s; z) + <br />

G10(k, k − (n − 1), s; 1 − z)<br />

s<br />

+ <br />

k ′ +k ′′ =k<br />

n ′ +n ′′ =n−1<br />

<br />

s ′<br />

s ′′<br />

s<br />

G0(k ′ , n ′ , s ′ ; z) <br />

G0(k ′′ , k ′′ − n ′′ , s ′′ ; 1 − z)<br />

s ′′<br />

(k > n, n > 1)<br />

and<br />

<br />

G10(k, 1, s; z) + <br />

G10(k, k − 1, s; 1 − z) (2.72)<br />

s<br />

+ <br />

k ′ +k ′′ =k<br />

n ′ +n ′′ =1<br />

<br />

These formul<strong>as</strong> yield <strong>the</strong> proposition.<br />

s ′<br />

s<br />

G0(k ′ , n ′ , s ′ ; z) <br />

G0(k ′′ , k ′′ − n ′′ , s ′′ ; 1 − z) = ζ(k).<br />

s ′′<br />

2.7 Functional relations derived from <strong>the</strong> connection<br />

formula between irregular solutions<br />

In this section, we consider functional relations <strong>of</strong> <strong>the</strong> multiple polylogarithms<br />

derived from <strong>the</strong> connection formula between solutions to <strong>the</strong> <strong>hypergeometric</strong><br />

<strong>equation</strong> on neighborhoods <strong>of</strong> z = 0 and ∞. The construction <strong>of</strong> solutions<br />

shown in Section 2.5 and Section 2.6 can be applied whe<strong>the</strong>r <strong>the</strong> solution is<br />

holomorphic or not, <strong>the</strong>n we can obtain <strong>the</strong> functional relations with respect<br />

to <strong>the</strong> (1, 2), (2, 1) and (2, 2)-element <strong>of</strong> <strong>the</strong> connection formula (2.21) between<br />

z = 0 and z = 1 and <strong>the</strong> connection formula (2.22) between z = 0 and<br />

z = ∞ in a similar way <strong>as</strong> above. But, in general, <strong>the</strong>se relations are too<br />

complicated to be described explicitly.<br />

In what follows, we give <strong>the</strong> functional relations derived from <strong>the</strong> limit<br />

<strong>of</strong> <strong>the</strong> (1, 1)-element <strong>of</strong> (2.22) <strong>as</strong> β tends to 0. The results include Euler’s<br />

inversion formula between z = 0 and ∞ and Riemann zeta values at even<br />

positive integers <strong>as</strong> <strong>the</strong> limit <strong>as</strong> z → 1.


2.7 Functional relations derived from z = 0, ∞ 51<br />

2.7.1 The fundamental solution matrix on <strong>the</strong> neighborhood<br />

<strong>of</strong> z = ∞<br />

Let u = 1<br />

z be a complex coordinate at z = ∞ and ρ∞ : X → M(2, C) be a<br />

<strong>representation</strong> defined by<br />

<br />

0<br />

ρ∞(X) = ρ0(Y ) − ρ0(X) =<br />

α<br />

<br />

<br />

−β<br />

0<br />

, ρ∞(Y ) = ρ0(Y ) =<br />

α + β<br />

α<br />

<br />

0<br />

.<br />

q<br />

(2.73)<br />

From d<br />

d 1<br />

u2<br />

= −u2 and = −u− , one can rewrite <strong>Gauss</strong> <strong>hypergeometric</strong><br />

dz du 1−z 1−u<br />

<strong>equation</strong> (2.33) <strong>as</strong><br />

<br />

d ρ∞(X)<br />

G =<br />

du u + ρ∞(Y<br />

<br />

)<br />

G, (2.74)<br />

1 − u<br />

and get <strong>the</strong> fundamental solution ρ∞(H0(u)). Comparing <strong>the</strong> <strong>as</strong>ymptotic<br />

properties, we have<br />

Lemma 18.<br />

<br />

1 1<br />

Φ∞ = ρ∞(H0(u))<br />

<br />

1 1<br />

Pro<strong>of</strong>. Put P =<br />

. We show<br />

Since ρ∞(X) = P<br />

− α<br />

β −1<br />

α 0<br />

0 β<br />

− α<br />

β<br />

Φ∞P −1 u −ρ∞(X) → I (u → 0).<br />

<br />

P −1 , we have<br />

u −ρ∞(X) = P<br />

Therefore by e<strong>as</strong>y computation, we obtain<br />

<br />

Φ∞P −1 u −ρ∞(X) = Φ∞P −1 P<br />

<br />

=<br />

− 1<br />

= β<br />

β − α<br />

→ β<br />

β − α<br />

u −α 0<br />

0 u −β<br />

<br />

−α u 0<br />

0 u−β <br />

P −1 .<br />

P −1 = Φ∞<br />

uαF∞0 uβF∞1 β uα (αF∞0 + uF ′ ∞0) − 1<br />

β uβ (βF∞1 + uF ′ ∞1)<br />

<br />

where ϕ (∞)<br />

0<br />

− 1<br />

α 1 − β<br />

−1<br />

<br />

. (2.75)<br />

<br />

−α u 0<br />

0 u−β <br />

P −1<br />

<br />

−α u 0<br />

0 u−β <br />

β<br />

β − α P<br />

F∞0 − α<br />

β F∞1<br />

F∞0 − F∞1<br />

β (αF∞0 + uF ′ ∞0 − αF∞1 − uF ′ ∞1) − 1<br />

β (αF∞0 + uF ′ ∞0 − βF∞1 − uF ′ ∞1)<br />

0 − 1<br />

β<br />

= u α F∞0, ϕ (∞)<br />

1<br />

and F ′ stands for d<br />

du<br />

F .<br />

<br />

0<br />

= I,<br />

(α − β)<br />

= u β F∞1, F∞0, F∞1 are holomorphic at u = 0,


52 Chapter. 2<br />

Then we obtain<br />

Φ −1<br />

∞ = β<br />

<br />

1 1<br />

β − α − α<br />

β −1<br />

<br />

ρ∞(H0(u)) −1 , (2.76)<br />

ρ∞(H0(u)) −1 = <br />

Li( ¯ S(w); u)ρ∞(W ).<br />

w<br />

In (2.76), it is difficult to calculate ρ∞(W ) concretely, but one can see<br />

<strong>the</strong> limit <strong>of</strong> ρ∞(W ) and ρ∞(H0(u)) −1 for β → 0 <strong>as</strong> follows.<br />

Lemma 19. For any word W = I in H, we have<br />

lim<br />

β→0 ρ∞(W<br />

⎧<br />

⎪⎨ α<br />

) =<br />

⎪⎩<br />

|W |−d(w) <br />

d(W )−1 0 0<br />

(α + p)<br />

α α + p<br />

α |W |−d(w) <br />

<br />

d(W )−1 0 0<br />

(α + p)<br />

α + p α + p<br />

Pro<strong>of</strong>. Under <strong>the</strong> limit <strong>as</strong> β tends to 0, we have<br />

<br />

0<br />

ρ∞(X) →<br />

α<br />

<br />

0<br />

,<br />

α<br />

<br />

0<br />

ρ∞(Y ) →<br />

α<br />

<br />

0<br />

.<br />

α + p<br />

Therefore<br />

ρ∞(X k1−1 Y · · · X kr−1 Y ) → α k1+···+kr−r<br />

= α k1+···+kr−r (α + p) r−1<br />

ρ∞(X k1−1 Y · · · X kr−1 Y X s ) → α k1+···+kr−r (α + p) r−1<br />

= α k1+···+kr−r+s (α + p) r−1<br />

hold. These results imply <strong>the</strong> lemma’s formul<strong>as</strong>.<br />

r<br />

(W ∈ HY )<br />

(W ∈ HX).<br />

(2.77)<br />

<br />

0 0<br />

α α + p<br />

<br />

0<br />

α<br />

<br />

0<br />

α<br />

<br />

0<br />

,<br />

α + p<br />

<br />

0<br />

α<br />

α + p<br />

s−1<br />

<br />

<br />

0<br />

α<br />

<br />

0 0<br />

<br />

0<br />

α<br />

α + p α + p<br />

Lemma 20. We denote by ˜g1(k, n) <strong>the</strong> sum <strong>of</strong> all words <strong>of</strong> h started with x<br />

which consists <strong>of</strong> (k − n) x’s and n y’s. Then we have<br />

˜g1(k, n) =<br />

<br />

(−1) k−n−1−i (x k−n−i y n ) x x i . (2.78)<br />

k−n−1<br />

i=0


2.7 Functional relations derived from z = 0, ∞ 53<br />

Pro<strong>of</strong>. We prove <strong>the</strong> lemma by induction on n. Put k ′ = k−n. By definition,<br />

we note that<br />

(x a y) x x b b<br />

<br />

a + s<br />

=<br />

x<br />

s<br />

a+s yx b−s<br />

(2.79)<br />

holds in general non-negative integer a, b. Thus, for n = 1, we have<br />

s=0<br />

k ′ −1<br />

(−1)<br />

i=0<br />

k′ −1−i k<br />

(x ′ k<br />

−i i<br />

y) x x =<br />

′ −1<br />

(−1)<br />

i=0<br />

k′ i<br />

<br />

′<br />

−1−i k − i + s<br />

x<br />

s<br />

s=0<br />

k′ −i+s i−s<br />

yx<br />

k<br />

=<br />

′ −1<br />

k<br />

s=0<br />

′ −1<br />

(−1)<br />

j=s<br />

j−s<br />

<br />

j + 1<br />

x<br />

s<br />

j+1 yx k′ −1−j<br />

k<br />

=<br />

′ −1<br />

(−1) j<br />

<br />

j<br />

(−1) s<br />

<br />

j + 1<br />

s<br />

<br />

j=0<br />

k<br />

=<br />

′ −1<br />

j=0<br />

s=0<br />

x j+1 yx k′ −1−j = ˜g1(k ′ + 1, 1).<br />

x j+1 yx k′ −1−j<br />

For general, we suppose that <strong>the</strong> claim holds for ˜g1(k ′ + 1, 1), . . . , ˜g1(k ′ +<br />

n, n). Therefore we obtain<br />

k ′ −1<br />

(−1) k′ −1−i k<br />

(x ′ −i n+1 i<br />

y ) x x<br />

i=0<br />

k<br />

=<br />

′ −1<br />

(−1) k′ −1−i k<br />

((x ′ k<br />

−i n i<br />

y ) x x )y +<br />

′ −1<br />

(−1) k′ −1−i k<br />

((x ′ −i n+1 i−1<br />

y ) x x )x<br />

i=0<br />

i=1<br />

= ˜g1(k ′ k<br />

+ n, n)y +<br />

′ −2<br />

(−1) k′ −2−i k<br />

((x ′ −i n+1 i<br />

y ) x x )x<br />

i=0<br />

= ˜g1(k ′ + n, n)y + ˜g1(k ′ k<br />

− 1 + n, n)yx +<br />

′ −3<br />

= · · ·<br />

<br />

(−1) k′ −3−i k<br />

((x ′ −i n+1 i 2<br />

y ) x x )x<br />

= ˜g1(k ′ + n, n)y + ˜g1(k ′ − 1 + n, n)yx + ˜g1(k ′ − 2 + n, n)yx 2 + · · ·<br />

i=0<br />

+ ˜g1(2 + n, n)yx k′ −2 + (xy n+1 x x 0 )x k ′ −1<br />

= ˜g1(k ′ + n, n)y + · · · + ˜g1(2 + n, n)yx k′ −2 + ˜g1(1 + n, n)yx k′ −1<br />

= ˜g1(k ′ + n + 1, n + 1).


54 Chapter. 2<br />

Proposition 21. The following formula holds:<br />

<br />

1 0<br />

where<br />

H21 =<br />

∞<br />

k=1<br />

+ p <br />

lim<br />

β→0 ρ∞(H0(u)) −1 =<br />

(−1) k logk u<br />

k! αk + <br />

k>n≥1<br />

H22 = <br />

(−1) k<br />

<br />

(−1)<br />

k≥n≥0<br />

k Li 1,...,1<br />

<br />

n times<br />

Pro<strong>of</strong>. From Lemma 19,<br />

−1<br />

lim ρ∞(H0(u))<br />

β→0<br />

= I + <br />

k≥n≥1 w∈hy<br />

|w|=k,d(w)=n<br />

+ <br />

= I + <br />

(−1)<br />

k≥n≥1<br />

k Li 1,...,1<br />

<br />

n times<br />

H21 H22<br />

k−n−1<br />

(−1)<br />

i=0<br />

k−n−1−i Lik−n−i+1, 1,...,1<br />

<br />

n−1 times<br />

<br />

<br />

k>n≥0 w∈hx<br />

|w|=k,d(w)=n<br />

<br />

k≥n≥1 w∈yh<br />

|w|=k,d(w)=n<br />

+ <br />

= I + <br />

<br />

k>n≥0 w∈xh<br />

|w|=k,d(w)=n<br />

k≥n≥1 w∈h<br />

|w|=k,d(w)=n<br />

+ <br />

<br />

k>n≥1 w∈xh<br />

|w|=k,d(w)=n<br />

+ <br />

k≥1<br />

<br />

(−1) k Li(x k ; u)α k<br />

<br />

, (2.80)<br />

(u) logk−n u<br />

(k − n)! αk−n+1 (α + p) n−1<br />

(2.81)<br />

(u) logi (u)<br />

i! αk−n (α + p) n−1 ,<br />

(u) logk−n u<br />

(k − n)! αk−n (α + p) n . (2.82)<br />

Li( ¯ S(w); u)α k−n (α + p) n−1<br />

<br />

0 0<br />

α α + p<br />

Li( ¯ S(w); u)α k−n (α + p) n−1<br />

<br />

0 0<br />

α + p α + p<br />

(−1) k Li(w; u)α k−n (α + p) n−1<br />

<br />

0 0<br />

<br />

α α + p<br />

(−1) k Li(w; u)α k−n (α + p) n−1<br />

<br />

0 0<br />

α + p α + p<br />

(−1) k Li(w; u)α k−n (α + p) n−1<br />

(−1) k Li(w; u)α k−n (α + p) n−1<br />

<br />

0 0<br />

1 1<br />

<br />

0 0<br />

<br />

α α + p<br />

<br />

0 0<br />

p 0


2.7 Functional relations derived from z = 0, ∞ 55<br />

= I + <br />

(−1)<br />

k≥n≥1<br />

k Li(x k−n x y n ; u)α k−n (α + p) n−1<br />

<br />

0 0<br />

α α + p<br />

+ <br />

(−1)<br />

k>n≥1<br />

k Li(˜g1(k, n); u)α k−n (α + p) n−1<br />

<br />

0 0<br />

p 0<br />

+ <br />

(−1)<br />

k≥1<br />

k Li(x k ; u)α k<br />

<br />

0 0<br />

1 1<br />

= I + <br />

(−1)<br />

k≥1<br />

k logk u<br />

k! αk<br />

<br />

0 0<br />

1 1<br />

+ <br />

(−1)<br />

k≥n≥1<br />

k logk−n u<br />

(k − n)! Li 1,...,1 (u)α<br />

<br />

n times<br />

k−n (α + p) n−1<br />

<br />

0 0<br />

α α + p<br />

+ <br />

(−1) k Li(˜g1(k, n); u)α k−n (α + p) n−1<br />

<br />

0 0<br />

.<br />

p 0<br />

k>n≥1<br />

By using Lemma 20, <strong>the</strong> third term can be calculated <strong>as</strong><br />

<br />

k>n≥1<br />

− <br />

k>n≥1<br />

= <br />

(−1) k Li(˜g1(k, n); u)α k−n (α + p) n−1<br />

k>n≥1<br />

(−1) k<br />

(−1) k<br />

Thus we have<br />

<br />

<br />

k−n−1<br />

i=0<br />

<br />

−1<br />

lim ρ∞(H0(u))<br />

β→0<br />

<br />

0 0<br />

p 0<br />

(−1) k−n−1−i Li((x k−n−i y n ) x x i ; u)α k−n (α + p) n−1<br />

k−n−1<br />

(−1)<br />

i=0<br />

k−n−1−i Lik−n−i+1, 1,...,1<br />

<br />

n−1 times<br />

<br />

11<br />

H21 = <br />

(−1) k logk u<br />

k≥1<br />

+ <br />

k≥n≥1<br />

+ p <br />

k>n≥1<br />

k! αk<br />

= 1,<br />

(−1) k logk−n u<br />

(−1) k<br />

(k − n)! Li 1,...,1<br />

<br />

<br />

(u) logi u<br />

i!<br />

<br />

<br />

−1<br />

lim ρ∞(H0(u)) = 0,<br />

β→0<br />

12<br />

n times<br />

(u)α k−n+1 (α + p) n−1<br />

k−n−1<br />

(−1)<br />

i=0<br />

k−n−1−i Lik−n−i+1, 1,...,1<br />

<br />

n−1 times<br />

<br />

0 0<br />

p 0<br />

α k−n (α + p) n−1<br />

<br />

0 0<br />

.<br />

p 0<br />

(u) logi u<br />

i! αk−n (α + p) n−1 ,


56 Chapter. 2<br />

H22 = 1 + <br />

(−1) k logk u<br />

k! αk + <br />

= <br />

k≥n≥0<br />

k≥1<br />

(−1) k logk−n u<br />

(k − n)! Li 1,...,1<br />

<br />

n times<br />

k≥n≥1<br />

(−1) k logk−n u<br />

(k − n)! Li 1,...,1<br />

<br />

(u)α k−n (α + p) n .<br />

n times<br />

(u)α k−n (α + p) n<br />

2.7.2 The functional relations derived from <strong>the</strong> (1, 1)element<br />

<strong>of</strong> <strong>the</strong> connection formula between z = 0<br />

and ∞<br />

Multiplying both sides <strong>of</strong> <strong>the</strong> (1, 1)-elements <strong>of</strong> <strong>the</strong> connection formula (2.22)<br />

(Φ −1<br />

∞ )11(Φ0)11 + (Φ −1<br />

∞ )12(Φ0)21 = (C 0∞ )11<br />

(2.83)<br />

by β−α<br />

and <strong>the</strong>n taking <strong>the</strong> limit <strong>as</strong> β → 0, we obtain <strong>the</strong> following functional<br />

β<br />

relations <strong>of</strong> multiple polylogarithms.<br />

Proposition 22. The <strong>equation</strong><br />

1 + H21 + H22 lim<br />

β→0<br />

<br />

1 d<br />

z F (α, β, γ; z)<br />

β dz<br />

<br />

−πiα Γ(1 − p)Γ(1 − α)<br />

= e<br />

Γ(1 − (α + p))<br />

(2.84)<br />

holds on |α|, |α + 1 − γ| are sufficiently small and yields functional relations<br />

<strong>of</strong> multiple polylogarithms. Especially, <strong>as</strong> <strong>the</strong> coefficient <strong>of</strong> α m (α + 1 − γ) n<br />

for any positive integers m, n and z ∈ U, we have<br />

m 1 log z<br />

m! −<br />

m−1 <br />

i=0<br />

m−1 <br />

(−1)<br />

i=0<br />

n+i+1 Lim−i, 1,...,1<br />

<br />

n times<br />

=<br />

<br />

Lim−i(z) + (−1) m−i Lim−i( 1<br />

z )<br />

+<br />

<br />

m−1 <br />

i=0<br />

m1+m2+m3=m<br />

n1+n2=n<br />

n<br />

j=0<br />

( 1<br />

z<br />

<br />

log i 1<br />

z<br />

i!<br />

(−2πi)<br />

= Bm<br />

m<br />

, (2.85)<br />

m!<br />

1<br />

z )logi<br />

i! +<br />

m−1<br />

(−1)<br />

i=0<br />

n+i+1 Lim−i+1, 1,...,1<br />

<br />

n−1 times<br />

(−1) m+n−j−1 Li 1,...,1<br />

m1 + n1<br />

m1<br />

(<br />

<br />

1<br />

z )<br />

n−j times<br />

× <br />

s<br />

( 1<br />

z<br />

)logi 1<br />

z<br />

i!<br />

j<br />

<br />

m − i − 1 + j − k<br />

k=0<br />

m − i − 1<br />

1 logi z<br />

G0(m − i + j, k + 1, s; z)<br />

i!<br />

<br />

(−1) m1 Pm1+n1(ζ)Pm2(ζ) (−πi)m3<br />

m3! Pn2(−ζ),<br />

(2.86)


2.7 Functional relations derived from z = 0, ∞ 57<br />

where Bm are <strong>the</strong> Bernoulli numbers, namely <strong>the</strong> real numbers introduced<br />

through <strong>the</strong> generating function tm<br />

m<br />

Bm m!<br />

Pro<strong>of</strong>. We try to compute <strong>the</strong> both sides <strong>of</strong><br />

= t<br />

e t −1 .<br />

β − α −1<br />

lim (Φ∞ )11(Φ0)11 + (Φ<br />

β→0 β<br />

−1 β − α<br />

∞ )12(Φ0)21 = lim<br />

β→0 β (C0∞ )11. (2.87)<br />

The right hand side is equal to<br />

β − α<br />

lim<br />

β→0 β (C0∞ )11 = lim<br />

β→0<br />

β − α Γ(γ)Γ(β − α)<br />

e−πiα<br />

β Γ(β)Γ(γ − α)<br />

−πiα Γ(γ)Γ(1 + β − α) Γ(γ)Γ(1 − α) Γ(1 − p)Γ(1 − α)<br />

= lim e = e−πiα = e−πiα<br />

β→0 Γ(1 + β)Γ(γ − α) Γ(1)Γ(γ − α) Γ(1 − (α + p))<br />

∞ (−πiα)<br />

=<br />

k=0<br />

k ∞<br />

Pk(ζ)p<br />

k!<br />

k=0<br />

k<br />

∞<br />

Pk(ζ)α<br />

k=0<br />

k<br />

∞<br />

Pk(−ζ)(α + p)<br />

k=0<br />

k<br />

= <br />

⎛<br />

⎜ <br />

<br />

m1 + n1<br />

⎝<br />

(−1) m1Pm1+n1(ζ)Pm2(ζ) (−πi)m3<br />

m3! Pn2(−ζ)<br />

⎞<br />

⎟<br />

⎠<br />

m,n<br />

m1+m2+m3=m<br />

n1+n2=n<br />

m1<br />

Especially if m = 0, we have<br />

from eπiα Γ(1−p)Γ(1−α)<br />

Γ(1−(α+p))<br />

is<br />

<br />

n1+n2=n<br />

Pn1(ζ)Pn2(−ζ) =<br />

<br />

0 (n ≥ 1)<br />

1 (n = 0),<br />

× α m (α + p) n .<br />

<br />

= 1. Fur<strong>the</strong>rmore, since <strong>the</strong> coefficient <strong>of</strong> (α + p) α→0 0<br />

−πiα Γ(1 + α − (α + p))Γ(1 − α)<br />

<br />

<br />

e = e<br />

Γ(1 − (α + p)) α+p→0<br />

−πiα Γ(1 + α)Γ(1 − α)<br />

= e −πiα πα<br />

2πiα<br />

= e−πiα<br />

sin(πα) eπiα 2πiα<br />

=<br />

− e−πiα e2πiα − 1<br />

= (2πi)<br />

Bm<br />

m<br />

m! αm ,<br />

m≥0<br />

we have immediately<br />

<br />

m1+m2+m3=m<br />

(−1) m1Pm1(ζ)Pm2(ζ) (−πi)m3<br />

m3!<br />

= Bm<br />

(2πi) m<br />

m!


58 Chapter. 2<br />

for n = 0.<br />

On <strong>the</strong> o<strong>the</strong>r hand, <strong>the</strong> left hand side <strong>of</strong> (2.87) is calculated <strong>as</strong><br />

β − α<br />

lim<br />

β→0 β<br />

= lim<br />

β→0<br />

β − α<br />

β<br />

−1<br />

(Φ∞ )11(Φ0)11 + (Φ −1 <br />

∞ )12(Φ0)21<br />

β<br />

β − α<br />

(ρ∞(H0)) −1<br />

11 + (ρ∞(H0)) −1<br />

21 (Φ0)11<br />

+ (ρ∞(H0)) −1<br />

12 + (ρ∞(H0)) −1<br />

22 (Φ0)21<br />

= (1 + H21) lim<br />

β→0 (Φ0)11 + H22 lim<br />

β→0 (Φ0)21.<br />

By Theorem 5, we have<br />

lim<br />

β→0 (Φ0)11 = 1,<br />

lim<br />

β→0 (Φ0)12 = lim<br />

(<br />

β→0 1 d<br />

z F (α, β, γ; z))<br />

β dz<br />

= α <br />

k>n≥1<br />

G0(k − 1, n, ∗; z)p k−n−1 (α + p) n−1 ,<br />

where G0(k, n, ∗; z) stands for <br />

s G0(k, n, s; z) <strong>the</strong> sum <strong>of</strong> all multiple polylogarithms<br />

which have weight k and depth n.<br />

Therefore we obtain<br />

(1 + H21) lim<br />

β→0 (Φ0)11 + H22 lim<br />

β→0 (Φ0)21<br />

=<br />

∞<br />

(−1) m logm u<br />

m! αm + <br />

m=0<br />

+ ((α + p) − α) <br />

⎛<br />

+ α ⎝ <br />

m,n≥0<br />

m,n≥1<br />

(−1) m+n Li 1,...,1<br />

×<br />

(−1)<br />

m≥0,n≥1<br />

m+n Li 1,...,1<br />

<br />

n times<br />

<br />

(−1) m+n<br />

m−1<br />

<br />

i=0<br />

(−1) m−1−i Lim−i+1, 1,...,1<br />

(u)<br />

<br />

logm u<br />

m! αm (α + p) n<br />

⎞<br />

⎠<br />

n times<br />

m,n≥1<br />

<br />

(u) logm u<br />

m! αm+1 (α + p) n−1<br />

<br />

n−1 times<br />

G0(m + n − 1, n, ∗; z)p m−1 (α + p) n−1<br />

(u) logi (u)<br />

i! αm (α + p) n−1


2.7 Functional relations derived from z = 0, ∞ 59<br />

=<br />

∞<br />

(−1) m logm u<br />

m=0<br />

+ <br />

m! αm<br />

(−1)<br />

m≥1,n≥0<br />

m+n Li 1,...,1<br />

<br />

n+1 times<br />

+ <br />

<br />

(u) logm−1 u<br />

(m − 1)! αm (α + p) n<br />

(−1)<br />

m,n≥1<br />

m+n<br />

m−1<br />

(−1)<br />

i=0<br />

m−1−i Lim−i+1, 1,...,1<br />

<br />

n−1 times<br />

+ <br />

m≥2,n≥0<br />

+ <br />

m≥1,n≥0<br />

<br />

(−1) m+n+1<br />

m−2<br />

m−1<br />

<br />

i=0<br />

n<br />

j=0<br />

×<br />

i=0<br />

(−1) m−2−i Lim−i, 1,...,1<br />

<br />

n−1 times<br />

(−1) m+n−i−j−1 Li 1,...,1<br />

<br />

n−j times<br />

k=0<br />

Thus <strong>the</strong> coefficient <strong>of</strong> α m (α + p) n is<br />

(u) logi (u)<br />

i! αm (α + p) n<br />

(u) logi (u)<br />

i! αm (α + p) n−1<br />

(u) logm−i−1 u<br />

(m − i − 1)!<br />

j<br />

(−1) i<br />

<br />

i + j + k<br />

<br />

G0(i + j + 1, k + 1, ∗; z)<br />

i<br />

α m (α + p) n .<br />

⎧<br />

1 (m = n = 0)<br />

0 (m = 0, n ≥ 1)<br />

(−1)<br />

⎪⎨<br />

m logm m−1<br />

u log<br />

− (−1)m<br />

m!<br />

i=0<br />

m−i−1 u<br />

(m − i − 1)! Lii+1(z) (m ≥ 1, n = 0)<br />

m−1 <br />

− (−1)<br />

i=0<br />

i Lim−i(u) logi u<br />

i!<br />

m−1 <br />

(−1) n+i+1 Lim−i, 1,...,1 (<br />

<br />

1<br />

1<br />

z )logi (m, n ≥ 1)<br />

z i!<br />

⎪⎩<br />

i=0<br />

+<br />

+<br />

n times<br />

m−1 <br />

(−1)<br />

i=0<br />

n+i+1 Lim−i+1, 1,...,1<br />

<br />

n−1 times<br />

m−1 n<br />

(−1) m+n−j−1 Li 1,...,1<br />

i=0<br />

j=0<br />

( 1<br />

z<br />

(<br />

<br />

1<br />

z )<br />

n−j times<br />

)logi 1<br />

z<br />

i!<br />

j<br />

<br />

m − i − 1 + j − k<br />

k=0<br />

Comparing <strong>the</strong>se results, we complete <strong>the</strong> pro<strong>of</strong>.<br />

m − i − 1<br />

× <br />

G1(m − i + j, k + 1, s; z)<br />

s<br />

1 logi z .<br />

i!


60 Chapter. 2<br />

Especially, if m is an even positive integer, we have<br />

(2πi)<br />

−2ζ(m) = Bm<br />

m<br />

m!<br />

(2.88)<br />

when z tends to 1 in (2.85). Fur<strong>the</strong>rmore, taking <strong>the</strong> limit in (2.86) in <strong>the</strong><br />

c<strong>as</strong>e <strong>of</strong> n = 1 and 2, we obtain <strong>the</strong> following corollary.<br />

Corollary 23. The following relations among MZVs holds.<br />

(m + 2)ζ(m + 1) = 2 <br />

ζ(i + 1)ζ(2k) (m : odd),<br />

and<br />

i+2k=m,<br />

i,k≥1<br />

2ζ(m, 1) = mζ(m + 1) − 2 <br />

i+2k=m,<br />

i,k≥1<br />

(2.89)<br />

ζ(i + 1)ζ(2k) (m : even),<br />

(2.90)<br />

(m + 2)ζ(m + 1, 1) = − π2 (m + 2)(m + 1)<br />

ζ(m) + ζ(m + 2) (2.91)<br />

2 2<br />

− <br />

(i + 1)ζ(i + 2)ζ(2k)<br />

i+2k=m,<br />

i,k≥1<br />

− <br />

i+j+2k=m,<br />

i,j,k≥1<br />

(i + 1)ζ(i + 1)ζ(j + 1)ζ(2k) (m : odd),<br />

2ζ(m, 1, 1) = −mζ(m + 1, 1) + π2 (m + 2)(m + 1)<br />

ζ(m) −<br />

2 2<br />

+ <br />

i+2k=m,<br />

i,k≥1<br />

+ <br />

i+j+2k=m,<br />

i,j,k≥1<br />

(i + 1)ζ(i + 2)ζ(2k)<br />

ζ(m + 2)<br />

(2.92)<br />

(i + 1)ζ(i + 1)ζ(j + 1)ζ(2k) (m : even).<br />

Pro<strong>of</strong>. From <strong>the</strong> definition <strong>of</strong> MPLs and choice <strong>of</strong> <strong>the</strong> branches <strong>of</strong> <strong>the</strong>m on <strong>the</strong><br />

interval I defined in Section 2.2, we have <strong>the</strong> following analytic continuation;<br />

Li1,...,1<br />

<br />

n times<br />

( 1 1<br />

) =<br />

z n! (Li1(z) + log z + πi) n<br />

(z ∈ I), (2.93)


2.7 Functional relations derived from z = 0, ∞ 61<br />

1<br />

lim Lik1,k2,...,kn( ) = lim<br />

z→1 on I z z→1 on I Lik1,k2,...,kn(z) = ζ(k1, k2, . . . , kn) (2.94)<br />

(k1 ≥ 2).<br />

Here we note that Li(w, z) log z → 0 <strong>as</strong> z → 1 on I for any word w =<br />

1 in h. By <strong>the</strong> analytic continuation <strong>of</strong> (2.86) and <strong>the</strong> algebraic relation<br />

Li(w1; z)Li(w2; z)=Li(w1 x w2; z), we obtain<br />

(<strong>the</strong> left hand side <strong>of</strong> (2.86) at n = 1)<br />

→<br />

z→1 (1 + (−1)mm)ζ(m + 1) + (1 + (−1) m )ζ(m + 1) + (−1) m ζ(m)πi,<br />

(<strong>the</strong> left hand side <strong>of</strong> (2.86) at n = 2)<br />

→<br />

z→1 ((−1)m+1 − 1)ζ(m, 1, 1) + ((−1) m m π2<br />

m − 1)ζ(m + 1, 1) + (−1)<br />

m+1 m(m + 1)<br />

+ (−1)<br />

2<br />

2 ζ(m)<br />

ζ(m + 2) + (−1) m+1 ζ(m, 1)πi + (−1) m mζ(m + 1)πi.<br />

On <strong>the</strong> o<strong>the</strong>r hand, differentiating <strong>the</strong> (1, 1)-element <strong>of</strong> C 0∞ by (α + 1 − γ),<br />

taking <strong>the</strong> limit <strong>as</strong> (α + 1 − γ) → 0 and applying <strong>the</strong> <strong>equation</strong> (2.88), we<br />

have<br />

(<strong>the</strong> right hand side <strong>of</strong> (2.86) at n = 1)<br />

= (−1) m ζ(m + 1) − 2 <br />

i+2k=m<br />

i,k≥1<br />

(<strong>the</strong> right hand side <strong>of</strong> (2.86) at n = 2)<br />

(−1) i ζ(i + 1)ζ(2k) + (−1) m ζ(m)πi,<br />

= (−1) m (m + 1)ζ(m + 2) + (−1) m+1 ζ(m + 1, 1)<br />

− <br />

i+2k=m<br />

i,k≥1<br />

(−1) i (i + 1)ζ(i + 2)ζ(2k) − <br />

i=1<br />

i+j+2k=m<br />

i,j,k≥1<br />

m 1<br />

m−2 <br />

<br />

+ (−1) ζ(i + 1)ζ(m − i) + mζ(m + 1) πi.<br />

2<br />

(−1) i+j ζ(i + 1)ζ(j + 1)ζ(2k)<br />

The real parts <strong>of</strong> <strong>the</strong>se results yield <strong>the</strong> <strong>equation</strong>s from (2.89) to (2.92). The<br />

imaginary parts <strong>of</strong> <strong>the</strong>m are nothing but <strong>the</strong> identity for n = 1 and <strong>the</strong><br />

relation (2.65) previously shown for n = 2.


Chapter 3<br />

General <strong>representation</strong>s and<br />

many variable <strong>formal</strong> <strong>KZ</strong><br />

<strong>equation</strong>s<br />

In <strong>the</strong> previous chapter, we debated on <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong><br />

1<strong>KZ</strong> <strong>equation</strong> with respect to <strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong>. It is <strong>the</strong> most<br />

fundamental c<strong>as</strong>e <strong>of</strong> non-trivial semisimple <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong><br />

<strong>equation</strong>. In this chapter, we discuss general <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong><br />

1<strong>KZ</strong> <strong>equation</strong> and <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables.<br />

The contents <strong>of</strong> this chapter are now under investigation, but are important<br />

and interested targets to apply our method to make relations <strong>of</strong> MPLs and<br />

MVZs introduced on Chapter 2.<br />

3.1 Generalized <strong>hypergeometric</strong> <strong>equation</strong>s and<br />

multiple zeta values<br />

In this section, we consider <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> (1.38)<br />

with respect to <strong>the</strong> generalized <strong>hypergeometric</strong> <strong>equation</strong>. Let α1, . . . , αn+1,<br />

β1, . . . , βn be complex parameters and λn : X → M(n + 1, C) be a <strong>representation</strong><br />

<strong>of</strong> X = C{X, Y } defined <strong>as</strong><br />

⎛<br />

⎞<br />

0 1 0 · · ·<br />

⎜<br />

⎜0<br />

0 1 0 · · · ⎟<br />

⎜<br />

..<br />

⎟<br />

λn(X) = ⎜<br />

. .<br />

. ⎟ , (3.1)<br />

⎜<br />

0 0 0 · · · 1 0 ⎟<br />

⎝0<br />

0 0 · · · 0 1 ⎠<br />

0 (−1) n+1 sn (−1) n sn−1 · · · −s2 s1<br />

63


64 Chapter. 3<br />

⎛<br />

⎜<br />

λn(Y ) = ⎜<br />

⎝<br />

0 0 0 · · · 0 0<br />

. .<br />

.<br />

. .<br />

0 0 0 · · · 0 0<br />

tn+1 tn + (−1) n+1 sn tn−1 + (−1) n sn−1 · · · t2 − s2 t1 + s1<br />

⎞<br />

⎟<br />

⎠ ,<br />

(3.2)<br />

where si stands for <strong>the</strong> i-th elementary symmetric functions <strong>of</strong> α1, . . . , αn+1<br />

and ti <strong>the</strong> i-th elementary symmetric functions <strong>of</strong> p1 = (1 − β1), . . . , pn =<br />

(1 − βn). One can show that <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong><br />

(1.38) by λn is equivalent to <strong>the</strong> <strong>equation</strong><br />

(ϑ(ϑ − p1) · · · (ϑ − pn) − z(ϑ + α1) · · · (ϑ + αn+1)) f = 0, (3.3)<br />

where ϑ = z d is Euler operator. This <strong>equation</strong> is known <strong>as</strong> <strong>the</strong> generalized<br />

dz<br />

<strong>hypergeometric</strong> <strong>equation</strong> n+1En and h<strong>as</strong> <strong>the</strong> solution holomorphic at z = 0<br />

n+1Fn<br />

α1, . . . , αn+1<br />

β1, . . . , βn<br />

<br />

; z :=<br />

∞<br />

k=0<br />

(α1)k · · · (αn+1)k<br />

(β1)k · · · (βn)kk! zk<br />

(3.4)<br />

<strong>the</strong> generalized <strong>hypergeometric</strong> function. For n = 1, 2E1 is nothing but<br />

<strong>Gauss</strong> <strong>hypergeometric</strong> <strong>equation</strong> and 2F1( α1,α2 ; z) = F (α1, α2, β1; z) appeared<br />

β1<br />

in Section 1.3.1.<br />

Thus <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution λn(H0(z)) is <strong>the</strong> fundamental<br />

solution matrix <strong>of</strong> n+1En on a neighborhood <strong>of</strong> z = 0, applying our<br />

method introduced in previous sections, we can an iterated integral expression<br />

<strong>of</strong> <strong>the</strong> generalized <strong>hypergeometric</strong> <strong>equation</strong> and can expect to obtain<br />

various relations <strong>of</strong> MPLs and MZVs.<br />

On <strong>the</strong> o<strong>the</strong>r hand, Li([Li]) and Aoki-Kombu-Ohno([AKO]) extended <strong>the</strong><br />

result <strong>of</strong> Ohno-Zagier. On <strong>the</strong>ir works, <strong>the</strong> generalized <strong>hypergeometric</strong> function<br />

with certain special parameters is correlated with MZVs.<br />

For index k = (k1, . . . , kr), <strong>the</strong> i-height <strong>of</strong> k is defined <strong>as</strong> hi(k) = #{l|kl ><br />

i}. 1-height is particular a height defined in Section 1.3.2 and 2.5. Let<br />

G10(k, n, h1, . . . , hr; z) be a sum <strong>of</strong> MPLs whose weight is k, depth is n, iheight<br />

is hi (i = 1, . . . , r) and <strong>the</strong> first index is greater than 1. Li said that<br />

<strong>the</strong> generating function<br />

Φ0 = <br />

k≥n+ P r<br />

j=1 hj<br />

n≥h1≥...≥hr≥0<br />

h1≥1<br />

G10(k, n, h1, . . . , hr; z)x k−n−P r<br />

1<br />

j=1 hj<br />

x n−h1<br />

2 x h1−h2<br />

3<br />

· · · x hr−1−hr<br />

r+1 x hr<br />

r+2


3.1 Generalized <strong>hypergeometric</strong> <strong>equation</strong>s 65<br />

can be expressed by using <strong>the</strong> generalized <strong>hypergeometric</strong> function <strong>as</strong><br />

⎛<br />

⎛<br />

1 ⎜r−1<br />

Φ0 =<br />

⎜<br />

xr+2 − x1xr+1<br />

⎝ AjBjz<br />

j=0<br />

j ⎜ a1<br />

× ⎜<br />

+ j, a2 + j, . . . , ar+1 + j ⎟<br />

r+1Fr ⎝<br />

; z⎟<br />

j + 1 − x1, j + 1, . . . , j + 1 ⎠<br />

<br />

r−1 times<br />

(3.5)<br />

where a1 + · · · + ar+1 = x2 − x1, <br />

1≤i1≤i2≤···≤ij≤r+1 ai1 · · · aij = xj+1 − x1xj<br />

for j = 2, . . . , r + 1 and<br />

r−1<br />

<br />

i<br />

r − 1<br />

Aj = (xr+2−i − x1xr+1−i) + x1x2 ,<br />

j<br />

j<br />

Bj =<br />

i=j<br />

1<br />

(1 − x1)j(j!) r−1<br />

r−1<br />

i=1<br />

<br />

(ai)j.<br />

The notation { i<br />

} stands for <strong>the</strong> Stirling number <strong>of</strong> <strong>the</strong> second kind.<br />

j<br />

Aoki-Kombu-Ohno([AKO]) defined multiple zeta star value (MVSV, for<br />

short) <strong>as</strong><br />

ζ ∗ (k1, . . . , kr) =<br />

<br />

m1≥m2≥···≥mr≥1<br />

1<br />

m k1<br />

1 · · · m kr<br />

r<br />

and multiple star polylogarithm (MSPL, for short) <strong>as</strong><br />

Li ∗<br />

k1,...,kr (z) =<br />

<br />

m1≥m2≥···≥mr≥1<br />

z m1<br />

m k1<br />

1 · · · m kr<br />

r<br />

⎞<br />

(3.6)<br />

. (3.7)<br />

MVSVs can be expressed <strong>as</strong> Q-linear combinations <strong>of</strong> MZVs and vice versa.<br />

Let G ∗ 10(z) be a sum <strong>of</strong> MSPLs defined <strong>as</strong><br />

G ∗ 10(k, n, s; z) = Li ∗ (g10(k, n, s); z)<br />

and Φ ∗ 0(z) be <strong>the</strong> generating function <strong>of</strong> G ∗ 10(z), namely<br />

Φ ∗ 0(z) = <br />

k,n,s<br />

G ∗ 10(k, n, s; z)u k−n−s v n−s w 2s−2 .<br />

According to Aoki-Kombu-Ohno, Φ ∗ 0(z) is <strong>the</strong> unique solution to<br />

z 2 (1 − z) d2f df<br />

+ z((1 − z)(1 − u) − v)<br />

dz2 dz + (uv − w2 )f = z (3.8)<br />

vanishing at z = 0 and <strong>the</strong> value Φ∗ 0(1) can be written <strong>as</strong><br />

Φ ∗ 1<br />

0(1) =<br />

(1 − v)(1 − β) 3F2<br />

<br />

1 − β, 1 − β + u, 1<br />

; 1 , (3.9)<br />

2 − v, 2 − β<br />

− A0<br />

⎞<br />

⎟<br />

⎠ ,


66 Chapter. 3<br />

where β is a solution <strong>of</strong> <strong>the</strong> quadratic <strong>equation</strong> t 2 − (u + v)t + w 2 = 0.<br />

For Li and Aoki-Kombu-Ohno’s results, we suggest <strong>the</strong> following conjecture.<br />

Conjecture 24. Li and Aoki-Kombu-Ohno’s results can be interpreted through<br />

<strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> by λn and can be extend <strong>as</strong><br />

relations <strong>of</strong> MPLs.<br />

3.2 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized<br />

1<strong>KZ</strong> <strong>equation</strong> and MZVs<br />

In this section, we try to extend <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> to <strong>the</strong> c<strong>as</strong>e with<br />

many regular singular points on P 1 .<br />

3.2.1 The <strong>formal</strong> generalized <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variables<br />

Let a1, . . . , am be non-zero mutually different complex numbers. We extend<br />

<strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong> <strong>as</strong> follows: Let X, Y1, . . . , Ym be non-commutative<br />

<strong>formal</strong> elements. We call <strong>the</strong> <strong>equation</strong><br />

dG<br />

dz =<br />

<br />

X<br />

z +<br />

m<br />

i=1<br />

aiYi<br />

1 − aiz<br />

<br />

G (3.10)<br />

<strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong>. For n = 1, a1 = 1, it is <strong>the</strong> ordinary<br />

<strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> one variable (1.38). This <strong>equation</strong> is <strong>the</strong> universal<br />

Fuchsian differential <strong>equation</strong> which h<strong>as</strong> m + 2 regular singular points at<br />

z = 0, 1 1 , . . . , , ∞ referred to <strong>as</strong> <strong>the</strong> <strong>formal</strong> Schlesinger system.<br />

a1 am<br />

Algebraic properties <strong>of</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong> and its fundamental<br />

solutions can be constructed in <strong>the</strong> same way <strong>as</strong> <strong>the</strong> <strong>formal</strong> 1<strong>KZ</strong><br />

<strong>equation</strong>. For <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong>, <strong>the</strong> infinitesimal pure<br />

braid Lie algebra X = C{X, Y1, . . . , Yn} is <strong>the</strong> free Lie algebra generated by<br />

X, Y1, . . . , Ym, and <strong>the</strong> dual Hopf algebra <strong>of</strong> U(X) is <strong>the</strong> free shuffle algebra<br />

h = (C〈x, y1, . . . , ym〉, x, 1) generated by x = dz<br />

z , y1 = a1dz<br />

1−a1z , . . . , ym = amdz<br />

1−amz .<br />

We denote by h0 = C1 + hy1 + · · · + hym <strong>the</strong> x subalgebra <strong>of</strong> h generated by<br />

words which is not ended with x and define <strong>the</strong> hyperlogarithm<br />

L(x k1−1<br />

yi1 · · · x kr−1<br />

yir; z) = L( k1ai1 · · · krair; z) :=<br />

z<br />

0<br />

x k1−1<br />

yi1 · · · x kr−1<br />

yir,<br />

(3.11)


3.2 Rep. <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong> eq. 67<br />

where i1, . . . , ir ∈ {1, . . . , m}. The hyperlogarithm is a many-valued analytic<br />

function on P1 − {0, 1 1 , . . . , , ∞} and h<strong>as</strong> a Taylor expansion<br />

a1 am<br />

L( k1 ai1 · · · kr air; z) =<br />

<br />

n1>n2>···>nr>0<br />

a n1−n2a<br />

i1<br />

n2−n3<br />

i2<br />

n k1<br />

1 · · · n kr<br />

r<br />

· · · a nr<br />

ir<br />

z n1 (3.12)<br />

on |z| < min{ 1<br />

|ai | 1 , . . . , 1<br />

|air |}. Moreover it converges <strong>as</strong> z tends to 1 − 0 for<br />

k1 > 1 or k1 = 1, i1 = 1. If each ai’s are primitive m-th root <strong>of</strong> 1, <strong>the</strong> values<br />

<strong>of</strong> <strong>the</strong> limit <strong>as</strong> z tends to 1 are multiple L values([ArK]).<br />

We can also show that h = h0 [x] <strong>as</strong> x multiplication and can define<br />

reg0 : h → h0 <strong>as</strong> taking <strong>the</strong> constant term <strong>of</strong> this decomposition. We define<br />

an extended hyperlogarithm <strong>as</strong><br />

L(wx s ; z) =<br />

s<br />

i=0<br />

L(reg 0 (wx s−i ); z) logi z<br />

, (3.13)<br />

i!<br />

<strong>the</strong>n <strong>the</strong> generating function <strong>of</strong> hyperlogarithms<br />

H0(z) = <br />

L(w; z)W (3.14)<br />

w∈h word<br />

is <strong>the</strong> fundamental solution to (3.10) normalized at z = 0.<br />

3.2.2 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong><br />

<strong>equation</strong><br />

Let ρ : X → M(n, C) be a <strong>representation</strong> <strong>of</strong> X. The <strong>representation</strong> <strong>of</strong> <strong>the</strong><br />

<strong>formal</strong> generalized 1<strong>KZ</strong> <strong>equation</strong> by ρ<br />

dG<br />

dz =<br />

<br />

ρ(X)<br />

z +<br />

m<br />

<br />

aiρ(Yi)<br />

G (3.15)<br />

1 − aiz<br />

is a Fuchsian differential <strong>equation</strong> which h<strong>as</strong> regular singular points at z =<br />

0, 1<br />

a1<br />

i=1<br />

, . . . , 1<br />

am , ∞ (exceptionally if ρ(X) = 0 (resp, ρ(Yi) = 0, ρ(−X − Yi) =<br />

0), <strong>the</strong> <strong>equation</strong> is holomorphic at z = 0 (resp. z = 1<br />

ai<br />

, z = ∞)). Un-<br />

fortunately, in general ρ, <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized 1<strong>KZ</strong><br />

<strong>equation</strong> h<strong>as</strong> accessory parameters (that is, parameters which are not determined<br />

from <strong>as</strong>ymptotic behaviors) and <strong>the</strong> connection problem is very<br />

complicated.<br />

Assume that <strong>the</strong> <strong>representation</strong> ρ is irreducible; namely if a subspace<br />

V ⊂ C n satisfies ρ(X)(V ) ⊂ V, ρ(Yi)(V ) ⊂ V for all i, <strong>the</strong>n V = {0} or C n .


68 Chapter. 3<br />

Katz([Kat]) said that <strong>the</strong> <strong>equation</strong> (3.15) h<strong>as</strong> no accessory parameters (in<br />

this c<strong>as</strong>e, we say that <strong>the</strong> <strong>equation</strong> is rigid) if and only if <strong>the</strong> relation<br />

(2 − (m + 1))n 2 + <br />

i=0,...,m,∞<br />

dim Z(Ai) = 2 (3.16)<br />

holds, where A0 = exp(2πiρ(X)), Ai = exp(2πiρ(Yi)), A∞ <br />

= exp(2πiρ(−X −<br />

Yi)), and Z(A) stands for <strong>the</strong> centralizer <strong>of</strong> A, namely <strong>the</strong> set <strong>of</strong> all matrices<br />

in M(n, C) which commute to A.<br />

If a <strong>representation</strong> ρ is irreducible and <strong>the</strong> <strong>equation</strong> (3.15) is rigid, <strong>the</strong><br />

connection coefficients <strong>of</strong> solutions to <strong>the</strong> <strong>equation</strong> are explicitly given due to<br />

Oshima([Os]). Fur<strong>the</strong>rmore, Katz([Kat]), Yokoyama([Y]) and Oshima([Os])<br />

gave <strong>the</strong> method for construct all irreducible rigid Fuchsian <strong>equation</strong>s by two<br />

manipulations called ”middle convolution” and ”addition”.<br />

It is very interested problem to consider rigid and irreducible Fuchsian<br />

<strong>equation</strong>s through our viewpoint <strong>of</strong> a <strong>representation</strong>s <strong>of</strong> <strong>the</strong> <strong>formal</strong> generalized<br />

1<strong>KZ</strong> <strong>equation</strong>.<br />

3.3 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables<br />

In this section, we discuss <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variable (<strong>the</strong> <strong>formal</strong><br />

2<strong>KZ</strong> <strong>equation</strong>, for short) and its <strong>representation</strong>s. We can define <strong>the</strong> <strong>formal</strong><br />

2<strong>KZ</strong> <strong>equation</strong> via <strong>the</strong> <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on M0,5 in Section 1.2.2. However,<br />

in this c<strong>as</strong>e, unfortunately <strong>the</strong> infinitesimal pure braid Lie algebra X is not<br />

free, thus <strong>the</strong> dual Hopf algebra B h<strong>as</strong> complicated structure. We introduce<br />

an algebraic structure on <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> and its solutions according<br />

to [OU].<br />

3.3.1 The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> <strong>of</strong> two variables<br />

The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on <strong>the</strong> moduli space M0,5 (1.31) can be written <strong>as</strong><br />

dG = <br />

ζ1Z1 + ζ11Z11 + ζ2Z2 + ζ22Z22 + ζ12Z12 G (3.17)<br />

on <strong>the</strong> cubic coordinate (z1, z2) defined by (1.30), where Z1 = X12 + X13 +<br />

X23, Z2 = X23, Z11 = −X14, Z22 = −X12, Z12 = −X24 are <strong>formal</strong> elements<br />

in <strong>the</strong> infinitesimal pure braid Lie algebra X and ζ1 = dz1<br />

z1 , ζ11 = dz1<br />

1−z1 , ζ2 =<br />

dz2<br />

z2 , ζ22 = dz2<br />

1−z2 , ζ12 = d(z1z2)<br />

1−z1z2<br />

are differential 1-forms. This <strong>equation</strong> is equal to


3.3 The <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> 69<br />

<strong>the</strong> following partial differential system;<br />

∂<br />

G =<br />

∂z1<br />

∂<br />

G =<br />

∂z2<br />

Z1<br />

z1<br />

<br />

Z2<br />

z2<br />

+ Z11<br />

+<br />

1 − z1<br />

z2Z12<br />

1 − z1z2<br />

+ Z22<br />

+<br />

1 − z2<br />

z1Z12<br />

1 − z1z2<br />

<br />

G, (3.18)<br />

<br />

G. (3.19)<br />

The <strong>equation</strong> h<strong>as</strong> <strong>the</strong> divisors D = {z1 = 0, 1, ∞} ∪ {z2 = 0, 1, ∞} ∪ {z1z2 =<br />

1} on P 1 × P 1 and D are normal crossing at (z1, z2) = (0, 0), (1, 0), (0, 1).<br />

The infinitesimal pure braid relation (1.32) for Z1, . . . , Z12 is written <strong>as</strong><br />

[Z1, Z2] = [Z11, Z2] = [Z1, Z22] = 0,<br />

[Z11, Z22] = −[Z11, Z12] = [Z22, Z12] = −[Z1 − Z2, Z12], (3.20)<br />

and <strong>the</strong> Arnold relation (1.33) for ζ1, . . . , ζ12 <strong>as</strong><br />

ζ1 ∧ ζ11 = ζ2 ∧ ζ22 = 0,<br />

ζ22 ∧ ζ11 + ζ11 ∧ ζ12 + ζ12 ∧ (ζ22 + ζ2) = 0,<br />

(ζ1 + ζ2) ∧ ζ12 = 0. (3.21)<br />

The universal enveloping algebra U(X) h<strong>as</strong> <strong>the</strong> following tensor product<br />

decomposition.<br />

Lemma 25 ([OU]). The decomposition<br />

holds for {i1, i2} = {1, 2}, where X (ik)<br />

i1⊗i2<br />

U(X) = U(X (i1)<br />

) ⊗ U(X(i2)<br />

) (3.22)<br />

i1⊗i2 i1⊗i2<br />

stands for <strong>the</strong> Lie subalgebra<br />

X (i1)<br />

i1⊗i2 = C{Zi1, Zi1i1, Z12}, X (i2)<br />

i1⊗i2 = C{Zi2, Zi2i2} ({i1, i2} = {1, 2}).<br />

3.3.2 The reduced bar algebra and iterated integral <strong>of</strong><br />

two variables<br />

Since X is a quotient <strong>of</strong> <strong>the</strong> free Lie algebra, <strong>the</strong> dual Hopf algebra B <strong>of</strong> U(X)<br />

is a subalgebra <strong>of</strong> <strong>the</strong> free shuffle algebra generated by ζ1, . . . , ζ12. We call<br />

<strong>the</strong> subalgebra B <strong>the</strong> reduced bar algebra and give a direct definition <strong>of</strong> B <strong>as</strong><br />

follows.<br />

We denote by A = {ζ1, ζ11, ζ2, ζ22, ζ12} a set <strong>of</strong> letters and S(A) <strong>the</strong> free<br />

shuffle algebra generated by A. We define that an element<br />

ϕ = <br />

cI ωi1 ◦ · · · ◦ ωis ∈ S(A)<br />

I={i1,...,is}


70 Chapter. 3<br />

(where each ωiα ∈ A, CI ∈ C) satisfies Chen’s integrability condition([C1])<br />

if, for all l (1 ≤ l < s),<br />

<br />

cI ωi1 ⊗ · · · ⊗ ωil ∧ ωil+1 ⊗ · · · ⊗ ωis = 0 (3.23)<br />

I<br />

holds <strong>as</strong> a multiple differential form and define B <strong>as</strong> a subspace <strong>of</strong> S(A)<br />

spanned by elements satisfying Chen’s integrability condition. B h<strong>as</strong> <strong>the</strong><br />

grading B = ∞<br />

s=0 Bs, Bs = B ∩ Ss(A) <strong>as</strong> vector space with respect to <strong>the</strong><br />

length <strong>of</strong> words. We obtain clearly B0 = C1, B1 = Cζ1 ⊕ Cζ11 ⊕ Cζ2 ⊕<br />

Cζ22 ⊕ Cζ12 and B2 is a 19 dimensional vector space given by<br />

B2 = <br />

Cω ◦ ω ⊕ <br />

Cζi ◦ ζii ⊕ <br />

ω∈A<br />

⊕ <br />

ω1=ζ1,ζ11<br />

ω2=ζ2,ζ22<br />

i=1,2<br />

i=1,2<br />

Cζii ◦ ζi<br />

C(ω1 ◦ ω2 + ω2 ◦ ω1) ⊕ <br />

ω∈A−{ζ12}<br />

C(ω ◦ ζ12 + ζ12 ◦ ω)<br />

⊕ C(ζ1 ◦ ζ12 + ζ2 ◦ ζ12) ⊕ C(ζ11 ◦ ζ12 + ζ22 ◦ ζ11 − ζ22 ◦ ζ12 − ζ2 ◦ ζ12)<br />

(3.24)<br />

by virtue <strong>of</strong> <strong>the</strong> AR (3.21). Fur<strong>the</strong>rmore we have<br />

Bs =<br />

s−1 <br />

j=1<br />

Bj ◦ Bs−j =<br />

s−2 <br />

j=0<br />

B1 ◦ · · · ◦ B1<br />

<br />

j times<br />

◦B2 ◦ B1 ◦ · · · ◦ B1<br />

<br />

s−j−2 times<br />

(3.25)<br />

for s > 2 due to Brown([B]). One can show that B is a Hopf subalgebra <strong>of</strong><br />

S(A) and is <strong>the</strong> dual Hopf algebra <strong>of</strong> U(X).<br />

According to Chen([C2]), for any element ϕ ∈ B, <strong>the</strong> value <strong>of</strong> <strong>the</strong> iterated<br />

integral <strong>of</strong> ϕ depends only on <strong>the</strong> homotopy cl<strong>as</strong>s <strong>of</strong> <strong>the</strong> integral contour.<br />

Thus we can define <strong>the</strong> iterated integral <strong>of</strong> two variable<br />

(z1,z2)<br />

ϕ <strong>as</strong> U(X)-<br />

(z0 1 ,z0 2 )<br />

valued many-valued analytic function on P1 × P1 − D for ϕ ∈ B. Especially<br />

(z1,z2)<br />

if ϕ ∈ B h<strong>as</strong> no terms terminated by ζ1 and ζ2, <strong>the</strong> iterated integral<br />

can be defined.<br />

We denote by S 0 (A) <strong>the</strong> subspace <strong>of</strong> S(A) spanned by elements which<br />

have no terms terminated by ζ1 and ζ2, and B 0 = B ∩ S 0 (A) <strong>the</strong> subspace <strong>of</strong><br />

B. Clearly S 0 (A) (resp. B 0 ) is a x-subalgebra <strong>of</strong> S(A) (resp. B).<br />

In what follows, we <strong>as</strong>sume that 0 < |z1|, |z2| < 1 and consider <strong>the</strong> following<br />

two contours C1⊗2, C2⊗1 : (0, 0) → (z1, z2).<br />

(0,0)<br />

ϕ


3.3 The <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> 71<br />

z2<br />

(0, 1) (1, 1)<br />

C (1)<br />

C1⊗2 = C<br />

1⊗2<br />

(1)<br />

1⊗2 ◦ C (2)<br />

1⊗2<br />

✻<br />

C (2)<br />

1⊗2<br />

C (1)<br />

2⊗1<br />

(z1, z2)<br />

✲<br />

✻<br />

C (2)<br />

2⊗1<br />

C2⊗1 = C (2)<br />

✲<br />

(0, 0) (1, 0)<br />

2⊗1 ◦ C (1)<br />

2⊗1<br />

C1⊗2 = C (1)<br />

1⊗2 ◦ C (2)<br />

1⊗2,<br />

C (2)<br />

1⊗2 : (0, 0) → (0, z2),<br />

C (1)<br />

1⊗2 : (0, z2) → (z1, z2).<br />

C2⊗1 = C (2)<br />

2⊗1 ◦ C (1)<br />

2⊗1,<br />

C (1)<br />

2⊗1 : (0, 0) → (z1, 0),<br />

C (2)<br />

2⊗1 : (z1, 0) → (z1, z2).<br />

3.3.3 The normalized fundamental solution to <strong>the</strong> <strong>formal</strong><br />

2<strong>KZ</strong> <strong>equation</strong><br />

Let L(z1, z2) be a solution to <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> (3.17). We call<br />

L(z1, z2) <strong>the</strong> fundamental solution normalized at (0, 0) if L(z1, z2) can be<br />

written <strong>as</strong><br />

z1<br />

L(z1, z2) = ˆ L(z1, z2)z Z1<br />

1 z Z2<br />

2 , (3.26)<br />

where ˆ L(z1, z2) is a U(X)-valued many-valued analytic function holomorphic<br />

on a neighborhood <strong>of</strong> (0, 0) and ˆ L(0, 0) = I.<br />

Proposition 26 ([OU]). The fundamental solution normalized at (0, 0) to<br />

<strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> (3.17) exists uniquely and expressed <strong>as</strong><br />

L(z1, z2) = ˆ L(z1, z2)z Z1<br />

1 z Z2<br />

2 , ˆ L(z1, z2) =<br />

ˆLs(z1, z2) =<br />

(z1,z2)<br />

(0,0)<br />

∞<br />

s=0<br />

ˆLs(z1, z2),<br />

ad(Ω0) + µ(Ω ′ ) s (1 ⊗ I), (3.27)<br />

where Ω0 = ζ1Z1 + ζ2Z2, Ω ′ = ζ11Z11 + ζ22Z22 + ζ12Z12, and <strong>the</strong> action ad<br />

and µ on S(A) ⊗ U(X) stand for<br />

ad(ω ⊗ X)(ϕ ⊗ F ) = (ω ◦ ϕ) ⊗ [X, F ],<br />

µ(ω ⊗ X)(ϕ ⊗ F ) = (ω ◦ ϕ) ⊗ XF


72 Chapter. 3<br />

for ϕ ⊗ F ∈ S(A) ⊗ U(X), ω ⊗ X ∈ B1 ⊗ X. Fur<strong>the</strong>rmore for any s ∈ Z≥0,<br />

ad(Ω0) + µ(Ω ′ ) s (1 ⊗ I) belongs to B 0 ⊗ U(X).<br />

The iterated integral on (3.27) can be calculate <strong>as</strong> follows. Let A (1)<br />

1⊗2 =<br />

{ζ1, ζ11, ζ (1)<br />

12 }, A (2)<br />

1⊗2 = {ζ2, ζ22}, A (2)<br />

2⊗1 = {ζ2, ζ22, ζ (2)<br />

12 } and A (1)<br />

2⊗1 = {ζ1, ζ11} be<br />

sets <strong>of</strong> letters and regard ζ (i)<br />

12 <strong>as</strong> <strong>the</strong> 1-form<br />

Proposition 27 ([OU]).<br />

ˆL(z1, z2)<br />

=<br />

<br />

=<br />

W ′ ∈W0 (X (1)<br />

1⊗2 )<br />

W ′′ ∈W0 (X (2)<br />

1⊗2 )<br />

<br />

W ′ ∈W 0 (X (2)<br />

2⊗1 )<br />

W ′′ ∈W 0 (X (1)<br />

2⊗1 )<br />

Here W 0 (X (ik)<br />

i1⊗i2<br />

ζ (1)<br />

12 = z2dz1<br />

, ζ<br />

1 − z1z2<br />

(2)<br />

12 = z1dz2<br />

. (3.28)<br />

1 − z1z2<br />

L(θ (1)<br />

1⊗2(W ′ ); z1)L(θ (2)<br />

1⊗2(W ′′ ); z2) α(W ′ )α(W ′′ )(I), (3.29)<br />

L(θ (2)<br />

2⊗1(W ′ ); z2)L(θ (1)<br />

2⊗1(W ′′ ); z1) α(W ′ )α(W ′′ )(I). (3.30)<br />

) is <strong>the</strong> set <strong>of</strong> all words <strong>of</strong> U(X(ik)<br />

) which does not ended with<br />

i1⊗i2<br />

Z1 and Z2, α stands for <strong>the</strong> algebra homomorphism α : U(X) → End(U(X))<br />

defined by<br />

α : (Z1, Z11, Z2, Z22, Z12) ↦→ (ad(Z1), µ(Z11), ad(Z2), µ(Z22), µ(Z12))<br />

and θ (ik)<br />

i1⊗i2<br />

: U(X(ik)<br />

) → S(A(ik)<br />

) is a map which gives <strong>the</strong> duality; it is<br />

i1⊗i2 i1⊗i2<br />

defined by <strong>the</strong> replacement <strong>of</strong> letters<br />

θ (ik)<br />

(Zik ) = ζik , θ(ik)<br />

(Zikik ) = ζikik , θ(i1)<br />

i1⊗i2 i1⊗i2 i1⊗i2 (Z12) = ζ (i1)<br />

12 .<br />

We note that L(θ (1)<br />

1⊗2(W ′ ); z1) (resp. L(θ (2)<br />

2⊗1(W ′ ); z2)) is a hyperlogarithm<br />

defined by Section 3.2.1 for m = 2, z = z1, a1 = 1, a2 = z2, x = ζ1, ξ1 =<br />

ζ11, ξ2 = ζ (1)<br />

12 (resp. m = 2, z = z2, a1 = 1, a2 = z1, x = ζ2, ξ1 = ζ22, ξ2 =<br />

ζ (2)<br />

12 ) and L(θ (2)<br />

1⊗2(W ′′ ); z2) (resp. L(θ (1)<br />

2⊗1(W ′′ ); z2)) is a multiple polylogarithm<br />

<strong>of</strong> one variable z2 (resp. z1).


3.3 The <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> 73<br />

3.3.4 Decomposition <strong>the</strong>orem <strong>of</strong> <strong>the</strong> normalized fundamental<br />

solution and <strong>the</strong> generalized harmonic<br />

product relation<br />

We consider <strong>the</strong> following four <strong>formal</strong> (generalized) 1<strong>KZ</strong> <strong>equation</strong>.<br />

dz1G(z1, z2) = Ω (1)<br />

1⊗2G(z1, z2), Ω (1)<br />

1⊗2 = ζ1Z1 + ζ11Z11 + ζ (1)<br />

12 Z12, (3.31)<br />

dz2G(z2) = Ω (2)<br />

1⊗2G(z2), Ω (2)<br />

1⊗2 = ζ2Z2 + ζ22Z22, (3.32)<br />

dz2G(z1, z2) = Ω (2)<br />

2⊗1G(z1, z2), Ω (2)<br />

2⊗1 = ζ2Z2 + ζ22Z22 + ζ (2)<br />

12 Z12, (3.33)<br />

dz1G(z1) = Ω (1)<br />

2⊗1G(z1), Ω (1)<br />

2⊗1 = ζ1Z1 + ζ11Z11, (3.34)<br />

where dz1 (resp. dz2) stands for <strong>the</strong> exterior differentiation by <strong>the</strong> variable<br />

z1 (resp. z2). The fundamental solution normalized at <strong>the</strong> origin to each<br />

<strong>equation</strong>s L (1)<br />

1⊗2, L (2)<br />

1⊗2, L (2)<br />

2⊗1, L (1)<br />

2⊗1 satisfies <strong>the</strong> conditions<br />

ˆL (ik)<br />

i1⊗i2 =<br />

∞<br />

L (ik)<br />

i1⊗i2 = ˆ L (ik)<br />

i1⊗i2 zZi k<br />

ik ,<br />

<br />

<br />

ˆL<br />

s=0<br />

(ik)<br />

i1⊗i2,s , L ˆ(ik) i1⊗i2,s<br />

zi =0<br />

k<br />

= 0 (s > 0), L ˆ(ik) i1⊗i2,0 = I.<br />

Proposition 28. The fundamental solution L(z1, z2) to (3.17) normalized<br />

at <strong>the</strong> origin decomposes to product <strong>of</strong> <strong>the</strong> normalized fundamental solutions<br />

<strong>of</strong> <strong>the</strong> (generalized) <strong>formal</strong> 1<strong>KZ</strong> <strong>equation</strong>s <strong>as</strong> follows:<br />

L(z1, z2) = L (1)<br />

1⊗2L (2)<br />

1⊗2 = ˆ L (1)<br />

1⊗2 ˆ L (2)<br />

1⊗2z Z1<br />

1 z Z2<br />

2<br />

= L (2)<br />

2⊗1L (1)<br />

2⊗1 = ˆ L (2)<br />

2⊗1 ˆ L (1)<br />

2⊗1z Z1<br />

1 z Z2<br />

2 .<br />

The expressions (3.29) and (3.30) mean that each decomposition in Proposition<br />

28 corresponds to <strong>the</strong> choice <strong>of</strong> <strong>the</strong> integral contours C1⊗2, C2⊗1. Fur<strong>the</strong>rmore,<br />

for <strong>the</strong> connection problem <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong>, we obtain<br />

Proposition 29. The connection formula between <strong>the</strong> fundamental solutions<br />

<strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> (3.17) normalized at (z1, z2) = (0, 0) and<br />

(1, 0), (0, 1) is given <strong>as</strong><br />

L (0,0) (z1, z2) = L (1,0) (z1, z2)Φ<strong>KZ</strong>(Z1, Z11), (3.35)<br />

L (0,0) (z1, z2) = L (0,1) (z1, z2)Φ<strong>KZ</strong>(Z2, Z22). (3.36)<br />

In addition, [OU] claimed <strong>the</strong> followings:


74 Chapter. 3<br />

Proposition 30 ([OU]). We define linear maps ι1⊗2 : B → S(A (1)<br />

1⊗2) ⊗<br />

S(A (2)<br />

1⊗2) by<br />

ιi1⊗i2 =<br />

<br />

Pr (1)<br />

1⊗2<br />

<br />

<br />

⊗ Pr<br />

B<br />

(2)<br />

<br />

<br />

1⊗2<br />

◦<br />

B<br />

¯ ∆, (3.37)<br />

where ¯ ∆ is <strong>the</strong> coproduct <strong>of</strong> <strong>the</strong> Hopf algebra S(A) (1.9) and Pr (1)<br />

1⊗2 : S(A) →<br />

S(A (1)<br />

1⊗2) (resp. Pr (1)<br />

1⊗2 : S(A) → S(A (2)<br />

1⊗2)) stands for a projection defined by<br />

(ζ1, ζ11, ζ2, ζ22, ζ12) ↦→ (ζ1, ζ11, 0, 0, ζ (1)<br />

12 ) (resp. (0, 0, ζ2, ζ22, 0)). Then<br />

and<br />

ι1⊗2 : B → S(A (1)<br />

1⊗2) ⊗ S(A (2)<br />

1⊗2)<br />

ι1⊗2| B 0 : B 0 → S 0 (A (1)<br />

1⊗2) ⊗ S 0 (A (2)<br />

1⊗2)<br />

are both x algebra isomorphism. We can also define <strong>the</strong> x isomorphism<br />

ι2⊗1 : B → A (2)<br />

2⊗1 ⊗ A (1)<br />

2⊗1 in <strong>the</strong> same way. Fur<strong>the</strong>rmore, <strong>the</strong> decomposition<br />

holds <strong>as</strong> x multiplication.<br />

B = B 0 [ζ1, ζ2] (3.38)<br />

For ψ1⊗ψ2 ∈ S0 (A (i1)<br />

i1⊗i2 )⊗S0 (A (i2)<br />

), we define <strong>the</strong> integral<br />

i1⊗i2<br />

by <br />

Ci 1 ⊗i 2<br />

ψ1 ⊗ ψ2 :=<br />

zi1 zi2 ψ1<br />

zi =0 1<br />

zi =0 2<br />

ψ2.<br />

<br />

Ci 1 ⊗i 2<br />

ψ1⊗ψ2<br />

The map ι1⊗2 (resp. ι2⊗1) picks up <strong>the</strong> terms <strong>of</strong> B 0 whose iterated integral<br />

along C1⊗2 (resp. C2⊗1) does not vanish, namely<br />

for ϕ ∈ B 0 .<br />

(z1,z2)<br />

(0,0)<br />

<br />

ϕ =<br />

C1⊗2<br />

Proposition 31 ([OU]). Putting<br />

<br />

ι1⊗2(ϕ) =<br />

C2⊗1<br />

ϕ(W ′ , W ′′ ) = ι −1<br />

1⊗2(θ (1)<br />

1⊗2(W ′ ) ⊗ θ (2)<br />

1⊗2(W ′′ )) ∈ B 0<br />

for W ′ ∈ W 0 (X (1)<br />

1⊗2), W ′′ ∈ W 0 (X (2)<br />

1⊗2), we have<br />

<br />

ι2⊗1(ϕ) (3.39)<br />

ι1⊗2(ϕ(W<br />

C1⊗2<br />

′ , W ′′ )) = L(θ (1)<br />

1⊗2(W ′ ); z1)L(θ (2)<br />

1⊗2(W ′′ ); z2), (3.40)


3.3 The <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> 75<br />

and<br />

ˆL(z1, z2) =<br />

Fur<strong>the</strong>rmore <strong>the</strong> <strong>equation</strong><br />

<br />

W ′ ∈W0 (X (1)<br />

1⊗2 )<br />

W ′′ ∈W0 (X (2)<br />

1⊗2 )<br />

(z1,z2)<br />

ϕ(W<br />

(0,0)<br />

′ , W ′′ ) α(W ′ )α(W ′′ )(I). (3.41)<br />

L(θ (1)<br />

1⊗2(W ′ ); z1)L(θ (2)<br />

1⊗2(W ′′ ); z2) =<br />

for each W ′ ∈ W0 (X (1)<br />

1⊗2), W ′′ ∈ W0 (X (2)<br />

<br />

C2⊗1<br />

ι2⊗1(ϕ(W ′ , W ′′ )) (3.42)<br />

1⊗2) yields <strong>the</strong> functional relation<br />

among hyperlogarithms referred to <strong>as</strong> <strong>the</strong> generalized harmonic product <strong>of</strong><br />

hyperlogarithm. The relations are equivalent to <strong>the</strong> relations derived from<br />

comparing coefficients <strong>of</strong> each element <strong>of</strong> U(X) on (3.29) and (3.30).<br />

These relations contains <strong>the</strong> harmonic product <strong>of</strong> MPLs which is <strong>the</strong><br />

relation expressing a product <strong>of</strong> MPLs <strong>as</strong> sum <strong>of</strong> MPLs <strong>of</strong> two variables by<br />

using series expression, for instance,<br />

Lik1(z1) Lil1(z2) = z<br />

m>0<br />

m 1<br />

mk1 z<br />

n>0<br />

n <br />

<br />

2<br />

= + l1 n<br />

m>n>0<br />

<br />

+<br />

m=n>0<br />

<br />

<br />

z<br />

n>m>0<br />

m 1 zn 2<br />

mk1nl1 = Lik1,l1(1, 1; z1, z2) + Lik1+l1(z1z2) + Lik1,l1(1, 1; z2, z1),<br />

where MPLs <strong>of</strong> two variables Lik1,...,ki+j (i, j; z1, z2) is a special c<strong>as</strong>e <strong>of</strong> hyperlogarithm<br />

defined by<br />

Lik1,...,ki+j (i, j; z1, z2) := L( k1 1 · · · ki 1 ki+1 z2 · · · ki+j z2; z1) (3.43)<br />

=<br />

<br />

n1>n2>···>nr>0<br />

z n1<br />

1 z ni+1<br />

2<br />

n k1<br />

1 · · · n kr<br />

r<br />

Taking limits <strong>of</strong> <strong>the</strong>m <strong>as</strong> z tends to 1, we obtain <strong>the</strong> harmonic product<br />

<strong>of</strong> MZVs (1.25). Thus <strong>the</strong> harmonic product <strong>of</strong> MZVs can be interpreted <strong>as</strong><br />

<strong>the</strong> connection problem <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong>.<br />

3.3.5 Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong><br />

Now we can consider a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> and its<br />

fundamental solution. Let ρ : X → M(n, C) be a <strong>representation</strong> <strong>of</strong> <strong>the</strong><br />

infinitesimal pure braid Lie algebra X. We call Pfaffian system<br />

dG =<br />

<br />

ρ(Z1) dz1<br />

z1<br />

+ ρ(Z11) dz1<br />

+ ρ(Z2)<br />

1 − z1<br />

dz2<br />

+ ρ(Z22)<br />

z2<br />

dz2<br />

+ ρ(Z12)<br />

1 − z2<br />

d(z1z2)<br />

.<br />

1 − z1z2<br />

(3.44)


76 Chapter. 3<br />

a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> and <strong>the</strong> <strong>formal</strong> sum<br />

ρ(L(z1, z2)) (3.45)<br />

a <strong>representation</strong> <strong>of</strong> <strong>the</strong> fundamental solution normalized at origin. A <strong>representation</strong><br />

<strong>of</strong> <strong>the</strong> fundamental solution can be expressed <strong>as</strong> a <strong>formal</strong> sum whose<br />

coefficients are product <strong>of</strong> hyperlogarithms and multiple polylogarithms. If<br />

it converges absolutely, it is a fundamental solution matrix <strong>of</strong> (3.44).<br />

Representations <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> contain <strong>the</strong> differential <strong>equation</strong>s<br />

satisfied by Appell <strong>hypergeometric</strong> function F1, F2 and F4 ([Ko],[Ka]).<br />

Indeed, let µ1 be a <strong>representation</strong> <strong>of</strong> X defined <strong>as</strong><br />

⎛<br />

⎞<br />

⎛<br />

⎞<br />

0 1 1<br />

0 0 0<br />

µ1(Z1) = ⎝0<br />

1 − γ 0 ⎠ , µ1(Z11) = ⎝αβ<br />

α + β + 1 − γ β⎠<br />

,<br />

0 0 1 − γ<br />

0 0 0<br />

⎛<br />

⎞<br />

⎛<br />

⎞<br />

0 0 1<br />

0 0 0<br />

µ1(Z2) = ⎝0<br />

0 −β ⎠ , µ1(Z22) = ⎝0<br />

β<br />

0 0 β + 1 − γ<br />

′ −β<br />

0 −β ′ ⎠ ,<br />

⎛<br />

⎞<br />

β<br />

µ1(Z12) = ⎝<br />

0 0 0<br />

0 0 0<br />

αβ ′ β ′ α + β + 1 − γ<br />

⎠ .<br />

Then <strong>the</strong> Appell <strong>hypergeometric</strong> function<br />

F1(α, β, β ′ ∞<br />

, γ; u, v) :=<br />

m,n=0<br />

(α)m+n(β)m(β ′ )n<br />

u<br />

(γ)m+nm!n!<br />

m v n<br />

(3.46)<br />

satisfies <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> (3.17) by µ1 through<br />

<strong>the</strong> blowing up (u, v) = (z1, z1z2).<br />

For o<strong>the</strong>r Appell <strong>hypergeometric</strong> function<br />

F2(α, β, β ′ , γ, γ ′ ; u, v) :=<br />

F4(α, β, γ, γ ′ ; u, v) :=<br />

∞<br />

m,n=0<br />

∞<br />

m,n=0<br />

(α)m+n(β)m(β ′ )n<br />

(γ)m(γ ′ )nm!n! um v n<br />

(α)m+n(β)m+n<br />

(γ)m(γ ′ )nm!n! um v n<br />

<strong>the</strong> correspondence <strong>representation</strong>s µ2, µ4 are given <strong>as</strong> follows.<br />

⎛<br />

−β − β<br />

⎜<br />

µ2(Z1) = ⎜<br />

⎝<br />

′ ⎞<br />

0 1 0<br />

⎟<br />

⎠ ,<br />

−βδ −β ′ + δ 0 0<br />

−β ′ δ ′ 0 −β + δ ′ 0<br />

0 −β ′ δ ′ −βδ −α + 1<br />

(u, v) = (z, zw) ,<br />

(3.47)<br />

(u, v) = (z 2 1w2, (1 − z1)(1 − z1z2)) ,<br />

(3.48)


3.3 The <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> 77<br />

⎛<br />

0 1 0 0<br />

⎜<br />

µ2(Z11) = ⎜0<br />

−β<br />

⎝<br />

′ ⎞<br />

+ α + δ − 1 0 0 ⎟<br />

0 0 0 0⎠<br />

, µ2(Z2)<br />

⎛<br />

−β<br />

⎜<br />

= ⎜<br />

⎝<br />

′ 0 1 0<br />

0 −β ′ ⎞<br />

0 1 ⎟<br />

⎠ ,<br />

0 −β ′ δ ′ 0 0<br />

−β ′ δ ′ 0 δ ′ 0<br />

0 −β ′ δ ′ 0 δ ′<br />

⎛<br />

0<br />

⎜<br />

µ2(Z22) = ⎜0<br />

⎝0<br />

0<br />

0<br />

0<br />

0<br />

0<br />

0<br />

0<br />

1<br />

1<br />

0 0 0 α + δ + δ ′ ⎞<br />

⎟<br />

⎠<br />

− 1<br />

, µ2(Z12)<br />

⎛<br />

0<br />

⎜<br />

= ⎜0<br />

⎝0<br />

0<br />

0<br />

0<br />

1<br />

0<br />

−β + α + δ<br />

0<br />

0<br />

′ − 1<br />

⎞<br />

⎟<br />

0⎠<br />

0 0 −βδ 0<br />

,<br />

⎛<br />

0<br />

⎜<br />

µ4(Z1) = ⎜0<br />

⎝0<br />

1 1<br />

1 − γ + ε 0<br />

0 1 − γ + ε<br />

0<br />

1<br />

1<br />

⎞<br />

⎟<br />

⎠<br />

0 0 0 2(1 − γ)<br />

,<br />

⎛<br />

0<br />

⎜<br />

µ4(Z11) = ⎜αβ<br />

⎝<br />

0<br />

γ<br />

0 0<br />

′ 0 0<br />

ε<br />

0<br />

0<br />

0<br />

0 0 (α + ε)(β + ε) γ ′<br />

⎞<br />

⎟<br />

⎠ , µ4(Z2)<br />

⎛<br />

0<br />

⎜<br />

= ⎜0<br />

⎝0<br />

0<br />

0<br />

0<br />

1<br />

ε<br />

1 − γ<br />

0<br />

1<br />

0<br />

⎞<br />

⎟<br />

⎠<br />

0 0 0 1 − γ<br />

,<br />

⎛<br />

0 0<br />

⎜<br />

µ4(Z22) = ⎜0<br />

−ε<br />

⎝0<br />

ε<br />

⎞<br />

0 0<br />

ε 0 ⎟<br />

−ε 0⎠<br />

0 0 0 0<br />

, µ4(Z12)<br />

⎛<br />

0<br />

⎜<br />

= ⎜ 0<br />

⎝αβ<br />

0<br />

0<br />

−ε<br />

0 0<br />

0 0<br />

−γ ′ 0<br />

0 (α + ε)(β + ε) 0 γ ′<br />

⎞<br />

⎟<br />

⎠ ,<br />

where δ = β + 1 − γ, δ ′ = β ′ + 1 − γ ′ and ε = γ + γ ′ − α − β − 1. We remark<br />

that <strong>the</strong> <strong>equation</strong> satisfied by Appell <strong>hypergeometric</strong> function<br />

F3(α, α ′ , β, β ′ , γ; u, v) :=<br />

∞<br />

m,n=0<br />

(α)m(α ′ )n(β)m(β ′ )n<br />

u<br />

(γ)m+nm!n!<br />

m v n<br />

(3.49)<br />

also can be expressed <strong>as</strong> a <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong>, however<br />

<strong>the</strong> expression is too complicated to write concretely.<br />

We call <strong>the</strong> <strong>representation</strong> <strong>of</strong> <strong>the</strong> <strong>formal</strong> 2<strong>KZ</strong> <strong>equation</strong> (3.17) by µi (i =<br />

1, 2, 4) Appell <strong>hypergeometric</strong> <strong>equation</strong> <strong>of</strong> type i. Proposition 27 and 29 give<br />

us <strong>the</strong> way to calculate <strong>the</strong> iterated integral expressions <strong>of</strong> Appell <strong>hypergeometric</strong><br />

functions and we can discuss connection problems and relationship<br />

to MZVs. Calculating <strong>the</strong> iterated integral expressions and considering correspondence<br />

relations <strong>of</strong> hyperlogarithms are issues in <strong>the</strong> future.


82 Bibliography<br />

[Oh] Y. Ohno, A generalization <strong>of</strong> <strong>the</strong> duality and sum formul<strong>as</strong> on<br />

<strong>the</strong> multiple zeta values, J. Number Theory 74 (1999), no. 1,<br />

39–43.<br />

[Ok] J. Okuda, Duality formul<strong>as</strong> <strong>of</strong> <strong>the</strong> special values <strong>of</strong> multiple<br />

polylogarithms, Bull. London Math. Soc. 37 (2005), no. 2, 230–<br />

242.<br />

[OkU] J. Okuda and K. Ueno, The sum formula for multiple zeta values<br />

and connection problem <strong>of</strong> <strong>the</strong> <strong>formal</strong> Knizhnik-Zamolodchikov<br />

<strong>equation</strong>, Zeta Functions, Topology and Quantum Physics, 145–<br />

170, Dev. Math., 14, Springer, New York, 2005.<br />

[OkU2] J. Okuda and K. Ueno, Relations for multiple zeta values and<br />

Mellin transforms <strong>of</strong> multiple polylogarithms, Publ. Res. Inst.<br />

Math. Sci. 40 (2004), no. 2, 537–564.<br />

[Os] T. Oshima, Cl<strong>as</strong>sification <strong>of</strong> Fuchsian systems and <strong>the</strong>ir connection<br />

problem, preprint (2008), arXiv:math.CA/0811.2916.<br />

[OS] P. Orlik and L. Solomon, Combinatorics and topology <strong>of</strong> complements<br />

<strong>of</strong> hyperplanes, Invent. Math. 56 (1980), no. 2, 167–189.<br />

[OT] P. Orlik and H. Terao, Arrangements <strong>of</strong> Hyperplanes,<br />

Grundlehren der Ma<strong>the</strong>matischen Wissenschaften, 300,<br />

Springer-Verlag, Berlin, 1992.<br />

[OU] S. Oi and K. Ueno, The <strong>formal</strong> <strong>KZ</strong> <strong>equation</strong> on <strong>the</strong> moduli space<br />

M0,5 and <strong>the</strong> harmonic product <strong>of</strong> multiple zeta values, preprint<br />

(2009), arXiv:math.QA/0910.0718.<br />

[OZ] Y. Ohno and D. Zagier, Multiple zeta values <strong>of</strong> fixed weight,<br />

depth, and height Indag. Math. (N.S.) 12 (2001), no. 4, 483–<br />

487.<br />

[R] G. Racinet, Doubles mélanges des polylogarithmes multiples aux<br />

racines de l’unité, Publ. Math. Inst. Hautes Études Sci. No. 95<br />

(2002), 185–231.<br />

[Re] C. Reutenauer, Free Lie Algebr<strong>as</strong>, London Ma<strong>the</strong>matical Society<br />

Monographs. New Series, 7. Oxford Science Publications. The<br />

Clarendon Press, Oxford University Press, New York, 1993.<br />

[T1] T. Ter<strong>as</strong>oma, Selberg integrals and multiple zeta values, Compositio<br />

Math. 133 (2002), no. 1, 1–24.


[T2] T. Ter<strong>as</strong>oma, Mixed Tate motives and multiple zeta values, Invent.<br />

Math. 149 (2002), no. 2, 339–369.<br />

[W] Z. Wojtkowiak, Monodromy <strong>of</strong> iterated integrals and nonabelian<br />

unipotent periods, Geometric Galois actions, 2, 219–<br />

289, London Math. Soc. Lecture Note Ser., 243, Cambridge<br />

Univ. Press, Cambridge, 1997.<br />

[WW] E.T. Whittaker and G.N. Watson, A course <strong>of</strong> modern analysis,<br />

Reprint <strong>of</strong> <strong>the</strong> fourth (1927) edition, Cambridge Ma<strong>the</strong>matical<br />

Library, Cambridge Univ. Press, Cambridge, 1996.<br />

[Y] T. Yokoyama, Construction <strong>of</strong> systems <strong>of</strong> differential <strong>equation</strong>s<br />

<strong>of</strong> Okubo normal form with rigid monodromy, Math. Nachr.,<br />

279 (2006), 327–348.<br />

[Z1] D. Zagier, Values <strong>of</strong> zeta functions and <strong>the</strong>ir applications, First<br />

European Congress <strong>of</strong> Ma<strong>the</strong>matics, Vol. II (Paris, 1992), 497–<br />

512, Progr. Math., 120, Birkhäuser, B<strong>as</strong>el, 1994.<br />

[Z2] D. Zagier, Multiple zeta values, preprint.<br />

83

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!