0807.4855/hf1.tex
1: 
2: \section{Introduction} 
3: 
4: \subsection{Summary} Let $X$ be a compact Kahler manifold of dimension $n$. 
5: Denote by ${\cal C}{\cal C}_\bullet(X)$ 
6: the cyclic homology complex of the reduced cohomology algebra of $X$. 
7: We define a linear map, the {\it Hodge correlator map}: 
8: \be \la{1}
9: {\rm Cor}_{\cal H}: H_0\Bigl({\cal C}{\cal C}_\bullet(X)\otimes H_{2n}(X)[-2]\Bigr) \lra \C.
10: \ee
11: Evaluating this map on cycles  
12: %which are given by (certain sums of) cyclic products 
13: %${\cal C}(h^1 \otimes ... \otimes h^m)$ of cohomology classes $h^i$ with
14: %$\sum_i ({\rm deg}(h^i) -1) =2n-2$, 
15: we get complex numbers, called 
16: {\it Hodge correlators}. 
17: 
18: 
19: We introduce 
20: a Feynman integral related to $X$. We show that 
21: its correlators, 
22: defined via the standard perturbative series 
23: expansion procedure, are well defined, and identify them with the Hodge correlators. 
24: We show that the Hodge correlator map defines a functorial real mixed Hodge structure of the 
25: rational homotopy type of $X$.   The Hodge correlators are the homotopy periods of $X$.
26: 
27: \vskip 2mm
28: The category of real mixed Hodge structures (MHS) is canonically identified 
29: with the category of representations of the Hodge Galois group. 
30: We show that the Hodge correlator map provides  
31: an explicit  action 
32: of the Hodge 
33: Galois group on the formal neighborhood of the trivial local system on $X$. 
34: It
35: encodes a functorial homotopy action 
36: of the Hodge 
37: Galois group on the rational homotopy type of $X$, 
38: describing the real MHS on the latter.  
39: 
40: A real MHS on the rational homotopy type of a  
41: complex algebraic variety was defined, by different methods, 
42:  by J. Morgan \cite{M} and R. Hain \cite{H}. Our construction should give
43:  the same real MHS. We prove this for the 
44: real MHS on the 
45: pronilpotent completion of the fundamental group of $X$. 
46: 
47: \vskip 3mm
48: Now let $X$ be a regular projective algebraic 
49: variety over a field $k$. Assuming the motivic formalism,  we define 
50: {\it motivic correlators of $X$}. They lie in the motivic Lie coalgebra of $k$. 
51: Given an embedding $k \hra \C$, the periods 
52: of the Hodge realization of the motivic correlators are the Hodge correlators. 
53: The motivic correlators come together with an explicit formula for their coproduct 
54: in the motivic Lie coalgebra of $k$. 
55: This is one of the main advantages of presenting homotopy periods 
56: of $X(\C)$ as periods of the motivic correlators. Using this one can perform 
57: {arithmetic analysis} 
58: of the homotopy periods of varieties defined over number fields. 
59: Summarising: 
60: 
61: \begin{figure}[ht]
62: \centerline{\epsfbox{hf10.eps}}
63: \label{hc1}
64: \end{figure}
65: A similar  picture for a smooth 
66: but not necessarily compact 
67: complex curve $X$ was obtained in \cite{G1}. 
68: In that case 
69: there are natural harmonic representatives for ${\rm gr}^WH^{\ast}(X)$. 
70: This allows to avoid discussion of DG objects, and makes the story simpler. 
71: 
72: The results of this paper admit a generalization where 
73: $X$ is replaced by an arbitrary regular complex algebraic variety, 
74: which we hope to  discuss elsewhere.  
75: %In Section 7 we work out a part of this picture, 
76: %defining a real MHS on the pronilpotent completion 
77: %of the fundamental group of an arbitrary smooth complex variety $Y$ via Hodge correlators 
78: %for $Y$. 
79: 
80: \vskip 2mm
81: In the rest of the Introduction we give a rather detailed account 
82: of our constructions and results. Complete details are available in
83:  the main body of the paper. 
84: 
85: \subsection{Hodge correlators for a compact Kahler manifold}  
86: Given a graded vector space $V$, denote by ${\cal C}_{V}$ the cyclic tensor envelope of 
87: $V$:
88: $$
89: {\cal C}_{V}:= \oplus_{m=0}^{\infty}\Bigl(\otimes^mV\Bigr)_{\Z/m\Z}
90: $$
91: where the subscript $\Z/m\Z$ denotes the coinvariants of the cyclic shift 
92: $$
93: v_1 \otimes v_2 \otimes ... \otimes v_m \lms 
94: (-1)^s v_m \otimes v_1 \otimes ... \otimes v_{m-1}, \qquad s = {\rm deg}(v_m)
95: \sum_{i=1}^{m-1}{\rm deg}(v_i). 
96: $$
97: We denote by $(v_1 \otimes v_2 \otimes ... \otimes v_m)_{\cal C}$ 
98: the projection of the element $v_1 \otimes v_2 \otimes ... \otimes v_m$ to ${\cal C}_{V}$. 
99: Let
100: $$
101: \overline {\rm H}_*(X):= \oplus_{i=1}^{2n-1}{\rm H}_i(X),
102: $$
103: be the reduced cohomology algebra of $X$. Let us shift it to the right by one:
104: $$
105: {\Bbb H}_*:= \overline {\rm H}_*(X)[-1]. 
106: $$
107: It sits 
108: in the degrees $[-(2n-1), -1]$.  
109: We also need the dual object:
110:   $$
111: {\Bbb H}^*:= \overline {\rm H}^*(X)[1], \qquad 
112: \overline {\rm H}^*(X):= \oplus_{i=1}^{2n-1}{\rm H}^i(X).
113: $$
114: The spaces ${\cal C}_{{\Bbb H}^*}$ and ${\cal C}_{{\Bbb H}_*}$ are graded dual 
115: to each other. 
116: 
117: \noindent
118: Finally, let us introduce the shifted by two fundamental cohomology class of $X$:
119: \be \la{sfc}
120:  {\cal H}:= H^{2n}(X)[2]. 
121: \ee
122:  Let  $ {\cal H}^{\vee}$ is its dual.
123: 
124: \vskip 3mm 
125: Choose a base point $a \in X$. 
126: Choose a splitting of the de Rham complex of $X$, bigraded in the usual way, into 
127: an arbitrary subspace ${\cal H}ar_X$ isomorphically projecting onto 
128: the cohomology of $X$ ("harmonic forms'') 
129: and its orthogonal complement. 
130: We take the $\delta$-function $\delta_a$ 
131: at the point $a \in X$ 
132: as a representative of the fundamental class. 
133: 
134: 
135: Our first goal is to define a degree zero linear map, 
136: a precursor of the Hodge correlator map: 
137: \be \la{1.17.08.10q}
138: {\rm Cor}^*_{{\cal H}, a}: {\cal C}_{{\Bbb H}^*} \otimes 
139: {\cal H}^{\vee} \lra \C.
140: \ee
141: 
142: 
143: \vskip 2mm
144: We need a Green current $G_a(x,y)$ on $X \times X$. 
145: It satisfies the differential equation 
146: \be \la{GEQ}
147: (2\pi i)^{-1}\overline \partial 
148: \partial G_a(x,y) = \delta_{\Delta} - P_{\rm Har}
149: \ee
150: where $\delta_{\Delta}$ is the $\delta$-function of the diagonal, and 
151: $P_{\rm Har}$ is the Schwarz kernel of the projector onto the space of harmonic forms, 
152: realized by an $(n,n)$-form on $X \times X$. Given a basis $\{\alpha_i\}$ 
153: in the space of ``harmonic forms'' of dimensions $\not = 0, 2n$, and the dual basis 
154: $\{\alpha^{\vee}_i\}$, we have 
155: $$
156: P_{\rm Har} = \delta_a \otimes 1 + 1 \otimes \delta_a + \sum \alpha^{\vee}_i \otimes 
157: \alpha_i, \qquad \int_X  \alpha_i \wedge 
158: \alpha_j^{\vee} = \delta_{ij}.
159: $$
160:  The two currents on the right hand side of (\ref{GEQ}) represent 
161: the same cohomology class, so the equation has a solution by the 
162: $\overline \partial \partial$-lemma. 
163: Let us choose such a solution. 
164: 
165: \vskip 2mm
166: \begin{figure}[ht]
167: \centerline{\epsfbox{hf5.eps}}
168: \caption{A plane  trivalent tree decorated by harmonic forms $\alpha_i$.}
169: \label{hf5}
170: \end{figure}
171: Let us pick a cyclic tensor product 
172: of harmonic forms
173: $$
174: W= (\alpha_0 \otimes \ldots \otimes \alpha_m)_{\cal C}.  
175: $$
176:  Take a plane trivalent tree 
177: $T$ decorated by $W$, see Fig \ref{hf5}. Let us  cook up 
178: a  top degree current 
179: on 
180: \be \la{4.30.08.1}
181: X ^{\{\mbox{\rm internal vertices of $T$}\}}.
182: \ee
183:  Let ${\cal A}^*_M[-1]$ be the de Rham complex of $M$, 
184: shifted by $1$ to the right. 
185: There is a degree zero linear map 
186: \be \la{xixi}
187: \xi: S^{*}({\cal A}^*_M[-1]) \to {\cal A}^*_M[-1], \qquad \varphi_0 \cdot \ldots \cdot
188: \varphi_m 
189: \lms {\rm Sym}_{m+1}\Bigl(\varphi_0 \wedge d^\C \varphi_1 \wedge \ldots  \wedge d^\C 
190: \varphi_m\Bigr). 
191: \ee 
192: Take a decorating harmonic form $\alpha_i$ assigned to 
193: an external edge $E_i$ of the tree $T$. Put the form $\alpha_i$ 
194:  to the copy of $X$ assigned  
195: to the internal vertex of the edge $E_i$. 
196: Apply the operator $\xi$  to the wedge product of 
197: the Green currents assigned to  
198: the edges of $T$. Finally,  multiply 
199: the obtained currents on (\ref{4.30.08.1}) 
200: with an appropriate sign, discussed in Section 2.3. One shows that we get a current. 
201: Integrating it over (\ref{4.30.08.1})  we 
202: get a number assigned to $T$. Taking the sum over all trees $T$, we get 
203: a complex number ${\rm Cor}^*_{{\cal H}, a}(W)$. Altogether, we 
204: get the map (\ref{1.17.08.10q}). One checks that its degree is zero. 
205: 
206: 
207: \vskip 3mm
208: The $\cup$-product in ${\rm H}^{*}(X)$, followed by the projection to $\overline {\rm H}^*(X)$, 
209: induces an algebra structure on $\overline {\rm H}^*(X)$. 
210: There is a differential $\delta$ on ${\cal C}_{{\Bbb  H}^*}$,  
211: provided by products of neighbors in a cyclic word:
212: $$
213: \delta (\overline \alpha_0 \otimes \ldots \otimes \overline \alpha_m)_{\cal C} = 
214: {\rm Cycle}_{m+1}
215: (-1)^{|\alpha_m|}(\overline \alpha_0 \otimes \ldots \otimes \overline \alpha_{m-2} 
216: \otimes \overline {\alpha_{m-1}\cup \alpha_m})_{\cal C}.   
217: $$
218: Here ${\rm Cycle}_{m+1}$ means the sum of  cyclic shifts, 
219: $\overline \alpha \in {\Bbb H}^*$ is the shifted by one 
220: element $\alpha \in \overline {\rm H}^*(X)$, and $|\overline \alpha|$ is its degree.
221: The complex $({\cal C}_{{\Bbb  H}^*}, \delta)$ is nothing else but 
222: the {\it cyclic homology complex} for the algebra 
223: $\overline {\rm H}^*(X)$. 
224: %The differential $\delta$ induces a differential on 
225: %${\cal C}_{{\Bbb  H}^*}\otimes {\cal H}^{\vee}$. 
226: 
227: 
228: \vskip 2mm
229: There is a subspace of {\it shuffle relations} in 
230: ${\cal C}_{{\Bbb H}_*}$  generated by the 
231: elements
232: $$
233: \sum_{\sigma \in \Sigma_{p,q}}\pm (v_0 \otimes v_{\sigma(1)} 
234: \otimes \ldots \otimes v_{\sigma(p+q)})_{\cal C}, 
235: \qquad p, q \geq 1, 
236: $$ 
237: where the sum is over all $(p,q)$-shuffles, and 
238: the signs are given by the standard sign rule. 
239: Set
240: $$ 
241: {\cal C}{{\cal L}ie}^{\vee}_{{\Bbb H}^*}:= 
242: \frac{ {\cal C}_{{\Bbb H}^*}}{\mbox{Shuffle relations}};
243: \qquad 
244: {\cal C}{{\cal L}ie}_{{\Bbb H}_*}:= \mbox{the dual of 
245: ${\cal C}{{\cal L}ie}^{\vee}_{{\Bbb H}^*}$}.
246: $$
247: 
248:   Let ${\cal L}ie_{{\Bbb H}_*}$ be the free graded Lie algebra 
249:  generated by ${\Bbb H}_*$. The space ${\cal C}{{\cal L}ie}_{{\Bbb H}_*}$ 
250: is the projection to the cyclic envelope of the space ${\cal L}ie_{{\Bbb H}_*} \otimes {\Bbb H}_*$. 
251: 
252: 
253: 
254: 
255: 
256: \bt \la{MAATH} The map (\ref{1.17.08.10q}) induces a well defined linear map, 
257: called the {\rm Hodge correlator map}: 
258: \be \la{1.17.08.10as}
259: {\rm Cor}^*_{{\cal H}, a}: H^0_{\delta}\Bigl(
260: {\cal C}{\cal L}ie^{\vee}_{{\Bbb H}^*}\otimes {\cal H}^{\vee}\Bigr) \lra \C. 
261: \ee
262: -- It 
263: does not depend on the choices involved in the definition of the map 
264: (\ref{1.17.08.10q}). 
265: \et
266: We prove this in Section \ref{hf4.1ref}, 
267: as a consequence of an equivalent Theorem 
268: \ref{3/11/07/101q} proved there. 
269: \vskip 3mm
270: Now let $X$ be the set of complex 
271: points of an algebraic variety $X_{/\Q}$ over $\Q$. Then 
272: the rational de Rham cohomology $H^*_{\rm DR}(X_{/\Q}, \Q)$
273:  induce a $\Q$-rational structure on the 
274: space on the left in (\ref{1.17.08.10as}).  So we get a collection 
275: of numbers, the images  
276: of the elements of this $\Q$-vector space under the Hodge correlator map. 
277: We  call them  {\it the real periods of the homotopy type of $X$}. 
278: 
279: 
280: \vskip 2mm
281: Dualising map (\ref{1.17.08.10as}) we get a homology class, 
282: {\it the Hodge correlator class}: 
283: \be \la{HCCL}
284: {\bf H}^*_{X, a}\in H_0^{\delta}\Bigl(
285: {\cal C}{\cal L}ie_{{\Bbb H}_*}\otimes {\cal H}\Bigr).
286: \ee
287: 
288: 
289: 
290: 
291: \paragraph{Describing special derivations.} 
292: Let 
293: $
294: \delta: {\Bbb H}_* \lra \Lambda^2{\Bbb H}_*
295: $ 
296: be the degree $1$ map  dualising the product on $\overline {\rm H}^*(X)$. 
297: See Section 3.2 for a discussion of signs. 
298: It gives rise to a differential $\delta$ 
299: of the free Lie algebra 
300:  ${\cal L}ie_{{\Bbb H}_*}$, providing it with a DG Lie algebra structure. 
301: 
302: 
303: The image of the shifted by one fundamental class ${\rm H}_{2n}(X)[-1]$ 
304: under the map dualising the commutative product map 
305: ${\rm H}^*(X) \otimes {\rm H}^*(X) \to {\rm H}^{2n}(X)$ 
306: provides a special element ${\rm S} \in {\cal L}ie_{{\Bbb H}_{*}}$, 
307: better understood as a map
308: $$
309: {\rm S}: {\rm H}_{2n}(X)[-1] \lra [{\Bbb H}_*(X), {\Bbb H}_*(X)] \subset {\cal L}ie_{{\Bbb H}_{*}}.
310: $$
311: 
312: \bd
313: A derivation of the Lie subalgebra ${\cal L}ie_{{\Bbb H}_{*}}$ 
314: is special if it kills the element ${\rm S}$. 
315: We denote by ${\rm Der}^S({\cal L}ie_{{\Bbb H}_*})$ the Lie algebra of all special derivations of 
316: the Lie algebra ${\cal L}ie_{{\Bbb H}_*}$. 
317: \ed
318: 
319: \vskip 2mm
320: 
321: 
322: The commutator with the differential $\delta$ on ${\cal L}ie_{{\Bbb H}_*}$  provides the Lie algebra 
323: ${\rm Der}^S{\cal L}ie_{{\Bbb H}_*}$ with a DG Lie algebra structure.  
324: In Section 3.2 we show the following
325: 
326: \bp \la{1.17.08.10as1sd} (i) There is a natural isomorphism 
327: \be \la{1.17.08.10as1}
328: \theta: {\cal C}{\cal L}ie_{{\Bbb H}_*}\otimes {\cal H} \stackrel{\sim}{\lra}
329: {\rm Der}^S{\cal L}ie_{{\Bbb H}_*}. 
330: \ee
331: (ii) There is a Lie algebra structure on  
332: ${\cal C}{\cal L}ie_{{\Bbb H}_*}\otimes {\cal H}$ which, 
333: together with the differential $\delta$, makes it into a DG Lie algebra. The 
334: isomorphism (\ref{1.17.08.10as1})
335: is an isomorphism of DG Lie algebras.
336: \ep
337: We prove it in Section 3.2. 
338: In particular, there is  an isomorphism of Lie algebras
339: \be \la{1.17.08.10as12}
340: \theta_0: H_0^\delta\Bigl({\cal C}{\cal L}ie_{{\Bbb H}_*}\otimes {\cal H}\Bigr) \stackrel{\sim}{\lra} 
341: H_0^\delta\Bigl({\rm Der}^S{\cal L}ie_{{\Bbb H}_*}\Bigr). 
342: \ee
343: It  is crucial for our definition of motivic correlators. 
344: 
345: 
346: 
347: Here is a sketch of constructions leading to a proof of Proposition \ref{1.17.08.10as1sd}. 
348: First, given an element $F\otimes {\cal H} \in {\cal C}_{{\Bbb H}_*}\otimes {\cal H}$, 
349: there is a derivation $\theta_{F\otimes {\cal H}}$ of the tensor algebra 
350: ${\rm T}_{{\Bbb H}_*}$ acting on the generators as follows:
351: $$
352: \theta_{F\otimes {\cal H}}: 
353: p \lms \sum_q \frac{\partial F}{\partial q}\langle q \cap {\cal H} \cap p \rangle. 
354: $$
355: Here $\frac{\partial F}{\partial q}$ is the non-commutative 
356: partial derivative, e.g. $$
357: \frac{\partial}{\partial q_1}(q_1 q_2 q_1 q_3)_{\cal C} = q_2 q_1 q_3 + q_3q_1 q_2 ,
358: $$ the sum is over a basis $\{q\}$ in ${\Bbb H}_*$, 
359: and $\langle q \cap {\cal H} \cap p \rangle$ 
360: is the iterated cap product with the fundamental class ${\cal H}$. 
361: Second, the condition  
362: $F \in {\cal C}{\cal L}ie_{{\Bbb H}_*}\otimes {\cal H}$ just means that 
363: $\theta_{F\otimes {\cal H}}$ preserves the Lie subalgebra ${\cal L}ie_{{\Bbb H}_*}$. 
364: Finally,  $\theta_{F\otimes {\cal H}}({\rm S}) =0$, i.e. it is a special derivation.
365: 
366: 
367: 
368: 
369: \vskip 2mm
370: There is a Lie cobracket on the space 
371: $
372: {\cal C}_{{\Bbb H}^*}\otimes {\cal H}^{\vee},  
373: $ 
374: making it into  a graded Lie coalgebra, and 
375: $({\cal C}_{{\Bbb H}^*}\otimes {\cal H}^{\vee}, \delta)$ into
376:  a DG Lie coalgebra. Namely, take a cyclic word, 
377: cut it into two pieces, insert the Casimir element 
378: $\sum_i\alpha_i^{\vee} \otimes \alpha_i$ 
379: and take the sum over all cuts with appropriate signs, as on Fig \ref{hf7}. 
380: \begin{figure}[ht]
381: \centerline{\epsfbox{hf7.eps}}
382: \caption{The Lie cobracket on ${\cal C}_{{\Bbb H}^*}\otimes {\cal H}^{\vee}$. 
383: The even factors ${\cal H}^{\vee}$ are not shown.}
384: \label{hf7}
385: \end{figure} 
386: Thus the dual 
387: ${\cal C}_{{\Bbb H}_*}\otimes {\cal H}$ is a DG Lie algebra. 
388: Furthermore, 
389: ${\cal C}{{\cal L}ie}^{\vee}_{{\Bbb H}^*} \otimes {\cal H}^{\vee}$ 
390: inherits a DG Lie coalgebra structure. 
391: So  
392: ${\cal C}{{\cal L}ie}_{{\Bbb H}_*}\otimes {\cal H}$ is a DG Lie algebra.
393: Therefore   
394: $H^0_{\delta}\Bigl({\cal C}{{\cal L}ie}^{\vee}_{{\Bbb H}^*}\otimes {\cal H}^{\vee}\Bigr)$  
395: is a Lie coalgebra. 
396: 
397: 
398: 
399: 
400: 
401: \paragraph{Functoriality of the Hodge correlator classes.} Given a map $f:X \to Y$ of Kahler manifolds, 
402: the map $f_*: H_*(X) \to H_*(Y)$ induces a Lie algebra map 
403: $$
404: f_*: {\cal C}{\cal L}ie_{{\Bbb H}_*(X)} \to 
405: {\cal C}{\cal L}ie_{{\Bbb H}_*(Y)}. 
406: $$
407: It commutes with the differentials (Lemma \ref{6.13.08.1}). Here is our key result:
408: 
409: 
410: \bt \la{PPPq} 
411: For any map $f:X \to Y$ of Kahler manifolds 
412: the derivation ${\bf H}^*_{Y, f(a)}$ preserves the 
413: Lie subalgebra $f_*({\cal L}ie_{{\Bbb H}_*(X)})$, and its restriction to the latter 
414: is  homotopic to  $f_*({\bf H}^*_{X, a})$. 
415: \et
416: 
417: Theorem 
418: \ref{MAATH} can be viewed as a special case of 
419: Theorem \ref{PPPq} when $f:X \to X$ is the identity map, but we use
420: two different splittings/Green currents to 
421: define the Hodge correlators on $X$. We prove Theorem \ref{PPPq} in Section \ref{hf4.2ref}. 
422: We prove a more precise Theorem \ref{7.12.08.1}, calculating the coboundary explicitly.\footnote{There 
423: is a different, less computational proof, which we discuss elsewhere. 
424: We also hope to give a different proof using the Feynman integral from Section \ref{hf1.6}  and 
425: the BV formalism elsewhere.} 
426: 
427: 
428: 
429: \vskip 2mm
430: 
431: {\bf Remark}. Let $({\cal L}_\bullet, \delta)$ be a DG Lie algebra with a differential 
432: $\delta$. Denote by ${\rm Der}{\cal L}_\bullet$ the Lie algebra of derivations 
433: of the Lie algebra ${\cal L}_\bullet$. The commutator with the differential $\delta$ is a 
434: differential on the Lie algebra ${\rm Der}{\cal L}_\bullet$, making it into a DG Lie algebra.
435:  Then one has 
436: the following elementary Lemma.
437: \bl 
438: The Lie algebra $H_0^{\delta}\Bigl({\rm Der}{\cal L}_\bullet\Bigr)$ consists of 
439: the derivations of the Lie algebra ${\cal L}_\bullet$ 
440: preserving the grading, commuting  with 
441: the differential $\delta$, and defined up to a homotopy. 
442: \el
443:  The degree zero cycles 
444: are the derivations preserving the grading and commuting 
445: with the differential. 
446: The degree zero boundaries are the differentials homotopic to zero. 
447: 
448: Applying the map (\ref{1.17.08.10as12}) to the Hodge correlator class (\ref{HCCL}) 
449: class we arrive at  an element
450: \be \la{6.7.08.1}
451: {\bf H}^*_{X, a}\in H_0^{\delta}\Bigl({\rm Der}^S{\cal L}ie_{{\Bbb H}_*}\Bigr).
452: \ee
453: 
454: 
455: In the next two subsections we explain how this element 
456: defines a functorial real mixed Hodge 
457: structure on the rational homotopy type of compact Kahler manifolds. 
458: 
459: 
460: 
461: 
462: 
463: 
464: 
465: \subsection{Rational homotopy type DG Lie algebras} 
466: Let $M$ be a smooth simply connected 
467: manifold, $a\in M$. According to Quillen \cite{Q} and Sullivan   
468: \cite{S}, 
469: the rational homotopy type of $M$ is described by a DG Lie algebra 
470: over $\Q$ concentrated in degrees 
471: $0, -1, -2, ...$. It is well defined up to a quasiisomorphism. 
472: There is a similar DG Lie algebra 
473: ${\rm L}(M, a)$ 
474: over $\Q$ in the case when $M$ is not necessarily simply 
475: connected. Its zero cohomology $H^0{\rm L}(M, a)$  is  a Lie algebra, 
476: isomorphic to  the pronilpotent completion 
477: of $\pi_1(M, a)$. 
478:  
479: 
480: \vskip 3mm
481: The DG Lie algebra ${\rm L}(M, a)\otimes \C$ appears as follows. 
482: Let ${\rm DGCom}$ (respectively ${\rm DGCoLie}$) be the category of 
483: differential graded commutative algebras (respectively Lie coalgebras) 
484: over $\Q$. 
485: Then there is a pair of adjoint functors 
486: (the Bar construction followed to the projection to the 
487: indecomposables, and the Chevalley complex)
488: $$
489: {\cal B}: {\rm DGCom} \lra {\rm DGCoLie}, \qquad  
490: {\cal S}: {\rm DGCoLie} \lra {\rm DGCom}, 
491: $$
492: Take the de Rham DGCom ${\cal A}^*(M)$ over $\C$. Given a point $a \in M$, 
493: the evaluation map  ${\rm ev}_a: 
494: {\cal A}^*(M)\to \C$ provides  a DGCom denoted 
495: ${\cal A}^*(M, a)$. Applying to it the 
496: functor ${\cal B}\otimes \C$ we get a DGCoLie quasiisomorphic to the 
497: Lie coalgebra dual to  
498: ${\rm L}(M, a)\otimes \C$. 
499: 
500: Suppose now that $M$ is a compact Kahler manifold. 
501: Then, thanks to the formality theorem \cite{DGMS}, 
502: the DG commutative algebra ${\cal A}^*(M, a)$ is quasiisomorphic to 
503: its cohomology, the reduced cohomology  $\widetilde H^*(M, \C)$ 
504: of $X$. So applying the 
505: functor ${\cal B}$ to the rational reduced cohomology $\widetilde H^*(M, \Q)$ 
506: we get the minimal model of the rational homotopy type DG Lie algebra. 
507: We reserve from now on the notation  ${\rm L}(M, a)$ to this DG Lie algebra. 
508: 
509: 
510: \vskip 3mm
511: Let $X$ be a smooth irreducible projective 
512: variety over a field $k$. Let us choose an 
513: embedding $\sigma: k \subset \C$. Then the
514:  rational homotopy type DG Lie algebra 
515: ${\rm L}(X(\C), a)$ is supposed to be 
516: an object of motivic origin: 
517: One should exist a DG Lie algebra ${\rm L}_{\cal M}(X, a)$ 
518: in the yet hypothetical abelian tensor 
519: category of mixed motives over $k$, 
520: whose Betti realization is the DG Lie algebra 
521: ${\rm L}(X(\C), a)$. The Hodge realization ${\rm L}_{\rm Hod}(X, a)$ is known thanks to 
522: Morgan \cite{M}. 
523: 
524: 
525: \vskip 2mm
526: Let ${\cal L}ie_{\widetilde {\Bbb H}_{*}}$ be 
527: the free graded Lie algebra 
528:  generated by 
529: $$
530: \widetilde {\Bbb H}_{*}:= {\rm H}_{>0}(X)[-1], 
531: $$ 
532: sitting in degrees $[-(2n-1), 0]$. The Lie algebra ${\cal L}ie_{\widetilde {\Bbb H}_{*}}$
533: has DG Lie algebra with the differential dualising the product on ${\rm H}^{>0}(X)$. 
534: The formality theorem \cite{DGMS} implies:
535: 
536: \begin{theorem} \label{5.12.07.1}
537: The associate graded 
538: for the weight filtration 
539: ${\rm gr}^W{\rm L}(X, a)$ 
540: of the rational homotopy type  DG Lie algebra  is 
541: quasiisomorphic to the DG Lie algebra ${\cal L}ie_{\widetilde {\Bbb H}_*}$.
542: \end{theorem}
543: 
544: The zero cohomology $H^0{\rm L}_{\cal M}(X, a)$ 
545: is supposed to be a Lie algebra 
546: isomorphic to the (conjectural) unipotent motivic fundamental group of $X$. 
547: The latter was defined unconditionally for 
548: unirational $X$'s over number fields \cite{DG}. 
549: 
550: 
551: 
552: 
553: \subsection{Hodge correlators and a real MHS on the rational homotopy type}
554: 
555: \paragraph{The real Hodge Galois group.} The category of real mixed Hodge structures is an abelian  
556: tensor category \cite{D}. 
557: The associate graded for the weight filtration provides a fiber functor 
558: $H \lms {\rm gr}^WH$
559:  on this category. Therefore it is canonically 
560: equivalent to the category of representations 
561: of a pro-algebraic group over $\R$, called the Hodge Galois group $G_{\rm Hod}$. 
562: It  
563: is isomorphic to 
564: a semidirect product of $\C^*_{\C/\R}$ and a prounipotent algebraic group 
565: $U_{\rm Hod}$: 
566: $$
567: 0 \lra U_{\rm Hod} \lra G_{\rm Hod} \lra \C^*_{\C/\R} \lra 0, \qquad 
568: \C^*_{\C/\R} \hra G_{\rm Hod}.  
569: $$
570: The action of the group $\C^*_{\C/\R}$ provides 
571: the Lie algebra of the unipotent group 
572: ${\rm U}_{\rm Hod}$ with a structure of a Lie algebra in the category of 
573: direct sums of pure 
574: real Hodge structures. We denote this Lie algebra by ${\rm L}_{\rm Hod}$. 
575: So $G_{\rm Hod}$-modules are nothing else but 
576: ${\rm L}_{\rm Hod}$-modules in the category of real Hodge structures. 
577:  
578: According to Deligne \cite{D2}, 
579: the Lie algebra ${\rm L}_{\rm Hod}$  
580: is generated by certain generators 
581: $g_{p,q}$, $p,q \geq 1$, satisfying the only relation $\overline g_{p,q} = -g_{q, p}$. 
582: In Section 4 of \cite{G1} we defined a 
583: different set of generators, denoted below 
584: by $G^*_{p,q}$ and called the canonical generators --  
585: they behave nicer in families.\footnote{There are, in fact, two natural normalizations of the generators, denoted by $G^*_{p,q}$ and $G_{p,q}$ -- 
586: they differ by certain binomial coefficients depending 
587: on $p,q$. Both play a role in our story. 
588: We use $\ast$ in the objects related to the $G_{p,q}$-generators, 
589: e.g. ${\rm Cor}_{\cal H}^*$. 
590:  The Hodge correlators 
591: related to the generators $G^*_{p,q}$ are defined by using the operator 
592: $\omega$, see Section 2.1, instead of the operator $\xi$ in (\ref{xixi}).} 
593: In the mixed Tate case (i.e. $p=q$) they essentially 
594: coincide with the ones defined 
595: in \cite{L}. 
596: So  a real mixed Hodge structure is nothing else but the following data: 
597: 
598: \begin{itemize} \la{!!}
599: \item
600: a real Hodge structure $V$, i.e. a real vector space $V$, 
601: whose complexification $V_\C$ is bigraded, 
602: 
603: \item a 
604: collection of imaginary operators  $\{G^*_{p,q}\}$ on $V_\C$, $p,q >0$, of 
605: the bidegree by $(-p, -q)$, such that $\overline G^*_{p,q} =  - G^*_{q, p}$. 
606: \end{itemize}
607: Indeed, a representation  of $\C^*_{\C/\R}$ in $V$ 
608: is the same thing as a real Hodge structure in $V$, and the operators  
609: $G^*_{p,q}$ are the images of the canonical generators 
610: of   ${\rm L}_{\rm Hod}$. We 
611: encode a collection of the operators $\{G^*_{p,q}\}$ by a single operator
612: $$
613: G^* = \sum_{p,q>0}G^*_{p,q}.
614: $$
615: 
616: \paragraph{The Hodge Galois group action on the rational homotopy type DG Lie algebra.}
617: \begin{theoremconstruction} \la{4.25.08.22}
618: There is a canonical map 
619: \be \la{4.25.08.21qq}
620: {\rm L}_{\rm Hod} \lra H_0^{\delta} 
621: \Bigl({\rm Der}^S{\cal L}ie_{{\Bbb H}_*}\Bigr), \qquad G^* \lms {\Bbb H}^*_{X,a}, 
622: \ee 
623: The action of the pure Hodge group ${\Bbb C}_{\C/\R}^*$ 
624: is provided by  the real  Hodge structure  
625: on $H^{*}(X)$.
626:  \end{theoremconstruction}
627: This 
628: leads to a real MHS on $H^{\delta}_*{\rm L}(X, a)$ as follows. 
629: 
630: \bd \la{spes}
631: A derivation of the graded Lie algebra ${{\cal L}ie}_{\widetilde {\Bbb H}_*}$ 
632: is {\it special} if it kills both the 
633: generator $H_{2n}(X)[-1]$ and the element ${\rm S} = \delta (H_{2n}(X)[-1])$. 
634: \ed
635: Denote by ${\rm Der}^S{{\cal L}ie}_{\widetilde {\Bbb H}_*}$ the Lie algebra of all 
636: special derivations. The inclusion 
637: ${\Bbb H}_*\hra \widetilde {\Bbb H}_*$ provides  an isomorphism of graded Lie algebras
638: \be \la{mapqc}
639: {\rm Der}^S{{\cal L}ie}_{{\Bbb H}_*}\stackrel{\sim}{\lra} {\rm Der}^S{{\cal L}ie}_{\widetilde {\Bbb H}_*}.
640: \ee
641: One easily checks that the differentials preserve the subalgebras of special derivations. 
642: %of the Lie algebras ${{\cal L}ie}_{ {\Bbb H}}$ 
643: %and ${{\cal L}ie}_{\widetilde {\Bbb H}}$. In other words, in both cases 
644: %the commutator of a special derivation with $\delta$ is a special derivation.  
645: %Indeed, let $D$ be a special derivation of the Lie algebra 
646: %${{\cal L}ie}_{\widetilde {\Bbb H}}$. Set 
647: %$h = H_{2n}(X)[-1]$. 
648: %Then $[\delta, D](h) = \delta D(h) - D\delta(h) =0$, and $[\delta, D](\delta(h)) = 
649: %\delta D(\delta(h))=0$. Similarly for ${{\cal L}ie}_{{\Bbb H}}$. 
650: So the map 
651: (\ref{mapqc}) is an isomorphism of DG Lie algebras.
652: Therefore it induces an isomorphism
653: $$
654: H_0^\delta\Bigl({\rm Der}^S{\cal L}ie_{{\Bbb H}_*}\Bigr) =  
655: H_0^\delta\Bigl({\rm Der}^S{{\cal L}ie}_{\widetilde {\Bbb H}_*}\Bigr). 
656: $$
657: The element ${\Bbb H}^*_{X,a}$ lives in the Lie algebra on the left. 
658: We transform it to an element $\widetilde {\Bbb H}^*_{X,a}$ of the Lie algebra on the right. 
659: So we get 
660: a map 
661: \be \la{4.221qqw}
662: {\rm L}_{\rm Hod} \lra H_0^{\delta} 
663: \Bigl({\rm Der}^S{\cal L}ie_{\widetilde {\Bbb H}_*}\Bigr), \qquad G^* \lms \widetilde {\Bbb H}^*_{X,a}, 
664: \ee 
665: The Lie algebra  $H_0^{\delta} 
666: \Bigl({\rm Der}^S{\cal L}ie_{\widetilde {\Bbb H}_*}\Bigr)$ acts 
667: by derivations of ${\cal L}ie_{\widetilde {\Bbb H}_*}$, defined up to a homotopy. 
668: Finally,  ${\cal L}ie_{\widetilde {\Bbb H}_*}$ is 
669: isomorphic to 
670:  ${\rm gr}^W{\rm L}(X, a)$ by Theorem \ref{5.12.07.1}. 
671: 
672: Summarising, we defined a real MHS on ${\rm L}(X, a)$. 
673: When the point $a$ varies, we get a variation of real MHS.  
674: The proof of this is completely similar to the one given in \cite{G1} 
675: in the case of curves, and thus is omitted. 
676:  
677: \vskip 3mm
678: In particular, we get a real MHS on the pronilpotent completion 
679: $\pi_1^{\rm nil}(X,a)$. 
680: \bt \la{FMHSR} Let $X$ be a complex regular projective  variety. Then 
681: the  real MHS on $\pi_1^{\rm nil}(X,a)$ provided by the 
682: Hodge correlators coincides with the classical one. 
683: \et
684: {\bf Proof}. We proved in \cite{G1} that this is true for any smooth curve $X$. 
685: The Hodge correlator real MHS on $\pi_1^{\rm nil}(X,a)$ is 
686: functorial by Theorem \ref{PPPq}. 
687: By the Lefschetz theorem there exists a plane section $Y$ of $X$ 
688: such that $\pi_1(Y)$ surjects onto $\pi_1(X)$. 
689: This implies Theorem \ref{FMHSR}.  
690: 
691: \vskip 3mm
692: The real MHS on $\pi_1^{\rm nil}(X,a)$ can be  described 
693: via Chen's iterated integrals. Theorem \ref{FMHSR} 
694: tells that the Hodge correlators provide the same 
695: real periods as iterated integrals. It would be interesting to 
696: relate 
697: the Hodge correlators to the iterated integrals directly. 
698: I know how to do this only in the simplest cases. 
699: 
700: 
701: 
702: 
703: \subsection{Motivic correlators for a regular projective variety}
704: \paragraph{Motivic Galois groups.} One expects that there exists an abelian tensor category 
705: ${\cal M}{\cal M}_k$ of mixed motives over a field $k$. 
706: Each object $M$ of this category should 
707: carry a canonical weight filtration $W_{\bullet}$. The functor 
708: $$
709: \omega: {\cal M}{\cal M}_k \stackrel{?}{\lra} {\cal P}{\cal M}_k, \qquad
710: M \lms {\rm gr}^{W}M
711: $$
712: should be a fiber functor with values in the semisimple tensor category 
713: ${\cal P}{\cal M}_k$ of pure motives. The algebraic group scheme of all 
714: automorphisms of this fiber functor is called the motivic Galois group 
715: $G_{\rm Mot}$. Thanks to the Tannakian formalism, the functor $\omega$ 
716: then provides a canonical equivalence between 
717: the category of all mixed motives and the category 
718: of representations of $G_{\rm Mot}$ in the category of pure motives. 
719: So the  motivic Galois group 
720: $G_{\rm Mot}$ contains all the information telling us how the 
721: mixed motives are obtained from the pure ones. 
722: 
723: The motivic Galois group $G_{\rm Mot}$ is a prounipotent algebraic group 
724: in the tensor category of pure motives. 
725: We denote by ${\rm L}_{\rm Mot}$ its Lie algebra, and by ${\cal L}_{\rm Mot}$ 
726: the dual Lie coalgebra -- see, say,  \cite{G4} for more details.  
727:  
728: 
729: 
730: 
731: \paragraph{Motivic correlators.} The motivic Galois group should act  
732: by symmetries of the rational motivic homotopy type of  a variety
733: $X$. The latter, however, is well defined only up to 
734: a quasiisomorphism. Here is a precise statement:
735: 
736: {\it One should exist a 
737: homomorphism  of Lie algebras in the category of pure motives} 
738: \be \la{4.25.08.21g}
739: {\rm L}_{\rm Mot} \stackrel{?}{\lra} H_0^{\delta} 
740: \Bigl({\rm Der}({\rm gr}^W{\rm L}_{\cal M}(X,a))\Bigr). 
741: \ee
742: 
743: 
744: 
745: The $\R$-Hodge realization of the 
746: motivic Lie algebra ${\rm L}_{\rm Mot}$ 
747: is the Hodge Lie algebra ${\rm L}_{\rm Hod}$. So the 
748:  $\R$-Hodge realization of the map (\ref{4.25.08.21g}) should give the map (\ref{4.25.08.21qq}). 
749: 
750: 
751: \vskip 2mm
752: Denote by ${\cal L}ie_{{\rm Mot}({\Bbb H}_*)}$ the free graded Lie algebra 
753: generated by the reduced 
754: homology motive of $X$ shifted to the right by $1$. It is a DG Lie algebra. 
755: Similarly, there is a slightly bigger DG Lie algebra 
756: ${\cal L}ie_{ {\rm Mot}(\widetilde {\Bbb H}_*)}$. Theorem \ref{5.12.07.1} 
757: tells that one should exist a canonical isomorphism
758: \be \la{CIS}
759: {\cal L}ie_{{\rm Mot}(\widetilde {\Bbb H}_*)}
760:  \stackrel{?}{=}{\rm gr}^W{\rm L}_{\cal M}(X,a).
761: \ee 
762: Just like in Definition \ref{spes}, one defines  special derivations of the Lie algebra 
763: on the left, and hence on the right of (\ref{CIS}).  
764: The motivic Galois Lie algebra should act by special derivations 
765: of the  Lie algebra 
766: ${\rm gr}^W{\rm L}^{\rm Mot}(X,a)$. 
767: There is a canonical isomorphism
768: \be \la{1.17.08.10as12s}
769: {\rm Der}^S{\cal L}ie_{{\rm Mot}({\Bbb H}_*)} \stackrel{\sim}{\lra} 
770: {\rm Der}^S{\cal L}ie_{{\rm Mot}(\widetilde {\Bbb H}_*)}. 
771: \ee
772: Therefore  the map 
773: (\ref{4.25.08.21g}) plus isomorphisms (\ref{1.17.08.10as12s}) and  
774: (\ref{CIS}) provide a canonical map of Lie algebras
775: \be \la{4.25.08.21q}
776: {\rm L}_{\rm Mot} \stackrel{?}{\lra} H_0^{\delta} 
777: \Bigl({\rm Der}^S{\cal L}ie_{{\rm Mot}({\Bbb H}_*)}\Bigr). 
778: \ee
779: Dualizing the map (\ref{4.25.08.21q}) and using the motivic version of 
780: isomorphism (\ref{1.17.08.10as12}) we get a morphism of Lie coalgebras, 
781: called the 
782: {\it motivic correlator map} for  $X$:
783: \be \la{4.25.08.21rty}
784: {\rm Cor}_{\rm Mot}: 
785: H^0_{\delta}\Bigl({\cal C}{\cal L}ie^{\vee}_{{\rm Mot}({\Bbb H}_*)}\otimes {\cal H}^{\vee} \Bigr)
786: \stackrel{?}{\lra}
787:  {\cal L}_{\rm Mot}.
788: \ee
789: Its source is easily 
790: defined in terms of the cohomology motive of $X$.  The target is remarkable: 
791: it carries all the information 
792: how the mixed motives are glued from the pure ones. The cobracket 
793: of an element in ${\cal L}_{\rm Mot}$  contains all the information about 
794: the complexity of the element. Therefore, since 
795: (\ref{4.25.08.21rty}) is a Lie coalgebra map,  and 
796: the cobracket in the left Lie coalgebra  
797: is very transparent, see Fig \ref{hf7}, we get a full control 
798: of the arithmetic complexity of motivic correlators. 
799: 
800: 
801: 
802: 
803: \paragraph{Relating motivic and Hodge correlators.} 
804: The hypothetical motivic Lie coalgebra ${\cal L}_{\rm Mot}$ 
805: is an inductive limit of objects in the  
806: semisimple category ${\cal P}{\cal M}_k$ 
807: of pure motives over $k$. 
808: One can decompose it into the isotipical components:
809: $$
810: {\cal L}_{\rm Mot} \stackrel{?}{=} \bigoplus_{M \in {\rm Iso}({\cal P}{\cal M}_k)}
811: {\cal L}_{M}\otimes M_{}^{\vee}, \qquad 
812: {\cal L}_{M} = {\rm Hom}_{{\cal P}{\cal M}_k}(M^{\vee}, {\cal L}_{\rm Mot}).
813: $$
814: Here $M$ runs through the isomorphism classes of simple objects in 
815: the category ${\cal P}{\cal M}_k$, $M^{\vee}$ is the dual motive, 
816: and ${\cal L}_{M}$ 
817: is a $\Q$-vector space. 
818: 
819: One can do this in the Hodge realization. 
820: Namely, denote by ${\cal L}_{\rm Hod}$ the $\R$-Hodge Lie coalgebra. 
821: Then there is a decomposition into the isotipical components:
822: $$
823: {\cal L}_{\rm Hod} = \bigoplus_{H \in {\rm Iso}({\cal P}{\rm Hod}_\R)}
824: {\cal L}_{H}\otimes H_{}^{\vee}.
825: $$ 
826: 
827: 
828: \bp \la{cpm} 
829: There is a canonical linear map, called {\it the real period map}: 
830: $$
831: {\rm P}_\R: {\cal L}_{\rm Hod} \lra i\R.
832: $$
833: \ep
834: 
835: The Hodge and motivic correlator are compatible as follows. 
836: \bp \la{10.21.08.1}
837: Let $X$ be a regular projective variety over $\R$. Then 
838: the composition 
839: $$
840: H^0_{\delta}\Bigl({\cal C}{\cal L}ie^{\vee}_{{\rm Hod}({\Bbb H}_*)}\otimes 
841: {\cal H}^{\vee} \Bigr)~
842: \stackrel{{\rm Cor}_{\rm Hod}}{\lra} ~{\cal L}_{\rm Hod} ~
843: \stackrel{{\rm P}_\R}{\lra} ~i\R.
844: $$
845: coincides with  the Hodge correlator map 
846: ${\rm Cor}_{\cal H}$. 
847: \ep
848: Therefore, given a regular projective variety $X$ over $\Q$, and 
849: assuming the existence the motivic correlator map  ${\rm Cor}_{\rm Mot}$, 
850: the Hodge correlator map coincides with the composition
851: $$
852: H^0_{\delta}\Bigl({\cal C}{\cal L}ie^{\vee}_{{\rm Mot}({\Bbb H}_*)}\otimes 
853: {\cal H}^{\vee} \Bigr)~
854: \stackrel{{\rm Cor}_{\rm Mot}}{\lra}~ {\cal L}_{\rm Mot} ~
855: \stackrel{r_{\rm Hod}}{\lra} ~
856: {\cal L}_{\rm Hod} ~\stackrel{{\rm P}_\R}{\lra}~ i\R.
857: $$
858: Here $r_{\rm Hod}$ is the $\R$-Hodge realization functor. 
859: 
860: \vskip 3mm
861: \subsection{A Feynman integral for Hodge correlators} \la{hf1.6} 
862: Let ${\cal L}_1$ and ${\cal L}_2$ be two DG Lie algebras. 
863: Let us forget for a moment about their differentials, treating them as graded Lie superalgebras. 
864: Then there is an affine ${\Bbb G}_m$-superscheme\footnote{That is, an affine
865:  superscheme equipped with an action of the algebraic group ${\Bbb G}_m$} of graded 
866: Lie superalgebra maps ${{\cal H}om}_{\rm Lie}({\cal L}_1, {\cal L}_2)$. 
867: Its points with values in a graded supercommutative algebra $A$ are defined by 
868: $$
869: {{\cal H}om}_{\rm Lie}({\cal L}_1, {\cal L}_2)(A):= {\rm Hom}_{\rm Lie}({\cal L}_1, {\cal L}_2\otimes A).
870: $$
871: Denote by ${\cal L}ie_{V}$ the free graded Lie superalgebra 
872: generated by a finite dimensional graded vector  space $V$. 
873: Let ${\cal G}$ be a graded Lie superalgebra. 
874: Then there is an isomorphism of affine ${\Bbb G}_m$-superschemes
875: \be \la{4.20.08.6}
876: {{\cal H}om}_{\rm Lie}({\cal L}ie_{V^{*}}, {\cal G}) = V \otimes {\cal G}.
877: \ee
878: Suppose now that ${\cal L}ie_{V^*}$ has a differential 
879: $\Delta$ providing it with a  DG Lie algebra structure. 
880: Since $\Delta$ is an odd (degree $1$) infinitesimal automorphism 
881: of the Lie superalgebra ${\cal L}ie_{V^*}$, 
882:  it is transformed by isomorphism (\ref{4.20.08.6}) 
883: to an infinitesimal 
884: automorphism of the affine  ${\Bbb G}_m$-superscheme $V \otimes {\cal G}$, that is 
885: to a homological vector field $v_{{\Delta}}$ on 
886: $V \otimes {\cal G}$, making it into an affine DG scheme -- 
887: see Section \ref{hf5.2ref} for background. 
888: 
889: Furthermore, let ${\cal D}$ be a derivation 
890: of the graded Lie algebra ${\cal L}ie_{V}$ commuting with  
891: $\Delta$. Then isomorphism (\ref{4.20.08.6}) transforms it  
892: to a homological vector field $v_{{\cal D}}$ commuting with 
893: $v_{{\Delta}}$. 
894:  
895: \vskip 3mm
896: {\bf Example.} 
897: Let $({\cal L}, d)$ be a DG Lie algebra. Then ${\cal L}[1]$ is a DG space 
898: with the homological vector field given by the Chern-Simons vector field 
899: $Q_{\rm CS}$ 
900: (see Section \ref{hf6.1ref}):
901: $$
902: \stackrel{{\cdot}}{\alpha} = Q_{\rm CS}(\alpha) = 
903: d \alpha + \frac{1}{2}[\alpha, \alpha]. 
904: $$
905: In particular,  given a DG commutative algebra $A$  and a Lie algebra ${\cal G}$, 
906: there is  a DG Lie algebra 
907: $A \otimes {\cal G}$ with the Lie bracket 
908: $[a_1 \otimes g_1, a_2 \otimes g_2]:= a_1a_2 \otimes [g_1, g_2]$, and hence 
909: a DG scheme  $A[1] \otimes {\cal G}$.
910: 
911:  
912: 
913: 
914: \vskip 3mm
915: The functor ${\cal B}: {\rm DGCom} \to {\rm DG Colie}$ takes an 
916: $A \in {\rm DGCom}$ to 
917:  the cofree Lie coalgebra ${\rm CoLie}(A[1])$ 
918: cogenerated by the graded space $A[1]$, with a certain differential $\Delta$.  
919: 
920: The functor ${\cal B}$  is characterised by the following universality property: 
921: \begin{itemize}
922: \item One has a functorial isomorphism of the affine DG schemes:
923: \be \la{DGISO}
924: {{\cal H}om}_{\rm CoLie}({\cal G}^{\ast}, {\cal B}(A)) = A[1] \otimes {\cal G}.
925: \ee
926: \end{itemize}
927: In other words, the functorial isomorphism (\ref{DGISO})  
928: transforms  
929: the differential $\Delta$ on ${\cal B}(A)$ 
930: to the 
931: Chern-Simons homological vector field $Q_{\rm CS}$ on $A[1] \otimes {\cal G}$.  
932: \vskip 3mm
933: 
934: Applying (\ref{DGISO}) to the algebra 
935: $A = \overline {\rm H}^*(X)$, we get  
936: an isomorphism of DG schemes
937: \be \la{4.20.08.5}
938: {{\cal H}om}_{\rm Lie}({\cal L}ie_{{\Bbb H}_*}, {\cal G}) = 
939: {\Bbb H}^*(X) \otimes {\cal G} =: {\cal G}_{{H}}[1].
940: \ee 
941: The differential $\delta$ on the DG Lie algebra ${\cal L}ie_{{\Bbb H}_*}$ 
942: is transformed to the Chern-Simons homological 
943: vector field $Q_{CS}$ on ${\cal G}_{{H}}[1]$. 
944: 
945: \vskip 3mm
946: Let ${\cal G}$ be a Lie 
947: algebra with a  non-degenerate 
948: scalar product $Q(\ast, \ast)$. 
949: For example one can take the Lie algebra ${\rm Mat}_N$ of 
950: $N\times N$ matrices 
951: with $Q(A,B):= {\rm Tr}(AB)$. 
952: 
953: The Hodge correlator ${\bf H}^*_{X,a}$, see (\ref{6.7.08.1}),  is a derivation of 
954: ${\cal L}ie_{{\Bbb H}_*}$ commuting with $\delta$ and defined up to a homotopy. 
955: So it is transformed by isomorphism (\ref{4.20.08.5}) 
956: to a homological vector field ${\rm Q}_{\rm Hod}$ on ${\cal G}_{{H}}[1]$, 
957: commuting with the Chern-Simons vector field ${\rm Q}_{\rm CS}$ and 
958: well defined up to commutators 
959: $[{\rm Q}_{\rm CS}, R]$. 
960: Our next goal 
961: is to show 
962: that we get a Hamiltonian vector field, whose Hamiltonian 
963: is given by a Feynman integral introduced below. 
964: 
965: 
966: \vskip 3mm
967: 
968: 
969: Given  ${\cal G}$-valued differential forms $\varphi_i$ on $X$, 
970: we define a differential  form
971: $$
972: \langle \varphi_1, \varphi_2,\varphi_3\rangle:= Q(
973: \varphi_1, [\varphi_2,\varphi_3]\rangle.
974: $$ 
975: 
976: Choose a splitting of the bigraded de Rham complex  ``harmonic'' forms ${\cal H}ar_X$ 
977: and its orthogonal complement.  
978: Let ${\rm Ker}~d^\C\subset {\cal A}^{*}(X)$ be the kernel of $d^\C$. 
979: Let us define, using the splitting,  a function on the functional space 
980: $$
981: {\rm Ker}~d^\C\otimes {\cal G}[1].   
982: $$ 
983: Let $\psi \in {\rm Ker}~d^\C\otimes {\cal G}[1]$. 
984: Write ${\rm Ker}~d^\C
985:  = {\cal H}ar_X \oplus {\rm Im}~d^\C$, and 
986: $$
987: \psi = \psi_0 + \alpha,  
988: \quad \psi_0 \in {\rm Im}~d^\C\otimes {\cal G}[1] , \quad \alpha \in 
989: {\cal H}ar_X\otimes {\cal G}[1].
990: $$
991: Choose $\varphi$ such that $d^\C\varphi = \psi_0$. Then there is an  action: 
992: \be \la{ALP}
993: S(\psi) = \int_X\frac{1}{2}( \varphi, d\psi ) + 
994: \frac{1}{6}\langle \varphi, \psi,
995:  \psi \rangle, \qquad d^\C\varphi = \psi_0.
996: \ee
997: Observe that $S(\psi)$ is independent of 
998: the choice of $\varphi$. Indeed, 
999:  $\int_X \langle d^\C \eta, d \psi\rangle  = 
1000: \int_X \langle d^\C \eta, \psi, \psi\rangle =0$ since $d^\C\psi=0$. 
1001: Notice that $d^\C$ (Lagrangian in (\ref{ALP})) is very close to 
1002: the Chern-Simons Lagrangian: 
1003: $$
1004: d^\C\Bigl(\frac{1}{2}(\varphi, d\psi ) + 
1005: \frac{1}{6}\langle \varphi, \psi,
1006:  \psi \rangle\Bigr) = \frac{1}{2}\langle \psi_0, d\psi \rangle + 
1007: \frac{1}{6}\langle \psi_0, \psi,
1008:  \psi \rangle
1009: $$
1010: We  would like to ``integrate'' 
1011: $e^{ iS(\psi)}$ along the fibers of the natural projection 
1012: provided by the splitting of the de Rham complex: 
1013: $$
1014: {\rm Ker}~d^\C\otimes {\cal G}[1] \lra H^*(X)\otimes {\cal G}[1],
1015:  \qquad \psi = \psi_0 + \alpha \lms \alpha.  
1016: $$
1017: However the quadratic part 
1018: $(\varphi, d\psi)$ of the Lagrangian 
1019: is highly degenerate. Precisely, by  the classical 
1020: Hodge theory 
1021: there exists a subspace $F \subset {\cal A}^{*,*}_X$ 
1022: such that there is an isomorphism of bicomplexes
1023: \be \la{3.3.08.10}
1024: {\cal A}^{*,*}_X = {\cal H}ar_X ~~\bigoplus ~~
1025: \begin{array}{ccc}
1026: d^\C F&\stackrel{d}{\lra}
1027:  &d^\C d F \\
1028: %&&\\
1029: d^\C \uparrow &&d^\C \uparrow \\
1030: %&&\\
1031: F &\stackrel{d}{\lra} & d F 
1032: \end{array}  
1033: \ee
1034: Here all arrows in the square are isomorphisms. 
1035: Then 
1036: $$
1037: {\rm Ker}~d^\C = {\cal H}ar_X \oplus d^\C F \oplus d^\C d F, 
1038: $$
1039: and kernel of the quadratic form 
1040: $(\varphi, d\psi)$ is the subspace 
1041: ${\cal H}ar_X \oplus 
1042: d^\C d F$. The form is non-degenerate 
1043: on $d^\C F$.  
1044: We integrate over  
1045: $\psi \in d^\C F$, i.e. over a ``a quarter $F$ of the de Rham complex'',
1046: getting a function, the Hodge correlator, which  
1047: depends on the decomposition (\ref{3.3.08.10}):
1048: \be \la{4.18.08.1}
1049: \widetilde {\rm Cor}_{\cal H}^{*}(\alpha):= 
1050: \int e^{iS(\alpha + d^\C \varphi)}{\cal D}\varphi, \qquad \varphi\in F.
1051: \ee
1052: Elaborating 
1053: $$
1054: S(\alpha + d^\C \varphi):= \int_X 
1055: \frac{1}{2}(\varphi, dd^\C \varphi) 
1056: + \frac{1}{6}
1057: \langle \varphi, d^\C  \varphi, d^\C  \varphi\rangle 
1058: + \frac{1}{3}\langle \alpha, \varphi, d^\C \varphi \rangle 
1059: + \frac{1}{6}\langle \alpha, \alpha, \varphi \rangle,  
1060: $$
1061: we use the perturbative series expansion 
1062: on the tree level to make sense of the integral, treating 
1063: $\alpha$ as a small parameter -- see Section \ref{hf5.3sec}. 
1064: It involves only plane trivalent trees decorated by 
1065: the harmonic form $\alpha$, i.e. it amounts to minimizing 
1066: the action:
1067: $$
1068: \widetilde {\rm Cor}_{\cal H}^{*}(\alpha)= {\rm Min}_{\varphi \in F}
1069: S(\alpha + d^\C \varphi).
1070: $$
1071: 
1072: %Let us define a function on the functional space 
1073: %$$
1074: %{\cal A}^*_{X}\otimes {\cal G} ~\bigoplus ~ {\cal G}_H[1].   
1075: %$$ 
1076: %Given  
1077: %$\varphi \in {\cal A}^*_{X}\otimes {\cal G}$ and 
1078: %$\alpha \in {\cal G}_H[1]$, set 
1079: %$$
1080: %S(\alpha, \varphi):= \int_X 
1081: %\frac{1}{2}(\varphi \wedge dd^\C \varphi) 
1082: %+ \frac{1}{6}
1083: %\langle \varphi, d^\C  \varphi, d^\C  \varphi \rangle 
1084: %+ \frac{1}{2}\langle \alpha, \varphi, d^\C \varphi \rangle 
1085: %+ \frac{1}{2}\langle \alpha, \alpha, \varphi \rangle. 
1086: %$$
1087: %We treat 
1088: %$\alpha$ as a small parameter, and would like to ``integrate'' 
1089: %$e^{ iS(\alpha, \varphi)}$ %$e^{\frac{i}{\hbar} S(\alpha, \varphi)}$
1090: %over the space of fields $\varphi$ in order to get a function 
1091: %${\rm Cor}^*_{\cal H}(\alpha)$ on the 
1092: %space ${\cal G}_H[1]$. 
1093: %However the quadratic part 
1094: %$(\varphi, dd^\C \varphi)$ of the Lagrangian 
1095: %is highly degenerate. Precisely, by  the classical 
1096: %Hodge theory 
1097: %there exists a subspace $F \subset {\cal A}^{*,*}_X$ 
1098: %such that there is an isomorphism of bicomplexes
1099: %\be \la{3.3.08.10}
1100: %{\cal A}^{*,*}_X = {\cal H}ar_X ~~\bigoplus ~~
1101: %\begin{array}{ccc}
1102: %d^\C F&\stackrel{d}{\lra}
1103: % &d^\C d F \\
1104: %%&&\\
1105: %d^\C \uparrow &&d^\C \uparrow \\
1106: %%&&\\
1107: %F &\stackrel{d}{\lra} & d F 
1108: %\end{array}  
1109: %\ee
1110: %Here ${\cal H}ar_X$ is a finite 
1111: %dimensional bigraded subspace of ``harmonic'' forms, and 
1112: %all arrows in the square are isomorphisms. 
1113: %The kernel of the quadratic form 
1114: %$(\varphi, dd^\C \varphi)$ is the subspace 
1115: %${\cal H}ar_X \oplus d F \oplus 
1116: %d^\C F  \oplus 
1117: %d^\C d F$. The form is non-degenerate 
1118: %on $F$.  
1119: %We integrate over  
1120: %$f \in F$, i.e. over a ``a quarter of the space of fields $\varphi$'',
1121: %getting a function, the Hodge correlator, which  
1122: %depends on the decomposition (\ref{3.3.08.10}):
1123: %\be \la{4.18.08.1}
1124: %{\rm Cor}_{\cal H}^{F,*}(\alpha):= 
1125: %\int e^{iS(\alpha, f)}{\cal D}f, \qquad f\in F.
1126: %\ee
1127: %We use the perturbative series expansion 
1128: %on the tree level to make sense of the integral -- see Section \ref{hf5.3sec}. 
1129: %It involves only plane trivalent trees decorated by 
1130: %the harmonic form $\alpha$, i.e. it amounts to minimizing 
1131: %the action:
1132: %$$
1133: %{\rm Cor}_{\cal H}^{F, *}(\alpha)= {\rm Min}_{f \in F}
1134: %S(\alpha, f).
1135: %$$
1136: 
1137: 
1138: 
1139: 
1140: 
1141: 
1142: 
1143: The space  
1144: ${\cal G}_{{\rm H}}[1]$ has an even  symplectic structure 
1145: $$
1146: \omega(\overline h_1 \otimes g_1, \overline h_2 \otimes g_2):=  (-1)^{|h_1|}Q(g_1, g_2) \int_X h_1 \wedge h_2.
1147: $$
1148:  So function (\ref{4.18.08.1}) gives rise to a 
1149: Hamiltonian vector field  on  
1150: ${\cal G}_{{\rm H}}[1]$:
1151: \be \la{18.4.08.8}
1152: \widetilde {\rm Q}^{*}_{\rm Hod}%_{X, {\cal G}, a}
1153:  \in S^{>0}({{\cal G}_{{\rm H}}[1]}^{\vee})\otimes
1154:  {\cal G}_{{\rm H}}[1].
1155: \ee
1156: Here on the right stands 
1157: the reduced Chevalley cochain complex 
1158: of the graded Lie algebra ${\cal G}_{{\rm H}}$ with the adjoint 
1159: coefficients. It has a differential $\delta$ \, 
1160: given by the commutator with $Q_{\rm CS}$.  
1161: 
1162: 
1163: 
1164: 
1165: 
1166: \begin{theorem} \la{18.4.08.10}
1167: i) The element (\ref{18.4.08.8}) provides a 
1168: well defined, functorial  cohomology class
1169: \be \la{CLass}
1170: {\rm Q}^*_{\rm Hod} \in H_0^{\delta}\Bigl({\rm Der}{\cal G}_{{\rm H}}\Bigr). 
1171: \ee
1172: 
1173: ii) The class  (\ref{CLass}) is the image of the class  ${\bf H}^*_{X,a}$ 
1174: under the isomorphism (\ref{4.20.08.5}). 
1175: \et
1176: 
1177: The part i) means that the cohomology class (\ref{18.4.08.8})
1178:  does not depend 
1179: on the choice of the subspaces $F$ and ${\cal H}ar_X$, as well as on 
1180: the choice of a Green current involved in 
1181: its perturbative series expansion 
1182: definition. The part i) is deduced  from ii) and  Theorem \ref{MAATH}. 
1183: The part ii) is obtained by comparing the  
1184: correlators from Sections 2.3/\ref{hf6.a} with the ones from Section \ref{hf5.4ref}. 
1185: 
1186: 
1187: 
1188: \vskip 3mm
1189: We view the class $Q^*_{\rm Hod}$ as a  vector field on the DG space ${\cal G}_H[1]$
1190:  -- we call it the {\it Hodge vector field}. 
1191: The part i) of Theorem \ref{18.4.08.10} just means that 
1192: it commutes with the Chern-Simons field ${\rm Q}_{CS}$, 
1193: and is well defined up to commutators with ${\rm Q}_{CS}$. Let $\varepsilon$ be an odd element 
1194: of degree $1$, so $\varepsilon^2=0$. Then the part i) of Theorem \ref{18.4.08.10} 
1195: is equivalent to the
1196: 
1197: \bt \la{DEFO}
1198: The vector field ${\rm Q}_{CS} + \varepsilon Q^*_{\rm Hod}$ is a homological vector field on the 
1199:  DG space ${\cal G}_H[1]$, well defined up to commutators $[{\rm Q}_{CS}, \ast]$.
1200: 
1201: So  a real mixed Hodge structure on the rational homotopy type of $X$ is the same thing as 
1202: a deformation of the DG space ${\cal G}_H[1]$ over the odd line ${\rm Spec}\R[\varepsilon]$,
1203:  functorial in ${\cal G}$, considered up to an isomorphism. 
1204: \et
1205: 
1206: %\input{hfbv.tex}
1207: 
1208: 
1209: 
1210: 
1211: \paragraph{The Hodge Galois group and complex local systems}
1212: 
1213: An $\R$-mixed Hodge structure on the pronilpotent fundamental group 
1214: $\pi_1^{\rm nil}(X,x)$ tell us  that there is an 
1215: extension of Lie algebras, where ${\cal G}_{\rm Hod}$ is the Lie algebra of the 
1216: Hodge Galois group $G_{\rm Hod}$: 
1217: \be \la{6.6.08.1}
1218: 0 \lra \pi_1^{\rm nil}(X,x) \lra \pi_1^{\rm Hod}(X,x)
1219: \lra {\cal G}_{\rm Hod} \lra 0.  
1220: \ee
1221: A  base point $x\in X$ provides a  splitting 
1222: $s_x: {\cal G}_{\rm Hod} \to \pi_1^{\rm Hod}(X,x)$, 
1223: and hence  an action of ${\cal G}_{\rm Hod}$ on 
1224: the Lie algebra $\pi_1^{\rm nil}(X,x)$, and thus a  
1225: real MHS on $\pi_1^{\rm nil}(X,x)$.  
1226: 
1227: 
1228: It follows that the Lie algebra ${\cal G}_{\rm Hod}$ acts  
1229: on representations of the Lie algebra  $\pi_1^{\rm nil}(X,x)$. The latter 
1230: are automatically nilpotent, and  
1231: form the formal neighborhood of the trivial local system. 
1232: So ${\cal G}_{\rm Hod}$ acts on 
1233: the formal neighborhood of the trivial local system. 
1234: In Section \ref{hf6.5ref} we describe this action explicitly by using
1235:  the class (\ref{CLass}). 
1236: %It extends 
1237: %to an action of the Lie algebra 
1238: %${\cal G}_{\rm Hod}$ by formal vector fields on the moduli space of 
1239: %complex local systems on $X$. 
1240: 
1241: 
1242: %\vskip 3mm
1243: %For a variety $X$ 
1244: %over $\Q$,  there is an $l$-adic analog of (\ref{6.6.08.1}),
1245: % given by the exact sequence
1246: %$$
1247: %0 \lra \pi_1^{(l)}(X,x) \lra \pi_1^{\rm arithm}(X,x)
1248: %\lra {\rm Gal}(\overline \Q/\Q) \lra 0. 
1249: %$$
1250: %Here $\pi_1^{(l)}(X,x)$ is the pro-$l$ completion of the 
1251: %fundamental group of $X$. 
1252: %A  base point $x\in X(\Q)$ provides a splitting 
1253: %of this exact sequence, and hence an action of the Galois group 
1254: %on $\pi_1^{(l)}(X,x)$. 
1255: 
1256: %In the $l$-adic situation the absolute Galois group acts 
1257: %on the category of $l$-adic sheaves, and in particular on $l$-adic local systems 
1258: %on $X$. One hopes for a similar picture in the Hodge situation:
1259: 
1260: 
1261: 
1262: %\vskip 3mm 
1263: % {\bf Example}. By the Hain-Zucker theorem \cite{HZ}, 
1264: %the category of good unipotent variations of real 
1265: %mixed Hodge structures on $X$ is equivalent to 
1266: %the category of  representations of the 
1267: %Hodge fundamental Lie algebra $\pi_1^{\rm Hod}(X,x)$, that is to the subcategory 
1268: %of $L_{\rm Hod}$-equivariant representations of $\pi_1^{\rm nil}(X,x)$.
1269: 
1270: 
1271: %\input{hf9.tex}
1272: 
1273: 
1274: 
1275: 
1276: 
1277: 
1278:  \paragraph{Acknowledgments.} I am very grateful to 
1279: Maxim  Kontsevich for useful discussions.
1280:  
1281: 
1282: A part of this work 
1283:  was done    
1284: during my visits to the IHES at the Summer of 
1285: 2007 and MPI(Bonn) at the Summer of 2008. 
1286: I am grateful to these institutions for the hospitality and support. 
1287: I was supported by the NSF grant DMS-0653721. 
1288: I am grateful to the referee for useful comments. 
1289: 
1290: 
1291: 
1292: 
1293: 
1294: %\vskip 3mm
1295: %\input{motivicgalois.tex}
1296: 
1297: %\vskip 3mm
1298: %\input{hfideas}
1299: