1:
2: \section{The virtual normal bundle for toric varieties}
3: \mylabel{sec:virtnormbdl}
4:
5: In this section we analyze Graber and Pandharipande's virtual
6: normal bundle to the fixed point components of the moduli space of
7: stable maps for the natural $(\C^*)^n$--action on a toric variety,
8: hence generalizing Graber and Pandharipande's example for
9: projective space $\C\P^n$ (\cite{gp97}), and we will derive our
10: main result. Contrary to their calculations for $\C\P^n$, however,
11: we will restrict ourselves here to genus zero stable maps.
12:
13: So let $\vt$ be a smooth projective complex variety.
14: Remember from section \ref{subsec:obstrcplxGW}, that for the cohomology sheaves of
15: the dual natural perfect obstruction
16: theory $E_\bullet$ for our moduli stack $\modsp$ of stable maps
17: \mycomment{eq:homolobcom}
18: \begin{equation}\label{eq:homolobcom}
19: 0\mapright \cT^0\mapright E_{0} \MRE{d}
20: E_{1}\mapright \cT^1\mapright 0,
21: \end{equation}
22: %
23: the sheaves $\cT^i$ are given by:
24: %
25: \[ \cT^i = \underline\Ext^i_\pi([f^*\Omega^1_X \longrightarrow
26: \Omega^1_{\ucmodsp/\modsp}(D)], \O_{\ucmodsp}),
27: \quad i=0,1. \]
28:
29: As before we will sometimes write $\cM$ and $\cC$ for the
30: moduli space $\modsp$ and its universal curve $\ucmodsp$, respectively,
31: if no confusion can arise.
32:
33: \begin{lemm}
34: The sheaves $\cT^i$ fit into the following exact sequence:
35: \mycomment{eq:gobcomlong}
36: \begin{equation}\label{eq:gobcomlong}
37: \begin{split}
38: 0&\mapright \underline\Hom_\pi(\Omega^1_{\cC/\cM}(D),
39: \O_{\cC}) \mapright
40: \underline\Hom_\pi(f^*\Omega^1_X, \O_{\cC}) \mapright \cT^0
41: \mapright\\
42: &\mapright \underline\Ext^1_\pi(\Omega^1_{\cC/\cM}(D),
43: \O_{\cC}) \mapright
44: \underline\Ext_\pi^1(f^*\Omega^1_X, \O_{\cC})
45: \mapright \cT^1\mapright 0.\\
46: \end{split}
47: \end{equation}
48: \end{lemm}
49:
50: \begin{proof}
51: Let $\cK^\bullet$ be the complex
52: $\cK^\bullet=[f^*\Omega^1_X\longrightarrow \Omega^1_{\ucmodsp/\modsp}(D)]$
53: indexed at $-1$ and $0$. It fits into the following short exact sequence:
54: \[
55: 0 \mapright \Omega^1_{\cC/\cM}(D) \mapright \cK^\bullet
56: \mapright f^*\Omega^1_X[1] \mapright 0.\]
57: %
58: The corresponding long exact sequence of higher direct image sheaves corresponding
59: to $\pi: \ucmodsp \longrightarrow \modsp$ is then (\ref{eq:gobcomlong} where
60: exactness on the left, \ie. the injectivity of the map
61: ${\underline\Hom}_\pi(\Omega^1_{\cC/\cM}(D), \O_{\cC})
62: \longrightarrow {\underline\Hom}_\pi(f^*\Omega^1_X, \O_{\cC})$
63: induced by the natural map $f^*\Omega^1_X\longrightarrow
64: \Omega^1_{\cC/\cM}\longrightarrow \Omega^1_{\cC/\cM}(D)$ is
65: equivalent to the stability of the map $f: \ucmodsp
66: \longrightarrow X$ (\cf. lemma \ref{lem:stableinj} and the remark
67: following the lemma). Exactness on the right of the long
68: exact sequence follows from the fact of the fibers of $\ucmodsp
69: \longrightarrow \modsp$ being curves.
70: \end{proof}
71:
72: Now, let $\cM_\mygraph$ be a fixed point component in the moduli
73: stack of stable maps $\modsp$, and $\pi_\mygraph:\cC_\mygraph
74: \longrightarrow \cM_\mygraph$ its universal curve. By lemma
75: \ref{lem:restrobth}, we know that the restriction of the long
76: exact sequence (\ref{eq:gobcomlong}) to the fixed point component
77: $\cM_\mygraph$ becomes:
78: \mycomment{eq:loccomlong}
79: \begin{equation}\label{eq:loccomlong}
80: \begin{split}
81: 0&\longrightarrow
82: \underline\Hom_{\pi_\mygraph}(\Omega^1_{\cC_\mygraph/\cM_\mygraph}(D),
83: \O_{\cC_\mygraph}) \longrightarrow
84: \underline\Hom_{\pi_\mygraph}(f_\mygraph^*\Omega^1_X,
85: \O_{\cC_\mygraph}) \longrightarrow \cT^0|_{\cM_\mygraph} \longrightarrow\\
86: &\longrightarrow
87: \underline\Ext^1_{\pi_\mygraph}(\Omega^1_{\cC_\mygraph/\cM_\mygraph}(D_\mygraph),
88: \O_{\cC_\mygraph}) \longrightarrow
89: \underline\Ext^1_{\pi_\mygraph}(f_\mygraph^*\Omega^1_X,
90: \O_{\cC_\mygraph}) \longrightarrow \cT^1|_{\cM_\mygraph}\longrightarrow 0.\\
91: \end{split}
92: \end{equation}
93: In particular, if we restrict to a single stable map $(C;
94: \underline{x}; f)\in \cM_\mygraph$, we get:
95: \mycomment{eq:obcomlong}
96: \begin{equation}\label{eq:obcomlong}
97: \begin{split}
98: 0&\longrightarrow \Ext^0(\Omega_C(\underline{x}),\O_C)\longrightarrow
99: H^0(C, f^*TX)\longrightarrow {\mathcal T}^0|_{\{C\}}
100: \longrightarrow\\
101: &\longrightarrow \Ext^1(\Omega_C(\underline{x}),\O_C)\longrightarrow
102: H^1(C, f^*TX)\longrightarrow {\mathcal T}^1|_{\{C\}}
103: \longrightarrow 0.\\
104: \end{split}
105: \end{equation}
106: %
107: To determine the virtual normal bundle for the torus action on
108: our toric variety, we will have to fulfill the moving parts of
109: $\cT^i_\mygraph=\cT^i|_{\cM_\mygraph}$, hence the moving parts in
110: the sequence (\ref{eq:loccomlong}). We will call the $i^{th}$ term of
111: this long exact sequence by $B_i$, and the moving part (with
112: respect to the induced torus action) by $B_i^\move$. Remember,
113: that the virtual normal bundle $N_\mygraph^\virt$ is the moving
114: part of the induced complex $E_{\bullet,\mygraph}$, that is
115: \[ e^{T_N}(N_\mygraph^\virt)=e^{T_N}(E_{0, \mygraph}^\move -E_{1,\mygraph}^\move)
116: =e^{T_N}({\mathcal T}_\mygraph^{0,\move}-{\mathcal T}_\mygraph^{1,\move}),
117: \]
118: where the second equation holds because of the exact sequence (\ref{eq:homolobcom}).
119: Now, applying the long exact sequence (\ref{eq:loccomlong}) we obtain for the
120: equivariant Euler class of the
121: virtual normal class $N^{vir}$ the following formula:
122: \mycomment{eq:eczerl}
123: \begin{equation}\label{eq:eczerl}
124: e^{T_N}(N_\mygraph^{\text{virt}})=\frac{e^{T_N}(B_2^\move)e^{T_N}(B_4^\move)}{e^{T_N}(B_1^\move)e^{T_N}(B_5^\move)}
125: \in H^*_{T_N}(X, \Q).
126: \end{equation}
127: %
128: The notation is indeed correct, since the $B_i^\move$,
129: $i=1,2,4,5$, are vector bundles on $\M_\mygraph$! This does not
130: apply in general to the fixed parts of these sheaves, or even to
131: the sheaves ${\mathcal T}^i_\mygraph$. So actually, at least for the
132: moving parts, we look at a long exact sequence of the kind of
133: (\ref{eq:obcomlong}).
134:
135: In the following, we will calculate the contributions of the four bundles
136: to the equivariant Euler class of the virtual normal bundle.
137:
138:
139: %-----------------------------------------------------------------------
140: \subsection{Computation of the equivariant Euler class of $B_1^\move$}
141: %-----------------------------------------------------------------------
142:
143: This bundle almost never contributes to the virtual normal bundle --- indeed
144: we will show the following lemma:
145:
146: \begin{lemm}
147: The $T_N$--equivariant Euler class of $B_1^\move$ is given by
148: \mycomment{eq:ec1}
149: \begin{equation}\label{eq:ec1}
150: e^{T_N}(B_1^\move)= \prod_{F\in\F_1} \we_F.
151: \end{equation}
152: \end{lemm}
153:
154: \begin{proof}
155: The bundle $B_1=\Ext^0(\Omega_C(D), \O_C)=\Aut_\infty(C)$
156: parameterizes infinitesimal automorphisms of the pointed domain.
157: The induced $T_N$--action on $\Aut(C)$ is obviously trivial on all
158: automorphism $\varphi$ of $C$ that restrict to the identity
159: $\varphi_{|C_e}=\id_{C_e}$ on all irreducible components $C_e$
160: corresponding to edges $e\in\Edge(\mygraph)$ in the graph
161: $\mygraph$. Thus, the moving part of $\Aut_\infty(C)$ splits into
162: \[ \Aut_\infty^\move(C) = \bigoplus_{e\in\Edge(\mygraph)} \Aut_\infty^\move(C_e).\]
163: Note in particular, that the bundle $\Aut_\infty^\move(C)$ is
164: topologically trivial on $\M_\mygraph$ since it only depends on
165: the irreducible components that are not mapped to a point, \ie.
166: that are rigid in $\M_\mygraph$.
167:
168: The $\Aut_\infty^\move(C_e)$ obviously depend on the special points
169: of the irreducible component $C_e$, \ie. on the way it is ``glued''
170: at the nodes corresponding to the vertices $\gotv_1(e)$ and $\gotv_2(e)$
171: of the edge $e$. Since we only look at moduli stacks of stable maps with
172: at least three marked points, we can exclude the two special cases
173: (\cf. figures \ref{fig:vnbb1f1} and \ref{fig:vnbb1f2}) where both
174: vertices are in $\Vertex_{1,0}$, or where one is in $\Vertex_{1,0}$ and
175: the other in $\Vertex_{1,1}$. We are left with two cases:
176: \begin{description}
177: \item[Case 1] {\bf One edge in $\Vertex_{1,0}$, the other in $\Vertex_{2,0}
178: \cup \Vertex_{\ge 3}$} --- Without loss of generality we can assume that
179: $\gotv_1(e)\in \Vertex_{1,0}$. In this case, $C_e$ corresponds to a
180: non--contracted $\P^1$ attached to
181: another non--contracted (or, in the $\Vertex_{\ge 3}$--case, contracted)
182: component (see figure \ref{fig:vnbb1f3}). Therefore we have to
183: look at M\"obius transformations that fix one point, infinity say:
184: \[
185: [x_1:x_2] \mapsto [ax_1+bx_2:x_2].
186: \]
187: Let $F\in\F_1$ be the flag corresponding to $\gotv_1$. We have seen above
188: that the induced $T_N$--action
189: on $C_e$ is given by (using again multi--index notation):
190: \[ t\cdot[x_1:x_2]=[t^{\we_F} x_1:x_2],\]
191: since the co--ordinate $x_1$ corresponds to the chart of the flag $F$
192: (while $x_2$ corresponds to the chart
193: around infinity, \ie. at the vertex $\gotv_2$. To determine the
194: $T_N$--action on the group $\Aut_\infty(C_e)$ of infinitesimal automorphisms
195: of $C_e$, we have to compute:
196: \begin{align*}
197: t\cdot(a,b)\cdot t^{-1}\cdot [x_1:x_2]&=t\cdot(a,b)\cdot[t^{-\we_F}x_1:x_2]\\
198: &=t\cdot[at^{-\we_F}x_1+bx_2:x_2]\\
199: &=[ax_1+t^{\we_F}bx_2:x_2]\\
200: &=(a,t^{\we_F}b)\cdot[x_1:x_2],
201: \end{align*}
202: hence the $T_N$--action on an automorphism of $C_e$ is given by:
203: \[
204: t\cdot(a,b) = (a, t^{\we_F}b).
205: \]
206: The moving part of $\Aut_\infty(C_e)$ is thus spanned by the second co--ordinate,
207: and the weight of the action there is $\we_F$ --- in this case we have:
208: \[ e^{T_N}(\Aut_\infty(C_e)^\move) = \we_F.\]
209:
210: \item[Case 2] {\bf Neither vertex is in $\Vertex_{1,0}$} ---
211: In this case, any automorphism of $C$ restricts to an automorphism on
212: $C_e$ that fixes the two points corresponding to the special points of
213: the vertices $\gotv_1$ and $\gotv_2$. Any such automorphism on $C_e$
214: (w.~l.~o.~g.\ we take the two points to be zero and infinity) has to look like
215: \[ [x_1:x_2] \DMRE{} [a x_1:x_2],\]
216: where $a\neq 0$ is a non-negative integer. With the same analysis as above
217: of the $T_N$--action on such automorphism $a$,
218: we see that the $T_N$--action on $\Aut_\infty(C_e)$ is trivial, \ie.
219: \[ \Aut_\infty^\move(C_e)=(0).\]
220: \end{description}
221: \end{proof}
222:
223: \vspace{0.5cm} \mycomment{fig:vnbb1f1-3}
224: \begin{figure}[ht]
225: \begin{minipage}[t]{.22\linewidth}
226: \begin{center}
227: \epsfig{figure=vnbb1f1.eps,width=\linewidth}
228: \caption{A graph with two vertices in $\V_{1,0}$.}\label{fig:vnbb1f1}
229: \end{center}
230: \end{minipage}\hfill
231: \begin{minipage}[t]{.22\linewidth}
232: \begin{center}
233: \epsfig{figure=vnbb1f2.eps,width=\linewidth}
234: \caption{A graph with the left vertex in $\V_{1,0}$ and the right vertex in $\V_{1,1}$.}\label{fig:vnbb1f2}
235: \end{center}
236: \end{minipage}
237: \hfill
238: \begin{minipage}[t]{.44\linewidth}
239: \begin{center}
240: \epsfig{figure=vnbb1f3.eps,width=\linewidth}
241: \caption{Two examples for parts of graphs that can occur in $\M_{0,m}(X,A)$ for $m\ge3$.}\label{fig:vnbb1f3}
242: \end{center}
243: \end{minipage}
244: \end{figure}
245:
246:
247: %----------------------------------------------------------------------------------
248: \subsection{The equivariant Euler class of $B_4^\move$}
249: %----------------------------------------------------------------------------------
250:
251: Here we are looking at the bundle $B_4=\Ext^1(\Omega_C(D),\O_C)=\operatorname{Def}(C)$ of deformations
252: of the pointed domain, that is deformations that vary some of the special points (varying the isomorphism
253: class of the curve) or that smooth some
254: double points.
255: Again, deformations of contracted components have obviously weight zero,
256: since the $T_N$--action on these components is trivial, so deformations coming from varying special
257: points do not contribute to $B_4^\move$.
258:
259: The other deformations of $C$ come from smoothing nodes of $C$ which join a non--contracted
260: component and a contracted or non--contracted component. Such a smoothing corresponds to
261: choosing an element of the tangent bundle at the double point. So let ${\mathcal L}_F$ be the
262: universal cotangent line (\cf. section \ref{sec:lbdms}) at the double point corresponding
263: to an $F\in\F_{\ge3}\cup\F_{2,0}$, and
264: write $e_F=e({\mathcal L}_F)=c_1({\mathcal L}_F)$ for the usual Euler class of this line bundle.
265:
266: If we look at the smoothing of a double point between a contracted and a non--contracted
267: component, \ie. if $F\in\F_{\ge3}$, let $F=(\gotv,e)$. In this case, the bundle $\cL_F$ is
268: the pull back to $\cM_\mygraph$ of the corresponding cotangent line on the Deligne-Mumford space
269: $\dmsp[\deg(\gotv)]$ of $\gotv$; on the other components of $\cM_\mygraph$, the bundle $\cL_F$
270: is trivial. The $T_N$-action on $\cM_\mygraph$ is trivial, and we have seen above that the
271: $T_N$--action on the bundle $\cL_F$ has weight $\we_F$. Hence in this case:
272: \[ e^{T_N}({\mathcal L}_F^*) = \we_F - e_F.\]
273:
274: In the second case, when we look at a vertex $\gotv\in\Vertex_{2,0}$ joining two non--contracted
275: components, we analogously obtain for the equivariant Euler class of the tangent line at this
276: node:
277: \[ e^{T_N}({\mathcal L}_F^*) = \we_{F_1} + \we_{F_2} - e_{F_1} - e_{F_2},\]
278: where $F_1, F_2\in \F_{2,0}$ are the two flags at $\gotv$.
279: However, ${\mathcal L}_{F_1}$ and ${\mathcal L}_{F_2}$ are topologically
280: trivial on $\M_\mygraph$ (since non--contracted components are
281: rigid in $\M_\mygraph$), so we have proven:
282:
283: \begin{lemm} The $T_N$--equivariant Euler class of the moving part of the bundle
284: $B_4$ is equal to:
285: \mycomment{eq:ec2}
286: \begin{equation}\label{eq:ec2}
287: e^{T_N}(B_4^\move)= \prod_{F\in\F_{\ge3}} (\we_F-e_F)
288: \prod_{\gotv\in\V_{2,0}} (\we_{F_1(\gotv)}+\we_{F_2(\gotv)}),
289: \end{equation}
290: where $F_1(\gotv), F_2(\gotv)$ denote the two different flags at
291: the vertex $\gotv\in\Vertex_{2,0}$.\proofend
292: \end{lemm}
293:
294: %-----------------------------------------------------------------------
295: \subsection{The equivariant Euler class of the quotient $B_2^\move-B_5^\move$}
296: %-----------------------------------------------------------------------
297:
298: Like Graber and Pandharipande, we will use the following exact
299: sequence to calculate the contribution coming from $H^*(f^*TX)$:
300: \[
301: 0\longrightarrow\O_C\longrightarrow
302: \bigoplus_{\gotv\in\V}\O_{C_\gotv}\oplus\bigoplus_{e\in\E}\O_{C_e}
303: \longrightarrow \bigoplus_{F\in\F} \O_{x_F}\longrightarrow 0.
304: \]
305:
306: Note, that a--priori it only makes sense to sum over $\V_{\ge3}$ in the
307: middle term, and over $\F_{\ge3}$ and $\V_{2,0}$ (instead of $\F$) in
308: the right term. However, we add the same in both terms, so they
309: cancel each other. By passing to the pullback under $f$ and taking
310: cohomology, we obtain:
311: \begin{multline*}
312: 0\rightarrow H^0(f^*T\vt)\rightarrow\bigoplus_{\gotv\in\V} H^0(C_\gotv, f^*T\vt)\oplus
313: \bigoplus_{e\in\E}H^0(C_e, f^*T\vt)\rightarrow\\
314: \rightarrow \bigoplus_{F\in\F}T_{p_{i(F)}}\vt \rightarrow H^1(f^*T\vt)\rightarrow
315: \bigoplus_{e\in\E}H^1(C_e, f^*T\vt)\rightarrow 0.\\
316: \end{multline*}
317:
318: Note that since we only look at genus zero curves, $H^1(C_v, f^*T\vt)=0$. On the
319: other hand, $H^1(C_e, f^*T\vt)$ is not necessarily zero for a toric variety $\vt$ as it is in general not
320: convex. Since $f^*T\vt$ is trivial on $C_\gotv$ for $\gotv\in\Vertex$,
321: $H^0(C_\gotv, f^*T\vt)=f_{p_{\sigma(\gotv)}}^*T\vt$. Hence we obtain the following formula:
322: \mycomment{eq:ec3a}
323: \begin{equation}\label{eq:ec3a}
324: \begin{split}
325: H^0(f^*T\vt)-H^1(f^*T\vt)&=+\, \bigoplus_{\gotv\in\V}T_{p_{\sigma(\gotv)}}\vt \, +\,
326: \bigoplus_{e\in\E} H^0(C_e, f^*T\vt)\\
327: &\phantom{{=}}-\,\bigoplus_{F\in\F} T_{p_{\sigma(F)}}\vt
328: -\,\bigoplus_{e\in\E} H^1(C_e, f^*T\vt).\\
329: \end{split}
330: \end{equation}
331:
332: To compute the equivariant Euler class of $H^0(C_e, f^*T\vt)$, we
333: again observe, that the bundle is constant.
334: To determine the weights of the induced action, we look at the following
335: Euler sequence on $\vt$:
336: \[ 0 \rightarrow \O^{n-d} \rightarrow \O(Z_1)\oplus\ldots\oplus\O(Z_n)
337: \rightarrow T\vt \rightarrow 0. \]
338: Pulling back to $C_e$ and taking cohomology gives
339: \[ 0 \rightarrow \C^{n-d} \rightarrow H^0(C_e, \O(\cocl_1))\oplus\ldots\oplus
340: H^0(C_e, \O(\cocl_n))\rightarrow H^0(C_e, f^*T\vt)\rightarrow 0,\]
341: where the tuple $(\cocl_1, \ldots, \cocl_n)$ describes the homology
342: class of $f_*[C_e]$.
343:
344: As in section \ref{sec:tvaction}, let $\partial e=\{\gotv_1,\gotv_2\}$ be the two
345: nodes at the ends of the edge $e$, $\sigma_i=\sigma(\gotv_i)\in\Sigma^{(d)}$ be the
346: two $d$--cones in the fan $\Sigma$ corresponding to the two nodes $\gotv_1$ and $\gotv_2$,
347: and let
348: \begin{align*}
349: \sigma_1&=\langle v_{i_1}, \ldots, v_{i_{d-1}}, v_{l_1(e)}\rangle\\
350: \sigma_2&=\langle v_{i_1}, \ldots, v_{i_{d-1}}, v_{l_2(e)}\rangle.
351: \end{align*}
352:
353: Using Proposition \ref{thm:homcoords} it is easy to show that for a one-dimensional
354: cone
355: $v_i\notin \{v_{i_1}, \ldots, v_{i_{d-1}}, v_{l_1(e)}, v_{l_2(e)}\}$, the
356: $T_N$--action on $\O_{C_e}(\cocl_i)$ is trivial, we can thus disregard it
357: with respect to moving part of $H^0(C_e, f^*T\vt)$. For an $i_j\in
358: \{i_1, \ldots, i_{d-1}\}$, $\O_{C_e}(\lambda_{i_j})$ is the direction of
359: the pull back $f^*T\vt$ of the tangent space that corresponds to $V_{\tau_{i_j}}$,
360: where $\tau_{i_j}=\langle v_{i_1}, \ldots, \widehat{v_{i_j}}, \ldots, v_{i_{(d-1)}}, v_{l_1(e)}
361: \rangle$.
362:
363: To compute $e^{T_N}(H^{0,\move}(C_e, \O(\cocl_{i_j}))$, let us fix $i_j$ and write
364: $\tau=\tau_{i_j}$ and $\lambda_e^\gamma=\lambda_{i_j}$.
365: If $\gamma\diamond\sigma_1$ such that $\gamma\cap\sigma_1=\tau$, then
366: the weight at $\sigma_1\in\vt$ of the action on the fiber is just $\we^{\sigma_1}_{\gamma}$, while
367: the weight at $\sigma_1$ of the action on the base $C_e$ is $\we_{F_1}$, where $F_1$ is the
368: flag $(\sigma_1(\gotv), e)$.
369:
370: By analysis of the induced action on holomorphic sections of the bundle $\O_{C_e}(\lambda_e^\gamma)$,
371: we see that its weights are:
372: \[ \we^{\sigma_1}_{\gamma_k} -\frac{b}{d_e}\we^{\sigma_1}_{\sigma_2},
373: \quad b=1,\ldots, \lambda_e^{\gamma_k}\]
374: which are all non--trivial.
375: For $\O(\lambda_{l_1(e)})$ and $\O(\lambda_{l_2(e)})$, there is in each case
376: exactly one zero weight among these weights:
377: \begin{align*}
378: \O_{C_e}(\lambda_{l_1(e)}):&\quad
379: \we^{\sigma_1}_{\sigma_2}, \ldots, \frac{1}{d_e}\we^{\sigma_1}_{\sigma_2}, 0\\
380: \O_{C_e}(\lambda_{l_2(e)}):&\quad
381: 0, -\frac{1}{d_e}\we^{\sigma_1}_{\sigma_2}, \ldots, -\we^{\sigma_1}_{\sigma_2}.
382: \end{align*}
383: Hence we have just proven the following lemma:
384:
385: \begin{lemm}\mylabel{lem:ec3p1}
386: The $T_N$--equivariant Euler class of the moving part of the trivial bundle
387: $H^0(C_e, f^*T\vt)$ is:
388: \begin{multline*}
389: e^{T_N} \left(H^{0,\move}(C_e, f^*T\vt)\right)=\\
390: (-1)^{d_e}\frac{(d_e!)^2}{d_e^{2d_e}}\left(\we^{\sigma_1}_{\sigma_2}\right)^{2d_e}
391: \prod_{\sigma_2\neq\gamma\face\sigma_1}
392: \prod_{b=0}^{\lambda_e^{\gamma}}
393: \left(\we^{\sigma_1}_\gamma - \frac{b}{d_e}\we^{\sigma_1}_{\sigma_2}\right).
394: \end{multline*}\proofend
395: \end{lemm}
396:
397: \begin{coro}\mylabel{coro:ec3p3}
398: The $T_N$--equivariant Euler class of the moving part of $H^1(C_e, f^*T\vt)$ is:
399: \[
400: e^{T_N}(H^{1,\move}(C_e, f^*T\vt))=\prod_{\sigma_2\neq\gamma\face\sigma_1}
401: \prod_{b=\lambda_e^\gamma}^{-2}
402: \left(\we^{\sigma_1}_\gamma - \frac{b+1}{d_e}\we^{\sigma_1}_{\sigma_2}\right).
403: \]
404: \end{coro}
405:
406: \begin{proof}
407: Just apply Serre duality:
408: \[ H^1(C_e, \O(\lambda_e^\gamma))=\Hom(\O(\lambda_e^\gamma)=H^0(C_e,\O(\lambda_e^\gamma)\otimes
409: \omega_{C_e}),\]
410: where $\omega_{C_e}$ is the dualizing bundle of $C_e\cong\P^1$.
411: \end{proof}
412:
413: So it only remains to compute the weights of the (trivial)
414: bundles $T_{p_{\sigma(\gotv)}}\vt$ and $T_{p_{\sigma(F)}}\vt$ which is
415: now straightforward:
416:
417: \begin{lemm}\mylabel{lem:ec3p2}
418: For a maximal cone $\sigma\in\Sigma^{(d)}$, the equivariant Euler
419: class of the trivial bundle $T_{p_\sigma}\vt$ is equal to
420: \[
421: e^{T_N}(T_{p_\sigma}\vt)=\prod_{\gamma\face\sigma} \we^\sigma_\gamma = \totwe{\sigma}.
422: \]
423: \end{lemm}
424:
425: \begin{coro}
426: The $T_N$--equivariant Euler class
427: of the moving part of the difference bundle $B_2-B_5$ is given by:
428: \mycomment{eq:ec3}
429: \begin{multline}\label{eq:ec3}
430: e^{T_N} (B_2^\move-B_5^\move) =
431: \prod_{\gotv\in\V} \left(\totwe{\sigma(\gotv)}\right)^{\val(\gotv)-1}\cdot\\
432: \prod_{\stackrel{\scriptstyle e\in\E}{\partial e=\{\gotv_1, \gotv_2\}}}
433: \left(
434: \frac{(-1)^d d^{2d}}{(d!)^2 \left(\we^{\sigma_1}_{\sigma_2}\right)^{2d}}
435: \prod_{\sigma_2\neq\gamma\face\sigma_1}
436: \frac{\displaystyle \prod_{i=\lambda^\gamma_e+1}^{-1}
437: \left( \we^{\sigma_1}_\gamma - \frac{i}{d}\cdot \we^{\sigma_1}_{\sigma_2}
438: \right)}{\displaystyle
439: \prod_{i=0}^{\lambda^\gamma_e}
440: \left(\we^{\sigma_1}_\gamma - \frac{i}{d}\cdot \we^{\sigma_1}_{\sigma_2}
441: \right)}
442: \right)_{\begin{array}{l} \scriptstyle \sigma_1=\sigma(\gotv_1)\\
443: \scriptstyle \sigma_2=\sigma(\gotv_2)\\
444: \scriptstyle d=d_e,
445: \end{array}}
446: \end{multline}
447: where $\val: \V \longrightarrow {\mathbb N}$ is the number
448: $\val(\gotv)$ of flags at a vertex $\gotv\in\V$.
449: \end{coro}
450:
451: \begin{proof}
452: Just assemble the long sequence (\ref{eq:ec3a}), Lemma \ref{lem:ec3p1},
453: Corollary \ref{coro:ec3p3}, and Lemma \ref{lem:ec3p2}. To obtain the
454: exponent of $\totwe{\sigma(\gotv)}$ just observe that there $\val(\gotv)$
455: flags at the vertex $\gotv$.
456: \end{proof}
457:
458: %--------------------------------------------------------------------
459: % MAIN THEOREM - THE FIXED POINT FORMULA
460: %--------------------------------------------------------------------
461: \subsection{The main theorem}
462: %--------------------------------------------------------------------
463:
464: In the previous subsections we have computed all contributions
465: (\ref{eq:ec1}), (\ref{eq:ec2}) and (\ref{eq:ec3}) entering the
466: formula (\ref{eq:eczerl}) for the equivariant Euler class of the
467: virtual normal bundle to $\cM_\mygraph$. Therefore, applying
468: Graber and Pandharipande's virtual Bott residue formula
469: (\ref{eq:bottres}) we obtain the following proposition
470: expressing the genus--zero Gromov--Witten invariants of a smooth
471: projective toric variety $\vt$ in terms of its $\modsp$--graphs
472: $\mygraph$ and the fan $\Sigma$:
473:
474: \mycomment{thm:maintheorem}
475: \begin{theo}\label{thm:maintheorem}
476: The genus--zero Gromov--Witten invariants for a toric variety $\vt$ are given
477: by
478: \mycomment{eq:mainthm}
479: \begin{equation}\label{eq:mainthm}
480: \Phi^{\vt}_{m,A}(Z^{l_1}, \ldots, Z^{l_m}) = \sum_\mygraph
481: \frac{1}{|{\mathbf A_\mygraph}|}\int_{\M_\mygraph}
482: \frac{\prod_{j=1}^m\prod_{k=1}^n
483: \left(\we^{\sigma(j)}_k\right)^{l_{j,k}}}{e^{T_N}(N_\mygraph^{\text{virt}})},
484: \end{equation}
485: where
486: \begin{itemize}
487: \item we use the convention $0^0=1$;
488: \item $Z^{l_i}=Z_1^{l_{i,1}}\cdot\ldots\cdot Z_n^{l_{i,n}}$;
489: \item $\sigma: \{1,\ldots, m\} \rightarrow \Sigma^{(d)}$, the image $\sigma(j)$ of $j$ corresponding
490: to the fixed point the marked point $j\in\{1, \ldots, m\}$ is mapped to:
491: \[ \exists \gotv\in\V(\mygraph)\, : \sigma(\gotv)=\sigma(j) \; \wedge \; j\in S(\gotv); \]
492: \item we define $\we^{\sigma(j)}_k:=\left\{
493: \begin{array}{ll}
494: 0&\quad \text{if $v_k \notin \Sigma^{(1)}_{\sigma(j)}$}\\
495: \we_\gamma^{\sigma(j)}&\quad \text{if $\gamma\face\sigma(j)$ and
496: $v_k\in\Sigma^{(1)}_{\sigma(j)}\backslash\Sigma^{(1)}_{\gamma};$}
497: \end{array}\right. $
498: \item the inverse of the Euler class of the virtual normal bundle is given by
499: \begin{multline*}
500: \frac{1}{e^{T_N}(N_\mygraph^{\text{virt}})} = \prod_{\gotv\in\V}
501: \left(\totwe{\sigma(\gotv)}\right)^{\val(\gotv)-1}\cdot
502: \prod_{F\in\F_{\ge3}}
503: \frac{1}{\we_F-e_F}\\\cdot\prod_{F\in\F_1}\we_F\cdot
504: \prod_{\gotv\in\V_{2,0}}\frac{1}{\we_{F_1(\gotv)}+\we_{F_2(\gotv)}}\cdot\\
505: \prod_{\stackrel{\scriptstyle e\in\E}{\partial e=\{\gotv_1,
506: \gotv_2\}}} \left(
507: \frac{(-1)^d d^{2d}}{(d!)^2 \left(\we^{\sigma_1}_{\sigma_2}\right)^{2d}}
508: \prod_{\sigma_2\neq\gamma\face\sigma_1}
509: \frac{\displaystyle \prod_{i=\lambda^\gamma_e+1}^{-1}
510: \left( \we^{\sigma_1}_\gamma - \frac{i}{d}\cdot \we^{\sigma_1}_{\sigma_2}
511: \right)}{\displaystyle
512: \prod_{i=0}^{\lambda^\gamma_e}
513: \left(\we^{\sigma_1}_\gamma - \frac{i}{d}\cdot \we^{\sigma_1}_{\sigma_2}
514: \right)}
515: \right)_{\begin{array}{l} \scriptstyle \sigma_1=\sigma(\gotv_1)\\
516: \scriptstyle \sigma_2=\sigma(\gotv_2)\\
517: \scriptstyle d=d_e.
518: \end{array}}
519: \end{multline*}
520: \end{itemize}
521: \end{theo}
522:
523: