1: \section{Introduction}
2:
3: Let $X$ be a non-singular algebraic variety and $E_0 \to E_1 \to
4: \cdots \to E_n$ a sequence of vector bundles and bundle maps over $X$.
5: A set of {\em rank conditions\/} for this sequence is a collection $r
6: = \{ r_{ij} \}$ of non-negative integers, for $0 \leq i < j \leq n$.
7: This data defines the quiver variety
8: \[ \Omega_r(E_\bull) = \{ x \in X \mid
9: \rank(E_i(x) \to E_j(x)) \leq r_{ij} ~\forall i < j \} \,,
10: \]
11: which comes with a natural structure of subscheme of $X$, given as the
12: intersection of the zero sections of the maps $\bigwedge^{r_{ij}+1}
13: E_i \to \bigwedge^{r_{ij}+1} E_j$. We demand that the rank conditions
14: can {\em occur\/} as the ranks over a point in $X$. If we set $r_{ii}
15: = \rank(E_i)$, then this is equivalent to the conditions that $r_{ij}
16: \leq \min ( r_{i,j-1}, r_{i+1,j} )$ for all $0 \leq i < j \leq n$ and
17: $r_{ij} + r_{i-1,j+1} \geq r_{i-1,j} + r_{i,j+1}$ for all $0 < i \leq
18: j < n$. In this case the expected codimension of the quiver variety
19: $\Omega_r(E_\bull)$ is the integer $d(r) = \sum_{i < j}
20: (r_{i,j-1}-r_{ij}) (r_{i+1,j}-r_{ij})$.
21:
22: In joint work with Fulton \cite{buch.fulton:chern} we established a
23: formula for the cohomology (or Chow) class of the quiver variety when
24: this codimension is attained. This was generalized in
25: \cite{buch:grothendieck} to the following formula for the structure
26: sheaf of a quiver variety in the Grothendieck ring $K(X)$ of algebraic
27: vector bundles on $X$:
28: \begin{equation} \label{E:orig_kqf}
29: [\OO_{\Omega_r(E_\bull)}] \ = \ \sum_\mu
30: c_\mu(r)\, G_{\mu_1}(E_1-E_0)\, G_{\mu_2}(E_2-E_1) \cdots
31: G_{\mu_n}(E_n-E_{n-1}) \,.
32: \end{equation}
33: Here the sum is over finitely many sequences $\mu =
34: (\mu_1,\dots,\mu_n)$ of partitions $\mu_i$ such that the sum $\sum
35: |\mu_i|$ of the weights is greater than or equal to $d(r)$. The
36: stable Grothendieck polynomials $G_{\mu_i}(E_i-E_{i-1})$ are defined
37: in section \ref{sec:grothpoly}. The {\em quiver coefficients\/}
38: $c_\mu(r)$ appearing in this formula are integers which are uniquely
39: determined by the condition that (\ref{E:orig_kqf}) is true for all
40: varieties $X$ and bundle sequences $E_\bull$, together with the
41: condition that $c_\mu(r) = c_\mu(r+m)$ holds for all $m \in \N$, where
42: $r+m = \{ r_{ij} + m \}$ is the rank conditions obtained by adding the
43: integer $m$ to the original rank conditions.
44:
45: A formula for quiver coefficients was also given in
46: \cite{buch.fulton:chern, buch:grothendieck}; in the case of
47: $K$-theory, this is based on the algebra of stable Grothendieck
48: polynomials constructed in \cite{buch:littlewood-richardson}.
49: Although the original formulas for quiver coefficients do not keep
50: track of their signs, it was conjectured that the cohomological quiver
51: coefficients (given by sequences $\mu$ such that $\sum |\mu_i| =
52: d(r)$) are non-negative, while the $K$-theoretic quiver coefficients
53: have signs that alternate with codimension, i.e.\ $(-1)^{\sum |\mu_i|
54: - d(r)} c_\mu(r) \geq 0$. Special cases of of these conjectures
55: have been proved in \cite{buch.fulton:chern, buch:on, buch:stanley,
56: buch:grothendieck, buch.kresch.ea:schubert,
57: buch.kresch.ea:grothendieck}.
58:
59: In their recent paper \cite{knutson.miller.ea:four}, Knutson, Miller,
60: and Shimozono deliver a breakthrough within the theory of quiver
61: formulas, and prove at least two explicit combinatorial formulas for
62: the cohomological quiver coefficients, which show that these
63: coefficients are non-negative. One of the important ideas in their
64: work is to reinterpret the {\em lace diagrams\/} of Abeasis and
65: Del-Fra \cite{abeasis.del-fra:degenerations} as sequences of partial
66: permutations. This interpretation is explained by a Gr{\"o}bner
67: degeneration of a quiver variety in a matrix space into a union of
68: products of matrix Schubert varieties. The {\em component formula\/}
69: of \cite{knutson.miller.ea:four} writes the cohomology class of a
70: quiver variety as a sum, over all `{\em minimal\/}' lace diagrams, of
71: the products of the Schubert polynomials for the corresponding partial
72: permutations. The proof that quiver coefficients are non-negative is
73: obtained by proving a stable version of this component formula, where
74: the Schubert polynomials are replaced with Stanley symmetric
75: functions. This is sufficient because Stanley symmetric functions are
76: known to be Schur positive \cite{edelman.greene:balanced,
77: lascoux.schutzenberger:structure}.
78:
79: Knutson, Miller, and Shimozono also prove a {\em ratio formula}, which
80: writes the class of a quiver variety as a quotient of two Schubert
81: polynomials. This formula follows from a careful analysis of the
82: Zelevinsky map \cite{zelevinsky:two, lakshmibai.magyar:degeneracy},
83: and is in fact established for both cohomology and $K$-theory. The
84: component formulas are proved using a combination of the Gr{\"o}bner
85: degeneration and the ratio formula, as well as a combinatorial study
86: of a double version of the ratio formula. In particular, it is proved
87: that a limit of the double ratio formula agrees with the {\em double
88: quiver functions\/} introduced in \cite{buch.fulton:chern,
89: buch:grothendieck} and named in \cite{knutson.miller.ea:four}. The
90: authors of \cite{knutson.miller.ea:four} have informed us that they
91: can generalize their methods to work in $K$-theory, although,
92: according to their own description, this approach is rather
93: complicated.\footnote{This work is now available in the paper
94: \cite{miller:alternating}, which proves a $K$-theoretic component
95: formula and a double version of it, but only in the stable case and
96: for large rank conditions.}
97: % (this work is now available \cite{miller:alternating}.)
98:
99: In this paper we give simpler proofs of the above mentioned formulas,
100: using methods that work equally well in $K$-theory. In particular, we
101: prove that the $K$-theoretic quiver coefficients have alternating
102: signs, and we derive an explicit combinatorial formula for these
103: coefficients. Starting from the ratio formula, we give combinatorial
104: proofs of $K$-theoretic generalizations of the component formulas,
105: where the Schubert polynomials and Stanley symmetric functions are
106: replaced with ordinary and stable Grothendieck polynomials. These
107: formulas are given in terms of sequences of partial permutations,
108: which we call {\em KMS-factorizations\/} of the {\em Zelevinsky
109: permutation\/} defined in \cite{knutson.miller.ea:four}. To
110: conclude that $K$-theoretic quiver coefficients have alternating
111: signs, we use Lascoux's result that stable Grothendieck polynomials
112: are linear combinations with alternating signs of Stable Grothendieck
113: polynomials for partitions \cite{lascoux:transition}. To make our
114: paper self-contained, we also give a short proof of the ratio formula.
115:
116: The {\em factor sequences conjecture\/} of \cite{buch.fulton:chern}
117: states that cohomological quiver coefficients count the number of
118: sequences of semistandard Young tableaux which can be generated by a
119: sequence of factorizations and multiplications of chosen tableaux
120: arranged in a {\em tableaux diagram}. A special case of this
121: conjecture, corresponding to a particular choice of tableaux diagram,
122: was proved in \cite{knutson.miller.ea:four}. However, so far there
123: has been no progress in generalizing this conjecture to $K$-theory.
124: In this paper we close this gap by showing that KMS-factorizations can
125: be defined by the same algorithm as defines factor sequences, except
126: that the tableau diagram is replaced with a diagram of permutations,
127: and the plactic product of tableaux is replaced with multiplication of
128: permutations in the degenerate Hecke algebra. A tableau-based version
129: of this type of factor sequences also exists; we briefly outline this
130: for cohomological quiver coefficients, and refer to \cite{bkty:stable}
131: for details and the general case.
132: %A tableau-based variant
133: %of this type of factor sequences will be discussed in a subsequent
134: %paper \cite{bkty:stable}.
135:
136: We remark that it was already known that the cohomological component
137: formulas can be derived combinatorially from the ratio formula, by
138: using a simplification of Yong \cite{yong:embedding}. However, Yong's
139: method still requires the analysis of the double ratio formula and its
140: limits from \cite{knutson.miller.ea:four}. The approach presented
141: here simplifies things further by working only with the single ratio
142: formula, by applying Fomin and Kirillov's construction of Grothendieck
143: polynomials based on solutions to Yang-Baxter equations
144: \cite{fomin.kirillov:yang-baxter, fomin.kirillov:grothendieck}, and by
145: observing that the stable component formula follows easily from the
146: non-stable component formula.
147:
148: Other simple proofs of the cohomological component formulas have also
149: surfaced. For example, they can be deduced very easily from the Thom
150: polynomial theory developed by Feh\'er and Rim\'anyi
151: \cite{feher.rimanyi:classes}, or deduced directly from the above
152: mentioned Gr{\"o}bner degeneration with a symmetry argument. This is
153: explained in \cite{buch.feher.ea:positivity}. While attempts to
154: generalize these methods to $K$-theory have not been successful, they
155: might hold more promise for quiver varieties of other types (see
156: \cite{feher.rimanyi:classes}).
157:
158: Some of the results proved in \cite{knutson.miller.ea:four} imply that
159: the cohomological double ratio formula for large rank conditions
160: satisfies nice properties, including {\em multi supersymmetry\/} and a
161: double version of the component formula. In the last section of this
162: paper, we establish these properties for the $K$-theoretic double
163: ratio formula given by arbitrary rank conditions. In particular, we
164: prove a conjecture from \cite{knutson.miller.ea:four} stating that the
165: double ratio formula satisfies a {\em rank stability\/} property.
166: This conjecture is equivalent to the statement that the polynomials
167: defined by the double ratio formula are specializations of the
168: original quiver formulas \cite{buch.fulton:chern, buch:grothendieck}.
169: Even though the double ratio formula is not needed for the proof of
170: alternating signs of quiver coefficients given in this paper, its
171: multi supersymmetry property has some nice applications. For example,
172: this property was used in \cite{knutson.miller.ea:four} to prove the
173: above mentioned case of the factor sequences conjecture. The multi
174: supersymmetry property also implies that general quiver coefficients
175: are special cases of the coefficients studied in
176: \cite{buch.kresch.ea:grothendieck}. In fact, quiver coefficients can
177: be realized as Schubert structure constants on flag varieties
178: \cite{bergeron.sottile:schubert, lenart.robinson.ea:grothendieck,
179: buch.sottile.ea:quiver}.
180:
181: This paper is organized as follows. In section \ref{sec:grothpoly} we
182: explain Fomin and Kirillov's construction of Grothendieck polynomials,
183: which is a key ingredient in the combinatorial parts of this paper.
184: Section \ref{sec:zel} gives a new construction of the Zelevinsky
185: permutation which is required for our proof that KMS-factorizations
186: can be viewed as factor sequences. Section \ref{sec:ratio} contains
187: the proof of the ratio formula. In section~\ref{sec:restrict} we
188: prove a formula for double Grothendieck polynomials applied to certain
189: rearrangements of the same set of variables, which in section
190: \ref{sec:component} is used to derive the non-stable component formula
191: from the ratio formula. In section \ref{sec:facseq} we establish the
192: factor sequences definition of KMS-factorizations and discuss its
193: consequences. As a corollary we obtain a rank stability property for
194: KMS-factorizations, which in section \ref{sec:altsigns} is used to
195: derive the stable component formula and deduce that quiver
196: coefficients have alternating signs. Section \ref{sec:double} finally
197: proves the above mentioned properties of the double ratio formula.
198:
199: We are very grateful to Rich\'ard Rim\'anyi for discussions at the
200: Banach Institute in Warsaw, which led to our observation that the
201: stable component formula can be deduced from the non-stable formula,
202: and which triggered our search for other simplifications to
203: \cite{knutson.miller.ea:four}. We also thank Feh{\'e}r, Kresch,
204: Sottile, Tamvakis, and Yong for inspiring collaboration on related
205: papers, and Fulton, Miller, and Sottile for helpful comments to our
206: paper. Finally, we thank Martin Guest, Anatol Kirillov, and the
207: Research Institute for Mathematical Sciences in Kyoto for their
208: hospitality while this paper was written.
209:
210:
211: %%% Local Variables:
212: %%% mode: latex
213: %%% TeX-master: "as2"
214: %%% End:
215: