1: \documentclass[noamsfonts]{amsproc}
2: \usepackage{aspmproc, amssymb}
3:
4: \title[On the cusp form motives in genus~$1$ and level~$1$]{On the cusp
5: form motives in genus~$1$ and level~$1$}
6:
7: \author[C. Consani and C. Faber]{Caterina Consani and Carel Faber}
8:
9: \address{
10: {\rm Caterina Consani}\\
11: Department of Mathematics \hfill Department of Mathematics\\
12: The Johns Hopkins University \hfill University of Toronto\\
13: 3400 North Charles Street \hfill Toronto, Ontario\\
14: Baltimore, MD 21218, U.S.A. \hfill Canada M5S 3G3\\
15: kc@math.jhu.edu \hfill kc@math.toronto.edu\\
16: \\
17: {\rm Carel Faber}\\
18: Institutionen f\"or Matematik \hfill Department of Mathematics\\
19: Kungliga Tekniska H\"ogskolan \hfill The Johns Hopkins University\\
20: 100 44 Stockholm \hfill 3400 North Charles Street\\
21: Sweden \hfill Baltimore, MD 21218, U.S.A.\\
22: faber@kth.se \hfill faber@math.jhu.edu}
23:
24: \rcvdate{April 20, 2005}% to be supplied by the volume editor
25: \rvsdate{December 23, 2005}% to be supplied by the volume editor
26:
27: \thanks{This paper owes a great debt to the work of
28: Ezra Getzler. We thank Mathematisches Forschungsinstitut Oberwolfach, where
29: we began this work during a ``Research in Pairs'' stay in May 2004.
30: This research was supported in part by NSERC grant 72024520 and by grants from
31: the Swedish Research Council and the G\"oran Gustafsson Foundation.}
32:
33: \setcounter{firstpage}{1} % to be supplied by the volume editor
34: \setcounter{lastpage}{18} % to be supplied by the volume editor
35:
36:
37: \hyphenation{Looijenga}
38: \hyphenation{Getzler}
39:
40: \newtheorem{pr}{Proposition}
41: \newtheorem{lm}{Lemma}
42: \newtheorem{tm}{Theorem}
43:
44:
45: \newcommand{\proj}{\mathbf P}
46: \newcommand{\barr}{\overline}
47: \newcommand{\rarr}{\rightarrow}
48: \newcommand{\oh}{{\mathcal{O}}}
49: \newcommand{\com}{\mathbb{C}}
50:
51: \newcommand{\D}{{\mathcal{D}}}
52: \newcommand{\E}{{\mathcal{E}}}
53: \newcommand{\SSS}{\mathsf{S}}
54:
55: \newcommand{\F}{\mathbf{F}}
56: \newcommand{\G}{\mathbf{G}}
57: \newcommand{\Q}{\mathbb{Q}}
58: \newcommand{\V}{\mathbf{V}}
59: \newcommand{\Z}{\mathbb{Z}}
60: \newcommand{\C}{\mathbb{C}}
61:
62: \newcommand{\lan}{\langle}
63: \newcommand{\ran}{\rangle}
64: \newcommand{\eqq}{\stackrel{\sim}{=}}
65: \newcommand{\deli}{\bigtriangleup}
66: \newcommand{\sumo}{\oplus}
67: \newcommand{\grad}{\nabla}
68:
69: \newcommand{\GG}{{\mathbf{g}}}
70: \newcommand{\NN}{{\mathbf{n}}}
71: \newcommand{\DD}{{\mathbf{d}}}
72: \newcommand{\mumu}{{\mathbf{\mu}}}
73: \newcommand{\M}{{\overline M}}
74: \newcommand{\len}{\ell}
75:
76: \newcommand{\mgn}{\overline{M}_{g,n}}
77:
78:
79: \def\scup{\mathbin{\text{\scriptsize$\cup$}}}
80: \def\scap{\mathbin{\text{\scriptsize$\cap$}}}
81:
82: \newcommand{\bpf}{\noindent {\em Proof.} }
83: \newcommand{\epf}{\qed \vspace{+10pt}}
84: \newcommand{\LL}{\mathbb{L}}
85: \newcommand{\hodge}{\mathbb{E}}
86: \newcommand{\rk}{\text{rk}}
87: \newcommand{\CM}{\mathcal{M}}
88: \newcommand{\UM}{\mathcal{C}}
89:
90: \newcommand{\mata}{\mathbf{A}}
91: \newcommand{\matb}{\mathbf{B}}
92: \newcommand{\por}{\stackrel{\rarr}{\mathcal{P}}}
93: \newcommand{\pord}{{\por}}
94: \newcommand{\ssym}{\text{sym}}
95: \newcommand{\matc}{\mathbf{C}}
96: \newcommand{\matd}{\mathbf{D}}
97:
98: \newcommand{\im}{\text{Im}}
99: \newcommand{\Aut}{\text{Aut}}
100: \newcommand{\qbar}{\hat{q}}
101: \newcommand{\pbar}{\hat{p}}
102: \newcommand{\rbar}{\hat{r}}
103: \newcommand{\qti}{\tilde{q}}
104: \newcommand{\pti}{\tilde{p}}
105: \newcommand{\rti}{\tilde{r}}
106:
107: \newcommand{\Vlm}{\V_{l,m}}
108: \newcommand{\bdy}{A_{1,1}}
109: \newcommand{\mbar}[1]{{\overline{M}}_{#1}}
110: \newcommand{\m}[1]{M_{#1}}
111: \newcommand{\bm}[1]{{\partial M}_{#1}}
112: \newcommand{\Sn}{\Sigma_n}
113: \newcommand{\sn}{\Sigma_{n-1}}
114: \newcommand{\bA}[1]{{\bf a}_{#1}}
115: \newcommand{\bB}[1]{{\bf b}_{#1}}
116: \newcommand{\Alt}{{\rm{Alt}}}
117: \newcommand{\p}{\partial}
118: \newcommand{\ch}{{\rm{ch}}}
119: \newcommand{\Ch}{{\rm{Ch}}}
120: \newcommand{\sgn}{{\rm{sgn}}}
121: \def\Lie{{\operatorname{\mathcal{L}\mathit{ie}}}}
122: \def\({(\!(}
123: \def\){)\!)}
124:
125: \newcommand{\ie}{{\it i.e.\/}\ }
126: \newcommand{\eg}{{\it e.g.\/}\ }
127: \newcommand{\cf}{{\it cf.\/}\ }
128: \newcommand{\opcit}{{\it op.cit.\/}\ }
129: \newcommand{\resp}{{\it resp.\/}\ }
130: \newcommand{\no}{\noindent}
131:
132: \newcommand{\De}[1]{\Delta_{#1}}
133: \newcommand{\Do}[1]{\Delta^{\circ}_{#1}}
134: \newcommand{\Ind}{{\rm Ind}}
135: \newcommand{\Sym}{{\rm Sym}}
136: \newcommand{\pp}{{\rm pr}}
137:
138: \begin{document}
139:
140: \begin{abstract}
141: We prove that the moduli space of stable $n$-pointed curves of genus~$1$
142: and the projector associated to the alternating representation of the
143: symmetric group on $n$ letters define (for $n>1$) the Chow motive
144: corresponding to cusp forms of weight $n+1$ for ${\rm SL}(2,\Z)$.
145: This provides an alternative (in level~$1$) to the construction of Scholl.
146: \end{abstract}
147:
148: \maketitle
149:
150: \setcounter{section}{0}
151: \section{Introduction}
152: In this paper we give an alternative construction of the Chow motives
153: $S[k]$ corresponding to cusp forms of weight $k$ for ${\rm SL}(2,\Z)$.
154: The Betti cohomology related to these cusp forms was initially studied
155: by Eichler and Shimura, after which Deligne constructed the corresponding
156: $\ell$-adic Galois representations. Using the canonical desingularization
157: of the fiber products of the compactified universal elliptic curve
158: constructed by Deligne, Scholl then defined projectors
159: such that the realizations of the associated Chow motives are
160: these parabolic cohomology groups.
161: The smooth projective varieties
162: used in this construction are called Kuga-Sato varieties.
163:
164: Instead of the Kuga-Sato varieties, we use the spaces $\mbar{1,n}$, the
165: Knudsen-Deligne-Mumford moduli spaces of stable $n$-pointed curves
166: of genus~$1$. The symmetric group $\Sn$ acts naturally on $\mbar{1,n}$,
167: by permuting the $n$ marked points. Let $\alpha$ denote its alternating
168: character. Our main result is that $\mbar{1,n}$ (for $n>1$) and the projector
169: $\Pi_{\alpha}$ corresponding to $\alpha$ define the Chow motive $S[n+1]$.
170: In other words, we have the following result.
171: \par\medskip
172: \noindent {\bf Theorem.} {\em
173: For $n>1$,}
174: $$\Pi_{\alpha}(H^*(\mbar{1,n},\Q))=\Pi_{\alpha}(H^n(\mbar{1,n},\Q))
175: =H^1_!(\m{1,1},\Sym^{n-1}R^1\pi_*\Q).$$
176: \noindent Here $\pi:\E\to\m{1,1}$ is the universal elliptic curve and
177: $H^i_!={\rm Im}(H^i_c\to H^i)$ denotes the parabolic cohomology.
178:
179: The cohomology $H^*(\mbar{g,n})$ of the moduli space of stable $n$-pointed
180: curves of genus~$g$ has been studied intensively in recent years, in
181: particular for $n>0$ through the connection with Gromov-Witten theory.
182: Since $\mbar{g,n}$ is a smooth projective stack over $\Z$, these groups
183: have arithmetic relevance as well. Getzler has initiated the study of the
184: cohomology $H^*(\m{g,n})$ of the moduli space of smooth $n$-pointed curves
185: of genus~$g$ as a representation of $\Sn$. Through the theory of
186: modular operads, as developed by Getzler and Kapranov, the $\Sn$-equivariant
187: Euler characteristics of the cohomology of
188: the spaces $\mbar{g,n}$ are expressed
189: in the $\Sn$-equivariant Euler characteristics of the cohomology of
190: the spaces $\m{g,n}$. The action of $\Sn$ is
191: crucial here. Another central idea
192: of Getzler is to express the $\Sn$-equivariant Euler characteristic of
193: $H^*(\m{g,n})$ in terms of the Euler characteristics of the cohomology
194: of irreducible symplectic local systems on $\m{g}$. Since these local systems
195: are pulled back from the moduli space $A_g$ of principally polarized
196: abelian varieties of dimension~$g$, this provides a connection with genus~$g$
197: Siegel modular forms.
198:
199: In genus~$1$, this connection is given by Eichler-Shimura theory. In higher
200: genus, despite the very important work of Faltings and Chai, much less is
201: known. Van der Geer and the second author have obtained an explicit
202: conjectural formula for the motivic Euler characteristics of these local systems
203: in genus~$2$.
204:
205: Our work is motivated by the desire to understand the motives underlying
206: Siegel modular forms and the cohomology of the corresponding local systems.
207: We expect that the results proved in this paper for genus~$1$, when
208: suitably generalized, will provide a major step towards this goal.
209:
210: Unbeknownst to us, Manin had suggested in \cite{Ma1},~0.2 and~2.5, that
211: it would be desirable to replace the Kuga-Sato varieties by
212: moduli spaces of curves of genus~$1$ with marked points and a level
213: structure. Cf.~\cite{Ma2},~3.6.2.
214:
215: In section~2, we determine the alternating part of the $\Sn$-equivariant
216: Euler characteristic of $\m{1,n}$. Section~3 deals with the
217: $\Sn$-equivariant cohomology of $\m{0,n}$; some of the results obtained
218: here may be of independent interest. The theory of modular operads
219: and the results of section~3 are used in section~4 to determine the
220: alternating part of the Euler characteristic of $\mbar{1,n}\setminus\m{1,n}$.
221: In section~5 we combine the results of sections~2 and 4 and prove our main
222: theorem.
223:
224: %\newpage
225: \section{The contribution of the interior}
226: In this section we determine the contribution of
227: $\m{1,n}$,
228: i.e., we determine
229: $$\lan s_{1^n},e_c^{\Sn}(\m{1,n})\ran,$$
230: the alternating part of the $\Sn$-equivariant Euler characteristic
231: of the compactly supported cohomology
232: of $\m{1,n}$.
233: Here, for a partition~$\lambda$ of~$n$, the notation~$s_{\lambda}$ is used
234: for the Schur function corresponding to the irreducible representation
235: of~$\Sn$ indexed by~$\lambda$, and~$\lan\,,\,\ran$ stands for the
236: standard inner product on the ring of symmetric functions, for which
237: the~$s_{\lambda}$ form an orthonormal basis.
238: We will usually not make a notational distinction between a (possibly
239: virtual) $\Sn$-representation~$V$ and its characteristic~$\ch_n(V)$,
240: the symmetric function corresponding to it (\cite{GK},~7.1).
241: The Euler characteristic is taken in $K_0$ of a convenient category, such as
242: the category of mixed Hodge structures or of $\ell$-adic Galois representations.
243:
244: Let $E=(E,0)$ be an elliptic curve. We may think of the points
245: of $E^{n-1}$ as $n$-tuples
246: $$(0,x_2,\dots,x_n)$$
247: (with $x_1=0$)
248: and by doing so we find a natural action of $\Sn$ on $E^{n-1}$
249: (combine the effect of a permutation $\sigma$ with a translation
250: of each coordinate over $-x_{\sigma^{-1}(1)}$).
251: We are interested in the subspace of $H^\bullet(E^{n-1})
252: =H^\bullet(E)^{\otimes(n-1)}$ where the induced action of $\Sn$
253: is via the alternating representation.
254:
255: Let $\sn\subset\Sn$ be the subgroup permuting the last $n-1$ entries.
256: \begin{lm} The subspace of $H^\bullet(E)^{\otimes(n-1)}$
257: where the induced action of $\sn$ is
258: via the alternating representation is isomorphic to
259: $$\oplus_{k=0}^{n-1}
260: \wedge^k H^{{\rm even}}(E)\otimes\Sym^{n-1-k}H^1(E).$$
261: \end{lm}
262:
263: \bpf
264: The subspace of $V=H^\bullet(E)^{\otimes(n-1)}$ where $\sn$ acts
265: alternatingly is generated by sums
266: $$\sum_{\sigma\in\sn}(-1)^{\sgn(\sigma)}\sigma^*(v)$$
267: with $v\in V$.
268: Clearly, we may restrict ourselves to pure tensors $v$ such that the first
269: $k$ factors are in $H^{{\rm even}}(E)$ and the remaining $n-1-k$ factors
270: are in $H^1(E)$, for some $k$. Fix $k$. It suffices now to consider the
271: action of $\Sigma_k\times\Sigma_{n-1-k}$ on such $v$. This leads to the
272: claimed isomorphism.\epf
273:
274: Note that only the terms with $k\le2$ in the direct sum above are
275: nonzero.
276: Thus it is concentrated in degrees $n-2$, $n-1$, and $n$.
277: \begin{pr} The subspace of $H^\bullet(E)^{\otimes(n-1)}$ where
278: the induced action of $\Sn$ is via the alternating representation
279: is $\Sym^{n-1}H^1(E)$.
280: \end{pr}
281:
282: \bpf Let $\tau\in\Sn$ be the transposition $(12)$. We need to show
283: that $\tau^*(\gamma)=-\gamma$ for all
284: $\gamma\in\Sym^{n-1}H^1(E)$, but that none of the
285: $\sn$-alternating vectors
286: %corresponding to
287: coming from %%CF Jan 5 06, to avoid overfull boxes
288: elements of $H^{{\rm
289: even}}(E)\otimes\Sym^{n-2}H^1(E)$ and $\wedge^2 H^{{\rm
290: even}}(E)\otimes\Sym^{n-3}H^1(E)$ have this property.
291:
292: As an example, consider the case $n=2$. Note that
293: $\tau(0,x_2)=(x_2,0)=(0,-x_2)$. Thus $\tau=-1_E$ and the
294: $(-1)$-eigenspace of $\tau^*$ on $H^\bullet(E)$ is $H^1(E)$.
295:
296: In the general case,
297: $$\tau(0,x_2,x_3,\dots,x_n)=(0,-x_2,x_3-x_2,\dots,x_n-x_2).$$
298: Denote by $\pp_i:E^{n-1}\to E$ the projection onto the $i$th
299: factor (with $2\le i\le n$) and by $\tau_i$ the composition
300: $\pp_i\circ\tau$. Then
301: $$\tau^*(\gamma_2\otimes\dots\otimes\gamma_n)=
302: \tau_2^*(\gamma_2)\cdot\ldots\cdot\tau_n^*(\gamma_n).$$ For
303: $k\ge3$ we have $\tau_k=m\circ((-\pp_2)\times\pp_k)$, where
304: $m:E\times E\to E$ denotes the group law. Observe now that
305: $$\tau_k^*(\zeta)=-\pp_2^*(\zeta)+\pp_k^*(\zeta)$$
306: for $\zeta\in H^1(E)$ and $k\ge3$.
307:
308: Denote by $p_i:E^{n-1}\to E^{n-2}$ the projection forgetting the
309: $i$th factor ($2\le i\le n$). Let $\gamma\in\Sym^{n-2}H^1(E)$ and
310: denote by
311: $$\Gamma=\sum_{i=2}^n (-1)^i p_i^*\gamma$$
312: the $\sn$-alternating vector corresponding to $1\otimes\gamma$.
313: Let $I$ be the ideal $\pp_2^*(H^1(E)\oplus H^2(E))$. Note that
314: $\Gamma\equiv p_2^*\gamma=1\otimes\gamma\mod I$. But
315: $\tau^*(1\otimes\gamma)\equiv 1\otimes\gamma\mod I$ by the above.
316: Thus $\tau^*\Gamma=-\Gamma$ implies $\gamma=0$.
317:
318: This shows that the $\sn$-alternating vectors corresponding to
319: elements of $H^0(E)\otimes\Sym^{n-2}H^1(E)$ are not
320: $\Sn$-alternating. We conclude that the alternating representation
321: of $\Sn$ does not occur in degree $n-2$. By duality, it does not
322: occur in degree $n$ either.
323:
324: Denote by $p_{ij}:E^{n-1}\to E^{n-3}$ the projection forgetting
325: the $i$th and $j$th factors ($2\le i<j\le n$). Let
326: $\gamma\in\Sym^{n-3}H^1(E)$ and denote by
327: $$\Xi=\sum_{i=2}^{n-1}\sum_{j=i+1}^n (-1)^{i+j}
328: p_{ij}^*\gamma\cdot(\pp_i^*p-\pp_j^*p)$$ the $\sn$-alternating
329: vector corresponding to $(1\wedge p)\otimes\gamma$ (here $p$ is
330: the class of a point). Then
331: $$(-1)^{n+1}p_{n*}\Xi=p_{n*}\big(\sum_{i=2}^{n-1}(-1)^i
332: p_{in}^*\gamma\cdot\pp_n^*p\big) =\sum_{i=2}^{n-1} (-1)^i
333: p_i^*\gamma,$$ the $\Sigma_{n-2}$-alternating vector in
334: $H^\bullet(E)^{\otimes(n-2)}$ corresponding to $1\otimes\gamma$.
335: Using that $p_n\circ\tau=\tau\circ p_n$, one shows that
336: $\tau^*\Xi=-\Xi$ implies $\gamma=0$. Thus the alternating
337: representation of $\Sn$ can occur only in $\Sym^{n-1}H^1(E)$.
338:
339: %Finally,
340: To conclude, %%CF Jan 5 06, to avoid overfull boxes
341: we show that these vectors are indeed $\Sn$-alternating.
342: Choose $\alpha$ and $\beta$ in $H^1(E)$ with $\alpha\cdot\beta=p$. Fix
343: $k$ and $l$ with sum $n-1$ and let
344: $\gamma=\gamma_2\otimes\dots\otimes\gamma_n$
345: with $k$ of the factors equal to $\alpha$ and the remaining $l$
346: equal to $\beta$.
347: If $\gamma_2=\alpha$, then $\gamma+\tau^*\gamma$ is a sum of $l$ terms;
348: each term arises from $\gamma$ by replacing $\gamma_2$ by $p$ and one
349: of the $\beta$'s by $1$. If $\gamma_2=\beta$, then $\gamma+\tau^*\gamma$
350: is a sum of $k$ terms; each term arises from $\gamma$ by
351: replacing $\gamma_2$ by $-p$ and one of the $\alpha$'s by $1$. It is
352: now easy to see that the symmetric tensor $\Gamma$ that is the sum of
353: all $\gamma$ satisfies $\tau^*\Gamma=-\Gamma$. This finishes the proof.\epf
354:
355: We may think of the fiber of $\m{1,n}$ over $[E]$ as the open
356: subset $D^\circ_n$ of $E^{n-1}$ where the $n$ points $0,x_2,\dots,x_n$
357: are mutually distinct, i.e., the complement of the $n-1$ zero
358: sections $x_i=0$ (with $2\le i\le n$) and the diagonals $x_i=x_j$
359: (with $2\le i<j\le n$). Clearly this open subset is $\Sn$-invariant.
360: \begin{lm}
361: The subspace of $H^\bullet_c(D^\circ_n)$
362: where the induced action of $\Sn$ is
363: via the alternating representation is canonically isomorphic to
364: the corresponding subspace of $H^\bullet(E^{n-1})$, thus to
365: $\Sym^{n-1}H^1(E)$.
366: \end{lm}
367:
368: \bpf Write $D_k$ for the closed subset of $E^{n-1}$ where $\{0,x_2,\dots,x_n\}$
369: has cardinality at most $k$ and $D_k^\circ=D_k\setminus D_{k-1}$ for its
370: open subset where $\{0,x_2,\dots,x_n\}$ has cardinality $k$. The subsets
371: $D_k$ and $D_k^\circ$ are $\Sn$-invariant. By induction on $k$ we show that
372: $H^\bullet_c(D_k)$ does not contain a copy of the alternating representation
373: for $k\le n-1$. Note that $D_1$ is a point. We may assume $n>2$. We have
374: exact sequences
375: $$H^{i-1}_c(D_{k-1})\to H^i_c(D_k^\circ) \to H^i_c(D_k) \to H^i_c(D_{k-1})$$
376: of $\Sn$-representations. By induction, the outer terms do not contain
377: alternating representations. Consider $D_k^\circ$ for $k\le n-1$. For every
378: connected component, there exists
379: a transposition
380: in $\Sn$ acting on it as the
381: identity. This shows that $H^\bullet_c(D_k^\circ)$ does not contain an
382: alternating representation, and the same holds for $H^\bullet_c(D_k)$.
383: The exact sequence above, with $k=n$, now gives the result.\epf
384:
385: For a variety $X$ with $\Sn$-action, denote $\lan s_{1^n},e_c^{\Sn}(X)\ran$
386: by $A_c(X)$. Clearly, we have
387: $$A_c(E^{n-1})=A_c(D_n^\circ)=(-1)^{n-1}\Sym^{n-1}H^1(E).$$
388: Let $\pi:\E\to S$ be a relative elliptic curve. We may consider the
389: $\Sn$-action on the relative spaces $\E^{n-1}/S$ and $\D_n^\circ/S$
390: and obtain
391: $$A_c(\E^{n-1}/S)=A_c(\D_n^\circ/S)=(-1)^{n-1}\Sym^{n-1}R^1\pi_*\Q$$
392: and similarly with $\Q_{\ell}$-coefficients. The Leray spectral sequence
393: gives then immediately
394: $$A_c(\E^{n-1})=A_c(\D_n^\circ)=(-1)^{n-1}e_c(S,\Sym^{n-1}R^1\pi_*\Q).$$
395: Applying this to the universal elliptic curve, we obtain in particular
396: $$A_c(\m{1,n})=(-1)^{n-1}e_c(\m{1,1},\Sym^{n-1}R^1\pi_*\Q).$$
397:
398: Let $n>1$. Then $H^i_c(\m{1,1},\Sym^{n-1}R^1\pi_*\Q)=0$ when $i\neq1$
399: or $n$ even. For $n$ odd,
400: $$H^1_c(\m{1,1},\Sym^{n-1}R^1\pi_*\Q)=S[n+1]+1,$$
401: cf.~\cite{Ge4}, Thm.~5.3 and below. Here we have written $1$ for the
402: trivial Hodge structure $\Q$ (or the corresponding $\ell$-adic Galois
403: representation) and $S[n+1]$ for Getzler's $\SSS_{n+1}$; this is an
404: equality in the Grothendieck group of our category.
405: We have proved the following result.
406: \begin{tm}
407: \label{interior}
408: The alternating part of the $\Sn$-equivariant Euler characteristic
409: of the compactly supported cohomology of~$\m{1,n}$ is given by the
410: following formula:
411: $$A_c(\m{1,n})=
412: \left\{
413: \begin{array}{rl}
414: -S[n+1]-1,&\quad n>1\,\,{\rm odd;}\\
415: 0,&\quad n\,\,{\rm even.}\\
416: \end{array}
417: \right.
418: $$
419: Here $S[n+1]=H^1_!(\m{1,1},\Sym^{n-1}R^1\pi_*\Q)$, the parabolic
420: cohomology of the local system~$\Sym^{n-1}R^1\pi_*\Q$, is the part
421: of the cohomology of~$\m{1,n}$ corresponding to cusp forms of weight~$n+1$.
422: \end{tm}
423: Of course $A_c(\m{1,1})=L$, the Hodge structure $\Q(-1)$. If we
424: formally define $S[2]=-L-1$, then the formula above holds for
425: $n=1$ as well.
426:
427:
428: %\newpage
429: \section{Cohomology of genus~$0$ moduli spaces and representations
430: of symmetric groups}
431:
432: In this section we study the cohomology groups $H^i(\m{0,n})$
433: as representations of the symmetric group $\Sn$. One of our main tools
434: is the following. Let $X$ be an algebraic variety, let $Y\subset X$
435: be a closed subvariety, and let $U=X\setminus Y$ denote the complement. Then
436: the long exact sequence of compactly supported cohomology
437: $$\dots\to H^k_c(U)\to H^k_c(X)\to H^k_c(Y)\to H^{k+1}_c(U)\to\dots$$
438: is a sequence of mixed Hodge structures. See~\cite{DK},~p.~282.
439:
440: \begin{lm} {\rm (Getzler)}
441: \label{wt2i}
442: The mixed Hodge structure on $H^i(\m{0,n})$ is pure of weight $2i$.
443: \end{lm}
444:
445: \bpf This is Lemma~3.12 in \cite{Ge1}. We wish to give a different proof
446: here. The case $n=3$ is trivial. For $n=4$, we use the sequence above,
447: with $X={\bf P}^1$, $Y=\{0,1,\infty\}$, and $U=\m{0,4}$. The sequence reads
448: $$0\to H^0_c(\m{0,4})\to H^0_c({\bf P}^1) \to H^0_c(\{0,1,\infty\})
449: \to H^1_c(\m{0,4})\to 0 \to 0 \to $$
450: $$\to H^2_c(\m{0,4})\to H^2_c({\bf P}^1) \to 0. $$
451: Note first that $H^0_c(\m{0,4})=0$. Clearly, $H^1_c(\m{0,4})$ has weight $0$
452: and $H^2_c(\m{0,4})$ has weight $2$. The statement follows by duality:
453: $$H^k_c(V)^{\vee}\cong H^{2m-k}(V)(m)$$
454: as mixed Hodge structures, for $V$ a nonsingular irreducible
455: %$m$-dimensional variety.
456: variety of dimension~$m$. %%CF Jan 5 06, to avoid overfull boxes
457:
458: For $n>4$ we have that $U=\m{0,n}$ is isomorphic to the complement in
459: $X=\m{0,n-1}\times \m{0,4}$ of the disjoint union
460: $$Y=\coprod_{i=4}^{n-1} \{x_i=x_n\}\,,$$
461: where we think of a $k$-pointed curve of genus~$0$ as given by a $k$-tuple
462: $(0,1,\infty,x_4,\dots,x_k)$ on ${\bf P}^1$. Thus,
463: $$H^{k-1}_c(Y)\to H^k_c(U)\to H^k_c(X)$$
464: is an exact sequence of mixed Hodge structures. By dualizing and
465: %tensoring with $\Q(3-n)$,
466: applying a Tate twist,
467: the same holds for
468: $$H^i(X)\to H^i(U) \to H^{i-1}(Y)(-1)$$
469: (with $i=2(n-3)-k$).
470: By the K\"unneth formula and induction on $n$, the terms on the left and right
471: have pure Hodge structures of weight $2i$. Hence the same holds for the
472: term in the middle.
473: \epf
474:
475: For $k\ge0$, denote by $\De{k}$ the closed part of $\mbar{0,n}$
476: corresponding to stable curves with at least $k$ nodes and denote
477: by $\Do{k}$ the open part $\De{k}\setminus\De{k+1}$ corresponding to
478: stable curves with exactly $k$ nodes. Put $d=n-3$. Clearly,
479: $\De{k}\neq\emptyset$ for $0\le k\le d$. In general, $\De{k}$ is
480: singular, with nonsingular irreducible components, all of
481: codimension~$k$. But $\De0=\mbar{0,n}$ and $\De{d}$ (a collection of
482: points) are nonsingular. All $\Do{k}$ are nonsingular. Of course
483: $\Do{0}=\m{0,n}$ and $\Do{d}=\De{d}$. We have the long exact
484: sequence
485: $$\dots\to H^{a-1}_c(\De{k+1})\to H^a_c(\Do{k})\to H^a_c(\De{k})\to
486: H^a_c(\De{k+1})\to \dots$$ of mixed Hodge structures. Since the
487: $\De{k}$ are invariant for the natural action of $\Sn$, it is also
488: a sequence of $\Sn$-representations.
489:
490: \begin{lm}
491: \label{rows0}
492: The cohomology groups $H^i(\m{0,n})$ vanish for $i>n-3$. For $0\le i\le n-3$,
493: the irreducible representations of $\Sn$ occurring in $H^i(\m{0,n})$
494: have Young diagrams with at most $i+1$ rows.
495: In particular, the irreducible representations of $\Sn$ occurring in
496: $H^\bullet(\m{0,n})$ have Young diagrams with at most $n-2$ rows.
497: \end{lm}
498:
499: \bpf The claimed vanishing is immediate. Let us abbreviate the
500: rest of the statement by ``$H^i(\m{0,n})$ has $\le i+1$ rows''. We
501: prove it by induction on $n$. The case $n=3$ is trivial. Assume
502: $n>3$.
503: % insertion Jan 6 06
504: Recall that $d=n-3$. %%
505: We require an analysis of the boundary strata:
506:
507: \smallskip \noindent
508: {\bf Claim.} Assume $d-b>0$. Then $H^a_c(\De{d-b})$ has $\le
509: d+1+b-a$ rows.
510:
511: \smallskip
512: \noindent We prove the claim by induction on $b$. We begin with
513: the case $b=0$. Since $\De{d}$ is a collection of points, $a=0$ may
514: be assumed. Each point corresponds to a stable curve with $d$
515: nodes, hence with $d+1$ components. Each component has exactly
516: three special points (nodes or marked points). Let $n_j$ be the
517: number of marked points on the $j$th component, for some numbering
518: of the components. By permuting the $n$ marked points on the
519: stable curve, we obtain a $\Sn$-representa\-tion $R$, which is a
520: direct summand of $H^0(\De{d})$. Note that $R$ is a
521: subrepresentation of the induced representation
522: $$\Ind^{\Sn}_{\prod_{j=1}^{d+1}\Sigma_{n_j}}{\bf 1}.$$
523: The induced representation has $\le d+1$ rows, hence $R$ does. Now
524: $H^0(\De{d})$ is a direct sum of representations analogous to $R$,
525: thus it has $\le d+1$ rows as well. This proves the claim in the
526: case $b=0$.
527:
528: Assume $b>0$. Observe that $H^a_c(\Do{k})\cong
529: H^{2(d-k)-a}(\Do{k})$ as $\Sn$-representa\-tions. Also, each
530: connected component of $\Do{k}$ is for $k\ge1$ a product of $k+1$
531: spaces $\m{0,m_j}$, with $m_j<n$. By induction on $n$ and the
532: K\"unneth formula, $H^a_c(\Do{k})$ has $\le 2(d-k)-a+k+1=2d-k-a+1$
533: rows, for $k\ge1$. Putting $k=d-b$, we find that $H^a_c(\Do{d-b})$
534: has $\le d+1+b-a$ rows.
535:
536: By induction on $b$, we have that $H^a_c(\De{d-b+1})$ has $\le
537: d+b-a$ rows. From the long exact sequence, we find that
538: $H^a_c(\De{d-b})$ has $\le d+1+b-a$ rows. This proves the claim.
539:
540: In particular, $H^a_c(\De1)$ has $\le 2d-a$ rows. Consider the
541: exact sequence
542: $$ H^{k-1}_c(\De1)\to H^k_c(\m{0,n})\stackrel{\alpha}{\to}
543: H^k_c(\mbar{0,n}).$$ From Lemma~\ref{wt2i} we know that
544: $H^k_c(\m{0,n})$ has weight $2k-2d$. But $H^k_c(\mbar{0,n})$ has
545: weight $k$. Thus $\alpha=0$ for $k<2d$. Hence $H^k_c(\m{0,n})$ has
546: $\le 2d+1-k$ rows for $k<2d$. Thus $H^i(\m{0,n})$ has $\le i+1$
547: rows for $i>0$. But it is obviously true for $i=0$ as well. This
548: finishes the proof.\epf
549:
550:
551: %\newpage
552: \section{The contribution of the boundary}\label{bdy}
553: In this section we determine the contribution of the boundary
554: $$\bm{1,n}=\mbar{1,n}\setminus\m{1,n},$$
555: i.e., we determine
556: $$\lan s_{1^n},e_c^{\Sn}(\bm{1,n})\ran.$$
557: We use the main result of \cite{Ge2}.
558: To state it, we introduce the following notations:
559: \begin{eqnarray*}
560: &&{\bf a}_g := \sum_{n>2-2g} \ch_n(e_c^{\Sn}(\m{g,n})),\quad{\text {and}}\\
561: &&{\bf b}_g := \sum_{n>2-2g} \ch_n(e_c^{\Sn}(\mbar{g,n})).
562: \end{eqnarray*}
563: Here $\ch_n$
564: %stands for
565: denotes %%CF Jan 5 06, to avoid overfull boxes
566: the characteristic of a finite-dimensional
567: $\Sn$-represen\-tation (\cite{GK},~7.1) and its extension by linearity
568: to virtual representations.
569: For a (formal) symmetric function $f$ (such as ${\bf a}_g$ and ${\bf b}_g$),
570: we also write
571: \begin{eqnarray*}
572: &&f'=\frac{\partial f}{\partial p_1}=p_1^\perp f,\\
573: &&\dot{f}=\frac{\partial f}{\partial p_2}=\frac12\,p_2^\perp f,\\
574: &&\psi_i(f)=p_i\circ f.
575: \end{eqnarray*}
576: Here $p_i$ is the symmetric function equal to the sum of the $i$th powers
577: of the variables,
578: $p_i^\perp$ is the adjoint of multiplication with $p_i$
579: with respect to the standard inner product,
580: and $\circ$ is the plethysm of symmetric functions (\cite{GK},~7.2).
581: We will denote the $i$th complete symmetric function by~$h_i$ and
582: the $i$th elementary symmetric function by~$e_i$.
583:
584: We can now state Getzler's result (Theorem 2.5 in \cite{Ge2}):
585: % ! numbering is different in the published version !
586: $$
587: {\bf b}_1= \biggl( {\bf a}_1 - \frac{1}{2} \sum_{n=1}^\infty
588: \frac{\phi(n)}{n} \log(1-\psi_n({\bf a}_0'')) +
589: \frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0+\tfrac14\psi_2({\bf a}_0'')}
590: {1-\psi_2({\bf a}_0'')} \biggr) \circ
591: (h_1+{\bf b}_0').
592: $$
593: The numerator of the third term inside the big
594: parentheses on the right-hand side
595: has been corrected here; there is a minor computational mistake in the
596: derivation of the theorem in line 4 on page 487, which affects the result
597: (but not Corollary 2.8). % ! numbering is different !
598:
599: As Getzler remarks, the term $\bA1\circ(h_1+\bB0')$
600: corresponds to the sum over graphs obtained by attaching a forest
601: whose vertices have genus~$0$ to a vertex of genus~$1$; in particular,
602: $\bA1\circ h_1=\bA1$, the contribution of smooth curves, is part of this term,
603: corresponding to graphs consisting of a single vertex of genus~$1$.
604: The remainder of this term, corresponding to graphs where at least one vertex
605: of genus~$0$ has been attached to a vertex of genus~$1$, is part of the
606: contribution of the boundary. We show that the alternating representation
607: does not occur here. For a symmetric function
608: $$f=\sum_{n=0}^\infty f_n,$$
609: we write
610: $$\Alt(f)=\sum_{n=0}^\infty \lan s_{1^n},f_n\ran t^n.$$
611:
612: \begin{lm}
613: \label{notrees1}
614: The alternating representation does not occur in the
615: contribution of the part of the boundary
616: of~$\m{1,n}$ corresponding to graphs where at least one vertex of
617: genus~$0$ has been attached to a vertex of genus~$1$. In terms of the
618: notation introduced above:
619: $$\Alt(\bA1\circ(h_1+\bB0'))=\Alt(\bA1).$$
620: \end{lm}
621:
622: \bpf
623: We choose to give a somewhat geometric proof instead of a proof using mostly
624: the language of symmetric functions.
625:
626: Observe first that a boundary stratum corresponding to a graph with a
627: genus~$1$ vertex is isomorphic to a product
628: $$\m{1,m}\times\prod_i \m{0,n_i};$$
629: i.e., it is not necessary to take the quotient by a finite group.
630: (The corresponding graph has no automorphisms: there is a unique
631: shortest
632: path from each of the $n$ legs to the vertex of genus~$1$, and
633: every vertex and every edge lie on such a path.)
634: By the K\"unneth formula,
635: the cohomology of such a product is isomorphic to the tensor product
636: of the cohomologies of the factors.
637:
638: Consider the $\Sn$-orbit of such a stratum. The direct sum of the cohomologies
639: of the strata in the orbit forms a $\Sn$-representation $V$.
640: It is induced from the cohomology of a single stratum, considered as a
641: representation $W$ of the stabilizer $G$ in $\Sn$ of the stratum.
642: By Frobenius Reciprocity, $V$ contains a copy of the
643: alternating representation if and only if $W$ contains a copy of the
644: restriction of the alternating representation to $G$.
645:
646: To each vertex of the graph, one associates the symmetric group corresponding
647: to the legs attached to the vertex.
648: The product over the vertices of these symmetric groups is a subgroup $H$
649: of $G$ and the further restriction of the alternating representation to
650: $H$ is the tensor product over the vertices of the alternating representations
651: of these symmetric groups.
652:
653: Consider a moduli space $\m{0,k}$ corresponding to an extremal vertex of
654: the graph corresponding to a boundary stratum as above.
655: The symmetric group associated to this vertex is a standard
656: subgroup $\Sigma_{k-1}\subset\Sigma_k$, permuting the $k-1$ legs
657: attached to the vertex and leaving the unique half-edge fixed.
658: By Lemma~\ref{rows0}, the irreducible $\Sigma_k$-representations occurring
659: in $H^\bullet(\m{0,k})$ have Young diagrams with at most $k-2$ rows.
660: The Young diagrams of the irreducible representations occurring in the
661: restriction to $\Sigma_{k-1}$ also have at most $k-2$ rows, as they are
662: obtained by removing one box. Therefore the alternating representation
663: does not occur here. It follows that $V$ does not contain a copy of the
664: alternating representation either.
665: \epf
666:
667: We return to Getzler's result. We need to evaluate
668: $$\Alt \biggl( \biggl( - \frac{1}{2} \sum_{n=1}^\infty
669: \frac{\phi(n)}{n} \log(1-\psi_n({\bf a}_0'')) +
670: \frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0+\tfrac14\psi_2({\bf a}_0'')}
671: {1-\psi_2({\bf a}_0'')} \biggr) \circ (h_1+{\bf b}_0') \biggr).
672: $$
673: Getzler remarks that the two terms inside the big inner parentheses may
674: be thought of as a sum over necklaces (graphs consisting of a single
675: circuit) and a correction term, taking into account the fact that
676: necklaces of~$1$ or~$2$ vertices have non-trivial involutions (while
677: those with more vertices do not).
678: The plethysm with $h_1+\bB0'$ stands again for attaching a forest whose
679: vertices have genus~$0$. We begin with the analogue of Lemma~\ref{notrees1}.
680:
681: \begin{lm}
682: \label{notrees0}
683: The alternating representation does not occur in the
684: contribution of the part of the boundary
685: of~$\m{1,n}$ corresponding to graphs where at least one vertex of
686: genus~$0$ has been attached to a necklace. In terms of the
687: notation introduced above:
688: $$
689: \Alt \biggl( \biggl( - \frac{1}{2} \sum_{n=1}^\infty
690: \frac{\phi(n)}{n} \log(1-\psi_n({\bf a}_0'')) +
691: \frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0+\tfrac14\psi_2({\bf a}_0'')}
692: {1-\psi_2({\bf a}_0'')} \biggr) \circ (h_1+{\bf b}_0') \biggr)
693: $$$$
694: =\Alt \biggl( - \frac{1}{2} \sum_{n=1}^\infty
695: \frac{\phi(n)}{n} \log(1-\psi_n({\bf a}_0'')) +
696: \frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0+\tfrac14\psi_2({\bf a}_0'')}
697: {1-\psi_2({\bf a}_0'')} \biggr).
698: $$
699: \end{lm}
700:
701: \bpf
702: In this case, each boundary stratum is isomorphic to a product
703: $$\biggl(\,\prod_{v\in\text{necklace}}\m{0,n(v)}\biggr)\big/I
704: \times \prod_{v\notin\text{necklace}}\m{0,n(v)}.
705: $$
706: The finite group $I$ is trivial when the necklace has at least $3$ vertices.
707: It has $2$ elements when the necklace has $1$ resp.~$2$ vertices and acts
708: by reversing the edge in the necklace resp.~by interchanging the two
709: edges of the necklace. In particular, $I$ acts trivially on the moduli
710: spaces corresponding to the vertices of the forest.
711:
712: Just as in the proof of Lemma~\ref{notrees1}, the alternating representation
713: does not occur in the cohomology of a moduli space $\m{0,k}$
714: corresponding to an extremal vertex of one of the trees of the forest.
715: It follows that the alternating representation does not occur in the
716: cohomology of a $\Sn$-orbit of boundary strata as soon as the forest
717: is nonempty.
718: \epf
719:
720: In order to determine the contribution of the part of the boundary
721: of~$\m{1,n}$ corresponding to necklaces without attached trees, we need
722: several lemmas.
723:
724: \begin{lm}
725: \label{a0''}
726: The restriction of the $\Sn$-representation $H^\bullet(\m{0,n})$
727: to the standard subgroup~$\Sigma_{n-2}$ contains the alternating
728: representation exactly once. In terms of the notation introduced above:
729: $$\Alt(\bA0'')=\frac{t}{1+t}\,.$$
730: \end{lm}
731:
732: \bpf
733: The Young diagrams corresponding to the irreducible representations
734: of $\Sigma_{n-2}$ occurring in $\bA0''$ are
735: obtained by removing $2$ boxes from a Young diagram occurring in
736: $\bA0$. To obtain a copy of the alternating representation of
737: $\Sigma_{n-2}$, one needs to start with a Young diagram with at least
738: $n-2$ rows. From Lemma~\ref{rows0}, only the top cohomology
739: $H^{n-3}(\m{0,n})$ can contribute.
740: Observe that $H^{n-3}(\m{0,n})\cong H_c^{n-3}(\m{0,n})\cong H_{n-3}(\m{0,n})$
741: as $\Sigma_n$-representations.
742: Thus
743: \begin{eqnarray*}
744: \Alt(\bA0'')&=&\Alt\biggl(\frac{\p^2}{\p p_1^2}\sum_{n=3}^\infty
745: \ch_n\bigl(e_c^{\Sn}(\m{0,n})\bigr)\biggr) \\
746: &=&\Alt\biggl(\frac{\p^2}{\p p_1^2}\sum_{n=3}^\infty
747: (-1)^{n-3}\ch_n\bigl(H_c^{n-3}(\m{0,n})\bigr)\biggr).
748: \end{eqnarray*}
749: Getzler shows in \cite{Ge1}, p.~213, l.~3 that
750: $$H_c^{n-3}(\m{0,n}) \cong \sgn_n\otimes \Lie\(n\).$$
751: Here $\sgn_n$
752: %is
753: denotes %%CF Jan 5 06, to avoid overfull boxes
754: the alternating representation and $\Lie\(n\)$
755: %is %%CF Jan 5 06, to avoid overfull boxes
756: the
757: $\Sn$-represen\-tation that is part of the cyclic Lie operad.
758: Getzler and Kapranov show in \cite{GK}, Example 7.24 that
759: $$
760: \Ch(\Lie):=\sum_{n=3}^\infty\ch_n(\Lie\(n\))
761: =(1-p_1)\sum_{n=1}^\infty\frac{\mu(n)}{n}\log(1-p_n)+h_1-h_2,
762: $$
763: where $\mu(n)$ is the M\"obius function.
764: Hence
765: $$
766: \sum_{n=3}^\infty(-1)^{n-3}\ch_n(\sgn_n\otimes\Lie\(n\))
767: =-(1+p_1)\sum_{n=1}^\infty\frac{\mu(n)}{n}\log(1+p_n)+h_1+e_2$$
768: and
769: \begin{eqnarray*}
770: &&
771: \frac{\p^2}{\p p_1^2}
772: \biggl(\,\sum_{n=3}^\infty(-1)^{n-3}\ch_n(\sgn_n\otimes\Lie\(n\))\biggr)
773: \\&&\qquad
774: =\frac{\p}{\p p_1}
775: \biggl(-\sum_{n=1}^\infty\frac{\mu(n)}{n}\log(1+p_n)-
776: (1+p_1)\frac{1}{1+p_1}\biggr)+1
777: \\&&\qquad
778: =1-\frac{1}{1+p_1}
779: =\frac{p_1}{1+p_1}\,.
780: \end{eqnarray*}
781: But
782: $$\Alt\biggl(\frac{p_1}{1+p_1}\biggr)=\frac{t}{1+t}\,,$$
783: since $\lan p_1^n,s_{1^n}\ran=1$.
784: \epf
785:
786: \begin{lm}
787: \label{pk}
788: %Let $f_n$ be a symmetric function of degree $n$. If
789: %$\lan s_{1^n},f_n\ran=0$, then $\lan s_{1^{nk}},p_k\circ f_n\ran=0$.
790: Let $f_n$ be a symmetric function of degree $n$. Assume that
791: $\lan s_{1^n},f_n\ran=0$. Then $\lan s_{1^{nk}},p_k\circ f_n\ran=0$.
792: %%CF Jan 5 06, to avoid overfull boxes
793: \end{lm}
794:
795: \bpf
796: Write $e(\lambda)$ resp.~$o(\lambda)$ for the number of even resp.~odd
797: parts of a partition $\lambda$.
798: For $\lambda$ a partition of $n$,
799: $$ o(\lambda)\equiv n\pmod{2} \qquad{\rm and}\qquad
800: \lan s_{1^n},p_{\lambda}\ran=(-1)^{e(\lambda)}.$$
801: Here $p_{\lambda}=\prod_i p_{\lambda_i}$ is the symmetric function
802: of degree~$n$ that is the product of the power sums corresponding
803: to the parts of~$\lambda$.
804: If $f_n=\sum_{\lambda} a_{\lambda}p_{\lambda}\,$, then $p_k\circ f_n
805: =\sum_{\lambda} a_{\lambda}p_{k\lambda}\,$, where $k\lambda$ is
806: the partition of $kn$ obtained from $\lambda$ by multiplying all parts
807: with $k$. For $k$ odd,
808: $$\sum_{\lambda} a_{\lambda}(-1)^{e(\lambda)}
809: =\sum_{\lambda} a_{\lambda}(-1)^{e(k\lambda)},$$
810: whereas for $k$ even,
811: $$e(k\lambda)=o(\lambda)+e(\lambda)\equiv n+e(\lambda) \pmod{2},$$
812: so that
813: $$\sum_{\lambda} a_{\lambda}(-1)^{e(k\lambda)}=(-1)^n
814: \sum_{\lambda} a_{\lambda}(-1)^{e(\lambda)}.$$
815: The result follows.
816: \epf
817:
818: \begin{lm}
819: \label{psik}
820: The occurrence of the alternating representation observed in
821: Lemma~\ref{a0''} is stable under plethysm with~$p_k$. In terms
822: of the notation introduced above:
823: $$\Alt(\psi_k(\bA0''))=\frac{-(-t)^k}{1-(-t)^k}\,.$$
824: \end{lm}
825:
826: \bpf
827: {From} Lemma~\ref{a0''},
828: $$\Alt(\bA0'')=\Alt\biggl(\frac{p_1}{1+p_1}\biggr)\,.$$
829: Applying Lemma~\ref{pk} and using
830: that $\lan s_{1^{kn}},p_k^n\ran=(-1)^{(k-1)n}$, we find
831: $$
832: \Alt(\psi_k(\bA0''))=\Alt\biggl(p_k\circ\biggl(\frac{p_1}{1+p_1}\biggr)
833: \biggr)=\Alt\biggl(\frac{p_k}{1+p_k}\biggr)
834: $$$$ %%CF Jan 5 06, to avoid overfull boxes
835: =\Alt\biggl(\sum_{n=1}^\infty(-1)^{n-1}p_k^n\biggr)
836: %$$$$
837: =\sum_{n=1}^\infty(-1)^{n-1}(-1)^{(k-1)n}t^{kn}
838: $$$$ %%CF Jan 5 06, to avoid overfull boxes
839: =-\sum_{n=1}^\infty ((-t)^k)^n
840: =\frac{-(-t)^k}{1-(-t)^k}\,.
841: $$
842: \epf
843:
844: \noindent For two symmetric functions $f$ and $g$, we have
845: $$\Alt(fg)=\Alt(f)\Alt(g).$$
846: This follows immediately from the Littlewood-Richardson rule
847: (cf.~\cite{FH}, p.~456). One may use this identity to shorten the proof
848: of Lemma~\ref{psik}. Similarly,
849: $$
850: \Alt(\log(1-\psi_k(\bA0'')))=\log\biggl(1-\frac{-(-t)^k}{1-(-t)^k}\biggr)
851: =-\log(1-(-t)^k)$$
852: and
853: $$
854: \Alt\biggl(-\sum_{n=1}^{\infty}\frac{\phi(n)}{n}\log(1-\psi_n(\bA0'')\biggr)
855: =\sum_{n=1}^{\infty}\frac{\phi(n)}{n}\log(1-(-t)^n)
856: $$$$
857: =-\sum_{n=1}^{\infty}\frac{\phi(n)}{n}\sum_{k=1}^{\infty}\frac{(-t)^{nk}}{k}
858: =-\sum_{m=1}^{\infty}\frac{(-t)^m}{m}\sum_{d|m}\phi(d)
859: $$$$ %%CF Jan 5 06, to avoid overfull boxes
860: =-\sum_{m=1}^{\infty}(-t)^m=\frac{t}{1+t}\,.
861: $$
862: It remains to evaluate the contribution of the correction term,
863: $$
864: \Alt\biggl(\frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0
865: +\tfrac14\psi_2({\bf a}_0'')}{1-\psi_2({\bf a}_0'')} \biggr).
866: $$
867: We need one more lemma.
868:
869: \begin{lm}
870: \label{a0dot}
871: The alternating part of the formal symmetric function
872: $$\dot{{\bf a}}_0=\frac{\p {\bf a}_0}{\p p_2}$$ is given
873: by the following formula:
874: $$\Alt(\dot{{\bf a}}_0)=\frac12\,\frac{t}{1-t}\,.$$
875: \end{lm}
876:
877: \bpf
878: In terms of Young diagrams, multiplication by $s_2$ is the operation
879: of adding two boxes, not in the same column, and multiplication by $s_{1^2}$
880: is the operation of adding two boxes, not in the same row. Now
881: $p_2=s_2-s_{1^2}$ and $\tfrac{\p}{\p p_2}=\tfrac12p_2^{\perp}$, where
882: $p_2^{\perp}$ is the adjoint of multiplication by $p_2$. Thus, to obtain
883: a copy of the alternating representation of $\Sigma_{n-2}$ in a term of
884: $\dot{{\bf a}}_0$, one needs to start with a Young diagram with at least
885: $n-2$ rows, just as in the proof of Lemma~\ref{a0''}. So
886: $$\Alt(\dot{{\bf a}}_0)=\Alt\biggl(\frac{\p}{\p p_2}\sum_{n=3}^\infty
887: (-1)^{n-3}\ch_n\bigl(H_c^{n-3}(\m{0,n})\bigr)\biggr).$$
888: We now find
889: $$
890: \frac{\p}{\p p_2}
891: \biggl(\,\sum_{n=3}^\infty(-1)^{n-3}\ch_n(\sgn_n\otimes\Lie\(n\))\biggr)
892: =\frac12\,\frac{1+p_1}{1+p_2}-\frac12
893: =\frac12\,\frac{p_1-p_2}{1+p_2}\,.
894: $$
895: But
896: $$\Alt\biggl(\frac12\,\frac{p_1-p_2}{1+p_2}\biggr)
897: =\frac12\,\frac{t+t^2}{1-t^2}=\frac12\,\frac{t}{1-t}\,.$$
898: \epf
899:
900: \noindent An easy calculation combining Lemmas~\ref{psik}
901: and~\ref{a0dot} gives
902: $$
903: \Alt\biggl(\frac{{\dot{{\bf a}}_0}^2+\dot{{\bf a}}_0
904: +\tfrac14\psi_2({\bf a}_0'')}{1-\psi_2({\bf a}_0'')} \biggr)
905: =\frac12\,\frac{t}{1-t}\,.
906: $$
907: The contribution from the necklaces becomes then
908: $$
909: \frac12\,\frac{t}{1+t}+\frac12\,\frac{t}{1-t}=\frac{t}{1-t^2}\,,$$
910: i.e., $1$ for $n$ odd and $0$ for $n$ even.
911: We have proved the following result.
912:
913: \begin{tm}
914: \label{boundary}
915: The alternating part of the $\Sn$-equivariant Euler characteristic
916: of the cohomology of~$\bm{1,n}$ is given by the
917: following formula:
918: $$\lan s_{1^n},e_c^{\Sn}(\bm{1,n})\ran=\left\{
919: \begin{array}{rl}
920: 1,&\quad n\,\,{\rm odd;}\\
921: 0,&\quad n\,\,{\rm even.}\\
922: \end{array}
923: \right.
924: $$
925: \end{tm}
926:
927: %\newpage
928: \section{The construction of the motive}
929: The main result of Section~\ref{bdy} (Theorem~\ref{boundary}) is
930: $$A_c(\bm{1,n})=\left\{
931: \begin{array}{rl}
932: 1,&\quad n\,\,{\rm odd;}\\
933: 0,&\quad n\,\,{\rm even.}\\
934: \end{array}
935: \right.
936: $$
937: Combining this with Theorem~\ref{interior}, we immediately obtain
938: $$A_c(\mbar{1,n})=\left\{
939: \begin{array}{rl}
940: -S[n+1],&\quad n\,\,{\rm odd;}\\
941: 0,&\quad n\,\,{\rm even.}\\
942: \end{array}
943: \right.
944: $$
945: Let $n>1$ be an odd integer. The pair consisting of
946: $\mbar{1,n}$ and the projector
947: $$\frac1{n!}\sum_{\sigma\in\Sn}(-1)^{\sgn(\sigma)}\sigma_*$$
948: defines a Chow motive, since $\mbar{1,n}$ is the quotient of a
949: smooth projective variety by a finite group~\cite{BP}.
950: We wish to show that it is pure of
951: degree $n$. This will conclude our construction of the motive
952: $S[n+1]$, an alternative (in level $1$ only) to Scholl's
953: construction. The arguments below are similar to those
954: in~\cite{Sc}, 1.3.4.
955:
956: First, $\lan s_{1^n},H^i_c(\m{1,n})\ran=0$ for $i\neq n$. As in~\cite{De},
957: proof of~5.3, the degeneration of the Leray spectral sequence at $E_2$ due
958: to Lieberman's trick implies that $\lan s_{1^n},H^i_c(\E^{n-1})\ran=0$
959: when $i\neq n$ for a relative elliptic curve $\E\to S$ and this implies
960: the statement.
961:
962: Thus we have an exact sequence
963: $$0 \to H_c^{n-1}(\mbar{1,n})(\alpha) \to H_c^{n-1}(\bm{1,n})(\alpha)
964: \to H^n_c(\m{1,n})(\alpha) \to $$
965: $$\to H^n_c(\mbar{1,n})(\alpha) \to H^n_c(\bm{1,n})(\alpha) \to 0$$
966: and isomorphisms $H^i(\mbar{1,n})(\alpha)\cong
967: H^i(\bm{1,n})(\alpha)$ for $i\notin\{n-1,n\}$. Here $V(\alpha)$
968: denotes the alternating part of a $\Sn$-representation $V$.
969: Therefore $H^i(\bm{1,n})(\alpha)$ is pure of weight $i$ for $i>n$.
970: But then all these spaces vanish, since $H^i(\bm{1,n})$ has weight
971: $\le i$ for all $i$ and since $A_c(\bm{1,n})$ has weight $0$.
972: Hence $H^i(\mbar{1,n})(\alpha)=0$ for $i>n$ and then by duality
973: for $i<n$ as well. This shows that
974: $H^n(\mbar{1,n})(\alpha)=S[n+1]$ and concludes the alternative
975: construction of these motives.
976:
977:
978: \begin{thebibliography}{[Ge4]}
979: \bibitem[BP]{BP} M. Boggi and M. Pikaart, {\em
980: Galois covers of moduli of curves.}
981: Compositio Math. {\bf 120} (2000), no. 2, 171--191.
982: \bibitem[DK]{DK} V. I. Danilov and A. G. Khovanski\u\i, {\em
983: Newton polyhedra and an algorithm for calculating Hodge-Deligne numbers.}
984: Math. USSR-Izv. {\bf 29} (1987), no. 2, 279--298.
985: \bibitem[De]{De} P. Deligne, {\em
986: Formes modulaires et repr\'esentations $\ell$-adiques.}
987: S\'eminaire Bourbaki 1968/69, no. 347, Lecture Notes in Math. 179,
988: Springer-Verlag, Berlin, 1971, 139--172.
989: \bibitem[FH]{FH} W. Fulton and J. Harris, {\em
990: Representation theory. A first course.}
991: Graduate Texts in Mathematics, 129. Readings in Mathematics.
992: Springer-Verlag, New York, 1991.
993: \bibitem[Ge1]{Ge1} E. Getzler, {\em
994: Operads and moduli spaces of genus~$0$ Riemann surfaces.}
995: The moduli space of curves (Texel Island, 1994), 199--230,
996: Progr. Math., 129, Birkh\"auser Boston, Boston, MA, 1995.
997: \bibitem[Ge2]{Ge2} E. Getzler, {\em
998: The semi-classical approximation for modular operads.}
999: Comm. Math. Phys. {\bf 194} (1998), no. 2, 481--492.
1000: \bibitem[Ge3]{Ge3} E. Getzler, {\em
1001: Topological recursion relations in genus~$2$}. Integrable systems and
1002: algebraic geometry (Kobe/Kyoto, 1997), 73--106, World Sci. Publishing,
1003: River Edge, NJ, 1998.
1004: \bibitem[Ge4]{Ge4} E. Getzler, {\em
1005: Resolving mixed Hodge modules on configuration spaces.}
1006: Duke Math. J. {\bf 96} (1999), no. 1, 175--203.
1007: \bibitem[GK]{GK} E. Getzler and M. M. Kapranov, {\em
1008: Modular operads.}
1009: Compositio Math. {\bf 110} (1998), no. 1, 65--126.
1010: \bibitem[Ma1]{Ma1} Yu. Manin, {\em
1011: Iterated integrals of modular forms and noncommutative modular symbols.}
1012: Preprint 2005, math.NT/0502576.
1013: \bibitem[Ma2]{Ma2} Yu. Manin, {\em
1014: Iterated Shimura integrals.}
1015: %Preprint 2005,
1016: %math.NT/0507438.
1017: Preprint, math.NT/0507438. %%CF Jan 5 06, to avoid overfull boxes
1018: \bibitem[Sc]{Sc} A. J. Scholl, {\em
1019: Motives for modular forms.} Invent. Math. {\bf 100} (1990),
1020: 419--430.
1021: \end{thebibliography}
1022: \end{document}
1023:
1024: