1: \documentclass[12pt]{article}
2:
3: \title{The local Gromov-Witten theory of curves}
4: \author{Jim Bryan and Rahul Pandharipande}
5: %\date{July 2006}
6:
7:
8:
9:
10: \usepackage{diagrams}
11:
12: \usepackage{texdraw}
13: \input{txdtools}
14: \input{bukser}
15: \input{tableaux}
16:
17:
18: %\usepackage{eepic,epic}
19:
20: \usepackage{verbatim}
21: \usepackage{amsmath,amsthm,amsfonts}
22: %\usepackage{amstex}
23:
24: \newcommand{\cnums} {{\mathbb C}} % complex numbers
25: \newcommand{\nnums} {{\mathbb N}} % natural numbers
26: \newcommand{\rnums} {{\mathbb R}} % real numbers
27: \newcommand{\znums} {{\mathbb Z}} % integers
28: \newcommand{\qnums} {{\mathbb Q}} % rationals
29:
30: \newcommand{\Hom}{\operatorname{Hom}}
31: \newcommand{\Ker}{\operatorname{Ker}}
32: \newcommand{\End}{\operatorname{End}}
33: %\newcommand{\deg}{\operatorname{deg}}
34: \newcommand{\Tr}{\operatorname{tr}}
35: \newcommand{\tr}{\operatorname{tr}}
36: \newcommand{\Coker}{\operatorname{Coker}}
37: \newcommand{\im}{\operatorname{Im}}
38:
39:
40:
41: \renewcommand{\O}{\mathcal{O}}
42: \renewcommand{\P}{\mathbb{P}}
43: \newcommand{\M}{\overline{M}^{\bullet}_{h}}
44: \newcommand{\E}{\mathbb{E}}
45: \newcommand{\smargin}[1]{\marginpar{\tiny{#1}}}
46: \newcommand{\sss}[1]{\scriptscriptstyle{#1}}
47:
48:
49: \newcommand{\combinatfactor}{\mathfrak{z}}
50: \newcommand{\Wmeas}{\mathbf{W}}
51: \newcommand{\vertline}{\operatorname{|}}
52: \newcommand{\Zhat}{\widehat{Z}}
53:
54: \newcommand{\vac}{v_\emptyset}
55: \DeclareMathOperator{\Aut}{Aut}
56: \DeclareMathOperator{\Hilb}{Hilb}
57: \DeclareMathOperator{\Sym}{Sym}
58: \newcommand{\zz}{{\combinatfactor}}
59: \newcommand{\cF}{\mathcal{F}}
60: \newcommand{\Q}{\mathbb{Q}}
61: \newcommand{\Z}{\mathbb{Z}}
62: \newcommand{\lv}{\left |}
63: \newcommand{\rv}{\right |}
64: \newcommand{\lang}{\left\langle}
65: \newcommand{\rang}{\right\rangle}
66: \newcommand{\MM}{\mathsf{M}}
67:
68: \newcommand{\TT}{T^\pm}
69:
70: \newtheorem{thm}{Theorem}[section]
71: \newtheorem{theorem}[thm]{Theorem}
72: \newtheorem{cor}[thm]{Corollary}
73: \newtheorem{lem}[thm]{Lemma}
74: \newtheorem{lemma}[thm]{Lemma}
75: \newtheorem{prop}[thm]{Proposition}
76: \newtheorem{proposition}[thm]{Proposition}
77: \newtheorem{ax}{Axiom}
78: \newtheorem{defn}[thm]{Definition}
79: \newtheorem{definition}[thm]{Definition}
80: \newtheorem{warning}[thm]{Warning}
81: \newtheorem{questions}{Question}
82: \newtheorem{conjecture}[thm]{Conjecture}
83: \newtheorem{assumption}{Assumption}
84: \newtheorem{remark}{Remark}
85:
86:
87:
88: \begin{document}
89:
90: \maketitle
91:
92: \begin{abstract}
93: The local Gromov-Witten theory of curves is solved by localization and
94: degeneration methods. Localization is used for the exact evaluation of
95: basic integrals in the local Gromov-Witten theory of $\P^1$. A TQFT
96: formalism is defined via degeneration to capture higher genus
97: curves. Together, the results provide a compete and effective
98: solution.
99:
100: The local Gromov-Witten theory of curves is equivalent to the local
101: Donaldson-Thomas theory of curves, the quantum cohomology of the
102: Hilbert scheme points of $\cnums^2$, and the orbifold quantum
103: cohomology the symmetric product of $\cnums^2$. The results of the
104: paper provide the local Gromov-Witten calculations required for the
105: proofs of these equivalences.
106: \end{abstract}
107:
108:
109:
110: \tableofcontents
111:
112:
113:
114: \section{Introduction}
115:
116:
117:
118:
119:
120:
121:
122:
123:
124: \subsection{Local Gromov-Witten theory}
125:
126: The Gromov-Witten theory of threefolds, particularly Calabi-Yau threefolds,
127: is a very rich subject. The study of \emph{local
128: theories}, Gromov-Witten theories of non-compact targets, has
129: revealed much of the structure. Let $X$ be a complete,
130: nonsingular, irreducible
131: curve of genus $g$ over $\cnums$, and let $$N\to X$$ be a rank 2
132: vector bundle with $\det N\cong K_{X}$. Then $N$ is a non-compact
133: Calabi-Yau\footnote{We call any quasi-projective
134: threefold with trivial canonical bundle Calabi-Yau.} threefold, and the
135: Gromov-Witten theory, defined and studied in
136: \cite{Br-Pa-rigidity, Br-Pa,Br-Pa-TQFT,
137: Fa-Pa,Pandharipande-degenerate-contributions}, is
138: called the \emph{local Calabi-Yau theory of $X$}. We study here the local
139: theory of curves without imposing the Calabi-Yau condition $\det N\cong
140: K_{X}$ on the bundle $N$.
141:
142: The study of non Calabi-Yau local theories has several advantages. The
143: calculations of \cite{Fa-Pa,Pandharipande-degenerate-contributions,
144: Pandharipande-ICM} predict a uniform structure for all threefold theories
145: closely related to the Calabi-Yau case. The introduction of non Calabi-Yau
146: bundles $N$ yields a more flexible mathematical framework in which new
147: methods arise. We present a complete solution of the local Gromov-Witten
148: theory of curves.
149: The result requires a nonsingularity statement
150: proven in the Appendix with C. Faber and A. Okounkov.
151:
152:
153: The space of curves in a
154: Calabi-Yau threefold $Y$ is always of virtual
155: dimension 0. After suitable (and certainly
156: non-algebraic) deformation of the geometry
157: of $Y$,
158: we may expect to
159: find only isolated curves and their multiple covers ---
160: though no complete statement has yet been proven.
161:
162: The Gromov-Witten theory of $Y$ may then be viewed as
163: an enumeration of the isolated curves {\em together} with
164: a Gromov-Witten theory of local type
165: for the multiple covers. When defined, the latter theory
166: should be closely related to the local Gromov-Witten theory
167: of curves studied here
168: \cite{Br-Pa-rigidity}.
169:
170:
171: \subsection{Equivalences}
172: The local Gromov-Witten theory of curves is of substantial interest
173: beyond the original motivations. The local theory may
174: be viewed as an exactly solved quantum deformation of the
175: Hurwitz question of enumerating
176: ramified coverings of curves. In fact, the
177: solution has been discovered to arise in many different
178: geometry contexts.
179:
180:
181: Our study of the local Gromov-Witten theory of curves
182: is a starting point for
183: several lines of inquiry:
184: \begin{enumerate}
185: \item[(i)]
186: The Gromov-Witten/Donaldson-Thomas correspondence of \cite{MNOP1,MNOP2} may
187: be naturally studied in the context of local theories.
188: Our results together with \cite{dtlc} prove the correspondence
189: for local theories of curves, see Section
190: \ref{subsec: GW/DT for local curves}.
191:
192:
193: \item[(ii)]
194: The local theory of the {\em trivial} rank 2 bundle over $\P^1$
195: is equivalent to the quantum cohomologies of the Hilbert
196: scheme $\Hilb^n(\cnums^2)$ and
197: the orbifold
198: $(\cnums^2)^{n}/S_n$.
199: Our results here together with
200: \cite{Bryan-Graber, Ok-Pan-Hilb} prove the equivalences,
201: see Section \ref{sec: further directions}.
202: \end{enumerate}
203: We expect further connections will likely be found in the future.
204:
205:
206:
207:
208:
209:
210:
211:
212: \subsection{Results}
213: Let $N$ be a rank 2 bundle on a curve $X$ of genus $g$.
214: We assume $N$ is decomposable as a direct sum of
215: line bundles,
216: \begin{equation}\label{ffg}
217: N=L_1\oplus L_2.
218: \end{equation}
219: The splitting determines a scaling action of a 2-dimensional torus
220: $$T=\cnums^* \times \cnums^*$$ on $N$.
221: The {\em level} of the splitting is the pair of integers
222: $(k_1,k_2)$ where,
223: $$k_i= {\text {deg}}(L_i).$$
224: Of course, the scaling action and the level
225: depend upon the splitting \eqref{ffg}.
226:
227:
228:
229:
230: The Gromov-Witten residue invariants of $N$, defined in
231: Section~\ref{subsec: GW residue invariants of N}, take values in the
232: localized equivariant cohomology ring of $T$ generated by $t_1$ and
233: $t_2$. The basic objects of study in our paper are the
234: \emph{partition functions}
235: \[
236: {\mathsf{GW}}_d(g\vertline k_{1},k_{2}) \in \qnums (t_{1},t_{2})((u)),
237: \]
238: the generating functions for the degree $d$ residue invariants of
239: $N$. Here, $u$ parameterizes the domain genus.
240:
241:
242:
243: The residue invariants specialize to the local invariants of $X$ in a
244: Calabi-Yau threefold defined in \cite{Br-Pa,Br-Pa-TQFT} if the level satisfies
245: \[
246: k_{1}+k_{2}=2g-2
247: \]
248: and the variables are equated,
249: \[
250: t_1=t_2.
251: \]
252: Equating the variables is equivalent to considering the residue theory
253: of $N$ with respect to the diagonal action of a 1-dimensional
254: torus.
255:
256:
257:
258:
259:
260:
261: For the Gromov-Witten residue invariants of $N$, we develop a gluing theory
262: in Section \ref{ctt}
263: following \cite{Br-Pa-TQFT}. The interpretation of the local
264: theory as TQFT is discussed in Section \ref{tqft}.
265: In Sections \ref{ss} - \ref{ctt} and the Appendix, the gluing relations,
266: together with a few basic integrals, are proven to
267: determine the full local theory of curves.
268: The level freedom of the theory
269: plays an essential role. We provide explicit formulas in Sections
270: \ref{sec: s1=-s2 limit} and \ref{sec: explicit formulas}.
271:
272: A parallel equivariant Donaldson-Thomas residue theory can be defined for
273: the threefold $N$. We conjecture a
274: Gromov-Witten/Donaldson-Thomas correspondence
275: for equivariant residues in the framework of
276: \cite{MNOP1,MNOP2}, see Section~\ref{sec: GW/DT correspondence for
277: residues}. An important consequence of our theory is Theorem
278: \ref{thm: Z is a rational function of q}. After suitable normalization,
279: ${\mathsf {GW}}_d(g\vertline k_{1},k_{2})$ is a
280: \emph{rational} function of the variables $t_1$, $t_2$, and
281: \[
282: q=-e^{iu}.
283: \]
284: The result verifies a
285: prediction of the GW/DT correspondence, see Conjecture~2R of
286: Section~\ref{sec: GW/DT correspondence for residues}.
287:
288:
289:
290: The residue invariants of $N$ are of special interest when the variable
291: reduction,
292: \[
293: t_1+t_2=0,
294: \]
295: is taken. The reduction is equivalent to considering the residue theory of
296: $N$ with respect to the {\em anti-diagonal} action of a 1-dimensional
297: torus. In Theorem \ref{thm: s1=-s2 general formula},
298: we obtain a general closed formula for the partition
299: function in the anti-diagonal case.
300:
301: If we additionally specialize to the Calabi-Yau case, our formula is
302: particularly attractive. The residue partition function here is simply a
303: $Q$-deformation of the classical formula for unramified covers (see
304: Corollary~\ref{cor: s1=-s2 limit, CY case}):
305: \[
306: {\mathsf{GW}}_d
307: (g\vertline k,2g-2-k)= (-1)^{d (g-1-k)}\sum _{\rho }\left(\frac{d!}{\dim
308: _{Q}\rho } \right)^{2g-2} Q^{-c_{\rho } (g-1-k)}
309: \]
310: where $Q=e^{iu}$ and the sum is over partitions. With the
311: anti-diagonal action, $N$ is \emph{equivariantly} Calabi-Yau.
312:
313:
314: Using the above formula, Aganagic, Ooguri, Saulina, and Vafa have
315: recently found that the local Gromov-Witten theory of curves is
316: closely related to $q$-deformed 2D Yang-Mills theory and bound states
317: of BPS black holes \cite{Aganagic-Ooguri-Saulina-Vafa,Vafa-04-2dYang-Mills}.
318:
319:
320: The anti-diagonal action is exactly {\em opposite} to the original
321: motivations of the project. It would be very interesting to find
322: connections between the anti-diagonal case and the original questions of
323: the Gromov-Witten theory of curves in Calabi-Yau threefolds.
324:
325: \subsection{Acknowledgments} The authors thank G. Farkas, T. Graber,
326: A. Greenspoon, S. Katz, J. Kock, C. Teleman, M. Thaddeus, C. Vafa, and
327: R. Vakil for valuable discussions. We thank J. Kock for the use of his
328: cobordism \LaTeX \@ macros. A first draft of the Appendix was
329: completed during a visit by C. Faber to Princeton in the summer of
330: 2004.
331:
332: J.~B. was partially supported by the NSERC, the
333: Clay Institute, and the Aspen Institute. R.~P. was partially supported by
334: the Packard foundation and the NSF.
335:
336:
337:
338:
339: \section{The residue theory}
340:
341: \subsection{Gromov-Witten residue invariants}
342:
343: Let $Y$ be a nonsingular, {\em quasi-projective}, algebraic threefold.
344: Let $\M (Y,\beta )$ denote the moduli space
345: of stable maps
346: \[
347: f:C\to Y
348: \]
349: of genus $h$ and degree $\beta\in H_{2} (Y,\znums )$.
350: The superscript $\bullet$ indicates the
351: possibility of disconnected domains $C$.
352: We require $f$ to be nonconstant on each connected component of $C$.
353: The genus, $h(C)$, is defined by
354: \[
355: h (C)= 1-\chi (\O _{C})
356: \]
357: and may be negative.
358: %We follow here the notation of \cite{Br-Pa-TQFT}.
359:
360:
361:
362:
363: Let $Y$ be equipped with an action by an algebraic torus $T$.
364: We will define Gromov-Witten residue invariants under the following
365: assumption.
366: \begin{assumption}\label{assuption: T fixed locus of GW space is compact}
367: The $T$-fixed point set $\M (Y,\beta )^{T}$ is compact.
368: \end{assumption}
369:
370: We motivate the definition of the residue invariants of $Y$
371: as follows. We would like to define the reduced Gromov-Witten partition
372: function ${\mathsf Z}' (Y)_\beta$ as a generating function of the
373: integrals of the identity class over the moduli spaces of maps,
374: \begin{equation}\label{eqn: moral defn of Zgw}
375: {\mathsf Z}' (Y)_\beta\text{ ``}=\text{'' }\sum _{h\in \znums
376: }u^{2h-2}\int _{[\M (Y,\beta ) ]^{vir}}1 .
377: \end{equation}
378: However, the integral on the right might not be well-defined if $Y$ is
379: not compact.
380:
381: If $Y$ has trivial canonical bundle and $\M (Y,\beta )$ is compact,
382: then the integral \eqref{eqn: moral defn of Zgw} is well-defined. The
383: resulting series ${\mathsf Z}' (Y)_\beta$ is then the usual
384: reduced partition function for the degree $\beta$ disconnected
385: Gromov-Witten invariants\footnote{We follow the notation of
386: \cite{MNOP1,MNOP2} for the reduced partition function. The prime
387: indicates the removal of the degree 0 contributions. In
388: \cite{MNOP1,MNOP2}, the moduli space $\M (Y,\beta )$ is denoted by
389: $\overline{M}'_{h} (Y,\beta )$. However, to maintain notational
390: consistency with \cite{Br-Pa-TQFT}, we will {\em not} adopt the latter
391: convention.} of $Y$. We can use the virtual localization formula to
392: express ${\mathsf Z}'(Y)_\beta$ as a residue integral over the
393: $T$-fixed point locus.
394:
395: More generally, under Assumption~\ref{assuption: T fixed locus of GW space
396: is compact}, the series ${\mathsf Z}'(Y)_\beta$ can be {\em defined}
397: via localization.
398:
399: \begin{definition}\label{defn: Zgw'}
400: The reduced partition function for the degree $\beta $ residue
401: Gromov-Witten invariants of $Y$ is defined by:
402: \begin{equation}\label{eqn: definition of Zgw}
403: {\mathsf Z}'(Y)_\beta =\sum _{h\in \znums }u^{2h-2}\int
404: _{[\M (Y,\beta )^{T}]^{vir}}\frac{1}{e (\operatorname{Norm}^{vir})}.
405: \end{equation}
406: \end{definition}
407:
408: The $T$-fixed
409: part of the perfect obstruction theory for $\M (Y,\beta )$ induces a
410: perfect obstruction theory for $\M (Y,\beta )^{T}$ and hence a virtual
411: class \cite{Gr-Pa}.
412: The
413: equivariant virtual normal bundle
414: of the embedding, $$\M (Y,\beta )^{T}\subset \M (Y,\beta ), $$ is
415: $\operatorname{Norm}^{vir}$ with
416: equivariant Euler class
417: $e(\operatorname{Norm}^{vir})$. The integral in \eqref{eqn: definition of Zgw}
418: denotes equivariant push-forward to a point.
419:
420:
421:
422:
423:
424: Let $r$ be the rank of $T$, and
425: let $t_{1}, \ldots, t_r$ be generators for the equivariant cohomology
426: of $T$,
427: \[
428: H^{*}_{T} (\text{pt})\cong \qnums [t_{1},\ldots,t _{r}].
429: \]
430: By Definition~\ref{defn: Zgw'}, $ {\mathsf Z}'(Y)_\beta $ is a Laurent
431: series in $u$ with coefficients given by rational functions of the
432: variables $t_{1},\ldots,t_{r}$ of homogeneous degree equal to minus the
433: virtual dimension of $\M(Y,\beta)$.
434:
435:
436:
437:
438:
439:
440:
441: \subsection{Gromov-Witten residue invariants of $N$} \label{subsec: GW residue invariants of N}
442:
443: Let $X$ be a nonsingular, irreducible, projective curve of genus $g$. Let
444: $$N=L_1\oplus L_2$$ be a rank 2 bundle on $X$. The residue invariants
445: of the threefold $N$ with respect to the 2-dimensional scaling torus
446: action can be written in terms of integrals over the moduli space of
447: maps to $X$.
448:
449:
450:
451: The residue theory may be considered for the 1-dimensional scaling
452: torus action on an \emph{indecomposible} rank 2 bundle $N$. Since every
453: rank 2 bundle is equivariantly deformation equivalent to a
454: decomposable bundle, the residue invariants of indecomposable bundles
455: are specializations of the split case.
456:
457:
458:
459: A stable map to $N$ which is $T$-invariant must factor through the zero
460: section. Hence,
461: \[
462: \M (N,d[X])^{T}\cong \M (X,d).
463: \]
464: Moreover, the $T$-fixed part of the perfect obstruction theory of $\M
465: (N,d[X])$, restricted to $\M (N,d[X])^{T}$, is exactly the usual perfect
466: obstruction theory for $\M (X,d)$. Hence,
467: \[
468: [\M (N,d[X])^{T}]^{vir}\cong [\M (X,d)]^{vir}.
469: \]
470:
471: The virtual normal bundle of $\M (N,d[X])^{T}\subset \M (N,d[X])$,
472: considered as an element of $K$-theory on $\M (X,d)$, is given by
473: \[
474: \operatorname{Norm}^{vir}=R^{\bullet }\pi _{*}f^{*} (L_{1}\oplus L_{2})
475: \]
476: where
477: \[
478: \begin{diagram}[height=0.8cm]
479: U&\rTo^{f}&X\\
480: \dTo^{\pi }&&\\
481: \M (X,d)&&
482: \end{diagram}
483: \]
484: is the universal diagram for $\M (X,d)$.
485:
486: %Using the equation
487: %\[
488: %-K_{N}\cdot d[X] = -d (2g-2-k_{1}-k_{2}),
489: %\]
490: %and the definitions,
491: The reduced Gromov-Witten partition function of the residue invariants
492: may be written in the following form via equivariant integration:
493: \[
494: {\mathsf Z}'_d(N)=\sum _{h\in \znums }u^{2h-2}\int
495: _{[\M (X,d)]^{vir}}e (-R^{\bullet }\pi _{*}f^{*} (L_{1}\oplus L_{2})).
496: \]
497: We will be primarily interested in a partition function with a shifted
498: exponent,
499: \[
500: {\mathsf{GW}}_d (g\vertline k_{1},k_{2})
501: =u^{d (2-2g+k_{1}+k_{2})}\ {\mathsf Z}'_d(N).
502: \]
503: The shift can be interpreted geometrically as
504: $$\int_{d[X]} c_1(T_N) = d(2-2g+k_1+k_2),$$
505: where $T_N$ is the tangent bundle of the threefold $N$.
506:
507:
508:
509: The explicit dependence on the equivariant parameters $t_{1}$
510: and $t_{2} $ may be written as follows. Let $b_{1}$ and $b_{2}$
511: be non-negative integers
512: satisfying
513: $$
514: b_{1}+b_{2}=2h-2+d (2-2g)$$
515: where $2h-2+d(2-2g)$ is the virtual dimension of $\M(X,d)$.
516: Let
517: \[
518: {\mathsf{GW}}^{b_{1},b_{2}}_d(g\vertline k_{1},k_{2})
519: =\int _{[\M (X,d)]^{vir}}c_{b_{1}} (-R^{\bullet }\pi _{*}f^{*}L_{1})c_{b_{2}}
520: (-R^{\bullet }\pi _{*}f^{*}L_{2}),
521: \]
522: where $\int$ here denotes {\em ordinary} integration.
523: The equivariant Euler class $e (-R^{\bullet }\pi _{*}f^{*} (L_{1}\oplus
524: L_{2}))$ is easily expressed in terms of the equivariant parameters and the
525: {\em ordinary} Chern classes of $-R^{\bullet }\pi _{*}f^{*} (L_{1})$ and
526: $-R^{\bullet }\pi _{*}f^{*} (L_{2})$,
527: \begin{multline*}
528: {\mathsf{GW}}_d (g\vertline k_{1},k_{2})= \\
529: u^{d (k_{1}+k_{2})}t_{1}^{d (g-1-k_{1})}t_{2}^{d (g-1-k_{2})} \sum
530: _{b_{1},b_{2}=0}^{\infty } u^{b_{1}+b_{2}}t_{1}^{\frac{1}{2}(b_{2}-b_{1})}
531: t_{2}^{\frac{1}{2}(b_{1}-b_{2})}{\mathsf{GW}}^{b_{1},b_{2}}_d
532: (g\vertline k_{1},k_{2}).
533: \end{multline*}
534: Since $b_{1}+b_{2}$ is even, the exponents of $t_{1}$ and $t_{2}$ are
535: integers. We see that ${\mathsf{GW}}_d (g\vertline k_{1},k_{2})$
536: is a Laurent
537: series in $u$ with coefficients given by rational functions of $t_{1}$
538: and $t_{2}$ of homogeneous degree $d (2g-2-k_{1}-k_{2})$.
539:
540:
541:
542:
543:
544:
545: \section{Gluing formulas}
546: \label{gfs}
547:
548: \subsection{Notation and conventions for partitions}\label{subsub:
549: conventions for partitions}
550:
551:
552:
553: By definition, a partition $\lambda $ is a finite
554: sequence of positive integers
555: \[
556: \lambda = (\lambda _{1}\geq \lambda _{2}\geq \lambda_{3} \geq \dots )
557: \]
558: where
559: $$|\lambda |=\sum _{i}\lambda _{i}=d.$$ We use the notation $\lambda
560: \vdash d$ to indicate that $\lambda $ is a partition of $d$.
561:
562: The number of parts of $\lambda $ is called the \emph{length} of
563: $\lambda $ and is denoted $l (\lambda )$. Let $m_{i} (\lambda )$ be
564: the number of times that $i$ occurs in the partition $\lambda $. We
565: may write a partition in the format:
566: \[
567: \lambda = (1^{m_{1}}2^{m_{2}}3^{m_{3}}\cdots )\ .
568: \]
569: The combinatorial factor,
570: \[
571: \combinatfactor (\lambda ) =\prod _{i=1}^{\infty }m_{i} (\lambda )!i^{m_{i}
572: (\lambda )},
573: \]
574: arises frequently.
575:
576:
577:
578:
579: A partition $\lambda $ is uniquely determined by the associated
580: Ferrers diagram, which is the collection of $d$ boxes located at
581: $(i,j)$ where $1\leq j\leq \lambda _{i}$. For example
582: \[
583: (3,2,2,1,1)= (1^{2}2^{2}3)=\tableau{3 2 2 1 1}.
584: \]
585: The \emph{conjugate partition} $\lambda '$ is obtained by reflecting the
586: Ferrers diagram of $\lambda $ about the $i=j$ line.
587:
588:
589:
590: In Section~\ref{sec: s1=-s2 limit}, we will require the following standard
591: quantities.
592: Given a box in the Ferrers diagram, $\Box \in \lambda $, define the
593: \emph{content} $c (\Box )$ to be $i-j$, and the \emph{hooklength} $h (\Box
594: )$ to be $ \lambda_{i} + \lambda_{j} '-i-j+1$. The total content
595: \[
596: c_{\lambda }=\sum _{\Box \in \lambda }c (\Box )
597: \]
598: and the total hooklength
599: \[
600: \sum _{\Box \in \lambda }h (\Box )
601: \]
602: satisfy the following identities (page 11 of \cite{Macdonald}):
603: \begin{equation}\label{eqn: identities for total content and hook length}
604: \sum _{\Box \in \lambda }h (\Box ) = n (\lambda )+n (\lambda ')+d,\quad \quad
605: \quad c_{\lambda }=n (\lambda' )-n (\lambda ),
606: \end{equation}
607: where
608: \[
609: n (\lambda )=\sum _{i=1}^{l (\lambda )} (i-1)\lambda _{i}.
610: \]
611:
612: \subsection{Relative invariants}
613:
614:
615: To formulate our gluing laws for the residue theory of rank 2 bundles on $X$,
616: we require relative versions of the
617: residue invariants.
618:
619:
620:
621: Motivated by the symplectic theory of A.-M. Li and Y. Ruan \cite{Li-Ruan},
622: J. Li has developed an algebraic theory of relative stable maps to a pair
623: $(X,B)$. This theory compactifies the moduli space of maps to $X$ with
624: prescribed ramification over a non-singular divisor $B\subset X$,
625: \cite{Li-relative1,Li-relative2}. Li constructs a moduli space of
626: relative stable maps together with a virtual fundamental cycle and
627: proves a gluing formula.
628:
629: Consider a degeneration of $X$ to $X_{1}\cup_{B} X_{2}$, the union of
630: $X_{1}$ and $X_{2}$ along a smooth divisor $B$. The gluing formula
631: expresses the virtual fundamental cycle of the usual stable map moduli
632: space of $X$ in terms of virtual cycles for relative stable maps of
633: $(X_{1},B)$ and $(X_{2},B)$. The theory of relative stable maps has
634: also been pursued in \cite{EGH, Ionel-Parker-Annals2003,Ionel-Parker00}.
635:
636: In our case, the target is a non-singular curve $X$ of genus $g$,
637: and the divisor $B$ is a collection of points $x_{1},\dots ,x_{r}\in X$.
638:
639: \begin{definition}\label{defn: rel stable maps}
640: Let $(X,x_{1},\dots x_{r})$ be a fixed non-singular genus $g$ curve with
641: $r$ distinct marked points. Let $\lambda_1, \ldots, \lambda_r$ be
642: partitions of $d$.
643: Let
644: \[
645: \M (X, \lambda _{1},\dots, \lambda_r)
646: \]
647: be the moduli space of genus $h$ relative stable maps (in the sense of
648: Li)\footnote{For a formal definition of relative stable maps, we refer to
649: \cite{Li-relative1} Section~4.} with target $(X,x_{1},\dots ,x_{r})$
650: satisfying the following:
651: \begin{enumerate}
652: \item[(i)] The maps have degree $d$.
653: \item[(ii)] The maps are ramified over $x_{i}$ with ramification type $\lambda
654: _{i}$.
655: \item[(iii)] The domain curves are possibly disconnected, but the map is not
656: degree 0 on any connected component.
657: \item[(iv)] The domain curves are \emph{not} marked.
658: \end{enumerate}
659: The partition $\lambda _{i}\vdash d$ determines a ramification type over
660: $x_{i}$ by requiring the monodromy of the cover (considered as a conjugacy
661: class of $S_{d}$) has cycle type $\lambda _{i}$.
662: \end{definition}
663:
664: Our moduli spaces of relative stable maps differ from Li's in a few
665: minor ways. For a complete discussion, see \cite{Br-Pa-TQFT}.
666:
667:
668: We define the relative reduced partition function
669: via equivariant integration over spaces of relative stable maps:
670: \[
671: {\mathsf Z}'(N)_{\lambda^1 \dots \lambda^r} =
672: \sum _{h\in \znums }
673: u^{2h-2}
674: \int _{[\M (X,\lambda ^{1},\dots ,\lambda ^{r})]^{vir}}e
675: (-R^{\bullet }\pi _{*}f^{*} (L_{1}\oplus L_{2})).
676: \]
677: Again, we will be primarily interested in a shifted generating function,
678: $$
679: {\mathsf{GW}}(g\vertline k_{1},k_{2})_{\lambda ^{1} \dots \lambda ^{r}}=
680: u^{d (2-2g+k_{1}+k_{2}-r)+\sum _{i=1}^{r}l (\lambda ^{i})} \
681: {\mathsf Z}'(N)_{\lambda^1 \dots \lambda^r}.$$
682: Since the degree $d$ is equal to $|\lambda^i|$, the degree
683: subscript is redundant in the relative theory.
684:
685:
686: The
687: exponent of $u$ in the partition function
688: ${\mathsf{GW}}_d(g\vertline k_1,k_2)$ of the non-relative theory
689: is
690: $$2h-2+ \int_{d[X]}c_1(T_N).$$
691: In the relative theory, the $2h-2$ term in the exponent is
692: replaced with $2h-2+\sum l (\lambda ^{i})$, the negative Euler
693: characteristic of the \emph{punctured} domain. The class $c_1(T_N)$
694: is replaced with the dual of the log canonical class of $N$ with
695: respect to the relative divisors.
696: The outcome is the modified exponent of $u$ in the partition function
697: ${\mathsf{GW}}(g\vertline k_{1},k_{2})_{\lambda ^{1} \dots \lambda ^{r}}.$
698:
699:
700:
701: As before, we can make the dependence on $t_{1}$ and $t_{2}$ explicit. Let
702: $$b_{1}+b_{2}=2h-2+d (2-2g)- \delta,$$
703: where
704: \[
705: \delta=\sum _{i=1}^{r} (d-l (\lambda ^{i})).
706: \]
707: Here, $b_1+b_2$ equals the virtual dimension
708: of $\M (X,\lambda ^{1}\dots \lambda ^{r}).$
709: Let
710: \[
711: {\mathsf{GW}}^{b_{1},b_{2}}
712: (g\vertline k_{1},k_{2})_{\lambda ^{1} \dots \lambda ^{r}}=\int
713: _{[\M (X,\lambda ^{1}, \dots ,\lambda ^{r})]^{vir}}c_{b_{1}} (-R^{\bullet
714: }\pi _{*}f^{*}L_{1})c_{b_{2}} (-R^{\bullet }\pi _{*}f^{*}L_{2}).
715: \]
716: Then, we have
717: \begin{multline}\label{eqn: expression for Zdgk1k2_lambdas}
718: {\mathsf{GW}} (g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}}=
719: u^{d (k_{1}+k_{2})}t_{1}^{d (g-1-k_{1})}t_{2}^{d (g-1-k_{2})} \\ \cdot \sum
720: _{b_{1},b_{2}=0}^{\infty }
721: u^{b_{1}+b_{2}}t_{1}^{\frac{b_{2}-b_{1}+\delta}{2}}
722: t_{2}^{\frac{b_{1}-b_{2}+\delta}{2}}
723: {\mathsf{GW}}^{b_{1},b_{2}}
724: (g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}}.
725: \end{multline}
726: Since the parity of $b_{1}+b_{2}$ is the same as $\delta$, the
727: exponents of $t_{1}$ and $t_{2}$ are integers.
728:
729: The partition function
730: ${\mathsf{GW}}(g\vertline k_{1},k_{2})_{\lambda ^{1} \dots
731: \lambda ^{r}}$ is a Laurent series in $u$ with coefficients given by
732: rational functions in $t_{1}$ and $t_{2}$ of homogeneous degree
733: \[
734: d (2g-2-k_{1}-k_{2})+\delta.
735: \]
736:
737:
738: In \cite{Br-Pa-TQFT}, the combinatorial factor $\combinatfactor
739: (\lambda )$ is used to raise the indices for the relative invariants. For the
740: residue invariants, an additional factor
741: $(t_{1}t_{2})^{l (\lambda )}$ must be included. We define:
742: \begin{equation}\label{eqn: index raising formula}
743: {\mathsf{GW}}
744: (g\vertline k_{1},k_{2})_{\mu ^{1}\dots \mu ^{s}}^{\nu^{1}\dots \nu
745: ^{t}}=
746: {\mathsf{GW}} (g\vertline k_{1},k_{2})_{\mu
747: ^{1}\dots \mu ^{s},\nu ^{1}\dots \nu ^{t}}
748: \left(\prod _{i=1}^{t} \combinatfactor (\nu ^{i})
749: (t_{1}t_{2})^{l (\nu ^{i})} \right).
750: \end{equation}
751:
752: \subsection{Gluing formulas}
753: The gluing formulas are determined by the following result.
754:
755: \begin{theorem}\label{thm: gluing formula}
756: For splittings
757: $g=g'+g''$ and $k_{i}=k_{i}'+k_{i}''$,
758: \[
759: {\mathsf{GW}} (g\vertline k_{1},k_{2})_{\mu ^{1}\dots \mu ^{s}}^{\nu ^{1}\dots
760: \nu ^{t}} =\sum _{\lambda \vdash d} {\mathsf{GW}} (g'\vertline
761: k_{1}',k_{2}')_{\mu ^{1}\dots \mu ^{s}}^{\lambda }
762: \mathsf{GW} (g''\vertline
763: k_{1}'',k_{2}'')_{\lambda }^{\nu ^{1}\dots \nu ^{t}}
764: \]
765: and
766: \[
767: {\mathsf{GW}}
768: (g\vertline k_{1},k_{2})_{\mu ^{1}\dots \mu ^{s}}=\sum _{\lambda \vdash
769: d}{\mathsf{GW}} (g-1\vertline k_{1},k_{2})^{\lambda }_{\mu ^{1}\dots \mu ^{s}\lambda }.
770: \]
771: \end{theorem}
772:
773: \begin{proof}
774: The proof follows the derivation of the gluing formulas in \cite{Br-Pa-TQFT}.
775: The only difference
776: is the modified metric term
777: $$\combinatfactor(\lambda) (t_1t_2)^{l(\lambda)}.$$ The first factor,
778: $\combinatfactor(\lambda)$, is obtained from the degeneration formula
779: for the virtual class \cite{Li-relative2} as in
780: \cite{Okounkov-Pandharipande-completed-cycles}.
781:
782: The second factor, $(t_1t_2)^{l(\lambda)}$,
783: arises from normalization sequences associated to the fractured domains.
784: Let
785: $$f:C \rightarrow X$$
786: be an element of $\M (X,\mu ^{1},\dots ,\mu ^{s}, \nu^1,\dots,\nu^t)$.
787: Consider a reducible degeneration of the target,
788: $$X=X'\cup X'',$$
789: over
790: which the line bundles $L_1$ and $L_2$ extend with degree splittings
791: $$k_1=k_1'+k_1'',$$
792: $$k_2=k_2'+k_2''.$$
793: In a degeneration of type $\lambda$,
794: the domain curve degenerates,
795: $$C= C' \cup C'',$$
796: into components lying over $X'$ and $X''$ and satisfying
797: $$|C' \cap C''| = l(\lambda).$$
798: For each line bundle $L_i$, we have a normalization sequence,
799: \begin{equation}
800: \label{rezz} 0 \rightarrow f^*(L_i)|_{C} \rightarrow f^*(L_i)|_{C'}
801: \oplus f^*(L_i)|_{C''} \rightarrow f^*(L_i)|_{C'\cap C''} \rightarrow
802: 0.
803: \end{equation}
804: The last term yields a trivial bundle of rank
805: $l(\lambda)$
806: with scalar torus action over the moduli space
807: of maps of degenerations of type $\lambda$.
808: The factor $(t_1t_2)^{l(\lambda)}$ is obtained from the
809: higher direct images of the normalization sequences \eqref{rezz}.
810: The analysis for irreducible degenerations of $X$ is identical.
811:
812: The exponent of $u$ in the series
813: ${\mathsf{GW}}
814: (g\vertline k_{1},k_{2})_{\mu ^{1}\dots \mu ^{s}}^{\nu ^{1}\dots \nu
815: ^{t}}$ of relative
816: invariants has been precisely chosen to respect the
817: gluing rules.
818: \end{proof}
819:
820:
821:
822:
823:
824:
825: \section{TQFT formulation of gluing laws}
826: \label{tqft}
827: \subsection{Overview}
828: The gluing structure of the residue theory of rank 2 bundles on curves
829: is most concisely formulated as a functor of tensor categories,
830: \[
831: \mathbf{{GW}}(-):2\mathbf{Cob}^{L_{1},L_{2}}\to R\mathbf{mod}.
832: \]
833:
834: The deformation invariance of the residue theory allows for a
835: topological formulation of the gluing structure.
836:
837: Our discussion follows Sections 2 and 4 of \cite{Br-Pa-TQFT} and draws from
838: Chapter~1 of \cite{Kock:FA-2DTQFT}. Modifications of the categories have
839: to be made to accommodate the more complicated objects studied here.
840:
841:
842: \subsection{ $2\mathbf{Cob}$ and $2\mathbf{Cob}^{L_{1},L_{2}}$}
843:
844: We first define the category $2\mathbf{Cob}$
845: of 2-cobordisms. The objects of
846: $2\mathbf{Cob}$
847: are compact oriented
848: 1-manifolds, or equivalently, finite unions of oriented circles. Let
849: $Y_1$ and $Y_2$ be objects of the category. A morphism,
850: \[
851: Y_{1}\rightarrow Y_{2},
852: \]
853: is an equivalence class of oriented cobordisms $W$ from $Y_{1}$ to
854: $Y_{2}$. Two cobordisms are equivalent if they are diffeomorphic by a
855: boundary preserving oriented diffeomorphism. Composition of morphisms
856: is obtained by concatenation of the corresponding cobordisms. The
857: tensor structure on the category is given by disjoint union.
858:
859: The category $2\mathbf{Cob}^{L_{1},L_{2}}$ is defined to have the same objects as
860: $2\mathbf{Cob}$. A morphism in $2\mathbf{Cob}^{L_{1},L_{2}},$
861: $$Y_{1} \rightarrow Y_{2},$$
862: is an equivalence class of triples $(W,L_{1},L_{2})$ where $W$
863: is an oriented cobordism from $Y_{1}$ to $Y_{2}$ and $L_{1},L_{2}$ are
864: complex line bundles on $W$, trivialized on $\partial W$. The triples
865: $(W,L_{1},L_{2})$
866: and $(W',L'_{1},L'_{2})$ are equivalent if there exists a boundary preserving
867: oriented diffeomorphism, $$f:W\to W',$$ and bundle isomorphisms $$L_{i}\cong
868: f^{*}L'_{i}.$$ Composition is given by concatenation of the cobordisms and
869: gluing of the bundles along the concatenation using the trivializations.
870:
871: The isomorphism class of $L_{i}$ is determined by the Euler class $$e
872: (L_{i})\in H^{2} (W,\partial W),$$ which assigns an integer to each
873: component of $W$. For a connected cobordism $W$, we refer to the pair of
874: integers $(k_{1},k_{2})$, determined by the Euler classes of $L_{1}$ and
875: $L_{2}$, as the \emph{level}. Under concatenation, the
876: levels simply add. For example:
877:
878:
879: \begin{center}
880: \begin{texdraw}\setunitscale 1.3
881: \SFA
882: \move (120 0)
883: \boundaryFA \multNB \comultNB
884: \move (120 30)\boundaryFA
885: \move (200 0)\boundaryFA
886: \move (200 30)\boundaryFA
887: \htext (100 15) {$=$}
888: \htext (15 -5) {$\scriptstyle{(-3,1)}$}
889: \htext (15 30) {$\scriptstyle{(2,0)}$}
890: \htext (65 0) {$\scriptstyle{(7,-3)}$}
891: \htext (65 35) {$\scriptstyle{(-4,3)}$}
892: \htext (160 15) {$\scriptstyle{(2,1)}$}
893: \end{texdraw}
894: \end{center}
895:
896:
897:
898:
899: The empty manifold is a distinguished object in $2\mathbf{Cob}$
900: and $2\mathbf{Cob}^{L_{1},L_{2}}$. A morphism in
901: $2\mathbf{Cob}^{L_{1},L_{2}}$ from the empty manifold to itself is given by
902: a compact, oriented, closed 2-manifold $X$ together with a pair of complex
903: line bundles $L_{1}\oplus L_{2}\to X$.
904:
905: The full subcategory of $2\mathbf{Cob}^{L_{1},L_{2}}$ obtained by
906: restricting to level $(0,0)$ line bundles is clearly isomorphic to the
907: category $2\mathbf{Cob}$.
908:
909: More generally, we obtain an embedding $2\mathbf{Cob}\subset
910: 2\mathbf{Cob}^{L_{1},L_{2}}$ for any fixed integers $(a,b)$ by requiring
911: the level of any connected cobordism to be $(a\chi ,b\chi )$ where $\chi $
912: is the Euler characteristic of the cobordism.
913:
914: If $a+b=-1$, such an embedding is termed {\em Calabi-Yau}
915: since the threefold $$L_{1}\oplus L_{2}\to X$$
916: has numerically trivial canonical class if $$\deg (L_{1})+ \deg (L_{2})=-\chi.$$
917:
918: \subsection{Generators for $2\mathbf{Cob}$ and
919: $2\mathbf{Cob}^{L_{1},L_{2}}$}
920:
921: The category $2\mathbf{Cob}$ is generated by the morphisms
922: \[
923: \begin{texdraw}\setunitscale 1.0
924: \unitFA
925: \move (30 0)
926: \comultFA
927: \move (90 -15)
928: \multFA
929: \move (150 0)
930: \counitFA
931: \move (180 0)
932: \idFA
933: \move (240 -15)
934: \twistFA
935: \end{texdraw}
936: \]
937:
938:
939:
940:
941: In other words, any morphism (cobordism) can be obtained by taking
942: compositions and tensor products (concatenations and disjoint unions) of
943: the above list (Proposition~1.4.13 of \cite{Kock:FA-2DTQFT}).
944:
945: The category $2\mathbf{Cob}^{L_{1},L_{2}}$ is then clearly generated by the
946: morphisms
947: \begin{center}
948: \begin{texdraw}\setunitscale 1.0
949: \unitFA
950: \move (30 0)
951: \comultFA
952: \move (90 -15)
953: \multFA
954: \move (150 0)
955: \counitFA
956: \move (180 0)
957: \idFA
958: \move (240 -15)
959: \twistFA
960: \htext (-10 15) {$\scriptstyle{(0,0)}$}
961: \htext (45 0) {$\scriptstyle{(0,0)}$}
962: \htext (115 0) {$\scriptstyle{(0,0)}$}
963: \htext (155 15) {$\scriptstyle{(0,0)}$}
964: \htext (200 0) {$\scriptstyle{(0,0)}$}
965: \htext (245 30) {$\scriptstyle{(0,0)}$}
966: \htext (275 30) {$\scriptstyle{(0,0)}$}
967: \end{texdraw}
968: \end{center}
969: along with the morphisms
970: \begin{center}
971: \begin{texdraw}\setunitscale 1.0
972: \unitFA
973: \move (60 0)
974: \unitFA
975: \move (120 0)
976: \unitFA
977: \move (180 0)
978: \unitFA
979: \htext (-10 15) {$\scriptstyle{(0,1)}$}
980: \htext (50 15) {$\scriptstyle{(1,0)}$}
981: \htext (110 15) {$\scriptstyle{(0,-1)}$}
982: \htext (170 15) {$\scriptstyle{(-1,0)}$}
983: \end{texdraw}
984: \end{center}
985:
986: Let $R$ be a commutative ring with unit, and let $R\mathbf{mod}$ be
987: the tensor category of $R$-modules. By a well-known result (see
988: Theorem 3.3.2 of \cite{Kock:FA-2DTQFT}), a 1+1 dimensional $R$-valued
989: TQFT, which is by definition a symmetric tensor functor
990: \begin{equation}
991: \label{ejf8923}
992: \mathbf{F}:2\mathbf{Cob}\to R\mathbf{mod},
993: \end{equation}
994: is equivalent to a commutative Frobenius algebra over $R$.
995:
996: Given a symmetric tensor functor \eqref{ejf8923},
997: the underlying
998: $R$-module of the Frobenius algebra is given by
999: \[
1000: H=\mathbf{F} (S^{1})
1001: \]
1002: and the Frobenius algebra structure is determined as follows:
1003: \[
1004: \begin{array}{lrl}
1005: \text{multiplication}&\mathbf{F} (\intextmultFA ):&H\otimes H\to H\\
1006: \text{unit}&\mathbf{F} (\intextunitFA ):&R\to H\\
1007: \text{comultiplication}&\mathbf{F} (\intextcomultFA ):&H\to H\otimes H\\
1008: \text{counit}&\mathbf{F} (\intextcounitFA ):&H\to R.
1009: \end{array}
1010: \]
1011:
1012:
1013: Let $\mathbf{F}$ be a symmetric tensor functor on the larger category
1014: $2\mathbf{Cob}^{L_1,L_2}$,
1015: \[
1016: \mathbf{F}:2\mathbf{Cob}^{L_{1},L_{2}}\to R\mathbf{mod}.
1017: \]
1018: Since the functor $\mathbf{F}$ is determined by the
1019: values on the generators of $2\mathbf{Cob}^{L_1,L_2}$, the functor
1020: $\mathbf{F}$
1021: is determined by the level $(0,0)$ Frobenius algebra together with the
1022: elements
1023:
1024: \[
1025: \mathbf{F}\left(\intextunitFA
1026: ^{\sss{(0,1)}} \right),\
1027: \mathbf{F}\left(\intextunitFA
1028: ^{\sss{(1,0)}} \right),\
1029: \mathbf{F}\left(\intextunitFA
1030: ^{\sss{(0,-1)}} \right),\
1031: \mathbf{F}\left(\intextunitFA
1032: ^{\sss{(-1,0)}} \right),\
1033: \]
1034: Since the latter two elements are the inverses in the Frobenius
1035: algebra of the first two, we obtain half of the following Theorem.
1036:
1037: \begin{theorem}\label{thm: Z:2cobL1L2-->Rmod <--> Frob Alg and 2 inv elts}
1038: A symmetric tensor functor
1039: \[
1040: \mathbf{F}:2\mathbf{Cob}^{L_{1},L_{2}}\to R\mathbf{mod}
1041: \]
1042: is uniquely determined by a commutative Frobenius algebra over $R$
1043: for the level $(0,0)$ theory
1044: and two
1045: distinguished, invertible elements
1046:
1047: \[
1048: \mathbf{F}\left(\intextunitFA
1049: ^{\sss{(0,-1)}} \right),\
1050: \mathbf{F}\left(\intextunitFA
1051: ^{\sss{(-1,0)}} \right).
1052: \]
1053: \end{theorem}
1054:
1055: \begin{proof}
1056: Uniqueness was proved above. The existence result will not be used in
1057: the paper. We leave the details to
1058: the reader.
1059: \end{proof}
1060:
1061:
1062:
1063: \subsection{The functor $\mathbf{GW} (-)$}
1064:
1065: Let $R$ be the ring of Laurent series in $u$ whose coefficients are rational
1066: functions in $s _{1}$ and $s _{2}$,
1067: \[
1068: R=\qnums (t_{1},t_{2})((u)).
1069: \]
1070: The collection of partition functions
1071: $\mathsf{GW} (g\vertline k_{1},k_{2})_{\lambda
1072: ^{1}\dots \lambda ^{r}}$ of degree $d$ gives rise to a functor
1073: \[
1074: \mathbf{GW}(-):2\mathbf{Cob}^{L_1,L_2}\to R\mathbf{mod}
1075: \]
1076: as follows. Define
1077: \[
1078: \mathbf{GW} (S^{1})=H=\bigoplus _{\lambda \vdash d}Re_{\lambda }
1079: \]
1080: to be the free $R$-module with basis $\{e_{\lambda } \}_{\lambda \vdash d}$
1081: labeled by partitions of $d$, and let
1082: \[
1083: \mathbf{GW} \left(S^{1} \coprod\dots \coprod S^{1} \right)
1084: =H\otimes \dots \otimes H.
1085: \]
1086:
1087: Let $W_{s}^{t} (g\vertline k_{1},k_{2})$ be the connected genus $g$ cobordism from a
1088: disjoint union of $s$ circles to a disjoint union of $t$ circles, equipped
1089: with line bundles $L_{1}$ and $L_{2}$ of level $(k_{1},k_{2})$.
1090: We define the $R$-module homomorphism
1091: \[
1092: \mathbf{GW} \left(W_{s}^{t}
1093: (g\vertline k_{1},k_{2})\right):H^{\otimes s}\to H^{\otimes t}
1094: \]
1095: by
1096: \[
1097: e_{\eta ^{1}}\otimes \dots \otimes e_{\eta ^{s}}\mapsto \sum _{\mu
1098: ^{1}\dots \mu ^{t}\vdash d} {\mathsf{GW}} (g\vertline k_{1},k_{2})_{\eta ^{1}\dots
1099: \eta ^{s}}^{\mu ^{1}\dots \mu ^{t}}e_{\mu ^{1}}\otimes \dots \otimes
1100: e_{\mu ^{t}}.
1101: \]
1102: We extend the definition of $\mathbf{GW} (-)$ to disconnected cobordisms
1103: using tensor products:
1104: \[
1105: \mathbf{GW} \left(W[{1}]\coprod \dots \coprod W[{n}]\right)=\mathbf{GW}
1106: \left(W[{1}])\otimes \dots \otimes \mathbf{GW} (W[{n}]\right).
1107: \]
1108:
1109:
1110:
1111: \begin{theorem}\label{thm: gluing defines TQFT functor}
1112: $\mathbf{GW} (-): 2\mathbf{Cob}^{L_{1},L_{2}}\to
1113: R\mathbf{mod}$ is a well-defined functor.
1114: \end{theorem}
1115: \textsc{Proof:} Following the proof of Proposition~4.1 of
1116: \cite{Br-Pa-TQFT},
1117: the gluing laws imply the following compatibility:
1118: \[
1119: \mathbf{GW} \big((W,L_{1},L_{2})\circ
1120: (W',L'_{1},L'_{2})\big)=\mathbf{GW}(W,L_{1},L_{2})\circ \mathbf{GW}
1121: (W',L'_{1},L'_{2}).
1122: \]
1123: We must also prove that $\mathbf{GW} (-)$ takes identity morphisms to
1124: identity morphisms. Since $W_{1}^{1} (0\vertline 0,0)$ is the identity
1125: morphism from $S^{1}$ to itself in $2\mathbf{Cob}^{L_{1},L_{2}}$, we
1126: require
1127: \begin{equation}\label{jjww}
1128: {\mathsf{GW}} (0\vertline 0,0)_{\mu }^{\nu }=\delta _{\mu}^{\nu }.
1129: \end{equation}
1130: Equation \eqref{jjww} will be proved in Lemma~\ref{lem: charge (0,0) tube}.\qed
1131:
1132:
1133:
1134: \section{Semisimplicity in level $(0,0)$} \label{ss}
1135: \subsection{Rings of definition}
1136: The partition functions for the level $(0,0)$ relative invariants lie
1137: in the ring of power series in $u$,
1138: $$\mathsf{GW}
1139: (g\vertline 0,0)_{\lambda ^{1}\dots\lambda ^{r}} \in \qnums
1140: (t_{1},t_{2})[[u]],$$
1141: since, by
1142: equation \eqref{eqn: expression for Zdgk1k2_lambdas}, no negative powers of
1143: $u$ appear.
1144: The level $(0,0)$ relative invariants therefore
1145: determine a commutative
1146: Frobenius algebra over the ring
1147: \[
1148: R=\qnums(t_{1},t_{2})[[u]].
1149: \]
1150:
1151:
1152:
1153: We will require formal square
1154: roots of $t_{1} $ and $t_{2}$. Let $\tilde{R}$ be the complete local ring
1155: of power series in $u$ whose coefficients are rational functions in
1156: $t_{1}^{\frac{1}{2}}$ and $t_{2}^{\frac{1}{2}}$,
1157: \[
1158: \tilde{R}=\qnums(t_{1}^{\frac{1}{2}},t^{\frac{1}{2}}_{2}) [[u]].
1159: \]
1160:
1161:
1162: \subsection{Semisimplicity} A commutative Frobenius algebra $A$ is
1163: {\em semisimple} if $A$ is isomorphic to a direct sum of 1-dimensional
1164: Frobenius algebras.
1165:
1166: \begin{proposition}\label{prop: Zd is semi-simple}
1167: The Frobenius algebra determined by the level $(0,0)$ sector of
1168: $\mathbf{GW} (-)$ in degree $d$ is semisimple over $\tilde{R}$.
1169: \end{proposition}
1170: \textsc{Proof:}
1171: $\tilde{R}$ is a complete local ring with maximal ideal $m$ generated by $u$.
1172: Let $F$ be the Frobenius algebra determined by the level $(0,0)$ theory
1173: in degree $d$.
1174: The underlying $\tilde{R}$-module of the Frobenius
1175: algebra $F$,
1176: $$H=\oplus _{\lambda \vdash d}\tilde{R}e_{\lambda },$$
1177: is freely generated.
1178: By Proposition~2.2 of
1179: \cite{Br-Pa-TQFT}, $F$ is semisimple if and only if $F/mF$ is
1180: semisimple over $\tilde{R}/m\tilde{R}$.
1181:
1182: The structure constants of the multiplication
1183: in $F/mF$ are given by the $u=0$ specialization of the invariants
1184: $\mathsf{GW}(0\vertline 0,0)_{\alpha \beta }^{\gamma }$.
1185: By \eqref{eqn: expression for Zdgk1k2_lambdas}, after the $u=0$ specialization,
1186: only the
1187: $$b_{1}=b_{2}=0$$
1188: terms remain. The latter are the expected dimension 0 terms with domain genus
1189: \[
1190: 2h-2 = d -l (\alpha )-l (\beta )-l (\gamma ).
1191: \]
1192: In the expected dimension 0 case, the moduli space
1193: $\M (\P ^{1},\alpha, \beta, \gamma )$
1194: is nonsingular of {\em actual} dimension 0. We conclude:
1195: \begin{align*}
1196: \mathsf{GW} (0\vertline 0,0)_{\alpha \beta }^{\gamma }|_{u=0}&= \combinatfactor(\gamma)
1197: (t_{1}t_{2})^{l (\gamma )}\
1198: \mathsf{GW} (0\vertline 0,0)_{\alpha \beta \gamma }|_{u=0}\\
1199: &=
1200: \combinatfactor(\gamma)
1201: (t_{1}t_{2})^{\frac{1}{2} (d-l (\alpha )-l (\beta )+l (\gamma ))}
1202: \ \int _{[\M (\P^1,\alpha,\beta ,\gamma)]}1\\
1203: &=\combinatfactor(\gamma)
1204: (t_{1}t_{2})^{\frac{1}{2} (d-l (\alpha )-l (\beta )+l (\gamma ))}
1205: \ \mathbf{H}_{d}^{\P ^{1}} (\alpha ,\beta ,\gamma ),
1206: \end{align*}
1207: where $\mathbf{H}_{d}^{\P ^{1}} (\alpha ,\beta ,\gamma )$ is the Hurwitz
1208: number of degree $d$ covers of $\P ^{1}$ with prescribed ramification
1209: $\alpha $, $\beta $, and $\gamma $ over the
1210: points $0,1,\infty \in \P^1$.
1211:
1212: Up to factors of $t_1$ and $t_2$, the quotient $F/mF$ is
1213: the Frobenius algebra associated to
1214: the TQFT studied by Dijkgraaf-Witten and Freed-Quinn
1215: \cite{Dijkgraaf-Witten90,Freed-Quinn}. The latter Frobenius algebra is
1216: isomorphic to $\qnums [S_{d}]^{S_{d}}$, the center of the group
1217: algebra of the symmetric group, and is well-known to be
1218: semisimple.
1219:
1220:
1221:
1222:
1223: We derive below an explicit idempotent basis for $F/mF$ analogous to the
1224: well-known idempotent basis for $\qnums [S_{d}]^{S_{d}}$.
1225: The formal square roots of $t_{1}$ and $t_{2}$ are required here.
1226:
1227:
1228: Let $\rho$ be an irreducible representation of $S_d$. The conjugacy
1229: classes of $S_d$ are indexed by partitions $\lambda$ of size $d$. Let
1230: $\chi^\rho_\lambda$ denote the trace of $\rho$ on the conjugacy class
1231: $\lambda$. The Hurwitz numbers are determined by the following
1232: formula:
1233: \[
1234: \mathbf{H}^{\P ^{1}}_{d} (\alpha ,\beta ,\gamma )=
1235: \sum _{\rho }\frac{d!}{\dim \rho }
1236: \frac{\chi^\rho_\alpha}{\combinatfactor (\alpha )}
1237: \frac{\chi^\rho
1238: _\beta}{\combinatfactor (\beta )}
1239: \frac{\chi^\rho_\gamma }{\combinatfactor (\gamma )},
1240: \]
1241: see, for example, \cite{Okounkov-Pandharipande-completed-cycles} equation {0.8}.
1242: The above sum is over all irreducible representations $\rho$ of $S_d$.
1243:
1244: The structure constants for multiplication in $F/mF$ are
1245: \begin{equation}\label{eqn: u=0 structure constants}
1246: \mathsf{GW} (0\vertline 0,0)^{\gamma }_{\alpha \beta }|_{u=0}
1247: = (t_{1}t_{2})^{\frac{1}{2}
1248: (d-l (\alpha )-l (\beta )+l (\gamma ))} \sum _{\rho }\frac{d!}{\dim \rho
1249: } \frac{\chi^\rho_\alpha}{\combinatfactor(\alpha)}
1250: \frac{\chi^\rho_\beta}{\combinatfactor
1251: (\beta )}\chi^\rho_\gamma.
1252: \end{equation}
1253: We define a new basis $\{v_{\rho }^{0} \}$ for $F/mF$ by
1254: \begin{equation}\label{eqn: v0 basis}
1255: v_{\rho }^{0}=\frac{\dim \rho }{d!}\sum _{\alpha }
1256: \left(t_{1}^{\frac{1}{2}}t_{2}^{\frac{1}{2}} \right)^{l (\alpha )-d}\chi
1257: ^{\rho }_\alpha e_{\alpha }.
1258: \end{equation}
1259: The elements $\{v_{\rho }^{0} \}$ form an idempotent basis:
1260: \[
1261: v_{\rho }^{0}\cdot v_{\rho '}^{0}=\delta _{\rho \rho '}^{\rho ''}v_{\rho
1262: ''}^{0}.
1263: \]
1264: By Proposition~2.2 of \cite{Br-Pa-TQFT}, there exists a unique idempotent basis
1265: $\{v_{\rho } \}$ of $F$, such that $v_{\rho }=v_{\rho }^{0} \mod m$. \qed
1266:
1267: \begin{remark}
1268: In general, $v_{\rho }\neq v_{\rho }^{0}$ but for the
1269: anti-diagonal specialization $$t_{1}=-t_{2},$$
1270: the equality $v_{\rho }=v_{\rho }^{0}$ holds (see Section~\ref{sec: s1=-s2
1271: limit}).
1272: \end{remark}
1273:
1274:
1275: \subsection{Structure}
1276: Semisimplicity leads to a basic structure result.
1277: \begin{theorem}\label{thm: structure theorem}
1278: There exist universal series,
1279: $\lambda_{\rho },\eta _{\rho
1280: }\in \tilde{R},$ labeled by partitions $\rho$,
1281: for which
1282: \[
1283: \mathsf{GW}_d
1284: (g\vertline k_{1},k_{2})
1285: =\sum _{\rho\vdash d} \lambda_{\rho} ^{g-1}\eta_{\rho}
1286: ^{-k_{1}}\overline{\eta }_{\rho}^{-k_{2}}.
1287: \]
1288: Here, $\overline{\eta }_{\rho}$ is obtained from $\eta _{\rho}$ by
1289: interchanging $t_{1}$ with $t_{2}$.
1290: \end{theorem}
1291:
1292: \textsc{Proof:} Let $\{v_{\rho } \}$ be an idempotent basis for the level
1293: $(0,0)$ Frobenius algebra of $\mathsf{GW} (-)$ in degree $d$.
1294:
1295: Define $\lambda _{\rho }$ to be the inverse of the counit evaluated on
1296: $v_{\rho }$:
1297: \[
1298: \lambda _{\rho }^{-1}=\mathsf{GW} \left(\intextcounitFA
1299: ^{\sss{(0,0)}} \right) (v_{\rho }).
1300: \]
1301: Equivalently, $\lambda _{\rho }$ is the eigenvalue for the eigenvector
1302: $v_{\rho }$ under the \emph{genus adding operator} $G$:
1303: \[
1304: G=\mathsf{GW} \left(\intexthandleFA^{\sss{(0,0)}} \right):H\to H.
1305: \]
1306:
1307: Let $\eta _{\rho }$ (respectively $\overline{\eta }_{\rho }$) be the
1308: coefficient of $v_{\rho }$ in the element
1309: \[
1310: \eta =\mathsf{GW} \left(\intextunitFA^{\sss{(-1,0)}} \right)\in H,\quad
1311: (\text{respectively }\overline{\eta }=\mathsf{GW}
1312: \left(\intextunitFA^{\sss{(0,-1)}} \right)\in H).
1313: \]
1314: Equivalently, $\eta _{\rho }$ (respectively $\overline{\eta }_{\rho }$) is
1315: the eigenvalue for the eigenvector $v_{\rho }$ under the \emph{left
1316: annihilation operator} (respectively \emph{right annihilation
1317: operator}):
1318: \[
1319: A=\mathsf{GW} \left(\intextidFA^{\sss{(-1,0)}} \right)\quad (\text{respectively
1320: }\overline{A}=\mathsf{GW} \left(\intextidFA^{\sss{(0,-1)}} \right)).
1321: \]
1322: The gluing rules imply:
1323: \[
1324: {\mathsf{GW}} (g\vertline k_{1},k_{2})_d=
1325: \tr (G^{g-1}A^{-k_{1}}\overline{A}^{-k_{2}}).
1326: \]
1327: The operators $G$, $A$, and $\overline{A}$ are simultaneously diagonalized
1328: by the basis $\{v_{\rho } \}$, so the Theorem is equivalent to the above
1329: formula.\qed
1330:
1331:
1332:
1333:
1334: \section{Solving the theory} \label{ctt}
1335: \subsection{Overview}
1336: The {\em full local theory of curves} is the set
1337: of all series
1338: \begin{equation}
1339: \mathsf{GW}(g\vertline k_1,k_2)_{\lambda^1\dots\lambda^r}.
1340: \end{equation}
1341: The functor ${\mathbf {GW}}$ contains the
1342: data of the full local theory.
1343: By Theorem \ref{thm: Z:2cobL1L2-->Rmod <--> Frob Alg and 2 inv elts},
1344: the full local theory is
1345: determined by the following {\em basic} series:
1346: \begin{equation}
1347: \label{eqn: basic series} \mathsf{GW}(0\vertline 0,0)_{\lambda \mu \nu }\ , \
1348: \mathsf{GW}(0\vertline 0,0)_{\lambda \mu}\ , \
1349: \mathsf{GW}(0\vertline 0,0)_\lambda\ ,
1350: \end{equation}
1351: $$ \ \mathsf{GW}(0\vertline -1,0)_\lambda\ , \
1352: \mathsf{GW}(0\vertline 0,-1)_\lambda.$$
1353:
1354:
1355:
1356: We present a recursive method for calculating the full local
1357: theory of curves using the TQFT formalism.
1358: Four of the basic series,
1359: \[
1360: \mathsf{GW}(0\vertline 0,0)_{\lambda \mu}\ ,
1361: \ \mathsf{GW}(0\vertline 0,0)_\lambda\ , \
1362: \mathsf{GW}(0\vertline -1,0)_\lambda\ , \
1363: \mathsf{GW}(0\vertline 0,-1)_\lambda\ ,
1364: \]
1365: are determined by closed formulas. The first two are easily
1366: obtained by dimension considerations (Lemmas \ref{lem: charge (0,0) tube}
1367: and \ref{lem: charge (0,0) cap}). The last two have been determined in
1368: \cite{Br-Pa-TQFT} in case the equivariant parameters $t_i$ are set to
1369: 1. The insertion of the equivariant parameters is straightforward
1370: (Lemma~\ref{lem: the CY cap}).
1371:
1372:
1373: The level $(0,0)$ pair of pants series
1374: \[
1375: \mathsf{GW}(0\vertline 0,0)_{\lambda \mu \nu }
1376: \]
1377: are much more subtle.
1378: The main result of the Appendix (with C. Faber and A. Okounkov) is
1379: the determination of {all} degree $d$ level $(0,0)$ pair of pants series
1380: from the {\em single} series
1381: \begin{equation}\label{nnqq}
1382: \mathsf{GW}(0\vertline 0,0)_{(d),(d),(1^{d-2}2)}
1383: \end{equation}
1384: using the TQFT associativity relations, level $(0,0)$ series of {\em lower
1385: degree}, and Hurwitz numbers of covering genus 0. A closed formula for
1386: \eqref{nnqq} is derived in Section~\ref{qw223}. The outcome is a
1387: computation of the full local theory of curves via recursions in degree
1388: (Theorem~\ref{thm: reconstruction of level (0,0) pants}).
1389:
1390:
1391:
1392:
1393:
1394:
1395: \subsection{The level $(0,0)$ tube and cap}\label{subsec: level 0 tube and
1396: cap} We complete the proof of Theorem \ref{thm: gluing defines TQFT
1397: functor} by calculating the series
1398: $\mathsf{GW}(0\vertline 0,0)_\alpha^\beta$.
1399:
1400:
1401: \begin{lemma}\label{lem: charge (0,0) tube}
1402: The invariants of the level $(0,0)$ tube are given by:
1403: \[
1404: \mathsf{GW}(0\vertline 0,0)_{\alpha \beta }=\begin{cases}
1405: \frac{1}{\combinatfactor (\alpha )
1406: (t_{1}t_{2})^{l (\alpha )}} &\text{if $\alpha =\beta $}\\
1407: 0&\text{if $\alpha \neq \beta $.}
1408: \end{cases}
1409: \]
1410: Consequently, we have
1411: \[
1412: \mathsf{GW}(0\vertline 0,0)_{\alpha }^{\beta }=\delta _{\alpha }^{\beta }
1413: \]
1414: as was required for $\mathsf{GW}(-)$ to be a functor.
1415: \end{lemma}
1416: \textsc{Proof:}
1417: %\[
1418: %Z^{b_{1},b_{2}}_{d} (0\vertline 0,0)_{\alpha,\beta}=\int
1419: %_{[\M (X,\lambda ^{1},\dots ,\lambda ^{r})]^{vir}}c_{b_{1}} (-R^{\bullet
1420: %}\pi _{*}f^{*}L_{1})c_{b_{2}} (-R^{\bullet }\pi _{*}f^{*}L_{2}).
1421: %\]
1422: The virtual dimension of the moduli space
1423: $\overline{M}_{h} (\P ^{1},\alpha ,\beta )$ with {\em connected} domains is
1424: \[
1425: 2h-2+l (\alpha )+l (\beta ).
1426: \]
1427: Let $\E^\vee$ be the rank $h$ dual Hodge bundle on
1428: $\overline{M}_h(\P^1,\alpha,\beta)$. Since the line bundles $L_i$ may be
1429: taken to be trivial,
1430: $$c(-R^\bullet \pi_*f^*(L_i))=c(\E^\vee),$$
1431: where the equality is of ordinary (non-equivariant) Chern classes.
1432: The integral
1433: \begin{multline}
1434: \label{ppw3}
1435: \int _{[\overline{M}_{h} (\P ^{1},\alpha ,\beta )]^{vir}}c_{b_{1}}
1436: ( -R^\bullet \pi_*f^*(L_1) )c_{b_{2}} ( -R^\bullet \pi_*f^*(L_2) )= \\
1437: \int _{[\overline{M}_{h} (\P ^{1},\alpha ,\beta )]^{vir}}c_{b_{1}}
1438: (\E^{\vee })c_{b_{2}} (\E^{\vee })
1439: \end{multline}
1440: is zero if
1441: \[
1442: 2h-2+l (\alpha )+l (\beta )>2h.
1443: \]
1444: The only possible non-zero integrals are for $l (\alpha )=l (\beta )=1$.
1445: For $h>0$,
1446: $$c_{h}(\E^\vee)^2=0,$$
1447: by Mumford's relation. Hence, the
1448: integral \eqref{ppw3} is zero unless $h=0$.
1449:
1450: Therefore, the only connected stable map which contributes to the
1451: integral \eqref{ppw3} is the unique degree $d$ map
1452: \[
1453: f_{d}:\P ^{1}\to \P ^{1}
1454: \]
1455: totally ramified over $0$ and $\infty $. The only
1456: disconnected maps which contribute are disjoint unions of genus 0
1457: totally ramified maps of lower degree.
1458: Given a partition $\alpha \vdash d$, let
1459: \[
1460: f_{\alpha }:\bigsqcup _{l (\alpha )}\P ^{1}\to \P ^{1}
1461: \]
1462: be the map determined by $f_{\alpha _{i}}$ on the $i$th component. The map
1463: $f_{\alpha }$ has ramification profile $\alpha $ over both $0$ and $\infty$.
1464: The map is isolated in moduli and has an automorphism group of order
1465: $\combinatfactor (\alpha )$. Thus
1466: \[
1467: \mathsf{GW}^{b_{1},b_{2}} (0\vertline 0,0)_{\alpha\beta}
1468: =\begin{cases}
1469: \frac{1}{\combinatfactor (\alpha )}&\text{if $b_{1}=b_{2}=0$ and $\alpha =\beta $}\\
1470: 0&\text{otherwise.}
1471: \end{cases}
1472: \]
1473: The Lemma then follows directly from equation~(\ref{eqn: expression for
1474: Zdgk1k2_lambdas}).\qed
1475:
1476:
1477: The level $(0,0)$ cap has a simple form obtained by a similar dimensional
1478: argument.
1479:
1480: \begin{lemma}\label{lem: charge (0,0) cap}
1481: The invariants of the level $(0,0)$ cap are given by
1482: \[
1483: \mathsf{GW} (0\vertline 0,0)_{\lambda }=\begin{cases}
1484: \frac{1}{d!(t_{1}t_{2})^{d}}& \text{if $\lambda= (1^{d})$}\\
1485: 0&\text{if $\lambda \neq (1^{d})$.}
1486: \end{cases}
1487: \]
1488: \end{lemma}
1489: \textsc{Proof:} The (connected domain) moduli space $\overline{M}_{h} (\P
1490: ^{1},\lambda )$ has virtual dimension
1491: \[
1492: 2h-2+d+l (\lambda ).
1493: \]
1494: Hence,
1495: \begin{equation}\label{eqn: charge 0 cap, connected integral}
1496: \int _{[\overline{M}_{h} (\P ^{1},\lambda )]^{vir}}c_{b_{1}}
1497: (\E^{\vee })c_{b_{2}} (\E^{\vee }) = 0
1498: \end{equation}
1499: if
1500: \[
1501: 2h-2+d+l (\lambda )>2h.
1502: \]
1503: In order for (\ref{eqn: charge 0 cap, connected integral}) to be non-zero, we
1504: must have $d=l (\lambda )=1$. The virtual dimension is then $2h$, which
1505: implies $h=0$ by Mumford's relation.
1506:
1507:
1508: The only connected stable map for which the integral
1509: (\ref{eqn: charge 0 cap, connected integral}) is non-zero is the
1510: isomorphism
1511: \[
1512: f:\P ^{1}\to \P ^{1}.
1513: \]
1514: The Lemma is then obtained from \eqref{eqn: expression for
1515: Zdgk1k2_lambdas} by accounting for disconnected
1516: covers.\qed
1517:
1518:
1519: \subsection{The Calabi-Yau cap}
1520: \begin{lemma}\label{lem: the CY cap}
1521: The invariants of the level $(-1,0)$ cap are given by
1522: \[
1523: \mathsf{GW} (0\vertline -1,0)_{\lambda }= (-1)^{|\lambda|}
1524: (-t_{2})^{-l (\lambda
1525: )}\frac{1}{\combinatfactor (\lambda )}\prod _{i=1}^{l (\lambda
1526: )}\left(2\sin \frac{\lambda _{i}u}{2} \right)^{-1}
1527: \]
1528: \end{lemma}
1529: \textsc{Proof:} The calculation has already been done by
1530: localization in the proof of Theorem 5.1 in \cite{Br-Pa-TQFT} in case
1531: $t_1=t_2=1$. We must insert the equivariant parameters. The
1532: relevant {\em connected} integrals are
1533: \[
1534: \int _{[\overline{M}_{h} (\P ^{1},\lambda )]^{vir}}c_{b_{1}} (-R^{\bullet
1535: }\pi _{*}f^{*}\O (-1))c_{b_{2}} (-R^{\bullet }\pi _{*}f^{*}\O ).
1536: \]
1537: The virtual dimension of the moduli space $\overline{M}_{h} (\P
1538: ^{1},\lambda )$ is
1539: \[
1540: 2h-2+d+l (\lambda ).
1541: \]
1542: The object $-R^{\bullet }\pi _{*}f^{*}\O (-1)$ is represented by a
1543: bundle of rank $h-1+d$. Similarly, $-R ^{\bullet }\pi _{*}f^{*}\O $ is
1544: represented by a bundle of rank $h$ (minus a trivial
1545: factor). Consequently, the integral is zero unless $b_{1}=h-1+d$,
1546: $b_{2}=h$, and $\lambda = (d)$. From equation~(\ref{eqn: expression
1547: for Zdgk1k2_lambdas}), we find that the insertion of the equivariant
1548: parameters yields a factor of $t_{2}^{-1}$.
1549:
1550: Since the disconnected invariant
1551: is a product of $l (\lambda )$ connected integrals, the invariant has the
1552: factor $t_{2}^{-l (\lambda )}$.
1553: \qed
1554: \vspace{+10pt}
1555:
1556:
1557:
1558: The series $\mathsf{GW}(0\vertline 0,-1)_\lambda$
1559: is obtained from $\mathsf{GW}(0\vertline -1,0)_\lambda$
1560: by exchanging $t_1$ and $t_2$.
1561:
1562:
1563: \subsection{The level $(0,0)$ pair of pants}
1564:
1565:
1566:
1567:
1568:
1569:
1570:
1571:
1572:
1573:
1574:
1575:
1576: \subsubsection{Normalization}\label{n3451}
1577: We will study the local theory of curves here with a slightly different
1578: normalization. Recall
1579: \[
1580: \delta =\sum _{i=1}^{r} (d-l (\lambda ^{i})).
1581: \]
1582: Let
1583: \begin{align}\label{eqn: Zhat}
1584: {\mathsf{GW}}^*
1585: (g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}}&=
1586: (-iu)^{d (2-2g+k_{1}+k_{2})-\delta } \
1587: {\mathsf Z}'(N)_{ \lambda^1 \dots \lambda^r}\\
1588: &= \ \ (-i)^{d (2-2g+k_{1}+k_{2})-\delta }\ \mathsf{GW}
1589: (g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}}.\nonumber
1590: \end{align}
1591: With the altered metric,
1592: \[
1593: {\mathsf{GW}}^* (g\vertline k_{1},k_{2})_{\mu ^{1}\dots \mu ^{s}}^{\nu
1594: ^{1}\dots \nu ^{t}}= \left(\prod _{i=1}^{t} \combinatfactor (\nu ^{i})
1595: (-t_{1}t_{2})^{l (\nu ^{i})} \right) {\mathsf{GW}}^* (g\vertline
1596: k_{1},k_{2})_{\mu
1597: ^{1}\dots \mu ^{s}\nu ^{1}\dots \nu ^{t}},
1598: \]
1599: the partition functions \eqref{eqn: Zhat} satisfy the same gluing rules
1600: as partition functions $\mathsf{GW}
1601: (g\vertline k_1,k_2)_{\lambda^1\dots \lambda^r}$.
1602: Moreover, a tensor functor,
1603: \[
1604: {\mathbf{GW^*}}:2\mathbf{Cob}^{L_{1},L_{2}}\to R\mathbf{mod}.
1605: \]
1606: is defined just as before.
1607:
1608: The reason for the altered normalization is the following
1609: result proven in the Appendix.
1610:
1611: \begin{theorem}\label{thm: Z is a rational function of q}
1612: The product
1613: \[
1614: e^{\frac{idu}{2} (2-2g+{k_{1}}+{k_{2}})}\
1615: \mathsf{GW}^*(g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}} ,
1616: \]
1617: is a \emph{rational function} of $t_{1}$,
1618: $t_{2}$, and $q=-e^{iu}$ with $\qnums$-coefficients.
1619: \end{theorem}
1620:
1621:
1622: The Theorem is closely related to the GW/DT correspondence discussed in
1623: Section~\ref{subsec: GW/DT for local curves}.
1624: The Calabi-Yau cap provides a good example:
1625: \begin{eqnarray*}
1626: e^{\frac{idu}{2}}\
1627: \mathsf{GW}^*
1628: (0\vertline -1,0)_{\lambda}& =& e^{\frac{idu}{2}} (-i)^{l(\lambda)}
1629: (-1)^{d} (-t_{2})^{-l (\lambda
1630: )}\frac{1}{\combinatfactor (\lambda )}\prod _{j=1}^{l (\lambda
1631: )}\left(2\sin \frac{\lambda _{j}u}{2} \right)^{-1} \\
1632: %& =&(-1)^d \frac{1}{\combinatfactor (\lambda )}\frac{1}{ s_2^{l(\lambda)}}
1633: %\prod _{j=1}^{l (\lambda
1634: %)} \frac{ie^{\frac{i\lambda_ju}{2}} }{2\sin \frac{\lambda _{j}u}{2}}\\
1635: % & = &
1636: %(-1)^{d-l(\lambda)} \frac{1}{\combinatfactor (\lambda )}\frac{1}{ s_2^{l(\lambda)}}
1637: %\prod _{j=1}^{l (\lambda
1638: %)} \frac{e^{\frac{i\lambda_ju}{2}} }{e^{\frac{i\lambda_j u}{2}} - e^{\frac{-i\lambda_ju}{2}}}
1639: % \\
1640: & = &
1641: (-1)^{d-l(\lambda)} \frac{1}{\combinatfactor (\lambda )}\frac{1}
1642: { t_2^{l(\lambda)}}
1643: \prod _{j=1}^{l (\lambda
1644: )} \frac{1}{1 - (-q)^{-\lambda_j}}.
1645: \end{eqnarray*}
1646:
1647:
1648: \subsubsection{The degree 1 case}
1649: The level $(0,0)$ tube and cap in degree 1 are:
1650: $$\mathsf{GW}^*
1651: (0\vertline 0,0)_{\tableau{1},\tableau{1}} = -\frac{1}{t_1t_2},
1652: \ \mathsf{GW}^*(0\vertline 0,0)_{\tableau{1}} = -\frac{1}{t_1t_2}.$$
1653: By the gluing formula,
1654: $$\mathsf{GW}^*
1655: (0\vertline 0,0)_{\tableau{1},\tableau{1},\tableau{1}}
1656: \ (-t_1t_2)\ \mathsf{GW}^*(0\vertline 0,0)_{\tableau{1}} =
1657: \mathsf{GW}^*(0\vertline 0,0)_{\tableau{1},\tableau{1}}.$$
1658: We conclude,
1659: $$\mathsf{GW}^*
1660: (0\vertline 0,0)_{\tableau{1},\tableau{1},\tableau{1}} = -\frac{1}{t_1t_2}.$$
1661: Hence, all the basic series in degree 1 are known.
1662:
1663: \subsubsection{The series
1664: $\mathsf{GW}^*(0\vertline 0,0)_{(d),(d),(1^{d-2}2)}$}
1665: \label{qw223}
1666: The degree $d\geq 2$ series
1667: $\mathsf{GW}(0\vertline 0,0)_{(d),(d),(1^{d-2}2)}$ plays a
1668: special role in the level $(0,0)$ theory.
1669:
1670: \begin{theorem}\label{thm: the dd2 integral}
1671: For $d\geq 2$,
1672: \begin{eqnarray*}
1673: \mathsf{GW}^* (0\vertline 0,0)_{(d), (d),
1674: (1^{d-2}2)}& = &-\frac{i}{2}\frac{t_{1}+t_{2}}{t_{1}t_{2}}\left(d\cot
1675: \left(\frac{du}{2}\right)-\cot \left(\frac{u}{2}\right) \right) \, .
1676: \end{eqnarray*}
1677: \end{theorem}
1678:
1679: \textsc{Proof:} We abbreviate the partition $(1^{d-2}2)$ by $(2)$. After
1680: adjusting
1681: equation \eqref{eqn: expression for Zdgk1k2_lambdas} for the new normalization, we
1682: find
1683: \begin{multline*}
1684: \mathsf{GW}^*
1685: (0\vertline 0,0)_{(d), (d), (2)}=\\
1686: -\frac{i} {(t_{1}t_{2})^{\frac{1}{2}}} \sum
1687: _{b_{1},b_{2}=0}^{\infty }u^{b_1+b_2}\left(\frac{t_{1}}{t_{2}}
1688: \right)^{\frac{b_{2}-b_{1}}{2}} \int _{[\M (\P ^{1},(d), (d), (2))]^{vir}}
1689: c_{b_{1}} (\E^{\vee })c_{b_{2}} (\E^{\vee }).
1690: \end{multline*}
1691: The domains of the maps in the moduli space $\M (\P ^{1},(d), (d),
1692: (2))$ are necessarily connected since there exists a point of total
1693: ramification. Since the virtual dimension is
1694: \[
1695: \operatorname{virdim}\overline{M}_{h} (\P ^{1},(d), (d), (2))=2h-1,
1696: \]
1697: the only values of $(b_{1},b_{2})$ which contribute to
1698: $\mathsf{GW}^*(0\vertline 0,0)_{(d), (d), (2)}$ are
1699: $(h,h-1)$ and $(h-1,h)$. We obtain:
1700: \[
1701: \mathsf{GW}^*
1702: (0\vertline 0,0)_{(d), (d), (2)}=-i\frac{t_{1}+t_{2}}{t_{1}t_{2}}\sum
1703: _{h=1}^{\infty }u^{2h-1}\int _{[\overline{M}_{h} (\P ^{1},(d), (d),
1704: (2))]^{vir}}\rho ^{*} (-\lambda _{h}\lambda _{h-1}).
1705: \]
1706: Here, $\lambda _{k}$ is the $k^{th}$ Chern class of the Hodge bundle on
1707: $\overline{M}_{h,2}$, and
1708: \[
1709: \rho :\overline{M}_{h} (\P ^{1},(d), (d), (2))\to \overline{M}_{h,2}
1710: \]
1711: is the natural map which takes a relative stable map to the domain marked
1712: by the two totally ramified points.
1713:
1714: Let
1715: $
1716: H_{d}\subset M_{h,2}
1717: $
1718: be the locus of curves admitting a degree $d$ map to $\P ^{1}$ which is totally
1719: ramified at the marked points. Equivalently, $H_{d}$ is the locus of curves
1720: $$[C,x_{1},x_{2}]$$ for which $\O (x_{1}-x_{2})$ is a nonzero $d$-torsion point in
1721: $\operatorname{Pic}^{0} (C)$. Let $$\overline{H}_{d}\subset
1722: \overline{M}_{h,2}$$ be the closure of $H_{d}$.
1723:
1724:
1725: Consider the locus of maps with nonsingular domains,
1726: $$M_h(\P^1,(d),(d),(2))\subset \overline{M}_h(\P^1,(d),(d),(2)),$$
1727: and let
1728: $$\partial \overline{M}_h(\P^1,(d),(d),(2))$$
1729: denote the complement.
1730: Let
1731: $$\partial \overline{M}_{h,1} \subset \overline{M}_{h,1}$$
1732: denote the nodal locus.
1733: Let $$\epsilon: \overline{M}_{h,2} \rightarrow\overline{M}_{h,1}$$
1734: be the map forgetting the first point.
1735: An elementary argument yields
1736: $$\rho\left( \partial \overline{M}_h(\P^1,(d),(d),(2)) \right) \subset \epsilon^{-1}(\partial
1737: \overline{M}_{h,1}).$$
1738:
1739: The restriction of the virtual class to
1740: $M_h(\P^1,(d),(d),(2))$ is well-known to equal the ordinary fundamental
1741: class of the moduli space, see \cite{pandharipande-toda2000}. Since
1742: $$\rho:M_h(\P^1,(d),(d),(2)) \rightarrow H_d$$
1743: is a proper cover of degree $2h$,
1744: we conclude
1745: \begin{equation}\label{eqn: pushforward of dd2 rel space to Mh2}
1746: \rho _{*}[\overline{M}_{h} (\P ^{1},(d), (d) ,(2))]^{vir}=2h[\overline{H}_{d}]+B
1747: \end{equation}
1748: where $B$ is a cycle supported on $\epsilon^{-1}(\partial
1749: \overline{M}_{h,1})$.
1750:
1751:
1752:
1753: Since $\lambda _{h}\lambda _{h-1}$ vanishes on cycles supported on the
1754: boundary of $\overline{M}_{h,1}$, we find
1755: \[
1756: \mathsf{GW}^*
1757: (0\vertline 0,0)_{(d), (d), (2)}=i\frac{t_{1}+t_{2}}{t_{1}t_{2}}\sum
1758: _{h=1}^{\infty }u^{2h-1} c_h(d),
1759: \]
1760: where
1761: \[
1762: c_{h} (d)=2h\int _{[\overline{H}_{d}]}\lambda _{h}\lambda _{h-1}.
1763: \]
1764:
1765: The cycle $[H_{d}]$ can be described as follows. Let
1766: \[
1767: \begin{diagram}[height=0.8cm]
1768: \operatorname{\mathcal{P}ic}^{0}\\
1769: \uTo^{s} \dTo_{\pi }\\
1770: M_{h,2}
1771: \end{diagram}
1772: \]
1773: be the universal Picard bundle with section
1774: \[
1775: s:[C,x_{1},x_{2}]\mapsto \O (x_{1}-x_{2}).
1776: \]
1777: Let $P_{d}\subset \operatorname{\mathcal{P}ic^{0}}$ be the locus of
1778: \emph{nonzero} $d$-torsion points. Then, by our previous characterization
1779: of $H_{d}$,
1780: \[
1781: [H_{d}]=\pi _{*}\left(s_{*}[M_{h,2}]\cap P_{d} \right)\in A_*(M_{h,2}).
1782: \]
1783: By a result of Looijenga using the Fourier-Mukai transform, the locus
1784: of $d$-torsion points of {\em any} family of Abelian varieties is a
1785: multiple of the zero section in the Chow ring
1786: \cite{Looijenga-invent}. Hence,
1787: \[
1788: [P_{d}]=\frac{d^{2h}-1}{2^{2h}-1}[P_{2}]
1789: \]
1790: and
1791: \[
1792: [H_{d}]=\frac{d^{2h}-1}{2^{2h}-1}[H_{2}].
1793: \]
1794: We conclude
1795: \[ c_{h}
1796: (d)=\frac{d^{2h}-1}{2^{2h}-1}c_{h} (2).
1797: \]
1798:
1799:
1800: Consider the $d=2$ case.
1801: In genus 1, the class $$[\overline{H}_{2}]\in A_*(\overline{M}_{1,2})$$
1802: pushes forward to $3[\overline{M}_{1,1}]$ under the map $\epsilon$.
1803: For genus $h>1$, let $\overline{H}\subset \overline{M}_{h}$ denote the
1804: hyperelliptic locus. There are
1805: $$(2h+2) (2h+1)$$ ways of marking two of the
1806: Weierstrass points on each curve in $H$. Consequently, the class
1807: $$[\overline{H}_{2}]\in A_*( \overline{M}_{h,2})$$ pushes forward to $(2h+2)
1808: (2h+1)[\overline{H}]$ under the forgetful map
1809: $$\overline{M}_{h,2}\to
1810: \overline{M}_{h}.$$
1811: We find
1812: \begin{align*}
1813: \mathsf{GW}^* (0\vertline 0,0)_{(2), (2), (2)}&
1814: =i\frac{t_{1}+t_{2}}{t_{1}t_{2}}\left(6u\int
1815: _{\overline{M}_{1,1}}\lambda _{1}+\sum _{h=2}^{\infty } \frac{(2h+2)!}{(2h-1)!}
1816: u^{2h-1}\int _{\overline{H}}\lambda _{h}\lambda _{h-1}\right)\\
1817: &=i\frac{t_{1}+t_{2}}{t_{1}t_{2}}\left(\frac{u^{4}}{96}+
1818: \sum _{h=2}^{\infty }u^{2h+2}\int _{\overline{H}}\lambda _{h}\lambda _{h-1} \right)'''\\
1819: &=i\frac{t_{1}+t_{2}}{t_{1}t_{2}}\left(u^{2}H (u) \right)'''
1820: \end{align*}
1821: where $H (u)$ is defined in \cite{Faber-Pandharipande-logarithmic} on page 222.
1822: By Corollary 2 of
1823: \cite{Faber-Pandharipande-logarithmic},
1824: \[
1825: (u^{2}H (u))''=-\log \left(\cos \left(\frac{u}{2} \right) \right),
1826: \]
1827: and thus
1828: \[
1829: \mathsf{GW}^*
1830: (0\vertline 0,0)_{(2), (2), (2)}=\frac{i}{2}
1831: \frac{t_{1}+t_{2}}{t_{1}t_{2}}\tan
1832: \left(\frac{u}{2} \right).
1833: \]
1834: We conclude
1835: \[
1836: \sum _{h=1}^{\infty }c_{h} (2)u^{2h-1}=\frac{1}{2}\tan\left(\frac{u}{2}
1837: \right).
1838: \]
1839:
1840:
1841: The function $\cot\left(\frac{u}{2} \right)$
1842: is an odd series in $u$ with a simple
1843: pole at $u=0$. We define $b_{h}$ by
1844: \[
1845: \cot\left(\frac{u}{2} \right) = \sum _{h=0}b_{h}u^{2h-1}.
1846: \]
1847: The identity
1848: \[
1849: \frac{1}{2}\tan\left(\frac{u}{2} \right) =
1850: \frac{1}{2}\cot\left(\frac{u}{2} \right)-\cot\left(2\frac{u}{2} \right)
1851: \]
1852: yields
1853: \[
1854: c_{h} (2)= (\frac{1}{2}-2^{2h-1})b_{h}.
1855: \]
1856: Hence,
1857: \[
1858: c_{h} (d)=\frac{1}{2} (1-d^{2h})b_{h}.
1859: \]
1860: We obtain
1861: \[
1862: \mathsf{GW}^* (0\vertline 0,0)_{(d), (d),
1863: (2)}=-\frac{i}{2}\frac{t_{1}+t_{2}}{t_{1}t_{2}}\left(
1864: d\cot\left(\frac{du}{2} \right)- \cot \left(\frac{u}{2}
1865: \right) \right)
1866: \]
1867: which concludes the proof.\qed
1868:
1869: We may write the series as a rational function in $-q=e^{iu}$,
1870: \begin{equation}\label{fferr}
1871: \mathsf{GW}^* (0\vertline 0,0)_{(d), (d),
1872: (2)} =
1873: \frac{1}{2} \frac{t_{1}+t_{2}}{t_{1}t_{2}} \left(
1874: d\frac{(-q)^d+1}{(-q)^d-1} - \frac{(-q)+1}{(-q)-1}\right).
1875: \end{equation}
1876:
1877:
1878: \subsection{Reconstruction for the level $(0,0)$ pair of pants}
1879: The main result proven in the Appendix (with C. Faber and A. Okounkov) is the
1880: following.
1881: \begin{theorem}\label{thm: reconstruction of level (0,0) pants}
1882: Let $d\geq 2$. The set of degree $d$, level $(0,0)$ pair of pants series
1883: \[
1884: \mathsf{GW}^*(0\vertline 0,0)_{\lambda \mu \nu}
1885: \]
1886: can be uniquely reconstructed from
1887: \[
1888: \mathsf{GW}^* (0\vertline 0,0)_{(d), (d),
1889: (2)}
1890: \]
1891: via the TQFT associativity relations, lower degree series of level $(0,0)$,
1892: and Hurwitz numbers of covering genus $0$.
1893: \end{theorem}
1894:
1895: The proof yields an effective method of computing the level $(0,0)$ pair of
1896: pants series via recursions in degree. Since all the basic series
1897: \eqref{eqn: basic series} can be computed, the full local theory of curves is
1898: effectively determined. Theorem \ref{thm: Z is a rational function of q}
1899: is obtained in the Appendix as a Corollary of Theorem~\ref{thm:
1900: reconstruction of level (0,0) pants}.
1901:
1902:
1903:
1904:
1905:
1906:
1907:
1908:
1909:
1910:
1911:
1912:
1913:
1914: \section{The anti-diagonal action}\label{sec: s1=-s2 limit}
1915: \subsection{Overview}
1916: We study a well-behaved special case of the local
1917: theory of curves. Consider the action of the
1918: anti-diagonal subgroup
1919: $$\TT = \{ (\xi, \xi^{-1})\ | \ \xi \in \cnums^*\} \subset T.$$
1920: on $N=L_{1}\oplus L_{2}$. The anti-diagonal action
1921: corresponds
1922: to the limit
1923: \[
1924: t_{1}+t_2=0
1925: \]
1926: in equivariant cohomology.
1927: The
1928: induced $\TT$-action on $K_{N}$ is trivial.
1929: Explicit formulas can be found since the level
1930: $(0,0)$ Frobenius algebra can be explicitly diagonalized in
1931: the anti-diagonal case.
1932:
1933: We define the $Q$-dimension of $\rho $, an irreducible representation
1934: of the symmetric group, indicated $\dim _{Q}\rho $, as follows:
1935: \[
1936: \frac{\dim _{Q}\rho }{d!}=
1937: \prod _{\Box \in \rho } i\left({ Q^{\frac{h (\Box)}{2}}
1938: -Q^{\frac{-h (\Box)}{2}}} \right)^{-1},
1939: \]
1940: see \cite{Okounkov-Pandharipande-unknot}.
1941: Under the substitution $Q=e^{iu}$, the $Q$-dimension can be expressed as:
1942: \[
1943: \frac{\dim _{Q}\rho }{d!}
1944: = \prod _{\Box \in \rho }
1945: \left(2 \sin\frac{ h (\Box)u}{2} \right)^{-1}.
1946: \]
1947: By the hook length formula for $\dim \rho $, the
1948: leading term in $u$ of the above expression is $\frac{\dim \rho}{d!} $.
1949:
1950:
1951: The main result
1952: here is a closed formula for the (absolute) local theory of curves
1953: with the anti-diagonal action.
1954:
1955: \begin{theorem}\label{thm: s1=-s2 general formula}
1956: Under the restrictions $t_{1}=t$ and $t_{2}=-t$,
1957: \begin{multline*}
1958: \mathsf{GW}_d(g\vertline k_{1},k_{2})
1959: =\\ (-1)^{d (g-1-k_{2})}t^{d (2g-2-k_{1}-k_{2})}\
1960: \sum
1961: _{\rho }\left(\frac{d!}{\dim \rho } \right)^{2g-2}\left(\frac{\dim \rho
1962: }{\dim _{Q}\rho } \right)^{k_{1}+k_{2}} Q^{\frac{1}{2}c_{\rho }
1963: (k_{1}-k_{2})}
1964: \end{multline*}
1965: where $Q=e^{iu}$ and $c_{\rho }$ is the total content of $\rho $
1966: (see Section \ref{subsub: conventions for partitions}).
1967: \end{theorem}
1968:
1969: \subsection{Corollaries}
1970:
1971: If $k_{1}+k_{2}=2g-2$, the threefold $N=L_{1}\oplus L_{2}$ is
1972: Calabi-Yau. As previously remarked, the $t_{1}+t_{2}=0$
1973: limit corresponds to
1974: the trivial $\TT$-action on the canonical
1975: bundle. In other words, $N$ is \emph{equivariantly} Calabi-Yau.
1976:
1977: \begin{cor}\label{cor: s1=-s2 limit, CY case}
1978: In the equivariantly Calabi-Yau case,
1979: \[
1980: \mathsf{GW}_d (g\vertline k,2g-2-k)
1981: = (-1)^{d (g-1-k)}\sum _{\rho }\left(\frac{d!}{\dim
1982: _{Q}\rho } \right)^{2g-2} Q^{-c_{\rho } (g-1-k)}.
1983: \]
1984: \end{cor}
1985: In particular, for the balanced splitting,
1986: $$k_1=k_2=g-1,$$
1987: the partition function is a
1988: $Q$-deformation of the classical formula for unramified covers.
1989: \begin{cor}\label{cor: s1=-s2, generic CY}
1990: In the balanced equivariantly Calabi-Yau case,
1991: \[
1992: \mathsf{GW}_d (g\vertline g-1,g-1)
1993: =\sum _{\rho }\left(\frac{d!}{\dim _{Q}\rho } \right)^{2g-2}.
1994: \]
1995: \end{cor}
1996: Another special Calabi-Yau case is when the base curve $X$ is elliptic.
1997: We obtain a formula recently derived by Vafa using string theoretic
1998: methods (page 8 of \cite{Vafa-04-2dYang-Mills}).
1999: \begin{cor}\label{cor: vafa's formula}
2000: Let $L\to E$ be a degree $k$ line bundle on an elliptic curve $E$. The
2001: partition function for the Calabi-Yau action on $L\oplus L^{-1}$ is
2002:
2003: \[
2004: \mathsf{GW}_d(1\vertline k,-k)=(-1)^{dk}\sum _{\rho } Q^{kc_{\rho }}.
2005: \]
2006: \end{cor}
2007:
2008: \subsection{Proof of Theorem \ref{thm: s1=-s2 general formula}}
2009:
2010: To derive the formula of Theorem~\ref{thm: s1=-s2 general formula}, we
2011: first explicitly diagonalize the level $(0,0)$ Frobenius algebra for the
2012: anti-diagonal action.
2013: \begin{lemma}\label{lem: s1=-s2 charge 0,0 structure constants}
2014: For the anti-diagonal action, the level $(0,0)$ series have no
2015: nonzero terms of positive degree in $u$.
2016: \end{lemma}
2017:
2018:
2019: \textsc{Proof of Lemma.} Let $\cnums _{t}$ denote $\TT$-representation
2020: given by the standard action of the projection of
2021: $$\TT \subset T= \cnums^* \times \cnums^*$$
2022: on the first factor, and let
2023: \[
2024: c_{1} (\cnums _{t})=t.
2025: \]
2026: The dual line bundle is $\cnums _{t}^{\vee }=\cnums
2027: _{-t}$.
2028:
2029: The level $(0,0)$ partition functions are built from the
2030: following integrals:
2031: \[
2032: \int _{[\M (X,\lambda ^{1},\dots ,\lambda ^{r})]^{vir}}
2033: e \left(-R^{\bullet }\pi
2034: _{*} (\O \otimes \cnums _{t}) \right)
2035: e (-R^{\bullet }\pi
2036: _{*} (\O \otimes \cnums _{-t})).
2037: \]
2038: For any vector bundle $E$, the equivariant Euler class $e (E\otimes \cnums
2039: _{t})$ is a polynomial in $t$ whose coefficients are the (ordinary) Chern
2040: classes of $E$. The above integrand is a weight factor times
2041: \[
2042: e (\E ^{\vee }\otimes \cnums _{t})e (\E^{\vee } \otimes \cnums _{-t}) =
2043: (-1)^{h} e \left(\left(\E ^{\vee }\oplus \E \right)\otimes \cnums _{t}
2044: \right).
2045: \]
2046: Since the Chern classes of $\E ^{\vee }\oplus \E $ all vanish by Mumford's
2047: relation, the last expression is pure weight. The only non-zero
2048: integrals occur when
2049: $$b_{1}=b_{2}=0$$
2050: in equation \eqref{eqn: expression for Zdgk1k2_lambdas}.
2051: Only the constant terms in $u$ are
2052: non-zero. In particular, $\mathsf{GW}
2053: (0\vertline 0,0)^{\gamma }_{\alpha \beta }$ is given
2054: by the $t_{1}+t_{2}=0$ limit of equation (\ref{eqn: u=0 structure
2055: constants}).\qed
2056: \vspace{+10pt}
2057:
2058: The structure
2059: constants for the level $(0,0)$ Frobenius algebra are given by:
2060: \[
2061: \mathsf{GW}
2062: (0\vertline 0,0)_{\alpha \beta }^{\gamma }=\left(-t^{2} \right)^{\frac{1}{2}
2063: (d-l (\alpha )-l (\beta )+l (\gamma ))}\sum _{\rho }\left(\frac{d!}{\dim
2064: \rho } \right)\frac{\chi^{\rho }_\alpha \chi^{\rho }_\beta}
2065: {\combinatfactor (\alpha )\combinatfactor (\beta )}\chi^{\rho}_
2066: \gamma.
2067: \]
2068: As a consequence of the Lemma, multiplication in the level $(0,0)$
2069: Frobenius algebra is diagonalized by the basis $v_{\rho }^{0}$ constructed
2070: in the proof of Proposition~\ref{prop: Zd is semi-simple}.
2071:
2072:
2073: In order to diagonalize the level $(0,0)$ Frobenius algebra,
2074: we had to enlarge the
2075: coefficient ring to $\tilde{R}$ to include the formal square roots
2076: $t_{1}^{\frac{1}{2}}$ and $t_{2}^{\frac{1}{2}}$. The specialization
2077: \[
2078: t_{1}^{\frac{1}{2}}=t^{\frac{1}{2}},\quad t_{2}^{\frac{1}{2}}
2079: =it^{\frac{1}{2}}
2080: \]
2081: is compatible with
2082: $$t_1=t, \quad t_2=-t.$$
2083: By Lemma \ref{lem: s1=-s2 charge 0,0 structure constants}, the idempotent
2084: basis $v_{\rho }^{0}=v_{\rho }$ given by equation \eqref{eqn: v0 basis} is:
2085: \begin{equation}\label{dfg}
2086: v_{\rho }=\frac{\dim \rho }{d!}\sum _{\alpha } (it)^{l (\alpha )-d}\chi
2087: ^{\rho }_\alpha e_{\alpha }.
2088: \end{equation}
2089:
2090:
2091: To apply Theorem \ref{thm: structure theorem}, we must
2092: compute $\lambda _{\rho }$, $\eta _{\rho }$, and $\overline{\eta }_{\rho
2093: }$. We compute $\lambda _{\rho }$ as follows:
2094: \begin{align*}
2095: \lambda _{\rho }^{-1}&=\mathbf{GW} \left(\intextcounitFA
2096: ^{\sss{(0,0)}} \right) (v_{\rho })\\
2097: &=\frac{\dim \rho }{d!}\sum _{\alpha } (it)^{l (\alpha )-d}\chi
2098: ^{\rho }_\alpha \ \mathsf{GW} (0\vertline 0,0)_{\alpha }\\
2099: &=\frac{\dim \rho }{d!} (it)^{l (1^{d} )-d}\chi
2100: ^{\rho }_{(1^{d})}\frac{1}{d!(-t^{2})^{d}}\\
2101: &=\left(\frac{\dim \rho }{d!} \right)^{2} (it)^{-2d}.
2102: \end{align*}
2103: Hence,
2104: $$\lambda _{\rho } =(it)^{2d}\left(\frac{d!}{\dim \rho } \right)^{2}.$$
2105:
2106: In order to compute $\eta _{\rho}$, we must express $\eta $ in
2107: terms of the basis
2108: $\{v_{\rho } \}$.
2109: \begin{align*}
2110: \eta& =\mathbf{GW}\left(\intextunitFA
2111: ^{\sss{(-1,0)}} \right)\\
2112: &=\sum _{\alpha }
2113: \mathsf{GW} (0\vertline -1,0)^{\alpha }e_{\alpha }\\
2114: &=\sum _{\alpha } (-1)^{d}t^{l (\alpha )}\left(\prod _{i=1}^{l (\alpha )}
2115: \frac{-1}{2\sin \frac{\alpha _{i}u}{2}} \right)e_{\alpha }\\
2116: &=\sum _{\alpha } (-1)^{d} (it)^{l (\alpha )}Q^{d/2}\left(\prod _{i=1}^{l (\alpha )}
2117: \frac{1}{1-Q^{\alpha _{i}}} \right)e_{\alpha }
2118: \end{align*}
2119: where $Q=e^{iu}$ as before. The expression
2120: \[
2121: \prod _{i=1}^{l (\alpha )}\frac{1}{1-Q^{\alpha _{i}}}
2122: \]
2123: arises in the theory of symmetric functions.
2124: The power sum symmetric functions are defined by:
2125: \begin{align*}
2126: p_{k} (x_{1},x_{2},x_3,\dots )&= x_{1}^{k}+x_{2}^{k}+x_3^k+\dots \\
2127: p_{\alpha }&=\prod _{i=1}^{l (\alpha )}p_{\alpha _{i}}.
2128: \end{align*}
2129: For the specialization
2130: $$x_{1}=1, \ x_{2}=Q, \ x_{3}=Q^2, \dots,$$
2131: we obtain
2132: $$p_{k} (Q)= (1-Q^{k})^{-1}.$$
2133: Hence,
2134: \[
2135: \eta =\sum _{\alpha } (-1)^{d} (it)^{l (\alpha )}Q^{d/2}p_{\alpha }
2136: (Q)\ e_{\alpha }
2137: \]
2138: and similarly
2139: \[
2140: \overline{\eta } =\sum _{\alpha } (-1)^{d} (it)^{l (\alpha
2141: )}Q^{d/2} (-1)^{l (\alpha )} p_{\alpha } (Q)\ e_{\alpha }.
2142: \]
2143:
2144:
2145:
2146: Inversion of \eqref{dfg} yields the following formula:
2147: \begin{equation}\label{kwqq3}
2148: e_{\alpha }=(it)^{d-l (\alpha )}\sum _{\rho }\frac{d!}{\dim \rho
2149: }\frac{\chi^{\rho }_\alpha}{\combinatfactor (\alpha )}v_{\rho }
2150: \end{equation}
2151: After substituting \eqref{kwqq3} in the expression for $\eta$, we find
2152: \begin{align*}
2153: \eta &=\sum _{\rho }v_{\rho} \left[ (-it)^{d}Q^{d/2}\frac{d!}{\dim
2154: \rho }\left(\sum _{\alpha }\frac{\chi^{\rho }_\alpha p_{\alpha }
2155: (Q)}{\combinatfactor (\alpha )} \right)\right],\\
2156: \overline{\eta } &=\sum _{\rho }v_{\rho} \left[ (+it)^{d}Q^{d/2}\frac{d!}{\dim
2157: \rho }\omega \left(\sum _{\alpha }\frac{\chi^{\rho }_\alpha p_{\alpha }
2158: (Q)}{\combinatfactor (\alpha )} \right)\right].
2159: \end{align*}
2160: Here, $\omega $ is the involution on the ring of symmetric
2161: functions defined by
2162: \[
2163: (-1)^{l (\alpha )}p_{\alpha }= (-1)^{d}\omega (p_{\alpha }).
2164: \]
2165:
2166:
2167: The sum over $\alpha $ in the
2168: latter expressions for $\eta$ and $\overline{\eta}$
2169: is equal to the Schur function $s_{\rho }
2170: (Q)$, see \cite{Macdonald} page 114.
2171: We have
2172: \[
2173: \omega (s_{\rho })=s_{\rho '}
2174: \]
2175: where $\rho '$ is the dual representation (or conjugate partition),
2176: \cite{Macdonald} page 42.
2177: Thus, we
2178: obtain
2179: \begin{align*}
2180: \eta _{\rho }&= (-it)^{d}Q^{d/2}\frac{d!}{\dim \rho }s_{\rho }
2181: (Q)\\
2182: \overline{\eta } _{\rho }&= (+it)^{d}Q^{d/2}\frac{d!}{\dim \rho }s_{\rho' }(Q).
2183: \end{align*}
2184: The Schur functions are easily expressed in terms of the
2185: $Q$-dimension. From \cite{Macdonald} page 45,
2186: \begin{align*}
2187: s_{\rho }&=Q^{n (\rho )}\prod _{\Box \in \rho
2188: }\frac{1}{1-Q^{h (\Box )}}\\
2189: &=Q^{n (\rho )-\frac{1}{2} (n (\rho )+n (\rho ')+d)} (-1)^{d}\prod
2190: _{\Box \in \rho }\left(Q^{h (\Box )/2}-Q^{-h (\Box )/2}
2191: \right)^{-1}\\
2192: &=Q^{-\frac{1}{2} (d+c_{\rho })}i^{d}\ \frac{\dim _{Q}\rho }{d!}.
2193: \end{align*}
2194: We have used \eqref{eqn: identities for total content and hook length} in the
2195: above formulas. We conclude
2196: \begin{align*}
2197: \eta _{\rho }&= (+t)^{d}\; Q^{-\frac{c_{\rho }}{2}}\; \frac{\dim _{Q}\rho }{\dim \rho },\\
2198: \overline{\eta } _{\rho }&= (-t)^{d}\; Q^{+\frac{c_{\rho }}{2}}\; \frac{\dim _{Q}\rho }{\dim \rho }.
2199: \end{align*}
2200: Theorem \ref{thm: s1=-s2 general formula} then follows directly from
2201: Theorem \ref{thm: structure theorem}.\qed
2202:
2203:
2204:
2205: \section{A degree 2 calculation}\label{sec: explicit formulas}
2206:
2207: The partition function $\mathsf{GW} (g\vertline k_{1},k_{2})$ in
2208: degree 2 is calculated here. The result was announced previously in
2209: \cite{Br-Pa-TQFT}.
2210:
2211: We abbreviate the level $(0,0)$ pair of pants by
2212: \[
2213: \mathsf{GW} (0\vertline 0,0)_{\lambda \mu \nu }=
2214: \mathsf{P}_{\lambda \mu \nu }
2215: \]
2216: and the Calabi-Yau cap by
2217: \[
2218: \mathsf{GW} (0\vertline -1,0)_{\lambda }=\mathsf{C}_{\lambda }.
2219: \]
2220: We apply the usual convention \eqref{eqn: index
2221: raising formula} for raising indices to
2222: $\mathsf{P}_{\lambda \mu \nu }$ and $\mathsf{C}_{\lambda }$.
2223:
2224: From the proof of Theorem~\ref{thm: structure theorem}, the partition
2225: function is
2226: \[
2227: \mathsf{GW}
2228: (g\vertline k_{1},k_{2})=
2229: \Tr \left(G^{g-1}A^{-k_{1}}\overline{A}^{-k_{2}} \right).
2230: \]
2231: The genus adding operator $G$ and the right annihilation operator $A$ can
2232: be computed in terms of
2233: $\mathsf{P}_{\lambda \mu \nu }$ and $\mathsf{C}_{\lambda }$ by the
2234: gluing formula.
2235: \begin{align}\label{eqn: A and G in terms of Z and C}
2236: G^{\mu }_{\nu }&=\sum _{\lambda ,\epsilon \vdash d}
2237: \mathsf{P}^{\mu }_{\lambda \epsilon }\mathsf{P}^{\lambda \epsilon }_{\nu },\\
2238: A^{\mu }_{\nu }&=\sum _{\lambda \vdash d}
2239: \mathsf{C}^{\lambda }\mathsf{P}^{\mu }_{\lambda \nu }.\nonumber
2240: \end{align}
2241: We obtain $\overline{A}$ from $A$ by switching $t_{1}$ and $t_{2}$.
2242:
2243: $\mathsf{P}_{\lambda \mu \nu }$ is determined recursively by Theorem~\ref{thm:
2244: reconstruction of level (0,0) pants} and
2245: $\mathsf{C}_{\lambda }$ is given explicitly
2246: by Lemma~\ref{lem: the CY cap}. We list their values for $d\leq 2$:
2247:
2248:
2249: \begin{equation*}
2250: \begin{array}{lll}
2251: \mathsf{C}_{\tableau{1}}=\frac{1}{t_{2}}\frac{1}{2\sin \frac{u}{2}},
2252: &\quad \mathsf{C}_{\tableau{1 1}}=
2253: \frac{1}{t_{2}^{2}}\frac{1}{2\left(2\sin \frac{u}{2} \right)^{2}}, &\quad
2254: \mathsf{C}_{\tableau{2}}=-\frac{1}{t_{2}}\frac{1}{4\sin u},\\ &&\\
2255: \mathsf{P}_{\tableau{1}\tableau{1}\tableau{1}}= \frac{1}{t_{1}t_{2}},
2256: &\quad
2257: \mathsf{P}_{\tableau{1 1}\tableau{1 1}\tableau{1 1}}=
2258: \frac{1}{2} \frac{1}{(t_{1}t_{2})^{2}},&\quad
2259: \mathsf{P}_{\tableau{1 1}\tableau{1 1}\tableau{2}}=0,\\ &&\\
2260: \mathsf{P}_{\tableau{1 1}\tableau{2}\tableau{2}}=
2261: \frac{1}{2} \frac{1}{t_{1}t_{2}}.&\quad
2262: \mathsf{P}_{\tableau{2}\tableau{2}\tableau{2}}=
2263: -\frac{1}{2} \frac{t_1+t_2}{t_{1}t_{2}}\tan\frac{u}{2}.&\quad
2264: \\
2265: \end{array}
2266: \end{equation*}
2267:
2268:
2269:
2270:
2271: \vspace{10pt}
2272:
2273:
2274: For $d=1$, we have
2275: \begin{align*}
2276: G^{\tableau{1}}_{\tableau{1}}&=
2277: \mathsf{P}^{\tableau{1}}_{\tableau{1}\tableau{1}}
2278: \mathsf{P}^{\tableau{1}\tableau{1}}_{\tableau{1}}=t_{1}t_{2}\\
2279: A^{\tableau{1}}_{\tableau{1}}&=
2280: \mathsf{C}_{\tableau{1}}
2281: \mathsf{P}_{\tableau{1}}^{\tableau{1}\tableau{1}}
2282: =t_{1} \left(2\sin \frac{u}{2} \right)^{-1}.
2283: \end{align*}
2284: Hence,
2285: \[
2286: \mathsf{GW}_{1} (g\vertline k_{1},k_{2})
2287: = (t_{1}t_{2})^{g-1}t_{1}^{-k_{1}}t_{2}^{-k_{2}}\left(2\sin \frac{u}{2} \right)^{k_{1}+k_{2}}.
2288: \]
2289:
2290: \vspace{10pt}
2291:
2292:
2293: For $d=2$, we compute the entries of $G$ and $A$ via \eqref{eqn: A
2294: and G in terms of Z and C} to obtain:
2295:
2296:
2297:
2298: \begin{align*}
2299: G&=
2300: \left(
2301: \begin{smallmatrix}
2302: 4 (t_{1}t_{2})^{2}&\quad \quad &-2 (t_{1}t_{2})^{2}
2303: (t_{1}+t_{2})\tan\frac{u}{2}\\&\quad \quad &\\
2304: -2 (t_{1}t_{2})(t_{1}+t_{2})\tan\frac{u}{2} &
2305: &4 (t_{1}t_{2})^{2}+2 (t_{1}t_{2}) (t_{1}+t_{2})^{2}\tan^{2}\frac{u}{2}
2306: \end{smallmatrix}
2307: \right),\\
2308: &\\
2309: A&=\left(
2310: \begin{smallmatrix}
2311: t_{1}^{2}\left(2\sin \frac{u}{2} \right)^{-2}&&-t_{1}^{2}
2312: t_{2}\left(2\sin u \right)^{-1}\\ &&\\
2313: -t_{1}\left(2\sin u \right)^{-1}
2314: &&t_{1} (t_{1}+t_{2})\left(2\cos \frac{u}{2} \right)^{-2}+
2315: t_{1}^{2}\left(2 \sin \frac{u}{2} \right)^{-2}
2316: \end{smallmatrix}
2317: \right).
2318: \end{align*}
2319: The matrices $G$, $A$, and $\overline{A}$ mutually commute and so we can
2320: simultaneously diagonalize them to obtain:
2321: \begin{equation}\label{eqn: d=2 full formula}
2322: \mathsf{GW}_{2} (g\vertline k_{1},k_{2})=
2323: \lambda _{+}^{g-1}\eta _{+}^{-k_{1}}\overline{\eta }_{+}^{-k_{2}}+\lambda _{-}^{g-1}\eta _{-}^{-k_{1}}\overline{\eta }_{-}^{-k_{2}},
2324: \end{equation}
2325: where
2326: \begin{align*}
2327: \lambda _{\pm}&=\frac{t_{1}t_{2}}{(1-q)^{2}}
2328: \left(-\Theta \pm (1+q) (t_{1}+t_{2})\sqrt{\Theta } \right)\\
2329: \eta _{\pm }&=\frac{qt_{1}}{2 (1-q^{2})^{2}}\left((t_{1}-t_{2})
2330: (1+q)^{2}-8t_{1}q\pm (1+q)\sqrt{\Theta } \right)\\
2331: \overline{\eta } _{\pm }&=\frac{qt_{2}}{2 (1-q^{2})^{2}}
2332: \left((t_{2}-t_{1}) (1+q)^{2}-8t_{2}q\pm (1+q)\sqrt{\Theta } \right)\\
2333: \Theta &= (t_{1}-t_{2})^{2} (1+q)^{2}+16qt_{1}t_{2}\\
2334: q&=-e^{iu}.
2335: \end{align*}
2336: For the specialization
2337: \[
2338: t_{1}=t_{2}=t,
2339: \]
2340: the above equations simplify to
2341: \begin{align*}
2342: \lambda _{\pm }&=\frac{4t^{4}}{1\mp \sin \frac{u}{2}}\\
2343: \eta _{\pm }=\overline{\eta }_{\pm }&=\frac{t^{2}}
2344: {4\sin ^{2}\frac{u}{2}\left(1\mp \sin \frac{u}{2} \right)}.
2345: \end{align*}
2346: Hence,
2347: \begin{align*}
2348: \mathsf{GW}_{2} (g\vertline k_{1},k_{2})\Big|_{t_{1}=t_{2}=t}
2349: =&t^{2 (2g-2-k_{1}-k_{2})}4^{g-1}
2350: \left(2\sin \frac{u}{2} \right)^{2 (k_{1}+k_{2})} \\
2351: &\cdot \left\{\left(1+\sin \frac{u}{2} \right)^{k_{1}+k_{2}+1-g}
2352: +\left(1-\sin \frac{u}{2} \right)^{k_{1}+k_{2}+1-g} \right\}.
2353: \end{align*}
2354: In particular, the local $d=2$ Calabi-Yau partition function
2355: % \cite{Br-Pa-TQFT}
2356: is given by:
2357: \begin{multline*}
2358: \mathsf{GW}_{2} (g\vertline g-1,g-1)\Big|_{t_1=t_2=t}=\\
2359: \left(2\sin \frac{u}{2} \right)^{4g-4}\left\{\left(4-4\sin
2360: \frac{u}{2} \right)^{g-1}+ \left(4+4\sin \frac{u}{2} \right)^{g-1} \right\},
2361: \end{multline*}
2362: in agreement with the announcement of \cite{Br-Pa-TQFT} up to
2363: $u$ shifting conventions.
2364:
2365: The partition function in degree 2
2366: is easily seen to satisfy the BPS integrality of
2367: Gopakumar-Vafa.
2368:
2369:
2370:
2371: \section{The GW/DT correspondence for residues}\label{sec: GW/DT
2372: correspondence for residues}
2373:
2374:
2375: \subsection{Overview} A
2376: Gromov-Witten/Donaldson-Thomas correspondence parallel to
2377: \cite{MNOP1,MNOP2} is conjectured here for equivariant residues in both the
2378: absolute and relative cases. The computation of the full local
2379: Gromov-Witten theory of curves together with the GW/DT correspondence
2380: predicts the full local Donaldson-Thomas theory of curves.
2381:
2382: \subsection{Residue Invariants in Donaldson-Thomas theory}
2383:
2384:
2385: Let $Y$ be a nonsingular, {\em quasi-projective}, algebraic threefold.
2386: Let $I_{n} (Y,\beta )$ denote the moduli space of ideal sheaves
2387: \[
2388: 0 \rightarrow I_{Z}\rightarrow \O _{Y}\rightarrow \O_Z \rightarrow 0
2389: \]
2390: of subschemes $Z$ of degree $\beta =[Z]\in H_{2}
2391: (Y,\znums )$ and Euler characteristic $n=\chi (\O _{Z})$.
2392: Though $Y$ may not be compact, we require $Z$ to have \emph{proper} support.
2393:
2394: Let $Y$ be equipped with an action by an algebraic torus $T$.
2395: The moduli space $I_n(Y,\beta)$ carries a $T$-equivariant perfect obstruction
2396: theory obtained from (traceless) $\text{Ext}_0(I,I)$, see
2397: \cite{t}. Though $Y$ is
2398: quasi-projective,
2399: $\text{Ext}_0(I,I)$ is well-behaved since the associated
2400: quotient scheme $Z\subset Y$ is proper. Alternatively, for any
2401: $T$-equivariant compactification,
2402: $$Y\subset \overline{Y},$$
2403: the obstruction theory on
2404: $$I_n(Y,\beta)\subset I_n(\overline{Y},\beta)$$
2405: is obtained by
2406: restriction.
2407:
2408:
2409:
2410:
2411: We will define
2412: Donaldson-Thomas residue invariants under the following assumption.
2413: \begin{assumption}\label{assuption: T fixed locus of DT space is compact}
2414: The $T$-fixed point set $I_{n} (Y,\beta )^{T}$ is compact.
2415: \end{assumption}
2416:
2417: The definition of the Donaldson-Thomas residue invariants of $Y$ follows
2418: the
2419: strategy of the Gromov-Witten case. We define ${\mathsf Z}_{DT}(Y)_{\beta
2420: } $ formally by:
2421: \begin{equation}\label{eqn: moral defn of Zdt}
2422: {\mathsf Z}_{DT}(Y)_{\beta }\text{ ``}=\text{'' }\sum _{n\in \znums
2423: }q^{n} \int _{[I_{n} (Y,\beta ) ]^{vir}}1 .
2424: \end{equation}
2425: The variable $q$ indexes the Euler number $n$.
2426: Under Assumption~\ref{assuption: T fixed locus of DT space is compact},
2427: the integral on
2428: the right of \eqref{eqn: moral defn of Zdt} is well-defined
2429: by the virtual localization formula as an equivariant residue.
2430:
2431:
2432:
2433:
2434: \begin{definition} The partition function for
2435: the degree $\beta $ Donaldson-Thomas residue invariants of $Y$ is defined by:
2436: \begin{equation}
2437: \label{defn: definition of Zdt}
2438: {\mathsf Z}_{DT}(Y)_{\beta } =\sum _{n\in \znums }q^{n}
2439: \int _{[I _{n} (Y,\beta )^{T}]^{vir}}\frac{1}{e
2440: (\operatorname{Norm}^{vir})}.
2441: \end{equation}
2442: \end{definition}
2443:
2444: The $T$-fixed
2445: part of the perfect obstruction theory for $I_n(Y,\beta )$ induces a
2446: perfect obstruction theory for $I_n (Y,\beta )^{T}$ and hence a virtual
2447: class \cite{Gr-Pa,MNOP1}.
2448: The
2449: equivariant virtual normal bundle
2450: of the embedding, $$I_n (Y,\beta )^{T}\subset I_n (Y,\beta ), $$ is
2451: $\operatorname{Norm}^{vir}$ with
2452: equivariant Euler class
2453: $e(\operatorname{Norm}^{vir})$. The integral in \eqref{defn: definition of Zdt}
2454: denotes equivariant push-forward to a point.
2455:
2456:
2457:
2458:
2459:
2460:
2461: As defined, ${\mathsf Z}_{DT}(Y)_{\beta }$ is {\em unprimed} since
2462: the degree $0$ contributions have not yet been removed. In
2463: Gromov-Witten theory, the degree 0 contributions are removed geometrically by
2464: forbidding such components in the moduli problem. Since a geometrical
2465: method of removing the degree 0 contribution from
2466: Donaldson-Thomas theory does not appear to be available, a formal
2467: method is followed.
2468:
2469: \begin{definition}\label{defn: reduced DT partition function Zdt'}
2470: The reduced partition function ${\mathsf Z}'_{DT}(Y)_{\beta }$ for the degree $\beta
2471: $ Donaldson-Thomas residue invariants of $Y$ is defined by:
2472: \begin{equation*}
2473: {\mathsf Z}'_{DT}(Y)_{\beta } = \frac{{\mathsf Z}_{DT}(Y)_{\beta } }
2474: {{\mathsf Z}_{DT}(Y)_0}.
2475: \end{equation*}
2476: \end{definition}
2477:
2478:
2479: Let $r$ be the rank of $T$, and let $t_1,\dots, t_r$ be generators of the
2480: equivariant cohomology of $T$. By definition, $ {\mathsf
2481: Z}'_{DT}(Y)_{\beta } $ is a Laurent series in $q$ with coefficients given
2482: by rational functions of $t_1,\ldots, t_r$ of homogeneous degree equal to
2483: minus the virtual dimension of $I_n(Y,\beta)$.
2484:
2485:
2486:
2487: \subsection{Conjectures for the absolute theory}
2488:
2489:
2490: The equivariant degree 0 series is conjecturally determined
2491: in terms of the MacMahon function,
2492: $$M(q)=\prod_{n\geq 1} \frac{1}{(1-q^n)^n},$$
2493: the generating series for 3-dimensional partitions.
2494:
2495:
2496: \vspace{+10pt}
2497: \noindent {\bf Conjecture 1.} The degree 0 Donaldson-Thomas partition function
2498: is determined by:
2499: \begin{equation*}
2500: {\mathsf Z}_{DT} (Y)_0=M (-q)^{\int _{Y}c_{3} (T_Y\otimes K_Y)},
2501: \end{equation*}
2502: where the integral in the exponent is defined via localization on $Y$,
2503: \[
2504: \int _{Y}c_{3} (T_Y\otimes K_Y) =\int _{Y^{T}}\frac{c_{3}(T_Y\otimes K_Y) }{e (N_{Y^{T}/Y})} \in
2505: {\mathbb Q}(t_1,\dots, t_r).
2506: \]
2507: \vspace{+10pt}
2508:
2509: The subvariety $Y^{T} $ is compact as a consequence of
2510: Assumption~\ref{assuption: T fixed locus of DT space is compact}. By
2511: Theorem 1 of \cite{MNOP2}, Conjecture 1 holds for toric $Y$.
2512:
2513: \vspace{+10pt}
2514: \noindent {\bf Conjecture 2.} The reduced series ${\mathsf Z}'_{DT}(Y)_{\beta }$
2515: is a rational function of the equivariant parameters $t_i$ and $q$.
2516: \vspace{+10pt}
2517:
2518:
2519: The GW/DT correspondence for absolute residue invariants can now be stated.
2520:
2521: \vspace{+10pt}
2522: \noindent {\bf Conjecture 3.}
2523: After the change of variables $e^{iu}=-q$,
2524: \[
2525: (-iu)^{\int_\beta c_1(T_Y)} \ {\mathsf Z}'_{GW}(Y)_{\beta } =(-q)^{-\frac{1}{2}\int_\beta c_1(T_Y)}
2526: \ {\mathsf Z}'_{DT}(Y)_{\beta}.
2527: \]
2528: \vspace{+10pt}
2529:
2530: Conjectures 1-3 are equivariant versions of the conjectures of \cite{MNOP1,MNOP2}.
2531: In \cite{MNOP2}, a GW/DT correspondence for primary and (certain) descendent
2532: field insertions is presented. The equivariant correspondence with insertions
2533: remains to be studied.
2534:
2535:
2536: \subsection{The relative conjectures}
2537: A Gromov-Witten/Donaldson-Thomas residue correspondence for relative theories
2538: may also be defined.
2539: Let
2540: $$S\subset Y$$
2541: be a nonsingular, $T$-invariant, divisor.
2542: Let
2543: $\beta\in H_2(Y,{\mathbb Z})$
2544: be a curve class
2545: satisfying
2546: $$\int_\beta[S] \geq 0.$$
2547: Let $\eta$ be a partition of $\int_\beta[S]$ weighted
2548: by the equivariant cohomology of $S$,
2549: $$H_T^*(S,{\mathbb Q}).$$ We follow here
2550: the notation of \cite{MNOP2}.
2551: The reduced Gromov-Witten partition function,
2552: \begin{equation*}
2553: {\mathsf Z}'_{GW}(Y/S)_{\beta,\eta} =\sum _{h\in \znums }u^{2h-2}\int
2554: _{[\M (Y/S,\beta,\eta )^{T}]^{vir}}\frac{1}{e (\operatorname{Norm}^{vir})},
2555: \end{equation*}
2556: is well-defined if $\M (Y/S,\beta,\eta )^{T}$ is compact.
2557: The weighted partition $\eta$ specifies the relative
2558: conditions imposed on the moduli space of maps.
2559:
2560:
2561: We refer the reader to \cite{MNOP2} for a discussion of relative
2562: Donaldson-Thomas theory.
2563: The relative Donaldson-Thomas partition function,
2564: \begin{equation*}
2565: {\mathsf Z}_{DT}(Y/S)_{\beta,\eta } =\sum _{n\in \znums }q^{n}
2566: \int _{[I _{n} (Y/S,\beta,\eta )^{T}]^{vir}}\frac{1}{e
2567: (\operatorname{Norm}^{vir})},
2568: \end{equation*}
2569: is well-defined if $I _{n} (Y/S,\beta,\eta )^{T}$ is compact.
2570: Let
2571: \begin{equation*}
2572: {\mathsf Z}'_{DT}(Y/S)_{\beta,\eta } = \frac{{\mathsf Z}_{DT}(Y/S)_{\beta,\eta } }
2573: {{\mathsf Z}_{DT}(Y/S)_0}.
2574: \end{equation*}
2575: denote the reduced relative partition function.
2576:
2577: The weighted partition $\eta$ in Donaldson-Thomas theory specifies the relative
2578: conditions imposed on the moduli space of ideal sheaves. The partition
2579: $\eta$
2580: determines an element of the
2581: Nakajima basis of
2582: the $T$-equivariant cohomology of the Hilbert scheme of
2583: points of $S$.
2584:
2585: \vspace{+10pt}
2586: \noindent {\bf Conjecture 1R.} The degree 0 relative Donaldson-Thomas partition function
2587: is determined by:
2588: \begin{equation*}
2589: {\mathsf Z}_{DT} (Y/S)_0=M (-q)^{\int _{Y}c_{3} (T_Y[-S]\otimes K_Y[S])},
2590: \end{equation*}
2591: where $T_Y$ is the sheaf of tangent fields with logarithmic zeros and
2592: $K_Y$ is the logarithmic canonical bundle.
2593: \vspace{+10pt}
2594:
2595:
2596: \vspace{+10pt}
2597: \noindent
2598: {\bf Conjecture 2R.} The reduced series ${\mathsf
2599: Z}'_{DT}(Y/S)_{\beta,\eta }$ is a rational function of the equivariant
2600: parameters $t_i$ and $q$.
2601: \vspace{+10pt}
2602:
2603:
2604:
2605: \noindent {\bf Conjecture 3R.}
2606: After the change of variables $e^{iu}=-q$,
2607: \[
2608: (-iu)^{\int_\beta c_1(T_Y)+l(\eta)-|\eta|}
2609: \ {\mathsf Z}'_{GW}(Y/S)_{\beta,\eta } =(-q)^{-\frac{1}{2}\int_\beta c_1(T_Y)}
2610: \ {\mathsf Z}'_{DT}(Y/S)_{\beta,\eta},
2611: \]
2612: where $|\eta|=\int_\beta[S]$.
2613: \vspace{+10pt}
2614:
2615: Conjectures 1R-3R are equivariant versions of the relative conjectures of
2616: \cite{MNOP2} without insertions.
2617:
2618: \subsection{The local theory of curves}
2619: \label{subsec: GW/DT for local curves}
2620: Let $X$ be a nonsingular curve of genus $g$.
2621: Let $N$ be a rank 2 bundle on a $X$ with a direct sum decomposition,
2622: \[
2623: N=L_{1}\oplus L_{2}.
2624: \]
2625: Let $k_i$ denote the degree of $L_i$ on $X$.
2626:
2627: Consider the Gromov-Witten residue theory of $N$
2628: relative to the
2629: $T$-invariant divisor
2630: $$S= \bigcup_{p\in D} N_{p}\ \subset N,$$
2631: where $D \subset X$ is a finite set of points. Since
2632: $$H_T^*(S) = \bigoplus_{p\in D} H^*_T(p),$$
2633: $\eta$ is simply a list of partitions indexed by $D$.
2634:
2635:
2636: \begin{theorem} \label{ko34}
2637: The GW/DT correspondence holds (Conjectures 1R-3R)
2638: for the local theory of curves.
2639: \end{theorem}
2640:
2641: Theorem \ref{ko34} is proven by matching the calculation of the
2642: local Gromov-Witten theory of curves here with the determination
2643: of the local Donaldson-Thomas theory of curves in \cite{dtlc}.
2644: The results of \cite{dtlc} depend upon foundational aspects
2645: of relative Donaldson-Thomas theory which have not yet been
2646: treated in the literature.
2647:
2648:
2649: The $\mathsf{GW}^*$-partition functions in relative Gromov-Witten
2650: theory are defined in Section \ref{n3451}. Similarly, the
2651: $\mathsf{DT}^*$-partition functions in relative Donaldson-Thomas
2652: theory are defined by:
2653: \[
2654: \mathsf{DT}^*(g\vertline k_1,k_2)_{\lambda^1\dots\lambda^r} =
2655: (-q)^{-\frac{d}{2}(2-2g+k_1+k_2)} \ {\mathsf Z}'_{DT}(N)_{d[X],
2656: \lambda^1 \dots \lambda^r}.
2657: \]
2658: The GW/DT correspondence for local curves can be
2659: conveniently restated as the equality,
2660: \begin{equation}\label{kk231}
2661: \mathsf{GW}^*(g\vertline k_1,k_2)_{\lambda^1\dots\lambda^r} =
2662: \mathsf{DT}^*(g\vertline k_1,k_2)_{\lambda^1\dots\lambda^r},
2663: \end{equation}
2664: after the variable change $e^{iu}=-q$,
2665:
2666:
2667: \begin{comment}
2668:
2669:
2670:
2671: \subsection{An example}
2672: We conclude with
2673: an example of the GW/DT correspondence for the local theory of curves.
2674: By \eqref{fferr},
2675: \begin{eqnarray*}
2676: \mathsf{GW}^* (0\vertline 0,0)_{(2), (2),
2677: (2)} & = &
2678: \frac{1}{2} \frac{t_{1}+t_{2}}{t_{1}t_{2}} \left(
2679: 2\frac{(-q)^2+1}{(-q)^2-1} - \frac{(-q)+1}{(-q)-1}\right).\\
2680: & = &
2681: \frac{1}{2} \frac{t_{1}+t_{2}}{t_{1}t_{2}} \left( -1 - 2q + \ldots \right).
2682: \end{eqnarray*}
2683: Hence, on the Donaldson-Thomas side, we expect
2684: \begin{equation}\label{lastt}
2685: \ {\mathsf Z}'_{DT}(0\vertline 0,0)_{2[X], (2),(2),(2)} = \frac{1}{2}
2686: \frac{t_{1}+t_{2}}{t_{1}t_{2}}
2687: \left(-q^2 - 2 q^3 + \ldots \right).
2688: \end{equation}
2689:
2690: We will verify the first coefficient of the Donaldson-Thomas series \eqref{lastt}.
2691: Let $N$ be trivial rank 2 bundle over $\P^1$,
2692: $$N=\cnums^2 \times \P^1.$$
2693: Let $S\subset N$ be the relative divisor determined by the points $p_1,p_2,p_3\in\P^1$.
2694: The moduli spaces $I_n(N/S,2[\P^1])$ are empty for $n<2$.
2695: Hence, the terms of ${\mathsf Z}'_{DT}(0\vertline 0,0)_{2[X], (2),(2),(2)}$
2696: of order less than $q^2$ vanish.
2697:
2698:
2699: The
2700: moduli space
2701: $I_2(N/S,2[\P^1])$ is isomorphic to the Hilbert scheme of 2 points of
2702: $\cnums^2$,
2703: \begin{equation}
2704: \label{kf56}
2705: I_2(N/S,2[\P^1]) \stackrel {\sim}{=} \Hilb (\cnums^2,2).
2706: \end{equation}
2707: Since $I_2(N/S,2[\P^1])$ is nonsingular of
2708: expected dimension 4, the virtual class reduces to the ordinary
2709: fundamental class of the moduli space.
2710:
2711: Under the isomorphism \eqref{kf56},
2712: the $T$-fixed locus of the moduli space consists of the two
2713: fixed double points at the origin of $\cnums^2$,
2714: $$I_2(N/S,2[\P^1])^T= \{f_1,f_2\}.$$
2715: The equivariant Euler classes of the normal bundles,
2716: $$e(\text{Norm}_{f_1})= t_1 t_2 (t_1-t_2) (2t_2),$$
2717: $$e(\text{Norm}_{f_2})= t_1 t_2 (t_2-t_1) (2t_1),$$
2718: are simply the tangent weights to the Hilbert scheme of points.
2719:
2720:
2721:
2722: We now consider the relative conditions on the moduli space of ideal sheaves
2723: $I_2(N/S,2[\P^1])$.
2724: Let
2725: \[
2726: \epsilon: I_2(N/S,2[\P^1]) \rightarrow \prod_{i=1}^3 \Hilb(\cnums^2_{p_i},2)
2727: \]
2728: be the intersection map to the Hilbert scheme of points of $S$,
2729: see\cite{MNOP2}. Let
2730: \[
2731: H_i \subset \Hilb(\cnums_{p_i}^2,2)
2732: \]
2733: denote the locus of double points.
2734: The Nakajima basis element corresponding to the relative condition $(2)$ over $p_i$
2735: is $[H_i]/2$ where the $H_i$ denote the double point locus.
2736:
2737: Let $H$ denote the double point locus in $\Hilb (\cnums^2,2)$.
2738: By definition, the $q^2$ coefficient of
2739: ${\mathsf Z}'_{DT}(0\vertline 0,0)_{2[X], (2),(2),(2)}$ is:
2740: \begin{multline*}
2741: \int_{I_2(N/S,2[\P^1])^T} \frac{\epsilon^*( \frac{[H_1]}{2} \cup
2742: \frac{[H_2]}{2} \cup \frac{[H_3]}{2})}{e(\text{Norm})} = \\
2743: \frac{[H]_{f_1}^3}{8 t_1 t_2 (t_1-t_2)(2t_2)} +
2744: \frac{[H]_{f_2}^3}{8 t_1 t_2 (t_2-t_1)(2t_1)}.
2745: \end{multline*}
2746: Using the restriction formulas
2747: $$[H]_{f_1}= 2t_2, \ \ [H]_{f_2}=2t_1,$$
2748: we conclude that the $q^2$ coefficient equals
2749: $$-\frac{1}{2} \frac{t_{1}+t_{2}}{t_{1}t_{2}}$$
2750: in agreement with \eqref{lastt}.
2751:
2752: \end{comment}
2753:
2754: \section{Further directions}\label{sec: further directions}
2755: \subsection{The Hilbert scheme $\Hilb^n(\cnums^2)$}
2756: %The level $(0,0)$ theory of $\P^1$ is proven in \cite{Ok-Pan-Hilb}
2757: %to {\em coincide} with the 3-point functions of the equivariant
2758: %Gromov-Witten theory of the Hilbert scheme $\Hilb^{n} (\cnums ^{2})$
2759: %of $n$ points in $\cnums^2$:
2760:
2761:
2762: Let the 2-dimensional torus $T$ act on $\cnums^2$ by scaling the factors.
2763: Consider the induced $T$-action on $\Hilb^n(\cnums^2)$.
2764: The $T$-equivariant cohomology of $\Hilb^n(\cnums^2)$,
2765: $$H^*_T(\Hilb ^{n} (\cnums ^{2}),\qnums),$$
2766: has
2767: a canonical Nakajima basis,
2768: $$\{ \ |\mu\rangle\ \}_{|\mu|=n}$$
2769: indexed by partitions of $n$.
2770: The degree of a curve in $\Hilb^n(\cnums^2)$ is determined
2771: by intersection with the divisor
2772: $$D = - |2,1^{n-2} \rangle.$$
2773:
2774: Define the series $\langle \lambda,\mu,\nu\rangle^{{\Hilb^n (\cnums^2)}}$ of
2775: 3-pointed, genus 0, $T$-equivariant Gromov-Witten invariants by a sum over curve degrees:
2776: $$\langle \lambda,\mu,\nu
2777: \rangle^{\Hilb^n(\cnums^2)} = \sum_{d\geq 0} q^d
2778: \langle \lambda,\mu,\nu
2779: \rangle^{\Hilb^n(\cnums^2)}_{0,3,d}.$$
2780:
2781: \begin{theorem}\label{gg345} A Gromov-Witten/Hilbert correspondence holds:
2782: \begin{equation*}
2783: \mathsf{GW}^*(0\vertline 0,0)_{\lambda\mu\nu}
2784: = (-1)^n\langle \lambda,\mu,\nu
2785: \rangle^{\Hilb^n(\cnums^2)},
2786: \end{equation*}
2787: after the variable change $e^{iu}=-q$.
2788: \end{theorem}
2789: The proof of Theorem \ref{gg345} is obtained by our
2790: determination of the series $\mathsf{GW}^*(0\vertline 0,0)_{\lambda\mu\nu}$
2791: together with the computation of the quantum cohomology of
2792: the Hilbert scheme in \cite{Ok-Pan-Hilb}.
2793:
2794:
2795:
2796: \subsection{The orbifold $\Sym (\cnums ^{2}) $} The 3-pointed, genus
2797: 0, $T$-equivariant Gromov-Witten invariants of the orbifold $\Sym
2798: (\cnums^2)=(\cnums^2)^n/S_n$ can be related to
2799: $\mathsf{GW}^*(0\vertline
2800: 0,0)_{\lambda\mu\nu}$, see \cite{Bryan-Graber}.
2801:
2802:
2803: The Hilbert scheme $\Hilb^n(\cnums^2)$ is a crepant resolution of the
2804: (singular) quotient variety $\Sym (\cnums ^{2})$. Theorem
2805: \ref{gg345} may be viewed as relating the
2806: $T$-equivariant quantum cohomology of the quotient {\em orbifold}
2807: $\Sym (\cnums^2)$ to the $T$-equivariant quantum cohomology of the
2808: resolution $\Hilb^n(\cnums^2)$. The correspondence requires extending
2809: the definition of orbifold quantum cohomology to include quantum
2810: parameters for twisted sectors \cite{Bryan-Graber}.
2811:
2812: Mathematical conjectures relating the quantum cohomologies of
2813: orbifolds and their crepant resolutions in the
2814: non-equivariant case have been pursued by
2815: Ruan (motivated by the physical predictions of Vafa and Zaslow).
2816: Theorem \ref{gg345} suggests that the correspondence also holds
2817: in the equivariant context.
2818:
2819:
2820:
2821: \vspace{1.5in}
2822: \begin{center}
2823: \scriptsize
2824: \begin{picture}(200,75)(-30,-50)
2825: \put(-100 ,-115 ){\line(1 ,0 ){340}}
2826: \put(-100 ,-115 ){\line(0 ,1 ){225}}
2827: \put(-100 ,110 ){\line(1 ,0 ){340}}
2828: \put(240 ,-115 ){\line(0 ,1 ){225}}
2829: \thicklines
2830: \put(25,25){\line(1,1){50}}
2831: \put(25,25){\line(1,-1){50}}
2832: \put(125,25){\line(-1,1){50}}
2833: \put(125,25){\line(-1,-1){50}}
2834: \put(75,-25){\line(0,1){100}}
2835: \put(25,25){\line(1,0){45}}
2836: \put(80,25){\line(1,0){45}}
2837: \put(75,95){\makebox(0,0){Equivariant quantum}}
2838: \put(75,85){\makebox(0,0){cohomology of $\Hilb (\cnums ^{2})$ }}
2839: \put(75,-35){\makebox(0,0){Equivariant orbifold quantum }}
2840: \put(75,-45){\makebox(0,0){cohomology of $\operatorname{Sym} (\cnums ^{2})$ }}
2841: \put(160,35){\makebox(0,0){Equivariant}}
2842: \put(160,25){\makebox(0,0){Gromov-Witten}}
2843: \put(160,15){\makebox(0,0){theory of $\P ^{1}\times \cnums ^{2}$}}
2844: \put(-15,35){\makebox(0,0){Equivariant}}
2845: \put(-15,25){\makebox(0,0){Donaldson-Thomas}}
2846: \put(-15,15){\makebox(0,0){theory of $\P ^{1}\times \cnums ^{2}$}}
2847: \end{picture}
2848: \end{center}
2849: \begin{quote}
2850: \scriptsize{ The four isomorphic theories described by
2851: $\mathsf{GW}^*(0\vertline 0,0)_{\lambda\mu\nu}$. The southern and
2852: eastern theories have parameter $u$, while the northern and western
2853: have parameter $q=-e^{iu}$. The horizontal equivalence is the
2854: equivariant DT/GW correspondence for $\P ^{1} \times \cnums ^{2}$. The
2855: vertical equivalence is the equivariant Crepant Resolution Conjecture
2856: for $\Hilb \cnums ^{2}\to \operatorname{Sym}\cnums ^{2}$. The four
2857: corners are computed in \cite{Bryan-Graber,dtlc,Ok-Pan-Hilb} and the
2858: present paper. }
2859: \end{quote}
2860:
2861:
2862:
2863:
2864:
2865: \normalsize
2866:
2867:
2868: \pagebreak
2869:
2870:
2871: \appendix
2872: %\section{\hspace{-26pt} ppendix: Reconstruction Result}\label{appendix}
2873: \section{Appendix: Reconstruction Result}\label{appendix}
2874:
2875: %\begin{center} { \large Appendix: Reconstruction Result} \end{center}
2876:
2877:
2878: %\begin{center} {By \large J. Bryan, C. Faber, A. Okounkov, and R. Pandharipande} \end{center}
2879:
2880: {By \large J. Bryan, C. Faber, A. Okounkov, and R. Pandharipande}
2881:
2882:
2883: %\setcounter{section}{0}
2884:
2885:
2886: %\vspace{15pt}
2887: %\noindent {\bf A.1 Overview}
2888: %\vspace{15pt}
2889:
2890: \subsection{Overview}
2891:
2892: We present a proof of Theorem~\ref{thm: reconstruction of level (0,0)
2893: pants} using a closed formula for the series
2894: $$\mathsf{GW}^*(0\vertline 0,0)_{\lambda,(2),\nu}$$
2895: obtained from Theorem \ref{thm: the dd2 integral}
2896: and the semisimplicity of the Frobenius algebra associated to the
2897: level $(0,0)$ theory. The proof was
2898: motivated by the study of the quantum cohomology of $\Hilb ^{n} (\cnums
2899: ^{2})$ in \cite{Ok-Pan-Hilb}.
2900:
2901:
2902: \subsection{Fock space}
2903:
2904: %\vspace{15pt}
2905: %\noindent {\bf A.2 Fock space}
2906: %\vspace{15pt}
2907:
2908: By definition, the Fock space $\cF$
2909: is freely generated over $\Q$ by commuting
2910: creation operators $$\alpha_{-k},\ \ k\in\Z_{>0},$$
2911: acting on the vacuum vector $\vac$. The annihilation
2912: operators
2913: $$\alpha_{k},\ \ k\in\Z_{>0},$$ kill the vacuum
2914: $$
2915: \alpha_k \cdot \vac =0,\quad k>0 \,,
2916: $$
2917: and satisfy the commutation relations
2918: $$
2919: \left[\alpha_k,\alpha_l\right] = k \, \delta_{k+l,0}\,.
2920: $$
2921:
2922:
2923: A natural basis of $\cF$ is given by
2924: the vectors
2925: %
2926: \begin{equation}
2927: \label{basis}
2928: \lv \mu \rang = \frac{1}{\zz(\mu)} \, \prod_{i=1}^{l (\mu )} \alpha_{-\mu_i} \, \vac \,.
2929: \end{equation}
2930: %
2931: indexed by partitions
2932: $\mu$.
2933: After extending scalars to $\Q(t_1,t_2)$,
2934: we define the following
2935: {\em nonstandard} inner product on $\cF$:
2936: %
2937: \begin{equation}
2938: \label{inner_prod}
2939: \lang \mu | \nu \rang =
2940: \frac{(-1)^{|\mu|-\ell(\mu)}}{(t_1 t_2)^{l(\mu)}}
2941: \frac{\delta_{\mu\nu}}{\zz(\mu)} \,.
2942: \end{equation}
2943:
2944:
2945: \subsection{The matrix $\MM _{2}$}
2946: %\vspace{15pt}
2947: %\noindent {\bf A.3 The matrix $\MM_2$}
2948: %\vspace{15pt}
2949:
2950: Define the linear transformation $\MM_2$ on $\cF$ by
2951: \[
2952: \lang \mu | \MM_2 | \nu \rang = (-1)^{|\mu|}
2953: \mathsf{GW}^*(0\vertline 0,0)_{\mu, (2), \nu}\ \delta_{|\mu|,|\nu|},
2954: \]
2955: after an extension of scalars to $\Q(t_1,t_2)[[u]]$.
2956:
2957:
2958: The matrix $\MM_2$ can be written in closed form in terms of creation and
2959: annihilation operators on Fock space:
2960: \begin{multline}
2961: \label{vv22}
2962: -\MM_2 = \frac{t_1+t_2}{2} \sum_{k>0} \left( {k} \frac{(-q)^k+1}{(-q)^k-1} -\frac{(-q)+1}{(-q)-1}\right) \, \alpha_{-k} \, \alpha_k + \\
2963: \frac12 \sum_{k,l>0}
2964: \Big[t_1 t_2 \, \alpha_{k+l} \, \alpha_{-k} \, \alpha_{-l} -
2965: \alpha_{-k-l}\, \alpha_{k} \, \alpha_{l} \Big] \,,
2966: \end{multline}
2967: where $-q=e^{iu}$.
2968: The above formula was studied in \cite{Ok-Pan-Hilb} as
2969: the matrix of quantum multiplication by the hyperplane class
2970: in the quantum cohomology of $\Hilb^n(\cnums^2)$.
2971:
2972: Formula~\eqref{vv22} can be obtained as follows. Using dimension counts
2973: similar to those in subsection~\ref{subsec: level 0 tube and cap}, the
2974: disconnected invariants $\mathsf{GW}^* (0\vertline 0,0)_{\mu, (2),\nu }$
2975: are easily reduced to connected invariants of one of two
2976: possible types. First, there are the (necessarily connected) invariants
2977: $\mathsf{GW}^*(0\vertline 0,0)_{(d) ,(2), (d)}$ (computed in
2978: Theorem~\ref{thm: the dd2 integral}), and second there are domain genus 0
2979: Hurwitz numbers. The combinatorics of writing disconnected invariants in
2980: terms of connected invariants is most efficiently handled with the Fock
2981: space formalism and yields Equation~\eqref{vv22}.
2982:
2983:
2984:
2985: The first summand of Equation~\eqref{vv22} gives the diagonal terms of the
2986: matrix $\MM _{2}$. The second summand gives the off diagonal terms with
2987: the $t_{1}t_{2}$ term of the summand appearing below the diagonal and the
2988: remaining term appearing above.
2989:
2990:
2991: \begin{lemma}\label{lem: eigenvals of M2 are distinct}
2992: The eigenvalues of $\MM _{2} $ are distinct.
2993: \end{lemma}
2994:
2995: The eigenvalues are symmetric functions in $t_{1}$ and $t_{2}$. In the
2996: $t_{1}t_{2}=0$ limit, $\MM _{2}$ is \emph{upper-triangular}. Hence it
2997: suffices to show that the diagonal entries are distinct. By
2998: Equation~\eqref{vv22}, the diagonal entry at a partition $\mu $ is
2999: \begin{equation}\label{eqn: diagonal entries of M}
3000: -\frac{t_{1}+t_{2}}{2}\sum _{k>0}km_{k} (\mu )F_{k}
3001: \end{equation}
3002: where $m_{k} (\mu )$ is the number of $k$'s in the partition $\mu $ and
3003: \[
3004: F_{k}=k\frac{(-q)^{k}+1}{(-q)^{k}-1}-\frac{(-q)+1}{(-q)-1}.
3005: \]
3006: The rational functions $\{F_{k} \}_{k>1}$ are easily seen to be linearly
3007: independent over $\qnums $ (by, for example, studying the poles of
3008: $F_{k}$), and hence the diagonal entries are distinct.\qed
3009:
3010:
3011: \begin{comment}
3012: Formula \eqref{vv22} is obtained from a direct calculation. Domain
3013: components of positive genus contribute {\em exactly} as in Theorem
3014: \ref{thm: the dd2 integral} and yield the first summand. Genus 0 domain
3015: components contribute as Hurwitz numbers and yield the second.
3016: \end{comment}
3017:
3018:
3019: \subsection{Proof of Theorem \ref{thm: reconstruction of level (0,0) pants}}
3020:
3021: Let $d>0$. We
3022: abbreviate a list $(2),\dots ,(2)$ of $r$ copies of $(2)$ by $(2)^{r}$.
3023: The gluing formula yields the equation
3024: \[
3025: \lang \mu \vertline \MM^r_2 \vertline \nu \rang
3026: =
3027: (-1)^d\mathsf{GW}^*(0\vertline 0,0)_{\mu, (2)^{r}, \nu}
3028: \]
3029: for partitions $\mu,\nu$ of $d$.
3030:
3031: A second application of the gluing formula yields the following computation:
3032: \begin{eqnarray*}
3033: \mathsf{GW}^*(0\vertline 0,0)_{\mu, (2)^r, \nu} & = &
3034: \sum_{\gamma\vdash d} \mathsf{GW}^*(0\vertline 0,0)_{\mu\gamma\nu}
3035: \mathsf{GW}^*(0\vertline 0,0)^{\gamma }_{ (2)^r} \\
3036: & = &
3037: \sum_{\gamma\vdash d}
3038: \mathsf{GW}^*(0\vertline 0,0)_{\mu\gamma\nu}
3039: \ \zz(\gamma)(-t_1t_2)^{l(\gamma)} (-1)^d \lang \gamma \vertline \MM^r_2 \vertline (1^d) \rang.
3040: \end{eqnarray*}
3041: The second equality uses the level $(0,0)$ cap calculation of
3042: Lemma~\ref{lem: charge (0,0) cap}.
3043:
3044: Taken together, the above equation provide a linear system for the
3045: degree $d$, level $(0,0)$ pair of pants integrals,
3046: \begin{equation}\label{ff897}
3047: \lang \mu \vertline \MM^r_2 \vertline \nu \rang =
3048: \sum_{\gamma\vdash d}
3049: \mathsf{GW}^*(0\vertline 0,0)_{\mu\gamma\nu}
3050: \ \zz(\gamma)(-t_1t_2)^{l(\gamma)}
3051: \ \lang \gamma \vertline \MM^r_2 \vertline (1^d) \rang.
3052: \end{equation}
3053: The linear equations have coefficients in the field $\Q(t_1,t_2,q)$.
3054: The proof of the Theorem is concluded by demonstrating the nonsingularity of
3055: the system \eqref{ff897}.
3056:
3057: Let $\cF_d \subset \cF$ be the subspace spanned by the vectors $|\mu
3058: \rangle$ satisfying $|\mu|=d$. The transformation $\MM_2$ preserves
3059: $\cF_d$.
3060:
3061: The eigenvectors for $\MM _{2}$ restricted to $\cF _{d}$ are the idempotent
3062: basis of the semisimple Frobenius algebra associated to the degree $d$,
3063: level $(0,0)$ theory. The identity vector $|(1^d)\rangle$ of the Frobenius
3064: algebra has the coefficient 1 in each component of the idempotent
3065: basis. Hence, the set of vectors
3066: \[
3067: \{ \ \MM_2^r\vertline (1^d) \rangle \ \}_{r\geq 0}
3068: \]
3069: have coefficients given by powers of the eigenvalues of $\MM _{2}$
3070: restricted to $\cF _{d}$. These eigenvalues are distinct by Lemma~\ref{lem:
3071: eigenvals of M2 are distinct}, thus the above set of vectors spans $\cF_d$
3072: and the linear system \eqref{ff897} is nonsingular. \qed
3073:
3074:
3075:
3076:
3077:
3078:
3079:
3080: \subsection{Proof of Theorem \ref{thm: Z is a rational function of q}}
3081:
3082: Since $-q=e^{-iu}$, we may disregard all integral terms in the exponent of the
3083: prefactor
3084: $$e^{\frac{idu}{2} (2-2g+{k_{1}}+{k_{2}})}.$$
3085: Consider
3086: the product
3087: \[
3088: e^{\frac{idu}{2} ({k_{1}}+{k_{2}})}\
3089: \mathsf{GW}^*(g\vertline k_{1},k_{2})_{\lambda ^{1}\dots \lambda ^{r}}.
3090: \]
3091: The series $\mathsf{GW}^*(g\vertline k_{1},k_{2})_{\lambda ^{1}\dots
3092: \lambda ^{r}}$ can be calculated by gluing in terms of the caps
3093: \begin{equation}\label{xsd}
3094: \mathsf{GW}^*
3095: (0\vertline \pm1,0)_\lambda, \ \ \mathsf{GW}^*(0\vertline 0,\pm1)_{\lambda }
3096: \end{equation}
3097: and the pair of pant series
3098: \[
3099: \mathsf{GW}^*(0\vertline 0,0)_{\lambda\mu\nu}.
3100: \]
3101: By the proof of Theorem
3102: \ref{thm: reconstruction of level (0,0) pants},
3103: the pair of pant series lie in $\Q(t_1,t_2,q)$.
3104: By the calculation of Section \ref{n3451},
3105: $$e^{-\frac{idu}{2}}\mathsf{GW}^*(0\vertline -1,0)_\lambda, \
3106: e^{-\frac{idu}{2}}\mathsf{GW}^*(0\vertline 0,-1)_\lambda \in
3107: \Q(t_1,t_2,q).$$ Since the opposite caps are inverses in the Frobenius
3108: algebra, we conclude
3109: $$e^{\frac{idu}{2}}\mathsf{GW}^*(0\vertline 1,0)_\lambda, \
3110: e^{\frac{idu}{2}}\mathsf{GW}^*(0\vertline 0,1)_\lambda \in \Q(t_1,t_2,q).$$
3111: The Theorem is proven by distributing a factor of $e^{\pm\frac{idu}{2}}$ to
3112: each cap of type \eqref{xsd} in the gluing formula. \qed
3113:
3114:
3115: \begin{thebibliography}{99}
3116:
3117:
3118: \bibitem{Aganagic-Ooguri-Saulina-Vafa}
3119: M.~Aganagic, H.~Ooguri, N.~Saulina, and C.~Vafa,
3120: {\em Black holes, $q$-deformed 2d Yang-Mills, and
3121: non-perturbative topological strings}, Nuclear Phys. B \textbf{715} (2005), 304--348. hep-th/0411280.
3122:
3123:
3124:
3125: \bibitem{Bryan-Graber} J. Bryan and T. Graber,
3126: {\em The crepant resolution conjecture}, math.AG/0610129.
3127:
3128: \bibitem{Br-Pa-rigidity}
3129: J. Bryan and R. Pandharipande,
3130: {\em On the rigidity of stable maps to Calabi-Yau threefolds},
3131: Proceedings of the BIRS workshop on the
3132: interaction of finite type and Gromov-Witten invariants, Geometry and Topology Monographs, Vol. \textbf{8} (2006), 97--104.
3133: math.AG/0405204.
3134:
3135: \bibitem{Br-Pa}
3136: J. Bryan and R. Pandharipande,
3137: {\em BPS states of curves in Calabi-Yau 3-folds},
3138: Geom. Topol. {\bf 5} (2001), 287--318.
3139:
3140: \bibitem{Br-Pa-TQFT} J. Bryan and R. Pandharipande,
3141: {\em Curves in Calabi-Yau 3-folds and Topological Quantum Field
3142: Theory}, Duke J. \textbf{126} (2005), 369--396,
3143: {math.AG/0306316}.
3144:
3145:
3146: \bibitem{Dijkgraaf-Witten90}
3147: R. Dijkgraaf and E. Witten,
3148: {\em Topological gauge theories and group cohomology},
3149: {Comm. Math. Phys.} {\bf{129}} (1990),
3150: {393--429},
3151:
3152: \bibitem{EGH}
3153: Y.~Eliashberg, A.~Givental, and H.~Hofer,
3154: \emph{Introduction to symplectic field theory,}
3155: Geom. Funct. Anal. (2000), 560--673.
3156:
3157:
3158: \bibitem{Fa-Pa} C. Faber and R. Pandharipande,
3159: {\em Hodge integrals and Gromov-Witten theory},
3160: {Invent. Math.} {\bf{139}} {2000},
3161: {173--199}.
3162:
3163: \bibitem{Faber-Pandharipande-logarithmic}
3164: {C. Faber and R. Pandharipande},
3165: {\em Logarithmic series and Hodge integrals in the tautological
3166: ring (with an appendix by D. Zagier)}
3167: {Michigan Math. J.} {\bf{48}} (2000),
3168: {215--252}.
3169:
3170: \bibitem{Freed-Quinn}
3171: {D. Freed and F. Quinn},
3172: {\em Chern-Simons theory with finite gauge group},
3173: {Comm. Math. Phys.}
3174: {\bf{156}} (1993), {435--472}.
3175:
3176:
3177: \bibitem{Gr-Pa}
3178: T.~Graber and R.~Pandharipande,
3179: {\em Localization
3180: of virtual classes}, Invent. Math. {\bf 135} (1999), 487--518.
3181:
3182:
3183:
3184: \bibitem{Ionel-Parker-Annals2003}
3185: E.-N. Ionel and T. Parker.
3186: \emph{Relative {G}romov-{W}itten invariants.}
3187: Ann. of Math. \textbf{157} (2003), 45--96.
3188:
3189: \bibitem{Ionel-Parker00}
3190: E.-N. Ionel and T. Parker,
3191: \emph{The symplectic sum formula for Gromov-Witten invariants},
3192: Ann. of Math. \textbf{159} (2004), 935--1025.
3193:
3194:
3195:
3196: \bibitem{Kock:FA-2DTQFT}
3197: {J. Kock},
3198: {\em Frobenius algebras and 2D topological quantum field
3199: theories},
3200: {London Mathematical Society Student Texts} {\bf {59}},
3201: {Cambridge University Press}:
3202: {Cambridge},
3203: {2004}.
3204:
3205:
3206:
3207: \bibitem{Li-Ruan}
3208: A.-M. Li and Y. Ruan,
3209: \emph{ Symplectic surgery and {G}romov-{W}itten invariants of {C}alabi-{Y}au
3210: 3-folds,}
3211: Invent. Math.\textbf{145} (2001),151--218.
3212:
3213: \bibitem{Li-relative1}
3214: J. Li,
3215: \emph{Stable morphisms to singular schemes and relative stable morphisms,}
3216: JDG \textbf{57} (2001), 509--578.
3217:
3218:
3219: \bibitem{Li-relative2}
3220: J. Li,
3221: \emph{A degeneration formula of GW-invariants}, JDG {{\bf 60}}
3222: (2002), 199--293.
3223:
3224:
3225: \bibitem{Looijenga-invent}
3226: E. Looijenga.
3227: \newblock \emph{On the tautological ring of {${M}\sb g$}.}
3228: \newblock { Invent. Math.}, \textbf{121} (1995), 411--419.
3229:
3230:
3231: \bibitem{Macdonald}
3232: I.~Macdonald,
3233: \emph{Symmetric functions and Hall polynomials},
3234: The Clarendon Press, Oxford University Press, New York, 1995.
3235:
3236:
3237:
3238: \bibitem{MNOP1} D.~Maulik, N.~Nekrasov, A.~Okounkov, and R.~Pandharipande, {\em Gromov-Witten
3239: theory and Donaldson-Thomas theory I}, math.AG/0312059. To appear in Comp. Math.
3240:
3241: \bibitem{MNOP2} D.~Maulik, N.~Nekrasov, A.~Okounkov, and R.~Pandharipande, {\em Gromov-Witten
3242: theory and Donaldson-Thomas theory II}, math.AG/0406092. To appear in Comp. Math.
3243:
3244:
3245:
3246:
3247: \bibitem{Okounkov-Pandharipande-unknot}
3248: A. Okounkov and R. Pandharipande, {\em Hodge integrals and invariants of the
3249: unknot}, Geom. Top. {\bf 8} (2004), 675-699.
3250:
3251: \bibitem{Okounkov-Pandharipande-completed-cycles} A. Okounov and
3252: R. Pandharipande,
3253: {\em Gromov-Witten theory, Hurwitz theory, and completed cycles},
3254: Annals of Math. (2) \textbf{163} (2006), 517--560. {math.AG/0204305}.
3255:
3256: \bibitem{Ok-Pan-Hilb} A. Okounkov and R. Pandharipande,
3257: {\em
3258: Quantum cohomology of the Hilbert scheme
3259: of points in the plane}, math.AG/0411120.
3260:
3261: \bibitem{dtlc} A. Okounkov and R. Pandharipande, {\em Local
3262: Donaldson-Thomas theory of curves}, math.AG/0512573.
3263:
3264: \bibitem{Pandharipande-degenerate-contributions}
3265: R. Pandharipande,
3266: {\em Hodge integrals and degenerate contributions},
3267: {Comm. Math. Phys.}
3268: {\bf {208}} (1999), {489--506}.
3269:
3270:
3271: \bibitem{pandharipande-toda2000}
3272: R. Pandharipande, {\em The Toda equation and the Gromov-Witten theory
3273: of the Riemann sphere}, Lett. Math. Physics {\bf{53}} (2000), 59--74.
3274:
3275: \bibitem{Pandharipande-ICM}
3276: {R. Pandharipande},
3277: {Three questions in Gromov-Witten theory},
3278: {Proceedings of the International Congress of Mathematicians,
3279: Vol. II},
3280: {503--512},
3281: {Higher Ed. Press}: {Beijing},
3282: {2002}.
3283:
3284:
3285: \bibitem{t} R. Thomas, {\em A holomorphic Casson invariant for Calabi-Yau
3286: 3-folds, and bundles on K3 fibrations}, JDG {\bf 54} (2000), 367--438.
3287:
3288: \bibitem{Vafa-04-2dYang-Mills} C. Vafa, \emph{Two dimensional Yang-Mills, black holes and topological strings}. hep-th/0406058.
3289:
3290:
3291: \end{thebibliography}
3292:
3293:
3294: %\bibliography{mainbiblio}
3295:
3296:
3297:
3298:
3299:
3300:
3301: \bibliographystyle{plain}
3302:
3303:
3304:
3305:
3306:
3307: \vspace{+10 pt}
3308: \noindent
3309: Department of Mathematics \\
3310: University of British Columbia \\
3311: Vancouver, BC, V6T 1Z2, Canada\\
3312: jbryan@math.ubc.ca \\
3313:
3314:
3315:
3316:
3317: \vspace{+10 pt}
3318: \noindent
3319: Department of Mathematics\\
3320: Princeton University\\
3321: Princeton, NJ 08544, USA\\
3322: rahulp@math.princeton.edu
3323:
3324:
3325: \end{document}
3326: