math0602161/CBS.tex
1: % Equivariant Chow rings of quot schemes
2: % Tom Braden
3: % Linda Chen
4: % Frank Sottile
5: %
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7: \documentclass[12pt]{amsart}
8: \usepackage{epsf,amssymb}
9: \usepackage{colordvi}
10: \usepackage{graphicx,epsf}
11: 
12: %%%%%%%%%%   Layouts    %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: %  US  Letter size
14: \headheight=8pt     \topmargin=20pt
15: \textheight=620pt     \textwidth=450pt
16: \oddsidemargin=15pt   \evensidemargin=15pt
17: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
18: \theoremstyle{plain}
19: \newtheorem{thm}{Theorem}
20: \newtheorem{lemma}[thm]{Lemma}
21: \newtheorem{prop}[thm]{Proposition}
22: \newtheorem{example}[thm]{Example}
23: \theoremstyle{definition}
24: \newtheorem*{rmk}{Remark}
25: %%%%%%%%%%%%%%%%%%%%%%%    Definitions
26: 
27: \newcommand{\Grkn}{{\mathit{Gr}(k,n)}}
28: 
29: \newcommand{\C}{{\mathbb{C}}}
30: \newcommand{\G}{{\mathbb{G}}}
31: \newcommand{\R}{{\mathbb{R}}}
32: \renewcommand{\P}{{\mathbb{P}}}
33: \newcommand{\Z}{{\mathbb{Z}}}
34: \newcommand{\Q}{{\mathbb{Q}}}
35: 
36: \newcommand{\cF}{{\mathcal{F}}}
37: \newcommand{\cM}{{\mathcal{M}}}
38: \newcommand{\cO}{{\mathcal{O}}}
39: \newcommand{\cQ}{{\mathcal{Q}}}
40: \newcommand{\cS}{{\mathcal{S}}}
41: \newcommand{\cT}{{\mathcal{T}}}
42: \newcommand{\cY}{{\mathcal{Y}}}
43: \newcommand{\cZ}{{\mathcal{Z}}}
44: 
45: \newcommand{\va}{{\mathbf a}}
46: \newcommand{\vb}{{\mathbf b}}
47: \newcommand{\vc}{{\mathbf c}}
48: \newcommand{\vj}{{\mathbf j}}
49: \newcommand{\vl}{{\mathbf l}}
50: \newcommand{\vm}{{\mathbf m}}
51: \newcommand{\vr}{{\mathbf r}}
52: \newcommand{\vdelta}{{\mbox{\boldmath$\delta$}}}
53: \newcommand{\sdelta}{{\mbox{\scriptsize\boldmath$\delta$}}}
54: \newcommand{\vzero}{{\mathbf 0}}
55: 
56: \newcommand{\be}{{\mathbf e}}
57: \newcommand{\bE}{{\mathbf E}}
58: \newcommand{\bff}{{\mathbf f}}
59: \newcommand{\bv}{{\mathbf v}}
60: 
61: \newcommand{\ol}{\overline}
62: 
63: \newcommand{\Hom}{\mathrm{Hom}}
64: \newcommand{\frakt}{\mathfrak{t}}
65: \newcommand{\frakm}{\mathfrak{m}}
66: \newcommand{\Lie}{\mathop{\mathrm{Lie}}}
67: \newcommand{\lcm}{\mathop{\mathrm{lcm}}}
68: \newcommand{\Span}{\mathop{\mathrm{Span}}}
69: \def\im{\mathop{\rm Im}\nolimits}
70: 
71: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
72: %%\def\publname{\scriptsize \Red{30 November 2006. {\tt math.AG/0602161}}
73: %\def\currentvolume{}
74: %\def\currentissue{}
75: %%\pagespan{1}{60}
76: %\PII{}}
77: %\copyrightinfo{}{}
78: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
79: \begin{document}
80: 
81: \title{The Equivariant Chow rings of quot schemes}  
82: 
83: \author{Tom Braden}
84: \address{Department of Mathematics\\
85:         University of Massachusetts\\
86:         Amherst, MA \ 01003\\
87:         USA}
88: \email{braden@math.umass.edu}
89: 
90: \author{Linda Chen}
91: \address{Department of Mathematics\\
92:         The Ohio State University\\
93:         Columbus, OH \ 43210\\
94:         USA}
95: \email{lchen@math.ohio-state.edu}
96: 
97: \author{Frank Sottile}
98: \address{Department of Mathematics\\
99:          Texas A\&M University\\
100:          College Station\\
101:          TX \ 77843\\
102:          USA}
103: \email{sottile@math.tamu.edu}
104: 
105: 
106: %\date{\today}
107: \thanks{Braden supported in part by NSF grant DMS-0201823}
108: \thanks{Chen supported in part by NSF VIGRE grant DMS-9810750 and NSF grant
109:   DMS-0432701} 
110: \thanks{Sottile supported in part by NSF CAREER
111:   grant DMS-0134860 and the Clay Mathematical Institute}
112: \subjclass{14C05, 14F43, 14M15, 55N91}
113: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
114: %  14C05  Parametrization (Chow and Hilbert schemes)
115: %  14M15  Grassmannians, Schubert varieties, flag manifolds
116: %  14F43  Intersection Theory
117: %  55N91  Equivariant homology and cohomology
118: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
119: \keywords{Quot scheme, Grassmannian, equivariant cohomology} 
120: 
121: \begin{abstract}
122:  We give a presentation for the (integral) torus-equivariant Chow ring of the
123:  quot scheme, a smooth compactification of the space of rational curves of
124:  degree $d$ in the Grassmannian.
125:  For this presentation, we refine Evain's extension of the method of
126:  Goresky, Kottwitz, and MacPherson
127:  to express the torus-equivariant Chow ring in terms of the torus-fixed points and 
128:  explicit relations coming from the geometry of families of torus-invariant curves.
129:  As part of this calculation, we give a complete description of the
130:  torus-invariant curves on the quot scheme and show that each family is a 
131:  product of projective spaces.
132: \end{abstract}
133: 
134: \maketitle
135: 
136: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
137: \section*{Introduction}
138: 
139: Let $k$ be an algebraically closed field 
140: and let $d,n,r$ be non-negative integers with $r<n$.
141: We study the quot scheme $\cQ_d:=\cQ_d(r,n)$ parametrizing
142: quotient sheaves  on $\P^1$
143: of the trivial vector bundle $\cO^n_{\P^1}$ which have rank $r$ and degree $d$. 
144: When $r>0$, this is a compactification of the space $\cM_d$ of
145: parametrized rational curves of degree $d$ on the Grassmannian $G(r,n)$
146: of $r$-dimensional quotients of $k^n$.  
147: Indeed, a morphism from $\P^1$ to $G(r,n)$ of degree $d$ is
148: equivalent to a quotient bundle $\cO^n_{\P^1}\rightarrow\cT$ of rank $r$
149: and degree $d$.  
150: 
151: Str{\o}mme \cite{St87} showed that $\cQ_d(r,n)$ is a smooth, projective, 
152: rational variety of dimension 
153: $r(n-r)+nd$. He described the decomposition into Bia{\l}ynicki-Birula 
154: cells induced by an action of a one-dimensional torus $T$ on $\cQ_d$,
155: thereby determining its Betti numbers.   
156: He also gave a presentation of its integral Chow ring (Theorem 5.3, {\it
157: loc. cit.}) in terms of generators and relations.  However, the set of
158: generators is far from minimal, and the relations are given by the
159: annihilator of a certain class, and are therefore non-explicit.  He also
160: gave a more elementary set of generators for its rational Chow ring.
161: We compute the corresponding equivariant 
162: classes in Sections \ref{classes} and \ref{S:ex}.
163: 
164: Later, 
165: Bertram used the geometry of $\cQ_d$ to determine the (small) quantum
166: cohomology ring of the Grassmannian~\cite{Bertram}.  
167: He used a recursive description of the boundary $\cQ_d\setminus\cM_d$ to
168: show that the $3$-point genus zero Gromov-Witten invariants of the
169: Grassmannian are equal to particular intersection numbers on $\cQ_d$.  By
170: studying certain types of intersections on quot schemes, he obtained a
171: quantum Schubert calculus for the Grassmannian.  However, he did not need
172: to compute the full cohomology ring of $\cQ_d$.
173: 
174: Our main result is a presentation
175: of the $T$-equivariant Chow ring $A_T^*(\cQ_d)$ for an
176: action of a torus $T$ on $\cQ_d$.   This  
177: determines the ordinary Chow ring $A^*(\cQ_d)$.  
178: Our presentation gives $A_T^*(\cQ_d)$ as an explicit
179: subring of a direct sum of polynomial rings.
180: When $k=\C$, the cycle class map induces an
181: isomorphism between Chow and cohomology rings, so our result
182: also determines the $T$-equivariant and ordinary cohomology rings of $\cQ_d$.
183: 
184: Our presentation arises from an analysis of the localization map.
185: When an algebraic torus $T$ acts on a smooth variety 
186: $X$ with finitely many fixed points $X^T$, 
187: the inclusion $i\colon X^T\hookrightarrow X$ induces the
188: localization map of (integral) equivariant Chow rings
189: %
190:  \begin{equation} \label{loc map}
191:    i^*\colon A^*_T(X) \to A^*_T(X^T)\ =\  
192:      \bigoplus_{p\in X^T} A^*_T(p)\,.
193:  \end{equation}
194: %
195: Each summand $A^*_T(p)$ is canonically isomorphic to the symmetric algebra
196: $S$ of the character group of $T$.
197: When $k=\C$, there is a similar localization for rational equivariant cohomology.  
198: 
199: 
200: 
201: 
202: When $k=\C$, Chang and Skjelbred~\cite{ChSk74} showed that the image of the
203: localization map in cohomology is cut out by the images of the $T$-equivariant
204: cohomology of components of the one-skeleton of $X$ (the points
205: fixed by some codimension one subtorus of $T$).
206: These components are closures of families of $T$-invariant curves.
207: In particular, when $X$ has finitely many one-dimensional $T$-orbits (whose
208: closures are $T$-invariant curves), Goresky, 
209: Kottwitz, and MacPherson~\cite{GKM98} used this to describe
210: the image of the localization map for equivariant cohomology.  
211: Each $T$-invariant curve gives a 
212: relation, and these GKM relations cut out the image.
213: Brion~\cite{brion} showed that this remains true for rational equivariant Chow
214: rings of varieties over any algebraically closed field.\smallskip
215: 
216: The torus $T=T_{k^n}\times T_{\P^1}$ acts on $\cQ_d$,
217: where $T_{k^n}= (\G_m)^n$ acts diagonally on $k^n$ and $T_{\P^1}=\G_m$
218: acts primitively on $\P^1$. 
219: This action has finitely many fixed points, but there are 
220: infinitely many $T$-invariant curves.
221: The GKM relations remain valid, but are now insufficient to cut out
222: the image of $i^*$. 
223: There will be extra relations coming from 
224: connected components of the one-skeleton of $\cQ_d$.
225: Brion~\cite{brion} adapted the result of Chang and Skjelbred to Chow
226: groups, showing that the relations given by these families are sufficient
227: to cut out $A^*_T(X)$, rationally.  Determining these relations explicitly
228: is more difficult than for the GKM relations, however, and few
229: cases have been worked out in detail.      
230: 
231: One case for which the relations are known is when $X$ is a Hilbert scheme of
232: points on a toric surface, which has families of $T$-invariant curves.
233: Following a sugestion of Brion,  Evain~\cite{evain} used 
234: Edidin and Graham's~\cite{EG98b} version of the 
235: Atiyah-Bott-Berline-Vergne
236: localization formula for equivariant Chow groups to give relations in terms of
237: ideal-membership. 
238: The relations come from elements of the $T$-equivariant Chow rings of
239: the families of $T$-invariant curves.
240: 
241: 
242: We discuss this in Section \ref{extended GKM theory}, and
243: give a more explicit formula for Evain's relations when each
244: component $Y$ of $X^{T'}$ has smooth $T$-invariant subvarieties $Z$
245: whose classes $[Z]$ generate $A^*_T(Y)$. 
246: We derive necessary and sufficient linear relations over $\Q$ from
247: Evain's ideal-membership relations. 
248: Lastly, we show that if the $T$-weights of the tangent space at each
249: fixed point are not too dependent (see Theorem~\ref{T:GKM_algebraic_Z}),
250: then Evain's relations also determine the integral Chow ring.
251: 
252: All of these additional
253: hypotheses hold for $\cQ_d$.  
254: In fact, the components $Y$ for $\cQ_d$ are relatively simple: all are products of 
255: projective spaces.  
256: As a result, we obtain explicit descriptions of the equivariant Chow
257: ring of $\cQ_d$, both rationally and integrally. 
258: \smallskip
259: 
260: 
261: To describe the combinatorics of fixed points in $\cQ_d^T$, we 
262: use the following notations.
263: For an element $\va = (a_1, \dots, a_n) \in \Z^n$, we define
264: $|\va|=\sum_i a_i$.  
265: We use addition and subtraction on $\Z^n$ considered as
266: an abelian group, and denote the identity element by 
267: $\vzero = (0,\dots,0)$. If 
268: $\va\in (\Z_{\ge 0})^n$, we set $\va! = \prod_i a_i!$, 
269: where $0! = 1$.  Finally we use the partial order $\va \le \vb$ to
270: mean $a_i \le b_i$ for all $i$.
271: 
272: In Section~\ref{fixed_points}, we give an
273: explicit parametrization of the fixed point set
274: $\cQ_d^T$ by a set $\cF$ of triples $(\vdelta, \va, \vb)$, where 
275: %
276: \begin{itemize}
277:  \item $\vdelta\in \{0,1\}^n$ takes the
278: value $1$ exactly $n{-}r$ times, so that $|\vdelta| = n-r$, and
279:  \item $\va,\vb$ are elements of $(\Z_{\ge 0})^n$ 
280:    which satisfy $|\va|+|\vb|=d$ and for which $\delta_i = 0$ 
281: implies $a_i = 0$ and $b_i = 0$.
282: 
283:  \end{itemize}
284: %
285: The fixed points are maximally degenerate quotient sheaves supported
286: at $0$ and $\infty$; the data $\va$ and $\vb$ describe the structure of 
287: the stalks as modules over $\cO_{\P^1}$ and as representations of $T$.
288: 
289: Recall that $A^*_T(p) = S$, the symmetric algebra of the character group of $T$.
290: We have $S=\Z[\be_1,\dots,\be_n,\bff]$, where 
291: $\be_1,\dots,\be_n$, and  $\bff$ are dual to the obvious basis coming from the
292: decomposition $T = T_{k^n}\times T_{\P^1}$.
293: In particular, $\bff$ restricts to the identity character
294: on $T_{\P^1}$ and to the trivial character of $T_{\C^n}$.
295: We write $S_\Q$ for $S\otimes_\Z\Q$ and sometimes $S_\Z$ for $S$, 
296: when we wish to emphasize our ring of scalars.
297: We write $S^\cF$ for the set of tuples of polynomials 
298: $(f_{(\sdelta,\va,\vb)}\in S \mid (\vdelta,\va,\vb)\in \cF)$.
299: Then $S^\cF =  A^*_T( Q_d^T)$, under the identification of $\cF$ with
300: $Q_d^T$.  
301: We exhibit the image of the localization map as a subring of $S^\cF$.
302: 
303: In Section \ref{Tcurves} we describe a finite set of $T$-invariant
304: curves which span the tangent space at each fixed point.  In Section
305: \ref{Tfamilies}, we describe the families of $T$-invariant curves
306: and their closures.  Based on this description, our relations for the
307: image of $i^*_{\cQ_d}$ are the following:
308: 
309: \begin{enumerate}
310: \item[{I.}] For any pair $(\vdelta,\va, \vb)$, $(\vdelta',\va', \vb')\in \cF$ with 
311: $\vdelta = \vdelta'$, $\va = \va'$ and $\vb = \vb'$ except in positions $i$ and $j$, and
312: $\delta_i = \delta'_j = 1$ and $\delta_j = \delta'_i = 0$, we have
313: %
314:  \[
315:      f_{(\sdelta,\va,\vb)}\  \equiv\  
316:      f_{(\sdelta',\va',\vb')} \mod \be_j - \be_i +(a'_j - a_i)\bff.
317:  \]
318: %
319:   (Note that $(a'_j - a_i) = -(b'_j - b_i)$, since $a_i + b_i = a'_j + b'_j$.)
320: 
321: \item[{II.}] 
322:  \begin{enumerate}
323:   \item For any pair $(\vdelta,\va, \vb)$, $(\vdelta,\va', \vb)\in \cF$
324:     with $\va$, $\va'$ agreeing except in positions $i$ and $j$,
325: we have
326: %
327:    \[
328:       f_{(\sdelta,\va,\vb)}\ \equiv\ f_{(\sdelta,\va',\vb)} \mod \be_j - \be_i +(a'_j -
329:       a_i)\bff.
330:    \]
331: %
332: \item For any pair $(\vdelta,\va, \vb)$, $(\vdelta,\va, \vb')\in \cF$
333:       with $\vb$, $\vb'$ agreeing except in positions $i$ and $j$,
334: we have 
335:   \[
336:      f_{(\sdelta,\va,\vb)}\ \equiv\ 
337:      f_{(\sdelta,\va,\vb')} \mod \be_j - \be_i + (b_i - b'_j)\bff.
338:   \]
339: %
340: \item If we have $(\vdelta,\va, \vb)$, $(\vdelta,\va', \vb)$, 
341:       $(\vdelta, \va, \vb')\in \cF$ satisfying both of the previous conditions
342:       (with the same $i$ and $j$), and in addition $a'_j - a_i = b_i - b'_j$,
343:       then 
344: %
345:   \[
346:       Df_{(\sdelta,\va,\vb)} - Df_{(\sdelta,\va',\vb)}
347:          - Df_{(\sdelta,\va,\vb')} + Df_{(\sdelta,\va',\vb')}\ \equiv\
348:       0 \mod \be_j - \be_i+(a'_j - a_i)\bff\,
349:   \]
350: where  $D$ is differentiation in the direction of $\be_j^\vee$, the
351: dual basis vector to $\be_j$.
352: 
353: \item[(c)$'$] Under the hypotheses of II(c),
354: %
355:   \[
356:       f_{(\sdelta,\va,\vb)} - f_{(\sdelta,\va',\vb)} 
357:          - f_{(\sdelta,\va,\vb')} + f_{(\sdelta,\va',\vb')}\ \equiv\ 
358:       0 \mod (\be_j - \be_i+(a'_j - a_i)\bff)^2\,.
359:   \]
360: %
361: \end{enumerate}
362: 
363: \item[{III.}] For every $(\vdelta,\va,\vb)\in\cF$ and every 
364: $\vzero \le \vb' < \vb$,
365:   \[
366:      \sum_{\vb'\le \vc \le \vb}  \frac{(-1)^{|\vc|}}{(\vb-\vc)!(\vc - \vb')!} 
367:       D^{|\vb|-|\vb'|-1}f_{(\sdelta,\va+\vc,\vb-\vc)} 
368: \equiv 0 \mod \bff\,,
369:   \]
370: %
371:     where $D$ is differentiation in the direction of $\bff^\vee$, the dual 
372: basis vector to $\bff$.
373:  
374: \item[{III$'$.}] For every $(\vdelta,\va,\vb)\in\cF$, $\vb \ne \vzero$,
375: \[ 
376: \sum_{0\leq \vc\leq \vb} \frac{(-1)^{|\vc|}}{\vc!(\vb-\vc)!} \, 
377: f_{(\sdelta,\va+\vc,\vb-\vc)} \equiv 0 \mod  \bff^{|\vb|}.
378: \]
379: 
380: \end{enumerate}
381: 
382: Relations I,  II(a), and II(b)
383: are standard GKM relations, while the rest come from 
384: families of $T$-invariant curves.    
385: In particular, relations II(c)/II(c)$'$ (respectively III/III$'$) 
386: come from certain families whose closures are isomorphic to $\P^1\times\P^1$ 
387: (respectively arbitrary products of projective spaces).  
388: 
389: \begin{thm} \label{Main_Theorem}
390: The rational equivariant Chow ring $A^*_T(\cQ_d)_\Q$ is isomorphic
391: to the set of tuples $f = (f_{(\sdelta,\va,\vb)}) \in S^\cF_\Q$
392: subject to the relations {\rm I}, {\rm II(a)(b)(c)}, and\/ {\rm III}.
393:   
394: The integral equivariant Chow ring $A^*_T(\cQ_d)$ is isomorphic
395: to the set of tuples $f = (f_{(\sdelta,\va,\vb)}) \in S^\cF_\Z$
396: subject to the relations {\rm I}, {\rm II(a)(b)(c)$'$}, and\/ {\rm III}$'$.  
397: \end{thm}
398:  
399: 
400: We prove Theorem \ref{Main_Theorem} in Section \ref{main proof}.
401: Since the equivariant Chow ring determines the ordinary 
402: Chow ring for smooth spaces, this gives in principle a complete
403: description of the Chow ring of $\cQ_d$.
404: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
405: %
406: %   Frank Added the following
407: %
408: The resulting computation of Betti numbers is the same as Stroome's
409:  computation.
410: %We note that the Betti numbers of $\cQ_d$ so obtained from
411: %Theorem~\ref{Main_Theorem} coincide with those computed by Str{\o}mme.
412: %Both computations use the same data, the $T$-fixed points of $\cQ_d$ and certain
413: %$T$-weights. 
414: %Str{\o}mme's calculation used the weights of $T$ acting on tangent spaces at the
415: %fixed points and Theorem~\ref{Main_Theorem} essentially uses the weights of 
416: %$T$ on certain invariant curves. 
417: %These data coincide, by Theorem~\ref{tangentdirections}.
418: 
419: 
420: 
421: 
422: 
423: The quot scheme $\cQ_d$ represents the functor which associates to a
424: scheme $X$ the set of flat families over $X$ of sheaves on $\P^1$ which are
425: quotients of $\cO_{\P^1}^n$ of rank $r$ and degree $d$.
426: Thus, there is a universal exact sequence
427: %
428:  \begin{equation}\label{E:definition}
429:     0\ \rightarrow\ \cS\ \rightarrow\
430:    \cO^n_{\P^1\times\cQ_d}\ 
431:    \rightarrow\  \cT\ \rightarrow 0
432:  \end{equation}
433: %
434: of sheaves on $\P^1\times\cQ_d$ where 
435: $\cT$ has rank $r$ and degree $d$.
436: This sequence of sheaves is flat, and flatness implies that 
437: $\cS$ is a vector bundle of degree $-d$ and rank $n{-}r$.
438: In particular, points of $\cQ_d$ are exact sequences of such sheaves on $\P^1$.
439: 
440: Str{\o}mme's generators of the rational cohomology ring of $\cQ_d$ were
441: K\"{u}nneth components of the Chern classes of the tautological vector bundle 
442: $\cS$ on $\P^1\times\cQ_d$. 
443: In Section~\ref{classes}, we describe the equivariant
444: Chern classes of $\cS$ in $A^*_T(\cQ_d)$ and thus lifts of Str\o mme's
445: generators to the equivariant Chow groups. 
446: In Section~\ref{S:ex} we work this out explicitly for
447: $\cQ_2(0,2)$, using Theorem~\ref{Main_Theorem} to describe
448: the equivariant and ordinary Chow rings
449: and giving explicit lifts of Str{\o}mme's generators as localized classes. 
450: \iffalse
451: In Section~\ref{classes}, we describe the equivariant
452: Chern classes of $\cS$ in $A^*_T(\cQ_d)$.
453: Since $\cS$ has a canonical $T$-equivariant structure, the 
454: localization of these classes are given by simple expressions in the 
455: $T$-weights of the $T$-invariant curves.  Choosing module generators
456: for the equivariant Chow ring of $\P^1$ yields a K\"unneth decomposition
457: of the equivariant Chern classes, and thus lifts of Str\o mme's
458: generators to the equivariant Chow groups.
459: 
460: In Section~\ref{S:ex} we work this out explicitly for
461: $\cQ_2(0,2)$, using Theorem~\ref{Main_Theorem} to describe
462: the equivariant and ordinary Chow rings
463: and giving explicit lifts of Str{\o}mme's generators as localized classes. 
464: \fi
465: 
466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
467: \section{Torus-fixed points of $\cQ_d$} \label{fixed_points}
468: 
469: 
470: The free module $\cO^n_{\P^1}=k^n\otimes\cO_{\P^1}$ has  
471: basis $\be_1,\dotsc,\be_n$.  
472: Write $T_{k^n}$ for the group of diagonal matrices in this basis.  
473: Let $[x,y]$ be
474: coordinates on $\P^1$ with $x$ vanishing at 0 and $y$ at $\infty$.
475: For $T_{\P^1}$ acting on $\P^1$ with fixed
476: points $0$ and $\infty$, the torus $T:= T_{k^n}\times T_{\P^1}$ acts
477: on $\cQ_d$ naturally as indicated by the given splitting.  
478: 
479: The $T$-fixed points 
480: are indexed by triples $(\vdelta,\va,\vb)$ in the
481: set $\cF$ of the Introduction.
482: The fixed point corresponding to $(\vdelta,\va,\vb)$ is the sequence of sheaves
483: on $\P^1$
484: %
485:  \[
486:    \cS_{(\sdelta,\va,\vb)}\ \longrightarrow\ \cO^n_{\P^1}\ 
487:    \longrightarrow\ \cT_{(\sdelta,\va,\vb)}
488:  \]
489: %
490: where $\cS_{(\sdelta,\va,\vb)}$ is the image of the map 
491: %
492:  \[
493:     \bigoplus_{i=1}^n\cO_{\P^1}(-a_i-b_i)\ 
494:     \xrightarrow{\ \mathrm{diag}(\delta_i x^{a_i} y^{b_i} )\ }\
495:     \cO^n_{\P^1}\,. 
496: \]
497: We identify this fixed point with the subsheaf
498: $\cS_{(\sdelta,\va,\vb)}$ of $\cO^n_{\P^1}$.
499: 
500: We introduce the following notation.
501: For natural numbers $a,b$, let $\cS_{a,b}$ be the subsheaf of $\cO_{\P^1}$
502: which is the image of the map
503: \[
504: \cO_{\P^1}(-a-b)\ \xrightarrow{\ x^ay^b\ }\ \cO_{\P^1}\,.
505: \]
506: Under the identification of modules over $\cO_{\P^1}$ with saturated graded
507: modules of the homogeneous coordinate ring $k[x,y]$,  $\cS_{a,b}$ is the ideal
508: of $k[x,y]$ generated by $x^ay^b$. 
509: The quotient $\cO_{\P^1}/\cS_{a,b}$ is the skyscraper sheaf 
510: %
511:  \[
512:   \cT_{a,b}\ =\ \cO_{\P^1}/\frakm_0^a\ \oplus
513:   \cO_{\P^1}/\frakm_\infty^b
514:  \]
515: %
516: on $\P^1$ supported at $0$ and at $\infty$.
517: Here, $\frakm_p$ is the sheaf of ideals cutting out the point $p\in\P^1$.
518: Then we have
519: %
520:  \begin{eqnarray*}
521:    \cS_{(\sdelta,\va,\vb)} &=& \bigoplus_{\delta_i=1}
522:         \cS_{a_i,b_i}\cdot\be_i\,,\qquad
523:         \mbox{\ and}\\
524:    \cT_{(\sdelta,\va,\vb)} &=& \bigoplus_{\delta_j=1} \cT_{a_j,b_j}\cdot\be_j\ \oplus\  
525:             \bigoplus_{\delta_j=0} \cO_{\P^1}\cdot\be_j\,.
526:  \end{eqnarray*}
527: %
528: In the sum,  $\delta_j=0$ means those $j$ in
529: $\{1,\dotsc,n\}$ with $\delta_j=0$, and the same for $\delta_j=1$.
530: 
531: 
532: The tangent space to $\cQ_d$ at this fixed point
533: is $\Hom(\cS_{(\sdelta,\va,\vb)},\,\cT_{(\sdelta,\va,\vb)})$.  
534: Let $\hat{\be}_1,\dotsc,\hat{\be}_n$ be the basis dual to
535: $\be_1,\dotsc,\be_n$.  
536: For each $i,j$, set
537: $\bE_{ij}:=\hat{\be}_i\otimes\be_j\in \Hom(k^n,k^n)$.  
538: These $\bE_{ij}$ form a basis for $\Hom(k^n,k^n)$.
539: 
540: \begin{thm}
541: \label{tangentspace}
542: The tangent space $T_{(\sdelta,\va,\vb)}\cQ_d$ is canonically
543: identified with
544: %
545:  \begin{equation}\label{Eq:Tangent_Space}
546:    \bigoplus_{\delta_i=1}\bigoplus_{\delta_j=0}
547:      \Hom(\cS_{a_i,b_i},\, \cO_{\P^1} )\cdot \bE_{ij} \quad \oplus\quad
548:    \bigoplus_{\delta_i=1}\bigoplus_{\delta_j=1}
549:      \Hom(\cS_{a_i,b_i},\,\cT_{a_j,b_j})\cdot \bE_{ij}\ .
550:  \end{equation}
551: %
552: \end{thm}
553: 
554: We now give $T$-bases (bases of $T$-eigenvectors) for these summands and
555: determine the corresponding weights.  
556: Fix a basis for the character group of $T$ as follows.
557: Extend the action of $T_{k^n}$ on $\cO^n_{\P^1}$ to 
558: $T$ by letting the factor $T_{\P^1}$ act trivially.  Then we
559: abuse notation and denote the character of $T$ acting on the
560: $i$th basis vector $\be_i$ by the same symbol $\be_i$.  Thus
561: the dual basis element $\hat{\be}_i$ has $T$-weight $-\be_i$.
562: 
563: 
564: Similarly, extend the action of $T_{\P^1}$ on $\P^1$ to
565: an action of 
566: $T$ by letting $T_{k^n}$ act trivially, and
567: denote by $\bff$ the character of $T$ corresponding to
568: the action on the dense orbit $\P^1 \setminus \{0,\infty\}$.
569: More precisely, we can let $T_{\P^1} \cong \C^*$ act on the
570: homogeneous coordinates
571: $k[x,y]$ of $\P^1$ by $q\cdot x = qx$ and $q\cdot y = y$.
572: Thus $T$ acts on the rational function $z:=x/y$ with
573: weight $\bff$, and on the monomial $z^a=x^ay^{-a}$ with weight $a\,\bff$.
574: 
575: The first sum of~\eqref{Eq:Tangent_Space} involves spaces 
576: of the form $\Hom(\cS_{a,b},\cO_{\P^1})=H^0(\cS^*_{a,b})$.
577: This space of sections has a monomial $T$-basis
578:  \[
579:    \{z^c\mid -a \le c \le b\}\,.
580:  \]
581: %
582: For example, $H^0(\cS^*_{1,2})=k\cdot\{z^{-1}, z^0=1, z, z^2\}$.
583: Thus if $\delta_i=1$ and $\delta_j=0$, then the piece
584: $\Hom(\cS_{a_i,b_i},\cO_{\P^1})\cdot\bE_{ij}$ of the tangent space
585: has a monomial $T$-basis
586: $z^c\cdot\bE_{ij}$ for all $-a_i\leq c\leq b_i$.
587: The basis element $z^c\cdot\bE_{ij}$ has $T$-weight 
588: %
589:  \begin{equation}\label{Eq:TI}
590:     \be_j-\be_i+ c\,\bff \,.
591:  \end{equation}
592: %
593: %  These give rise to type I curves
594: %
595: 
596: 
597: The second sum of~\eqref{Eq:Tangent_Space} involves spaces of the form
598: $\Hom(\cS_{a,b},\cT_{\alpha,\beta})$. 
599:   Since 
600: $\cT_{\alpha,\beta}$ is a skyscraper sheaf supported at $0$ and
601: $\infty$, a map $\phi\in\Hom(\cS_{a,b},\cT_{\alpha,\beta})$ is determined  
602: by its actions at $0$ and at $\infty$.
603: At $0$, $z = x/y$ is a local parameter, so the map $\phi$
604: becomes
605: %
606:   \[
607:     \phi\ \colon\ z^a\C[z] \rightarrow \C[z]/{\langle z^\alpha \rangle}\,,
608:   \]
609: %
610: and thus has the form $z^{-a}f(z)$ where $f(z)$ has degree
611: less than $\alpha$.
612: At $\infty$, $z^{-1}$ is a local parameter, and
613: the map $\phi$ has the form $z^{b}g(z^{-1})$ where $g(z)$ is a
614: polynomial of degree less than $\beta$.
615: Thus $\Hom(\cS_{a,b},\cT_{\alpha,\beta})$ has the monomial $T$-basis
616: \[
617:    k\cdot \{z^{(\alpha-c)-a}\mid 1\leq c\leq\alpha\}\; \oplus \;
618:    k\cdot \{z^{b-(\beta-c)}\mid  1\leq c\leq\beta\}\,,
619: \]
620: where elements in the first summand act by zero on the stalk at
621: $\infty$, and elements of the second summand act by zero at $0$.
622: 
623: Thus, if $\delta_i=\delta_j=1$, then the summand
624: $\Hom(\cS_{a_i,b_i},\cT_{a_j,b_j})\cdot\bE_{ij}$  
625: of \eqref{Eq:Tangent_Space} has a monomial $T$-basis
626: %
627:  \[
628:    \{ z^{(a_j-c)-a_i}\cdot\bE_{ij} \mid 1\leq c\leq a_j \}\, \cup \,
629:    \{ z^{(b_j-c)-b_i}\cdot\bE_{ij} \mid 1\leq c\leq b_j\}
630:  \]
631: %
632: with corresponding $T$-weights
633: %
634:  \begin{equation}\label{Eq:TII_III}
635:    \be_j-\be_i + ((a_j-c) - a_i)\bff \quad\mbox{ and }\quad 
636:    \be_j-\be_i + (b_i - (b_j-c))\bff \,.
637:  \end{equation}
638: %
639: % These give curves of type III if i=j and type II if i\neq j
640: %
641: 
642: 
643: We note that this discussion gives a basis for $T_{(\sdelta,\va,\vb)}\cQ_d$
644: consisting of
645: %
646:  \begin{eqnarray*}
647:   \sum_{\delta_j=0}\sum_{\delta_i=1} (a_i+b_i+1)
648:   \quad +\quad  
649:   \sum_{\delta_i=1}\sum_{\delta_j=1} (a_j+b_j)
650:   &=& (n-r)\cdot(d+r) + r\cdot d\\
651:   &=& r\cdot(n-r) + nd\ =\ \dim\cQ_d
652:  \end{eqnarray*}
653: %
654: elements, which shows that $\cQ_d$ is smooth at the $T$-fixed point
655: $\cS_{(\sdelta,\va,\vb)}$, and hence everywhere. 
656: In the next section, we will describe $T$-invariant curves
657: in $\cQ_d$ incident on $\cS_{(\sdelta,\va,\vb)}$ whose
658: tangent directions at $\cS_{(\sdelta,\va,\vb)}$ coincide with this given
659: $T$-basis of $T_{(\sdelta,\va,\vb)}\cQ_d$.
660: 
661: 
662: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
663: 
664: \begin{example}\label{example_one}
665: {\rm 
666:  The  quot scheme $\cQ:=\cQ_2(0,2)$ of rank 0 and degree 2
667:  quotients of $\cO^2_{\P^1}$ has dimension $r(n{-}r)+dn=2\cdot 0+2\cdot 2=4$.
668:  Note that the associated Grassmannian is a point. 
669:  Since $r=0$, the index $\vdelta$ is the same for each fixed point,
670:  $\delta_1=\delta_2=1$,  and so the fixed points are indexed by quadruples
671:  $(a_1,a_2,b_1,b_2)$ of non-negative integers whose sum is 2. 
672:  Thus there are ten fixed points. 
673:  We represent the fixed point $(a_1,a_2,b_1,b_2)$ by two
674:  columns of boxes superimposed on a horizontal line where the $i$th
675:  column has $a_i$ boxes above the horizontal line and $b_i$ boxes below it.
676:  For example, 
677: \[
678:   \raisebox{-7.5pt}{\includegraphics[width=25pt]{figures/11.20.00.eps}}\ 
679:       \leftrightarrow\ (2,0,0,0)\,,\quad
680:   \raisebox{-5pt}{\includegraphics[width=25pt]{figures/11.11.00.eps}}\ 
681:       \leftrightarrow\ (1,1,0,0)\,,
682:    \qquad\mbox{and}\qquad
683:   \raisebox{-7pt}{\includegraphics[width=25pt]{figures/11.10.01.eps}}\ 
684:       \leftrightarrow\ (1,0,0,1)\,.
685: \]
686: 
687:  The fixed point corresponding to 
688:  \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.20.00.eps}}
689:  is the exact sequence of sheaves on $\P^1$
690: \[ 
691:    \begin{array}{ccccccccccc}
692:     &&\cO(-2)\cdot\be_1&\xrightarrow{\ x^2\ }&\cO\cdot\be_1
693:       &\longrightarrow &\cO/\frakm_0^2\cdot\be_1\\
694:    0&\longrightarrow &\oplus&&\oplus&&\oplus&\longrightarrow &0\,,\\
695:     &&\cO\cdot\be_2&\xrightarrow{\ \,1\,\ }&\cO\cdot\be_2&\longrightarrow& 0
696:    \end{array}
697: \]
698:  where $\cO=\cO_{\P^1}$.
699:  The tangent space at this fixed point is the sum of the two 2-dimensional
700:  $T$-invariant spaces of homomorphisms having the indicated $T$-bases.
701: %
702:  \begin{eqnarray*}
703:     \Hom(\cO, \cO/\frakm_0^2)\cdot\bE_{21}& =&
704:      k\{1, z\}\cdot\bE_{21}\\
705:     \Hom(x^2\cO, \cO/\frakm_0^2)\cdot\bE_{11}& =&
706:     k\{z^{-2},z^{-1}\}\cdot\bE_{11}\
707:  \end{eqnarray*}
708: %
709:  As before, $z:=x/y$ is a local parameter at $0$ and $z^{-1}$ is a local
710:  parameter at $\infty$.
711:  These basis elements have four distinct $T$-weights
712: \[
713:    \be_1-\be_2,\ \be_1-\be_2+\bff,\ \mbox{and}\  -\bff,\ -2\bff\,.
714: \]
715: 
716:  The fixed point corresponding to 
717:  \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.10.01.eps}}
718:  is the exact sequence of sheaves on $\P^1$,
719: \[ 
720:    \begin{array}{ccccccccccc}
721:     &&\cO(-1)\cdot\be_1&\xrightarrow{\ x\ }&\cO\cdot\be_1
722:       &\longrightarrow&\cO/\frakm_0\cdot\be_1\\
723:    0&\longrightarrow&\oplus&&\oplus&&\oplus&\longrightarrow&0\,.\\
724:     &&\cO(-1)\cdot\be_2&\xrightarrow{\ y\ }&\cO\cdot\be_2
725:       &\longrightarrow&\cO/\frakm_\infty\cdot\be_2
726:    \end{array}
727: \]
728:  The tangent space at this fixed point is the sum of four 1-dimensional
729:  $T$-invariant spaces of homomorphisms having bases and weights as
730:  indicated.
731: \[
732:   \begin{array}{|l|l|r|}\hline
733:     \mbox{$T$-eigenspace} & \mbox{basis} &\mbox{$T$-weight}\\ \hline
734:     \Hom(x\cO,\cO/\frakm_0)\cdot\bE_{11}&  z^{-1}\cdot\bE_{11}&
735:              -\bff\ \raisebox{-3pt}{\rule{0pt}{15pt}} \\ \hline
736:     \Hom(x\cO,\cO/\frakm_\infty)\cdot\bE_{12}& 1\cdot\bE_{12}&
737:             \be_2-\be_1 \raisebox{-3pt}{\rule{0pt}{15pt}}  \\\hline
738:     \Hom(y\cO,\cO/\frakm_0)\cdot\bE_{21}& 1\cdot\bE_{21}&
739:               \be_1-\be_2 \raisebox{-3pt}{\rule{0pt}{15pt}} \\ \hline
740:     \Hom(y\cO,\cO/\frakm_\infty)\cdot\bE_{22}& z\cdot\bE_{22}&
741:               \bff\ \raisebox{-3pt}{\rule{0pt}{15pt}} \\\hline
742:   \end{array}
743: \]
744: 
745: }
746: \end{example}
747: 
748: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
749: \section{$T$-invariant curves} 
750: \label{Tcurves}
751: 
752: We describe a collection of $T$-invariant curves whose
753: tangent directions at a $T$-fixed point $(\vdelta,\va,\vb)$ form a basis for
754: $T_{(\sdelta,\va,\vb)}\cQ_d$.
755: A $T$-invariant curve on $\cQ_d$ is, by definition, a 
756: flat family $\cS\to\P^1$ with a $T$-action whose
757: fibre $\cS(s,t)$ over a point $[s,t]\in\P^1$ is a free subsheaf of
758: $\cO^n_{\P^1}$ of rank $n-r$ and degree $-d$.
759: We exhibit each curve as a subsheaf $\cS$ of $\cO^n_{\P^1\times\P^1}$ with a
760: $T$-action that has rank $n-r$ and degree $(-d,-1)$ with
761: $\cS(1,0)=\cS_{(\sdelta,\va,\vb)}$ and $\cS(0,1)=\cS_{(\sdelta',\va',\vb')}$, where
762: $(\vdelta',\va',\vb')$ is some other $T$-fixed point on $\cQ_d$.
763: Then $\cS$ defines a $T$-invariant curve on $\cQ_d$ connecting the two fixed
764: points. 
765: 
766: Each sheaf $\cS$ has one of three types: I, II, or III.
767: We describe them below and then argue that they have the desired
768: properties.  
769: We write $\cO$ for $\cO_{\P^1\times\P^1}$.
770: We use the correspondence between sheaves over $\cO$ and saturated modules over
771: the bihomogeneous coordinate ring $k[x,y][s,t]$ of
772: $\P^1_{[x,y]}\times\P^1_{[s,t]}$.
773: Then $\cO^n$ is the rank $n$ free module with basis $\be_1,\be_2,\dotsc,\be_n$.
774: \medskip
775: 
776: \noindent{\bf Type I:}
777: Let $(\vdelta',\va',\vb')$ be another fixed point where the data
778: $(\vdelta,\va,\vb)$ and $(\vdelta,\va',\vb')$ agree except in positions $i$ and
779: $j$, with $\delta_i=\delta'_j=1$ and $\delta_j=\delta'_i=0$.
780: Note that $a_i+b_i=a'_j+b'_j$.  
781: Let $\cS$ be the subsheaf of $\cO^n$ which agrees 
782: with both $\cS_{(\sdelta,\va,\vb)}$ and $\cS_{(\sdelta',\va',\vb')}$
783: except for its
784: component in  $\cO\cdot\be_i+\cO\cdot\be_j$, where it is the rank 1 and
785: degree $-(a_i{+}b_i)$ subsheaf  generated by the single element
786: %
787:  \begin{equation}\label{Eq:typeI}
788:    s x^{a_i}y^{b_i}\cdot\be_i + t x^{a'_j}y^{b'_j}\cdot\be_j \, .\smallskip
789:  \end{equation}
790: 
791: 
792: \noindent{\bf Type II:}
793: Let $(\vdelta',\va',\vb')$ be another $T$-fixed point where
794: $\vdelta=\vdelta'$, $\vb=\vb'$, and the data $\va$ and $\va'$ agree except in
795: positions $i$ and $j$ with $i\neq j$.
796: We suppose that $i$ and $j$ have been chosen so that $a_i<a'_i$.
797: Then $a_j>a'_j$ and $c:=a'_i-a_i=a_j-a'_j>0$.
798: Set $\gamma:=a_i+b_i+c-a_j-b_j$.
799: 
800: If $\gamma\geq 0$, let $\cS$ be the subsheaf of $\cO^n$ which 
801: agrees with both $\cS_{(\sdelta,\va,\vb)}$ and $\cS_{(\sdelta',\va',\vb')}$
802: except for its components in $\cO\cdot\be_i+\cO\cdot\be_j$,
803: where it is the subsheaf
804: generated by 
805: %
806:  \begin{equation}\label{Eq:typeII}
807:    x^{a_i+c}y^{b_i}\cdot\be_i\,,\quad
808:    x^{a_j}y^{b_j}\cdot\be_j\,,\quad \mbox{ and }\quad
809:    s x^{a_i}y^{b_i}\cdot\be_i + t x^{a_j-c}y^{b_j+\gamma}\cdot\be_j \, .
810:  \end{equation}
811: %
812: When $\gamma<0$, replace the third generator by 
813: $s x^{a_i}y^{b_i-\gamma}\cdot\be_i + t x^{a_j-c}y^{b_j}\cdot\be_j$.
814: 
815: The remaining sheaves of type II are obtained by interchanging the roles of
816: $\va$ and $\vb$.
817: That is, $\vdelta=\vdelta'$, $\va=\va'$ and $\vb,\vb'$ agree except in positions 
818: $i\neq j$, and we further have that $c:=b'_i-b_i=b_j-b'_j>0$.
819: Set $\gamma:=a_i+b_i+c-a_j-b_j$.
820: If $\gamma\geq 0$, let $\cS$ be as before, except with the
821: generators~\eqref{Eq:typeII} replaced by 
822: %
823:  \begin{equation}\label{Eq:typeII-b}
824:    x^{a_i}y^{b_i+c}\cdot\be_i\,,\quad
825:    x^{a_j}y^{b_j}\cdot\be_j\,,\quad\mbox{ and }\quad
826:    s x^{a_i}y^{b_i}\cdot\be_i + t x^{a_j+\gamma}y^{b_j-c}\cdot\be_j \, .
827:  \end{equation}
828: %
829: If $\gamma<0$, then the third generator will be 
830: $s x^{a_i-\gamma}y^{b_i}\cdot\be_i + t x^{a_j}y^{b_j-c}\be_j$.
831: \smallskip
832: 
833: 
834: \noindent{\bf Type III:}
835: Let $(\vdelta',\va',\vb')$ be another $T$-fixed point where
836: $\vdelta=\vdelta'$ and the data $(\va,\vb)$ and $(\va',\vb')$ agree except in
837: position $i$.  Thus $\delta_i\neq 0$ and $a_i+b_i=a'_i+b'_i$.
838: Let $\cS$ be the subsheaf of $\cO^n$ which 
839: agrees with both $\cS_{(\sdelta,\va,\vb)}$ and
840: $\cS_{(\sdelta',\va',\vb')}$, except for
841: its component in $\cO\cdot\be_i$, where it is the 
842: rank 1 and degree $-(a_i+b_i)$ subsheaf generated by
843: %
844:  \begin{equation}\label{Eq:typeIII}
845:    s x^{a_i}y^{b_i}\cdot \be_i + t x^{a'_i}y^{b'_i}\cdot\be_i\,.\smallskip
846:  \end{equation}
847: %
848: 
849: 
850: \begin{thm}\label{Th:Curves_are_correct}
851:  The subsheaves $\cS$ of\/ $\cO^n_{\P^1\times\P^1}$ of types {\rm I}, 
852:  {\rm II}, and\/ {\rm III} are $T$-invariant
853:  and free of rank $n{-}r$ and degree $(-d,-1)$.
854:  They satisfy  $\cS(1,0)=\cS_{(\sdelta,\va,\vb)}$ and 
855:  $\cS(0,1)=\cS_{(\sdelta',\va',\vb')}$, and hence define $T$-invariant
856:  curves on $\cQ_d$.
857: \end{thm}
858: 
859: \begin{proof}
860:   The generators of $\cS$ are $T$-invariant, except for those described
861:  by~\eqref{Eq:typeI}, \eqref{Eq:typeII}, \eqref{Eq:typeII-b},
862:  and~\eqref{Eq:typeIII}.  
863:  But $T$ acts transitively on those generators for $s\cdot t\neq0$.
864: Therefore, each sheaf $\cS$ is $T$-invariant.
865: In all cases, $\cS$ has degree $-1$ with respect to $\P^1_{[s,t]}$.
866: 
867:  The theorem is clear for the sheaves of types I and
868:  III, as they are constant on $\P^1_{[s,t]}$, except for the rank 1
869:  components~\eqref{Eq:typeI} and~\eqref{Eq:typeIII}, each of which has
870:  degree $(-(a_i+b_i),-1)$.
871:  Specializing these generators at $[s,t]=[1,0]$ and
872:  $[0,1]$ shows that $\cS(1,0)=\cS_{(\sdelta,\va,\vb)}$ and 
873:  $\cS(0,1)=\cS_{(\sdelta',\va',\vb')}$.
874: 
875:  We use a Gr\"obner basis argument for the sheaves of type II.
876:  The Hilbert function for a submodule $M$ of $\cO^2$ equals the Hilbert function
877:  for the module of leading terms of any Gr\"obner basis of $M$.
878:  As explained in~\cite[Ch.~15]{E95}, a weight $\omega$ selecting these leading
879:  terms induces a $\G_m$-action on $\cO^2$ whose 
880:  restriction to the Gr\"obner basis of $M$ generates a flat
881:  family over $\mathbb{A}^1$ whose special fibre is the module of
882:  leading terms. 
883: 
884:  For now, set $s=t=1$.
885:  If $\be_i>\be_j$, then the 
886:  generators~\eqref{Eq:typeII} form a Gr\"obner basis for any
887:  position-over-monomial ordering, and the third generator
888:  has leading term $x^{a_i}y^{b_i}\cdot\be_i$.
889:  As $c>0$, the module of leading terms is generated by
890:  $x^{a_i}y^{b_i}\cdot\be_i$ and $x^{a_j}y^{b_j}\cdot\be_j$, and so
891:  it has rank 2 and degree $-(a_i{+}b_i{+}a_j{+}b_j)$.
892:  The weight $\omega$ with $\omega(\be_i)=0$ and $\omega(\be_j)=-1$ induces the
893:  leading terms and has corresponding $\G_m$-action
894:  $t.(\be_i,\be_j)=(\be_i,t\be_j)$, for $t\in\G_m$.
895:  This action on $\cS(1,1)$ is the flat family of modules over $\mathbb{A}^1$
896:  generated by 
897: \[
898:    x^{a_i+c}y^{b_i}\cdot\be_i\,,\quad
899:    x^{a_j}y^{b_j}\cdot\be_j\,,\quad\mbox{and}\quad 
900:    x^{a_i}y^{b_i}\cdot\be_i + t x^{a_j-c}y^{b_j+\gamma}\cdot\be_j \ ,
901: \]
902:  which is just the part of $\cS$ in
903:  $\cO\cdot\be_i+\cO\cdot\be_j$ restricted to the affine subset $U$ of
904:  $\P^1_{[s,t]}$ where $s\neq 0$.
905:  Thus $\cS|_U$ is a flat family over $U$ of free subsheaves of $\cO^n_{\P^1}$
906:  of rank $n{-}r$ and degree $-d$, and $\cS(1,0)=\cS_{(\sdelta,\va,\vb)}$.
907: 
908:  When $s=t=1$ the generators~\eqref{Eq:typeII} form a Gr\"obner basis when
909:  $\be_i<\be_j$, where the third generator has leading term
910:  $x^{a_j-c}y^{b_j+\gamma}\cdot\be_j$. 
911:  The module of leading terms is generated by 
912: \[
913:    x^{a_i+c}y^{b_i}\cdot\be_i\,,\quad
914:    x^{a_j}y^{b_j}\cdot\be_j\,,\quad\mbox{and}\quad 
915:    x^{a_j-c}y^{b_j+\gamma}\cdot\be_j \ .
916: \]
917:  Since $\gamma\geq 0$ and $c>0$, saturating the ideal of $k[x,y]$ 
918:  generated by $x^{a_j}y^{b_j}$ and $x^{a_j-c}y^{b_j+\gamma}$ by the irrelevant
919:  maximal ideal generated by $x$ and $y$ gives
920:  the ideal generated by $x^{a_j-c}y^{b_j}$. 
921:  Thus the module of leading terms is generated by 
922: \[
923:    x^{a_i+c}y^{b_i}\cdot\be_i\,,\quad\mbox{and}\quad 
924:    x^{a_j-c}y^{b_j}\cdot\be_j \ .
925: \]
926:  As before, restricting $\cS$ to the affine set of points $[s,t]$ of $\P^1$ where 
927:  $t\neq 0$ gives a flat family of subsheaves of $\cO^n_{\P^1}$ of rank $n{-}r$
928:  and degree $-d$ with special fibre $\cS(0,1)=\cS_{(\sdelta',\va',\vb')}$. 
929:  The same arguments suffice for the module generated by~\eqref{Eq:typeII-b}.
930: \end{proof}
931: 
932: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
933: %
934: %    Frank Changed the statement of this Theorem
935: %
936: %
937: \begin{thm} \label{tangentdirections}
938:   For any $T$-fixed point  $\cS_{(\sdelta, \va, \vb)}$ in $\cQ_d$, 
939:   the set of tangent directions to the $T$-invariant curves induced 
940:   by the sheaves $\cS$ of types {\rm I}, {\rm II}, and\/ {\rm III}
941:   corresponds to the $T$-basis of $T_{(\sdelta, \va, \vb)}\cQ_d$ defined
942:   in Section~$\ref{fixed_points}$, and this correspondence is $T$-equivariant,
943:   respecting the weights.
944:   More specifically, at the $T$-fixed point $\cS_{(\sdelta, \va, \vb)}$,
945: \begin{enumerate}
946:  \item[I.]  The weight of the type {\rm I} curve~$\eqref{Eq:typeI}$ is
947:      \[    \be_j-\be_i+(a'_j-a_i)\bf\,,\]
948:      and such curves correspond to the first summand of~$\eqref{Eq:Tangent_Space}$.
949: 
950:  \item[II.] The weight of the type {\rm II} curve~$\eqref {Eq:typeII}$ is 
951:     \[   \be_j-\be_i + (a'_j-a_i)\bf\,,\]
952:      the weight of the type  {\rm II} curve~$\eqref {Eq:typeII-b}$ is 
953:     \[   \be_j-\be_i + (b_i-b'_j)\bf\,,\]
954:       and such curves correspond to the second summand
955:       of~$\eqref{Eq:Tangent_Space}$ when $i\neq j$.
956:   
957:  \item[III.]  The weight of the type {\rm III} curve~$\eqref{Eq:typeIII}$ is
958:      \[    (a'_j-a_i)\bf\,,\]
959:       and such curves correspond to the second summand
960:       of~$\eqref{Eq:Tangent_Space}$ when $i= j$.
961: 
962: \end{enumerate}
963:   
964: \end{thm}
965: 
966: \begin{proof}
967:  In each of these curves, the $T$-fixed point $\cS_{(\sdelta,\va,\vb)}$ is the point
968:  where $s=1$ and $t=0$.
969:  In what follows, we work locally, setting $s=1$.
970:  
971:  Note that the generator~\eqref{Eq:typeI} of a type I sheaf may be rewritten
972: \[
973:     x^{a_i}y^{b_i}\cdot
974:     ( \be_i + t\cdot x^{a'_j-a_i}y^{b'_j-b_i}\bE_{ij}(\be_i))\,.
975: \]
976:  This shows that the tangent space to the curve when $t=0$ is the $T$-basis
977:  element $x^{a'_j-a_i}y^{b'_j-b_i}\bE_{ij}$ of 
978:  $\Hom(\cS_{a_i,b_i}\cdot\be_i,\, \cO_{\P^1}\cdot\be_j)$.
979:  Thus the tangent spaces of type I curves at $\cS_{(\sdelta,\va,\vb)}$ span the
980:  component of $T_{(\sdelta,\va,\vb)}\cQ_d$ given by the first summand
981:  of~\eqref{Eq:Tangent_Space}. 
982:  (Recall that in type I, we have $\delta_i=\delta'_j=1$ and $\delta_j=\delta'_i=0$.)
983: 
984:  A similar analysis shows that the tangent space at $t=0$ of the type III curve
985:  defined by~\eqref{Eq:typeIII} is spanned by $x^{a'_i-a_i}y^{b'_i-b_i}\bE_{ii}$,
986:  and so the tangent spaces  at $\cS_{(\sdelta,\va,\vb)}$ of type III curves span the
987:  component of $T_{(\sdelta,\va,\vb)}\cQ_d$ given by the second summand
988:  of~\eqref{Eq:Tangent_Space} when $i=j$.
989:  
990: 
991:  For a curve of type II, note that the family of sheaves described
992:  by~\eqref{Eq:typeII} is constant in a neighborhood of $\infty$.
993:  In a neighborhood of $0$, it is given by
994: \[
995:    x^{a_i+c}\be_i,\quad x^{a_j}\be_j,\quad\mbox{ and }\quad 
996:    x^{a_i}\cdot\bigl(\be_i + t x^{(a_j-c)-a_i}\bE_{ij}(\be_i)\bigr)\,.
997: \]
998:  Thus $x^{(a_j-c)-a_i}\bE_{ij}$ spans the tangent space at $t=0$.
999:  A similar argument near $\infty$ for the sheaves described
1000:  by~\eqref{Eq:typeII-b} shows that the tangent spaces of type II curves at
1001:  $\cS_{(\sdelta,\va,\vb)}$ span the component of
1002:  $T_{(\sdelta,\va,\vb)}\cQ_d$ given by the 
1003:  second summand of~\eqref{Eq:Tangent_Space} when $i\neq j$.
1004: \end{proof}
1005: 
1006: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1007: 
1008: A {\em moment graph} of a $T$-variety is a graph 
1009: whose vertices correspond to $T$-fixed points and whose edges correspond to
1010: $T$-invariant curves, embedded into $\R\otimes\Hom(T,\Z)$ so that the edge
1011: corresponding to a $T$-invariant curve is parallel to 
1012: the weight of the action of $T$ on the curve.   
1013: More specifically, if $C$ is a $T$-invariant curve joining fixed points $p$ and
1014: $q$, then the edge from $p$ to $q$ in the moment graph is a positive 
1015: multiple of the $T$-weight of $T_pC$.
1016: When $k=\C$ and we fix a K\"ahler form, there is a moment map
1017: $\mu\colon\cQ_d\to {\mathfrak t}^*$ and the image of the
1018: $T$-fixed points and $T$-invariant curves is a moment graph.
1019: 
1020: When there are finitely many $T$-invariant curves, the Goresky-Kottwitz-MacPherson
1021: method to compute equivariant cohomology is conveniently expressed in 
1022: terms of a moment graph, with one relation for each edge.  
1023: When there are infinitely many $T$-invariant curves, there are additional 
1024: relations coming from families of $T$-invariant curves, so it is 
1025: better to work with the moment {\em multigraph\/}, where each family of 
1026: $T$-invariant curves (which will appear as a connected component of parallel
1027: edges in the moment graph) is considered to form a single multiedge with
1028: more than $2$ vertices, given by the fixed points in the closure of the 
1029: family.  
1030: To have a structure which determines the equivariant cohomology
1031: or Chow groups, we should label each multiedge with the topological
1032: type of the corresponding family.
1033: 
1034: Guillemin and Zara \cite{GZ01,GZ02,GZ03} have explored the combinatorial
1035: properties of moment graphs.
1036: 
1037: 
1038: 
1039: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1040: \begin{example}\label{example_two}
1041: {\rm 
1042: 
1043: Figure~\ref{F:Q202} represents a moment multigraph of $\cQ_2(0,2)$.
1044: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1045: \begin{figure}[htb]
1046: \[
1047:   \begin{picture}(210,220)(0,-10)
1048: %fig2dev -Leps -m0.8 Q202.skeleton.fig Q202.skeleton.eps
1049:    \put( 40, 10){\includegraphics[height=180pt]{figures/Q202.skeleton.eps}}
1050:    \put( 10,140){\includegraphics[width=25pt]{figures/11.20.00.eps}}
1051:    \put( 88,193){\includegraphics[width=25pt]{figures/11.11.00.eps}}
1052:    \put(168,140){\includegraphics[width=25pt]{figures/11.02.00.eps}}
1053:    \put( 10, 90){\includegraphics[width=25pt]{figures/11.10.10.eps}}
1054:    \put(168, 90){\includegraphics[width=25pt]{figures/11.01.01.eps}}
1055:    \put( 10, 39){\includegraphics[width=25pt]{figures/11.00.20.eps}}
1056:    \put( 88, -8){\includegraphics[width=25pt]{figures/11.00.11.eps}}
1057:    \put(168, 39){\includegraphics[width=25pt]{figures/11.00.02.eps}}
1058:    \put(108, 78){\includegraphics[width=25pt]{figures/11.10.01.eps}}
1059:    \put( 70,108){\includegraphics[width=25pt]{figures/11.01.10.eps}}
1060: 
1061:    \thicklines
1062: 
1063:    \put(-20,5){\vector(1,0){50}} \put(-12,-5){$\be_1{-}\be_2$}
1064:    \put(-20,5){\vector(0,1){30}} \put(-30, 5){$\bff$}
1065:   
1066:   \end{picture}
1067: \]
1068: \caption{A moment multigraph of $\mathcal{Q}_2(0,2)$}
1069: \label{F:Q202}
1070: \end{figure}
1071: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1072: Since the $T$-fixed points  
1073: \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.10.01.eps}}
1074: and 
1075: \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.01.10.eps}}
1076: have the same image in this and in any moment multigraph, 
1077: we displace their images from their true positions for clarity. 
1078: Similarly, some images of  $T$-invariant curves are displaced or
1079: drawn as arcs.
1080: 
1081: The $T$-basis of the tangent space at  
1082: \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.20.00.eps}}
1083: has weights $\{ \be_1-\be_2, \be_1-\be_2+\bff, -\bff, -2\bff\}$.
1084: These correspond to the following four $T$-invariant curves:
1085: \[
1086:   \begin{array}{|l|c|c|c|c|}\hline
1087:    \mbox{Submodules of $\be_1\cO\oplus\be_2\cO$}&
1088:     t=0&s=0&\mbox{Weight}&\mbox{Type}\\\hline 
1089:     x^2\be_1, \ x^2\be_2,\ s\be_1+t\be_2& x^2\be_1,\ \be_2 
1090:       \rule{0pt}{12pt}  & \be_1,\ x^2\be_2& \be_1-\be_2& {\rm II}\\\hline
1091:     x^2\be_1,\ x^2\be_2,\ sx\be_1+ty\be_2&x^2\be_1,\ \be_2 & x\be_1,\ x\be_2
1092:       \rule{0pt}{12pt}  & \be_1-\be_2+\bff&{\rm II}\\\hline
1093:     (sx^2+txy)\be_1,\  \be_2 &x^2\be_1,\ \be_2 
1094:       \rule{0pt}{12pt}  & xy\be_1 ,\ \be_2&-\bff&{\rm III}\\\hline
1095:     (sx^2+ty^2)\be_1,\ \be_2 &x^2\be_1,\ \be_2 
1096:      \rule{0pt}{12pt}   & y^2\be_1,\ \be_2&-2\bff&{\rm III}\\\hline
1097:    \end{array}
1098: \]
1099: The $T$-basis to the tangent space at 
1100: \raisebox{-5pt}{\includegraphics[width=20pt]{figures/11.10.01.eps}}
1101: has weights $\{ \pm\bff, \pm(\be_2-\be_1)\}$.
1102: These correspond to the following four $T$-invariant curves:
1103: \[
1104:   \begin{array}{|l|c|c|c|c|}\hline
1105:    \mbox{Submodules of $\be_1\cO\oplus\be_2\cO$}&
1106:     t=0&s=0&\mbox{Weight}&\mbox{Type}\\ \hline
1107:     (sx+ty)\be_1,\ y\be_2 &x\be_1,\ y\be_2 & y\be_1,\ y\be_2
1108:        &-\bff&{\rm III}\\\hline
1109:     xy\be_1,\ y\be_2,\ sx\be_1+tx\be_2& x\be_1,\ y\be_2 
1110:                        & xy\be_1,\ \be_2& \be_2-\be_1&{\rm II}\\\hline
1111:     x\be_1,\ xy\be_2,\ ty\be_1+sy\be_2&x\be_1,\ y\be_2 & \be_1,\ xy\be_2
1112:                        & \be_1-\be_2&{\rm II}\\\hline
1113:     x\be_1,\ (sy+tx)\be_2 &x\be_1,\ y\be_2 & x\be_1,\ x\be_2
1114:        &\bff&{\rm III}\\\hline
1115:    \end{array}
1116: \] 
1117:  
1118: }\end{example}
1119: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1120: 
1121: 
1122: \section{Families of $T$-invariant curves}
1123: \label{Tfamilies}
1124: 
1125: 
1126: Suppose that $Z$ is a $T$-invariant curve on $\cQ_d$.
1127: Let $T'$ be the identity component of the stabilizer of $Z$ in $T$, a
1128: codimension one subtorus of $T$. 
1129: Then the action of $T$ on $Z$ factors through the quotient
1130: \[
1131:    T\ \longrightarrow\ T/T'\ \simeq\ \G_m\,.
1132: \]
1133: This composition $\chi\colon T\to \G_m$ is a primitive weight parallel to the
1134: weight of the action of $T$ on $Z$.
1135: Let $Y$ be the component of the $T'$-fixed point locus $\cQ_d^{T'}$ which
1136: contains $Z$.
1137: Then $Y\setminus Y^T$ is foliated by one-dimensional orbits of $T$
1138: whose closures are $T$-invariant curves.
1139: We call $Y$ the family of $T$-invariant curves on $\cQ_d$ which contains $Z$.
1140: If $p\in Y^T\subset\cQ_d^T$ is a $T$-fixed point of $Y$, then $T_pY$ is a
1141: $T$-invariant linear subspace of $T_p\cQ_d$ which is fixed pointwise by $T'$.
1142: In particular, all weights of the $T$-action on $T_pY$ are multiples of $\chi$. 
1143: 
1144: To classify families of $T$-invariant curves, we first 
1145: determine which $T$-weights of $T_p\cQ_d$ are parallel.
1146: If the weight of a curve is not parallel to the weight of any other curve, it is
1147: isolated. 
1148: By Theorem~\ref{tangentdirections}, it suffices to identify those $T$-invariant
1149: curves incident on a fixed point with parallel $T$-weights. 
1150: 
1151: \begin{thm}\label{T:parallel}
1152:   Two $T$-invariant curves $\cS$, $\cS'$ of types {\rm I}, {\rm II}, 
1153:   or\/ {\rm III} containing the fixed point $\cS_{(\sdelta,\va,\vb)}$ have
1154:   parallel $T$-weights if and only if either 
1155:  \begin{enumerate}
1156:   \item both $\cS$ and $\cS'$ have type {\rm III} or
1157:   \item both $\cS$ and $\cS'$ have type {\rm II}, and 
1158:    \begin{enumerate}
1159:     \item $\cS$  connects $\cS_{(\sdelta,\va,\vb)}$ to $\cS_{(\sdelta,\va',\vb)}$,
1160:     \item $\cS'$ connects $\cS_{(\sdelta,\va,\vb)}$ to $\cS_{(\sdelta,\va,\vb')}$,
1161:     \item $\va$ and $\va'$ agree except in positions $i$ and $j$ with 
1162: $i\neq j$; $\vb$ and $\vb'$  agree except in positions $i$ and $j$ (same $i,j$);
1163:  and $a_i+b_i=a'_j+b'_j$.
1164: \iffalse
1165:      \begin{itemize}
1166:        \item[--] $\va$ and $\va'$ agree except in positions $i$ and $j$ with $i\neq j$,
1167:        \item[--] $\vb$ and $\vb'$  agree except in positions $i$ and $j$ (same $i,j$), and 
1168:        \item[--] $a_i+b_i=a'_j+b'_j$.
1169:      \end{itemize}
1170: \fi 
1171:    \end{enumerate}
1172:  \end{enumerate}
1173: \end{thm}
1174: 
1175: \begin{proof}
1176:  For (1), note that the weight  of $T$ on a type III curve is parallel to
1177:  $\bff$.
1178:  
1179:  If a curve does not have type III, then its weight has the form
1180:  $\be_j-\be_i+c\,\bff$, where $\delta_i=1$ and either $\delta_j=0$ if it has type
1181:  I or $\delta_j=1$ if it has type II.
1182:  Thus $\cS$ and $\cS'$ have the same weight and type, $\vdelta=\vdelta'$, and
1183:  the indices $i$ and $j$ in their definitions coincide.
1184: 
1185:  Weights of curves of types I and II correspond to~\eqref{Eq:TI}
1186:  and~\eqref{Eq:TII_III}, respectively. 
1187:  Inspecting~\eqref{Eq:TI}, shows that no two curves of type I can have the same weight.
1188:  Inspecting~\eqref{Eq:TII_III} reveals that either a given curve $\cS$ of type
1189:  II has a unique weight, or else there is exactly one other type II curve $\cS'$ with
1190:  the same weight, and the two curves are as described in the statement of the
1191:  theorem. 
1192: \end{proof}
1193: 
1194: We show that all type III curves at a fixed point lie in a single family of
1195: $T$-invariant curves, and if two type II curves
1196: have the same weight then they lie in a 2-dimensional family.
1197: Together with the isolated $T$-invariant curves,
1198: this shows that the tangent spaces at a given fixed point
1199: $T$ to families of $T$-invariant curves are the subspaces of $T_p\cQ_d$
1200: which are stabilized by codimension 1 subtori of $T$.
1201: It follows that these families contain all $T$-invariant curves in $\cQ_d$.
1202: \medskip
1203: 
1204: \noindent{\bf Vertical families.}  
1205: A \emph{vertical family} containing $\cS_{(\sdelta,\va,\vb)}$ is parametrized by
1206: the product of projective spaces 
1207: \[
1208:     \prod_{\delta_i=1} \P H^0(\cO(a_i+b_i))\ \simeq\ 
1209:     \prod_{\delta_i=1} \P^{a_i+b_i}\,.
1210: \]  
1211: It contains exactly the fixed points $\cS_{(\sdelta',\va',\vb')}$ where
1212: $\vdelta=\vdelta'$, and $\va'+\vb'=\va+\vb$, along with all
1213: type III curves which connect them.  
1214: These include all type III curves at each of these fixed points.
1215: 
1216: Consider the family $\cS$ of submodules of $\cO^n_{\P^1}$ generated by
1217: \[
1218:    \{  \be_i s_i  \mid \delta_i=1,\quad s_i\in H^0(\cO(a_i+b_i))\}\,.
1219: \]
1220: The base of this family is $\prod_{\delta_i=1} \P H^0(\cO(a_i+b_i))$,
1221: all subsheaves have rank $n-r$ and degree $-d$, and the foliation by
1222: $T$-invariant curves is given by the $T$-action on the base.
1223: \medskip
1224: 
1225: \noindent{\bf Horizontal families.}
1226: If there exist $i,j,c,c'$ such that $1 \le c \le a_j$, 
1227: $1 \le c' \le b_j$, and $a_i+b_i+c+c'=a_j+b_j$, then the point
1228: $\cS_{(\sdelta,\va,\vb)}$ 
1229: lies in a \emph{horizontal family} parametrized by the product of two
1230: projective lines.  Let $a'_i=a_i+c$, $a'_j=a_j-c$, $b'_i=b_i+c'$, 
1231: $b'_j=b_j-c'$.
1232: Let $([s,t], [\sigma,\tau])$ be the coordinates of
1233: $\P^1\times\P^1$, and let $\cS$ be the submodule of $\cO^n_{\P^1}$
1234: which, except for its components in $\cO\cdot\be_i + \cO\cdot\be_j$,
1235: agrees with $\cS_{(\sdelta,\va,\vb)}$.
1236: The component of $\cS$ in $\cO\cdot\be_i + \cO\cdot\be_j$ is the subsheaf
1237: generated by 
1238: %
1239:  \[
1240:    \be_jx^{a_j}y^{b_j}, \ \  
1241:    s\be_ix^{a'_i}y^{b_i}+t\be_jx^{a_j}y^{b'_j},\ \ 
1242:    \sigma \be_ix^{a_i}y^{b'_i}+\tau\be_jx^{a'_j}y^{b_j},\ \ 
1243:    \be_ix^{a'_i}y^{b'_i}\,.
1244:  \]
1245: %
1246: Similar reasoning as for Theorem~\ref{Th:Curves_are_correct} shows that 
1247: this defines a family of $T$-invariant
1248: curves over the base $\P^1\times\P^1$ with coordinates $([s,t], [\sigma,\tau])$.  
1249: It contains
1250: four $T$-fixed points:  setting $t=\tau=0$ gives the fixed point
1251: $\cS_{(\sdelta,\va,\vb)}$, setting $t=\sigma=0$ gives the fixed point
1252: $\cS_{(\sdelta,\va',\vb)}$, setting  $s=\tau=0$ gives the fixed point
1253: $\cS_{(\sdelta,\va,\vb')}$, and  setting $s=\sigma=0$ gives the fixed point
1254: $\cS_{(\sdelta,\va',\vb')}$.  This family also contains 
1255: the four Type II curves connecting these four fixed points, given by setting 
1256: exactly one of $s,t,\sigma$, or $\tau$ equal to zero.
1257: Furthermore, the data $(\vdelta,\va,\vb)$ and $(\vdelta',\va',\vb')$ satisfy 
1258: Theorem~\ref{T:parallel}(2)(c), and any two $T$-invariant curves $\cS$ and
1259: $\cS'$ as in Theorem~\ref{T:parallel}(2) lie in a unique horizontal family.
1260: 
1261: 
1262: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1263: 
1264: \section{Algebraic extension of GKM theory}\label{extended GKM theory}
1265: We discuss equivariant localization and an extension
1266: of the Goresky-Kottwitz-Mac\-Pherson relations 
1267: when there are finitely many fixed points
1268: but infinitely many $T$-invariant curves.  We work with 
1269: equivariant Chow rings; similar results hold for equivariant
1270: cohomology.
1271: In fact, when $k=\C$, $X$ is smooth and projective, and $X^T$ is finite,
1272: the two theories coincide.
1273: 
1274: We first recall some properties of $T$-equivariant Chow rings as
1275: developed by Edidin and Graham~\cite{EG98a} and Brion~\cite{brion}.
1276: Next, we outline Evain's~\cite{evain} development of ideas of Brion which 
1277: extends the GKM relations to 
1278: describe the $T$-equivariant Chow ring of a smooth variety with finitely
1279: many $T$-fixed points when there are infinitely many $T$-invariant curves.
1280: This description involves ideal-membership relations, one for each generator of
1281: the equivariant Chow ring of each family of $T$-invariant curves.
1282: When the generators are given by smooth subvarieties, these 
1283: relations may be expressed in terms of tangent weights.
1284: This gives one form of our presentation for $A^*_T(\cQ_d)$ in
1285: Theorem~\ref{Main_Theorem}. 
1286: We next give a variant of these relations using 
1287: differential operators, which gives the other form  of our
1288: presentation for $A^*_T(\cQ_d)$.  
1289: We then compute these relations for products of projective spaces, 
1290: and finally deduce Theorem~\ref{Main_Theorem}.
1291: 
1292: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1293: \subsection{Torus equivariant Chow rings}
1294: 
1295: When a linear algebraic group $G$ acts on a smooth scheme $X$, 
1296: Edidin and Graham~\cite{EG98a} defined the equivariant Chow ring
1297: $A^*_G(X)$, using Totaro's algebraic approximation to the classifying space 
1298: of $G$.   It satisfies 
1299: functorial properties under equivariant maps analogous to 
1300: those for ordinary Chow rings~\cite{Fu98}, including proper pushforwards
1301: and pullbacks by local complete intersection morphisms.
1302: 
1303: When the group is a torus $T$, Brion \cite{brion} gave an alternative 
1304: development of this theory which includes versions of the localization
1305: theorems that hold for equivariant cohomology. 
1306: He gave the following presentation for the equivariant Chow ring, 
1307: analogous to the usual presentation of Chow groups.
1308: The equivariant Chow ring $A^*_T(p)$ of a point $p$ is the integral symmetric
1309: algebra $S$ of the character group $\hat{T}$ of $T$.
1310: Equivariant pullback makes $A^*_T(X)$ into an $S$-module.
1311: 
1312: \begin{prop}[Theorem~2.1~\cite{brion}]\label{P:Chow_Gens}
1313:  The $S$-module $A^*_T(X)$ is defined by generators $[Y]$, for each 
1314:  $T$-invariant subvariety $Y$, and by relations 
1315:  $[\mbox{div}_Y(f)]-\chi[Y]$  for each rational function $f$ on $Y$ which is
1316:  a $T$-eigenvector of weight $\chi$; here $\chi$ is considered as
1317:  an element of $S$ in degree $1$.
1318: \end{prop}
1319: 
1320: It follows immediately that the usual Chow ring may be recovered 
1321: from the $S$-module $A^*_T(X)$, as the quotient
1322: by the ideal $S^+$ of $S$ generated by the character group $\hat{T}$.
1323: 
1324: \begin{prop}[Corollary~2.3.1~\cite{brion}]\label{P:Chow}
1325:   We have
1326: \[
1327:    A^*(X)\ =\ A^*_T(X)\otimes_S \Z\ =\ A^*_T(X)/S^+ A^*_T(X)\,.
1328: \]
1329: \end{prop}
1330: 
1331: The analogous statement in equivariant cohomology requires stronger 
1332: hypotheses. 
1333: 
1334: When $k=\C$,  the connection between 
1335: Chow groups and cohomology is given by the cycle map
1336: %
1337:  \[
1338:    A^*_T(X)\ \to\  H^*_{T,c}(X, \Z)
1339:  \]
1340: %
1341: to compactly supported (Borel-Moore) equivariant cohomology.
1342: If $X$ is projective and the fixed point set $X^T$ is finite, then
1343: the cycle map is an isomorphism.
1344: 
1345: Some statements below hold only for the rational
1346: equivariant Chow ring $A^*_T(X)_\Q := A^*_T(X) \otimes_\Z \Q$.
1347: This is a module over the rational equivariant Chow ring 
1348: $A^*_T(p)_\Q$ of a point $p$, which is the symmetric algebra $S_\Q$ of
1349: $\hat{T}_\Q := \hat{T}\otimes_\Z \Q$.  
1350: 
1351: \subsection{Localization}
1352: We now assume that $X^T$ is finite, and that $X$ has a 
1353: decomposition into $T$-invariant affine cells $C_1,\dotsc,C_m$ which can be
1354: ordered so that for $i=1,\dotsc,m$, the union $C_1\cup\dotsb\cup C_i$ 
1355: is Zariski open.
1356: We will call such varieties {\em filtrable}; this is close to the terminology 
1357: Brion used in \cite{brion}, but he did not require that
1358: $X^T$ be finite, and his cells were allowed to be vector bundles over components
1359: of $X^T$. 
1360: If $X$ is smooth and projective and $X^T$ is finite, then 
1361: Bia{\l}ynicki-Birula~\cite{BB73} showed that it is filtrable.
1362: 
1363: Let $i\colon X^T\to X$ be the inclusion of the subscheme of
1364: $T$-fixed points of $X$.  
1365: \begin{prop}[Corollary 3.2.1~\cite{brion}] 
1366: The $S$-module $A^*_T(X)$ is free.  The map
1367: \[i^*\colon A^*_T(X) \to A^*_T(X^T)\]
1368: is an injection.  
1369: \end{prop}
1370: 
1371: Brion also established Chow ring versions of results of Chang and Skjelbred
1372: and of Goresky, Kottwitz, and MacPherson concerning the image of the
1373: localization map.
1374: \begin{prop}[Sections 3.3 and 3.4~\cite{brion}]\label{P:GKM_algebraic}
1375: \mbox{\ }
1376:  \begin{enumerate}
1377:   \item[(a)]
1378:    The image of the localization map $i^* A^*_T(X)_\Q \to A^*_T(X^T)_\Q$ is the
1379:    intersection of the images of the localization maps
1380: \[
1381:    i^*_{T'}\ \colon\ A^*_T(X^{T'})_\Q\ \to\  A^*_T(X^T)_\Q
1382: \]
1383:   where $T'$ runs over all codimension one subtori of $T$.
1384: 
1385:  \item[(b)]
1386:   When $T$ acts with finitely many fixed points and has finitely
1387:   many invariant curves, then the image of the localization map 
1388:   $i^*\colon A^*_T(X)_\Q \to A^*_T(X^T)_\Q\simeq (S_\Q)^{X^T}$
1389:   is the set of all tuples $(f_p)_{p\in X^T} \in (S_\Q)^{X^T}$ such that
1390:   whenever $p$ and $q$ belong to the
1391:   same irreducible $T$-invariant curve $C$, we have 
1392:   $f_p\equiv f_q$ modulo $\chi$, where $\chi$ is the weight of
1393:   the action of $T$ on $T_pC$.
1394:  \end{enumerate}
1395: \end{prop}
1396: 
1397: Statement (a) is analogous to a theorem of 
1398: Chang and Skjelbred \cite{ChSk74} for equivariant cohomology.
1399: This result, together with the easy calculation of the equivariant
1400: Chow groups of $\P^1$, immediately gives (b), which is the Chow analog of 
1401: the GKM relations for equivariant cohomology.
1402: 
1403: In general this result does not hold with $\Z$ coefficients.  For instance, 
1404: suppose that $\dim X = 2$,  $x\in X^T$, and the weights of $T$ on the
1405: tangent space $T_xX$ are $a\chi$ and $a'\chi'$, where 
1406: $\chi, \chi'\in \hat{T}$ are linearly independent primitive characters.  
1407: Then condition (b) would say that if $(f_p) \in A^*_T(X^T)$ 
1408: has $f_p = 0$ for $p \ne x$, then it is in the image of $i^*$ if $f_x$ is a 
1409: multiple of $\lcm(a,a')\chi\chi'$.  In fact, $f_x$ must
1410: be a multiple of $aa'\chi\chi'$.
1411: 
1412: This is essentially the only obstruction to working with $\Z$ coefficients,
1413: at least if the fixed point set is finite.  We say that the tangent weights
1414: of a $T$-variety $X$ are {\em almost coprime} if whenever two 
1415: $T$-weights of $T_pX$ for $p\in X^T$ are divisible by the same 
1416: integer $a > 1$, then they are parallel. 
1417: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1418: %
1419: %   Frank's addition, amended by Tom
1420: %
1421: With this added hypothesis, Brion's proof
1422: of Proposition~\ref{P:GKM_algebraic} given by Brion works over $\Z$.
1423: 
1424: \begin{thm} \label{T:GKM_algebraic_Z}
1425:  Let $X$ be a smooth filtrable $T$-variety whose 
1426:  tangent weights are almost coprime.
1427:  Then Proposition~$\ref{P:GKM_algebraic}$ holds with rational Chow groups
1428:  replaced by integral Chow groups.
1429: \end{thm}
1430: 
1431: %\begin{proof} We follow Brion's argument for Proposition \ref{P:GKM_algebraic}, 
1432: %which is by induction on the number of strata.  
1433: %The result is immediate when $X$ has only one stratum.  
1434: %If $X$ has more than one stratum, let $F$ be a closed stratum
1435: %and set $U := X \setminus F$.  
1436: %Then there is an exact sequence~\cite[Proposition~3.2]{brion} 
1437: %%
1438: % \begin{equation} \label{exact sequence} 
1439: %  0\ \to\ A^*_T(X)\  
1440: %  \xrightarrow{\, \rho \,}\ A^*_T(U) \times A^*_T(F) \ 
1441: %   \xrightarrow{\, (0,\pi) \,}\ A^*_T(F)/(e^T(N))\ \to\ 0\,,
1442: % \end{equation}
1443: %%
1444: %where $e^T(N)$ is the equivariant Euler class of the normal bundle $N$
1445: %to $F$ in $X$.  Under the isomorphism $A^*_T(F) \cong S$, this is just
1446: %the product of all the $T$-weights of $N$.  
1447: %The components of $\rho$ are the restriction maps, while the map
1448: %$\pi$ is the natural quotient.
1449: %
1450: %Let $\alpha\in A^*_T(X^T)$ lie in the image of 
1451: %$i^*_{T'}$ for all codimension one subtori $T'$ of $T$.
1452: %By the induction hypothesis applied to $U$, the restriction
1453: %$\alpha|_{U^T}$ is in the image of $A^*_T(U)\to A^*_T(U^T)$.
1454: %Since $A^*_T(X)\to A^*_T(U)$ is surjective,
1455: %if $\beta\in A^*_T(X)$ is a lift of $\alpha|_{U^T}\in A^*_T(U)$,
1456: %then $\alpha = i^*\beta + \gamma$ where $\gamma=\alpha- i^*\beta$
1457: %restricts to zero in $A^*_T(U)$.
1458: %Then $\gamma$ lies in the image of $i^*_{T'}$
1459: %for all codimension one subtori $T'$ of $T$, and it is enough to show
1460: %that $\gamma \in\im i^*$.  
1461: %By \eqref{exact sequence}, this holds if and only if 
1462: %$\gamma|_F \in A^*_T(F) \cong S$ is a multiple of $e^T(N)$.  
1463: %
1464: %To see this, write $e^T(N)=\prod_\chi e_\chi$, the product over all 
1465: %primitive characters $\chi \in \hat{T}$, where  
1466: %$e_\chi$ is the product of all characters in 
1467: %$N$ which are parallel to $\chi$.  Then 
1468: %$\ker \chi$ is a codimension one subtorus of $T$ and
1469: %$e_\chi$ is the top equivariant Chern class of the
1470: %normal bundle to $F^{\ker \chi}$ in $X^{\ker \chi}$.
1471: %
1472: %Since $\gamma \in \im i^*_{\ker \chi}$ for every 
1473: %primitive character $\chi$, we can apply the 
1474: %sequence \eqref{exact sequence} to $X^{\ker \chi}$
1475: %to conclude that $\gamma|_F$ is a multiple of
1476: %$e_\chi$ for all $\chi$ (note that $A^*_T(F^{\ker \chi})$ 
1477: %and $A^*_T(F)$ are identified with $S$ in such a way that
1478: %the restriction is the identity).  
1479: %
1480: %The hypothesis on the divisibility of weights on 
1481: %tangent spaces then implies that $\gamma|_F$ is a 
1482: %multiple of $e^T(N)$, as required.
1483: %\end{proof}
1484: 
1485: 
1486: 
1487: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1488: \subsection{Evain's relations}
1489: When $T$ does not have finitely many invariant curves on $X$, then
1490: statement (b) of Proposition~\ref{P:GKM_algebraic} fails, but by (a) we
1491: can still compute $A^*_T(X)$ if we know the images of $i^*_{T'}$ for 
1492: all codimension one subtori $T'$ of $T$.  A finite set of such $T'$
1493: suffices, namely those which fix at least one $T$-invariant curve
1494: pointwise.  Brion \cite{brion} and Goldin and Holm \cite{GH} have 
1495: computed cases where the components of $X^{T'}$ are low-dimensional.
1496: Evain \cite{evain} recently described relations in the general case:
1497: we recall his results.
1498: 
1499: Let $Y = X^{T'}$ for $T'$ a codimension one subtorus of $T$.
1500: By \cite{I}, $Y$ is smooth.
1501: For $p\in Y^T = X^T$, let $e^T_p(Y) = e^T(TY)|_p$ be the localization of the
1502: equivariant Euler class of $TY$ at $p$.  
1503: Under the identification $A^*_T(p) = S$, this 
1504: is the product of the $T$-weights on the tangent space $T_pY$.
1505: 
1506: \begin{prop}[Corollary 27~\cite{evain}]\label{P:C27Ev}
1507:  A class $\alpha=(\alpha_p)_{p\in Y^T}$ in $S^{Y^T}$ 
1508:  lies in $i^*_{T'}A^*_T(Y)$ if and only if 
1509: %
1510: \begin{equation}\label{E:Evain}
1511:   \sum_{p\in Y^T} \frac{\alpha_p\beta_p}{e^T_p(Y)}\ \ \in\ S
1512: \end{equation}
1513: %
1514:  for every $\beta\in i^*_{T'}A^*_T(Y)$.
1515: \end{prop}
1516: 
1517: \noindent{\bf Remark on Evain's proof.}
1518: The condition~\eqref{E:Evain} is necessary, since
1519: if $\pi$ is the %$T$-equivariant 
1520: projection of $Y$ to a point,
1521:  then  %$\pi_*$ is a map of $S$-modules and 
1522:  the sum is simply $\pi_*(\alpha\cdot\beta)$, by the integration
1523:  formula of Edidin and Graham~\cite{EG98b}.
1524: Note that since $\pi_*$ is $S$-linear, it is enough to take
1525: $\beta$ in a generating set of the $S$-module
1526: $i^*A^*_T(Y)$.
1527: 
1528:  By Bia{\l}ynicki-Birula \cite{BB73}, there are two
1529:  $T$-invariant cell decompositions $C_p^+$ and $C_p^-$ for $p\in Y^T$ of $X$ and
1530:  an ordering of the fixed points $Y^T$ such that the matrix with entries in $S$
1531:  whose $(p,q)$-entry is 
1532: \[
1533:    \pi_*([C^+_p]\cdot [C^-_q])
1534: \]
1535:  is unitriangular.
1536:  Either set of classes $[C^+_p]$ or $[C^-_p]$ forms a basis for the
1537:  $S$-module $A^*_T(X)$, and
1538:  expressing the elements $\alpha$ and $\beta$ in these two bases proves
1539:  sufficiency.  \medskip
1540: 
1541: Combining Proposition~\ref{P:C27Ev} with Proposition \ref{P:GKM_algebraic}
1542: and Theorem~\ref{T:GKM_algebraic_Z} gives the
1543: following criterion for membership in $i^*A^*_T(X)$.
1544: 
1545: \begin{thm}\label{T:E_Criterion}
1546:   A class $\alpha=(\alpha_p)_{p\in X^T} \in (S_\Q)^{X^T}$ lies in the image
1547:   $i^*A^*_T(X)_\Q$ of the localization map if and only if
1548: for all $Y = X^{T'}$ for $T'$ a codimension one subtorus of $T$ we have
1549: \[
1550:    \sum_{p\in Y^T} \frac{\alpha_p\beta_p}{e^T_p(Y)}\ \ \in\ S_\Q\,
1551: \]
1552: for all $\beta$ in a set of $S_\Q$-module generators 
1553: for $i^*_{T'}A^*_T(Y)_\Q$.
1554: 
1555: %where $Y = X^{T'}$  all $T'$-fixed subvarieties for codimension one subtori $T'$
1556: %of $T$ and $\beta$ runs over a set of $S_\Q$-module generators 
1557: %for $i^*_{T'}A^*_T(Y)_\Q$.
1558: 
1559: If the tangent weights of $X$
1560: are almost coprime, the same statement holds over $\Z$.
1561: \end{thm}
1562: 
1563: \begin{rmk}
1564:   When $X$ is smooth, the relations in
1565:   Theorem~\ref{T:E_Criterion} can also be taken for $Y$ running over
1566:   all irreducible components of the union of the fixed points
1567:   and the $T$-invariant curves, since such $Y$ are just the connected
1568:   components of the $T'$-fixed loci $X^{T'}$ for some codimension one subtorus
1569:   $T'$ of $T$.  
1570:   We call this union of fixed points and $T$-invariant curves the 
1571:   {\it one-skeleton of $X$}. 
1572: \end{rmk}
1573: 
1574: To apply Theorem~\ref{T:E_Criterion}, we need to know
1575: explicit generators of $A^*_T(Y)$, or more precisely their localizations
1576: to $Y^T$.
1577: By Proposition~\ref{P:Chow_Gens}, one class of generators are the equivariant
1578: fundamental cycles $[Z]$ of $T$-invariant subvarieties $Z$ of the components
1579: $Y$.  These are easy to compute when $Z$ is smooth, since if $p\in Z^T$ we have 
1580: $[Z]_p = e^T_p(N_ZY)$, the equivariant
1581: Euler class of the normal bundle to $Z$ in $Y$, while
1582: if $p\in Y^T \setminus Z^T$, then $[Z]_p = 0$.
1583: It follows that $\frac{[Z]_p}{e^T_p(Y)}=\frac{1}{e^T_p(Z)}$ if $p\in Z^T$.
1584: 
1585: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
1586: %
1587: %   Remove the following if we resend it to JAG
1588: %
1589: To see this, note that $[Z]_p\in A^*_T(p)$ is the pullback of $[Z]$ along
1590: the regular embedding $i_{p,Y}\colon p\to Y$.
1591: We factor $i_{p,Y}$ as the composition 
1592: \[
1593:    p\ \xrightarrow{\ i_{p,Z}\ }\ Z
1594:       \xrightarrow{\ i_Z\ }\ Y\,.
1595: \]
1596: The class $[Z]\in A^*_T(Y)$ is the pushforward along $i_Z$ of the unit
1597: class $1=[Z]\in A^*_T(Z)$, and so we have
1598: \[
1599:   [Z]_p\ =\ i^*_{p,Z} i^*_Z i_{Z,*} 1\ =\ 
1600:   i^*_{p,Z} e^T(N_ZY) =\ e^T_p(N_ZY)\,,
1601: \]
1602: by the self-intersection formula for Chow rings. 
1603: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1604: 
1605: Thus if we can find for each $Y$ a collection $\cZ_Y$ of smooth $T$-invariant
1606: subvarieties of $Y$ so that the classes $[Z]$ for $Z\in \cZ_Y$ generate
1607: $A^*_T(Y)$ as an $S$-module, we get the following more explicit version of
1608: Theorem \ref{T:E_Criterion}. 
1609: 
1610: \begin{thm}\label{T:toric_criterion}
1611:  A class $\alpha=(\alpha_p)_{p\in X^T}\in (S_\Q)^{X^T}$ lies in
1612:  $i^*A^*_T(X)_\Q$ if and only if 
1613: %
1614:  \begin{equation} \label{relation for smooth Z}
1615:    \sum_{p\in Z^T} \frac{\alpha_p}{e^T_p(Z)}\ \in\ S_\Q\,
1616:  \end{equation}
1617: %
1618:  for all $Z \in \cZ_Y$ and all components $Y$ of the one-skeleton of $X$. 
1619:  If the tangent weights of $X$ are almost coprime, the same statement holds over $\Z$. 
1620: \end{thm}
1621: 
1622:  The necessity of \eqref{relation for smooth Z} 
1623:  does not require the argument above,
1624:  since if $(\alpha_p) = i_*\alpha$ for $\alpha \in A^*_T(X)$, then the sum
1625:  is just $\pi_*(\alpha|_Z)$, where $\pi$ is the projection of $Z$ to a
1626:  point.
1627: 
1628: Obvious candidates for the subvarieties $[Z]$ are the
1629: closures of the Bia{\l}ynicki-Birula cells, since their classes
1630: form  an $S$-basis for $A^*_T(Y)$.  Unfortunately, they are not  
1631: in general smooth---this was the case for Evain.
1632: However, for the quot schemes we study, they are smooth, as the 
1633: connected components $Y$ are products of projective spaces.  
1634: More generally we can ask that for each component $Y$ of 
1635: $X^{T'}$ there is a torus $T_Y$ containing $T$ which acts on $Y$ 
1636: with finitely many orbits, so that $Y$ is a smooth toric variety. 
1637: The closures of the cells will be $T_Y$-orbit closures, and therefore
1638: smooth.
1639: 
1640: The relations of Theorem \ref{T:toric_criterion} 
1641: are the same as those found by Goldin and Holm \cite{GH}
1642: for equivariant cohomology of Hamiltonian $T$-spaces.
1643: Their result applied where the spaces $X^{T'}$ 
1644: are at most four-dimensional (over $\R$).  Since this
1645: implies that the components of $X^{T'}$ are toric manifolds, 
1646: %our theorem 
1647: Theorem~\ref{T:toric_criterion} 
1648: recovers their result. 
1649: 
1650: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1651: \subsection{Evain's relations as differential operators}
1652: We rewrite this algebraic criterion in a different form.
1653: Suppose that $Y$ is a smooth component of $X^{T'}$ and $Z\subset Y$ is a smooth
1654: $T$-invariant subvariety.  
1655: The action of $T$ on $Y$ factors through a character $\eta\colon T \to \C^*$, 
1656: so the weights of
1657: $T$ on $T_pZ$ for $p\in Z^T$ are non-zero scalar multiples of $\eta$.
1658: Thus there exist numbers $d_p=d_p(Z)$ so that 
1659: \[
1660:    e^T_p(Z) = d_p(Z)\cdot \eta^{\dim Z}\,.
1661: \]
1662: The terms in \eqref{relation for smooth Z} have a common
1663: denominator $\eta^{\dim Z}$, and so we may rewrite it
1664: as
1665: \[
1666:     \sum_{p\in Z^T} \frac{\alpha_p}{d_p(Z)}\ \in\ {\eta^{\dim Z}}S_\Q\,.
1667: \] 
1668: 
1669: We can rewrite this condition using a linear differential operator. 
1670: The ring $S_\Q$ is the symmetric algebra of $\hat{T}_\Q$, or dually
1671: the ring of polynomial functions on ${\hat{T}_\Q}^*$.
1672: Choose $\zeta\in {\hat{T}_\Q}^*$
1673: for which $\zeta(\eta)\neq 0$.  Then 
1674: the operator $D=D_\zeta$ of differentiation in the direction of
1675: $\zeta$ acts on $S_\Q$.  If $f\in S_\Q$ is divisible by $\eta$, then
1676: $\eta^k$ divides $f$ if and only if $\eta^{k-1}$ divides $Df$, so
1677: the relation \eqref{relation for smooth Z} is equivalent to
1678: %
1679:  \[ 
1680:    \sum_{p\in Z^T} d_{p}(Z)^{-1}D^j \alpha_p \equiv 0
1681:    \mod\eta,
1682:     \quad\mbox{for all}\ 0 \le j < \dim Z.
1683: \]
1684: 
1685: We give a variant of Theorem~\ref{T:toric_criterion} which uses the
1686: last relation, but only with the maximum order derivative $j = \dim Z - 1$.
1687: In exchange, we must apply it using more subvarieties $Z$. 
1688: 
1689: Let $Y$ be a component of the one-skeleton of $X$, let 
1690: $\eta$ be the associated character of $T$, and consider
1691: the two Bia{\l}ynicki-Birula cell decompositions
1692: $\{C^-_p\mid p\in Y^T\}$ and $\{C^+_p\mid p\in Y^T\}$ 
1693: induced by the $T$-action.
1694: Each cell $C=C^+_p, C^-_p$ is isomorphic to the $T$-vector space 
1695: $T_pC\subset T_pY$.
1696: Suppose that within each cell $C=C^+_p$ 
1697: we can find $T$-invariant affine subspaces $C_{p,1}, \dots C_{p,\dim C}$ with
1698: $\dim C_{p,i} = i$ and which have smooth closures $Z_{p,i} = \overline{C_{p,i}}$.
1699: As before, this will be true if each $Y$ is a toric variety for a larger torus 
1700: $T_Y$ containing $T$, since we can take each $Z_{p,i}$ to be the closure of a $T_Y$-orbit.
1701: 
1702: \begin{thm} \label{derivative theorem}
1703: With these assumptions, a 
1704: class $\alpha=(\alpha_p)_{p\in X^T}\in (S_\Q)^{X^T}$ lies in
1705: $i^*A^*_T(X)_\Q$ if and only if 
1706: %
1707:  \begin{equation} \label{top degree derivative relation}
1708:   \sum_{q\in Z^T} d_{q}(Z)^{-1}D^{\dim Z - 1} \alpha_q \equiv 0 \mod\eta .
1709:  \end{equation}
1710: %
1711: for all $Z = Z_{p,i}$ and for all components $Y$ of the one-skeleton of $X$.
1712: \end{thm}
1713: 
1714: \begin{proof}
1715: The necessity of the conditions \eqref{top degree derivative relation} follows from the 
1716: previous discussion.  
1717: 
1718: To show they are sufficient, let $U$ be an open union of the %opposite 
1719: %Bruhat 
1720: %Bia{\l}ynicki-Birula
1721: %
1722: cells $C^-_p$ and  note that
1723: $Z_{p,i}\subset U$ if and only if $p \in U$.  
1724: We use induction on the number of cells in $U$ to show that the image of 
1725: $i_U^*\colon A^*_T(U)_\Q \to A^*_T(U^T)_\Q$ is the set of $(\alpha_x)|_{x\in U^T}$
1726: satisfying \eqref{top degree derivative relation} for all $Z_{p,i}\subset U$.
1727: 
1728: When $U$ is a single cell, this is immediate, as $i^*_U$ is an isomorphism
1729: and there are no $Z_{p,i}$'s contained in $U$.  
1730: Otherwise, suppose
1731: $\alpha = (\alpha_x)|_{x\in U^T}$ satisfies \eqref{top degree derivative relation}
1732: for all $Z_{p,i}\subset U$.
1733: Let $C^-_p \subset U$ be a closed cell, and put $U' = U \setminus C^-_p$.
1734: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1735: %
1736: %   Frank's changes
1737: %
1738: There is an exact sequence\cite[Proposition~3.2]{brion} 
1739: %
1740:  \begin{equation} \label{exact sequence} 
1741:   0\ \to\ A^*_T(U)_\Q\  
1742:   \xrightarrow{\, \rho \,}\ A^*_T(U')_\Q \times A^*_T(C^-_p)_\Q \ 
1743:    \to\ A^*_T(C^-_p)_\Q/(e^T(N))\ \to\ 0\,,
1744:  \end{equation}
1745: %
1746: where $e^T(N)$ is the equivariant Euler class of the normal bundle $N$
1747: to $C^-_p$ in $X$.  Under the isomorphism $A^*_T(C^-_p) \cong S$, this is just
1748: the product of all the $T$-weights of $N$.  
1749: The components of $\rho$ are the restriction maps, while the map
1750: $A^*_T(C^-_p)_\Q \to A^*_T(C^-_p)_\Q/(e^T(N))$ is the natural quotient.
1751: 
1752: %Then $\alpha|_{U'}$ is in the image of $i^*_{U'}$, by the inductive hypothesis.
1753: By the inductive hypothesis, $\alpha|_{(U')^T}$ lies in the image of $i^*_{U'}$.
1754: %As in the proof of Theorem~\ref{T:GKM_algebraic_Z}, t
1755: The exact
1756: sequence~\eqref{exact sequence}  implies that $A^*_T(U) \to A^*_T(U')$
1757: is surjective  and so we can write $\alpha = i^*_U\beta + \gamma$, with
1758: $\beta \in A^*_T(U)$ and $\gamma|_{(U')^T} = 0$. 
1759: Since $i^*_U\beta$ satisfies the relations 
1760: \eqref{top degree derivative relation}, so does $\gamma$.
1761: It will be enough to show that $\gamma$ is in the image of $i^*_U$.  But using 
1762: the exact sequence \eqref{exact sequence}, we see that this holds 
1763: if and only if $\gamma_p$ is a multiple of $c_d^T(N)$, which is a non-zero multiple
1764: of $\eta^d$, where $d = \mathop{\mathrm{codim}} C^-_p = \dim C^+_p$.  But 
1765: the relation \eqref{top degree derivative relation} implies that 
1766: $D^{k}\gamma_p \equiv 0 \pmod \eta$ for $0 \le k < d$.  The result follows. 
1767: \end{proof}
1768: 
1769: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1770: 
1771: \begin{example}
1772: \label{Example_1}
1773: {\rm 
1774:  Let $T = \G_m$ act on $X = \P^r$ by 
1775: %
1776:  \[
1777:    t\cdot[x_0:x_1:\dots:x_r]\ =\ [x_0: tx_1: \dots: t^rx_r]
1778:  \]
1779: %
1780: in homogeneous coordinates, where $t\in T$.
1781: For each $0\le j \le r$, let $p_j \in X^T$ denote the $T$-fixed
1782: point corresponding to the $j$th standard basis vector $\be_j$.  The
1783: tangent space $T_{p_j}X$ is $\C^n/\C\cdot\be_j$, with the action of
1784: $T$ given by $t\cdot \overline{\be}_k = t^{k-j}\overline{\be}_k$.
1785: Thus $e_{p_j}(X) = (-1)^j j!(r{-}j)!\eta^r$, where 
1786: $\eta$ is the identity character.  
1787: More generally, if $0\le l\le n \le r$, let 
1788: $Z_{l,m} = \P\Span\{\be_l, \be_{l+1},\dots,\be_m\}$.  The same 
1789: calculation gives 
1790: %
1791:  \[
1792:     e_{p_j}(Z_{l,m})\ =\ (-1)^{j-l} (j-l)!(m-j)!\eta^{m-l}\,.
1793: \]
1794: 
1795: We can apply Theorem \ref{T:toric_criterion} using the smooth
1796: subvarieties $Z_{0,l}$, for $1\le l \le r$.
1797: Then $i^*A^*_T(X) \subset A^*_T(X^T)$
1798: is the set of tuples $\alpha = (\alpha_0,\dots,\alpha_r)$ where
1799: %
1800:  \begin{equation}\label{example relation 1} 
1801:      \sum_{0\le j\le l} \frac{(-1)^j\alpha_j}{j!(l-j)!}\ \in\ \eta^l S\,
1802:  \end{equation}
1803: %
1804: for all $1 \le l \le r$.
1805: 
1806: On the other hand, we can apply Theorem \ref{derivative theorem}
1807: using all the subvarieties $Z_{l,m}$. 
1808: If $D$ is differentiation on $S$ in the direction of $\eta^\vee$, then
1809: $\alpha$ lies in the image of $i^*$ if and only if 
1810: % 
1811:  \begin{equation}\label{example relation 2}
1812:    \sum_{l\le j\le m}
1813:    \frac{(-1)^{j-l}D^{m-l-1}\alpha_j}{(j-l)!(m-j)!}
1814:      \ \in\ \eta S
1815:  \end{equation}
1816: %
1817: for all $0 \le l < m \le r$.  
1818: We could take one more derivative and
1819: ask that the resulting sums vanish, but this would not generalize
1820: to actions of higher dimensional tori.
1821: 
1822: When $X = \P^1$ we get exactly the GKM relation for a primitive action.
1823: }
1824: \end{example}
1825: 
1826: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
1827: \begin{example}\label{Example_2}
1828: {\rm 
1829:  These same arguments apply to products of projective spaces.   
1830: Let $\vr = (r_1,\dots,r_n)$, and let $X = \P^{r_1} \times \dots \times 
1831: \P^{r_n}$, where the action of $t \in T = \G_m$ on a point
1832: $([x^1_0:x^1_1:\dots:x^1_{r_1}], \dots,[x^n_0:x^n_1:\dots:x^n_{r_n}])$ is
1833: given by multiplying $x^i_j$ by $t^j$.  The fixed points have
1834: the form $p_{\vj} = (p_{j_1},\dots,p_{j_n})$, where $p_{j_i}$ is the
1835: $j_i$th fixed point in $\P^{r_i}$, in the notation of Example~\ref{Example_1},
1836: and $\vj = (j_1,\dots,j_n)$ satisfies $0\le \vj \le \vr$, meaning that
1837: $0\le j_i \le r_i$ for all $1 \le i \le n$.
1838: 
1839: For $\vl,\vm \in \Z^n$ with $0 \le l_i \le m_i \le r_i$ for all $i$, set
1840: $Z_{\vl,\vm} = Z_{l_1,m_1}\times\dots \times Z_{l_n,m_n}$.  
1841: For each $\vl \le \vj \le \vm$, the tangent space to $Z_{\vl,\vm}$ at
1842: $p_{\vj}$ is $\bigoplus_{i = 1}^n T_{p_{j_i}}Z_{l_i,m_i}$.  
1843: Using the computation from Example~\ref{Example_1},
1844: we see that 
1845: %
1846:  \[
1847:    e^T_{p_{\vj}}(Z_{\vl,\vm})\ =\
1848:    (-1)^{|\vj|}(\vj-\vl)!(\vm-\vj)!\eta^{|\vm|-|\vl|}\,.
1849:  \]
1850: %
1851: Recall that 
1852: for an $n$-tuple $\va = (a_1,\dots,a_n)$, we put $\va! = a_1!\cdots a_n!$.
1853: 
1854: As in Example~\ref{Example_1}, we can either apply 
1855: Theorem \ref{T:toric_criterion} using the 
1856: subvarieties $Z_{\vzero,\vl}$, or Theorem \ref{derivative theorem} using 
1857: all the $Z_{\vl,\vm}$.  The resulting conditions for a tuple 
1858: $\alpha = (\alpha_\vj)_{\vzero\le \vj\le \vr}$
1859: to be in the image of the localization map are just 
1860: \eqref{example relation 1} and \eqref{example relation 2}, 
1861: where the variables now represent elements of $\Z^n$ rather than scalars.
1862: }
1863: \end{example}
1864: 
1865: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1866: \subsection{Proof of Theorem \ref{Main_Theorem}}\label{main proof}
1867: We combine these localization results with the 
1868: geometry of the quot scheme from Sections
1869: \ref{fixed_points}, \ref{Tcurves}, and \ref{Tfamilies} 
1870: to produce a proof of Theorem \ref{Main_Theorem}.
1871: 
1872: The only non-primitive tangent weights are those with 
1873: $i=j$ in~\eqref{Eq:TII_III}, which are multiples of $\bff$;
1874: These correspond to Type III curves.  Thus the 
1875: tangent weights are almost coprime, so 
1876: Theorem \ref{T:toric_criterion} gives a
1877: correct description of the integral equivariant Chow ring.
1878: 
1879: The relations I, II(a) and II(b) are the GKM relations
1880: for the $T$-invariant curves of types I and II, as described
1881: in Section \ref{Tcurves} using the identification
1882: of the $T$-weights of these curves with the tangent weights
1883: given by Theorem \ref{tangentdirections}.
1884: 
1885: The relations II(c)/II(c)$'$ come from the horizontal families of 
1886: $T$-invariant curves of type II.  As described in 
1887: Section~\ref{Tfamilies}, these are isomorphic to $\P^1 \times \P^1$,
1888: where the action on 
1889: each factor is by the same 
1890: primitive character.  As in Example~\ref{Example_2},
1891: we can apply the relation in Theorem~\ref{T:toric_criterion}
1892: with $Y = \P^1 \times \P^1$ to get the relations II(c)/II(c)$'$.  
1893: We get one new relation using $Z = Y$; smaller
1894: $T$-invariant subvarieties contained in $Y$ are either $T$-invariant
1895: curves, whose relations are already covered by II(a) and II(b), 
1896: or points, which give no relation.  
1897: 
1898: Finally, the relations III and III$'$ come from the vertical families.  
1899: As described in Section \ref{Tfamilies}, the vertical 
1900: family containing $\cS_{(\sdelta,\va,\vb)}$ is isomorphic
1901: to 
1902: \[
1903:     \prod_{\delta_i=1} \P H^0(\cO(a_i+b_i))\ \simeq\ 
1904:     \prod_{\delta_i=1} \P^{a_i+b_i}\,,
1905: \]   
1906: and the fixed points in the family are those $\cS_{(\sdelta,\va',\vb')}$
1907: with $\va' + \vb' = \va + \vb$.  
1908: The codimension one subtorus $T_{k^n}\subset T$
1909: acts trivially on the family, and the remaining action of 
1910: $T_{\P^1}$ is the one described in Example~\ref{Example_2}, using the
1911: monomial basis of $H^0(\cO(a_i+b_i))$. 
1912: Example~\ref{Example_2} then gives exactly the relations III and III$'$.
1913: This proves Theorem \ref{Main_Theorem}. \hfill\qed
1914: 
1915: 
1916: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1917: \section{Equivariant Chern classes on $\cQ_d$}\label{classes}
1918: 
1919: 
1920: Recall that $\P^1\times\cQ_d$ has a universal exact sequence of sheaves
1921: \[
1922:    0\ \rightarrow\ \cS\ \rightarrow\
1923:    \cO^n_{\P^1\times\cQ_d}\ 
1924:    \rightarrow\  \cT\ \rightarrow 0
1925: \]
1926: with $\cS$ the tautological vector bundle of rank $n-r$.
1927:   
1928: 
1929: Since both $\cQ_d$ and $\P^1$ have cell decompositions, we have a
1930: K\"unneth decomposition of Chow rings, 
1931: $A^*(\P^1\times\cQ_d)\simeq A^*(\P^1)\otimes_\Z A^*(\cQ_d)$.
1932: Let $\pi\colon \P^1\times\cQ_d\to \cQ_d$ be the projection. 
1933: For each $1\leq i\leq n{-}r$ we may decompose the Chern class $c_i(\cS)$ 
1934: %
1935:  \begin{equation}\label{Eq:Chern_Decomposition}
1936:    c_i(\cS)\ =\ \pi^*t_i+h\pi^*u_{i-1}\,,
1937:  \end{equation}
1938: %
1939: where $t_i,u_i\in A^i(\cQ_d)$ and 
1940: $h$ is the class of a point in $A^1(\P^1)$ pulled back to 
1941: $\P^1\times\cQ_d$.
1942: Note that $u_0=-d$.
1943: Str\o mme~\cite{St87} proved that $A^*(\cQ_d)_\Q$ is generated by the classes 
1944: \[
1945:    \{t_1,\ldots,t_k,u_1,\ldots,u_{k-1}\}\,.
1946: \]
1947: 
1948: 
1949: For each $1\leq i\leq n{-}r$, the equivariant Chern
1950: class $c_i^T(\cS)$ localizes at a fixed point $p$ to  
1951: the $i$th elementary symmetric polynomial $e_i$ in the $T$-weights of the
1952: fibre of $\cS$ at $p$.
1953: The fixed points of $\P^1\times\cQ_d$ correspond to $\{0,\infty\}\times \cF$.
1954: For $(\vdelta, \va, \vb)\in\cF$, the bundle $\cS_{(\sdelta, \va, \vb)}$ on
1955: $\P^1$ is a sum of line bundles $\cS_{a_j,b_j}\be_j$ for $\delta_j=1$.
1956: Since $\cS_{a,b}$ has weight $a\bff$ at $0$ and $-b\bff$ at $\infty$, 
1957: the localizations of 
1958: $c_i^T(\cS)$ are  
1959: %
1960:  \begin{eqnarray*}
1961:   c^T_{i,0,(\sdelta, \va, \vb)}&=&
1962:     e_i(\{\be_j+a_j\bff \mid \delta_j=1\})
1963:   \quad \mbox{at } (0,(\vdelta, \va, \vb))\\
1964:   c^T_{i,\infty,(\sdelta, \va, \vb)}&=&
1965:     e_i(\{\be_j-b_j\bff \mid \delta_j=1\})
1966:   \quad \mbox{at } (\infty,(\vdelta, \va, \vb))\,.
1967:  \end{eqnarray*}
1968: %
1969: 
1970: We also have a K\"unneth decomposition in equivariant Chow cohomology:
1971: \[
1972:    A^*_T(\P^1\times\cQ_d)\ \xleftarrow{\ \sim\ }\  A^*_T(\P^1)\otimes_S
1973:    A^*_T(\cQ_d)\,.
1974: \]
1975: To see this, just imitate the argument for ordinary Chow cohomology;
1976: the equivariant Chow cohomology of a variety with an algebraic
1977: cell decomposition will be a free $S$-module, with a module basis
1978: given by the closures of the cells.
1979: 
1980: An equivariant K\"unneth decomposition of $c^T_i(\cS)$ analogous to 
1981: \eqref{Eq:Chern_Decomposition} requires the choice of 
1982: a lift of the class of a point to $A^1_T(\P^1)$.
1983: Localizing a class $x\in A^1_T(\P^1)$ gives an ordered pair 
1984: $(x_0,x_\infty)\in \Z\bff\oplus\Z\bff$.
1985: Lifts of classes from $A^1(\P^1)$ 
1986: are only well-defined modulo the span of $(\bff,\bff)$.
1987: Three possible choices for lifting the class of a point are:
1988: %
1989:  \begin{equation}\label{Eq:lifts}
1990:   (i)\  (-\bff,0)\,,\qquad 
1991:   (ii)\  (0,\bff)\,,\quad\mbox{and}\quad
1992:   (iii)\ {\textstyle\frac{1}{2}}(-\bff,\bff)\,.
1993:  \end{equation}
1994: %
1995: The symmetric lift $(iii)$ requires rational coefficients.
1996: We will use this lift to express our formulas.
1997: 
1998: Given a lift $h\in A^1_T(\P^1)$ of the class of a point, 
1999: the formula~\eqref{Eq:Chern_Decomposition} defines equivariant lifts of
2000: the classes $t_i,u_{i-1}$.
2001: Str\o mme's result together with Proposition~\ref{P:Chow} implies that 
2002: these classes generate $A^*_T(X)_\Q$ as an $S_\Q$-algebra. 
2003: 
2004: 
2005: \begin{prop}\label{Prop:kunneth}
2006:  The symmetric choice $(iii)$ of lift $h\in A^1_T(\P^1)$ of the class 
2007:  of a point give the formula for the K\"unneth components $t_i, u_{i-1}$ of the
2008:  equivariant Chern class $c_i^T(\cS)$ of the bundle $\cS$.
2009:  We express this in terms of its localization at the fixed point of $\cQ_d$
2010:  indexed by $p:=(\vdelta, \va, \vb)$.
2011: \[
2012:    t_{i,p}\ =\ \frac{1}{2}(c^T_{i,0,p}+c^T_{i,\infty,p})\qquad
2013:    u_{i-1,p}\ =\ \frac{1}{\bff}(c^T_{i,\infty,p}-c^T_{i,0,p})\,.
2014: \]
2015: \end{prop} 
2016: 
2017: 
2018: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2019: \section{Equivariant Chow ring of the quot scheme
2020:   $\cQ_2(0,2)$}\label{S:ex} 
2021: 
2022: We use Theorem~\ref{Main_Theorem} to describe the equivariant Chow
2023: ring of $\cQ_2:=\cQ_2(0,2)$.
2024: We first give a basis for $A^*_T(\cQ_2)$ as a module over 
2025: $S=\Z[\be_1,\be_2,\bff]$.
2026: 
2027: The equivariant Chow ring is the collection of tuples
2028: $(f_p\mid p\in \cQ_2^T)\in S^{ \cQ_2^T}$ which satisfy 
2029: the relations of Theorem~\ref{Main_Theorem}.
2030: If $p$ and $q$ are connected by an edge in the moment multigraph
2031: (see Figure~\ref{F:Q202}) with weight $\chi$,
2032: then $f_p-f_q$ lies in the ideal generated by $\chi$.
2033: These are the standard GKM relations.
2034: 
2035: There are two
2036: multiedges with four vertices in the multigraph, namely the vertical and horizontal 
2037: lines of symmetry.
2038: They should be seen as flattened quadrangles, since 
2039: they are images of subvarieties isomorphic to $\P^1 \times \P^1$.  
2040: Each 
2041: gives rise to an additional relation, as follows.
2042: Suppose that the quadrangle has four vertices $a,b,c,d$:
2043: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2044: \[
2045:   \begin{picture}(150,26)(-50,-1)
2046:    \put(-50,8){\vector(1,0){20}}\put(-45,12){$\chi$}
2047:    \put( 0,10){$a$}\put(42,2){$b$}\put(42,17){$c$}\put(84,10){$d$}
2048:    \put( 9,14){\line(6, 1){30}} \put( 9, 10){\line(6,-1){30}}
2049:    \put(52,19){\line(6,-1){30}} \put(52,  5){\line(6, 1){30}}
2050:   \end{picture}
2051: \]
2052: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2053: (Here, the edges are parallel with direction $\chi$.)
2054: Then the tuples $(f_p)$ must satisfy
2055: \[
2056:   f_a - f_b - f_c + f_d \ \in\ \chi^2 S\,.
2057: \]
2058: These are relations of types II(c)$'$ (horizontal) and III$'$ (vertical) of
2059: Theorem~\ref{Main_Theorem}. 
2060: 
2061: The remaining multiedges with more than two vertices are the left 
2062: and right vertical edges, both with three vertices.  They should
2063: be seen as flattened triangles, since they come from subvarieties
2064: isomorphic to $\mathbb{P}^2$.
2065: The additional relations they induce are described as follows.
2066: Let the three fixed points on the multiedge 
2067: be $a, b, c$, with $b$ between $a$
2068: and $c$.  Then 
2069: \[
2070:   f_a - 2 f_b + f_c\ \in\ \bff^2 S\,.
2071: \]
2072: This is relation of type III$'$  of Theorem~\ref{Main_Theorem}. 
2073: 
2074: When $k=\C$, we can construct an $S$-module basis for equivariant cohomology
2075: using equivariant Morse theory; as is  well-known, a generic projection of the
2076: moment map to a line will give a Morse function which is perfect for equivariant
2077: cohomology.  
2078: This results in an inductive algorithm to produce a basis, which is nicely
2079: expounded in~\cite{Tym} (see also \cite{GZ01}). 
2080: Pick a direction vector $\bv$ (corresponding to an element of the Lie algebra
2081: $\mathfrak t_\R$) which does not
2082: annihilate the direction vector of any edge.
2083: Orient each edge to have positive pairing with $\bv$, this is the Hasse diagram
2084: of a partial order on the fixed points induced by $\bv$.
2085: Then, using the relations described above, we can inductively construct a 
2086: triangular basis with respect to this ordering.
2087: That is, if $f=f(p)$ corresponds to the fixed point $p$, it vanishes at $q$
2088: ($f_q=0$) unless $p < q$, and $f_p$ is the product of weights of
2089: edges pointing down from $p$.
2090: While this algorithm was motivated by Morse theory, 
2091: it makes sense over any field $k$, if $\bv$ is a linear
2092: function on the character group of $T$ which does not annihilate any edge of the
2093: moment graph. 
2094: 
2095: 
2096: Set $\be:=\be_1-\be_2$ and pick the vector $\bv = \bff +\epsilon \be$, where
2097: $\epsilon>0$ is small.
2098: One basis element is the identity 
2099: $f(\includegraphics[width=14pt]{figures/11.00.11.eps})$,
2100: which localizes to 1 at each fixed point.
2101: We display each of the remaining nine in Figure~\ref{F:module} as a
2102: localization diagram, writing its localizations on a copy of a moment
2103: multigraph. 
2104: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2105: \begin{figure}[htb]
2106: \[
2107:   \begin{picture}(101,130)(-9,-11)
2108:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2109:                       \put(26,109){$2\bff$}
2110:      \put(-25,78){$\bff{-}\be$} \put(22,63){$\bff$} \put(66,78){$\bff{+}\be$}
2111:      \put(-25,55){$\bff{-}\be$}                     \put(66,55){$\bff{+}\be$}
2112:      \put(-25,32){$\bff{-}\be$} \put(36,46){$\bff$} \put(66,32){$\bff{+}\be$}
2113:                     \put(28.5,  0){0}
2114:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.00.20.eps}})$}
2115:      \thicklines \put(3.4,35.6){\Green{\circle{4}}}
2116:      \put(-50,20){\vector(1,4){7}}
2117:      \put(-58,28){$\bv$}
2118:   \end{picture}
2119:  \qquad
2120:   \begin{picture}(101,130)(-9,-11)
2121:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2122:                       \put(26,109){$2\bff$}
2123:      \put(-13,78){$2\bff$}\put(22,63){$\bff$} \put(66,78){$2\bff$}
2124:      \put(-8,55){$\bff$}                      \put(66,55){$\bff$}
2125:      \put(-8,32){0}       \put(36,46){$\bff$} \put(66,32){0}
2126:                     \put(28.5,  0){0}
2127:       \thicklines \put(3.4,58.5){\Green{\circle{4}}}
2128:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.10.10.eps}})$}
2129:   \end{picture}
2130:  \qquad
2131:   \begin{picture}(101,130)(-9,-11)
2132:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2133:                       \put(27,109){$\bff^2$}
2134:      \put(-12,78){$\bff^2$} \put(22,63){0} \put(66,78){$\bff^2$}
2135:      \put(-8,55){0}                      \put(66,55){0}
2136:      \put(-8,32){0}       \put(36,46){0} \put(66,32){0}
2137:                     \put(28.5, 0){0}
2138:      \thicklines \put(3.4,81.3){\Green{\circle{4}}}
2139:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.20.00.eps}})$}
2140:   \end{picture}
2141: \]
2142: \[
2143:   \begin{picture}(101,132)(-9,-11)
2144:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2145:                       \put(12,109){$\bff(\bff+\be)$}
2146:      \put(-9,78){0}\put(16,63){$\be\bff$}\put(66,78){$2\be\bff$}
2147:      \put(-9,55){0}                      \put(66,55){$\be\bff$}
2148:      \put(-9,32){0}       \put(36,46){0} \put(66,32){0}
2149:                     \put(28.5,  0){0}
2150:      \thicklines \put(30.3,60){\Green{\circle{4}}}
2151:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.10.eps}})$}
2152:   \end{picture}
2153: \qquad
2154:   \begin{picture}(101,132)(-9,-11)
2155:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2156:                       \put(12,109){$\bff(\bff+\be)$}
2157:      \put(-9,78){0} \put(22,63){0} \put(66,78){$2\be\bff$}
2158:      \put(-9,55){0}                      \put(66,55){$\be\bff$}
2159:      \put(-9,32){0} \put(36,46){$\be\bff$} \put(66,32){0}
2160:                     \put(28.5,  0){0}
2161:       \thicklines \put(33.3,57.2){\Green{\circle{4}}}
2162:     \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.10.01.eps}})$}
2163:   \end{picture}
2164: \qquad
2165:   \begin{picture}(101,132)(-9,-11)
2166:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2167:                       \put(28.5,109){0}
2168:      \put(-9,78){0} \put(22,63){0} \put(66,78){$\be(\be{-}\bff)$}
2169:      \put(-9,55){0}                \put(66,55){$\be^2$}
2170:      \put(-9,32){0} \put(36,46){0} \put(66,32){$\be(\be{+}\bff)$}
2171:                     \put(28.5,  0){0}
2172:       \thicklines \put(60.2,35.7){\Green{\circle{4}}}
2173:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.00.02.eps}})$}
2174:   \end{picture}
2175: \]
2176: \[
2177:   \begin{picture}(101,132)(-9,0)
2178:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2179:                       \put(10,109){$\bff^2(\bff+\be)$}
2180:      \put(-9,78){0} \put(22,63){0} \put(66,78){$2\be^2\bff$}
2181:      \put(-9,55){0}                \put(66,55){$\be^2\bff$}
2182:      \put(-9,32){0} \put(36,46){0} \put(66,32){0}
2183:                     \put(28.5,  0){0}
2184:       \thicklines \put(60.2,58.5){\Green{\circle{4}}}
2185:       \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.01.eps}})$}
2186:   \end{picture}
2187: \qquad
2188:   \begin{picture}(101,132)(-9,0)
2189:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2190:                       \put(10,109){$\bff^2(\bff+\be)$}
2191:      \put(-9,78){0} \put(22,63){0} \put(66,78){$2\be\bff^2$}
2192:      \put(-9,55){0}                \put(66,55){0}
2193:      \put(-9,32){0} \put(36,46){0} \put(66,32){0}
2194:                     \put(28.5,  0){0}
2195:       \thicklines \put(60.3,81.4){\Green{\circle{4}}}
2196:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.02.00.eps}})$}
2197:   \end{picture}
2198: \qquad
2199:   \begin{picture}(101,132)(-9,0)
2200:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2201:                       \put(5,109){$\bff^2(\bff^2-\be^2)$}
2202:      \put(-9,78){0} \put(22,63){0} \put(66,78){0}
2203:      \put(-9,55){0}                \put(66,55){0}
2204:      \put(-9,32){0} \put(36,46){0} \put(66,32){0}
2205:                     \put(28.5,  0){0}
2206:      \thicklines \put(31.7,104){\Green{\circle{4}}}
2207:      \put(-19,3){$f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.11.00.eps}})$}
2208:   \end{picture}
2209: \]
2210: \caption{An $S$-module basis for $A^*_T(\mathcal{Q}_2(0,2))$.}
2211: \label{F:module}
2212: \end{figure}
2213: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2214: 
2215: Set $x:=f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.10.10.eps}})-
2216:  \bff $, 
2217: $y:=f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.00.20.eps}})-
2218:  \bff $, 
2219: and 
2220: $z:=f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.10.01.eps}})-
2221:     f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.10.eps}})$.
2222: Figure~\ref{F:generators} shows their localization diagrams.
2223: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2224: \begin{figure}[htb]
2225: \[
2226:   \begin{picture}(100,130)(-9,0)
2227:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2228:                       \put(29,109){$\bff$}
2229:      \put(-9,78){$\bff$} \put(22,63){0} \put(66,78){$\bff$}
2230:      \put(-9,55){0}                \put(66,55){0}
2231:      \put(-19,32){$-\bff$} \put(36,46){0} \put(66,32){$-\bff$}
2232:                     \put(22,  0){$-\bff$}
2233:   \put(0,10){$x$}
2234:   \end{picture}
2235: \qquad
2236:   \begin{picture}(100,130)(-9,0)
2237:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2238:                       \put(29,109){$\bff$}
2239:      \put(-19,78){$-\be$} \put(22,63){0} \put(66,78){$\be$}
2240:      \put(-19,55){$-\be$}                \put(66,55){$\be$}
2241:      \put(-19,32){$-\be$} \put(36,46){0} \put(66,32){$\be$}
2242:                     \put(22,  0){$-\bff$}
2243:   \put(0,10){$y$}
2244:   \end{picture}
2245: \qquad
2246:   \begin{picture}(100,130)(-9,0)
2247:    \put(0,11){\includegraphics[height=95pt]{figures/Q202.skeleton.eps}}
2248:                       \put(28.5,109){0}
2249:      \put(-9,78){0} \put(15,63){$\be\bff$} \put(66,78){0}
2250:      \put(-9,55){0}                \put(66,55){0}
2251:      \put(-9,32){0} \put(35,46){$-\be\bff$} \put(66,32){0}
2252:                     \put(28.5,  0){0}
2253:   \put(0,10){$z$}
2254:   \end{picture}
2255: \]
2256: \caption{Generators for the $S_\Q$-algebra $A^*_T(\cQ_2(0,2))_\Q$.}
2257: \label{F:generators}
2258: \end{figure}
2259: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2260: We show that they generate $A^*_T(\cQ_2)_\Q$ as an $S_\Q$-algebra
2261: by showing that each basis element $f(p)$ of degree greater than 2
2262: lies in $S_\Q[x,y,z]$. 
2263: Since
2264: %
2265:  \begin{eqnarray}
2266:     f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.20.00.eps}})
2267:     &=&{\textstyle\frac{1}{2}} x(x{+}\bff)\,,\nonumber\\
2268:     f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.00.02.eps}})
2269:     &=& {\textstyle\frac{1}{2}}(y{+}\be)(y{-}x)\,,\ \
2270:     \mbox{and}\label{Eq:classes}\\
2271:      f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.10.01.eps}})
2272:    + f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.10.eps}})\,,
2273:     &=& (y{+}\be)(x{+}\bff)\,,\nonumber\
2274:  \end{eqnarray}
2275: %
2276: the four degree 2 basis elements lie in $S_\Q[x,y,z]$.
2277: The remaining three basis elements also lie in  $S_\Q[x,y,z]$,
2278: \[
2279:   y f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.10.eps}})\ =\ 
2280:     f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.01.eps}})\,,
2281:   \quad\quad
2282:   x f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.10.eps}})\ =\ 
2283:    f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.02.00.eps}})\,,
2284:   \quad\mbox{and}\quad
2285:    (y-\be) f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.01.01.eps}})\ =\ 
2286:    f(\raisebox{-3pt}{\includegraphics[width=14pt]{figures/11.00.11.eps}})\,.
2287: \]
2288: 
2289: Inspecting the localization diagrams of $x$, $y$, and $z$, shows that the
2290: following 5 expressions vanish in $A^*_T(\cQ_2)$
2291: %
2292:  \begin{equation}\label{E:GB}
2293:    xz\,,\quad yz\,,\quad x(x^2-\bff^2)\,,\quad
2294:    (y^2-\be^2)(y-x)\,,\quad \mbox{ and }\quad z^2-(y^2-\be^2)(x^2-\bff^2)\,.
2295:  \end{equation}
2296: %
2297: In the lexicographic term order where $z>y>x>e>f$, these five polynomials
2298: form a Gr\"obner basis for the ideal ${\mathcal I}$ of 
2299: $\Q[\be_1,\be_2,\bff, \ x, y, z]$ they generate 
2300: with leading terms $ xz$, $yz$, $x^3$, $y^3$, and $z^2$.
2301: There are ten standard monomials 
2302: \[
2303:    1,\ x,\ y,\ x^2,\ xy,\ y^2,\ z,\ x^2y,\ xy^2, \ \mbox{and}\ x^2y^2\,.
2304: \]
2305: Since $A^*_T(\cQ_2)_\Q$ is free over $S_\Q=\Q[\be_1,\be_2,\bff]$ of rank 10, we
2306: conclude that  
2307: \[
2308:     A^*_T(\cQ_2)_\Q \ \simeq\ \Q[\be_1,\be_2,\bff, \ x, y, z]/{\mathcal I}\,.
2309: \] 
2310: Using Proposition~\ref{P:Chow}, we obtain the presentation of the rational Chow
2311: ring 
2312: \[
2313:     A^*(\cQ_2)_\Q \ \simeq\ \Q[x, y, z]/\langle 
2314:     xz,\ yz,\ x^3,\ xy^2-y^3,\ z^2-x^2y^2 \rangle\,.
2315: \] 
2316: By~\eqref{Eq:classes}, the integral Chow ring has a more complicated
2317: presentation. 
2318: 
2319: 
2320: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2321: 
2322: We now consider Str\o mme's generators.
2323: Figure~\ref{F:Stromme1} shows the localization diagram of the first Chern class
2324: of $\cS$.
2325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2326: \begin{figure}[htb]
2327: \[
2328:   \begin{picture}(195,190)(-52,-24)
2329:    \put(-1,11){\includegraphics[height=140pt]{figures/Q202.skeleton.eps}}
2330:     \put(21,155){$\be_1{+}\be_2{+}2\bff$}
2331:     \put(-53,112){$\be_1{+}\be_2{+}2\bff$}
2332:     \put( 93,112){$\be_1{+}\be_2{+}2\bff$} 
2333:     \put(-50, 78){$\be_1{+}\be_2{+}\bff$} 
2334:     \put( 95, 78){$\be_1{+}\be_2{+}\bff$}
2335:     \put(-33, 44){$\be_1{+}\be_2$}
2336:     \put( 93, 44){$\be_1{+}\be_2$}
2337: 
2338:     \put(10,98){$\be_1{+}\be_2$}
2339:     \put(21,86){${+}\bff$}  
2340:     \put(50,70){$\be_1{+}\be_2$}
2341:     \put(61,58){${+}\bff$} 
2342:     \put(30, 1){$\be_1{+}\be_2$}
2343:     \put(-40,-24){Localization of $c_1^T(\cS)$ at $0\in\P^1$}
2344:   \end{picture}
2345:  \qquad
2346:   \begin{picture}(195,190)(-52,-24)
2347:    \put(-1,11){\includegraphics[height=140pt]{figures/Q202.skeleton.eps}}
2348:     \put( 30,155){$\be_1{+}\be_2$}
2349:     \put(-33,112){$\be_1{+}\be_2$}
2350:     \put( 93,112){$\be_1{+}\be_2$} 
2351:     \put(-50,78){$\be_1{+}\be_2{-}\bff$} 
2352:     \put( 95,78){$\be_1{+}\be_2{-}\bff$}
2353:     \put(-53,44){$\be_1{+}\be_2{-}2\bff$}
2354:     \put( 93,44){$\be_1{+}\be_2{-}2\bff$}
2355: 
2356:     \put(10,98){$\be_1{+}\be_2$}
2357:     \put(21,86){${-}\bff$}  
2358:     \put(50,70){$\be_1{+}\be_2$}
2359:     \put(61,58){${-}\bff$} 
2360:     \put(21, 1){$\be_1{+}\be_2{-}2\bff$}
2361:     \put(-40,-24){Localization of $c_1^T(\cS)$ at $\infty\in\P^1$}
2362:    \end{picture}
2363: \] 
2364:  \caption{First Chern class of $\cS$.} \label{F:Stromme1}
2365: \end{figure}
2366: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2367: By the formula of Proposition~\ref{Prop:kunneth}, we have
2368: \[
2369:    c^T_1(\cS)\ =\ \be_1+\be_2 + x \,-\, 2h\,,
2370: \]
2371: so that
2372: \[
2373:   t_1\ =\ \be_1+\be_2 + x\qquad\mbox{and}\qquad u_0\ =\ -2\,.
2374: \]
2375: 
2376: Figure~\ref{F:Stromme2} shows the localization diagram of the second Chern
2377: class of $\cS$.
2378: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2379: \begin{figure}[htb]
2380: \[
2381:   \begin{picture}(194,190)(-52,-24)
2382:    \put(-1,11){\includegraphics[height=140pt]{figures/Q202.skeleton.eps}}
2383:     \put(-6,155){$\be_1\be_2{+}(\be_1{+}\be_2)\bff{+}\bff^2$}
2384:     \put(-55,112){$\be_1\be_2{+}2\be_2\bff$}
2385:     \put( 93,112){$\be_1\be_2{+}2\be_1\bff$} 
2386:     \put(-52, 78){$\be_1\be_2{+}\be_2\bff$} 
2387:     \put( 95, 78){$\be_1\be_2{+}\be_1\bff$}
2388:     \put(-25, 44){$\be_1\be_2$}
2389:     \put( 93, 44){$\be_1\be_2$}
2390: 
2391:     \put(15,100){$\be_1\be_2$}
2392:     \put(15, 88){${+}\be_1\bff$}  
2393:     \put(50, 70){$\be_1\be_2$}
2394:     \put(50, 58){${+}\be_2\bff$} 
2395:     \put(37,  1){$\be_1\be_2$}
2396:     \put(-40,-24){Localization of $c_2^T(\cS)$ at $0\in\P^1$}
2397:   \end{picture}
2398: \qquad
2399:   \begin{picture}(194,190)(-52,-24)
2400:    \put(-1,11){\includegraphics[height=140pt]{figures/Q202.skeleton.eps}}
2401:     \put( 37,155){$\be_1\be_2$}
2402:     \put(-25,112){$\be_1\be_2$}
2403:     \put( 93,112){$\be_1\be_2$} 
2404:     \put(-52, 78){$\be_1\be_2{-}\be_2\bff$} 
2405:     \put( 95, 78){$\be_1\be_2{-}\be_1\bff$}
2406:     \put(-55, 44){$\be_1\be_2{-}2\be_2\bff$}
2407:     \put( 93, 44){$\be_1\be_2{-}2\be_1\bff$}
2408:     \put(-6,  1){$\be_1\be_2{-}(\be_1{+}\be_2)\bff{+}\bff^2$}
2409: 
2410:     \put(15,100){$\be_1\be_2$}
2411:     \put(15, 88){${-}\be_2\bff$}  
2412:     \put(50, 70){$\be_1\be_2$}
2413:     \put(50, 58){${-}\be_1\bff$} 
2414:     \put(-40,-24){Localization of $c_2^T(\cS)$ at $\infty\in\P^1$}
2415:   \end{picture}
2416: \]
2417:  \caption{Second Chern class of $\cS$.} \label{F:Stromme2}
2418: \end{figure}
2419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2420: By the formula of Proposition~\ref{Prop:kunneth}, we have
2421: \[
2422:    c^T_2(\cS)\ =\ \be_1\be_2 + 
2423:     {\textstyle\frac{1}{2}}x(y+ \be_1+\be_2) + {\textstyle\frac{1}{2}}z
2424:      \,-\, h(y+\be_1+\be_2)\,,
2425: \]
2426: so that 
2427: \[
2428:   t_2\ =\ \be_1\be_2 + 
2429:     {\textstyle\frac{1}{2}}x(y+ \be_1+\be_2) + {\textstyle\frac{1}{2}}z
2430:      \qquad\mbox{and}\qquad u_1\ =\ -(y+\be_1+\be_2)\,.
2431: \]
2432: 
2433: 
2434:  The corresponding classes in $A^*(\cQ_2)$ are
2435: \[
2436:    t_1\ =\ x\,,\quad
2437:    u_1\ =\ -y\,,\quad\mbox{and}\quad
2438:    t_2\ =\ {\textstyle\frac{1}{2}}(z+xy)\,.
2439: \]
2440:  Note that this shows that the claim in~\cite[Theorem (5.3)]{St87} 
2441: that the classes
2442:  $t_i,u_{i-1}$ generate the integral Chow ring is false.
2443: 
2444: \iffalse
2445: \begin{rmk}
2446:  Had we chosen the asymmetric (but integral) lift $(ii)$~$h=(0,\bff)$
2447:   from~\eqref{Eq:lifts} of the class of a point, the formula in
2448:  Proposition~\ref{Prop:kunneth} would be
2449: \[
2450:    t_{i,p}\ =\ c^T_{i,0,p}\qquad\mbox{and}\qquad
2451:    u_{i-1,p}\ =\ \frac{1}{\bff}(c^T_{i,\infty,p}-c^T_{i,0,p})\,.
2452: \]
2453:  This gives equivariant K\"unneth components
2454: \begin{eqnarray*}
2455:  t_1 &=& x + \be_1 + \be_2 + \bff\\
2456:  u_1 &=& y + \be_1 + \be_2\\
2457:  t_2 &=& {\textstyle\frac{1}{2}}z + 
2458:          {\textstyle\frac{1}{2}}(x+\bff)(y + \be_1 + \be_2) 
2459:          + \be_1\be_2\,.
2460: \end{eqnarray*}
2461: \end{rmk}
2462: \fi
2463: 
2464: %
2465: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2466: \providecommand{\bysame}{\leavevmode\hbox to3em{\hrulefill}\thinspace}
2467: \providecommand{\MR}{\relax\ifhmode\unskip\space\fi MR }
2468: % \MRhref is called by the amsart/book/proc definition of \MR.
2469: \providecommand{\MRhref}[2]{%
2470:   \href{http://www.ams.org/mathscinet-getitem?mr=#1}{#2}
2471: }
2472: \providecommand{\href}[2]{#2}
2473: \begin{thebibliography}{10}
2474: 
2475: \bibitem{Bertram}
2476: Aaron Bertram, \emph{Quantum {S}chubert calculus}, Adv. Math. \textbf{128}
2477:   (1997), no.~2, 289--305.
2478: 
2479: \bibitem{BB73}
2480: Andrzej Bia{\l}ynicki-Birula, \emph{Some theorems on actions of algebraic
2481:   groups}, Ann. of Math. (2) \textbf{98} (1973), 480--497.
2482: 
2483: \bibitem{brion}
2484: Michel Brion, \emph{Equivariant {C}how groups for torus actions}, Transform.
2485:   Groups \textbf{2} (1997), no.~3, 225--267.
2486: 
2487: \bibitem{ChSk74}
2488: Theodore Chang and Tor Skjelbred, \emph{The topological {S}chur lemma and
2489:   related results}, Ann. of Math. (2) \textbf{100} (1974), 307--321.
2490: 
2491: \bibitem{EG98a}
2492: Dan Edidin and William Graham, \emph{Equivariant intersection theory}, Invent.
2493:   Math. \textbf{131} (1998), no.~3, 595--634.
2494: 
2495: \bibitem{EG98b}
2496: \bysame, \emph{Localization in equivariant intersection theory and the {B}ott
2497:   residue formula}, Amer. J. Math. \textbf{120} (1998), no.~3, 619--636.
2498: 
2499: \bibitem{E95}
2500: David Eisenbud, \emph{Commutative algebra}, Graduate Texts in Mathematics, vol.
2501:   150, Springer-Verlag, New York, 1995.
2502: 
2503: \bibitem{evain}
2504: Laurent Evain, \emph{{The Chow ring of punctual Hilbert schemes of toric
2505:   surfaces}}.  {arXiv:math.AG/0503697}.
2506: 
2507: \bibitem{Fu98}
2508: William Fulton, \emph{Intersection theory}, second ed., Ergebnisse der
2509:   Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in
2510:   Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series
2511:   of Modern Surveys in Mathematics], vol.~2, Springer-Verlag, Berlin, 1998.
2512: 
2513: \bibitem{GH}
2514: Rebecca Goldin and Tara~S. Holm, \emph{The equivariant cohomology of
2515:   {H}amiltonian {$G$}-spaces from residual {$S\sp 1$} actions}, Math. Res.
2516:   Lett. \textbf{8} (2001), no.~1-2, 67--77.
2517: 
2518: \bibitem{GKM98}
2519: Mark Goresky, Robert Kottwitz, and Robert MacPherson, \emph{Equivariant
2520:   cohomology, {K}oszul duality, and the localization theorem}, Invent. Math.
2521:   \textbf{131} (1998), no.~1, 25--83.
2522: 
2523: \bibitem{GZ01}
2524: V.~Guillemin and C.~Zara, \emph{1-skeleta, {B}etti numbers, and equivariant
2525:   cohomology}, Duke Math. J. \textbf{107} (2001), no.~2, 283--349.
2526:   
2527: 
2528: \bibitem{GZ02}
2529: Victor Guillemin and Catalin Zara, \emph{Combinatorial formulas for products of
2530:   {T}hom classes}, Geometry, mechanics, and dynamics, Springer, New York, 2002,
2531:   pp.~363--405.
2532: 
2533: \bibitem{GZ03}
2534: \bysame, \emph{The existence of generating families for the cohomology ring of
2535:   a graph}, Adv. Math. \textbf{174} (2003), no.~1, 115--153.
2536: 
2537: \bibitem{I}
2538: Birger Iversen, \emph{A fixed point formula for action of tori on algebraic
2539:   varieties}, Invent. Math. \textbf{16} (1972), 229--236.
2540: 
2541: \bibitem{St87}
2542: Stein~Arlid Str{\o}mme, \emph{On parameterized rational curves in {G}rassmann
2543:   varieties}, Space Curves (F.~Ghione, C.~Peskine, and E.~Sernesi, eds.),
2544:   Lecture Notes in Math., vol. 1266, Springer-Verlag, 1987, pp.~251--272.
2545: 
2546: \bibitem{Tym}
2547: Julianna~S. Tymoczko, \emph{An introduction to equivariant cohomology and
2548:   homology, following {G}oresky, {K}ottwitz, and {M}ac{P}herson}, Snowbird
2549:   lectures in algebraic geometry, Contemp. Math., vol. 388, Amer. Math. Soc.,
2550:   Providence, RI, 2005, pp.~169--188.
2551: 
2552: \end{thebibliography}
2553: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2554: \end{document}
2555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2556: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2557: \bibliographystyle{amsplain}
2558: \bibliography{quot}
2559: 
2560: \end{document}
2561: 
2562: