1: %\documentstyle{article}
2: \documentclass[11pt]{article}
3: \textwidth 160mm
4: \textheight 220mm
5: \topmargin -5mm
6: \oddsidemargin -10mm
7: \input amssymb.sty
8: \input amssym.def
9: \input amssym
10: \input epsf
11:
12:
13:
14: \newtheorem{theorem}{Theorem}[section]
15: \newtheorem{lemma}[theorem]{Lemma}
16: \newtheorem{proposition}[theorem]{Proposition}
17: \newtheorem{corollary}[theorem]{Corollary}
18: \newtheorem{conjecture}[theorem]{Conjecture}
19: \newtheorem{definition}[theorem]{Definition}
20: \newtheorem{remark}[theorem]{Remark}
21: \newtheorem{example}[theorem]{Example}
22:
23: \begin{document}
24: \newcommand{\be}{\begin{equation}}
25: \newcommand{\ee}{\end{equation}}
26: \newcommand{\bt}{\begin{theorem}}
27: \newcommand{\et}{\end{theorem}}
28: \newcommand{\bd}{\begin{definition}}
29: \newcommand{\ed}{\end{definition}}
30: \newcommand{\bp}{\begin{proposition}}
31: \newcommand{\ep}{\end{proposition}}
32: \newcommand{\bl}{\begin{lemma}}
33: \newcommand{\el}{\end{lemma}}
34: \newcommand{\bc}{\begin{corollary}}
35: \newcommand{\ec}{\end{corollary}}
36: \newcommand{\bcon}{\begin{conjecture}}
37: \newcommand{\econ}{\end{conjecture}}
38: \newcommand{\la}{\label}
39: \newcommand{\Z}{{\Bbb Z}}
40: \newcommand{\R}{{\Bbb R}}
41: \newcommand{\Q}{{\Bbb Q}}
42: \newcommand{\C}{{\Bbb C}}
43: \newcommand{\hra}{\hookrightarrow}
44: \newcommand{\lra}{\longrightarrow}
45: \newcommand{\lms}{\longmapsto}
46:
47: \begin{titlepage}
48: %\title{Feynman integrals and Hodge rational homotopy type}
49: \title{Hodge correlators}
50: \author{A.B. Goncharov }
51:
52: \date{\it To Alexander Beilinson for his 50th birthday}
53:
54: \end{titlepage}
55: %\stepcounter{page}
56: \maketitle
57:
58: \tableofcontents
59:
60: \section{Introduction}
61:
62:
63:
64: \subsection{Summary}
65:
66: Let $X$ be a smooth compact complex curve, $v_0$ a non-zero
67: tangent vector
68: at a point $s_0$ of $X$, and $S^* = \{s_1, \ldots , s_m\}$
69: a collection of distinct points of $X $
70: different from $s_0$. We introduce {\it Hodge correlators}
71: related to this datum. They are complex numbers, given by
72: integrals of certain differential forms over products of copies of $X $.
73: Let $S = S^* \cup \{s_0\}$. When the data $(X, S, v_0)$
74: varies, Hodge correlators satisfy
75: a system of non-linear quadratic differential equations.
76: Using this,
77: we show that Hodge correlators
78: encode a variation of real mixed Hodge structures, called below mixed $\R$-Hodge structures. We prove that
79: it coincides
80: with the standard mixed $\R$-Hodge structure
81: on the pronilpotent completion $\pi^{\rm nil}_1(X -S, v_0)$
82: of the fundamental group of
83: $X -S$. The latter was defined, by
84: different methods, by J. Morgan \cite{M}, R. Hain \cite{H} and A.A. Beilinson (cf.
85: \cite{DG}).
86: The real periods of $\pi^{\rm nil}_1(X -S, v_0)$
87: have a well known description via Chen's iterated integrals.
88: Hodge correlators give a completely different
89: way to describe them.
90: \vskip 3mm
91:
92: Mixed Hodge structures are relatives
93: of $l$-adic Galois representations.
94: There are two equivalent definitions of mixed $\R$-Hodge structures:
95: as vector spaces with the weight and Hodge filtrations
96: satisfying some conditions \cite{D},
97: and as representations of the {\it Hodge Galois group} \cite{D2}.
98:
99: The true analogs of $l$-adic representations are
100: Hodge Galois group modules. However
101: usually we describe mixed $\R$-Hodge structures
102: arising from geometry by constructing the weight and Hodge filtrations,
103: getting the Hodge Galois group modules only {\it a posteriori}.
104: The Hodge correlators
105: describe the mixed
106: $\R$-Hodge structure on $\pi_1^{\rm nil}(X-S, v_0)$ directly as a module over
107: the Hodge Galois group.
108:
109:
110:
111:
112:
113: \vskip 3mm
114:
115: We introduce a Feynman integral related to $X$. It does not have
116: a rigorous mathematical meaning. However
117: the standard perturbative series
118: expansion procedure
119: provides a collection of its {\it correlators} assigned to
120: the data $(X,S, v_0)$,
121: which turned out to be
122: convergent finite dimensional integrals. We show that they
123: coincide with the Hodge correlators,
124: thus explaining the name of the latter. Moreover:
125:
126: \begin{figure}[ht]
127: \centerline{\epsfbox{hc1.eps}}
128: %\caption{}
129: \label{hc1}
130: \end{figure}
131:
132: \vskip 3mm
133: We define
134: {\it motivic correlators}. Their periods are the Hodge correlators.
135: Motivic correlators lie in the {\it motivic Lie coalgebra}, and describe the motivic fundamental group of $X-S$.
136: %\begin{figure}[ht]
137: %\centerline{\epsfbox{hc0.eps}}
138: %\caption{}
139: %\label{hc0}
140: %\end{figure}
141: The coproduct in the {motivic Lie coalgebra} is a new feature,
142: which is missing when we work just with numbers.
143: We derive a simple explicit formula for the coproduct
144: of motivic correlators.
145: It allows us to perform {\it arithmetic analysis} of Hodge correlators.
146: This is one of the essential advantages of the Hodge correlator description
147: of the real periods of $\pi_1^{\rm nil}(X-S)$, which has a lot of arithmetic
148: applications. It was available before
149: only for the rational curve case \cite{G7}.
150:
151: \vskip 3mm
152:
153: The Lie algebra of the unipotent part of the Hodge Galois group
154: is a free graded
155: Lie algebra. Choosing a set of its generators we arrive at
156: a collection of periods of a real MHS.
157: We introduce new generators of the Hodge Galois group,
158: which differ from Deligne's generators \cite{D2}.
159: The periods of variations mixed $\R$-Hodge structures
160: corresponding to these generators satisfy non-linear quadratic
161: Maurer-Cartan type
162: differential equations. For the subcategory of Hodge-Tate structures
163: they were defined by A. Levin \cite{L}. The Hodge correlators
164: are the periods for this set of generators.
165:
166:
167: We introduce a DG Lie coalgebra ${\cal L}_{{\cal H}, X}^*$.
168: The category of ${\cal L}_{{\cal H}, X}^*$-comodules
169: is supposed to be a DG-enhancement
170: of the category of {\it smooth $\R$-Hodge sheaves}, i.e.
171: the subcategory of Saito's mixed $\R$-Hodge sheaves whose cohomology are
172: variations of mixed $\R$-Hodge structures. We show that
173: the category of comodules over the Lie coalgebra $H^0{\cal L}_{{\cal H}, X}^*$
174: is equivalent to the category of variations of mixed $\R$-Hodge structures.
175:
176: \vskip 3mm
177:
178:
179:
180: The simplest Hodge correlators for
181: the rational and elliptic curves deliver
182: single-valued versions of the classical polylogarithms
183: and their elliptic counterparts, the classical
184: Eisenstein-Kronecker series \cite{We}. The latter were
185: interpreted by A.A. Beilinson and A. Levin \cite{BL} as periods of
186: variations of mixed $\R$-Hodge structures.
187: More generally, when $X= \C{\Bbb P}^1$ the Hodge correlators are real periods of
188: variations of mixed Hodge structures related to multiple polylogarithms.
189: For an elliptic curve $E$ they deliver
190: the multiple Eisenstein-Kronecker series defined in \cite{G1}.
191:
192:
193: When $X$ is a modular curve and $S$ is the set of its cusps,
194: Hodge correlators
195: generalize the Rankin-Selberg integrals.
196: Indeed, one of the simplest of them is the Rankin-Selberg
197: convolution of a pair $f_1, f_2$ of weight two cuspidal Hecke eigenforms
198: with an Eisenstein series. It
199: computes, up to certain
200: constants, the special value $L(f_1\times f_2, 2)$.
201: A similar Hodge correlator gives
202: the Rankin-Selberg convolution
203: of a weight two cuspidal Hecke eigenform $f$
204: and two Eisenstein series, and computes, up to certain
205: constants, $L(f,2)$.
206: (A generalization to higher weight modular forms will appear elsewhere).
207:
208:
209:
210: The simplest motivic correlators on modular curves deliver
211: Beilinson's elements in motivic cohomology, e.g.
212: the Beilinson-Kato Euler system in $K_2$.
213: We use motivic correlators to define {\it motivic multiple L-values}
214: related to: Dirichlet characters of $\Q$
215: (for $X= \C^*-\mu_N$, where $\mu_N$ is the group of $N$-th roots of unity);
216: Hecke Gr\"ossencharacters imaginary quadratic fields
217: (for CM elliptic curves minus torsion points);
218: Jacobi Gr\"ossencharacters of cyclotomic fields (for affine Fermat curves);
219: the weight two modular forms.
220:
221: \vskip 3mm
222: The Hodge correlators considered in this paper admit
223: a generalization when $X-S$ is replaced by
224: an arbitrary regular complex projective variety.
225: We prove that they describe the mixed $\R$-Hodge structure on the rational
226: motivic homotopy type of $X$, defined by a different method
227: in \cite{M}. We describe them in a separate paper since
228: the case of curves is more transparent, and so far has more applications\footnote{The case of a smooth compact Kahler manifold was treated in \cite{GII}}.
229: \vskip 3mm
230: Key constructions of this paper were outlined in Sections 8-9 of
231: \cite{G1},
232: which may serve as an introduction.
233:
234: \subsection{Arithmetic motivation I: special values of $L$-functions}
235: Beilinson's conjectures imply that special values of
236: $L$-functions of motives over $\Q$
237: at the integral points to the left of the critical line are periods of mixed motives over $\Q$.
238: This means that they can be written as integrals
239: $$
240: \int_{\Delta_B}\omega_A
241: $$
242: where $A$ and $B$ are normal crossing divisors over $\Q$
243: in a smooth $n$-dimensional projective variety $X$ over $\Q$, $\omega_A\in \Omega_{\rm log}^n(X-A)$ is an $n$-form with logarithmic singularities at $A$, and
244: $\Delta_B$ is an $n$-chain with the boundary on $B(\C)$.
245: Moreover, the special values are periods of their motivic avatars --
246: the motivic $\zeta$-elements -- which play a crucial role in arithmetic applications.
247:
248: The classical example is given by the Leibniz formula for the special values of the Riemann $\zeta$-function:
249: $$
250: \zeta(n) = \int_{0 \leq t_1 \leq \ldots \leq t_n \leq 1}\frac{dt_1}{1-t_1}\wedge \frac{dt_2}{t_2}.
251: \wedge \ldots \wedge \frac{dt_n}{t_n}.
252: $$
253: The corresponding motivic $\zeta$-elements were extensively studied (\cite{BK}, \cite{BD}, \cite{HW}).
254: Beilinson's conjectures are known
255: for the special values of the Dedekind $\zeta$-function of a number field thanks to the work of Borel
256: \cite{Bo1}, \cite{Bo2}.
257:
258: \vskip 3mm
259: Here is another crucial example. Let $f(z)$ be a weight two cuspidal Hecke eigenform.
260: So $f(z)dz$ is a holomorphic $1$-form on a compactified modular curve $\overline M$.
261: Recall that by the Manin-Drinfeld theorem the image in the Jacobian of
262: any degree zero cuspidal divisor $a$
263: on $\overline M$ is torsion. So there is a modular unit
264: $g_a \in {\cal O}^*(M)$ such that ${\rm div}(g_a)$ is a multiple of
265: the divisor $a$. According to Bloch and Beilinson \cite{B} we get an element
266: \be \la{BEE}
267: \{g_a, g_b\}\in K_2(\overline M)\otimes \Q.
268: \ee
269: Applying the regulator map to this element, and evaluating it on the
270: $1$-form $f(z)dz$, we get an integral
271: \be \la{BEERS}
272: \int_{\overline M(\C)}\log|g_a|\overline \partial \log|g_b| f(z)dz.
273: \ee
274: It is a Rankin-Selberg convolution integral. Indeed, $\log|g_a|$ is the value of a
275: non-holomorphic Eisenstein series at $s=1$, and $\partial \log|g_b|$ is a holomorphic
276: weight two Eisenstein series. Therefore according to the Rankin - Selberg method,
277: the integral is proportional to the product of special values of $L(f, s)$ at $s=1$ and $s=2$.
278: Moreover (see \cite{SS} for a detailed account) one can find cuspidal divisors $a$ and $b$ such that the proportionality coefficient is non-zero,
279: i.e.
280: \be \la{BEERS213}
281: \int_{\overline M(\C)}\log|g_a|\overline \partial \log|g_b| f(z)dz \sim L(f,2).
282: \ee
283: Using the functional equation for $L(f,s)$,
284: one can easily deduce that
285: $L'(f,0)$ is a period.
286: Finally, suitable modifications
287: of Beilinson's $\zeta$-elements (\ref{BEE})
288: give rise to Kato's Euler system \cite{Ka}.
289:
290: There are similar results, due to Beilinson \cite{B2} in the weight two case, and
291: Beilinson (unpublished) and, independently, Scholl (cf. \cite{DS}) for the special values of L-functions of
292: cuspidal Hecke eigenforms of arbitrary weight $w\geq 2$ at any integral point to the
293: left of the critical strip.
294:
295: \vskip 3mm
296: This picture, especially the
297: motivic $\zeta$-elements, seem to come out of the blue.
298: Furthermore, already for $L''({\rm Sym}^2f, 0)$, related by the functional equation to
299: $L({\rm Sym}^2f, 3)$
300: we do not know in general how to prove that it is a period.
301:
302: One may ask whether there is a general framework,
303: which delivers naturally
304: both Rankin-Selberg integrals (\ref{BEERS}) and Beilinson's $\zeta$-elements (\ref{BEE}),
305: and tells where to look for generalizations related to non-critical
306: special values $L({\rm Sym}^mf, n)$.
307:
308: We suggest that one should look at the motivic fundamental group of the
309: universal modular curve. We show that the simplest Hodge correlator
310: for the modular curve coincides with the Rankin-Selberg integral (\ref{BEERS}), and
311: the corresponding motivic correlator is Beilinson's motivic $\zeta$-element (\ref{BEE}) --
312: see Section \ref{mccurves} for an elaborate discussion.
313:
314: We will show elsewhere that the Rankin-Selberg integrals related to the
315: values at non-critical
316: special points of $L$-functions
317: of arbitrary cuspidal Hecke eigenforms and the corresponding motivic $\zeta$-elements
318: appear naturally as the Hodge and motivic correlators related
319: to the standard local systems on the modular curve:
320: the present paper deals with the trivial local system.
321:
322:
323:
324:
325: \subsection{Arithmetic motivation II: arithmetic analysis of periods}
326:
327:
328:
329: The goal of {\it arithmetic analysis} is to investigate periods without actually computing them, using instead
330: arithmetic theory of mixed motives.\footnote{The latter so far mostly is a conjectural theory, so
331: we often arrive to conjectures rather then theorems.} Here is
332: the simplest non-trivial example, provided by the values of the dilogarithm at rational numbers.
333: Recall that the classical dilogarithm
334: $$
335: {\rm Li}_2(z):= -\int_0^z\log (1-t) \frac{dt}{t}
336: $$
337: is a multivalued function on $\C{\Bbb P}^1-\{0, 1, \infty\}$. Its value at $z$ depends on the homotopy class of the path
338: from $0$ to $z$ on $\C{\Bbb P}^1-\{0, 1, \infty\}$ used to define the integral. Computing the monodromy
339: of the dilogarithm, we see that ${\rm Li}_2(z)$ is well defined modulo the subgroup of $\C$ generated by
340: $(2\pi i)^{2}$ and $2\pi i\log(z)$.
341: So the values ${\rm Li}_2(z)$ at $z\in \Q$ are well defined modulo the subgroup
342: $2\pi i\log \Q^*$
343: spanned by $\Z(2):= (2\pi i)^{2}\Z$ and
344: $2\pi i\log q$, $q\in \Q^*$ -- the latter is well defined modulo $\Z(2)$.
345:
346:
347:
348: Consider a map of abelian groups
349: \be \la{3.27.10.1}
350: \Delta: \Z[\Q^*-\{1\}] \lra \Q^*\otimes \Q^*, \qquad \{z\} \lms -(1-z) \otimes z.
351: \ee
352: \bcon \la{3.24.01.10}
353: Let $z_i \in \Q, a_j, b_j\in \Q^*$ and $n_i \in \Z$. Then one has
354: \be \la{3.27.10.1a}
355: \sum_i n_i{\rm Li}_2(z_i) + \sum_j \log a_j \cdot \log b_j = 0 ~ \mbox{\rm mod}~(2\pi i\log \Q^*)
356: \ee
357: if and only if
358: \be \la{3.27.10.1b}
359: \Delta\Bigl(\sum_in_i\{z_i\}\Bigr) + \sum_j( a_j \otimes b_j + b_j\otimes a_j) =0 ~~
360: \mbox{\rm in}~~ (\Q^*\otimes \Q^*)\otimes_{\Z}\Q.
361: \ee
362: \econ
363: This gives a complete conjectural description
364: of the $\Q$-linear relations between the
365: values of the dilogarithm and the products of two logarithms at rational numbers.
366:
367:
368: \bt \la{3.25.10.1}
369: The condition (\ref{3.27.10.1b}) implies (\ref{3.27.10.1a}).
370: \et
371: This tells the identities between integrals without actually computing the integrals.
372:
373: \vskip 3mm
374: {\bf Example}. One has
375: \be \la{5teq}
376: {\rm Li}_2(\frac{1}{3}) +{\rm Li}_2(-\frac{1}{2}) + \frac{1}{2}(\log \frac{3}{2})^2
377: =0 ~~\mbox{\rm mod}~(2\pi i\log \Q^*).
378: \ee
379: Indeed, modulo $2$-torsion in $\Q^* \otimes \Q^*$ (notice that $3/2 \otimes (-1)$ is a $2$-torsion) one has:
380: \be \la{5teq1}
381: (1-\frac{1}{3})\otimes \frac{1}{3} +(1+\frac{1}{2})\otimes (-\frac{1}{2})
382: = \frac{2}{3}\otimes \frac{1}{3} +
383: \frac{3}{2}\otimes \frac{1}{2} = \frac{3}{2} \otimes \frac{3}{2}.
384: \ee
385:
386: \paragraph{Motivic avatars of the logarithm and the dilogarithm.}
387: The values of the dilogarithm
388: and the product of two logarithms at rational arguments are
389: weight two mixed Tate periods over $\Q$. In fact they describe all such periods.
390: Here is the precise meaning of this.
391: \vskip 2mm
392: The category of mixed Tate motives over a number field $F$
393: is canonically equivalent to the category of graded comodules over
394: a Lie coalgebra ${\cal L}_{\bullet}(F)$ over $\Q$, graded by positive integers, the weights (cf. \cite{DG}).
395: The graded Lie coalgebra ${\cal L}_{\bullet}(F)$ is free, with the space of generators in degree $n$
396: isomorphic to $K_{2n-1}(F)\otimes \Q$. Equivalently, the kernel of the coproduct map
397: $$
398: \delta: {\cal L}_{\bullet}(F) \lra \Lambda^2 {\cal L}_{\bullet}(F)
399: $$
400: is isomorphic to the graded space $\oplus_{n>0}K_{2n-1}(F)\otimes \Q$,
401: where $K_{2n-1}(F)\otimes \Q$ is in degree $n$.
402: One has $K_1(F) =F^*$, and the group $K_{2n-1}(F)$ is of finite rank, which is
403: $r_1+r_2$ for odd and $r_2$ for even $n$ \cite{Bo2}.\footnote{Here $r_1$ and $r_2$
404: are the numbers of real and complex places of $F$, so that $[F:\Q]=2r_2+r_1$.}
405: It follows that ${\cal L}_{1}(F) = F^* \otimes_{\Z}\Q$. Let us describe the $\Q$-vector space ${\cal L}_{2}(F)$.
406:
407:
408:
409: For any field $K$, consider a version of the map (\ref{3.27.10.1}):
410: $$
411: \delta: \Z[K^*-\{1\}] \lra K^*\wedge K^*, \qquad \{z\} \lms - (1-z) \wedge z.
412: $$
413: Let $r(*,*,*,*)$ be the cross-ratio of four points on ${\Bbb P}^1$, normalized by
414: $r(\infty, 0, 1, x) = x$. Let $R_2(K)$ be the
415: subgroup of $\Z[K^*-\{1\}]$ generated by the elements
416: \be \la{5trel}
417: \sum_{i=1}^5 (-1)^i\{r(x_1, ..., \widehat x_i, ..., x_5)\}, \qquad x_i \in P^1(F),
418: \quad x_i \not = x_j.
419: \ee
420: Then $\delta (R_2(K)) =0$. Let $B_2(K):= \Z[K^*-\{1\}]/R_2(K)$ be the Bloch group of $K$.
421: The following proposition is deduced in Section 1 of \cite{G10} from a theorem of Suslin \cite{S}.
422:
423: \bp \la{3.25.10.2}
424: Let $F$ be a number field. Then one has ${\cal L}_{2}(F) = B_2(F) \otimes_{\Z}\Q$.\footnote{The same
425: should be true for any field $F$. However the Lie coalgebra ${\cal L}_\bullet(F)$ in that case is a conjectural object.}
426: \ep
427:
428: The dual of the universal enveloping algebra
429: of Lie coalgebra ${\cal L}_{\bullet}(F)$ is a commutative
430: graded Hopf algebra. Its weight two component is isomorphic to
431: is
432: \be \la{Motavat}
433: {\cal L}_{2}(F) \oplus {\rm Sym}^2{\cal L}_{1}(F) = (B_2(F) \oplus {\rm Sym}^2F^* )\otimes_{\Z}\Q.
434: \ee
435: The coproduct is provided by the map
436: \be \la{Motavat1}
437: \Delta_2: B_2(F) \oplus {\rm Sym}^2F^* \lra (F^*\otimes F^*)\otimes_\Z \Q
438: \ee
439: $$ \{z\}_2 \lms
440: \frac{1}{2}\Bigl(z \otimes (1-z)-(1-z) \otimes z \Bigr) , \quad
441: a\cdot b \lms a \otimes b + b \otimes a.
442: $$
443:
444: Let us return to our example, when $F=\Q$. The period homomorphism
445: $$
446: {\cal L}_{1}(\Q) \lra \C/\Z(1), \qquad q \lms \log q
447: $$ is known to be injective.
448: So $q\in \Q^*$ is the motivic avatar of $\log q$.
449:
450: \bp \la{3.25.10.3}
451: There is a weight two period map
452: \be \la{3.25.10.4}
453: {\cal L}_{2}(\Q) \oplus {\rm Sym}^2{\cal L}_{1}(\Q) \lra \C/2\pi i\log \Q^*, \qquad
454: \ee
455: $$
456: \{q\}_2 \lms {\rm Li}_2(q) + \frac{1}{2}\log(1-q)\log(q), \quad a\otimes b + b \otimes a \lms \log a \log b.
457: $$
458: \ep
459: The point is that the period map kills the
460: subgroup $R_2(\C)$, which is equivalent to Abel's five term relation for the dilogarithm.
461:
462: The period map (\ref{3.25.10.4}) is conjectured to be injective:
463: this is a special case of a Grothendieck type conjecture on periods.
464: Finally, ${\rm Ker}\Delta_2 = K_3(\Q)\otimes \Q$,
465: which is zero by Borel's theorem. The injectivity of $\Delta_2$
466: plus Propositions \ref{3.25.10.2} and \ref{3.25.10.3} imply Theorem \ref{3.25.10.1},
467: and, modulo the injectivity of the period map (\ref{3.25.10.4}), Conjecture \ref{3.24.01.10}.
468:
469: Although relation (\ref{5teq}) follows from the five
470: term relation for the dilogarithm, there is no procedure to write the element $\{\frac{1}{3}\} +
471: \{-\frac{1}{2}\}$ as a sum of the five term relations (\ref{5trel}).
472: The only effective way to know that they exist is
473: to calculate the coproduct, as we did in (\ref{5teq1}).
474:
475: Let us now discuss the case when $F$ is an arbitrary number field.
476: \vskip 3mm
477: {\it Real periods.} There is a single valued version of the dilogarithm, the Bloch-Wigner function
478: $$
479: {\cal L}_2(z):= -{\rm Im}\Bigl(\int_0^z\log (1-t) \frac{dt}{t} + \int_0^z\frac{dt}{1-t}\cdot \log|z|\Bigr).
480: $$
481: where both integrals are defined by using the same integration
482: path from $0$ to $z$. It satisfies the five term relation, and thus
483: provides a group homomorphism
484: $$
485: B_2(\C) \lra \R, \qquad \{z\}_2 \to {\cal L}_2(z).
486: $$
487: Using the (conjectural) isomorphism ${\cal L}_2(\C) = B_2(\C)$, it can be interpreted as
488: the {\it real period map} ${\cal L}_2(\C)\to \R$. Notice that
489: ${\cal L}_2(z) =0$ for $z\in \R$. So the real period map looses a lot of
490: information about the motivic Lie algebra. It relates, however,
491: the kernel of the coproduct map with the special values of $L$-functions.
492: Here is how it works in our running example.
493:
494: Let $F$ be a number field.
495: Then by Borel's theorem \cite{Bo2} the real period map provides an
496: injective regulator map
497: $$
498: K_3(F)\otimes \Q \stackrel{\sim}{=} {\rm Ker}\delta \subset B_2(F)\otimes \Q \lra \R^{r_2},
499: \qquad \{z\}_2
500: \lms
501: \Bigl({\cal L}_2(\sigma_1(z)), ..., {\cal L}_2(\sigma_{r_2}(z))\Bigr).
502: $$
503: Its image is a rational lattice -- that is, a lattice tensor $\Q$ --
504: in $\R^{r_2}$, whose covolume, well defined up to $\Q^*$,
505: is a $\Q^*$-multiple of $\zeta_F(-1)$.
506:
507: \vskip 2mm
508: So to perform the arithmetic analysis of the values of the dilogarithm / products of two logarithms
509: at rational arguments we
510: upgrade them to their motivic avatars lying in (\ref{Motavat}),
511: and determine their coproduct there. The kernel of the coproduct is captured by the
512: regulator map.
513:
514: The weight $m$ periods of mixed Tate motives over
515: a number field $F$ are studied similarly using the iterated coproduct in $\otimes^mF^*$,
516: see Section 4 of \cite{G7}.
517:
518: \vskip 2mm
519: In this paper we develope a similar picture for the periods of
520: the pronilpotent completions of fundamental groups of curves. Namely, we introduce motivic correlators,
521: which are the motivic avatars of the periods of fundamental groups of curves.
522: They span a Lie subcoalgebra in the motivic Lie coalgebra. The key point is that the coproduct
523: of a motivic correlator is given explicitly in terms of the motivic correlators. We prove that
524: their real periods
525: are given by the Hodge correlators.
526:
527:
528: \subsection{Pronilpotent completions of
529: fundamental groups of curves}
530:
531: The fundamental
532: group $\pi_1 = \pi_1(X - S, v_0)$ is a free group
533: with generators provided by loops around the
534: rest of the punctures $s_i \not = s_0$ and loops
535: generating $H_1(X )$.
536: Let ${\cal I}:= {\rm Ker}(\Q[\pi]\to \Q)$ be the augmentation ideal of the group algebra of
537: $\pi_1$. Then there is
538: a complete cocommutative Hopf algebra over $\Q$
539: \begin{equation} \label{10/25/04/1}
540: {\rm A}^{\rm Betti}(X - S, v_0):=
541: \lim_{{\longleftarrow}}\Q[\pi_1]/{\cal I}^n.
542: \end{equation}
543: Its coproduct is induced by the map $g \lms g \otimes g$, $g \in \pi_1$.
544: It is called the {\it fundamental Hopf algebra of $X - S$} with the tangential base point $v_0$.
545: The subset of its primitive elements is a free pronilpotent Lie
546: algebra over $\Q$, the Maltsev completion of $\pi_1$. It is denoted
547: $\pi_1^{\rm nil}(X-S, v_0)$ %or ${\rm L}^{\rm Betti}(X - S, v_0)$
548: and called the {\it fundamental Lie algebra}
549: of $X -S$.
550: The Hopf algebra (\ref{10/25/04/1})
551: is its universal enveloping algebra.
552:
553:
554: Denote by ${\rm T}(V)$ the tensor algebra of a vector space $V$.
555: The associated graded for the ${\cal I}$-adic filtration is isomorphic to the
556: tensor algebra
557: of $H_1(X -S, \Q)$:
558: $$
559: {\rm gr}^{\cal I}{\rm A}^{\rm Betti}(X - S, v_0) =
560: \oplus_{n=0}^{\infty} {\cal I}^n/{\cal I}^{n+1} = {\rm T}\Bigl(H_1(X -S, \Q)\Bigr).
561: $$
562:
563: There is a weight filtration on $H_1(X -S, \Q)$, given by the
564: extension
565: \begin{equation} \label{6.11.06.9}
566: 0 \lra \Q(1)^{|S|-1} \lra H_1(X -S, \Q) \lra H_1(X , \Q) \lra 0,
567: \end{equation}
568: where the right arrow is provided by the embedding
569: $%\begin{equation} \label{ex}
570: X -S \hookrightarrow X .
571: $
572: The Hopf algebra (\ref{10/25/04/1}) is equipped with a weight filtration $W$ compatible with
573: the Hopf algebra structure.
574: The corresponding
575: associated graded of the Hopf algebra (\ref{10/25/04/1}) is isomorphic to the tensor algebra
576: of ${\rm gr}^WH_1(X -S, \Q)$:
577: \begin{equation} \label{6.11.06.10}
578: {\rm gr}^{W}{\rm A}^{\rm Betti}(X - S, v_0) = {\rm T}\left( {\rm gr}^WH_1(X -S, \Q)
579: \right).
580: \end{equation}
581:
582:
583: \subsection{Hodge correlators} Recall $S^*= S-\{s_0\}$. Set
584: $$
585: {\rm V}_{X, S^*} := H_1(X, \C) \oplus \C[S^*], \qquad {\rm A}_{X, S^*} :=
586: {\rm T}({\rm V}_{X, S^*}).
587: $$
588: One sees from (\ref{6.11.06.9}), (\ref{6.11.06.10}) that there are canonical isomorphisms
589: of vector spaces
590: $$
591: {\rm V}_{X, S^*} = {\rm gr}^{W}H_1(X -S, \C), \quad {\rm A}_{X, S^*} =
592: {\rm gr}^{W}{\rm A}^{\rm Betti}(X - S, v_0)\otimes \C.
593: $$
594: We also need the dual objects:
595: $$
596: {\rm V}^{\vee}_{X, S^*} := H^1(X, \C) \oplus \C[S^*],
597: \qquad {\rm A}^{\vee}_{X, S^*} := {\rm T}({\rm V}^{\vee}_{X, S^*}).
598: $$
599:
600: For an associative algebra $A$ over $k$, let $A^+:= {\rm Ker}(A \to k)$ be the augmentation ideal. Let $[A^+, A^+]$ be the subspace (not the ideal) of $A^+$ generated by
601: commutators in $A^+$. The quotient space ${\cal C}(A):= A^+/[A^+,A^+]$
602: is called the {\it cyclic envelope}
603: of $A$. If $A$ is
604: freely generated by a set ${\cal S}$ then the vector space
605: ${\cal C}(A)$ has a basis parametrised by cyclic words in ${\cal S}$.
606:
607:
608: We use the shorthand ${\cal C}{\rm T}(V)$ for
609: ${\cal C}({\rm T}(V))$.
610: We define the subspace of {\it shuffle relations} in
611: ${\cal C}{\rm T}({\rm V})$ as the subspace generated by the
612: elements
613: $$
614: \sum_{\sigma \in \Sigma_{p,q}}(v_0 \otimes v_{\sigma(1)} \otimes \ldots \otimes v_{\sigma(p+q)}),
615: \qquad p, q \geq 1,
616: $$
617: where the sum is over all $(p,q)$-shuffles.
618: Set
619: \be \la{mvlca}
620: {\cal C}^{\vee}_{X, S^*}:= {\cal C}({\rm A}^{\vee}_{X, S^*}), \qquad
621: {\cal C}{{\cal L}ie}^{\vee}_{X, S^*}:=
622: \frac{ {\cal C}^{\vee}_{X, S^*}}{\mbox{Shuffle relations}};
623: \ee
624: \be \la{mvla}
625: {\cal C}_{X, S^*}:= {\cal C}({\rm A}_{X, S^*})
626: \qquad
627: {\cal C}{{\cal L}ie}_{X, S^*}:= \mbox{the dual of
628: ${\cal C}{{\cal L}ie}^{\vee}_{X, S^*}$}.
629: \ee
630: We show that ${\cal C}^{\vee}_{X, S^*}$ is a Lie coalgebra,
631: ${\cal C}{{\cal L}ie}^{\vee}_{X, S^*}$ is its quotient Lie coalgebra.
632: Equivalently,
633: ${\cal C}_{X, S^*}$ is a Lie algebra,
634: ${\cal C}{{\cal L}ie}_{X, S^*}$ is its Lie subalgebra.
635: .
636:
637: \vskip 3mm
638: Given a mass one volume form $\mu$ on $X $,
639: our Feynman integral construction in Section \ref{hc2sec}
640: provides a linear map, called the
641: {\it Hodge correlator map}
642: \be \la{1.15.08.1}
643: {\rm Cor}_{{\cal H}, \mu}: {\cal C}{{\cal L}ie}^{\vee}_{X, S^*} \lra \C.
644: \ee
645: {\bf Example.} For a cyclic word
646: \begin{equation} \label{11:42}
647: W = {\cal C}\Bigl(
648: \{a_0\} \otimes \{a_{1}\} \otimes ... \otimes \{a_{n}\} \Bigr),
649: \qquad a_i \in S^*,
650: \end{equation}
651: the Hodge correlator is given by a
652: sum over all plane trivalent trees $T$ whose external edges are decorated by
653: elements $a_0, \ldots , a_n$, see Fig \ref{feyn36}. To define the integral
654: corresponding to such a tree $T$ we proceed as follows.
655: The volume form $\mu$ provides a Green function $G(x,y)$
656: on $X ^2$. Each edge $E$ of $T$
657: contributes a Green function on $X ^{\{\mbox{\rm vertices of $E$}\}}$,
658: which we lift to a function
659: on $X ^{\{\mbox{\rm vertices of $T$}\}}$. Further, given
660: any $m+1$ smooth functions on a complex manifold $M$, there is a canonical
661: linear map $\omega_m: \Lambda^{m+1}{\cal A}^0_M \to {\cal A}^m_M$, where
662: ${\cal A}^k_M$ is the space of smooth $k$-forms on $M$ (Section 2.2).
663: Applying it to the Green functions assigned to
664: the edges of $T$, we get a differential form
665: of the top degree
666: on $X ^{\{\mbox{\rm internal vertices of $T$}\}}$. Integrating it, we
667: get the integral assigned to $T$. Taking the sum over all trees $T$, we get
668: the number ${\rm Cor}_{{\cal H}, \mu}(W)$. The shuffle relations result
669: from taking the sum over all trees with a given decoration.
670: \begin{figure}[ht]
671: \centerline{\epsfbox{feyn36.eps}}
672: \caption{A plane trivalent tree decorated by ${\cal C}(\{a_0\} \otimes \ldots \otimes \{a_5\})$.}
673: \label{feyn36}
674: \end{figure}
675:
676: Dualising the Hodge correlator map (\ref{1.15.08.1}), we get a {\it Green operator}
677: \begin{equation} \label{5.22.06.1}
678: {\Bbb G}_{\mu} \in {\cal C}{{\cal L}ie}_{X, S^*}.
679: \end{equation}
680:
681: Let
682: ${\rm L}_{X,S^*}$ be the free Lie algebra generated by the space
683: ${\rm V}_{X,S^*}$. There is a canonical isomorphism
684: \be \la{varl}
685: {\rm L}_{X,S^*} = {\rm gr}^W\pi_1^{\rm nil}(X-S, v_0)\otimes \C.
686: \ee
687: The Green operator can be viewed as a derivation of the Lie algebra ${\rm L}_{X,S^*}$ as follows.
688:
689:
690:
691: Let $(p_i, q_i)$ be a symplectic basis of $H_1(X)$. Denote by $X_s$ the generator of
692: ${\rm L}_{X,S^*}$ assigned to $s\in S^*$. There is a canonical generator
693: of ${\rm gr}^W\pi_1^{\rm nil}(X-S, v_0)$ corresponding to a loop around $s_0$.
694: Its projection to ${\rm L}_{X,S^*}$ is written as
695: \begin{equation} \label{6.15.06.1}
696: X_{s_0}:= -\sum_{s\in S^*}X_{s}+ \sum[p_i, q_i].
697: \end{equation}
698:
699: Say that a derivation of
700: the Lie algebra ${\rm L}_{X,S^*}$ is {\it special} if it kills
701: the generator (\ref{6.15.06.1}) and
702: preserves the conjugacy classes
703: of the generators $X_{s}$, where $s\in S^*$.
704:
705: This definition is motivated by
706: the $l$-adic picture (Section \ref{hc8sec}.2): If $X$ is a curve
707: over a field $F$ and $S \subset X(F)$,
708: the Galois group ${\rm Gal}(\overline F/F(\mu_{l^\infty}))$
709: acts by special automorphisms
710: of ${\rm L}_{X,S^*}\otimes \Q_l$. Indeed, by the comparison theorem
711: the latter is identified
712: with the associate graded
713: for the weight filtration on the Lie algebra of
714: the pro-$l$-completion $\pi_1^{(l)}(X-S, v_0)$ of the fundamental group.
715:
716: Denote the Lie algebra of special derivations by
717: ${\rm Der}^S{\rm L}_{X,S^*}$.
718: We show that the Lie algebra
719: ${\cal C}{{\cal L}ie}_{X,S^*}$ acts by special derivations on
720: the Lie algebra ${\rm L}_{X,S^*}$. This generalizes constructions of Drinfeld \cite{Dr1}
721: (when $X$ is of genus zero)
722: and Kontsevich \cite{K} (when $S$ is empty).
723: Namely, let $F$ be a cyclic polynomial in non-commuting variables
724: $Y_i$, i.e. $F \in {\cal C}{\rm T}(Y)$
725: where $Y$ is a vector space with a basis $\{Y_i\}$.
726: Then there are ``partial derivatives'' maps
727: $$
728: \partial/\partial Y_i: {\cal C}{\rm T}(Y) \lra {\rm T}(Y), \qquad F \lms {\partial F}/{\partial Y_i},
729: $$
730: defined by deleting one of $Y_i$'s from $F$, getting as a result a non-commutative polynomial,
731: and taking the sum over all possibilities. For instance if $F = {\cal C}(Y_1Y_2Y_1Y_3)$ then
732: ${\partial F}/{\partial Y_1} = Y_2Y_1Y_3+ Y_3Y_1Y_2$.
733:
734: Let $F$ be a cyclic polynomial in non-commuting variables
735: $X_{s}, p_i, q_i$, where $s\in S^*$.
736: Then the derivation $\kappa_F$ assigned to $F$ acts on the generators by
737: $$
738: p_i \lms -\frac{\partial F}{\partial q_i}, \quad q_i \lms \frac{\partial F}{\partial p_i}, \quad
739: X_s \lms [X_s, \frac{\partial F}{\partial X_s}], \quad s\in S^*.
740: $$
741: {\it A priori} it is a derivation of the associative algebra
742: ${\rm A}_{X, S^*}$. However if $F$ annihilates the shuffle relations,
743: it is a derivation of the Lie algebra ${\rm L}_{X, S^*}$.
744: We show that we get an isomorphism
745: %we get a map
746: %$
747: %{\cal C}_{X,S^*} \stackrel{}{\lra} {\rm Der}^S{\rm L}_{X,S^*}, ~~F \lms \kappa_F.
748: %$
749: %surjective with the kernel given by $\oplus_{s\in S^*}\Q[X_s]$.
750: %Denote by $\widetilde {\cal C}_{X,S^*}$ the quotient by this kernel. Then
751: %we arrive at an isomorphism
752: %\begin{equation} \label{9.17.05.5}
753: %\widetilde {\cal C}_{X,S^*} \stackrel{\sim}{\lra} {\rm Der}^S{\rm L}_{X,S^*}, \quad F \lms \kappa_F.
754: %\end{equation}
755: \begin{equation} \label{9.17.05.5}
756: {\cal C}{{\cal L}ie}_{X,S^*} \stackrel{\sim}{\lra} {\rm Der}^S{\rm L}_{X,S^*}, \quad F \lms \kappa_F.
757: \end{equation}
758:
759: Combining (\ref{5.22.06.1}) and (\ref{9.17.05.5}), we conclude that
760: the Green
761: operator can be viewed as an element
762: \begin{equation} \label{5.22.06.1qa}
763: {\Bbb G}_{\mu} \in {\rm Der}^S{\rm L}_{X,S^*}.
764: \end{equation}
765: We show that it encodes a mixed $\R$-Hodge structure using a
766: construction outlined in Section \ref{1.5}.
767:
768:
769: \subsection{A Feynman integral for Hodge correlators} \la{1.4}
770: Let $\varphi$ be a smooth function
771: on $X $ with values in $N\times N$ complex matrices. We say that
772: $\{\varphi\}$ is our space of fields.
773: Given $N$, consider the following correlator corresponding to $W\in {\cal C}^{\vee}_{X,S}$,
774: formally defined via
775: a Feynman integral
776: \begin{equation} \label{3.12.05.3asd}
777: {\rm Cor}_{X, N, h}(W):=
778: \int {\cal F}_W(\varphi)
779: e^{iS(\varphi)}{\cal D}\varphi,
780: \end{equation}
781: where
782: $$
783: S(\varphi):= \frac{1}{2\pi i}\int_{X } {\rm Tr}
784: \Bigl( \frac{1}{2}\partial \varphi \wedge \overline \partial\varphi +
785: \frac{1}{6}\hbar \cdot
786: \varphi [\partial \varphi, \overline \partial \varphi] \Bigr).
787: $$
788: and ${\cal F}_W(\varphi)$ is a function on the space of fields, see
789: Section \ref{hc12sec}.
790: For example, for the cyclic word (\ref{11:42})
791: we set
792: \begin{equation} \label{7.28.06.1}
793: {\cal F}_W(\varphi):= {\rm Tr}\Bigl(\varphi(a_0) \ldots \varphi(a_n) \Bigr).
794: \end{equation}
795: Formula (\ref{3.12.05.3asd}) does not have a precise mathematical meaning.
796: We understand it by postulating the perturbative series
797: expansion with respect to a small parameter $h$,
798: using the standard Feynman rules, and then taking the leading term in the
799: asymptotic expansion as $N \to \infty, \hbar=N^{-1/2}$. This way we get
800: a sum over a finite number of Feynman diagrams given by plane trivalent trees
801: decorated by the factors of $W$. One needs to specify
802: a volume form $\mu$ on $X $ to determine a measure ${\cal D}\varphi$ on the space of fields.
803: Mathematically, we need $\mu$ to specify the Green function which is used
804: to write the perturbative series
805: expansion.
806:
807:
808:
809:
810:
811: We prove in Theorem \ref{1.10.05.1} that these Feynman intregral correlators
812: coincide with the suitably normalised Hodge correlators.
813:
814:
815: \vskip 2mm
816: {\bf Problem}. The Hodge correlators provide only the leading term of the $N \to \infty$
817: asymptotics of the Feynman integral correlators.
818: The next terms in the asymptotics are given by the correlators corresponding to higher genus
819: ribbon graphs. They are typically divergent. Can they be renormalised? What is their role in the Hodge theory?
820: %\vskip 3mm
821: %\input{hc1.5.tex}
822:
823:
824: \subsection{Variations of mixed $\R$-Hodge structures by twistor connections} \la{1.5}
825: \la{hc1.5}
826:
827: A {\it real Hodge structure} is a real vector space $V$
828: whose complexification is equipped with a bigrading $V_\C = \oplus V^{p,q}$ such that
829: $\overline V^{p,q} = V^{q,p}$. A real Hodge structure is {\it pure} if $p+q$
830: is a given number, called the weight.
831: So a real Hodge structure is a direct sum of
832: pure ones of different weights. The category of real
833: mixed Hodge structures is equivalent to the category of representations of the
834: algebraic group $\C^*_{\C/\R}$ (i.e. the group $\C^*$ understood as a real group).
835:
836: According to P. Deligne \cite{D}, a {\it mixed $\R$-Hodge structure}
837: is a real vector space $V$ equipped with a weight filtration $W_{\bullet}$,
838: and a Hodge filtration $F^{\bullet}$ of its complexification $V_\C$,
839: satisfying the following condition. The filtration $F^{\bullet}$
840: and its conjugate $\overline F^{\bullet}$ induce on ${\rm gr}_n^WV$ a pure
841: weight $n$ real Hodge structure:
842: $$
843: {\rm gr}^WV_\C = \oplus_{p+q=n}F_{(n)}^{p}\cap \overline F_{(n)}^{q}.
844: $$
845: Here $F_{(n)}^{\bullet}$ is the filtration on ${\rm gr}_n^WV_\C$ induced by
846: $F^{\bullet}$, and similarly $\overline F_{(n)}^{\bullet}$.
847:
848: \vskip 3mm
849: The category of mixed $\R$-Hodge structures is a
850: Tannakian category
851: with
852: a canonical
853: fiber functor
854: $\omega: H \lms {\rm gr}^WH$ to the category of $\R$-Hodge structures.
855: Thus, thanks to the Tannakian formalism,
856: it is canonically equivalent to the category of representations
857: of a pro-algebraic group, the {\it Hodge Galois group} $G_{\rm Hod}$. The group $G_{\rm Hod}$ is a
858: semidirect product
859: $$
860: 0\lra {U}_{\rm Hod} \lra G_{\rm Hod} \lra \C^*_{\C/\R}\lra 0.
861: $$
862: of a real prounipotent algebraic group ${U}_{\rm Hod}$ and $\C^*_{\C/\R}$.
863: There is an embedding $s: \C^*_{\C/\R} \hra G_{\rm Hod}$ corresponding to the
864: functor $H \to {\rm gr}^WH$.
865: Let ${\rm L}_{\rm Hod}$ be the Lie algebra of the group ${U}_{\rm Hod}$.
866: The action of $s(\C^*_{\C/\R})$ provides the Lie algebra ${\rm L}_{\rm Hod}$ with
867: a structure of a Lie algebra
868: in the category of $\R$-Hodge structures.
869: The category of mixed $\R$-Hodge structures is equivalent to the category of
870: representations of the Lie algebra ${\rm L}_{\rm Hod}$ in the category of $\R$-Hodge structures.
871: The Lie algebra ${\rm L}_{\rm Hod}$ is a free Lie algebra in the category of $\R$-Hodge structures.
872: Thus the Lie algebra ${\rm L}_{\rm Hod}$ is generated by the $\R$-Hodge structure
873: $$
874: \oplus_{[H]}{\rm Ext}_{\R-{\rm MHS}}^1(\R(0), H)^{\vee}\otimes H.
875: $$
876: where the sum is over the set of isomorphism classes of simple $\R$-Hodge structures $H$.
877: The ones with non-zero ${\rm Ext}^1$ are parametrised by pairs of
878: negative integers $(-p,-q)$ up to a permutation (the Hodge degrees of $H$),
879: and one has ${\rm dim}_\R{\rm Ext}_{\R-{\rm MHS}}^1(\R(0), H)=1$
880: for those $H$.
881:
882: It follows that the complex Lie algebra ${\rm L}_{\rm Hod}\otimes_\R\C$ is generated by
883: certain elements
884: $n_{p, q}$, $p,q \geq 1$, of bidegree $(-p, -q)$,
885: with the only relation $\overline n_{p, q}= -n_{q, p}$.
886: Such a generators were defined by Deligne \cite{D2}. We call them Deligne's generators.
887:
888:
889:
890:
891: This just means that
892: a mixed $\R$-Hodge structure can be described by a pair $(V, g)$, where
893: $V$ is a real vector space whose complexification is equipped with a bigrading $V_\C = \oplus V^{p,q}$ such that
894: $\overline V^{p,q} = V^{q,p}$, and
895: $g = \sum_{p,q\geq 1}g_{p,q}$ is an imaginary operator on $V_\C$, so
896: that $g_{p,q}$ is of bidegree $(-p,-q)$ and
897: $\overline g_{p,q} = -g_{q,p}$. The operators $g_{p,q}$ are the images of
898: Deligne's
899: generators
900: $n_{p,q}$ in the representation $V$.
901: One can describe variations of real Hodge structures
902: using the operators $g_{p,q}$ fiberwise, but the Griffiths transversality condition
903: is encoded by complicated nonlinear
904: differential equations on $g_{p,q}$.
905:
906: A.A. Beilinson emphasized that
907: Deligne's generators may not be
908: the most natural ones, and asked whether one should exist
909: a canonical choice of the generators.
910: For the subcategory of mixed $\R$-Hodge-Tate structures
911: (i.e. when the Hodge numbers are zero unless $p=q$)
912: a different set of generators $n_p$
913: was suggested by
914: A. Levin \cite{L}.
915:
916:
917: \vskip 3mm
918: Let us decompose the Hodge correlator ${\Bbb G} = \sum
919: {\Bbb G}_{p,q}$ according to the Hodge bigrading.
920: Then it is tempting to believe that the elements ${\Bbb G}_{p,q}$
921: are images of certain generators of the Lie algebra ${\rm L}_{\rm Hod}$,
922: thus providing the mixed $\R$-Hodge structure on $\pi_1^{\rm nil}(X-S, v_0)$.
923: Furthermore, we proved (Section \ref{hc6sec})
924: that, when the data $(X, S, v_0)$ varies,
925: the components ${\Bbb G}_{p,q}$ satisfy a Maurer-Cartan type
926: nonlinear quadratic differential equations.
927:
928:
929:
930: Motivated by this,
931: we show that a variation of mixed $\R$-Hodge structures can be
932: described by a direct sum of variations of pure Hodge structures
933: plus a {\it Green datum} $\{G_{p,q}, \nu\}$ consisting of
934: operator-valued functions $G_{p,q}$, $p,q \geq 1$ and a $1$-form $\nu$,
935: satisfying a system of quadratic nonlinear differential equations.
936: The Green operators $G_{p,q}$ are
937: nonlinear modifications
938: of Deligne's operators $g_{p,q}$,
939: providing a canonical set of generators of the Hodge
940: Lie algebra ${\rm L}_{\rm Hod}$.
941: For the Hodge-Tate structures we recover
942: the Green operators of \cite{L}, although even in this case our
943: treatment is somewhat different,
944: based on the twistor transform defined in Section \ref{hc5sec}.
945:
946:
947:
948: \vskip 3mm
949: Let us describe our construction.
950: Let ${\cal V}$ be a direct sum of variations of pure real Hodge structures on $X $.
951: It is given by a real local system ${\cal V}$, whose complexification ${\cal V}_{\C}$ has a
952: canonical decomposition
953: \begin{equation} \label{8.20.05.1i}
954: {\cal V}_{\C} = \oplus_{p,q}{\cal V}^{p,q}.
955: \end{equation}
956: Tensoring ${\cal V}$ with the sheaf of smooth complex functions on $X $, we get
957: a $C^{\infty}$-bundle ${\cal V}_{\infty}$ with a flat
958: connection ${\bf d}$ satisfying the
959: Griffiths transversality condition.
960: Pick a real closed $1$-form
961: $$
962: \nu\in \Omega^1\otimes
963: {\rm End}^{-1, 0}{\cal V}_\infty \oplus \overline \Omega^1 \otimes
964: {\rm End}^{0, -1}{\cal V}_\infty, \qquad {\bf d}\nu =0, \quad \overline \nu = \nu,
965: $$
966: and a collection of imaginary operators
967: $$%\begin{equation} \label{8.20.05.6qaz}
968: G_{p,q}\in {\rm End}^{-p, -q}{\cal V}_{\infty}, \quad p,q \geq 1, \qquad \overline G_{p,q} = - G_{p,q}.
969: $$%\end{equation}
970: Let ${\bf d} = \partial' + \partial''$ be the decomposition
971: into the holomorphic and antiholomorphic components. Set
972: ${\bf d}^\C = \partial'- \partial''$. Consider the following $1$-form
973: with values in the local system ${\rm End}{\cal V}$:
974: $$
975: \psi:= {\bf d}^\C G + \nu.
976: $$
977:
978: Let $\varphi^{a,b}_{p,q}$ be an ${\rm End}^{-p, -q}{\cal V}$-valued
979: $(a,b)$-form.
980: We define its {\it Dolbeaux bigrading} $(s,t)$ by setting $s:= a-p$ and $t:=b-q$.
981: Denote by $\varphi_{s,t}$ the
982: component of $\varphi$
983: of the Dolbeaux bidegree $(-s, -t)$.
984:
985: \paragraph{Twistor connections.} Consider a ``twistor line'' $\R$ with a coordinate $u$,
986: the product $X\times \R$, and the canonical projection
987: $p: X \times \R \to X$.
988: Starting from a datum $\{G_{p,q}, \nu\}$, we
989: introduce a {\it twistor connection} $\nabla_{\cal G}$ on $p^*{\cal V}_{\infty}$:
990: \be \la{TCON}
991: \nabla_{\cal G}:= {\bf d}+
992: \sum_{s,t\geq 0}(1-u)^{s}(1+u)^{t}\Bigl((s+t+1)G_{s+1,t+1}du+ \psi_{s, t}\Bigr).
993: \ee
994:
995: The structure of the twistor connection
996: can be clarified by writing it in a homogeneous form as follows.
997: Consider a ``twistor plane'' $\C^2$ with coordinates $(z,w)$,
998: the product $X\times \C^2$, and the projection
999: $\pi: X \times \C^2 \to X$. The {\it homogeneous twistor connection}
1000: is a connection on $\pi^*{\cal V}_{\infty}$: \be \la{TCONH}
1001: \widehat \nabla_{\cal G}:= {\bf d}+
1002: \sum_{s,t\geq 0}z^{s}w^{t}\Bigl((s+t+1)G_{s+1,t+1}
1003: (zdw-wdz)/2
1004: + \psi_{s, t} \Bigr).
1005: \ee
1006: The group $\C^* \times \C^*$
1007: acts on the twistor plane: $(z,w) \lms (\lambda_1z, \lambda_2w)$, as well as
1008: on the Dolbeaux complex of the local system ${\rm End}{\cal V}$ on $X$,
1009: providing the Dolbeaux bigrading: $\varphi_{s,t} \lms
1010: \lambda_1^{-s}\lambda_2^{-t}\varphi_{s,t}$.
1011: The connection $\widehat \nabla_{\cal G}$
1012: is invariant under the action of the group $\C^* \times \C^*$.
1013: Restricting to ``complex twistor line'' $z+w=2$ parametrised by
1014: $z=1-u, w=1+u$, we recover (\ref{TCON}).
1015:
1016:
1017: The twistor line $\R$ is the set of fixed points of the
1018: involution
1019: $\sigma: (z,w) \lms (\overline w, \overline z)$
1020: of the twistor plane $\C^2$,
1021: restricted to the line $z+w=2$, so in fact $u \in i\R$.
1022: The connection $\widehat \nabla_{\cal G}$
1023: is invariant under the composition
1024: of the involution $\sigma$ with the
1025: complex conjugation $c$. Moreover,
1026: $$
1027: (c\circ \sigma)^*\widehat \nabla_{\cal G} = \widehat \nabla_{\cal G} ~~
1028: \mbox{if and only if} ~~ \overline G_{p,q} = -G_{p,q}, \quad \overline \nu = \nu.
1029: $$
1030:
1031: We show that $\nabla_{\cal G}$ is flat if and only if the
1032: datum $\{G_{p,q}, \nu\}$
1033: satisfies a Maurer-Cartan type system of
1034: quadratic nonlinear differential equations.
1035: We call it then a {\it Green datum}.
1036: Green data, or, equivalently, flat twistor connections,
1037: form an abelian tensor category in an obvious way.
1038:
1039:
1040: \bt \la{2.22.08.1}
1041: The abelian tensor
1042: category of flat twistor connections is canonically equivalent to
1043: the category of variations of mixed $\R$-Hodge
1044: structures on $X$.
1045: \et
1046:
1047:
1048: A flat twistor connection gives rise to
1049: a variation of mixed $\R$-Hodge structures on $X$ as follows.
1050: Restricting $\nabla_{\cal G}$ to $X \times \{0\}$
1051: we get a flat connection
1052: $\nabla^{0}_{\cal G}$ on
1053: ${\cal V}_{\infty}$. We equip the local system
1054: $({\cal V}_{\infty}, \nabla^{0}_{\cal G})$ with
1055: a structure of a variation of mixed $\R$-Hodge structures.
1056: The weight filtration
1057: is the standard weight filtration on a bigraded object:
1058: $$
1059: W_n{\cal V}:= \oplus_{p,q\leq n}{\cal V}^{p,q}.
1060: $$
1061: Let $P$ be the operator of parallel transport
1062: for the connection $\nabla_{\cal G}$ along the $\R$-factor
1063: from $X \times \{1\}$ to $X \times \{0\}$.
1064: Take the standard Hodge filtration on the bigraded object
1065: $F^p_{\rm st}{\cal V}_\C:= \oplus_{i\geq p}{\cal V}^{i, *}$
1066: on the restriction of $p^*{\cal V}$ to $X \times \{1\}$.
1067: We define the Hodge filtration $F^{\bullet}$ on ${\cal V}_{\C}$ by applying
1068: the operator of parallel transport $P$ to the standard Hodge filtration on $p^*{\cal V}_{\C}$ at $u=1$:
1069: $$
1070: F^{p}{\cal V}_\C := P(F^p_{\rm st}{\cal V}_\C).
1071: $$
1072: We prove that $({\cal V}_{\infty}, \nabla^{0}_{\cal G}, W_{\bullet}, F^{\bullet})$
1073: is a variation of real
1074: mixed Hodge structures on $X $, and that any variation
1075: is obtained this way.
1076:
1077: \vskip 3mm
1078: Just recently M. Kapranov \cite{Ka} gave a neat interpretation of the
1079: category of mixed $\R$-Hodge structures as the
1080: category of $\C^*_{\C/\R}$-equivariant (not necessarily flat) connections
1081: on $\C$. His description is equivalent to
1082: the one by twistor connections when $X$ is a point. Precisely, homogeneous twistor
1083: connections (\ref{TCONH}), restricted to $\C = (\C^2)^{\sigma}$,
1084: are natural representatives of the
1085: gauge equivalence classes of $\C^*_{\C/\R}$-equivariant connections
1086: on $\C$.
1087:
1088:
1089: \subsection{The twistor transform, Hodge DG algebra,
1090: and variations of real MHS} \la{hc1.6}
1091: Theorem \ref{2.22.08.1} describes the category of
1092: variations of mixed $\R$-Hodge structures via twistor connections.
1093: The next step would be to describe complexes of
1094: variations of mixed $\R$-Hodge structures. Precisely,
1095: we want an alternative
1096: description of the subcategory ${\rm Sh}^{\rm sm}_{{\rm Hod}}(X)$
1097: of Saito's (derived) category ${\rm Sh}_{{\rm Hod}}(X)$ of mixed $\R$-Hodge sheaves on $X$
1098: consisting of complexes
1099: whose cohomology are smooth, i.e. are
1100: variations of mixed $\R$-Hodge structures.
1101: We are going to define a DG Lie coalgebra,
1102: and conjecture that the DG-category of comodules over it
1103: is a DG-enhancement of the triangulated category of smooth mixed $\R$-Hodge sheaves on a compact complex manifold $X$.
1104: To include the case of a smooth open complex variety $X$, one needs to impose
1105: conditions on the behavior of the forms forming the DG Lie coalgebra at infinity.
1106: \vskip 3mm
1107:
1108:
1109: {\it The semi-simple tensor category ${\rm Hod}_X$}.
1110: Let $X$ be a regular complex projective variety.
1111: Denote by ${\rm Hod}_X$ the
1112: category of variations of real Hodge structures on $X(\C)$.
1113: Its objects are direct sums of variations of real pure
1114: Hodge structures of possibly different weights. The category
1115: ${\rm Hod}_X$
1116: is a semi-simple abelian
1117: tensor
1118: category. Denote by $w({\cal L})$ the weight of a pure variation
1119: ${\cal L}$.
1120: The commutativity morphism is given by
1121: $
1122: {\cal L}_1 \otimes {\cal L}_2 \lra
1123: (-1)^{w({\cal L}_1)w({\cal L}_2)}{\cal L}_2 \otimes {\cal L}_1.
1124: $
1125:
1126:
1127:
1128:
1129: \vskip 3mm
1130: {\it Hodge complexes}.
1131: Given a variation ${\cal L}$ of pure Hodge structures on
1132: $X$,
1133: we introduce a complex ${\cal C}^{\bullet}_{\cal H}({\cal L})$ and show that it
1134: calculates ${\rm RHom}_{{\rm Sh}_{{\rm Hod}}(X)}(\R(0), {\cal L})$
1135: in the category of
1136: mixed $\R$-Hodge sheaves on $X$.
1137: We call it the {\it Hodge complex} of ${\cal L}$.
1138: When ${\cal L} = \R(n)$ it is
1139: the complex calculating the weight $n$ Beilinson-Deligne
1140: cohomology of $X$.
1141:
1142: \vskip 3mm
1143: The smooth de Rham complexes of variations of real
1144: Hodge structures on $X$ can be organized into a commutative
1145: DG algebra in the category
1146: ${\rm Hod}_X$, called the de Rham DGA:
1147: $$
1148: {\cal A}_{X}:= \oplus_{{\cal L} } {\cal A}^{\bullet}({\cal L})\bigotimes
1149: {\cal L}^{\vee},
1150: $$
1151: where ${\cal L}$ runs through the isomorphism
1152: classes of simple objects in the category
1153: ${\rm Hod}_X$. The product is given by the wedge
1154: product of differential forms and tensor product of variations.
1155:
1156:
1157: We are going to define a commutative
1158: DG algebra ${\cal D}_{X}$ in the
1159: category ${\rm Hod}_X$. As an object, it is
1160: a direct sum
1161: of complexes
1162: \be \la{complex}
1163: {\cal D}_{X}:= \oplus_{{\cal L} } {\cal C}^{\bullet}_{\cal H}({\cal L})\bigotimes
1164: {\cal L}^{\vee},
1165: \ee
1166: where ${\cal L}$ runs through the isomorphism
1167: classes of simple objects in the category
1168: ${\rm Hod}_X$. To define a DGA structure on ${\cal D}_{X}$, we
1169: recall the projection $p:X \times \R \to X$, and introduce
1170: our main hero:
1171: \vskip 3mm
1172: {\it The twistor transform}.
1173: It is a linear map
1174: $$
1175: \gamma: \mbox{The Hodge complex of ${\cal V}$ on $X$} \lra
1176: \mbox{The de Rham complex of $p^*{\cal V}$ on $X\times \R$},
1177: $$
1178: Its definition is very similar to the formula
1179: (\ref{TCON}) / (\ref{TCONH}) for
1180: the twistor connection --
1181: see Definition \ref{12.09.ups.15}.
1182: The twistor transform is evidently injective, and gives
1183: rise to an injective linear map
1184: \be \la{12.16.ups.1}
1185: \gamma: {\cal D}_X \hra {\cal A}_X.
1186: \ee
1187: Our key result in Section \ref{hc3sec} is the following theorem.
1188: \bt \la{12.8.ups.1s} Let $X$ be a complex manifold. Then the image of the
1189: twistor transform (\ref{12.16.ups.1}) is
1190: closed under the differential and product, and thus
1191: is a DG subalgebra of the de Rham DGA.
1192: \et
1193:
1194: Theorem \ref{12.8.ups.1s} provides ${\cal D}_X$ with
1195: a structure of a commutative DGA in the category ${\rm Hod}_X$.
1196: We call it the {\it Hodge DGA}
1197: of the complex variety $X$.
1198: The
1199: differential $\delta$ on ${\cal D}_X$ is obtained
1200: by conjugation by a degree-like operator $\mu$, see (\ref{mu}), of the
1201: natural differential on (\ref{complex}).
1202:
1203: It is surprising that one can
1204: realize the Hodge DGA inside of the de Rham DGA on
1205: $X\times \R$.
1206:
1207: \vskip 3mm
1208: Given a semi-simple abelian tensor category ${\cal P}$, denote by
1209: ${DGCom}_{\cal P}$ and $DGCoLie_{\cal P}$ the categories of DG commutative
1210: and Lie coalgebras in the category ${\cal P}$.
1211: Recall the two standard functors
1212: $$
1213: {\cal B}: DGCom_{\cal P} \lra DGCoLie_{\cal P}, \qquad {\cal C}:
1214: DGCoLie_{\cal P} \lra DGCom_{\cal P}.
1215: $$
1216: The functor ${\cal B}$ is the bar construction followed
1217: by projection to the indecomposables. The functor ${\cal C}$ is
1218: given by the Chevalley standard complex:
1219: $$
1220: ({\cal G}^{\bullet}, d) \lms {\cal C}({\cal G}):= ({\rm Sym}^*({\cal G}^{\bullet}[1]), \delta), \qquad
1221: \delta:= d_{Ch} + d.
1222: $$
1223:
1224: Applying the functor ${\cal B}$ to the commutative DGA
1225: ${\cal D}_{X}$ in the category
1226: ${\rm Hod}_X$ we get a DG Lie coalgebra ${\cal L}_{{\cal H}; X}^*:=
1227: {\cal B}({\cal D}_{X})$ in the same
1228: category. Let ${\cal L}_{{\cal H}; X}:= H^0({\cal L}_{{\cal H}; X}^*)$ be the Lie coalgebra
1229: given by its zero cohomology.
1230: We show that
1231: Theorem \ref{2.22.08.1} is
1232: equivalent to
1233: \bt\la{12.09.ups.20}
1234: The category of comodules over the Lie coalgebra ${\cal L}_{\cal H; X}$ in the category
1235: ${\rm Hod}_X$ is canonically equivalent to the
1236: category of variations of mixed $\R$-Hodge structures on $X$.
1237: The equivalence is given by the functor
1238: ${\rm gr}^W$ of
1239: the associate graded for the weight filtration.
1240: \et
1241:
1242: When $X$ is a point it provides a
1243: canonical set of generators for the Lie algebra of the Hodge Galois group.
1244: The DGA ${\cal D}_X$ contains as a
1245: sub DGA the Hodge-Tate algebra of A.Levin \cite{L}, and our
1246: description of variations of Hodge-Tate structures essentially
1247: coincides with the one
1248: in {\it loc. cit}.
1249:
1250:
1251:
1252:
1253: \bcon \la{12.1.ups.7}
1254: The category ${\rm Sh}^{\rm sm}_{\rm Hod}(X)$ of smooth
1255: complexes of
1256: real Hodge sheaves on $X$
1257: is equivalent to the DG-category of DG-modules over the DG Lie coalgebra
1258: ${\cal L}^*_{{\cal H}; X}$.
1259: \econ
1260:
1261:
1262:
1263: \subsection{Mixed $\R$-Hodge structure on $\pi_1^{\rm nil}$ via Hodge correlators}
1264:
1265:
1266: Recall that ${\rm L}_{X,S^*}$, see (\ref{varl}),
1267: is a free Lie algebra generated by
1268: ${\rm gr}^WH^1(X-S)$. When the data $(X, S, v_0)$ varies, the latter
1269: forms a
1270: variation ${\cal L}$ of real Hodge structures over the
1271: enhanced moduli space ${\cal M}'_{g, n}$, $n=|S|$,
1272: (we add a tangent vector $v_0$ to the standard data).
1273: We are going to equip it with a Green datum, thus getting a variation
1274: mixed $\R$-Hodge structures.
1275:
1276: \vskip 3mm
1277: A point $s_0\in X $ determines a $\delta$-current
1278: given by the evaluation of a test
1279: function at $s_0$. Viewed as
1280: a generalized volume form $\mu$, it provides a Green function $G(x,y)$
1281: up to a constant. We use the tangent vector $v_{0}$ at $s_0$
1282: to specify the constant. Namely, let
1283: $t$ be a local parameter at $s_0$ with $\langle dt, v_0\rangle=1$.
1284: We normalize the Green function so that
1285: $G(s_0, s) - \log |t|$ vanishes at $s=s_0$.
1286:
1287: Let ${\Bbb G}_{v_0}$ be the corresponding Green
1288: operator (\ref{5.22.06.1qa}). It
1289: is an endomorphism of the smooth bundle ${\cal L}_{\infty}$.
1290: Thus the construction described in Section \ref{1.5} provides
1291: a collection of mixed $\R$-Hodge structures in the fibers of ${\cal L}$.
1292:
1293: Furthermore, the Hodge correlators in families
1294: deliver in addition to the Green operator ${\Bbb G}_{v_0}$
1295: an operator valued $1$-form $\nu$.
1296: We prove in Section \ref{hc6sec}
1297: that the pair ${\bf G}_{v_0}:= ({\Bbb G}_{v_0}, \nu)$
1298: satisfies the Green data differential equations. Therefore
1299: we get
1300: a variation of mixed $\R$-Hodge structures.
1301:
1302: \begin{theorem} \label{9.16.05.2} When the data
1303: $(X, S, v_0)$ varies,
1304: the
1305: variation of mixed $\R$-Hodge structures given by the Green datum
1306: ${\bf G}_{v_0}$
1307: is isomorphic to the
1308: variation formed by the standard
1309: mixed $\R$-Hodge structures on $\pi_1^{\rm nil}(X -S, v_0)$.
1310: \end{theorem}
1311:
1312: \vskip 3mm
1313: The standard MHS on $\pi_1^{\rm nil}(X -S, v_0)$ is described by
1314: using Chen's theory of iterated integrals \cite{Ch}.
1315: Our approach is different: it is
1316: given by integrals of non-holomorphic differential forms over products
1317: of copies of $X $. Theorem \ref{9.16.05.2} implies that the two sets
1318: of periods obtained in these two descriptions coincide. For example
1319: the periods of the real MHS on $\pi_1^{\rm nil}({\Bbb P}^1-\{0,1,\infty\}, v_{0})$,
1320: where $v_0 = \partial/\partial t$ at $t=0$,
1321: are given by the multiple $\zeta$-values \cite{DG}. So Theorem
1322: \ref{9.16.05.2} implies that
1323: the Hodge correlators in this case are $\Q$-linear combinations of the
1324: multiple $\zeta$-values, which is far from being obvious from their
1325: definition.
1326:
1327: Here is another benefit of our approach: for a modular curve $X$,
1328: it makes obvious that the Rankin-Selberg convolutions are periods of the
1329: motivic fundamental group of $X- \{\mbox{the cusps}\}$.
1330:
1331:
1332:
1333: \input{hc1.tex} %Introduction
1334: \input{hc2.tex} %\label{hc2sec} Hodge correlators
1335: \input{hc3.tex} %\label{hc5asec} %The Hodge DG algebra
1336: \input{hc4.tex} %\label{hc4sec} %Variations of real MHS via Green operators
1337: \input{hc5.tex} %\label{hc5sec} %Plane trees and Lie coalgebras
1338: \input{hc6.tex} %\label{hc6sec} Differential equations for HC
1339: \input{hc7.tex} %\label{hc9sec} Cyclic words and special derivations
1340: \input{hc9.tex} %\label{hc12sec} Proof of Theorem 2.4
1341: \input{hc8.tex} %\label{hc10sec}Motivic correlators
1342: \input{hc10.tex}%\label{hc13sec} Examples.
1343: \input{hc11.tex}%\label{hc14sec} Motivic correaltors on modular curves
1344: \input{hc12.tex}%\label{fey12sec}Hodge correlators are Feynman integral correlators.
1345: %\input{hc13.tex}%\label{hc13sec}The Hodge vector fields
1346: %\input{hc2a.tex}
1347: % Hodge correlators are correlators of a Feynman integral
1348: %\input{hcm3.tex} \label{hc3sec}%Properties of the Hodge correlators
1349: %\input{hc5aa.tex}\label{hc5aasec}%Deformation picture
1350: %Differential equations for the Hodge correlators on curves
1351: %\input{hcm7.tex}%\label{hc7sec} The rational homotopy type DG Lie algebras
1352: %\input{hcm8.tex}%\label{hc8sec}
1353: %$\R$-mixed Hodge structure on rational homotopy type
1354: %\input{hcapp.tex} Green function
1355: %\section{Polylogarithms on curves and special values
1356: %$L({ S}^n{\cal H}(1), n+1)$} %\input{feyc.tex}
1357:
1358:
1359: \begin{thebibliography}{BL}
1360: \bibitem[B]{B} Beilinson A.A.: {\it Regulators and values of $L$-functions}. (Russian) Current problems in mathematics, Vol. 24, 181--238, Itogi Nauki i Tekhniki, Akad. Nauk SSSR, Vsesoyuz.
1361: Inst. Nauchn. i Tekhn. Inform., Moscow, 1984.
1362: \bibitem[B1]{B1} Beilinson A.A.: {\it Height pairings between algebraic cycles}. $K$-theory, arithmetic and geometry (Moscow, 1984--1986), 1--25, Lecture Notes in Math., 1289, Springer, Berlin, 1987.
1363: \bibitem[B2]{B2} Beilinson A.A.: {\it Higher regulators of modular curves}.
1364: Contemp. Math. 55, 1-34 (1986)
1365: \bibitem[BD]{BD} Beilinson A.A., Deligne P, {\it Interpr\'etation
1366: motivique de la conjecture de Zagier}. in Symp. in Pure Math., v. 55, part
1367: 2, 1994,
1368: \bibitem[BL]{BL} Beilinson A.A., Levin A.M. :
1369: {\it The elliptic polylogarithms},
1370: Proc. Symp. in Pure Math, vol. 55, (1994), part 2, 126-196.
1371: \bibitem[BD]{BD} Beilinson A.A., Deligne P. Unpublished preprint on motivic polylogarithms.
1372: \bibitem[BK]{BK} Bloch S., Kato K.: {$L$-functions and Tamagawa numbers of motives}. Grothendieck Festschrift,
1373: Vol 1, 333-400. Progress Math. 86, Birkh\"auser, Boston (1990)
1374: \bibitem[Bo1]{Bo1} Borel A.: {\it Cohomologie des espaces fibr\'{e}s
1375: principaux}, Ann.\ Math.\ 57 (1953), 115--207.
1376: \bibitem[Bo2]{Bo2} Borel A.: {\it Cohomologie de $SL_{n}$ et valeurs de
1377: fonctions z\^{e}ta aux points entiers}. Annali Scuola Normale
1378: Superiore Pisa 4 (1977), 613--636.
1379: \bibitem[Ch]{Ch} Chen, Kuo Tsai:
1380: {\it Iterated path integrals}. Bull. Amer.
1381: Math. Soc. 83 (1977), no. 5, 831--879.
1382: \bibitem[D]{D} Deligne P.: {\it Hodge theory II}.
1383: Publ. IHES No. 40 (1971), 5--57;
1384: \bibitem[D1]{D1} Deligne P.: {\it Le group fondamental de la droite projective moine trois points}, In: Galois groups over $\Bbb Q$. Publ. MSRI, no. 16 (1989) 79-298.
1385: \bibitem[D2]{D2} Deligne P.: {\it Structures des Hodges mixte r\'eeles}. Proc. Symp. in Pure Math,
1386: vol. 55, (1994), part 1, 509-516.
1387: \bibitem[DG]{DG} Deligne P., Goncharov, A. {\it Groupes fondamentaux
1388: motiviques de Tate
1389: mixte}.
1390: Ann. Sci. Ecole Normale
1391: Sup. math.NT/0302267.
1392: \bibitem[Den1]{Den1} Deninger Ch.: {\it Higher regulators and Hecke
1393: $L$-series of
1394: imaginary quadratic fields II}, Ann. of Math, 132, N1 (1990), 131-158.
1395: \bibitem[DS]{DS} Deninger Ch. Scholl, A.: {\it The Beilinson conjectures}. $L$-functions and arithmetic (Durham, 1989),
1396: 173--209, London Math. Soc. Lecture Note Ser., 153, Cambridge Univ. Press, Cambridge, 1991.
1397: %\bibitem[Den2]{Den2} Deninger Ch.:
1398: %{\it Higher operations in Deligne cohomology}.
1399: %Invent. Math. 120 (1995) N2, 289-315.
1400: \bibitem[Dr1]{Dr1} Drinfeld V. G.: {\it On quasitriangular quasi-Hopf algebras and on a group that is closely connected with ${\rm Gal}(\overline Q/Q)$}. Algebra i Analiz 2 (1990), no. 4, 149--181;
1401: Leningrad Math. J. 2 (1991), no. 4, 829--860.
1402: \bibitem[Dr2]{Dr2} Drinfeld V. G.: {\it Elliptic modules}. Mat. Sb. (N.S.) 94(136) (1974), 594--627.
1403: \bibitem[GK]{GK} Ginzburg V.; Kapranov M.: {\it Koszul duality for operads}. Duke Math. J. 76 (1994), no. 1, 203--272.
1404: \bibitem[G1]{G1} Goncharov A.B.:
1405: {\it Multiple $\zeta$-values, Galois groups and
1406: geometry of modular varieties}, European Congress of
1407: Mathematics, Vol. I (Barcelona,
1408: 2000), 361--392, Progr. Math., 201, Birkhauser, Basel.
1409: math.AG/0204102.
1410: \bibitem[G2]{G2} Goncharov A.B.:
1411: {\it Chow polylogarithms and regulators}, Math. Res. Letters,
1412: vol. 2. (1995), N 1, pp. 95-112.
1413: \bibitem[G3]{G3} Goncharov A.B.: {\it Mixed elliptic motives}. Galois representations
1414: in Arithmetic Algebraic geometry, London Math. Soc. Lect.
1415: Note Ser., vol. 254, Cambridge University Press,
1416: 1998. 147-221.
1417: \bibitem[G4]{G4} Goncharov A.B.: {\it The dihedral Lie algebras and Galois
1418: symmetries of
1419: $\pi^{(l)}_{1}({\Bbb P}^{1} \backslash \{0, \mu_N,\infty \}, v_{1} )$},
1420: Duke Math. J., vol 100, N3, (2001), pp. 397-487.
1421: math.AG/0009121.
1422: \bibitem[G5]{G5} Goncharov A.B.:
1423: {\it Galois groups, geometry of modular varieties and graphs}.
1424: Proc. of Arbeitstagung, June 1999, in the MPI preprint
1425: (http//www.mpim-bonn.mpg.de/).
1426: \bibitem[G6]{G10} Goncharov A.B.:
1427: {\it Euler complexes and geometry of modular varieties.}
1428: GAFA (2008). Special volume dedicated to J.Bernstein's 60th birthday.
1429: arXive: math/0510310.
1430: \bibitem[G7]{G7} Goncharov A.B.: {\it Galois symmetries of fundamental groupoids and noncommutative geometry}. Duke Math. J. 128 (2005), no. 2, 209--284, ArXive
1431: math.AG/0208144,
1432: \bibitem[G8]{G8} Goncharov A.B.: {\it Polylogarithms, regulators,
1433: and Arakelov motivic complexes}. JAMS (2005), N1, 1-60.
1434: \bibitem[G9]{G9} Goncharov A.B.: {\it Multiple polylogarithms, cyclotomy and modular complexes}. Math. Res. Lett. 5 (1998), no. 4, 497--516
1435: \bibitem[G10]{G10} Goncharov A.B.: {\it Polylogarithms and motivic Galois groups} Motives, (1995).
1436: \bibitem[GII]{GII} Goncharov A.B.: {\it Hodge correlators II}. Moscow Math Journal. Special issue dedicated
1437: to Deligne's 65th birthday. N1, (2010). ArXiv:0807.4855.
1438: \bibitem[GZ]{GZ} Goncharov A.B., Zhao J.: {\it The Grassmannian trilogarithm}.
1439: Compositio Math. 127 (2001), no. 1, 83--108.
1440: \bibitem[H]{H} Hain R.: {\it The geometry of the mixed Hodge structure on the fundamental group}. Algebraic geometry, Bowdoin, 1985, 247--282, Proc. Sympos. Pure Math., 46, Part 2, Amer. Math. Soc., Providence, RI, 1987.
1441: \bibitem[H1]{H1} Hain R.: {\it Classical polylogarithms}. Proceedings of Symp. in Pure Math., Volume 55, (1994), vol. 2, 3-42.
1442: \bibitem[HZ]{HZ} Hain R, Zucker S.:
1443: {\it Unipotent variations of mixed Hodge structure}. Invent. Math. 88 (1987), no. 1, 83--124.
1444: \bibitem[HW]{HW} Huber A., Wildeshaus J. : {\it Classical Motivic Polylogarithm according to Beilinson and Deligne}.
1445: Documenta Mathematica 3: 27-133, 1998.
1446: \bibitem[Ka]{Ka} Kapranov M.: {\it Real mixed Hodge structures.} Preprint 2008,
1447: arXiv:0802.0215.
1448: \bibitem[KM]{KM} Kashiwara M.: {\it A study of variation of mixed Hodge structure}. Publ. Res. Inst. Math. Sci. 22 (1986), no. 5, 991--1024.
1449: \bibitem[K]{K} Kontsevich M.: {\it Formal (non)commutative symplectic
1450: geometry}. The Gelfand mathematical seminars, Birkhauser, 1993, p. 173-187.
1451: \bibitem[K2]{K2} Kontsevich M.: {\it Operads and motives in deformation quantization}.
1452: Lett. Math. Phys., 48 (1999) N1, 35-72.
1453: \bibitem[L]{L} Levin A.: {\it Variations of $\R$-Hodge-Tate structures}. Preprint MPI-2001.
1454: \bibitem[SS]{SS} Schappacher N., Scholl A.: {\it Beilinson's theorem on modular curves}. Beilinson's
1455: conjectures on special values of $L$-functions, 273--304, Perspect. Math., 4, Academic Press, Boston, MA, 1988.
1456: \bibitem[S]{S} Suslin A.A.: {\it $K_{3}$ of a field and Bloch's
1457: group}, Proceedings of the Steklov Institute of Mathematics 1991, Issue 4.
1458: \bibitem[M]{M} Morgan J.: {\it The algebraic topology of smooth algebraic varieties}. Inst. Hautes ˇĆ‰tudes Sci. Publ. Math. No. 48 (1978), 137--204.
1459: \bibitem[Z1]{Z1} Zagier D.: {\it Polylogarithms, Dedekind zeta functions and the algebraic $K$-theory of fields}. Arithmetic algebraic geometry
1460: (Texel, 1989), 391--430, Progr. Math., 89, BirkhˇĆˇ¤user Boston, MA, 1991.
1461: \bibitem[We]{We} Weil A.: {\it Elliptic functions according to Eisenstein and
1462: Kronecker}. Ergebnisse der Mathematik, 88, Springer 1977.
1463: %Lecture Notes in Mathematics, 1650. Springer-Verlag, Berlin, 1997
1464: %\bibitem[W2]{W2} Wildeshaus J. : {\it On an elliptic analog of Zagier's conjecture}.
1465: % Duke Math. J. 87 (1997), no. 2, 355--407.
1466: \bibitem[Wit]{Wit} Witten E.: {\it Perturbative quantum field theory}.
1467: In: Quantum fields and Strings: A course for mathematicians. Volume 1.
1468: P. Deligne, \ldots editors. AMS (1999).
1469: \end{thebibliography}
1470:
1471: %\input{fey7.tex}
1472: %\input{hc7.tex}
1473: \end{document}
1474:
1475:
1476:
1477: 1. hc1.eps
1478: 2. feyn36.eps
1479: 3. feyn-10.eps
1480: 4. feynleit.eps
1481: 5. feyn9.eps
1482: 6. hc7.eps
1483: 7. feyn6.eps
1484: 8. hc11.eps
1485: 9. feyn17.eps
1486: 10. feyn18.eps
1487: 11. feyn19.eps
1488: 12. feyn20.eps
1489: 13. feyn37.eps
1490: 14. feyn23.eps
1491: 15. feyn24.eps
1492: 16. feyn25.eps
1493: 17. feyn17a.eps
1494: 18. feyn38.eps
1495: 19. hc8.eps
1496: 20. hc10.eps
1497: 21. hc12.eps
1498: 22. fey100.eps
1499: 23. hc2.eps
1500: 24. feyn15.eps
1501: 25. hc16.eps
1502: 26. feyn.eps
1503: 27. hc17.eps
1504: 28. hc19.eps
1505: 29. feyn16.eps
1506: 30. hc21.eps
1507: 31. feyn32.eps
1508: 32. feyn10.eps
1509: 33. feyn33a.eps
1510: 34. feyn11.eps
1511: 35. feyn12.eps
1512: 36. feyn26.eps
1513: