1: \documentclass[10pt,reqno]{amsart}
2: \usepackage{amssymb,accents}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \def\noi{\noindent}
5:
6: \newtheorem{Thm}{Theorem}[section]
7: \newtheorem{Def}[Thm]{Definition}
8: \newtheorem{Lm}[Thm]{Lemma}
9: \newtheorem{Prop}[Thm]{Proposition}
10: \newtheorem{Cor}[Thm]{Corollary}
11: \newtheorem{Rem}[Thm]{Remark}
12: \newtheorem{state}{Theorem}
13: \def\baselinestretch{1.2}
14: \setlength{\parskip}{4pt}
15:
16: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
17:
18: \def\cal{\mathcal}
19: \def\Bbb{\mathbb}
20: \def\mf{\mathfrak}
21:
22: \def\<{\langle}
23: \def\>{\rangle}
24:
25: \def\a{\alpha}
26: \def\b{\beta}
27: \def\d{\delta}
28: \def\D{\Delta}
29: \def\th{\theta}
30: \def\eps{\varepsilon}
31: \def\l{\lambda}
32: \def\L{\Lambda}
33: \def\Li{\cal L}
34: \def\vphi{\varphi}
35:
36: \def\Re{\Bbb R}
37: \def\F{\Bbb F}
38: \def\C{\Bbb C}
39: \def\Z{\Bbb Z}
40: \def\N{\Bbb N}
41:
42: \def\X{\mf X}
43: \def\T{\mf T}
44: \def\R{\mf R}
45: \def\S{\cal S}
46: \def\P{\cal P}
47: \def\H{\cal H}
48: \def\O{\mf O}
49: \def\K{\mf K}
50: \def\A{\cal A}
51: \def\q{\cal C}
52: \def\D{\cal D}
53: \def\HH{\tilde \cal H}
54: \def\G{\mf h}
55:
56:
57:
58: \def\W{\ring W}
59: \def\w{\ring w}
60: \def\RR{\ring R}
61: \def\Q{\ring Q}
62: \def\h{\ring {\cal H}}
63: \def\Gc{\ring {\mf h}}
64: \def\E{\ring E}
65: \def\bo{\ring {\mf b}}
66: \def\p{\mf p}
67: \def\t{\mf t}
68: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
69: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
70: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
71:
72:
73: \begin{document}
74: \title[]
75: {A weight multiplicity formula for Demazure modules}
76: \author{Bogdan Ion}\thanks{Supported in part by a Rackham Faculty
77: Research Fellowship}
78: %\thanks{Department of Mathematics, University of Michigan, Ann
79: %Arbor, MI 48109.}
80: %\thanks{E-mail address: {\tt bogdion@umich.edu}}
81: \address{Department of Mathematics, University of Michigan, Ann
82: Arbor, MI 48109} \email{bogdion@umich.edu}
83: \begin{abstract}
84: We establish a formula for the weight multiplicities of Demazure
85: modules (in particular for highest weight representations) of a
86: complex connected algebraic group in terms of the geometry of its
87: Langlands dual.
88: \end{abstract}
89: \maketitle
90:
91: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
92: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
93: \thispagestyle{empty}
94: \section*{Introduction}
95:
96: Let $G$ be a complex, connected, simple algebraic group and let
97: $B$ and $T$ be a Borel subgroup and maximal torus of $G$ contained
98: in $B$. For an integral dominant weight $\l$, denote by $V_\l$ the
99: integrable highest weight $G$-module with highest weight $\l$. The
100: Demazure modules $D_{w(\l)}$ (with $w$ running over the Weyl group
101: of $G$) form a filtration of $V_\l$. The module $D_{w(\l)}$ is
102: defined as the $B$--submodule of $V_\l$ generated by the extremal
103: weight space of weight $w(\l)$ (which are one--dimensional). In
104: particular, if $w_\circ$ is the longest element in the Weyl group
105: of $G$, $D_{w_\circ(\l)}$ is precisely $V_\l$. More importantly,
106: the Demazure modules can be realized as spaces of global sections
107: of certain line bundles over Schubert varieties (see Section
108: \ref{demazure} for details).
109:
110: The main goal of this note is to explain how the recent
111: developments \cite{ion2}, \cite{ion3} in the theory of
112: nonsymmetric Macdonald polynomials have as consequence a geometric
113: formula for weight multiplicities in Demazure modules and, in
114: particular, irreducible $G$--modules. We will briefly state our
115: result.
116:
117: Let ${G^\vee}$ be the unique complex, connected, reductive group
118: whose root datum is dual to that of $G$. Let us also consider
119: $B^\vee$, a Borel subgroup of $G^\vee$. The group $G^\vee$ is also
120: defined over $\cal O:=\C[[x]]$. We denote $G^\vee(\cal O)$ by $K$
121: and let $I$ be the subgroup of $K$ defined as the inverse image of
122: $B^\vee(\C)$ under the reduction map $K\to G^\vee(\C)$. The space
123: $G^\vee(\C((x)))/I$ is endowed with a structure of ind--variety
124: over $\C$. For any integral weights $\l$ and $\mu$ let
125: $M_{\l,\mu}(\C)$ be the finite dimensional subvariety of
126: $G^\vee(\C((x)))/I$ defined in Section \ref{varieties}.
127:
128: \begin{state}\label{thm1}
129: Let $\l$ and $\mu$ be two integral weights such that $\mu\leq \l$.
130: The weight multiplicity $m_{\l,\mu}$ of the weight $\l$ in the
131: Demazure module $D_\l$ equals the number of top dimensional
132: irreducible components of $M_{\l,\mu}(\C)$.
133: \end{state}
134:
135: Other geometric formulas for weight multiplicities in irreducible
136: $G$--modules exist in the literature \cite{lusztig}, \cite{mv} but
137: no such formula was available for Demazure modules. Our result is
138: similar to one consequence \cite[Corollary 7.4]{mv} of the recent
139: work of Mirkovi\' c and Vilonen on the geometric Satake
140: isomorphism (see \ref{remarks} on further comments about this
141: connection). However, we should keep in mind that the Demazure
142: modules are generally just $B$--modules and their result covers
143: only the case of reductive groups; it would be interesting to
144: investigate if the ideas in \cite{mv} could be used to cover a
145: larger class of groups.
146:
147: \noi {\sl Acknowledgement:} I would like to thank Brian Conrad for
148: a clarifying conversation.
149: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
150: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
151: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
152: \section{Preliminaries}\label{prelim}
153:
154: \subsection{}
155: Let ${\mf g}$ be the complex, simple Lie algebra of rank $n$ and
156: let $\mf h$ and $\mf b$ be a Cartan subalgebra and, respectively,
157: a Borel subalgebra containing $\mf h$. Let $\RR$ be root system of
158: $\mf g$ with respect to $\mf h$ and denote by $R^+$ the set of
159: roots of $\mf b$ with respect to $\mf h$. The set of positive
160: simple roots is denoted by $\{\a_1,\dots,\a_n \}$ and
161: $\{\l_1,\cdots,\l_n\}$ will denote the fundamental weights. As
162: usual, $\Q$ denotes the root lattice of $\RR$ and $P$ denotes its
163: weight lattice.
164:
165:
166:
167: The vector space ${\mf h}_\Re^*$ (the real vector space spanned by
168: the roots) has a canonical scalar product $(\cdot, \cdot)$ which
169: we normalize such that it gives square length 2 to the short roots
170: in $\RR$ (if there is only one root length we consider all roots
171: to be short). We will use $\RR_s$ and $\RR_\ell$ to refer to the
172: short and respectively long roots in $\RR$. We will identify the
173: vector space $\mf h_\Re$ (the real vector space spanned by the
174: coroots) with its dual using this scalar product. Under this
175: identification $\a^\vee=2\a/(\a,\a)$ for any root $\a$. The root
176: $\th$ is defined as the {\sl highest short root} in $\RR$. Also,
177: let us consider $$\rho=\frac{1}{2}\sum_{\a\in R}\a^\vee$$
178:
179: To the finite root system $\RR$ we will associate an {\sl affine
180: root system} $R$. Let ${\rm Aff}(\mf h_\Re)$ be the space of
181: affine linear transformations of $\mf h_\Re$. As a vector space,
182: it can be identified to $\mf h_\Re^*\oplus \Re\d$ via
183: $$
184: (f+c\d)(x)=f(x)+c, \quad \text{for~} f\in \mf h_\Re^*, ~x\in\mf
185: h_\Re \text{~and~} c\in \Re
186: $$
187: Let $r$ denote the maximal number of laces connecting two vertices
188: in the Dynkin diagram of $\RR$. Then,
189: $$R:=(\RR_s+\Z\d)\cup(\RR_\ell +r\Z\d)\subset \mf h_\Re^*\oplus \Re\d$$
190:
191:
192: The set of affine positive roots $R^+$ consists of affine roots of
193: the form $\a+k\d$ such that $k$ is non--negative if $\a$ is a
194: positive root, and $k$ is strictly positive if $\a$ is a negative
195: root. The affine simple roots are $\{\a_i\}_{0\leq i\leq n}$,
196: where we set $\a_0:=\d-\th$.
197:
198: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
199:
200: \subsection{}
201: The scalar product on $\mf h_\Re^*$ can be extended to a
202: non--degenerate bilinear form on the real vector space
203: $$H:=\mf h_\Re^*\oplus \Re\d\oplus \Re\L_0$$ by requiring that $(\d,\mf h_\Re^*\oplus \Re\d)=
204: (\L_0,\mf h_\Re^*\oplus \Re\L_0)=0$ and $(\d,\L_0)=1$. Given
205: $\a\in R$ and $x\in H$ let
206: $$
207: s_\a(x):=x-\frac{2(x,\a)}{(\a,\a)}\a\
208: $$
209: The {\sl affine Weyl group} $W$ is the subgroup of ${\rm GL}(H)$
210: generated by all $s_\a$ (the simple reflections $s_i=s_{\a_i}$ are
211: enough). The {\sl finite Weyl group} $\W$ is the subgroup
212: generated by $s_1,\dots,s_n$. It is easy to see that $\mf
213: h_\Re^*+\Re\d+\L_0$ is stable under the action of $W$. Therefore,
214: if we identify $\mf h_\Re^*$ with $\mf h_\Re^*+\Re\d+\L_0$ we
215: obtain an affine action of $W$ on $\mf h_\Re^*$; we denote by
216: $w\cdot x$ the affine action of $w\in W$ on $x\in \mf h_\Re^*$.
217: For example, the affine action of $s_0$ can be described as
218: follows
219: \begin{equation}
220: s_0\cdot x =s_\th(x)+\th\
221: \end{equation}
222:
223: We define the fundamental affine chamber as
224: \begin{equation}\label{affinechamber}
225: \cal C:=\{ x\in \mf h_\Re^*\ | \ (x+\L_0,\a_i^\vee)\geq 0\ ,\
226: 0\leq i\leq n\}
227: \end{equation}
228: The non-zero elements of $\cal O_P:=P\bigcap \cal C$ are the
229: so-called minuscule weights. Let us remark that each orbit of the
230: affine action of $W$ on $P$ contains the origin or a unique
231: element of $\cal O_P$.
232:
233: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
234: \subsection{}
235:
236:
237: For each $w$ in $W$ let $\ell(w)$ be the length of a reduced (i.e.
238: shortest) decomposition of $w$ in terms of simple reflections. The
239: length of $w$ can be also geometrically described as follows. For
240: any affine root $\a$, denote by $H_\a$ the affine hyperplane
241: consisting of fixed points of the affine action of $s_\a$ on $\mf
242: h_\Re^*$. Then, $\ell(w)$ equals the number of affine hyperplanes
243: $H_\a$ separating $\cal C$ and $w\cdot\cal C$. For any affine
244: transformation of $\mf h_\Re^*$ which preserves the set of
245: hyperplanes $\{H_\a\}_{\a\in R}$, we can use the geometric point
246: of view to define the length of that transformation. For example,
247: for a weight $\l$ we define the following affine transformation of
248: $\mf h_\Re^*$
249: $$
250: \tau_\l(x)=x+\l
251: $$
252: It is easy to check that $\tau_\l$ has the required properties to
253: allow us to talk about its length. In fact a concrete formula for
254: its length is available (see, for example, \cite[(5)]{ion2})
255: \begin{equation}\label{length}
256: \ell(\tau_\l)=\sum_{\a\in\RR} |(\l,\a^\vee)|
257: \end{equation}
258:
259:
260: For each weight $\l$ define $\l_-$, respectively $\tilde\l$, to be
261: the unique element in $\W(\l)$, respectively $W\cdot\l$, which is
262: an anti-dominant weight, respectively an element of $\cal O_P$
263: (that is a minuscule weight or zero), and
264: $\w_\l\hspace{-0.2cm}^{-1}\in\ \W$, respectively $w_\l^{-1}\in W$,
265: to be the unique minimal length element by which this is achieved.
266: Also, for each weight $\l$ define $\l_+$ to be the unique element
267: in $\W(\l)$ which is dominant and denote by $w_\circ$ the maximal
268: length element in $\W$. It was shown in \cite[Lemma 1.7 (3)]{ion2}
269: that the following equality holds for any weight $\l$
270: \begin{equation}\label{eq3}
271: \ell(\tau_\l)=\ell(w_\l)+\ell(\w_\l)
272: \end{equation}
273: For later use we record the following.
274: \begin{Lm}\label{lemma1}
275: Let $\l$ be a weight. Then
276: $$
277: 2\<\l,\rho\>\leq \ell(w_\l)+\ell(\w_\l)
278: $$
279: \end{Lm}
280: \begin{proof}
281: From (\ref{length}) we know that $2\<\l,\rho\>\leq \ell(\tau_\l)$.
282: Now, (\ref{eq3}) implies the desired result.
283: \end{proof}
284:
285: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
286:
287: \subsection{}
288: The Bruhat order is a partial order on any Coxeter group defined
289: in way compatible with the length function. For an element $w$ we
290: put $w<s_iw$\ if and only if \ $\ell(w)<\ell(s_iw)$. The
291: transitive closure of this relation is called the Bruhat order.
292: The terminology is motivated by the way this ordering arises for
293: Weyl groups in connection with inclusions among closures of Bruhat
294: cells for a corresponding semisimple algebraic group.
295:
296: We can use the Bruhat order on $W$ to define a partial order on
297: the weight lattice which will also be called the Bruhat order. For
298: any $\l,\mu\in P$ we write
299: \begin{equation}\label{order}
300: \l<\mu\ \ \text{if and only if}\ \ w_\l<w_\mu
301: \end{equation}
302: The minimal elements of the weight lattice with respect to this
303: partial order are the minuscule weights and if $\l<\mu$ then
304: necessarily $\tilde \l=\tilde \mu$.
305:
306: \begin{Lm}\label{lemma2}
307: Let $\l$ and $\mu$ be two weights such that $\mu\leq \l$. Define
308: the rational number
309: \begin{equation}\label{n}
310: n_{\l,\mu}:=\frac{1}{2}\ell(w_\l)-\frac{1}{2}\ell(\w_\l)+\ell(w_\circ)-\<\mu,\rho\>
311: \end{equation}
312: Then $n_{\l,\mu}$ is a positive integer.
313: \end{Lm}
314: \begin{proof} Let us argue first that $n_{\l,\mu}$ is positive.
315: We can use Lemma \ref{lemma1} to obtain
316: $$
317: n_{\l,\mu}\geq
318: \frac{1}{2}\ell(w_\l)-\frac{1}{2}\ell(\w_\l)+\ell(w_\circ)-
319: \frac{1}{2}\ell(w_\mu)-\frac{1}{2}\ell(\w_\mu)
320: $$
321: Keeping in mind that $\ell(w_\l)\geq\ell(w_\mu)$ (which is a
322: consequence of the hypothesis) and that $w_\circ$ is the maximal
323: length element in $\W$, our claim immediately follows.
324:
325: To show that $n_{\l,\mu}$ is integer, let us remark that it is
326: enough to check that
327: $\frac{1}{2}\ell(w_\l)+\frac{1}{2}\ell(\w_\l)$ is integer. This
328: fact follows from (\ref{eq3}) together with
329: $$\ell(\tau_\l)=\ell(\tau_{\l_+})=2\<\l_+,\rho\>$$
330: \end{proof}
331: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
332: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
333: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
334:
335:
336: \section{Demazure modules}\label{demazure}
337:
338: Let $G$ be a complex, connected, simple algebraic group with Lie
339: algebra $\mf g$ and denote by $T$ and $B$ the maximal torus and
340: the Borel subgroup of $G$ with Lie algebras $\mf h$ and $\mf b$,
341: respectively.
342:
343: For an integral dominant weight $\l$, denote by $V_\l$ the
344: integrable highest weight ${\mf g}$-module with highest weight
345: $\l$. It is well known that $V_\l$ is an irreducible $\mf
346: g$--module. Hence, for any $w$ in $\W$ the $w(\l)$--weight space
347: $V_{\l,w(\l)}$ is one--dimensional. The {\sl Demazure module}
348: $D_{w(\l)}$ is defined as the ${\mf b}$--module generated by
349: $V_{\l, w(\l)}$. Since $\l$ is integral $D_{w(\l)}$ is also a
350: $B$--module. The Demazure modules associated to a fixed integral
351: dominant weight $\l$ form a filtration (with respect to the Bruhat
352: order on $\W$) of $V_\l$. In particular, $D_\l=V_{\l,\l}$ and
353: $D_{w_\circ(\l)}=V_\l$.
354:
355:
356: There exists an important geometrical construction of Demazure
357: modules which relates them with the geometry of Schubert varieties
358: in the flag variety $G/B$. For any $w$ in $\W$ let
359: $$
360: X_w=\overline{BwB/B}\subseteq G/B
361: $$
362: denote the Schubert variety associated to $w$. The Schubert
363: varieties are closed finite dimensional projective irreducible
364: subvarieties of $G/B$. For any integral weight $\l$ we denote by
365: $e^\l$ the character of $B$ obtained from $T$ via the isomorphism
366: $T\simeq B/[B,B]$. Consider the fiber product $\cal
367: L_\l=G\times_B\C_\l$, where $\C_\l$ denotes $\C$ equipped with the
368: $B$--action given by the character $e^\l$. The natural projection
369: $G\times\C_\l\to G$ induces a well defined $G$--equivariant
370: holomorphic map $\cal L_\l\to G/B$; in other words $\cal L_\l$
371: becomes a $G$--equivariant holomorphic line bundle over $G/B$. By
372: restriction we obtain a line bundle $\cal L_{\l,w}$ over $X_w$.
373: Since $X_w$ is $B$--invariant, the space of holomorphic sections
374: $H^0(X_w,\cal L_{\l,w})$ admits a $B$--module structure. The
375: relation with the Demazure module $D_{w(\l)}$ is the following.
376: \begin{Thm} Let $\l$ be a dominant integral weight and let $w$ be an element of $W$.
377: Then, $D_{w(\l)}$ and $H^0(X_w,\cal L_{-\l, w})^*$ are isomorphic
378: as $B$--modules.
379: \end{Thm}
380: The result holds in the more general setting of generalized flag
381: varieties of Kac--Moody groups. For a proof see, for example,
382: Corollary 8.1.26 in \cite{kumar}.
383:
384: Let $\l$ be an arbitrary integral weight. As $T\subset B$, the
385: Demazure module $D_{\l}$ is also a $T$--module. Its
386: $T$--character will be denoted by $\chi_\l$. Let us write
387: \begin{equation}\label{eq1}
388: \chi_\l=\sum_{\mu\in X^*(T)}m_{\l,\mu}e^\mu
389: \end{equation}
390: where we denoted by $X^*(T)$ the character group of the torus $T$.
391: The non--negative integers $m_{\l,\mu}$ appearing in the above
392: formula are the multiplicities of the weights $\mu$ in the
393: Demazure module $D_\l$. As remarked before, if $\l$ is dominant,
394: $\chi_\l$ equals $e^\l$ and $\chi_{w_\circ(\l)}$ is the character
395: of $V_\l$.
396: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
397: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
398: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
399:
400: \section{Nonsymmetric Macdonald polynomials}\label{nonsymm}
401:
402: \subsection{}
403:
404: Let us introduce a field $\C_{q,t}$ (of parameters) as follows.
405: Let $q$ and $t$ be two formal parameters and let $m$ be the lowest
406: common denominator of the rational numbers $\{(\a_j,\l_k)\ |\
407: 1\leq j,k\leq n \}$. The field $\C_{q,t}$ is defined as the field
408: of rational functions in $q^{\frac{1}{m}}$ and $t^{\frac{1}{2}}$
409: with complex coefficients. We will also use the field of rational
410: functions in $t^{\frac{1}{2}}$ denoted by $\C_t$. The algebra
411: $\cal R_{q,t}=\C_{q,t}[e^\l;\l\in P]$ is the group
412: $\C_{q,t}$-algebra of the lattice $P$. Similarly, the algebra
413: $\cal R_t=\C_t[e^\l;\l\in P]$ is the group $\C_t$-algebra of the
414: lattice $P$.
415:
416: The nonsymmetric Macdonald polynomials $E_\l(q,t)$ associated to
417: the root system $\RR$ are remarkable family of polynomials
418: indexed by the weight lattice $P$ and which form a basis for $\cal
419: R_{q,t}$. They were defined by Opdam and Macdonald for various
420: specializations of the parameters and in full generality by
421: Cherednik \cite{c}. We refer the reader to \cite{c} for a detailed
422: account of their construction and basic properties. We only
423: mention at this point that they satisfy the following
424: triangularity property with respect to the Bruhat order
425: $$
426: E_\l(q,t)\in {\rm span}\<~e^\mu~|~\mu\leq\l~\>
427: $$
428:
429: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
430:
431: \subsection{}
432: Although, a priori, the coefficients of the polynomials
433: $E_\l(q,t)$ are just rational functions in $q$ and $t$ it was
434: proved in \cite[Section 3.1]{ion2} that in fact each coefficient
435: has a finite limit as $q$ approaches infinity and, in consequence,
436: $$E_\l(t)=\lim_{q\to \infty}E_\l(q,t)$$ is well defined as an
437: element of $\cal R_t$.
438:
439: Another important fact is that the coefficients of $E_\l(t)$ are
440: polynomials in $t^{-1}$ and in consequence their limit as $t$
441: approaches infinity is finite. Moreover, the following is true.
442: Let $G$, $B$ and $T$ be a complex connected simple algebraic
443: group, a Borel subgroup and a maximal torus as in Section
444: \ref{demazure}. We can certainly regard $X^*(T)$, the character
445: group of the torus $T$, as being a sublattice of $P$. Therefore,
446: the Demazure character $\chi_\l$ associated to an integral weight
447: $\l$ becomes an element of $\cal R_t$, and the formula (\ref{eq1})
448: will be regarded as an identity in $\cal R_t$. The following
449: result was proved in \cite[Corollary 3.8]{ion2}.
450:
451: \begin{Thm}\label{limit} Let $\l$ be an arbitrary integral weight for $G$.
452: Then,
453: \begin{equation}
454: \chi_\l=\lim_{t\to\infty}E_\l(t)
455: \end{equation}
456: \end{Thm}
457:
458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459:
460: \subsection{} For special values of the parameter $t$ the
461: coefficients of the polynomials $E_\l(t)$ have a rather different
462: interpretation. To be able to state the result we first need to
463: introduce more notation.
464:
465: Given $\l$ in $P$ let us define the following normalization
466: factor
467: \begin{equation}\label{eq4}
468: j_\l=%t^{\ell(w_\l)}t^{-\frac{1}{2}(\ell(w_\l)+\ell(\w_\l))}=
469: t^{\frac{1}{2}(\ell(w_\l)-\ell(\w_\l))}
470: \end{equation}
471: where $w_\l$ and $\w_\l$ are the Weyl group elements defined in
472: Section \ref{prelim}. Let us remark that, as it follows from
473: \cite[Corollary 3.4]{ion2}, the normalization factor denoted by
474: the same symbol in \cite[(12)]{ion3} is, for equal values of the
475: parameters, precisely the element defined above.
476:
477: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
478: \subsection{}
479:
480:
481:
482: Let ${\bf G^\vee}$ be the unique split, connected, reductive group
483: scheme whose root datum is dual to that of $G$ (the Chevalley
484: group scheme). Let ${\bf B^\vee}$ be a Borel subgroup of ${\bf
485: G^\vee}$ and ${\bf T^\vee}$ a maximal split torus of $\bf G^\vee$
486: contained in ${\bf B^\vee}$. The Borel ${\bf B^\vee}$ has the Levi
487: decomposition ${\bf B=T^\vee U}$, where ${\bf U}$ is the unipotent
488: radical of ${\bf B^\vee}$. The unique Borel subgroup of $\bf
489: G^\vee$ which is opposed to $\bf B^\vee$ with respect to $\bf
490: T^\vee$ will be denoted by $\overline{\bf B}^\vee$ and
491: $\overline{\bf U}$ denotes its unipotent radical. We have
492: deliberately ignored the field of definition since all above
493: groups are defined over $\Z$ (the structure constants involve only
494: integers).
495:
496: Let $p$ be a prime number and $\mf t$ a positive integer power of
497: $p$. We will denote by $\F_{\mf t}$ the finite field with $\mf t$
498: elements. For the moment let $k$ denote an arbitrary field. Let
499: $x$ be a formal parameter and let $F:=k((x))$ be the quotient
500: field of $\cal O:=k[[x]]$ (formal power series with coefficients
501: in $k$). Of course, $F$ is a $p$--adic field, $\cal O$ is its ring
502: of integers and $k$ is the residue field.
503:
504: The $F$--rational points of ${\bf G^\vee}$ will be denoted by
505: $G^\vee(F)$ (or simply by $G^\vee$ if the field $F$ is implicitly
506: understood) and the same type of notation will be used for all the
507: linear algebraic groups defined above. Each $\l\in X^*(T)$
508: becomes an element of $X_*(T^\vee)$ (the cocharacter group of
509: $T^\vee$) and therefore determines a morphism $\l:F^\times\to
510: T^\vee$. We will denote by $x^\l$ the image of $x$ under the above
511: morphism. The groups $\bf G^\vee$, $\bf T^\vee$, $\bf U$,
512: $\overline{\bf U} $ are also defined over $\cal O$. We will denote
513: $G^\vee(\cal O)$ by $K$ and denote by $I$ the Iwahori subgroup
514: defined as the inverse image of $B^\vee(k)$ under $G^\vee(\cal
515: O)\to G^\vee(k)$. Of course, all the above considerations hold if
516: we replace $k$ by $\Z$.
517:
518:
519:
520: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
521: \subsection{}
522: In this subsection we will assume that the field $k$ is $\F_{\mf
523: t}$. Let us remark that the affine root system associated to
524: $(G^\vee,T^\vee)$ is in fact $R^\vee$ and that $K$ is a maximal
525: compact subgroup of $G^\vee$.
526:
527: The group $G^\vee$ is unimodular as opposed to the Borel subgroup
528: $B^\vee$ which is not. We choose a Haar measure on $G^\vee$
529: normalized such that the Iwahori subgroup $I$ has volume one. The
530: modular function of the Borel subgroup $\d:T^\vee\to \Re^\times_+$
531: is defined by
532: $\d(a)=|{\rm det}(Ad_{|U}(a))|$ for any $a$ in $T^\vee$;
533: we denoted by $Ad_{|U}$ the automorphism of the Lie algebra of
534: $U$ given by the adjoint representation and we denoted by
535: $|\cdot|$ the usual metric on $F$ induced by the valuation. Since
536: $G^\vee$ is split, the formula for the modular function takes the
537: following form on the elements $x^\l$, for $\l\in X^*(T)$
538: \begin{equation}\label{eq2}
539: \d{}(x^\l)={\mf t}^{2\<\l,\rho\>}
540: \end{equation}
541:
542: For an element $f=f(t)$ in $\cal R_t$ we will write $f(\mf t)$ to
543: refer to the element of $\C[e^\l;\l\in P]$ obtained by
544: substituting the positive integer $\mf t$ for the parameter $t$.
545: We are now ready to state a result connecting the coefficients of
546: nonsymmetric Macdonald polynomials with the geometry of the group
547: $G^\vee$. The Theorem stated below was proved in \cite{ion3} (see
548: Theorem 5.10 and formula (25)).
549: \begin{Thm}\label{padic}
550: Let $\l$ be an element of $X^*(T)$ (or, equivalently, of
551: $X_*(T^\vee))$. The coefficients appearing in
552: $$
553: E_\l(\mf t)=\sum_{\mu\leq \l} c_{\l,\mu} e^\mu
554: $$
555: are given by
556: $$
557: c_{\l,\mu}=\frac{vol(\overline{U} x^{-\mu}I\cap
558: Kx^{-\l}I)}{j_\l(\mf t)\mf
559: t^{\ell(w_\circ)}\d^{\frac{1}{2}}(x^{w_\circ(\mu)})}
560: $$
561: \end{Thm}
562: We would prefer to make the denominator in the above formula as
563: explicit as possible.
564: \begin{Lm}
565: Let $\l$ and $\mu$ be two elements of $X^*(T)$ for which $\mu\leq
566: \l$. Then,
567: $$
568: j_\l(\mf t)\mf
569: t^{\ell(w_\circ)}\d^{\frac{1}{2}}(x^{w_\circ(\mu)})=\mf
570: t^{n_{\l,\mu}}
571: $$
572: \end{Lm}
573: \begin{proof}
574: Straightforward from (\ref{n}), (\ref{eq4}) and (\ref{eq2}).
575: \end{proof}
576:
577: By combining Theorem \ref{limit} and Theorem \ref{padic} we can see that the
578: weight multiplicities (as defined by (\ref{eq1})) of the Demazure module
579: $D_\l$ can be computed as follows.
580: \begin{Cor}\label{cor1}
581: Let $\l$ and $\mu$ be two elements of $X^*(T)$ for which $\mu\leq
582: \l$. Then,
583: \begin{equation}\label{eq5}
584: m_{\l,\mu}=\lim_{\mf t\to\infty}\frac{vol(\overline{U}
585: x^{-\mu}I\cap Kx^{-\l}I)}{\mf t^{n_{\l,\mu}}}
586: \end{equation}
587: \end{Cor}
588: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
589: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
590: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
591: \section{The multiplicity formula}
592: \subsection{The varieties $ M_{\l,\mu}$}\label{varieties}
593:
594:
595: Let us assume for a moment that $k=\C$. It is well--known (see,
596: for example, \cite{mv} and the references therein) that the space
597: $G^\vee/K$ is an {\sl ind--variety} (i.e. admits an increasing
598: filtration with varieties such all inclusion maps among them are
599: closed embeddings) defined over $\C$. We refer the reader to
600: \cite[Chapter IV]{kumar} for a brief introduction to
601: ind--varieties. In our case, the members of the filtration on
602: $G^\vee/K$ can be constructed from a filtration of $G^\vee$
603: obtained by bounding the number of poles with respect to $x$ of
604: the matrix coefficients for a faithful representation of $G^\vee$.
605:
606: The space $ G^\vee/K$ is usually referred to as the {\sl affine
607: Grassmannian} of $G^\vee(\C)$. However, because $ G^\vee$ and $K$
608: are still defined for $k=\Z$, the same geometric constructions go
609: through in this case, and $G^\vee/K$ acquires a structure of
610: ind--variety defined over $\Z$ (see also the remarks in Section 14
611: of \cite{mv}). In a completely similar way, keeping in mind that
612: $I$ is defined for $k=\Z$, we endow $G^\vee/I$ with a structure of
613: ind--variety defined over $\Z$.
614:
615: Let us consider the map $\pi:G^\vee\to G^\vee/I$. The group $K$
616: acts on $G^\vee/I$ by finite dimensional orbits. Indeed, the
617: orbits are all of the form $\pi(Kx^\nu I)$, and the number of
618: poles of the elements in $Kx^\nu I$ is bounded by $$\max_{1\leq
619: i\leq n}|\<\nu,\a_i^\vee\>|$$ Therefore, $\pi(Kx^\nu I)$ is
620: included in one member of the filtration on $G^\vee/I$ and hence
621: it is finite dimensional. For any $\l$ and $\mu$ in $X_*(T)$, let
622: us define the variety $M_{\l,\mu}$ as $\pi(\overline{U}
623: x^{-\mu}I\cap Kx^{-\l}I)$. From the above considerations it is
624: clear that the varieties $M_{\l,\mu}$ are finite dimensional and
625: defined over $\Z$.
626:
627: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
628:
629: \subsection{Proof of Theorem \ref{thm1}}
630: We will show that the right hand side of the formula (\ref{eq5})
631: equals the number of irreducible components of top dimension of
632: $M_{\l,\mu}(\C)$. Let us recall that Corollary \ref{cor1} holds
633: under the hypothesis $k=\F_{\mf t}$. Because the Iwahori subgroup
634: $I$ has volume one we can regard the volume of the set
635: $\overline{U} x^{-\mu}I\cap Kx^{-\l}I$ as the number of right
636: $I$--cosets in $\overline{U} x^{-\mu}I\cap Kx^{-\l}I$ or,
637: equivalently, as the number of points in $M_{\l,\mu}(\F_{\mf t})$
638: (which we denote by $|M_{\l,\mu}(\F_{\mf t})|$).
639:
640: The Lefschetz fixed point formula (for the Frobenius automorphism)
641: and Theorem 1 in \cite{deligne} (for $\cal F_0=\Bbb Q_\ell$,
642: $S_0=Spec(\F_{\mf t})$ and $X_0=M_{\l,\mu}$) let us conclude that
643: $$\lim_{\mf t\to\infty}\frac{|M_{\l,\mu}(\F_{\mf t})|}{\mf
644: t^{n_{\l,\mu}}}$$ is indeed the number of irreducible components
645: of top dimension of $M_{\l,\mu}(\C)$.
646:
647: As an immediate consequence we obtain the following.
648: \begin{Cor}
649: Let $\l$ and $\mu$ be two elements of $X^*(T)$ for which $\mu\leq
650: \l$. The dimension of $M_{\l,\mu}(\C)$ is $n_{\l,\mu}$.
651: \end{Cor}
652:
653: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
654: \subsection{Final remarks}\label{remarks}
655:
656: If we keep in mind that for anti--dominant $\l$ the Demazure
657: module $D_\l$ is in fact $V_{\l_+}$, the irreducible $G$--module
658: with highest weight $\l_+=w_\circ(\l)$ (or lowest weight $\l$),
659: our result gives a geometric formula for the weight
660: multiplicities of $V_{\l_+}$. However, if we would be only
661: interested in a formula for weight multiplicities in $V_{\l}$
662: (here $\l$ is dominant), this could be obtained by closely
663: following the argument given above but replacing in Theorem
664: \ref{limit} the nonsymmetric Macdonald polynomials with symmetric
665: Macdonald polynomials $P_\l(t)$ (at $q\to\infty$) and Theorem
666: \ref{padic} with corresponding result stating the fact that
667: $P_\l(\mf t)$ arises as the Satake transform of the characteristic
668: function of the set $Kx^\l K$. If we denote by $Z_{\l,\mu}(\C)$
669: the $\C$--points of the image of $Ux^\mu K\cap Kx^\l K$ through
670: the map $$\tilde\pi:G^\vee\to G^\vee/K$$ the corresponding result
671: read as follows.
672: \begin{Thm}\label{thm2} Let $\l$ and $\mu$ be two dominant integral weights
673: such that $\mu\leq \l$. The weight multiplicity $m_{\l,\mu}$ of
674: the weight $\mu$ in the irreducible highest weight module $V_\l$
675: equals the number of top dimensional irreducible components of
676: $Z_{\l,\mu}(\C)$.
677: \end{Thm}
678:
679: The above statement is quite close to one of results in \cite{mv}
680: which we will briefly recall. For the following statements we
681: assume that $k=\C$. First, Theorem 3.2 in \cite{mv} shows that
682: $\tilde\pi(Ux^\mu K)\cap \overline{\tilde\pi(Kx^\l K)}$ (the
683: closure is in $G^\vee/K$) is pure dimensional (which, of course,
684: implies that $Z_{\l,\mu}$ is pure dimensional). Second, as a
685: consequence of the equivalence of tensor categories between the
686: category of representations of $G$ and the category of
687: $K$--equivariant perverse sheaves on the affine Grassmannian
688: $G^\vee/K$ with $\C$--coefficients, Corollary 7.4 in \cite{mv}
689: states that the weight multiplicity $m_{\l,\mu}$ equals the number
690: of irreducible components of $\tilde\pi(Ux^\mu K)\cap
691: \overline{\tilde\pi(Kx^\l K)}$. The absence of the closure in
692: Theorem \ref{thm2} slightly improves this statement.
693:
694: The main result in \cite{mv} (the above mentioned equivalence of
695: categories) was proved for reductive groups. Keeping in mind that
696: Theorem \ref{thm1} holds for Demazure modules (which are just
697: $B$--modules) and that $B$ is a solvable group, our result seems
698: to suggest that the results of Mirkovi\' c and Vilonen might
699: eventually be extended to a larger class of groups and
700: representations.
701: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
702: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
703: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
704: \begin{thebibliography}{10}
705:
706: %\bibitem{andersen}
707: %{\sc H. H. Andersen}, {Schubert varieties and Demazure's character
708: %formula}, {\sl Invent. Math.} {\bf 79} (1985), no. 3, 611--618.
709:
710: \bibitem{c}
711: {\sc I. Cherednik}, {Nonsymmetric Macdonald polynomials.} {\sl
712: Internat. Math. Res. Notices} {\bf 1995}, no.10, 483--515.
713:
714:
715: \bibitem{deligne}
716: {\sc P. Deligne}, La conjecture de Weil. II. {\sl Inst. Hautes \'
717: Etudes Sci. Publ. Math.} {\bf 52} (1980), 137--252.
718:
719: \bibitem{demazure}
720: {\sc M. Demazure}, {D\' esingularisation des vari\' et\' es de
721: Schubert g\' en\' eralis\' ees}, {\sl Ann. Sci. \' Ecole Norm.
722: Sup. (4)} {\bf 7} (1974), 53--88.
723:
724: \bibitem{ion1}
725: {\sc B. Ion}, {Nonsymmetric Macdonald polynomials and Demazure
726: characters.} {\sl Duke Math. J.} {\bf 116} (2003), no. 2,
727: 299--318.
728:
729: \bibitem{ion2}
730: {\sc B. Ion}, A Kato--Lusztig formula for nonsymmetric Macdonald
731: polynomials. Preprint, math.QA/0406060.
732:
733: \bibitem{ion3}
734: {\sc B. Ion}, Nonsymmetric Macdonald polynomials and matrix
735: coefficients for unramified principal series. Preprint,
736: math.QA/0407044.
737:
738: \bibitem{kumar}
739: {\sc S. Kumar}, Kac-Moody groups, their flag varieties and
740: representation theory. {\sl Progress in Mathematics} {\bf 204}.
741: Birkh\" auser Boston, Inc., Boston, MA, 2002.
742:
743: \bibitem{lusztig}
744: {\sc G. Lusztig}, {Singularities, character formulas, and a
745: $q$--analog of weight multiplicities. Analysis and topology on
746: singular spaces, II, III (Luminy, 1981)}, {\sl Ast\' erisque} {\bf
747: 101--102}, 208--229, Soc. Math. France, Paris, 1983.
748:
749: \bibitem{mv}
750: {\sc I. Mirkovi\' c and K. Vilonen}, Geometric Langlands duality
751: and representations of algebraic groups over commutative rings.
752: Preprint, math.RT/0401222.
753:
754: \end{thebibliography}
755:
756: \end{document}
757: