1: \documentclass{amsart}

2: \usepackage{amsmath,amssymb,amsxtra,amscd}

3: \usepackage[all]{xy}

4: %\usepackage{hyperref}

5: \usepackage{verbatim}

6: \usepackage[vcentermath,enableskew]{rcyoungtab}

7: \usepackage{pstricks}

8: \usepackage{pst-node}

9: 

10: \begin{document}

11: 

12: \newcommand{\ul}[1]{\underline{#1}}

13: 

14: % youngtab declarations

15: \Yboxdim{13pt}\Yinterspace{1pt}

16: 

17: % special symbols to be used in tableaux

18: \newcommand{\bl}{{}}

19: 

20: \newcommand{\ba}{\bar{1}}

21: \newcommand{\bb}{\bar{2}}

22: \newcommand{\bc}{\bar{3}}

23: \newcommand{\bd}{\bar{4}}

24: 

25: \newcommand{\ca}{\check{1}}

26: \newcommand{\cb}{\check{2}}

27: \newcommand{\cc}{\check{3}}

28: \newcommand{\cd}{\check{4}}

29: \newcommand{\ce}{\check{5}}

30: \newcommand{\cf}{\check{6}}

31: \newcommand{\cg}{\check{7}}

32: \newcommand{\ch}{\check{8}}

33: \newcommand{\ci}{\check{9}}

34: \newcommand{\cj}{\check{a}}

35: \newcommand{\ck}{\check{b}}

36: \newcommand{\cl}{\check{c}}

37: \newcommand{\cm}{\check{d}}

38: 

39: \newcommand{\ha}{\hat{1}}

40: \newcommand{\hb}{\hat{2}}

41: \newcommand{\hc}{\hat{3}}

42: \newcommand{\hd}{\hat{4}}

43: \newcommand{\he}{\hat{5}}

44: \newcommand{\hf}{\hat{6}}

45: \newcommand{\hg}{\hat{7}}

46: \newcommand{\hh}{\hat{8}}

47: \newcommand{\hi}{\hat{9}}

48: \newcommand{\hj}{\hat{a}}

49: \newcommand{\hk}{\hat{b}}

50: \newcommand{\hl}{\hat{c}}

51: \newcommand{\hm}{\hat{d}}

52: 

53: \newcommand{\sta}{1^*}

54: \newcommand{\stb}{2^*}

55: \newcommand{\stc}{3^*}

56: \newcommand{\std}{4^*}

57: \newcommand{\ste}{5^*}

58: \newcommand{\stf}{6^*}

59: \newcommand{\stg}{7^*}

60: \newcommand{\sth}{8^*}

61: \newcommand{\sti}{9^*}

62: \newcommand{\stj}{a^*}

63: \newcommand{\stk}{b^*}

64: \newcommand{\stl}{c^*}

65: \newcommand{\stm}{d^*}

66: 

67: \newcommand{\pmt}[2]{\begin{pmatrix} #1 \\ #2 \end{pmatrix}}%

68: \newcommand{\twocol}[1]{\multicolumn{2}{c}{#1}}

69: 

70: \newcommand{\shape}{\mathrm{shape}}

71: \newcommand{\dt}{\cdot}

72: 

73: %amsmath declarations

74: \numberwithin{equation}{section}

75: 

76: \newtheorem{theorem}{Theorem}

77: \newtheorem{prop}[theorem]{Proposition}

78: \newtheorem{conjecture}[theorem]{Conjecture}

79: \newtheorem{lemma}[theorem]{Lemma}

80: \newtheorem{cor}[theorem]{Corollary}

81: \theoremstyle{definition}

82: \newtheorem{remark}[theorem]{Remark}

83: \newtheorem{question}[theorem]{Question}

84: \newtheorem{example}[theorem]{Example}

85: \newtheorem{assumption}[theorem]{Assumption}

86: 

87: \newcommand{\anti}{\#}

88: \newcommand{\bilet}[2]{\begin{pmatrix} #2 \\ #1 \end{pmatrix}}

89: \newcommand{\CA}{\mathcal{C}'}

90: \newcommand{\CC}{\mathcal{C}}

91: \newcommand{\cck}{\check{c}}

92: \newcommand{\cht}{\widehat{c}}

93: \newcommand{\cocharge}{\overline{c}}

94: \newcommand{\coenergy}{\Db}

95: \newcommand{\Db}{\overline{D}}

96: \newcommand{\dck}{\check{d}}

97: \newcommand{\dckh}{\widehat{\dck}}

98: \newcommand{\dcm}{\check{d}_-}

99: \newcommand{\dcp}{\check{d}_+}

100: \newcommand{\eh}{\hat{e}}

101: \newcommand{\ev}{\mathrm{e}}

102: \newcommand{\fh}{\hat{f}}

103: \newcommand{\gb}{\overline{\gf}}

104: \newcommand{\gf}{\mathfrak{g}}

105: \newcommand{\Hb}{\overline{H}}

106: \newcommand{\Image}{\mathrm{Image}}

107: \newcommand{\Kb}{\overline{K}}

108: \newcommand{\KK}{f}

109: \newcommand{\Knuth}{\sim}

110: \newcommand{\La}{\Lambda}

111: \newcommand{\la}{\lambda}

112: \newcommand{\Lab}{\overline{\La}}

113: \newcommand{\lad}{\la^{\bullet}}

114: \newcommand{\LR}{\mathrm{LR}}

115: \newcommand{\Mb}{\overline{M}}

116: \newcommand{\nn}{\varnothing}

117: \newcommand{\nuh}{{\hat{\nu}}}

118: \newcommand{\Par}{\mathcal{P}}

119: \newcommand{\Pb}{\overline{P}}

120: \newcommand{\Pcl}{P_{cl}}

121: \newcommand{\Pns}{P^{\mathrm{ns}}}

122: \newcommand{\Pt}{\tilde{P}}

123: \newcommand{\Pv}{P^v}

124: \newcommand{\Rv}{R^v}

125: \newcommand{\Qt}{\tilde{Q}}

126: \newcommand{\rs}{\mathrm{rs}}

127: \newcommand{\rsh}{\hat{\rs}}

128: \newcommand{\se}{\diamondsuit}

129: \newcommand{\swch}{s}

130: \newcommand{\thet}[3]{{}^{#1}\theta_{#2}^{#3}}

131: \newcommand{\Tt}{\tilde{T}}

132: \newcommand{\Vh}{\hat{V}}

133: \newcommand{\ve}{\varepsilon}

134: \newcommand{\vn}{\varnothing}

135: \newcommand{\vp}{\varphi}

136: \newcommand{\VX}{VX}

137: \newcommand{\word}{\mathrm{word}}

138: \newcommand{\wt}{\mathrm{wt}}

139: \newcommand{\Xb}{\overline{X}}

140: \newcommand{\Z}{\mathbb{Z}}

141: \newcommand{\zv}{z^\vee}

142: \newcommand{\inner}[2]{\langle#1,#2\rangle}

143: 

144: \newcommand{\hdom}{{\begin{picture}(8,4)

145: \multiput(0,0)(4,0){3}{\line(0,1){4}}%

146: \multiput(0,0)(0,4){2}{\line(1,0){8}}\end{picture}}}

147: 

148: \newcommand{\vdom}{{\begin{picture}(4,8)

149: \multiput(0,0)(0,4){3}{\line(1,0){4}}%

150: \multiput(0,0)(4,0){2}{\line(0,1){8}}\end{picture}}}

151: 

152: \newcommand{\cell}{{\begin{picture}(4,4)

153: \multiput(0,0)(4,0){2}{\line(0,1){4}}%

154: \multiput(0,0)(0,4){2}{\line(1,0){4}}\end{picture}}}

155: 

156: \newcommand{\theset}{\{\nn,\,\cell\,,\,\hdom\,,\,\vdom\,\}}

157: \newcommand{\theneset}{\{\cell\,,\,\hdom\,,\,\vdom\,\}}

158: \newcommand{\ourset}{\{\nn,\cell\,,\hdom\,\}}

159: \newcommand{\onetwoset}{\{\cell\,,\hdom\,\}}

160: 

161: \title{On the $X=M=K$ conjecture}

162: \author{Mark Shimozono}

163: \address{Department of Mathematics \\

164: Virginia Tech \\

165: Blacksburg, VA 24061-0123 USA}

166: \thanks{Supported in part by NSF grant DMS-0401012}

167: 

168: \begin{abstract} In the large rank limit,

169: for any nonexceptional affine algebra, the graded branching

170: multiplicities known as one-dimensional sums, are conjectured to

171: have a simple relationship with those of type $A$, which are known

172: as generalized Kostka polynomials. This is called the $X=M=K$

173: conjecture. It is proved for tensor products of the ``symmetric

174: power" Kirillov-Reshetikhin modules for all nonexceptional affine

175: algebras except those whose Dynkin diagrams are isomorphic to that

176: of untwisted affine type $D$ near the zero node. Combined with

177: results of Lecouvey, this realizes the above one-dimensional sums of

178: affine type $C$, as affine Kazhdan-Lusztig polynomials (and

179: conjecturally for type $D$).

180: \end{abstract}

181: 

182: \maketitle

183: 

184: \section{Introduction}

185: 

186: \subsection{$X=M$ conjecture} \label{ss:X=M} Motivated by the

187: study of two-dimensional solvable lattice models, the seminal paper

188: \cite{KMN92} introduced the path model for the crystal graph of an

189: irreducible integrable highest weight module over a quantum affine

190: algebra $U'_q(\gf)$. A path is a semi-infinite sequence of elements

191: taken from the crystal graph of a suitable finite-dimensional

192: $U'_q(\gf)$-module. To apply the path model these suitable modules

193: and crystal graphs must be constructed. Unlike irreducible

194: integrable highest weight modules over a quantum Kac-Moody algebra,

195: it is rare for a finite-dimensional $U'_q(\gf)$-module to have a

196: crystal base. A number of suitable modules and crystal graphs had

197: been constructed in the literature

198: \cite{JMO00,KKM94,KMN92a,Ko99,Y98}.

199: 

200: Inspired by the work of Kirillov and Reshetikhin \cite{KR90} on

201: finite-dimensional representations of Yangians, the papers

202: \cite{HKOTT01,HKOTY99} conjectured the existence of a suitable

203: family of finite-dimensional $U'_q(\gf)$-modules $\{W_s^{(r)} \}$

204: called Kirillov-Reshetikhin (KR) modules. The module $W_s^{(r)}$ is

205: conjectured to be irreducible and to have a crystal base $B^{r,s}$.

206: The KR modules are indexed by pairs $(r,s)$ where $r$ is a node of

207: the Dynkin diagram of $\gb$ and $s\in \Z_{>0}$. The family of KR

208: modules is complete in the following sense: it is expected that

209: every finite-dimensional irreducible $U'_q(\gf)$-module with affine

210: crystal base, is a tensor product of KR modules \cite{Kas03}.

211: 

212: Let $W^L = \bigotimes (W_s^{(r)})^{\otimes L_s^{(r)}}$ be a finite

213: tensor product of KR modules, where $L=(L_s^{(r)})$ is a collection

214: of nonnegative integers. The tensor product $W^L$ has a

215: $U_q(\gb)$-equivariant grading by the coenergy function

216: \cite{HKOTT01,KMN92,NY97}. The \textbf{one-dimensional sum}

217: $\Xb_{L,\la}(t)$ is by definition the graded multiplicity in $W^L$

218: of the irreducible $U_q(\gb)$-module $V(\la)$ of highest weight

219: $\la$. The one-dimensional sum may be defined solely in terms of the

220: combinatorics of the KR crystals $B^{r,s}$. For affine type A, the

221: one-dimensional sums are known as \textbf{generalized Kostka

222: polynomials} \cite{NY97,ScWa99,S02} and are well-understood

223: combinatorially.

224: 

225: Based on considerations arising from the Bethe Ansatz \cite{KR90},

226: it was conjectured in \cite{HKOTT01,HKOTY99} that the

227: one-dimensional sum $\Xb_{L,\la}(t)$ is equal to a specific sum of

228: products of $t$-binomial coefficients, known as the

229: \textbf{fermionic formula} $\Mb_{L,\la}(t)$. This is the

230: \textbf{$\mathbf{X=M}$ conjecture}. The fermionic formula may be

231: expressed as a weighted sum over a set of combinatorial objects

232: called rigged configurations. For untwisted affine type $A$, $X=M$

233: was proved by exhibiting a grade-preserving bijection from classical

234: highest weight vectors in the affine crystal graph, to rigged

235: configurations \cite{KKR88,KR88,KSS02}. The $X=M$ conjecture has

236: also been proved for tensor products of the ``symmetric power" KR

237: crystal $B^{1,s}$ for nonexceptional affine algebras \cite{ScS04}

238: based partially on previous work for tensor powers of the ``vector"

239: KR crystal $B^{1,1}$ \cite{OSS03a}. $X=M$ has also been proved for

240: tensor products of the ``exterior power" KR crystals (those of the

241: form $B^{r,1}$) for the root systems $C_n^{(1)}$, $A_{2n-2}^{(2)}$,

242: $D_{n+1}^{(2)}$ \cite{OSS03} and $D_n^{(1)}$ \cite{Sci04}. In the

243: ungraded case $t=1$, $X=M$ is known in many cases (for signed

244: binomial coefficients in the $M$ formula) by combining \cite{Ch01}

245: and \cite{HKOTT01,HKOTY99}.

246: 

247: \subsection{Large rank}

248: \label{ss:limit} We show that for large rank the $X$ or $M$

249: polynomials have a surprisingly simple expression (the $K$ formula)

250: that involves affine type $A$ only.

251: 

252: Let $\{\gf_n\}$ be a family of nonexceptional affine algebras. They

253: are depicted in Figure \ref{fig:rootsys}. Let $\gb_n$ be the simple

254: Lie subalgebra of $\gf_n$ whose Dynkin diagram is obtained by

255: removing the zero node. The subscript $n$ indicates that $\gb_n$ has

256: rank $n$. Fix a partition $\la=(\la_1,\la_2,\dotsc,\la_r)$. For $n$

257: large, the partition $\la$ may be identified with the dominant

258: $\gb_n$-weight $\sum_i (\la_i-\la_{i+1}) \omega_i$, where $\omega_i$

259: is the $i$-th fundamental weight of $\gb_n$. Since the rank is

260: assumed to be large with respect to $\la$, $\la$ involves no spin

261: weights.

262: 

263: \begin{figure}

264: \begin{tabular}[t]{|c|c|c|} \hline

265: $\se$ & $X_N^{(r)}$ & \text{Dynkin diagram} \\ \hline $\nn$ &

266: \begin{minipage}{4em}

267: \begin{center}

268: $A_n^{(1)}$ \\ $(n \ge 2)$

269: \end{center}

270: \end{minipage} &

271: \raisebox{-7mm}{%

272: \scalebox{.9}{%

273: \psset{unit=1cm}%

274: \pspicture(-.25,-.5)(5.25,1.35)%

275: \rput(0,0){\circlenode{A1}{}}%

276: \rput(1,0){\circlenode{A2}{}}%

277: \rput(1.75,0){\pnode{A3}}%

278: \rput(2.5,0){$\dotsm\dotsm$}%

279: \rput(3.25,0){\pnode{A4}}%

280: \rput(4,0){\circlenode{A5}{}}%

281: \rput(5,0){\circlenode{A6}{}}%

282: \rput(2.5,.75){\circlenode{B}{}}%

283: \ncline{A1}{A2}%

284: \ncline{A2}{A3}%

285: %\ncline[linestyle=dashed]{A3}{A4}%

286: \ncline{A4}{A5}%

287: \ncline{A5}{A6}%

288: \ncline{A1}{B}%

289: \ncline{A6}{B}%

290: \rput(0,-.3){1}%

291: \rput(1,-.3){2}%

292: \rput(4,-.3){n-1}%

293: \rput(5,-.3){n}%

294: \rput(2.5,1.1){0}%

295: \endpspicture}}%

296: \\ \hline %

297: $\cell$ &

298: \begin{minipage}{4em}

299: \begin{center}

300: $D_{n+1}^{(2)}$\\$(n \ge 2)$

301: \end{center}

302: \end{minipage}&

303: \raisebox{-4.5mm}{%

304: \scalebox{.9}{%

305: \psset{unit=1cm}%

306: \pspicture(-.25,-.6)(6.25,.6)%

307: \rput(0,0){\circlenode{A0}{}}%

308: \rput(1,0){\circlenode{A1}{}}%

309: \rput(2,0){\circlenode{A2}{}}%

310: \rput(2.75,0){\pnode{A3}}%

311: \rput(3.5,0){$\dotsm\dotsm$}%

312: \rput(4.25,0){\pnode{A4}}%

313: \rput(5,0){\circlenode{A5}{}}%

314: \rput(6,0){\circlenode{A6}{}}%

315: \ncline[doubleline=true]{<-}{A0}{A1}%

316: \ncline{A1}{A2}%

317: \ncline{A2}{A3}%

318: %\ncline[linestyle=dashed]{A3}{A4}%

319: \ncline{A4}{A5}%

320: \ncline[doubleline=true]{->}{A5}{A6}%

321: \rput(0,-.3){0}%

322: \rput(1,-.3){1}%

323: \rput(2,-.3){2}%

324: \rput(5,-.3){n-1}%

325: \rput(6,-.3){n}%

326: \endpspicture}}

327: \\ \hline

328: $\cell$ &

329: \begin{minipage}{4em}

330: \begin{center}

331: $A_{2n}^{(2)}$ \\ $(n \ge 2)$

332: \end{center}

333: \end{minipage}&

334: \raisebox{-4.5mm}{%

335: \scalebox{.9}{%

336: \psset{unit=1cm}%

337: \pspicture(-.25,-.6)(6.25,.6)%

338: \rput(0,0){\circlenode{A0}{}}%

339: \rput(1,0){\circlenode{A1}{}}%

340: \rput(2,0){\circlenode{A2}{}}%

341: \rput(2.75,0){\pnode{A3}}%

342: \rput(3.5,0){$\dotsm\dotsm$}%

343: \rput(4.25,0){\pnode{A4}}%

344: \rput(5,0){\circlenode{A5}{}}%

345: \rput(6,0){\circlenode{A6}{}}%

346: \ncline[doubleline=true]{<-}{A0}{A1}%

347: \ncline{A1}{A2}%

348: \ncline{A2}{A3}%

349: %\ncline[linestyle=dashed]{A3}{A4}%

350: \ncline{A4}{A5}%

351: \ncline[doubleline=true]{<-}{A5}{A6}%

352: \rput(0,-.3){0}%

353: \rput(1,-.3){1}%

354: \rput(2,-.3){2}%

355: \rput(5,-.3){n-1}%

356: \rput(6,-.3){n}%

357: \endpspicture}}

358: \\ \hline

359: $\hdom$ &

360: \begin{minipage}{4em}

361: \begin{center}

362: $C_n^{(1)}$\\$(n \ge 2)$

363: \end{center}

364: \end{minipage}&

365: \raisebox{-4.5mm}{%

366: \scalebox{.9}{%

367: \psset{unit=1cm}%

368: \pspicture(-.25,-.6)(6.25,.6)%

369: \rput(0,0){\circlenode{A0}{}}%

370: \rput(1,0){\circlenode{A1}{}}%

371: \rput(2,0){\circlenode{A2}{}}%

372: \rput(2.75,0){\pnode{A3}}%

373: \rput(3.5,0){$\dotsm\dotsm$}%

374: \rput(4.25,0){\pnode{A4}}%

375: \rput(5,0){\circlenode{A5}{}}%

376: \rput(6,0){\circlenode{A6}{}}%

377: \ncline[doubleline=true]{->}{A0}{A1}%

378: \ncline{A1}{A2}%

379: \ncline{A2}{A3}%

380: %\ncline[linestyle=dashed]{A3}{A4}%

381: \ncline{A4}{A5}%

382: \ncline[doubleline=true]{<-}{A5}{A6}%

383: \rput(0,-.3){0}%

384: \rput(1,-.3){1}%

385: \rput(2,-.3){2}%

386: \rput(5,-.3){n-1}%

387: \rput(6,-.3){n}%

388: \endpspicture}}

389: \\ \hline

390: $\hdom$ &

391: \begin{minipage}{4em}

392: \begin{center}

393: $A_{2n}^{(2)\dagger}$\\$(n \ge 2)$

394: \end{center}

395: \end{minipage}&

396: \raisebox{-4.5mm}{%

397: \scalebox{.9}{%

398: \psset{unit=1cm}%

399: \pspicture(-.25,-.6)(6.25,.6)%

400: \rput(0,0){\circlenode{A0}{}}%

401: \rput(1,0){\circlenode{A1}{}}%

402: \rput(2,0){\circlenode{A2}{}}%

403: \rput(2.75,0){\pnode{A3}}%

404: \rput(3.5,0){$\dotsm\dotsm$}%

405: \rput(4.25,0){\pnode{A4}}%

406: \rput(5,0){\circlenode{A5}{}}%

407: \rput(6,0){\circlenode{A6}{}}%

408: \ncline[doubleline=true]{->}{A0}{A1}%

409: \ncline{A1}{A2}%

410: \ncline{A2}{A3}%

411: %\ncline[linestyle=dashed]{A3}{A4}%

412: \ncline{A4}{A5}%

413: \ncline[doubleline=true]{<-}{A5}{A6}%

414: \rput(0,-.3){0}%

415: \rput(1,-.3){1}%

416: \rput(2,-.3){2}%

417: \rput(5,-.3){n-1}%

418: \rput(6,-.3){n}%

419: \endpspicture}}

420: \\ \hline

421: %

422: $\vdom$ &

423: \begin{minipage}{4em}

424: \begin{center}

425: $B_n^{(1)}$\\$(n \ge 3)$

426: \end{center}

427: \end{minipage}&

428: \raisebox{-7mm}{%

429: \scalebox{.9}{%

430: \psset{unit=1cm}%

431: \pspicture(-.25,-.6)(6.25,1.3)%

432: \rput(0,0){\circlenode{A1}{}}%

433: \rput(1,0){\circlenode{A2}{}}%

434: \rput(1,1){\circlenode{A0}{}}%

435: \rput(2,0){\circlenode{A3}{}}%

436: \rput(2.75,0){\pnode{A4}}%

437: \rput(3.5,0){$\dotsm\dotsm$}%

438: \rput(4.25,0){\pnode{A5}}%

439: \rput(5,0){\circlenode{A6}{}}%

440: \rput(6,0){\circlenode{A7}{}}%

441: \ncline{A1}{A2}%

442: \ncline{A2}{A0}%

443: \ncline{A2}{A3}%

444: \ncline{A3}{A4}%

445: %\ncline[linestyle=dashed]{A4}{A5}%

446: \ncline{A5}{A6}%

447: \ncline[doubleline=true]{->}{A6}{A7}%

448: \rput(0,-.3){1}%

449: \rput(1,-.3){2}%

450: \rput(1.3,1){0}%

451: \rput(2,-.3){3}%

452: \rput(5,-.3){n-1}%

453: \rput(6,-.3){n}%

454: \endpspicture}}

455: \\ \hline

456: %

457: $\vdom$ &

458: \begin{minipage}{4em}

459: \begin{center}

460: $D_n^{(1)}$\\$(n \ge 4)$

461: \end{center}

462: \end{minipage}&

463: \raisebox{-7mm}{%

464: \scalebox{.9}{%

465: \psset{unit=1cm}%

466: \pspicture(-.25,-.6)(6.25,1.3)%

467: \rput(0,0){\circlenode{A1}{}}%

468: \rput(1,0){\circlenode{A2}{}}%

469: \rput(1,1){\circlenode{A0}{}}%

470: \rput(2,0){\circlenode{A3}{}}%

471: \rput(2.75,0){\pnode{A4}}%

472: \rput(3.5,0){$\dotsm\dotsm$}%

473: \rput(4.25,0){\pnode{A5}}%

474: \rput(5,0){\circlenode{A6}{}}%

475: \rput(6,0){\circlenode{A7}{}}%

476: \rput(5,1){\circlenode{A8}{}}%

477: \ncline{A1}{A2}%

478: \ncline{A2}{A0}%

479: \ncline{A2}{A3}%

480: \ncline{A3}{A4}%

481: %\ncline[linestyle=dashed]{A4}{A5}%

482: \ncline{A5}{A6}%

483: \ncline{A6}{A8}%

484: \ncline{A6}{A7}%

485: \rput(0,-.3){1}%

486: \rput(1,-.3){2}%

487: \rput(1.3,1){0}%

488: \rput(2,-.3){3}%

489: \rput(5,-.3){n-2}%

490: \rput(6,-.3){n-1}%

491: \rput(5.3,1){n}%

492: \endpspicture}}

493: \\ \hline

494: %

495: $\vdom$ &

496: \begin{minipage}{4em}

497: \begin{center}

498: $A_{2n-1}^{(2)}$\\$(n \ge 3)$

499: \end{center}

500: \end{minipage}&

501: \raisebox{-7mm}{%

502: \scalebox{.9}{%

503: \psset{unit=1cm}%

504: \pspicture(-.25,-.6)(6.25,1.3)%

505: \rput(0,0){\circlenode{A1}{}}%

506: \rput(1,0){\circlenode{A2}{}}%

507: \rput(1,1){\circlenode{A0}{}}%

508: \rput(2,0){\circlenode{A3}{}}%

509: \rput(2.75,0){\pnode{A4}}%

510: \rput(3.5,0){$\dotsm\dotsm$}%

511: \rput(4.25,0){\pnode{A5}}%

512: \rput(5,0){\circlenode{A6}{}}%

513: \rput(6,0){\circlenode{A7}{}}%

514: \ncline{A1}{A2}%

515: \ncline{A2}{A0}%

516: \ncline{A2}{A3}%

517: \ncline{A3}{A4}%

518: %\ncline[linestyle=dashed]{A4}{A5}%

519: \ncline{A5}{A6}%

520: \ncline[doubleline=true]{<-}{A6}{A7}%

521: \rput(0,-.3){1}%

522: \rput(1,-.3){2}%

523: \rput(1.3,1){0}%

524: \rput(2,-.3){3}%

525: \rput(5,-.3){n-1}%

526: \rput(6,-.3){n}%

527: \endpspicture}}

528: \\ \hline

529: \end{tabular}

530: \caption{\label{fig:rootsys} Dynkin diagrams for nonexceptional

531: affine root systems}

532: \end{figure}

533: 

534: Let $L=(L^{(r)}_s)$ be a collection of nonnegative integers, only

535: finitely many of which are nonzero, indexed by $(r,s)\in\Z_{>0}^2$.

536: For $n$ large, one may associate to $L$ the tensor product of KR

537: modules $W^L=\bigotimes_{r,s} W_s^{(r) \otimes L^{(r)}_s}$.

538: 

539: \begin{lemma} \label{lem:afflimit} \cite{SZ04} Let $\la$ and $L=(L^{(r)}_s)$

540: be as above, and $\mathcal{F}=\{\gf_n\}$ a nonexceptional family

541: of affine algebras. Then there is a limiting polynomial

542: \begin{equation}

543: \Mb^{\mathcal{F}}_{L,\la}(t) =

544:   \lim_{n\rightarrow\infty} \Mb^{\gf_n}_{L,\la}(t)

545: \end{equation}

546: called the \textbf{stable fermionic formula}. Moreover, there are

547: precisely four distinct families $M^\se_{L,\la}(t)$ of such

548: polynomials, corresponding to the partitions $\se\in

549: \{\nn,\cell\,,\hdom\,,\vdom\,\}$ of size at most two.

550: \end{lemma}

551: 

552: See Figure \ref{fig:rootsys}. Note that two affine families have

553: the same associated partition $\se$ if and only if the

554: neighborhoods of the zero nodes of their Dynkin diagrams are

555: isomorphic.

556: 

557: \begin{remark}

558: The $M$ formula is used in the statement of Lemma \ref{lem:afflimit}

559: because it is well-defined for all $L$, whereas the $X$ formula

560: depends on the conjectural existence of KR crystals. The polynomial

561: $\Mb^{\gf_n}_{L,\la}(t)$ is equal to $M(W^L,\la,t^{-1})$ in the

562: notation of \cite{HKOTT01}. In the cases that the $X$ formula is

563: defined, it is not hard to see that the $X$ formula also has a large

564: rank limit. We call this polynomial $\Xb^\se_{L,\la}(t)$ the

565: \textbf{stable one-dimensional sum}.

566: \end{remark}

567: 

568: \subsection{The ubiquity of type $A$ in large rank}

569: Originally the $K$ formula was found by experimentation involving

570: graded analogues of certain creation operators for universal

571: characters of classical type \cite{SZ04}. The purpose of this

572: subsection is to give a representation-theoretic motivation for the

573: $K$ formula. It is the philosophy of M. Kleber, that the characters

574: of KR modules $W_s^{(r)}$ with $r$ nonspin, should behave like Schur

575: functions (type $A$ characters) \cite{Kl02}. One strong piece of

576: supporting evidence is that the KR characters conjecturally satisfy

577: a system of relations called the $Q$-system

578: \cite{HKOTT01,HKOTY99,KR90}. The $Q$-system relates KR characters

579: with others whose indices $r$ are nearby in the Dynkin diagram. In

580: the large rank limit the $Q$-system only involves the part of the

581: Dynkin diagram that looks like one of type $A$.

582: 

583: The implications of Kleber's idea for the $X=M$ conjecture,

584: can be made precise by the following two conjectures.

585: 

586: Given any dominant $\gb$-weight $\la$ involving no spin nodes, let

587: $W^\la$ be Chari and Pressley's minimal affinization \cite{CP95}. It

588: is a finite-dimensional $U'_q(\gf)$-module that need not have a

589: crystal base. In type $A^{(1)}$, $W^\la$ is isomorphic to the

590: $U_q(\gb)$-irreducible $V^\la$. In any nonexceptional type, the

591: character of the minimal affinization $W^\la$ should behave like the

592: Schur function $s_\la$.

593: 

594: \begin{conjecture} \label{conj:Atensor}

595: Fix $L=(L^{(r)}_s)$. For any nonexceptional affine algebra

596: of sufficiently large rank,

597: \begin{equation*}

598:   W^L \cong \bigoplus_\tau \Xb_{L,\tau}^A(1) W^\tau

599: \end{equation*}

600: up to filtration, where $\Xb_{L,\tau}^A(t)$ is the type $A$ stable

601: one-dimensional sum.

602: \end{conjecture}

603: Due to the definition of $X$ as a tensor product multiplicity,

604: this holds for type $A$, at least on the level of characters.

605: 

606: For $\se\in \theset$, let $\Par^\se$ denote the set of partitions

607: whose Ferrers diagrams may be tiled by the shape $\se$. Explicitly,

608: $\Par^\nn$ is the singleton set consisting of the empty partition,

609: $\Par^\cell$ is the set $\Par$ of all partitions, $\Par^\hdom$ is

610: the set of partitions with even rows, and $\Par^\vdom$ the set of

611: partitions with even columns.

612: 

613: \begin{conjecture} \label{conj:restrict} \cite{CK00} Let $\se\in\theset$ and

614: $\{\gf_n\}$ a nonexceptional family of affine algebras associated

615: with $\se$ as in Figure \ref{fig:rootsys}. Then for $n$ sufficiently

616: large and partitions representing dominant weights as above, one has

617: \begin{equation*}

618:   W^\la \cong \bigoplus_\tau \left(\sum_{\mu\in \Par^\se} c^\tau_{\la\mu}\right) V^\la

619: \end{equation*}

620: as $U_q(\gb)$-modules, where $c^\tau_{\la\mu}$ is the

621: Littlewood-Richardson (LR) coefficient or stable type A tensor

622: multiplicity (see subsection \ref{ss:LR}).

623: \end{conjecture}

624: 

625: Suppose $\la$ is the rectangular partition having $r$ rows and $s$

626: columns; this corresponds to the weight $s\omega_r$. It is expected

627: that the KR module $W^{(r)}_s$ is the minimal affinization

628: $W^{s\omega_r}$. In this case the above decomposition into

629: $U_q(\gb)$-modules, agrees with that which is prescribed by

630: \cite{HKOTT01,HKOTY99} for the KR module $W^{(r)}_s$.

631: 

632: Using the definition of $X$ as a branching tensor multiplicity,

633: Conjectures \ref{conj:Atensor} and \ref{conj:restrict} imply that

634: \begin{equation*}

635:   \Xb_{L,\la}^\se(1) = \sum_\tau \Xb^A_{L,\tau}(1) \sum_{\mu\in \Par^\se} c^\tau_{\la\mu}.

636: \end{equation*}

637: The $K$ formula may be obtained by putting a grading parameter into

638: this formula.

639: 

640: \subsection{The $K$ formula}

641: Define the polynomials $\Kb^\se_{L,\la}(t)\in\Z_{\ge0}[t]$ by

642: \cite{SZ04}

643: \begin{equation} \label{eq:K} \Kb^\se_{L,\la}(t) =

644: t^{|L|-|\la|} \sum_\tau \Xb^A_{L,\tau}(t^2) \sum_{\mu\in \Par^\se}

645: c^\tau_{\la\mu}

646: \end{equation}

647: where $|\la|=\sum_i \la_i$ and $|L|=\sum_{r,s} rs L^{(r)}_s$.

648: 

649: \begin{conjecture} \label{conj:create} \cite{SZ04} ($\mathbf{X=M=K}$)

650: For all $\se\in\{\nn,\,\cell\,,\,\vdom\,,\,\hdom\,\}$,

651: \begin{equation}

652: \label{eq:affch} \Kb^\se_{L,\la}(t) = \Mb^\se_{L,\la}(t^2)

653: \end{equation}

654: \end{conjecture}

655: 

656: The conjecture was previously known for $\se=\nn$ by combining the

657: results of a number of papers; see \cite{SZ04} for an explanation.

658: Our main result is:

659: 

660: \begin{theorem} \label{th:KX}

661: $\Xb^\se_{L,\la}(t^2)=\Kb^\se_{L,\la}(t)$ for $\se\in\onetwoset$ and

662: for $L$ such that $L_s^{(r)}=0$ for $r>1$, that is, when all tensor

663: factors have the form $B^{1,s}$.

664: \end{theorem}

665: 

666: Combining this with \cite{ScS04} we have $X=M=K$ for tensor products

667: of $B^{1,s}$ for large rank in types $\se\in\ourset$.

668: 

669: \begin{remark} \label{rem:xi}

670: Let

671: \begin{equation} \label{eq:xi}

672: \xi = \begin{cases} 2 & \text{if $\se=\cell\,$} \\

673: 1 & \text{otherwise.}

674: \end{cases}

675: \end{equation}

676: Our coenergy function $\Db$ has value equal to $-\frac{1}{\xi}$

677: times the energy function $D$ used in \cite{HKOTT01}. Our $K$

678: formula in \cite{SZ04} uses energy instead of coenergy. The notation

679: $R$ in \cite{SZ04} refers to a list of rectangular partitions whose

680: multiplicities are given by $L$. The precise relationship is

681: \begin{equation} \label{eq:Kco}

682: \Kb^\se_{L,\la}(t) = t^{||L||+|L|-|\la|} K^\se_{\la,R}(t)

683: \end{equation}

684: where $||L||=\sum_{i,j\ge1} \binom{r_{ij}}{2}$ and

685: $r_{ij}=\sum_{p\ge i}\sum_{q\ge j} L_q^{(p)}$.

686: \end{remark}

687: 

688: \subsection{Combinatorial proof sketch} \label{ss:sketch}

689: By definition the $X$-formula has the form

690: \begin{equation} \label{eq:X}

691:   \Xb^\se_{L,\la}(t) = \sum_{b\in P_\se(L,\la)} t^{\Db(b)}

692: \end{equation}

693: where $P_\se(L,\la)$ is a finite set and

694: $\Db:P_\se(L,\la)\rightarrow \frac{1}{\xi}\Z_{\ge0}$ is a function.

695: They are described explicitly later.

696: 

697: Let $\LR(\tau;\la,\mu)$ be an appropriate set of cardinality

698: $c^\tau_{\la\mu}$; see section \ref{ss:LR}.

699: 

700: To prove Theorem \ref{th:KX} it suffices to exhibit a bijection

701: \begin{equation}\label{eq:bij}

702: \begin{split}

703:   \eta_L: P_\se(L,\la) &\rightarrow \bigcup_\tau \bigcup_{\mu\in \Par^\se}

704:   P_\nn(L,\tau) \times \LR(\tau;\la,\mu) \\

705: b\qquad &\mapsto \qquad(c,Z)

706: \end{split}

707: \end{equation}

708: such that

709: \begin{equation} \label{eq:bijstat}

710:   2\, \Db_\se(b) = |L| - |\la| + 2\, \Db_\nn(c).

711: \end{equation}

712: 

713: \subsection{Outline}

714: Sections \ref{sec:crystals} and \ref{sec:A} review affine crystal

715: theory and in particular that of type $A$. Section \ref{sec:VXR}

716: defines the map $\eta_L$ using the virtual crystal construction of

717: \cite{OSS03} and the combinatorial $R$-matrices. The virtual crystal

718: theory guarantees that this map respects the grading, provided it is

719: well-defined. The well-definedness and surjectivity of the map

720: require some work. Crystal embeddings called right-splitting are

721: used to reduce to the case of tensor powers of the vector KR crystal

722: $B^{1,1}$. In this case bijectivity and well-definedness are

723: established by considering in section \ref{sec:DDF} a completely

724: different realization of the bijection which we call the DDF map,

725: since its main ingredient is a bijection due to Delest, Dulucq, and

726: Favreau \cite{DDF88}. The DDF map has the advantage of being

727: obviously bijective, but it is not at all clear that it should

728: respect the grading. The proof is completed by observing inductively

729: that the VXR and DDF maps agree. In section \ref{sec:Roby} a more

730: intrinsic definition of the DDF map is given, extending the

731: interpretation of the original DDF bijection due to Roby \cite{R95}.

732: 

733: Section \ref{sec:end} contains a brief discussion about the case not

734: addressed in this article, namely, $\se=\vdom\,$.

735: 

736: While this manuscript was being completed, a proof appeared for the

737: ungraded $t=1$ case (with signed binomials) of the $X=M$ conjecture

738: \cite{Her05}. We were also informed by Cedric Lecouvey of his paper

739: \cite{Le04a}, which proves the special case of $X=K$ (Theorem

740: \ref{th:KX}) for tensor powers of $B^{1,1}$ in type $\hdom\,$. He

741: also proved in \cite{Le04} that for types $\hdom\,,\vdom\,$ and for

742: tensor products with factors of the form $B^{1,s}$, the $K$

743: polynomials are the affine Kazhdan-Lusztig polynomials of types $C$

744: and $D$ respectively, given by Lusztig's q-analogue of weight

745: multiplicity \cite{Lu83}. Combined with Theorem \ref{th:KX}, the

746: one-dimensional sums $X$ in type $\hdom$ given by tensor products of

747: crystals of the form $B^{1,s}$, are affine KL polynomials of type

748: $C$. Conjecturally the similar one-dimensional sums $X$ of type

749: $\vdom$ are affine KL polynomials of type $D$.

750: 

751: 

752: Many thanks to Cedric Lecouvey, Masato Okado, Anne Schilling, and

753: Mike Zabrocki for fruitful discussions and collaborations related to

754: this project.

755: 

756: \section{Review of affine crystal graphs}

757: \label{sec:crystals}

758: 

759: The definitions for general crystal graphs follow Kashiwara

760: \cite{Kas95}. The fundamental concepts for the study of crystal

761: graphs of finite-dimensional modules over affine algebras were

762: established in \cite{KMN92}.

763: 

764: \subsection{Affine algebras} Let $\gf\supset \gf'\supset \gb$

765: be an affine Kac-Moody algebra \cite{Kac90} with Dynkin vertex set

766: $I=\{0,1,\dotsc,n\}$, its derived subalgebra, and the simple Lie

767: subalgebra with Dynkin vertex set $J=I\backslash\{0\}$. Write

768: $U_q(\gf)\supset U'_q(\gf)\supset U_q(\gb)$ for the corresponding

769: quantized universal enveloping algebras \cite{KMN92}. For $\gf$ let

770: $\{\alpha_i\mid i\in I\}$ be the simple roots, $\{h_i\mid i\in I\}$

771: the simple coroots, $\{\Lambda_i\mid i\in I\}$ the fundamental

772: weights and $\delta$ the null root. Let

773: $P=\Z\delta\oplus\bigoplus_{i\in I} \Z\La_i$ and

774: $\Pb=\bigoplus_{i\in J} \Z\omega_i$ be the weight lattices for $\gf$

775: and $\gb$ respectively, where $\la\mapsto\bar{\la}$ denotes the

776: natural projection $P\rightarrow\Pb$ and $\{\omega_i=\bar{\La}_i\mid

777: i\in J\}$ are the fundamental weights for $\gb$.

778: 

779: \subsection{Affine crystal graphs}

780: Suppose $M$ is a finite-dimensional $U'_q(\gf)$-module with a

781: crystal base $B$. Then $B$ has the structure of a directed graph

782: with vertex set $B$ and directed edges colored by the set $I$, such

783: that:

784: \begin{enumerate}

785: \item If all edges are removed except those colored $i$, then the

786: connected components are finite directed paths called the

787: $i$-strings of $B$. Given $b\in B$, define $f_i(b)$ (resp. $e_i(b)$)

788: to be the vertex following (resp. preceding) $b$ in its $i$-string;

789: if there is no such vertex, declare the result to be the special

790: symbol $\emptyset$, which represents the zero element in the module

791: $M$. Define $\vp_i(b)$ (resp. $\ve_i(b)$) to be the number of arrows

792: from $b$ to the end (resp. beginning) of its $i$-string.

793: \item There is a weight function $\wt:B\rightarrow \Pb$ such that

794: for all $i\in I$,

795: \begin{equation*}

796: \begin{split}

797: \wt(f_i(b))&=\wt(b)-\overline{\alpha}_i \qquad\text{if $f_i(b)\not=\emptyset$}\\

798: \vp_i(b)-\ve_i(b) &= \inner{h_i}{\wt(b)}

799: \end{split}

800: \end{equation*}

801: \end{enumerate}

802: We shall call a colored directed graph $B$ with the above properties

803: a $U'_q(\gf)$-crystal.

804: 

805: A morphism $g:B\rightarrow B'$ of $U'_q(\gf)$-crystals is a map

806: $B\cup\{\emptyset\}\rightarrow B'\cup\{\emptyset\}$ such that for

807: any $b\in B$, either $g(b)\in B'$ or $g(b)=\emptyset$, and

808: $g(f_i(b))=f_i(g(b))$ and $g(e_i(b))=e_i(g(b))$ for all $b\in B$ and

809: $i\in I$ where by definition $g(\emptyset)=\emptyset$. An

810: isomorphism of crystals is a morphism of crystals which is a

811: bijection whose inverse is also a morphism of crystals.

812: 

813: The direct sum of crystals is the disjoint union.

814: 

815: \subsection{Crystal graphs for simple Lie algebras}

816: If $M$ is a finite-dimensional $U_q(\gb)$-module with crystal base

817: $B$ then its directed edges are colored by the set

818: $J=I\backslash\{0\}$, and for $i\in J$ it satisfies the properties

819: stated previously for affine crystals. In this case we call $B$ a

820: $U_q(\gb)$-crystal.

821: 

822: $b\in B$ is a \textbf{classical highest weight vector} if

823: $\ve_i(b)=0$ for all $i\in J$.

824: 

825: For $\la\in \Pb^+$ let $B(\la)$ be the crystal graph of the

826: irreducible highest weight $U_q(\gb)$-module of highest weight

827: $\la$. It has a unique classical highest weight vector and this

828: vector has weight $\la$. For $\gb$ of classical type $B(\la)$ was

829: given explicitly in \cite{KN94}.

830: 

831: Every finite-dimensional $U_q(\gb)$-module has a crystal graph which

832: is isomorphic to a direct sum of crystal graphs of the form

833: $B(\la)$.

834: 

835: A \textbf{classical component} of a $U'_q(\gf)$-crystal $B$ is a

836: connected component of the $U_q(\gb)$-crystal obtained from $B$ by

837: removing all the $0$-arrows.

838: 

839: \subsection{Tensor products}

840: \begin{remark}

841: We use the opposite of Kashiwara's tensor product convention to be

842: consistent with the $U_q(A_n)$-crystal graph combinatorics of Young

843: tableaux.

844: \end{remark}

845: 

846: If $B_1,B_2$ are crystal bases of the $U'_q(\gf)$-modules $M_1,M_2$

847: then the tensor product $M_2 \otimes M_1$ has a crystal base denoted

848: $B_2\otimes B_1$. Its vertex set is just the cartesian product $B_2

849: \times B_1$. Its edges are given by

850: \begin{equation} \label{eq:ftensor}

851: \begin{split}

852:   f_i(b_2\otimes b_1) &= \begin{cases}

853:   f_i(b_2)\otimes b_1 &\text{if $\ve_i(b_2)\ge\vp_i(b_1)$} \\

854:   b_2 \otimes f_i(b_1) &\text{otherwise.}

855: \end{cases} \\

856:   e_i(b_2\otimes b_1) &= \begin{cases}

857:   e_i(b_2)\otimes b_1 &\text{if $\ve_i(b_2)>\vp_i(b_1)$} \\

858:   b_2 \otimes e_i(b_1) &\text{otherwise,}

859: \end{cases}

860: \end{split}

861: \end{equation}

862: where the result is declared to be $\emptyset$ if either of its

863: tensor factors are. The weight function on $B_2\otimes B_1$ is

864: $\wt(b_2\otimes b_1)=\wt(b_2)+\wt(b_1)$.

865: 

866: The tensor product operation on crystals is associative.

867: 

868: \subsection{Contragredient dual} \label{ss:dual}

869: Given the crystal base $B$ of a $U'_q(\gf)$-module $M$, its

870: contragredient dual module $M^\vee$ has a crystal base $B^\vee$,

871: which has vertices written $b^\vee$ for $b\in B$, with

872: $\wt(b^\vee)=-\wt(b)$ and with all arrows reversed \cite{Kas95}:

873: \begin{equation} \label{eq:veeraise}

874:   f_i(b^\vee) = e_i(b)^\vee\qquad\text{for all $i\in I$ and $b\in

875:   B$.}

876: \end{equation}

877: There is a natural $U'_q(\gf)$-crystal isomorphism

878: \begin{equation} \label{eq:veetensor}

879: \begin{split}

880:  (B_L\otimes \dotsm\otimes B_1)^\vee &\cong B_1^\vee \otimes\dotsm\otimes

881:  B_L^\vee \\

882:  (b_L\otimes\dotsm\otimes b_1)^\vee &\mapsto b_1^\vee \otimes\dotsm\otimes

883:  b_L^\vee

884: \end{split}

885: \end{equation}

886: 

887: \subsection{Representative affine families}

888: For each partition $\se\in \{\nn,\cell\,,\hdom\,,\vdom\,\}$, we fix

889: a representative family of affine root systems. An affine algebra in

890: this family and its distinguished simple Lie subalgebra shall be

891: denoted $\gf_n^\se$ and $\gb_n^\se$ respectively. The subscript $n$

892: indicates the rank of $\gb_n^\se$. See Figure \ref{fig:crystals}.

893: \begin{figure}

894: \begin{tabular}{|c|c|c|c|}

895: \hline

896: $\se$ & $\gf_n^\se$ & $\gb_n^\se$ & $B_\se^1$ \\

897: \hline \hline

898: %

899: %A1

900: $\nn$  & $A_n^{(1)}$ & $A_n$ & \raisebox{-3mm}{%

901: \scalebox{.7}{%

902: \psset{unit=1.5cm,labelsep=.5mm}

903: \pspicture(0.75,0.7)(5.25,1.7)%

904: \rput(1,1){\rnode{A1}{\psframebox{1}}}%

905: \rput(2,1){\rnode{A2}{\psframebox{2}}}%

906: \rput(3,1){\rnode{A3}{\psframebox{3}}}%

907: \rput(4,1){\rnode{A4}{$\dotsm$}}%

908: \rput(5,1){\rnode{A5}{\psframebox{n+1}}}%

909: \ncline{->}{A1}{A2}%

910: \nbput{1}%

911: \ncline{->}{A2}{A3}%

912: \nbput{2}%

913: \ncline{->}{A3}{A4}%

914: \nbput{3}%

915: \ncline{->}{A4}{A5}%

916: \nbput{n}%

917: \nccurve[angleA=160,angleB=20]{->}{A5}{A1}%

918: \nbput{0}%

919: \endpspicture

920: }}%

921: \\ \hline

922: %

923: %D2

924: $\cell$ & $D_{n+1}^{(2)}$ & $B_n$ & \raisebox{-5mm}{%

925: \scalebox{.7}{%

926: \psset{unit=1.5cm,labelsep=.5mm}

927: \pspicture(0.75,0.7)(9.25,2)%

928: \rput(1,1){\rnode{A1}{\psframebox{1}}}%

929: \rput(2,1){\rnode{A2}{\psframebox{2}}}%

930: \rput(3,1){\rnode{A3}{$\dotsm$}}%

931: \rput(4,1){\rnode{A4}{\psframebox{n}}}%

932: \rput(5,1){\rnode{A5}{\psframebox{0}}}%

933: \rput(6,1){\rnode{A6}{\psframebox{$\overline{n}$}}}%

934: \rput(7,1){\rnode{A7}{$\dotsm$}}%

935: \rput(8,1){\rnode{A8}{\psframebox{$\overline{2}$}}}%

936: \rput(9,1){\rnode{A9}{\psframebox{$\overline{1}$}}}%

937: \rput(5,1.75){\rnode{B}{\psframebox{$\varnothing$}}}%

938: \ncline{->}{A1}{A2}%

939: \nbput{1}%

940: \ncline{->}{A2}{A3}%

941: \nbput{2}%

942: \ncline{->}{A3}{A4}%

943: \nbput{n-1}%

944: \ncline{->}{A4}{A5}%

945: \nbput{n}%

946: \ncline{->}{A5}{A6}%

947: \nbput{n}%

948: \ncline{->}{A6}{A7}%

949: \nbput{n-1}%

950: \ncline{->}{A7}{A8}%

951: \nbput{2}%

952: \ncline{->}{A8}{A9}%

953: \nbput{1}%

954: \nccurve[angleA=160,angleB=0]{->}{A9}{B}%

955: \nbput{0}%

956: \nccurve[angleA=180,angleB=20]{->}{B}{A1}%

957: \nbput{0}%

958: \endpspicture

959: }}%

960: 

961: \\ \hline

962: %C1

963: %

964: $\hdom$ & $C_n^{(1)}$ & $C_n$ & \raisebox{-4mm}{%

965: \scalebox{.7}{%

966: \psset{unit=1.5cm,labelsep=.5mm}

967: \pspicture(0.75,0.7)(8.25,1.9)%

968: \rput(1,1){\rnode{A1}{\psframebox{1}}}%

969: \rput(2,1){\rnode{A2}{\psframebox{2}}}%

970: \rput(3,1){\rnode{A3}{$\dotsm$}}%

971: \rput(4,1){\rnode{A4}{\psframebox{n}}}%

972: \rput(5,1){\rnode{A5}{\psframebox{$\overline{n}$}}}%

973: \rput(6,1){\rnode{A6}{$\dotsm$}}%

974: \rput(7,1){\rnode{A7}{\psframebox{$\overline{2}$}}}%

975: \rput(8,1){\rnode{A8}{\psframebox{$\overline{1}$}}}%

976: \ncline{->}{A1}{A2}%

977: \nbput{1}%

978: \ncline{->}{A2}{A3}%

979: \nbput{2}%

980: \ncline{->}{A3}{A4}%

981: \nbput{n-1}%

982: \ncline{->}{A4}{A5}%

983: \nbput{n}%

984: \ncline{->}{A5}{A6}%

985: \nbput{n-1}%

986: \ncline{->}{A6}{A7}%

987: \nbput{2}%

988: \ncline{->}{A7}{A8}%

989: \nbput{1}%

990: \nccurve[angleA=160,angleB=20]{->}{A8}{A1}%

991: \nbput{0}%

992: \endpspicture

993: }}%

994: \\ \hline

995: 

996: %D1

997: $\vdom$ & $D_n^{(1)}$ & $D_n$ & \raisebox{-1cm}{%

998: \scalebox{.7}{%

999: \psset{unit=1.5cm,labelsep=.5mm}

1000: \pspicture(-0.25,-1.1)(7.75,1.2)%

1001: \rput(0,0){\rnode{A1}{\psframebox{1}}}%

1002: \rput(1,0){\rnode{A2}{\psframebox{2}}}%

1003: \rput(2,0){\rnode{A3}{$\dotsm$}}%

1004: \rput(3,0){\rnode{A4}{\psframebox{n-1}}}%

1005: \rput(3.75,.5){\rnode{B1}{\psframebox{n}}}%

1006: \rput(3.75,-.5){\rnode{B2}{\psframebox{$\overline{n}$}}}%

1007: \rput(4.5,0){\rnode{A5}{\psframebox{$\overline{n\!\!-\!\!1}$}}}%

1008: \rput(5.5,0){\rnode{A6}{$\dotsm$}}%

1009: \rput(6.5,0){\rnode{A7}{\psframebox{$\overline{2}$}}}%

1010: \rput(7.5,0){\rnode{A8}{\psframebox{$\overline{1}$}}}%

1011: \ncline{->}{A1}{A2}%

1012: \nbput{1}%

1013: \ncline{->}{A2}{A3}%

1014: \nbput{2}%

1015: \ncline{->}{A3}{A4}%

1016: \nbput{n-2}%

1017: \ncline{->}{A4}{B1}%

1018: \naput{n-1}%

1019: \ncline{->}{A4}{B2}%

1020: \nbput{n}%

1021: \ncline{->}{B1}{A5}%

1022: \naput{n}%

1023: \ncline{->}{B2}{A5}%

1024: \nbput{n-1}%

1025: \ncline{->}{A5}{A6}%

1026: \nbput{n-2}%

1027: \ncline{->}{A6}{A7}%

1028: \nbput{2}%

1029: \ncline{->}{A7}{A8}%

1030: \nbput{1}%

1031: \nccurve[angleA=150,angleB=30]{->}{A8}{A2}%

1032: \nbput{0}%

1033: \nccurve[angleA=-150,angleB=-30]{->}{A7}{A1}%

1034: \naput{0}%

1035: \endpspicture

1036: }

1037: }%

1038: \\ \hline

1039: \end{tabular}\vspace{4mm}

1040: \caption{\label{fig:crystals} Representative affine families and

1041: $B_\se^1$}

1042: \end{figure}

1043: 

1044: \subsection{Categories of affine crystals}

1045: \label{ss:assume} We use two categories of affine crystal graphs:

1046: the category $\CC(A_n^{(1)})$ of tensor products of arbitrary KR

1047: crystals of type $A_n^{(1)}$ and the category $\CC^1(\gf^\se_n)$ of

1048: tensor products of KR crystals of the form $B_\se^s=B^{1,s}$ for

1049: $\gf_n^\se$.

1050: 

1051: Let $\CC=\CC(A_n^{(1)})$ or $\CC=\CC^1(\gf_n^\se)$. $\CC$ satisfies

1052: the following properties \cite{AK97,KKM94}:

1053: \begin{enumerate}

1054: \item Every $B\in \CC$ is the crystal base of an irreducible

1055: finite-dimensional $U'_q(\gf)$-module.

1056: \item $\CC$ is closed under tensor product.

1057: \item For every $B\in\CC$ there is a $\la_0\in \Pb^+$ such that there is a

1058: unique element $u(B)\in B$ of weight $\la_0$, and such that all

1059: other weights occurring in $B$ are in the convex hull of

1060: $\overline{W} \la_0$ where $\overline{W}$ is the Weyl group of

1061: $\gb$. We call $u(B)$ the \textbf{leading vector} of $B$ and $\la_0$

1062: the \textbf{leading weight} of $B$.

1063: \end{enumerate}

1064: The leading weight of $B^{r,s}$ is $s\omega_r$.

1065: 

1066: \subsection{KR crystals} \label{ss:KR}

1067: We give the KR crystals in the categories $\CC$.

1068: 

1069: \subsubsection{Vector KR crystals} \label{sss:boxKR}

1070: For $\gf=\gf_n^\se$, the KR crystal $B_\se^1$ is depicted in Figure

1071: \ref{fig:crystals}. It is a $U'_q(\gf)$-crystal with

1072: $U_q(\gb)$-decomposition given by

1073: \begin{equation}

1074:   B_\se^1 = \begin{cases}

1075:   B(\omega_1)\oplus B(0) & \text{if $\se=\cell\,$} \\

1076:   B(\omega_1) &\text{otherwise.}

1077:   \end{cases}

1078: \end{equation}

1079: Removing zero arrows from $B_\se^1$ and in the case $\se=\cell\,$

1080: removing the element $\nn\in B(0)$, one obtains the

1081: $U_q(\gb)$-crystal graph $B(\omega_1)$.

1082: 

1083: Define a partial order on $B(\omega_1)$ by $x<y$ if there is a

1084: directed path in $B(\omega_1)$ from $x$ to $y$.

1085: 

1086: The tensor product rule \eqref{eq:ftensor} gives the

1087: $U_q(\gb)$-crystal structure on tensor powers of $B(\omega_1)$.

1088: 

1089: \subsubsection{Row KR crystals}

1090: \label{sss:rowKR} More generally the KR crystal $B_\se^s$ is a

1091: $U'_q(\gf)$-crystal with $U_q(\gb)$-decomposition given by

1092: \cite{KKM94}

1093: \begin{equation} \label{eq:Bdecomp}

1094:   B^s_\se =

1095: \begin{cases}

1096: B(s\omega_1) & \text{if $\se\in\{\nn,\vdom\,\}$} \\

1097: \bigoplus_{r=0}^s B((s-r)\omega_1) &\text{if $\se=\cell\,$} \\

1098: \bigoplus_{r=0}^{\lfloor s/2\rfloor} B((s-2r)\omega_1) &\text{if

1099: $\se=\hdom\,$.}

1100: \end{cases}

1101: \end{equation}

1102: To give the $U_q(\gb)$-crystal structure on $B_\se^s$ it suffices to

1103: define such a structure for its classical components, all of which

1104: have the form $B(m\omega_1)$. As a set $B(m\omega_1)$ consists of

1105: the weakly increasing sequences of length $m$ of elements in

1106: $B(\omega_1)$, with the additional constraints that $0$ may only

1107: occur once for $\se=\cell\,$ and that $n$ and $\bar{n}$ may not both

1108: appear for $\se=\vdom\,$. The $U_q(\gb)$-crystal structure on

1109: $B(m\omega_1)$ is given by observing that the following map is a

1110: $U_q(\gb)$-crystal embedding

1111: \begin{equation} \label{eq:rowtensor}

1112: \begin{split}

1113: B(m\omega_1)&\overset{\iota_m}{\longrightarrow} B(\omega_1)^{\otimes m} \\

1114: b_m\dotsm b_1&\mapsto b_m\otimes\dotsm\otimes b_1

1115: \end{split}

1116: \end{equation}

1117: where $b_j\in B(\omega_1)$.

1118: 

1119: To specify the full $U'_q(\gf)$-crystal structure on $B_\se^s$ one

1120: must also know the $0$-arrows. This is given explicitly in

1121: \cite{KKM94,OSS03b}. Rather than recalling this here we shall define

1122: it indirectly using the virtual crystal construction

1123: \cite{OSS03,OSS03b}. This is postponed until section \ref{ss:Vrow}.

1124: 

1125: 

1126: 

1127: 

1128: \subsection{Combinatorial $R$-matrix} Given any $B_1,B_2\in \CC$

1129: with underlying $U'_q(\gf)$-modules $M_1,M_2$, the universal

1130: $R$-matrix induces a $U'_q(\gf)$-module isomorphism $M_2\otimes

1131: M_1\rightarrow M_1\otimes M_2$. By assumption the crystal graphs of

1132: these tensor products are connected. It follows that there is a

1133: unique $U'_q(\gf)$-crystal isomorphism $R=R_{B_2,B_1}:B_2\otimes B_1

1134: \rightarrow B_1\otimes B_2$ called the \textbf{combinatorial

1135: $R$-matrix}. Since $R$ preserves weights, leading vectors must

1136: correspond: $u(B_2\otimes B_1)=u(B_2)\otimes u(B_1)\mapsto

1137: u(B_1\otimes B_2)=u(B_1)\otimes u(B_2)$. In particular uniqueness

1138: implies that

1139: \begin{equation} \label{eq:R=}

1140: R_{B,B}=1_{B\otimes B}.

1141: \end{equation}

1142: For $\CC^1(\gf)$ for $\gf$ nonexceptional, the combinatorial

1143: $R$-matrices are given by \cite{HKOT00,HKOT02}. For $\CC(A_n^{(1)})$

1144: the combinatorial $R$-matrix is discussed in section \ref{ss:KRA}.

1145: 

1146: \subsection{Local coenergy} To agree with the convention in

1147: \eqref{eq:affch} and \eqref{eq:xi}, let $\xi=1$ except if, in the

1148: Dynkin diagram of $\gf$ there is a double arrow pointing toward the

1149: $0$ node, in which case $\xi=2$. The only nonexceptional affine

1150: families with $\xi=2$ are $D_{n+1}^{(2)}$ and $A_{2n}^{(2)}$; these

1151: are also precisely the affine families associated with $\cell\,$ in

1152: Figure \ref{fig:rootsys}. By \cite{KMN92}, for all $B_1,B_2\in \CC$

1153: there is a unique function $\Hb=\Hb_{B_2\otimes B_1}:B_2\otimes

1154: B_1\rightarrow \frac{1}{\xi}\Z_{\ge0}$ called the \textbf{local

1155: coenergy function}, such that

1156: \begin{enumerate}

1157: \item $\Hb$ is constant on classical components of $B_2\otimes B_1$.

1158: \item By \eqref{eq:ftensor}, for $b_2\otimes b_1\in B_2\otimes

1159: B_1$, $e_0(b_2\otimes b_1)$ equals either $e_0(b_2)\otimes b_1$ or

1160: $b_2\otimes e_0(b_1)$. The same is true for $e_0(b_1'\otimes b_2')$

1161: where $R_{B_2,B_1}(b_2\otimes b_1)=b_1'\otimes b_2'$. Say that $LL$

1162: (resp. RR) holds if $e_0$ acts on the left (resp. right) tensor

1163: factor for both $b_2\otimes b_1$ and $b_1'\otimes b_2'$. Then

1164: \begin{equation} \label{eq:H}

1165:   \Hb(e_0(b_2\otimes b_1)) = \Hb(b_2\otimes b_1)+ \dfrac{1}{\xi}

1166:   \cdot

1167: \begin{cases}

1168: 1 & \text{if $LL$} \\

1169: -1 & \text{if $RR$} \\

1170: 0 &\text{otherwise.}

1171: \end{cases}

1172: \end{equation}

1173: \item $\Hb(u(B_2)\otimes u(B_1))=0$.

1174: \end{enumerate}

1175: By definition

1176: \begin{equation} \label{eq:HR}

1177: \Hb_{B_1\otimes B_2} = \Hb_{B_2\otimes B_1} \circ R_{B_1,B_2}.

1178: \end{equation}

1179: If $B_1=B_2=B$ then by \eqref{eq:R=}, \eqref{eq:H} simplifies to

1180: \begin{equation}\label{eq:H=}

1181:   \Hb(e_0(b_2\otimes b_1))=\Hb(b_2\otimes b_1)+

1182: \dfrac{1}{\xi}

1183:   \cdot\begin{cases}

1184: 1 &\text{if $e_0(b_2\otimes b_1)=e_0(b_2)\otimes b_1$} \\

1185: -1 &\text{otherwise.}

1186: \end{cases}

1187: \end{equation}

1188: For $\CC^1(\gf)$ the local coenergy function is given explicitly in

1189: \cite{HKOT00,HKOT02} and for $\CC(A_n^{(1)})$ it is given in

1190: \cite{S02}.

1191: 

1192: \begin{example} \label{ex:Hbox}

1193: Using the partial order on $B(\omega_1)$ given in section

1194: \ref{sss:boxKR} we have

1195: \begin{equation} \label{eq:HB11}

1196: \begin{split}

1197:   \Hb_{B_\nn\otimes B_\nn}(x \otimes y) &=

1198:   \begin{cases}

1199:     0 & \text{if $x\le y$} \\

1200:     1 & \text{otherwise}

1201:   \end{cases} \\

1202:   \Hb_{B_\hdom\,\otimes B_\hdom}(x \otimes y) &=

1203:   \begin{cases}

1204:     0 & \text{if $x\le y$} \\

1205:     1 & \text{otherwise}

1206:   \end{cases} \\

1207:   \Hb_{B_\vdom\,\otimes B_\vdom}(x \otimes y) &=

1208:   \begin{cases}

1209:     2& \text{if $(x,y)=(\bar{1},1)$} \\

1210:     0& \text{if $x\le y$} \\

1211:     1& \text{otherwise}

1212:   \end{cases} \\

1213:   \Hb_{B_\cell\, \otimes B_\cell}(x\otimes y) &=

1214:   \begin{cases}

1215:      1&\text{if $x>y$ or $(x,y)=(\nn,\nn)$ or $(x,y)=(0,0)$} \\

1216:      \frac{1}{2}&\text{if exactly one of $x$ or $y$ is $\nn$} \\

1217:      0&\text{otherwise.}

1218:   \end{cases}

1219: \end{split}

1220: \end{equation}

1221: \end{example}

1222: 

1223: \subsection{Yang-Baxter equation} \label{ss:YB}

1224: This section follows \cite{KMN92}.  Given $B_1,B_2,B_3\in \CC$,

1225: there is a unique isomorphism $B_3 \otimes B_2 \otimes

1226: B_1\rightarrow B_1\otimes B_2\otimes B_3 $ induced by the algebraic

1227: universal $R$-matrix. By uniqueness the combinatorial $R$-matrices

1228: satisfy the Yang-Baxter equation

1229: \begin{equation} \label{eq:YB}

1230:   R_{B_3,B_2} \circ R_{B_3,B_1}\circ R_{B_2,B_1} =

1231:   R_{B_2,B_1} \circ R_{B_3,B_1} \circ R_{B_3,B_2}.

1232: \end{equation}

1233: Using sloppier notation, let $R_i$ denote the $R$-matrix which acts

1234: in the $i$-th and $(i+1)$-st positions from the right in a tensor

1235: product. Then we may rewrite the Yang-Baxter equation as the braid

1236: relation $R_1 R_2 R_1=R_2 R_1 R_2$.

1237: 

1238: Again by uniqueness one has

1239: \begin{equation} \label{eq:R21}

1240: \begin{split}

1241:   R_{(B_1 \otimes B_2) \otimes B_3} &= (R_{B_1\otimes B_3} \otimes

1242:   1_{B_2}) \circ (1_{B_1} \otimes R_{B_2\otimes B_3}) \\

1243:   R_{B_1 \otimes (B_2 \otimes B_3)} &=

1244:   (1_{B_2} \otimes R_{B_1\otimes B_3}) \circ

1245:   (R_{B_1 \otimes B_2} \otimes 1_{B_3})

1246: \end{split}

1247: \end{equation}

1248: 

1249: Let $H_i$ denote the value of the local coenergy function, evaluated

1250: at the $i$-th and $(i+1)$-st positions from the right in a tensor

1251: product. Then one has

1252: \begin{equation} \label{eq:HYB}

1253: \begin{split}

1254:   H_1 + H_2 R_1 &= H_1 R_2 + H_2 R_1 R_2 \\

1255:   H_2 + H_1 R_2 &= H_2 R_1 + H_1 R_2 R_1

1256: \end{split}

1257: \end{equation}

1258: 

1259: The following consequence of \eqref{eq:YB} and \eqref{eq:HYB} is

1260: well-known.

1261: 

1262: \begin{prop} \label{pp:Hbraided}

1263: Let $b\in B=B_L\otimes\dotsm\otimes B_1$ and $b'\in

1264: B'=B_M'\otimes\dotsm \otimes B_1'$.

1265: \begin{enumerate}

1266: \item $R_{B \otimes B'}(b\otimes b')$ may be computed by a composition of

1267: $R$-matrices of the form $R_{B_i \otimes B_j'}$ to shuffle the

1268: element of $B_i$ to the right past that in $B_j'$.

1269: \item $\Hb_{B \otimes B'}(b\otimes b')$ is the sum of

1270: the values $\Hb_{B_i \otimes B_j'}$ evaluated at the pairs of

1271: elements in $B_i \otimes B_j'$ that must be switched with an

1272: $R$-matrix $R_{B_i \otimes B_j'}$ in the computation of $R_{B

1273: \otimes B'}(b\otimes b')$.

1274: \end{enumerate}

1275: \end{prop}

1276: 

1277: \subsection{Intrinsic coenergy}

1278: \label{ss:D} We give a construction introduced in \cite{HKOTT01} as

1279: described in \cite{OSS03} for $\CC$ as in section \ref{ss:assume}.

1280: For $B\in \CC$, we shall define the intrinsic coenergy function

1281: $\Db_B:B\rightarrow \frac{1}{\xi}\Z_{\ge0}$. It satisfies the

1282: following rules.

1283: \begin{enumerate}

1284: \item $\Db_B$ is constant on classical components.

1285: \item $\Db_B(e_0(b))\le 1/\xi+\Db_B(b)$ for all $b\in B$.

1286: \item $\Db_B(u(B))=0$.

1287: \end{enumerate}

1288: 

1289: \begin{example} \label{ex:KRD}

1290: For KR crystals $B^{r,s}$ the intrinsic coenergy functions are

1291: prescribed in \cite{HKOTT01,HKOTY99}. For $\se\in\onetwoset$,

1292: $\gf_n^\se$ as in Figure \ref{fig:crystals}, and $B_\se^s$ with

1293: classical components as in \eqref{eq:Bdecomp}, the $r$-th tensor

1294: factor has intrinsic coenergy $r/\xi$ with $\xi$ as in

1295: \eqref{eq:xi}. In particular,

1296: \begin{equation} \label{eq:DB11}

1297:   \Db_{B_\se^1}(x) = \begin{cases}

1298:    \frac{1}{2} &\text{if $\se=\cell\,$ and $x=\nn$} \\

1299:    0 &\text{otherwise.}

1300:   \end{cases}

1301: \end{equation}

1302: \end{example}

1303: 

1304: Suppose the intrinsic coenergy functions $D_1$ and $D_2$ of

1305: $B_1,B_2\in\CC$ have already been defined. Define the intrinsic

1306: coenergy function for $B_2\otimes B_1$ as follows. Let $\Db^1_{B_i}$

1307: denote the function $\Db_{B_i}$ applied to the rightmost tensor

1308: factor. Then

1309: \begin{equation} \label{eq:D2}

1310:   \Db_{B_2\otimes B_1} =

1311:   \Hb_{B_2\otimes B_1} +

1312:   \Db_{B_1}^1 +

1313:   \Db_{B_2}^1 \circ R_{B_2,B_1}.

1314: \end{equation}

1315: 

1316: \begin{prop} \label{pp:Dtensorcat} \cite{OSS03} The above

1317: construction of $\Db$ is associative.

1318: \end{prop}

1319: 

1320: Using this construction we may define $\Db_B$ for all $B\in \CC$. By

1321: Proposition \ref{pp:Dtensorcat} the function $\Db_B$ doesn't depend

1322: on the way that $B$ is built up from two-fold tensor products.

1323: 

1324: For $B_1,\dotsc,B_L\in\CC$, the above construction yields the

1325: formula  \cite{HKOTT01,NY97,OSS03}

1326: \begin{equation} \label{eq:Dn}

1327:   \Db_{B_L\otimes\dotsm \otimes B_1} =

1328: \sum_{1\le i<j\le L} \Hb_i R_{i+1} R_{i+2} \dotsm R_{j-1} +

1329: \sum_{j=1}^L \Db_{B_j}^1 R_1 R_2 \dotsm R_{j-1}.

1330: \end{equation}

1331: If $B_j=B$ for all $j$ then by \eqref{eq:R=} this reduces to

1332: \begin{equation} \label{eq:Dn=}

1333:   \Db_{B^{\otimes L}} =

1334: \sum_{i=1}^{L-1} (L-i)\, \Hb_i + L \,\Db_B^1.

1335: \end{equation}

1336: 

1337: \begin{prop} \label{pp:DR} \cite{OSS03} Suppose

1338: $B$ and $B'$ are any two tensor products with the same collection of

1339: tensor factors $B_1,\dotsc,B_L\in\CC$ in some order. Let

1340: $R:B\rightarrow B'$ be the unique affine crystal isomorphism. Then

1341: \begin{equation} \label{eq:DR}

1342:   \Db_B = \Db_{B'}\circ R

1343: \end{equation}

1344: \end{prop}

1345: 

1346: \subsection{The $X$ formula}

1347: Let $B\in\CC^1(\gf_n^\se)$ and $\la\in\Pb^+$. Let $P(B)$ be the set

1348: of classical highest weight vectors in $B$ and $P(B,\la)$ those of

1349: highest weight $\la$. For type $\se$ we write $P_\se(B,\la)$. If

1350: $B=B^{1\otimes L}$ we write $P_\se(1^L,\la)$.

1351: 

1352: \begin{example} \label{ex:standardpath}

1353: Let $B=B_\se^{1\otimes L}$. Let $\gf$ be of type

1354: $A_{n-1}^{(1)},D_{n+1}^{(2)},C_n^{(1)},D_n^{(1)}$ so that $\gb$ has

1355: type $A_{n-1},B_n,C_n,D_n$ for types $\se=\theset$ respectively.

1356: 

1357: We use the identification of weights and elements of $\Z^n$ given in

1358: section \ref{ss:limit}. Let $m_x(b)$ be the number of times the

1359: symbol $x$ occurs in the word $b$. Define the weight function

1360: $\wt:B_\se^{1\otimes L}\rightarrow\Z^n$ by

1361: \begin{equation}

1362: \begin{split}

1363:   \wt(b) &=

1364:   (m_1(b)-m_{\bar{1}}(b),m_2(b)-m_{\bar{2}}(b),\dotsc,m_n(b)-m_{\bar{n}}(b))

1365: \end{split}

1366: \end{equation}

1367: For rank large with respect to $L$ and $\la$ it is easy to check

1368: that $b\in P_\se(1^L,\la)$ if and only if $\la^{(i)}=\wt(b_i\dotsm

1369: b_1)$ is a partition for all $i$, with $\la^{(L)}=\la$. In type $A$

1370: such classical highest weight vectors are called \textbf{Yamanouchi

1371: words}.

1372: \end{example}

1373: 

1374: The \textbf{one-dimensional sum} is defined by

1375: \begin{equation*}

1376:   \Xb(B,\la)(t) = \sum_{b\in P(B,\la)} t^{\Db_B(b)}.

1377: \end{equation*}

1378: 

1379: \begin{remark} \label{rem:Border}

1380: By Proposition \ref{pp:DR} this depends only on the collection of

1381: tensor factors in $B$ and not on their order. Let $L$ be the

1382: multiplicity array for $B$, that is, $B$ contains $L_i^{(a)}$ tensor

1383: factors equal to $B^{a,i}$. We write $\Xb(B,\la)(t)=\Xb_{L,\la}(t)$.

1384: \end{remark}

1385: 

1386: \begin{example} \label{ex:X} We give examples of summands in

1387: one-dimensional sums. Tensor symbols are dropped and subscripts are

1388: added for convenience. In these examples $\Db$ is calculated using

1389: \eqref{eq:Dn=}, \eqref{eq:HB11}, and \eqref{eq:DB11}.

1390: \begin{equation*}

1391: c=1_13_22_31_42_51_61_7\in P_\nn(1^7,(4,2,1)) \qquad

1392: b=\bar{2}_11_2\nn_3\bar{1}_42_51_61_7 \in P_\cell\,(1^7,(2))

1393: \end{equation*}

1394: \begin{equation*}

1395: \begin{split}

1396:  \Db_\nn(c) &= 1\cdot0+2\cdot1+3\cdot1+4\cdot0+5\cdot1+6\cdot0+7\cdot 0  = 10 \\

1397:  \Db_\cell\,(b) &= 1\cdot1+2\cdot\frac{1}{2}+3\cdot\frac{1}{2}+4\cdot1+5\cdot1+6\cdot0+7 \cdot 0=25/2.

1398: \end{split}

1399: \end{equation*}

1400: \end{example}

1401: 

1402: \section{Type A crystal graphs}

1403: \label{sec:A}

1404: 

1405: Let $N=2n$ for the remainder of the paper. In this section

1406: $\gf=A_{N-1}^{(1)}$ and $\gb=A_{N-1}$. We specialize to this case

1407: because of the virtual crystal construction in section

1408: \ref{sec:VXR}.

1409: 

1410: \subsection{$U_q(A_{N-1})$-crystals}

1411: \label{ss:Aclassical} The $U_q(A_{N-1})$-crystal structure on

1412: $B(\omega_1)^{\otimes m}$ is given in section \ref{sss:boxKR}; it is

1413: the set of words of length $m$ in the totally-ordered set

1414: $\{1<2<\dotsm<N\}$.

1415: 

1416: The crystal graph $B(\la)$ of the highest weight

1417: $U_q(A_{N-1})$-module of highest weight $\la$, is given by the set

1418: of (semistandard) Young tableaux of shape $\la$ with entries in the

1419: set $B(\omega_1)=\{1,2,\dotsc,N\}$ \cite{KN94}; see \cite{Ful97} for

1420: the definition of a tableau and its row word.

1421: 

1422: Let $|\la|=m$ be the number of cells in the diagram of the partition

1423: $\la$. The $U_q(A_{N-1})$-crystal structure on $B(\la)$ is given by

1424: declaring that the map $B(\la)\rightarrow B(\omega_1)^{\otimes m}$

1425: that sends a tableau to its row word, is an embedding of

1426: $U_q(A_{N-1})$-crystals.

1427: 

1428: Let $u\in B(\omega_1)^{\otimes m}$. There is a unique dominant

1429: $A_{N-1}$-weight $\la$ such that the component of $u$ in

1430: $B(\omega_1)^{\otimes m}$ is isomorphic to $B(\la)$. Moreover the

1431: isomorphism is unique. Denote by $P(u)\in B(\la)$ the image of $u$

1432: under this isomorphism. It is well-known that $P(u)$ is Schensted's

1433: $P$-tableau \cite{Sc61}.

1434: 

1435: \subsection{Knuth equivalence}

1436: We shall identify a tableau with its row word. Knuth defined an

1437: equivalence relation on words with letters in the totally-ordered

1438: finite set $A$, which is generated by relations of the form

1439: \begin{align*}

1440: uxzyv&\Knuth uzxyv && \text{for $x<y\le z$} \\

1441: uyxzv&\Knuth uyzxv && \text{for $x\le y<z$}

1442: \end{align*}

1443: where $u$ and $v$ are words with letters in $A$ and $x,y,z\in A$

1444: \cite{Kn70}.

1445: 

1446: One may show that for each word $u$, there is a unique tableau of

1447: partition shape that is Knuth-equivalent to $u$; we denote it by

1448: $[u]$.

1449: 

1450: The following results are well-known.

1451: 

1452: \begin{prop} \label{pp:KnuthA} For any word $u$, $P(u)=[u]$.

1453: \end{prop}

1454: 

1455: \begin{prop} \label{pp:Knuthres} Suppose $u\Knuth v$ where $u$ and $v$

1456: are words with letters in a totally ordered set $A$ and $B$ is a

1457: subinterval of $A$. Then $u|_B\Knuth v|_B$ where $u|_B$ is obtained

1458: from $u$ by erasing all letters not in $B$.

1459: \end{prop}

1460: 

1461: \subsection{Littlewood-Richardson Rule}

1462: \label{ss:LR} The Littlewood-Richardson (LR) rule \cite{LR34} is a

1463: combinatorial description of the structure constants of the ring of

1464: symmetric functions with respect the basis of Schur functions. See

1465: \cite{FG93} for an exposition of various combinatorial viewpoints on

1466: this rule. It is well-known that $c^\alpha_{\beta\gamma}$ is the

1467: multiplicity of the irreducible $U_q(A_{N-1})$-crystal $B(\alpha)$

1468: in the tensor product $B(\beta)\otimes B(\gamma)$ for $N$ large with

1469: respect to the partitions $\alpha,\beta,\gamma$.

1470: 

1471: Let $y$ be a word of length $m$ in the alphabet $\{1,2,\dotsc,N\}$,

1472: that is, $y$ is in the $U_q(A_{N-1})$-crystal $B(\omega_1)^{\otimes

1473: m}$. The word $y$ is \textbf{Yamanouchi} if it is a highest weight

1474: vector. This is equivalent to saying that the weight of each right

1475: factor of $y$ is a partition (and in particular $\wt(y)$ is a

1476: partition). Say that a word $w$ is $\alpha/\beta$-Yamanouchi if for

1477: any Yamanouchi word $y$ of weight $\beta$, $wy$ is Yamanouchi of

1478: weight $\alpha$.

1479: 

1480: \begin{theorem} \label{th:LR} The Littlewood-Richardson

1481: coefficient $c^\tau_{\la\mu}$ is the cardinality of any of the

1482: following sets:

1483: \begin{enumerate}

1484: \item The tableaux of skew shape $\tau/\la$ that are Yamanouchi of

1485: weight $\mu$ \cite{LR34}.

1486: \item The tableaux of shape $\mu$ that are $\tau/\la$-Yamanouchi.

1487: \item Given a fixed tableau $P$ of shape $\tau$,

1488: the set of pairs of tableaux $(T,S)$ of respective shapes $\la$ and

1489: $\mu$, such that $P=[TS]$ \cite{LS81}.

1490: \end{enumerate}

1491: \end{theorem}

1492: 

1493: \subsection{Type A KR crystals}

1494: \label{ss:KRA} The KR $U'_q(A_{N-1}^{(1)})$-crystal $B^{r,s}$ is

1495: isomorphic as a $U_q(A_{N-1})$-module to $B(s\omega_r)$

1496: \cite{KMN92a}. By section \ref{ss:Aclassical} $B^{r,s}$ consists of

1497: the semistandard tableaux with entries in $\{1,2,\dotsc,N\}$ whose 0

1498: shape is the $r\times s$ rectangle. The $0$-arrows for $B^{r,s}$ are

1499: given in \cite{KMN92a,S02}.

1500: 

1501: Since $B^{r,s} \otimes B^{r',s'}$ is multiplicity-free as a

1502: $U_q(A_{N-1})$-crystal, the above combinatorial $R$-matrix

1503: $R_{B^{r,s},B^{r',s'}}$ may be characterized as the unique

1504: $U_q(A_{N-1})$-crystal isomorphism $B^{r,s}\otimes

1505: B^{r',s'}\rightarrow B^{r',s'}\otimes B^{r,s}$ and can therefore be

1506: computed using the Robinson-Schensted-Knuth correspondence

1507: \cite{S02}.

1508: 

1509: We recall from \cite{S02} that for $b_1\otimes b_2\in B^{r,s}

1510: \otimes B^{r',s'}$, the value of the local coenergy function

1511: $\Hb(b_1 \otimes b_2)$ is the number of cells in the shape of the

1512: tableau $[b_1 b_2]$ that are strictly to the right of the $k$-th

1513: column where $k=\max(s,s')$.

1514: 

1515: Since $B^{r,s}\in\CC(A_{N-1}^{(1)})$ is irreducible as a

1516: $U_q(A_{N-1})$-crystal, we have:

1517: \begin{equation} \label{eq:AD}

1518:   \Db_{B^{r,s}}=0 \qquad\text{in type $A_{N-1}^{(1)}$.}

1519: \end{equation}

1520: 

1521: \subsection{Dual KR crystals of type $A$}

1522: It follows from \cite{S02} that there is a

1523: $U'_q(A_{N-1}^{(1)})$-crystal isomorphism

1524: \begin{equation} \label{eq:AKRdual}

1525: B^{r,s\vee}\cong B^{N-r,s}.

1526: \end{equation}

1527: Let $B^1=B^{1,1}\in\CC(A_{N-1}^{(1)})$. Its crystal graph is

1528: pictured in Figure \ref{fig:crystals} except that $n$ should be

1529: replaced by $N-1$. As ordered sets with respect to the partial order

1530: given in section \ref{sss:boxKR} we have

1531: \begin{align}

1532: B^1&=\{1<2<\dotsm<N\} \\

1533: \label{eq:Bvorder} B^{1\vee}&=\{N^\vee <\dotsm < 2^\vee < 1^\vee \}

1534: \end{align}

1535: As in section \ref{sss:rowKR} $B^{s\vee}$ is the set of weakly

1536: increasing words of length $s$ with letters in $B^{1\vee}$, and its

1537: $U_q(A_{N-1})$-crystal structure is determined from that of

1538: $B^{1\vee}$ using a version of the embedding $\iota$ of

1539: \eqref{eq:rowtensor}. As a special case of \eqref{eq:AD},

1540: \begin{equation} \label{eq:Ddual}

1541:   \Db_{B^{s\vee}}=0

1542: \end{equation}

1543: 

1544: \begin{prop} \label{pp:Rvee} The category $\CC=\CC(A_{N-1}^{(1)})$

1545: is closed under $\vee$. Moreover, if $B_1,B_2\in

1546: \CC=\CC(A_{N-1}^{(1)})$, then the following diagram commutes:

1547: \begin{equation}

1548: \begin{CD}

1549:   B_2 \otimes B_1 @>{R}>> B_1 \otimes B_2 \\

1550:   @V{\vee}VV @VV{\vee}V \\

1551:   B_1^\vee \otimes B_2^\vee @>>{R}> B_2^\vee \otimes B_1^\vee

1552: \end{CD}

1553: \end{equation}

1554: where $\vee:B_2\otimes B_1\rightarrow B_1^\vee \otimes B_2^\vee$

1555: sends $b_2\otimes b_1$ to $b_1^\vee \otimes b_2^\vee$.

1556: \end{prop}

1557: \begin{proof} $\CC$ is closed under duals by \eqref{eq:AKRdual}.

1558: The composition $b_1^\vee\otimes b_2^\vee \mapsto b_2\otimes b_1

1559: \mapsto R(b_2\otimes b_1):=b_1'\otimes b_2' \mapsto b_2^{\prime\vee}

1560: \otimes b_1^{\prime\vee}$ is a $U'_q(A_{N-1}^{(1)})$-crystal

1561: isomorphism $B_1^\vee \otimes B_2^\vee\rightarrow B_2^\vee \otimes

1562: B_1^\vee$ by \eqref{eq:veeraise}. The commutation follows by the

1563: uniqueness of the $R$-matrix.

1564: \end{proof}

1565: 

1566: Using section \ref{ss:KRA} we calculate some combinatorial

1567: $R$-matrices and local coenergy functions explicitly.

1568: 

1569: \begin{prop} \label{pp:RBBv}

1570: \begin{align}

1571: \label{eq:RBBv}

1572:   R_{B^1,B^{1\vee}}(i \otimes j^\vee) &=

1573:   \begin{cases}

1574:    j^\vee \otimes i & \text{if $i\not=j$} \\

1575:     (i+1)^\vee \otimes (i+1) & \text{if $i=j<N$} \\

1576:     1^\vee \otimes 1 & \text{if $i=j=N$.}

1577:   \end{cases} \\

1578: \label{eq:RBvB}

1579:   R_{B^{1\vee},B^1}(j^\vee \otimes i) &=

1580:   \begin{cases}

1581:    i \otimes j^\vee & \text{if $i\not=j$} \\

1582:     (i-1)^\vee \otimes (i-1)^\vee & \text{if $i=j>1$} \\

1583:     N\otimes N^\vee & \text{if $i=j=1$.}

1584:   \end{cases} \\

1585: \label{eq:HBBv}

1586:   H_{B^1 \otimes B^{1\vee}}(i \otimes j^\vee) &=

1587:   \begin{cases}

1588:   1 & \text{if $i=j=2n$} \\

1589:   0 &\text{otherwise}

1590:   \end{cases} \\

1591: \label{eq:HBvB}

1592:   H_{B^{1\vee} \otimes B^1}(j^\vee \otimes i) &=

1593:   \begin{cases}

1594:   1 & \text{if $i=j=1$} \\

1595:   0 &\text{otherwise.}

1596:   \end{cases}

1597: \end{align}

1598: \end{prop}

1599: 

1600: By Proposition \ref{pp:Hbraided}, the $R$-matrix

1601: $R_{q,p^\vee}:=R_{B^{1\otimes q},B^{1\vee\otimes p}}$ may be

1602: computed using $R_{1,1^\vee}=R_{B^1,B^{1\vee}}$ repeatedly.

1603: 

1604: \begin{prop} \label{pp:rowRBBv}

1605: Let $b_2\otimes b_1\in B^t \otimes B^{s\vee}\in\CC(A_{N-1}^{(1)})$,

1606: $m=\min(m_N(b_2),m_{N^\vee}(b_1))$, $b_2=\hat{b_2}N^m$, $b_1 =

1607: N^{\vee m} \hat{b_1}$, and $\hat{b_1}'\otimes \hat{b_2}' =

1608: R_{t-m,(s-m)^\vee}(\hat{b_2}\otimes \hat{b_1})$. Then $R_{B^t

1609: \otimes B^{s\vee}}(b_2\otimes b_1) = \hat{b_1}'1^{\vee m} \otimes

1610: 1^m \hat{b_2}'$.

1611: \end{prop}

1612: 

1613: \begin{prop} \label{pp:Rhalf} Let $b_2\otimes b_1\in B^t \otimes

1614: B^{s\vee}$ and $b_1'\otimes b_2' = R_{B^t,B^{s\vee}}(b_2\otimes

1615: b_1)$. Then each $i^\vee \in b_1$ either is unchanged in passing to

1616: $b_1'$ or it changes into $(i+1)^\vee$, including the case of

1617: $N^\vee$ staying the same or changing to $1^\vee$. Similarly each

1618: $i\in b_2$ either remains an $i$ in $b_2'$ or changes to an $(i+1)$

1619: in $b_2'$.

1620: \end{prop}

1621: \begin{proof} This follows from the description of

1622: $R_{B^t,B^{s\vee}}$, \eqref{eq:RBBv}, and the fact that $B^t$ and

1623: $B^{s\vee}$ consist of weakly increasing words.

1624: \end{proof}

1625: 

1626: \begin{prop} \label{pp:Knuthdual} Let $b,c\in B^{1\vee\otimes m}$

1627: with the total ordering on $B^{1\vee}$ given by \eqref{eq:Bvorder}.

1628: Then $b\Knuth c$ if and only if $b^\vee \Knuth c^\vee$.

1629: \end{prop}

1630: 

1631: \subsection{Dynkin reversal} \label{ss:Dynkinrev}

1632: Let $N=2n$, $\gf=A_{N-1}^{(1)}$ and $\gb=A_{N-1}$. There is an

1633: involution on the $U_q(A_{N-1})$-crystal $B(\omega_1)$ given by

1634: $j\mapsto j^*:=N+1-j$. It satisfies

1635: \begin{equation} \label{eq:fstar}

1636:   f_i(b^*) = e_{N-i}(b)^*\qquad\text{for all $1\le i\le N-1$.}

1637: \end{equation}

1638: Given any $U_q(A_{N-1})$-crystals $B_1,\dotsc,B_m$ which all have

1639: involutions $*:B_i\rightarrow B_i$ satisfying \eqref{eq:fstar}, one

1640: may define the map

1641: \begin{equation} \label{eq:startensor}

1642: \begin{split}

1643: B_m\otimes\dotsm\otimes B_1 &\overset{*}{\rightarrow}

1644: B_1\otimes\dotsm\otimes B_m \\

1645: (b_m\otimes\dotsm \otimes b_1)^* &= b_1^*\otimes\dotsm\otimes b_m^*

1646: \end{split}

1647: \end{equation}

1648: which also satisfies \eqref{eq:fstar}. Taking $B_i=B(\omega_1)$ for

1649: all $i$, one obtains an involution $*$ on $B(\omega_1)^{\otimes m}$.

1650: This restricts to an involution $*$ on $B(m\omega_1)$ (and hence on

1651: $B^m$) via the embedding $\iota_m$ in \eqref{eq:rowtensor}.

1652: 

1653: Similarly one may define $*$ on $B^{1\vee}$ by

1654: $j^{\vee*}=(N+1-j)^\vee$ and extend the definition of $*$ to

1655: $B^{s\vee}$ as before.

1656: 

1657: Finally, using \eqref{eq:startensor}, $*$ may be defined on a tensor

1658: product of KR $U_q(A_{N-1}^{(1)})$-crystals of the form $B^t$ and

1659: $B^{s\vee}$; it reverses the order of the tensor factors.

1660: 

1661: \begin{prop} \label{pp:R*} Let $B_1,B_2\in \CC=\CC(A_{N-1}^{(1)})$.

1662: Then the diagram commutes:

1663: \begin{equation} \label{eq:R*}

1664: \begin{CD}

1665:   B_2 \otimes B_1 @>{R}>> B_1 \otimes B_2 \\

1666:   @V{*}VV @VV{*}V \\

1667:   B_1 \otimes B_2 @>>{R}> B_2\otimes B_1

1668: \end{CD}

1669: \end{equation}

1670: \end{prop}

1671: \begin{proof} The proof is similar to that of Proposition

1672: \ref{pp:Rvee}.

1673: \end{proof}

1674: 

1675: \section{VXR map}

1676: \label{sec:VXR}

1677: 

1678: The virtual crystal method of \cite{OSS03} realizes

1679: $U'_q(\gf_n^\se)$-crystals for $\se\in\onetwoset$ by embedding them

1680: into $U'_q(A_{2n-1}^{(1)})$-crystals. The data for the desired map

1681: \eqref{eq:bij} may be obtained using the virtual crystal embedding

1682: followed by an $R$-matrix. The name VXR stands for ``Virtual X and

1683: R-matrix".

1684: 

1685: 

1686: \subsection{Virtual crystals for $B^s_\se$} \label{ss:Vrow}

1687: The KR crystals $B_\se^s\in\CC^1(\gf_n^\se)$ were defined in section

1688: \ref{ss:KR}. To compute the $X$ formula explicitly for

1689: $B\in\CC^1(\gf_n^\se)$ we need the $U'_q(\gf)$-crystal structure on

1690: $B_\se^s$ \cite{KKM94}, the intrinsic coenergy for $B_\se^s$

1691: \cite{HKOTT01,HKOTY99}, and the combinatorial $R$-matrix for

1692: $B_\se^s\otimes B_\se^t$ \cite{HKOT00,HKOT02,NY97}.

1693: 

1694: Instead of recounting all of these things, we use the method of

1695: virtual crystals \cite{OSS03} to reduce to computations in type $A$.

1696: The virtual X (VX) formula is a formula for the one-dimensional sums

1697: for an arbitrary affine algebra $\gf_X$ expressed in terms of a

1698: simply-laced affine algebra $\gf_Y$. They are known to be valid for

1699: $B\in \CC^1(\gf)$ where $\gf$ is a nonexceptional affine algebra

1700: \cite{OSS03b}. We require the case that $\gf=\gf_n^\se$ for

1701: $\se\in\onetwoset$; see Figure \ref{fig:crystals}. In both cases the

1702: VX formula is stated using the simply-laced affine root system

1703: $A_{2n-1}^{(1)}=A_{N-1}^{(1)}$.

1704: 

1705: With $\xi$ as in \eqref{eq:xi}, there is an isometric (up to a

1706: scalar multiple) embedding $\Psi$ of the weight lattice of

1707: $\gf_n^\se$ into that of $A_{N-1}^{(1)}$ defined by

1708: \begin{equation} \label{eq:embwt}

1709: \begin{split}

1710: \Psi(\La_i)&=\begin{cases}

1711:     \La_i^A + \La_{N-i}^A & \text{if $i\in\{1,\dotsc,n-1\}$} \\

1712:     \frac{2}{\xi} \La_i^A & \text{if $i\in\{0,n\}$.}

1713: \end{cases} \\

1714: \Psi(\delta) &= \dfrac{2}{\xi} \,\delta^A

1715: \end{split}

1716: \end{equation}

1717: where $\{\La_i\mid 0\le i\le n\}$ and $\delta$ are the fundamental

1718: weights and null root of $\gf_n^\se$ and $\{\La_i^A\mid 0\le i\le

1719: N-1 \}$ and $\delta^A$ are those of $A_{N-1}^{(1)}$. From this it

1720: follows that

1721: \begin{equation}\label{eq:embroot}

1722: \Psi(\alpha_i)=\begin{cases}

1723:     \alpha_i^A + \alpha_{N-i}^A & \text{if $i\in\{1,\dotsc,n-1\}$} \\

1724:     \frac{2}{\xi} \alpha_i^A & \text{if $i\in\{0,n\}$.}

1725: \end{cases}

1726: \end{equation}

1727: There is an induced map between the weight lattices of $\gb_n^\se$

1728: and $A_{N-1}$.

1729: 

1730: Define the virtual lowering operators to be the composite operators

1731: on a crystal of type $U'_q(A_{N-1}^{(1)})$ defined by

1732: \begin{equation*}

1733: \fh_i = \begin{cases}

1734:   f_i f_{N-i} &\text{if $i\in\{1,2,\dotsc,n-1\}$} \\

1735:    f_i^{\frac{2}{\xi}} & \text{if $i\in\{0,n\}$.}

1736: \end{cases}

1737: \end{equation*}

1738: Define the virtual raising operators $\eh_i$ with $e$ replacing $f$

1739: in the above definition.

1740: 

1741: For our purposes, an \textbf{aligned virtual crystal} of type $\se$

1742: is a pair $(V,\Vh)$ where $\Vh$ is a tensor product of KR crystals

1743: for $U'_q(A_{2n-1}^{(1)})$, $V\subset \Vh$ is a subset containing

1744: $u(\Vh)$ that is closed under the virtual operators $\fh_i$ and

1745: $\eh_i$ for $i\in \{0,\dotsc,n\}$, such that there is a tensor

1746: product $B$ of KR crystals of type $\se$ and a bijection

1747: $\Psi:B\rightarrow V$ such that

1748: \begin{enumerate}

1749: \item $\Psi(u(B))=\Psi(u(\Vh))$.

1750: \item For all $b\in B$, $\Psi(f_i b)=\fh_i \Psi(b)$

1751: and $\Psi(e_i b)=\eh_i \Psi(b)$.

1752: \end{enumerate}

1753: Moreover, the following must also hold for all $b\in B$, $i\in

1754: \{1,\dotsc,n-1\}$, and $j\in \{0,n\}$:

1755: \begin{equation} \label{eq:aligned}

1756: \begin{split}

1757: \vp_i(b)&=\vp_i(\Psi(b))=\vp_{N-i}(\Psi(b)) \\

1758: \ve_i(b)&=\ve_i(\Psi(b))=\ve_{N-i}(\Psi(b)) \\

1759: \frac{2}{\xi} \, \vp_j(b)&=\vp_j(\Psi(b)) \\

1760: \frac{2}{\xi} \, \ve_j(b)&=\ve_j(\Psi(b))

1761: \end{split}

1762: \end{equation}

1763: From this it follows that

1764: \begin{equation} \label{eq:Psiwt}

1765: \wt(\Psi(b))=\Psi(\wt(b)) \qquad\text{for all $b\in B$.}

1766: \end{equation}

1767: Define $\Vh^s,\Vh^{s\prime}\in \CC(A_{N-1}^{(1)})$ by

1768: \begin{equation} \label{eq:rowV}

1769: \Vh^s = B^{s\vee} \otimes B^s \qquad\qquad \Vh^{s\prime} = B^s

1770: \otimes B^{s\vee}

1771: \end{equation}

1772: and maps $\Psi_s:B_\se^s\rightarrow \Vh^s$ and

1773: $\Psi_s':B_\se^s\rightarrow \Vh^{s\prime}$ by:

1774: \begin{equation} \label{eq:Vrow}

1775: \begin{aligned}

1776:  \Psi_s(u_{B_\se^s}) &= u(\Vh^s) = N^{\vee s} \otimes 1^s &\qquad \Psi_s'(u_{B_\se^s}) &= u(\Vh^{s\prime})

1777:     = 1^s \otimes N^{\vee s}  \\

1778:  \Psi_s(f_i(b)) &= \fh_i \Psi_s(b) &\qquad \Psi_s'(f_i(b)) &= \fh_i \Psi_s'(b)

1779: \end{aligned}

1780: \end{equation}

1781: for all $0\le i\le n$ and $b\in B_\se^s$. By \cite{OSS03b} $\Psi_s$

1782: and $\Psi_s'$ are well-defined injections. Let $V^s=\Image(\Psi_s)$

1783: and $V^{s\prime}=\Image(\Psi_s')$. Then $(V^s,\Vh^s)$ and

1784: $(V^{s\prime},\Vh^{s\prime})$ are aligned virtual crystals with

1785: bijections $\Psi_s:B_\se^s \rightarrow V^s$ and $\Psi_s':B_\se^s

1786: \rightarrow V^{s\prime}$. The map $\Psi_s$ and set $V^s$ are

1787: explicitly given in \cite{OSS03b}. One also has

1788: \begin{equation} \label{eq:PsiR}

1789:   \Psi_s' = R_{B^{s\vee},B^s} \circ \Psi_s.

1790: \end{equation}

1791: On classical highest weight vectors one has

1792: \begin{equation} \label{eq:embhwv}

1793:   \Psi(1^r) = N^{\vee r} 1^{\vee(s-r)} \otimes 1^s

1794: \end{equation}

1795: for $r$ as in \eqref{eq:Bdecomp}.

1796: 

1797: \begin{example} \label{ex:PsiBox}

1798: For $s=1$ the maps $\Psi$ and $\Psi'$ are given explicitly by

1799: \begin{equation} \label{eq:D2toA}

1800: \begin{aligned}

1801:   \Psi(i) &= i^{*\vee} \otimes i &\qquad \Psi'(i) &= i\otimes i^{*\vee} \\

1802:   \Psi(\bar{i}) &= i^\vee \otimes i^* &\qquad \Psi'(\bar{i}) &= i^*\otimes i^\vee \\

1803:   \Psi(0) &= (n+1)^\vee \otimes (n+1) &\qquad \Psi'(0) &= n \otimes n^\vee \\

1804:   \Psi(\nn) &= 1^\vee \otimes 1 &\qquad \Psi'(\nn) &= N \otimes

1805:   N^\vee

1806: \end{aligned}

1807: \end{equation}

1808: for $1\le i\le n$ where $i^*$ and $i^{*\vee}=i^{\vee*}$ are defined

1809: in section \ref{ss:Dynkinrev}.

1810: \end{example}

1811: 

1812: Aligned virtual crystals form a tensor category \cite{OSS03}: if

1813: $(V_1,\Vh_1)$ and $(V_2,\Vh_2)$ are aligned virtual crystals of type

1814: $\se$ with bijections $\Psi_i:B_i\rightarrow V_i$, then $(V_2\otimes

1815: V_1,\Vh_2\otimes \Vh_1)$ is an aligned virtual crystal of type $\se$

1816: with bijection $\Psi_2\otimes \Psi_1:B_2\otimes B_1 \rightarrow V_2

1817: \otimes V_1$.

1818: 

1819: The virtual crystal for an arbitrary $B\in\CC$ is obtained by

1820: tensoring together the virtual crystals for its tensor factors. We

1821: introduce notation for such virtual crystals. Let

1822: $\nu=(\nu_1,\nu_2,\dotsc,\nu_m)$ be a sequence of positive integers.

1823: Define $B_\se^\nu\in \CC^1(\gf_n^\se)$ and

1824: $\Vh^\nu,\Vh^{\nu\prime}\in\CC(A_{2n-1}^{(1)})$ by

1825: \begin{equation} \label{eq:VB}

1826: \begin{split}

1827: B_\se^\nu&=B_\se^{\nu_m}\otimes\dotsm\otimes B_\se^{\nu_1}  \\

1828: \Vh^\nu &= \Vh^{\nu_m} \otimes\dotsm\otimes \Vh^{\nu_1}

1829: \\

1830: \Vh^{\nu\prime} &= \Vh^{\nu_m\prime} \otimes\dotsm\otimes

1831: \Vh^{\nu_1\prime}

1832: \end{split}

1833: \end{equation}

1834: Then $\Psi_\nu:B_\se^\nu\rightarrow \Vh^\nu$ defined by

1835: $\Psi_\nu=\Psi_{\nu_m} \otimes\dotsm\otimes \Psi_{\nu_1}$ gives the

1836: virtual crystal $(V^\nu,\Vh^\nu)$ where $V^\nu=\Image(\Psi_\nu)$.

1837: $\Psi_\nu':B_\se^\nu\rightarrow \Vh^{\nu\prime}$ is defined

1838: similarly. Again

1839: \begin{equation} \label{eq:PsinuR}

1840:   \Psi_\nu' = R \circ \Psi_\nu

1841: \end{equation}

1842: where $R$ is the tensor product of $R_{B^{\mu_i\vee},B^{\mu_i}}$.

1843: 

1844: Let $L$ be the multiplicity array for $B_\se^\nu$, that is,

1845: $L_i^{(1)}$ is the number of times the part $i$ occurs in $\nu$ and

1846: $L_i^{(a)}=0$ for $a>1$. For $\la\in\Pb^+$, define

1847: $\Pv(B_\se^\nu,\la)$ to be the set of classical highest weight

1848: vectors in $V^\nu\subset \Vh^\nu$ of weight $\Psi(\la)$. The VX

1849: formula is given by

1850: \begin{equation*}

1851: \VX(B_\se^\nu,\la)(t) = \sum_{b\in \Pv(B_\se^\nu,\la)}

1852: t^{\frac{1}{\xi}\Db_{\Vh^\nu}(b)}.

1853: \end{equation*}

1854: The coenergy function is computed in the

1855: $U'_q(A_{N-1}^{(1)})$-crystal $\Vh^\nu$.

1856: 

1857: \begin{theorem} \cite{OSS03b} For $B_\se^\nu\in \CC^1(\gf_n^\se)$

1858: and $L$ as above and for $\la\in\Pb^+$,

1859: $$X_{L,\la}^\se(t) = \VX(B_\se^\nu,\la)(t).$$

1860: \end{theorem}

1861: 

1862: For later we need the following fact: the image of $\Psi_\nu$

1863: consists of elements that are fixed by the map $d\mapsto d^{\vee*}$

1864: up to $R$-matrices.

1865: 

1866: \begin{prop} \label{pp:Psiselfdual} \cite{OSS03} For all

1867: $b\in B_\se^\nu$,

1868: \begin{equation} \label{eq:Psiselfdual}

1869:   \Psi_\nu'(b) = \Psi_\nu(b)^{\vee*}.

1870: \end{equation}

1871: \end{prop}

1872: \begin{proof} One may reduce to the case of a single tensor factor using

1873: \eqref{eq:veetensor} and \eqref{eq:startensor}. For a single tensor

1874: factor \eqref{eq:Psiselfdual} follows from \eqref{eq:RBBv}.

1875: \end{proof}

1876: 

1877: The $R$-matrix in type $\se\in\onetwoset$ is achieved by a

1878: composition of $R$-matrices of type $\nn$ in the virtual crystal.

1879: Suppose $(V_1,\Vh_1)$ and $(V_2,\Vh_2)$ are aligned virtual crystals

1880: of type $\se$ with bijections $\Psi_i:B_i\rightarrow V_i$. The

1881: combinatorial $R$-matrix $R_{\Vh_2,\Vh_1}:\Vh_2\otimes

1882: \Vh_1\rightarrow \Vh_1\otimes \Vh_2$, restricts to a bijection

1883: called the \textbf{virtual combinatorial $R$-matrix} $\Rv:V_2\otimes

1884: V_1\rightarrow V_1\otimes V_2$ such that the diagram commutes

1885: \cite{ScS04}:

1886: \begin{equation} \label{eq:VR}

1887: \begin{CD}

1888: B_2\otimes B_1 @>R>> B_1 \otimes B_2 \\

1889: @V{\Psi_2\otimes \Psi_1}VV @VV{\Psi_1\otimes \Psi_2}V \\

1890: V_2 \otimes V_1 @>>{\Rv}> V_1 \otimes V_2

1891: \end{CD}

1892: \end{equation}

1893: This is a bijective realization of the invariance of the $\VX$

1894: formula under permutations of the parts of $\nu$.

1895: 

1896: \subsection{The VXR map}

1897: For a composition $\nu$ we define a map $\eta_\nu$ as in

1898: \eqref{eq:bij}. Let $L$ be the multiplicity array for $B_\se^\nu\in

1899: \CC^1(\gf_n^\se)$. Define $B^\nu,B_\vee^\nu \in \CC(A_{N-1}^{(1)})$

1900: by

1901: \begin{equation} \label{eq:BBvee}

1902: \begin{split}

1903: B^\nu&=B^{\nu_m}\otimes\dotsm\otimes B^{\nu_1} \\

1904: B_\vee^\nu &= B^{\nu_m \vee} \otimes\dotsm\otimes B^{\nu_1\vee}

1905: \end{split}

1906: \end{equation}

1907: We have a commutative diagram

1908: \begin{equation} \label{eq:VXRrow}

1909: \xymatrix{

1910:   {B_\se^\nu} \ar[d]_{\Psi_\nu} \ar[r]^= &

1911:     {B_\se^\nu} \ar[d]^{\Psi_\nu'} \\

1912:   {\Vh^\nu} \ar[d]_{R_+} \ar[r]^{R_0} &

1913:  {\Vh^{\nu\prime}} \ar[d]^{R_-} \\

1914:   {B_\vee^\nu \otimes B^\nu} \ar[r]_R &

1915:   {B^\nu \otimes B_\vee^\nu}

1916: }

1917: \end{equation}

1918: where $R_+^{-1},R_0,R_-,R$ are all compositions of $R$-matrices of

1919: the form $R_{ B^{s\vee} \otimes B^t }$ applied at adjacent tensor

1920: positions. Let $b\in P_\se(B_\se^\nu,\la)$ and

1921: \begin{equation} \label{eq:dcdef}

1922: \dck \otimes c = R_+(\Psi_\nu(b)) \qquad d \otimes \cck =

1923: R_-(\Psi_\nu'(b))

1924: \end{equation}

1925: 

1926: \begin{example} \label{ex:dckc} In the running example it is convenient to take

1927: $n=3$ even though this is smaller than the generally necessary bound

1928: for $n$. With $b$ as in Example \ref{ex:X},

1929: \begin{equation*}

1930: \begin{tabular}{ccccccccc}

1931: $b$ & = & $\bb$ & 1 & $\nn$ & $\ba$ & 2 & 1 & 1 \\

1932: $\Psi(b)$ & = & $\cb5$ & $\cf1$ & $\ca1$ & $\ca6$ & $\ce2$ & $\cf1$

1933: & $\cf1$ \\

1934: $\Psi'(b)$ & = & $5\cb$ & $1\cf$ & $6\cf$ & $6\ca$ & $2\ce$ & $1\cf$

1935: & $1\cf$

1936: \end{tabular}

1937: \end{equation*}

1938: The map $R_+$ sending $\Psi(b)$ to $\dck \otimes c$ is computed

1939: below. The first line is $\Psi(b)$. To obtain each successive line,

1940: the $R$-matrix \eqref{eq:RBBv} is applied to each pair $i \otimes

1941: j^\vee$. The last line is $\dck\otimes c$.

1942: \begin{equation*}

1943: \begin{array}{cccccccccccccc}

1944: \twocol{7}&\twocol{6}& \twocol{5} & \twocol{4} & \twocol{3} & \twocol{2} & \twocol{1} \\ \cline{1-14} \\[-8pt]

1945: \cb&5&\cf&1&\ul{\ca}&\ul{1}&\ca&6&\ce&2&\cf&1&\cf&1 \\

1946: \cb&\cf&5&\cb&2&\cb&2&\ce&\ul{6}&\ul{\cf}&2&\cf&1&1 \\

1947: \cb&\cf&\cb&5&\cc&3&\ce&2&\ul{\ca}&\ul{1}&\cf&2&1&1 \\

1948: \cb&\cf&\cb&\cc&5&\ce&3&\ca&2&\cf&1&2&1&1 \\

1949: \cb&\cf&\cb&\cc&\cf&6&\ca&3&\cf&2&1&2&1&1 \\

1950: \cb&\cf&\cb&\cc&\cf&\ca&\ul{6}&\ul{\cf}&3&2&1&2&1&1 \\

1951: \cb&\cf&\cb&\cc&\cf&\ca&\ul{\ca}&\ul{1}&3&2&1&2&1&1

1952: \end{array}

1953: \end{equation*}

1954: Therefore $\dck=\cb\cf\cb\cc\cf\ca\ca$ and $c=1321211$.

1955: \end{example}

1956: 

1957: \begin{lemma} \label{lem:classres} $\dck \otimes c$ and $c$ are

1958: classical highest weight vectors.

1959: \end{lemma}

1960: \begin{proof} Since $b$ is a highest weight vector, $\ve_i(b)=0$ for

1961: all $1\le i\le n$. By \eqref{eq:aligned} $\ve_i(\Psi(b))=0$ for

1962: $1\le i\le N-1$, that is, $\Psi(b)$ is a classical highest weight

1963: vector. Since $R_+$ is an affine crystal isomorphism, $\dck\otimes c

1964: = R_+(\Psi(b))$ is also a classical highest weight vector. By

1965: \eqref{eq:ftensor}, $c$ is a classical highest weight vector.

1966: \end{proof}

1967: Let

1968: \begin{equation} \label{eq:Rtaudef}

1969:   \tau=\wt(c)

1970: \end{equation}

1971: \begin{equation} \label{eq:Bhalf}

1972: \begin{split}

1973:   B^{1\vee}_- &=\{N^\vee<\dotsm<(n+1)^\vee\} \\

1974:   B^{1\vee}_+ &=\{n^\vee<\dotsm<1^\vee \}

1975: \end{split}

1976: \end{equation}

1977: and let $\dcp$ (resp. $\dcm$) be the word obtained by concatenating

1978: the tensor factors in $\dck$ and then restricting to the subalphabet

1979: $B^{1\vee}_+$ (resp. $B^{1\vee}_-$).

1980: 

1981: Define

1982: \begin{equation} \label{eq:Zmudef}

1983:   Z = [\dcp^\vee] \qquad\qquad \mu=\shape(Z)

1984: \end{equation}

1985: 

1986: \begin{example} \label{ex:Zmu} For $n=3$, $B^{1\vee}_-=\{\cf<\ce<\cd\}$ and

1987: $B^{1\vee}_+=\{\cc<\cb<\ca\}$. With $\dck$ as in Example

1988: \ref{ex:dckc},

1989: \Yboxdim{10pt}%

1990: \begin{equation*}

1991: \dck = \begin{pmatrix} 7654321 \\ \cb\cf\cb\cc\cf\ca\ca

1992: \end{pmatrix} \quad

1993: \dcp=\begin{pmatrix} 75421 \\ \cb\cb\cc\ca\ca

1994: \end{pmatrix} \quad

1995: \dcm=\begin{pmatrix} 63 \\ \cf\cf

1996: \end{pmatrix} \quad

1997: \dcp^\vee = 11322 \qquad Z=\young(1122,3)

1998: \end{equation*}

1999: \end{example}

2000: 

2001: 

2002: The VXR map is defined by

2003: \begin{equation} \label{eq:VXR}

2004: \begin{split}

2005:  \eta_\nu: P_\se(B_\se^\nu,\la) &\rightarrow \bigcup_\tau

2006:  \bigcup_{\mu\in \Par^\se} P_\nn(B^\nu,\tau) \times

2007:  \LR(\tau;\la,\mu) \\

2008:  b &\mapsto (c,Z)

2009: \end{split}

2010: \end{equation}

2011: where the Littlewood-Richardson data $Z$ takes the form of a

2012: $\tau/\la$-Yamanouchi tableau of shape $\mu$; see Theorem

2013: \ref{th:LR}. To prove Theorem \ref{th:KX}, by the discussion in

2014: section \ref{ss:sketch} it suffices to establish the following

2015: result.

2016: 

2017: \begin{theorem} \label{th:VXR}

2018: $\eta_\nu$ is a well-defined bijection \eqref{eq:bij} satisfying

2019: \eqref{eq:bijstat}.

2020: \end{theorem}

2021: 

2022: \begin{lemma} \label{lem:details}

2023: \begin{enumerate}

2024: \item $\tau\supset \la$.

2025: \item $\dcm$ is a highest weight vector of weight

2026: $-w_0\la=(0^n,-\la_n,\dotsc,-\la_1)$ where $w_0$ is the longest

2027: element in the Weyl group of type $A_{N-1}$.

2028: \item The word $\dcp^\vee$ is $\tau/\la$-Yamanouchi.

2029: \end{enumerate}

2030: In particular $c\in P_\nn(B^\nu,\tau)$ for some partition $\tau$ and

2031: $Z$ is a $\tau/\la$-Yamanouchi tableau of some partition shape

2032: $\mu$.

2033: \end{lemma}

2034: \begin{proof} By the definition of $\Psi$ and the fact that

2035: $R$-matrices are isomorphisms,

2036: \begin{equation} \label{eq:embedwt} \wt(\dck\otimes c) =

2037: \wt(\Psi(b)) =\la-w_0\la = (\la_1,\dotsc,\la_n,-\la_n,\dotsc,-\la_1)

2038: \end{equation}

2039: Since the assertions to be proved only involve the

2040: $U_q(A_{N-1})$-crystal structure, using the $U_q(A_{N_1})$-crystal

2041: embeddings $B^s\rightarrow B^{1\otimes s}$ and $B^{s\vee}\rightarrow

2042: B^{1\vee\otimes s}$ of \eqref{eq:rowtensor} we may regard

2043: $\dck\otimes c$ as a word in the alphabet $B^1\cup B^{1\vee}$.

2044: Consider the sequence of weights of the right factors of

2045: $\dck\otimes c$ as it is scanned from right to left, letter by

2046: letter. They are all dominant weights by Lemma \ref{lem:classres},

2047: starting with the zero weight and ending at $\la-w_0\la$. Consider

2048: the first $n$ parts of the changing weights. They go from $(0^n)$ to

2049: $\tau$, adding $1$ to some part of the weight as $c$ is scanned from

2050: right to left, and then to $\la$ by removing cells corresponding to

2051: letters in $\dcp$. Consider the last $n$ parts of the weights. They

2052: remain fixed at $(0^n)$ as $c$ is scanned, and then go to

2053: $(-\la_n,\dotsc,-\la_1)$, subtracting $1$ from some part of the

2054: weight, for each letter in $\dcm$. This proves the various

2055: assertions.

2056: \end{proof}

2057: 

2058: \begin{remark} The splitting of all of the factors $B^s$ into $B^1$

2059: that was used in the above proof, preserves classical crystal

2060: structure but destroys the affine structure and therefore the

2061: coenergy. It turns out that splitting the rightmost tensor factor

2062: preserves the coenergy even though it does not respect the affine

2063: crystal structure. This is exploited in the next section.

2064: \end{remark}

2065: 

2066: Theorem \ref{th:VXR} is proved by descending induction on the number

2067: of parts of $\nu$. We shall reduce to the case $\nu=(1,1,\dotsc,1)$

2068: using a system of crystal embeddings.

2069: 

2070: \subsection{Right splitting}

2071: \label{ss:rs}

2072: 

2073: \begin{prop} \label{pp:split} \cite{S02a,ScS04} Let $\CC$ be either

2074: $\CC(A_n^{(1)})$ or $\CC^1(\gf_n^\se)$ and $B^{r,s}\in \CC$. Then

2075: there is a $U_q(\gb)$-crystal embedding

2076: \begin{equation} \label{eq:rs}

2077:  \rs_{r,s}: B^{r,s} \rightarrow B^{r,s-1}\otimes B^{r,1}

2078: \end{equation}

2079: called \textbf{right splitting}, such that, for any $B\in \CC$, the

2080: map $1 \otimes \rs_{r,s}:B \otimes B^{r,s} \rightarrow B \otimes

2081: B^{r,s-1} \otimes B^{r,1}$ is a $U_q(\gb)$-crystal embedding that

2082: preserves intrinsic coenergy.

2083: \end{prop}

2084: 

2085: We need explicit formulas for $\rs$ \cite{ScS04}.

2086: 

2087: For $A_{N-1}^{(1)}$, we write $\rs_s:B^s\rightarrow B^{s-1}\otimes

2088: B^1$ for $\rs_{1,s}$, and $\rs_{s\vee}: B^{s\vee}\rightarrow

2089: B^{(s-1)\vee} \otimes B^{1\vee}$ for $\rs_{N-1,s}$; see

2090: \eqref{eq:AKRdual}. Explicitly $\rs_s(b_s\dotsm b_1)=b_s\dotsm b_2

2091: \otimes b_1$ where $b_i\in B^1$. The same rule works for

2092: $\rs_{s\vee}$ except that $b_i\in B^{1\vee}$. These maps literally

2093: split off the rightmost letter.

2094: 

2095: For $\gf_n^\se$ we write $\rs^\se_s:B^s_\se \rightarrow B^{s-1}_\se

2096: \otimes B^1_\se$ for $\rs_{1,s}$. Explicitly, if $b\in

2097: B(r\omega_1)\subset B^s_\se$ for $r\ge1$, then one may write $b=ux$

2098: with $x\in B(\omega_1)$ and $u\in B((r-1)\omega_1)$. In this case

2099: $\rs_s^\se(b)=u\otimes x$. Otherwise $r=0$ and $b=\nn\in B(0)\subset

2100: B^s_\se$ and $\rs_s^\se(b)=\bar{1}\otimes 1$.

2101: 

2102: \begin{theorem} \label{th:Vsplit} \cite{ScS04} Let

2103: $B_\se\in\CC^1(\gf_n^\se)$ have the virtual crystal $(V,\Vh)$ with

2104: bijection $\Psi_B:B_\se\rightarrow V$. Then the following diagram

2105: commutes:

2106: \begin{equation} \label{eq:Vsplit}

2107: \begin{CD}

2108:   B_\se \otimes B_\se^s @>{\Psi_B\otimes \Psi_s}>> V \otimes V^s

2109:   \\

2110:   @V{1\otimes \rs^\se}VV @VV{1\otimes \rs'}V \\

2111:   B_\se \otimes B_\se^{s-1} \otimes B_\se^1 @>>{\Psi_B\otimes \Psi_{s-1}\otimes \Psi_1}> V \otimes V^{s-1}

2112:   \otimes V^1

2113: \end{CD}

2114: \end{equation}

2115: where $\rs':V^s\rightarrow V^{s-1} \otimes V$ is a composition of

2116: $R$-matrices and embeddings $\rs_\nn$.

2117: \end{theorem}

2118: 

2119: \begin{example} \label{ex:rs'}

2120: The map $\rs':V^s\rightarrow V^{s-1} \otimes V^1$ is given by the

2121: restriction to $V^s$ of the map $\rsh:\Vh^s\rightarrow \Vh^{s-1}

2122: \otimes \Vh^1$ defined by \cite{ScS04}

2123: \begin{equation*}

2124: \xymatrix{ {B^{s\vee} \otimes B^s} \ar[r]^-{R} &{B^s\otimes

2125: B^{s\vee}} \ar[r]^-{1\otimes \rs_{s\vee}} & {B^s\otimes

2126: B^{(s-1)\vee} \otimes B^{1\vee}} \ar[r]^-{R} &}

2127: \end{equation*}

2128: \begin{equation*}

2129: \xymatrix{ {B^{(s-1)\vee} \otimes B^\vee \otimes B^s}

2130: \ar[r]^-{1\otimes 1\otimes \rs_s} &{B^{(s-1)\vee} \otimes B^{1\vee}

2131: \otimes B^{s-1} \otimes B^1} \ar[r]^-{R} & }

2132: \end{equation*}

2133: \begin{equation*}

2134: \xymatrix{ {B^{(s-1)\vee} \otimes B^{s-1} \otimes B^{1\vee}  \otimes

2135: B^1} }

2136: \end{equation*}

2137: \end{example}

2138: 

2139: \subsection{Reduction from rows to boxes}

2140: We introduce coenergy-preserving embeddings of tensor products of KR

2141: modules into other such tensor products. In type $A$ ($\se=\nn$),

2142: under the identification of $P(L,\la)$ with semistandard tableaux of

2143: shape $\la$ and weight specified by $L$, one may compose such

2144: embeddings and combinatorial $R$-matrices to define an embedding

2145: which coincides with the map of Lascoux \cite{L91}, which sends a

2146: semistandard tableau to a standard one of the same shape and

2147: cocharge. The generalization of this embedding for a general tensor

2148: product of KR crystals in type $A_n^{(1)}$ is given in \cite{S02a}.

2149: Using virtual crystals this can be generalized to the affine types

2150: that embed in type $A^{(1)}$.

2151: 

2152: Let $\nu=(\nu_1,\dotsc,\nu_m)$ be a sequence of positive integers.

2153: One may pass from $\nu$ to $(1^{|\nu|})$ using two operations:

2154: \begin{enumerate}

2155: \item Exchanging adjacent parts, say, the $r$-th and $r+1$-th, resulting in

2156: $s_r \nu$.

2157: \item Passing to $\rs(\nu) := (1,\nu_1-1,\nu_2,\dotsc,\nu_m)$ if

2158: $\nu_1 \ge 2$.

2159: \end{enumerate}

2160: Define the map

2161: \begin{equation}

2162:   \thet{\se}{\nu}{s_r\nu}:

2163:   P_\se(B_\se^\nu,\la) \rightarrow P_\se(B_\se^{s_r \nu},\la)

2164: \end{equation}

2165: by an $R$-matrix acting in the $r$-th and $(r+1)$-th positions. This

2166: is a weight- and grade-preserving bijection; see Remark

2167: \ref{rem:Border}. If $\nu_1\ge2$ then define

2168: \begin{equation}

2169:   \thet{\se}{\nu}{\rs(\nu)}: P_\se(B_\se^\nu,\la) \rightarrow

2170:   P_\se(B_\se^{\rs(\nu)},\la)

2171: \end{equation}

2172: by $\rs$ acting on the rightmost tensor factor. This is a weight-

2173: and grade-preserving embedding by Theorem \ref{th:Vsplit}.

2174: 

2175: We reduce the proof of Theorem \ref{th:VXR} to the case

2176: $\nu=(1,1,\dotsc,1)$. The large rank assumption is used here in an

2177: essential way. See also Remark \ref{rem:tauparts}.

2178: 

2179: \begin{prop} \label{pp:nu} Let $\rho=s_r\nu$ or $\rho=\rs(\nu)$ if $\nu_1\ge2$.

2180: Suppose $\eta_\rho$ is a bijection. Then

2181: $\eta_\nu$ is a bijection such that the diagram commutes:

2182: \begin{equation} \label{eq:theta}

2183: \begin{CD}

2184:   P_\se(B_\se^\nu,\la) @>{\eta_\nu}>> \displaystyle{\bigcup_\tau

2185:   \bigcup_{\mu\in \Par^\se}

2186:   P_\nn(B^\nu,\tau) \times \LR(\tau;\la,\mu)} \\

2187:   @V{\thet{\se}{\nu}{\rho}}VV @VV{\bigcup\, (\thet{\nn}{\nu}{\rho} \times 1)}V \\

2188: P_\se(B_\se^\rho,\la) @>>{\eta_\rho}> \displaystyle{\bigcup_\tau

2189: \bigcup_{\mu\in \Par^\se} P_\nn(B^\rho,\tau) \times

2190:   \LR(\tau;\la,\mu)}

2191: \end{CD}

2192: \end{equation}

2193: \end{prop}

2194: \begin{proof} The map $\eta_\nu$ is well-defined by

2195: Lemma \ref{lem:details} except for the condition

2196: \begin{equation} \label{eq:seshape}

2197:   \mu\in \Par^\se

2198: \end{equation}

2199: which is nontrivial in the case $\se=\hdom\,$.

2200: 

2201: Let $b\in P_\se(B_\se^\nu,\la)$, $b'=\thet{\se}{\nu}{\rho}(b)$,

2202: $\dck \otimes c = R_+^\nu(\Psi_\nu(b))$, $\dck' \otimes c' =

2203: R_+^\rho(\Psi_\rho(b'))$, $\eta_\nu(b)=(c,Z)$, and

2204: $\eta_\rho(b')=(c',Z')$. By assumption there is a partition $\tau'$

2205: and a partition $\mu'\in\Par^\se$ such that $c'\in

2206: P_\nn(B^\rho,\tau')$ and $Z'$ is $\tau'/\la$-Yamanouchi of shape

2207: $\mu'$. To show that $\eta_\nu$ is a well-defined injection, it

2208: suffices to prove that

2209: \begin{align}

2210: \label{eq:cc'}

2211:  \thet{\nn}{\nu}{\rho}(c)&=c' \\

2212: \label{eq:ZZ'} Z&=Z'.

2213: \end{align}

2214: Suppose that $\rho=s_r \nu$. Then \eqref{eq:cc'} follows from

2215: \eqref{eq:VR} and the uniqueness of the crystal isomorphisms coming

2216: from compositions of $R$-matrices. Since $\dck$ and $\dck'$ are

2217: related by an $R$-matrix, $\dck\Knuth \dck'$ by Proposition

2218: \ref{pp:KnuthA}. By Proposition \ref{pp:Knuthres} with alphabet

2219: $B^{1\vee}_+$, $\dck|_{B^{1\vee}_+}\Knuth \dck'_{B^{1\vee}_+}$.

2220: Equation \eqref{eq:ZZ'} follows by taking duals and applying

2221: Proposition \ref{pp:Knuthdual} and \eqref{eq:Zmudef}.

2222: 

2223: Let $\rho=\rs(\nu)$ and $\nu=(s,\nuh)$. The diagram

2224: \begin{equation} \label{eq:diagsplitembed}

2225: \xymatrix{%

2226: {B_\se^{\nuh} \otimes B_\se^s } \ar[r]^{1\otimes \rs^\se}

2227: \ar[d]_{\Psi_\nu} & {B_\se^{\nuh} \otimes B_\se^{s-1} \otimes

2228: B_\se^1} \ar[d]^{\Psi_\rho} \\ %

2229: {\Vh^\nuh\otimes B^{s\vee}\otimes B^s} \ar[r]^(.4){1 \otimes \rsh}

2230: \ar[d]_{R_+^\nuh\otimes 1\otimes1} & {\Vh^\nuh\otimes

2231: B^{(s-1)\vee}\otimes B^{s-1} \otimes B^{1\vee} \otimes B^1}

2232: \ar[d]^{R_+^{\nuh}\otimes1\otimes R_{B^{s-1},B^{1\vee}}\otimes 1} \\ %

2233: {B^{{\nuh}\vee} \otimes B^{\nuh} \otimes B^{s\vee} \otimes B^s}

2234: \ar[r]^(.4){1\otimes 1\otimes \rsh'} \ar[d]_{1\otimes R\otimes 1} &

2235: {B^{\nuh\vee} \otimes B^{\nuh} \otimes B^{(s-1)\vee} \otimes

2236: B^{1\vee} \otimes B^{s-1} \otimes B^1} \ar[d]^{1\otimes R'\otimes 1} \\

2237: {B^{{\nuh}\vee} \otimes B^{s\vee}\otimes B^{\nuh}  \otimes B^s}  &

2238: {B^{{\nuh}\vee} \otimes B^{(s-1)\vee} \otimes B^{1\vee} \otimes

2239: B^{\nuh} \otimes B^{s-1} \otimes B^1} }

2240: \end{equation}

2241: commutes, where $\rsh$ is defined in Example \ref{ex:rs'} and

2242: $\rsh': B^{s\vee} \otimes B^s \rightarrow B^{(s-1)\vee} \otimes

2243: B^{1\vee} \otimes B^{s-1} \otimes B^1$ is defined by $\rsh' = (1

2244: \otimes R_{B^{s-1},B^{1\vee}} \otimes 1) \circ \rsh$. The first

2245: square commutes by Theorem \ref{th:Vsplit} and the second by the

2246: fact that $R^\nuh_+$ and $\rsh$ act on disjoint tensor positions.

2247: 

2248: The composite map down the left (resp. right) column of

2249: \eqref{eq:diagsplitembed} is $R_+^\nu \circ \Psi_\nu$ (resp.

2250: $R_+^\rho \circ \Psi_\rho$). Let $b\in P_\se(B_\se^\nu)$. One may

2251: write

2252: \begin{equation}

2253: \begin{split}

2254: \dckh \otimes \cht \otimes w \otimes 1^s &\in B^{\nuh\vee} \otimes B^\nuh \otimes B^{s\vee} \otimes B^s \\

2255: \dckh \otimes \cht \otimes v \otimes x \otimes 1^{s-1} \otimes 1

2256: &\in B^{\nuh\vee} \otimes B^\nuh \otimes B^{(s-1)\vee} \otimes

2257: B^{1\vee} \otimes B^{s-1} \otimes B^1

2258: \end{split}

2259: \end{equation}

2260: for the image of $b$ in the sets in the third row of

2261: \eqref{eq:diagsplitembed}. The rightmost factor of $b$ is a

2262: classical highest weight vector in $1^r\in B_\se^s$ with $r$ as in

2263: \eqref{eq:Bdecomp}. By explicit calculation

2264: \begin{align} \label{eq:hwvform}

2265: &\begin{aligned} w &= 1^{\vee s} \\ v \otimes x &= 1^{\vee(s-1)}

2266: \otimes 1^\vee

2267: \end{aligned}

2268: &\qquad&\text{if $r=0$} \\

2269: &\begin{aligned} w &= N^{\vee r} 1^{\vee(s-r)} \\ v \otimes x &=

2270: N^{\vee (r-1)} 1^{\vee(s-r)}\otimes N^\vee

2271: \end{aligned}

2272: &\qquad&\text{if $r\ge1$.}

2273: \end{align}

2274: 

2275: One may write the images of $b$ in the fourth row of

2276: \eqref{eq:diagsplitembed} in the form

2277: \begin{equation}

2278: \begin{split}

2279: \dckh \otimes w'\otimes \cht' \otimes 1^s &\in B^{\nuh\vee} \otimes B^{s\vee}\otimes B^\nuh  \otimes B^s \\

2280: \dckh \otimes v'\otimes x' \otimes \cht'' \otimes 1^{s-1} \otimes 1

2281: &\in B^{\nuh\vee}  \otimes B^{(s-1)\vee} \otimes B^{1\vee} \otimes

2282: B^\nuh \otimes B^{s-1} \otimes B^1

2283: \end{split}

2284: \end{equation}

2285: To prove \eqref{eq:cc'} and \eqref{eq:ZZ'} it suffices to show that

2286: \begin{equation} \label{eq:thestuff}

2287: \begin{split}

2288:   \cht'&=\cht'' \\

2289:   w'|_{B^{1\vee}_-}&\Knuth v'x'|_{B^{1\vee}_-} \\

2290:   w'|_{B^{1\vee}_+}&\Knuth v'x'|_{B^{1\vee}_+}

2291: \end{split}

2292: \end{equation}

2293: During the $R$-matrices going from the third to fourth rows of

2294: \eqref{eq:diagsplitembed}, in passing from $v\otimes x$ to $v'

2295: \otimes x'$ and from $w$ to $w'$, letters in the subalphabet

2296: $B^{1\vee}_+$ remain in $B^{1\vee}_+$. Some $N^\vee$'s may become

2297: $1^\vee$'s and thereafter remain in $B^{1\vee}_+$; all other letters

2298: in $B^{1\vee}_-$ will remain in $B^{1\vee}_-$. The reason for this

2299: is Proposition \ref{pp:Rhalf} and the large rank assumption; each

2300: letter in $w$ or $v\otimes x$ can only change by one during each

2301: $R$-matrix of the form $R_{B^t,B^{s\vee}}$, and we may assume that

2302: the number $\ell(\nuh)=\ell(\nu)-1$ of tensor factors in $\cht\in

2303: B^{\nuh}$ is strictly less than $n$. Equations \eqref{eq:thestuff}

2304: follow from these considerations and the explicit form of the

2305: highest weight vectors \eqref{eq:hwvform}.

2306: 

2307: This proves the well-definedness and injectivity of $\eta_\nu$. Note

2308: that the maximum number of actual changes that can occur is at most

2309: the number of nonzero parts of $\tau$; see Remark

2310: \ref{rem:tauparts}.

2311: 

2312: For the surjectivity of $\eta_\nu$, let $\tau$ be a partition,

2313: $\mu\in \Par^\se$, $c\in P_\nn(B^\nu,\tau)$, and $Z$ a

2314: $\tau/\la$-Yamanouchi tableau of shape $\mu$. Let

2315: $c'=\thet{\nn}{\nu}{\rho}(c)$. By the surjectivity of $\eta_\rho$

2316: there is a $b'\in P_\se(B_\se^\rho,\la)$ such that

2317: $\eta_\rho(b')=(c',Z)$, with $R_+^\rho(\Psi_\rho(b'))=\dck'\otimes

2318: c'$.

2319: 

2320: Suppose first that $\rho=s_r\nu$. Let $b\in B_\se^\nu$ be such that

2321: $R_j^\se(b)=b'$ where $R_j^\se$ exchanges the $r$-th and $(r+1)$-th

2322: factors in $b$. Let $\eta_\nu(b)=(c'',Z'')$ with

2323: $R_+^\nu(\Psi_\nu(b))=\dck''\otimes c''$. By \eqref{eq:VR}

2324: $R_j^\nn(c'')=c'$ where $R_j^\nn$ exchanges the $r$-th and

2325: $(r+1)$-th tensor factors in $c''$. By definition for $\rho=s_r\nu$,

2326: $\thet{\nn}{\nu}{\rho}=R_j^\nn$, so that $R_j^\nn(c)=c'$. Since

2327: $R_j^\nn$ is an isomorphism $c''=c$. Similarly $\dck''$ and $\dck'$

2328: are related by an $R$-matrix. The desired equality $Z''=Z'$ follows

2329: by Propositions \ref{pp:KnuthA}, \ref{pp:Knuthres}, and

2330: \ref{pp:Knuthdual}.

2331: 

2332: Otherwise suppose $\nu_1\ge2$ and $\rho=\rs(\nu)$. Let

2333: $\nu=(s,\nuh)$ with $s\ge2$ and $b'=y\otimes v\otimes x\in

2334: B_\se^\rho=B_\se^{\nuh}\otimes B_\se^{s-1}\otimes B_\se^1$. Since

2335: $b'$ is a classical highest weight vector, so is $v\otimes x$. We

2336: list the possibilities for $v\otimes x$, together with the resulting

2337: two rightmost tensor factors in $c'$.

2338: \begin{equation*}

2339: \begin{array}{|c||c|} \hline

2340: v\otimes x & c'_2 \otimes c'_1 \\ \hline \hline %

2341: 1^r \otimes 1 & 1^r \otimes 1 \\ \hline %

2342: 1^r 2\otimes 1 & 1^r 2\otimes 1 \\ \hline %

2343: \raisebox{-1pt}{\vphantom{L} $1^r \bar{1} \otimes 1$} & 1^r 2\otimes 1 \\ \hline %

2344: 1^r \otimes \nn & 1^{r-1}2\otimes 1 \\ \hline %

2345: \nn\otimes \nn & 2\otimes 1 \\ \hline %

2346: \end{array}

2347: \end{equation*}

2348: Now $c'$ is in the image of $\thet{\nn}{\nu}{\rho}=1_{B^{\nuh}}

2349: \otimes \rs_s$. By the definition of $\rs$ in type $A$, it follows

2350: that the only possible $c'$ is $1^r \otimes 1$. Define $b=y \otimes

2351: 1^{r+1}\in B_\se^\nu$. It satisfies $\thet{\se}{\nu}{\rho}(b)=b'$.

2352: By the commutativity of the diagram \eqref{eq:theta} and the

2353: injectivity of the vertical maps, it follows that

2354: $\eta_\nu(b)=(c,Z)$ as desired.

2355: \end{proof}

2356: 

2357: 

2358: \begin{remark} \label{rem:tauparts}

2359: The proof only requires that $\tau$ have at most $n$ nonzero parts.

2360: \end{remark}

2361: 

2362: \subsection{Grading}

2363: \label{ss:grading}

2364: 

2365: We show that the VXR map $\eta_\nu$ satisfies \eqref{eq:bijstat}.

2366: The proof uses the large rank assumption in that it is assumed that

2367: $\tau$ has at most $n$ nonzero parts.

2368: 

2369: \begin{prop} \label{pp:stat} For

2370: $b\in P_\se(B_\se^\nu,\la)$ and $c$ derived from $b$ by

2371: \eqref{eq:dcdef} the equation \eqref{eq:bijstat} holds.

2372: \end{prop}

2373: 

2374: \begin{example} \label{ex:stat} Let $L=7$, $\se=\cell\,$, $\tau=(4,2,1)$, and

2375: $\la=(2)$. Let $b\in P_\se(1^L,\la)$ and $c\in P_\nn(1^L,\tau)$ be

2376: as in Example \ref{ex:dckc}. Using Example \ref{ex:X},

2377: \eqref{eq:bijstat} says $2(25/2)=25 = (7 - 2) + 2\cdot 10$.

2378: \end{example}

2379: We have \cite{OSS03}

2380: \begin{equation} \label{eq:Dembed}

2381:   \Db_\nn(\Psi(b)) = 2 \Db_\se(b).

2382: \end{equation}

2383: For the following calculation we assume that $\se=\cell\,$ since

2384: $\se=\hdom$ is essentially a special case.

2385: 

2386: Since the embeddings $\thet{\se}{\nu}{\rho}$ preserve highest

2387: weights and coenergy, by Proposition \ref{pp:nu} we may reduce to

2388: the case that $\nu=(1,1,\dotsc,1)=1^L$.

2389: 

2390: Let $B_2=B^{1\vee \otimes L}$ and $B_1=B^{1\otimes L}$. By

2391: Proposition \ref{pp:DR} and \eqref{eq:D2} we have

2392: \begin{equation} \label{eq:Dwork}

2393:   \Db_\nn(\Psi(b)) = \Db_\nn(\dck \otimes c) =

2394:   \Hb_{B_2 \otimes B_1}(\dck \otimes c) +

2395:   \Db_{B_1}(c) + \Db_{B_2}(\cck)

2396: \end{equation}

2397: where $\cck$ is defined by \eqref{eq:dcdef}. To establish

2398: \eqref{eq:bijstat} it suffices by \eqref{eq:Dembed} and

2399: \eqref{eq:Dwork} to show that

2400: \begin{align}

2401: \label{eq:HH} \Hb_{B_2\otimes B_1}(\dck\otimes c) &= L -|\la| \\

2402: \label{eq:Dccheck} \Db_{B_1}(c) &= \Db_{B_2}(\cck).

2403: \end{align}

2404: To verify \eqref{eq:Dccheck} we compute the $R$-matrix $B_2\otimes

2405: B_1\rightarrow B_1 \otimes B_2$ using the composition $R_- \circ R_0

2406: \circ R_+$ where $R_0=R_{B^{1\vee},B^1}^{\otimes L}:(B^{1\vee}

2407: \otimes B^1)^{\otimes L}\rightarrow (B^1 \otimes B^{1\vee})^{\otimes

2408: L}$.

2409: 

2410: By Propositions \ref{pp:Rvee}, \ref{pp:R*}, and \ref{pp:Psiselfdual}

2411: and equations \eqref{eq:veetensor} and \eqref{eq:startensor}, we

2412: have

2413: \begin{equation*}

2414:   d \otimes \cck = (\dck \otimes c)^{\vee*} = (\dck)^{\vee*}

2415:   \otimes c^{\vee*}.

2416: \end{equation*}

2417: In particular $\cck = c^{\vee*}$. To prove \eqref{eq:Dccheck} it

2418: suffices to prove the following result.

2419: 

2420: \begin{prop} \label{pp:DAdual} For any $c\in B^{1\otimes L}$,

2421: \begin{equation} \label{eq:DAdual}

2422:   \Db(c) = \Db(c^{\vee*}).

2423: \end{equation}

2424: \end{prop}

2425: \begin{proof} Let $c= c_L\dotsm c_1$ with $c_i\in B^1$. Then $c^{\vee*} =

2426: c_L^{\vee*} \dotsm c_1^{\vee*}$ by \eqref{eq:veetensor} and

2427: \eqref{eq:startensor}. Note that for $i,j\in B^1$, $i<j$ if and only

2428: if $i^{\vee*}<j^{\vee*}$ in $B^{1\vee}$. Computing $\Db(c)$ and

2429: $\Db(c^{\vee*})$ using \eqref{eq:Dn=}, \eqref{eq:DB11},

2430: \eqref{eq:HB11}, and \eqref{eq:Ddual}, equation \eqref{eq:DAdual}

2431: follows.

2432: \end{proof}

2433: 

2434: To read the rest of this proof it is useful to look at Example

2435: \ref{ex:dckc}. We now prove \eqref{eq:HH}. From the definitions it

2436: is easy to see that

2437: \begin{equation} \label{eq:DLla}

2438:   L-|\la| = 2m + m_\nn(b)

2439: \end{equation}

2440: where $m$ is the number of barred letters in $b$ and $m_\nn(b)$ is

2441: the number of symbols $\nn$ in $b$. By Proposition \ref{pp:Hbraided}

2442: and \eqref{eq:HBBv}, $\Hb_{B_2 \otimes B_1}(\dck\otimes c)$ is equal

2443: to the number of times that the transformation

2444: \begin{equation} \label{eq:trans}

2445: 1^\vee \otimes 1 \rightarrow N \otimes N^\vee

2446: \end{equation}

2447: occurs during the application of $R_{B_2\otimes B_1}$ that sends

2448: $\dck\otimes c$ to $d\otimes \cck$. Therefore it suffices to show

2449: that

2450: \begin{enumerate}

2451: \item During $\dck \otimes c\rightarrow \Psi(b)$ \eqref{eq:trans}

2452: occurs $m$ times.

2453: \item During $\Psi(b)\rightarrow \Psi'(b)$ \eqref{eq:trans} occurs

2454: $m_\nn(b)$ times.

2455: \item During $\Psi'(b)\rightarrow d\otimes \cck$ \eqref{eq:trans}

2456: occurs $m$ times.

2457: \end{enumerate}

2458: (2) is immediate. By symmetry (3) follows from (1). For (1) it is

2459: equivalent to show that in passing from $\Psi(b)$ to $\dck\otimes c$

2460: by $R_+^{-1}$ there are $m$ applications of the inverse of

2461: \eqref{eq:trans}. To see this, note that for every barred letter

2462: $b_i$ in $b$ the corresponding plain letter in $\Psi(b_i)$ has value

2463: greater than $n$. As this plain letter is switched to the right by

2464: applications of $R^{-1}$, its value increases until it has value

2465: $N=2n$ and it participates in the inverse of \eqref{eq:trans}. It

2466: must do so, for it ends up as a letter in the highest weight vector

2467: $c$, whose weight $\tau$ is a partition with at most $n$ nonzero

2468: parts. Moreover this letter can only participate once in the inverse

2469: of \eqref{eq:trans} since it switches with at most $n$ dual letters.

2470: Conversely, the plain letter in $\Psi(b_i)$ for $b_i$ unbarred,

2471: being of value at most $n$, cannot increase in value to more than

2472: $N$ when it gets switched to the right, and hence cannot participate

2473: in the inverse of \eqref{eq:trans}.

2474: 

2475: This proves Proposition \ref{pp:stat}.

2476: 

2477: 

2478: \subsection{Incremental VXR} \label{ss:incVXR} For the eventual

2479: proof of the case $\nu=(1,1,\dotsc,1)$ of Theorem \ref{th:VXR} we

2480: describe the change in the various data in passing from $b'$ to $b$

2481: where $b=x \otimes b'$ with $x=b_L\in B_\se^1$ and $b'\in

2482: P_\se(1^{L-1},\la')$. Denote the data corresponding to $b'$ by

2483: primes.

2484: 

2485: To compare with the DDF bijection, we define tableaux $\Pt,\Qt,\Tt$

2486: as follows. $\Pt$ is the standard tableau associated with $c$. That

2487: is, $c_i=r$ if and only if the letter $i$ is in the $r$-th row of

2488: $\Pt$. Let $\Qt$ be the injective tableau of shape $\tau/\la$ on the

2489: alphabet $A^*$ where $A$ is the set of positions in $\dck$ occupied

2490: by the subword $\dcp$, such that the $i$-th letter of the biword

2491: $\dcp$ is $r^\vee$ if and only if $i^*$ is in the $r$-th row of

2492: $\Qt$. The tableau $\Qt$ is well-defined by Lemma \ref{lem:details}.

2493: Define the tableau $\Tt$ to be the injective tableau of shape $\la$

2494: whose alphabet is the complement of $A$, such that the biletter

2495: $(i^{*\vee},j)$ is in $\dcm$ if and only if $j$ is in the $i$-th row

2496: of $\Tt$. It is well-defined by Lemma \ref{lem:details}.

2497: 

2498: The map $R_+$ can be computed by the composition

2499: \begin{equation*}

2500: (B^{1\vee} \otimes B^1)^{\otimes L} \rightarrow (B^{1\vee} \otimes

2501: B^1) \otimes B^{1\vee\otimes (L-1)} \otimes B^{1\otimes (L-1)}

2502: \rightarrow (B^{1\vee})^{\otimes L} \otimes B^{1\otimes L}.

2503: \end{equation*}

2504: This sends $\Psi(b) \mapsto \Psi(x) \otimes \dck' \otimes c'\mapsto

2505: \dck \otimes c$. From this it follows that $c=c_L c'$ and hence that

2506: $\Pt'=\Pt|_{<L}$, with extra cell $s$ in the addable cell in the

2507: $c_L$-th row. Let $\Psi(x)=\zv \otimes y$ for some $y,z\in B^1\in

2508: \CC(A_{2n-1}^{(1)})$. We must compute the map $\zv \otimes y \otimes

2509: \dck' \mapsto \dck \otimes c_L$. We see that $\dck$ has last letter

2510: $\zv$. We watch the process that changes $y$ into $c_L$ as it moves

2511: past $\dck'$ by $R$-matrices. By taking sufficiently large $n\ge

2512: \ell(\tau)$ we may assume that $c_L\le n$. Since the moving plain

2513: letter can only interact with $L-1$ dual letters, the moving letter

2514: can change less than $n$ times.

2515: 

2516: \subsubsection{Suppose $x=r$ for $1\le r\le n$.}

2517: Then $y=r$, $z=r^*=2n+1-r$, and $\la$ is obtained by adding a corner

2518: cell $s'$ to $\la'$ in the $r$-th row. In this case no letters of

2519: $\dcm'$ are changed and $\dcm=z^\vee \dcm'$ where $z^\vee$ has index

2520: $L$. This adjoins the letter $L$ to the $r$-th row of $\Tt'$ to give

2521: $\Tt$. The $R$-matrix computation finds the biletter $(r^\vee,i_r)$

2522: in $\dcp'$ with $i_r$ maximal and changes it into

2523: $((r+1)^\vee,i_r)$. This changes the moving plain letter to $r+1$.

2524: Then it finds the biletter $((r+1)^\vee,i_{r+1})$ in $\dcp'$ with

2525: $i_{r+1}<i_r$ maximal. The algorithm continues in this way, ending

2526: with the plain letter $c_L$ and having changed biletters

2527: $(r^\vee,i_r),\dotsc,((c_L-1)^\vee,i_{c_L-1})$ by changing their

2528: first components from $j^\vee$ to $(j+1)^\vee$. There is no biletter

2529: in $\dcp'$ of the form $((c_L)^\vee,i_{c_L})$ with

2530: $i_{c_L}<i_{c_L-1}$. This is equivalent to the internal insertion

2531: $\Qt=I_{s'}(\Qt')$.

2532: 

2533: \subsubsection{Suppose $x=\bar{r}$ for $1\le r\le n$.}

2534: Then $z=r$ and $y=r^*=2n+1-r>n$. Since the moving plain letter $y$

2535: is transformed into a letter $c_L\le n$, it follows that there is a

2536: sequence of biletters

2537: $((2n+1-r)^\vee,j_r),((2n-r)^{\vee},j_{r-1}),\dotsc,((2n)^\vee,j_1)$

2538: in $\dcm'$ with $j_r>j_{r-1}>\dotsm>j_1$, each index maximal, and

2539: biletters $(1^\vee,i_1),(2^\vee,i_2),\dotsc$,

2540: $((c_L-1)^\vee,i_{c_L-1})$ in $\dcp'$ with

2541: $j_1>i_1>i_2>\dotsm>i_{c_L-1}$ with indices maximal. These biletters

2542: change such that their first letter $j^\vee$ is sent to

2543: $(j+1)^\vee$, with the exception that

2544: $((2n)^\vee,j_1)\mapsto(1^\vee,j_1)$. This means that $\Tt$ is

2545: obtained from $\Tt'$ by the reverse row insertion at the corner cell

2546: in the $r$-th row, with ejected letter $j_1$. $\dcp$ is obtained by

2547: the above changes in the letters $(j^\vee,i_j)$ in $\dcp'$, with an

2548: extra letter $r^\vee$ at the left end, indexed by $L$. Therefore

2549: $\Qt$ is obtained from $\Qt'$ by the external insertion of $j_1^*$,

2550: together with the adjoining of $L^*$ to the inside at the addable

2551: cell in the $r$-th row of $\la'$.

2552: 

2553: \subsubsection{Suppose $x=\nn$.} Then $z=y=1$. The moving plain

2554: letter $y=1$ finds the biletter $(1^\vee,j_1)$ with $j_1$ maximal,

2555: then $(2^\vee,j_2)$ with $j_2<j_1$ maximal, etc., ending with

2556: $((c_L-1)^\vee,j_{c_L-1})$. Thus $\Tt=\Tt'$ and $\Qt$ is obtained

2557: from $\Qt'$ by the external insertion of $L^*$.

2558: 

2559: 

2560: \section{The bijection via DDF}

2561: \label{sec:DDF}

2562: 

2563: Let $\se\in \{ \cell\,,\hdom\,\}$ and $\nu=1^L$. This section

2564: defines the DDF map, which is a form of the desired bijection

2565: \eqref{eq:bij} under the present assumptions. The DDF map is the

2566: composition of a number of bijections: Schensted row and column

2567: insertion \cite{Sc61}, the Delest-Dulucq-Favreau (DDF) bijection

2568: \cite{DDF88}, and the Burge correspondence \cite{B74}.

2569: 

2570: \subsection{Highest weight vectors and sequences of partitions}

2571: We first translate the definition of $P_\se(1^L,\la)$ given in

2572: Example \ref{ex:standardpath} into the language of

2573: multiplicity-space tableaux, which are certain sequences of

2574: partitions.

2575: 

2576: Recall the encoding of dominant weights by partitions in subsection

2577: \ref{ss:limit}. We shall use this to identify a dominant integral

2578: weight for one of the classical types $A_{n-1}$, $B_n$, $C_n$,

2579: $D_n$, with an element of $\Z^n$.

2580: 

2581: For each $b\in P_\se(1^L,\la)$ consider the associated sequence of

2582: partitions

2583: $$\lad=(\nn=\la^{(0)}\subset\la^{(1)}\subset\dotsm\subset\la^{(L)}=\la).$$

2584: The sequences of partitions that arise in this manner are

2585: characterized as follows. Every partition $\la^{(i)}$ is obtained

2586: from the previous partition $\la^{(i-1)}$ by

2587: \begin{enumerate}

2588: \item Adding a cell to the $r$-th row if $b_i=r$. This is allowed in all types.

2589: \item Removing a cell from the $r$-th row if $b_i=\bar{r}$.

2590: This is allowed in all types but $\se=\nn$.

2591: \item Doing nothing if $b_i=\nn$. This is allowed only in type $\se=\cell\,$.

2592: \end{enumerate}

2593: These sequences of partitions define three kinds of

2594: multiplicity-space tableaux $\lad$ of shape $\la$ and length $L$.

2595: For type $\nn$ they are called \textbf{standard tableaux}. For types

2596: $\hdom\,,\vdom\,$ they are called \textbf{oscillating tableaux}. For

2597: type $\cell\,$ we call them \textbf{Motzkin tableaux}.

2598: 

2599: \begin{remark} For large rank, the symbol $0$ does not appear in a

2600: classical highest weight vector of type $\cell\,$. Similarly in type

2601: $\vdom\,$ the symbols $n$ and $\bar{n}$ don't appear. Otherwise

2602: these tableau definitions must be modified.

2603: \end{remark}

2604: 

2605: \begin{example} \label{ex:hwv} The elements $c\in P_\nn(1^7,(4,2,1))$ and $b\in P_\cell(1^7,(2))$

2606: in Example \ref{ex:X} have corresponding standard and Motzkin

2607: tableaux given by \Yboxdim{5pt}

2608: \begin{align*}

2609: c&=1321211

2610:   &&(\nn, \quad \yng(1), \quad\yng(2),

2611: \quad\yng(2,1),\quad\yng(3,1),\quad\yng(3,2),\quad\yng(3,2,1),\quad\yng(4,2,1))

2612: \\

2613: b&=\bar{2}1\nn\bar{1}211

2614:   &&(\nn,\quad \yng(1), \quad \yng(2),\quad  \yng(2,1),\quad

2615: \yng(1,1),\quad \yng(1,1),\quad  \yng(2,1),\quad\yng(2)).

2616: \end{align*}

2617: \end{example}

2618: 

2619: \subsection{Robinson-Schensted}

2620: Let $w=w_1w_2\dotsm w_L$ be a word. The row insertion version of the

2621: Robinson-Schensted correspondence \cite{Sc61} is a bijection

2622: $w\mapsto (P,Q)$ between words in an alphabet $A$, and pairs of

2623: tableaux of the same partition shape with $P$ semistandard in the

2624: alphabet $A$ and $Q$ standard. Let $P=[w]$ and let $Q$ be defined by

2625: the sequence of partitions $\lad$ where

2626: $\la^{(i)}=\shape([w_1w_2\dotsm w_i])$. This is denoted

2627: \begin{equation} \label{eq:rowins}

2628:   \bilet{P}{Q} = \left( \bilet{\nn}{\nn} \leftarrow

2629:   \begin{pmatrix}

2630:   1&2&\dotsm&L \\

2631:   w_1&w_2&\dotsm&w_L

2632:   \end{pmatrix} \right).

2633: \end{equation}

2634: The column insertion version of the Robinson-Schensted

2635: correspondence \cite{Sc61} is a bijection between the same sets as

2636: above, sending the word $u=u_L\dotsm u_2 u_1$ to $(P,Q)$ such that

2637: $P=[u]$ and $Q$ is given by the sequence of partitions given by the

2638: shapes of the tableaux $[u_i\dotsm u_2 u_1]$. This is written

2639: \begin{equation} \label{eq:colins}

2640:   \bilet{P}{Q} = \left(

2641:   \begin{pmatrix}

2642:   L&\dotsm&2&1 \\

2643:   u_L&\dotsm&u_2&u_1

2644:   \end{pmatrix} \rightarrow \bilet{\nn}{\nn}

2645:   \right)

2646: \end{equation}

2647: Here the letters of $u$ get inserted started from the \textit{right}

2648: end.

2649: 

2650: \begin{remark} \label{rem:rowcol} By definition the bumping

2651: (``P'') tableaux for the row and the column insertion of a word $w$,

2652: are equal, as both are defined by $[w]$.

2653: \end{remark}

2654: 

2655: The Robinson-Schensted correspondence has the following well-known

2656: involution symmetry.

2657: 

2658: \begin{prop} \label{pp:inv}

2659: Suppose that $w$ is a permutation of the set $\{1,2,\dotsc,L\}$ with

2660: $w(i)$ written $w_i$. Let $w\mapsto (P,Q)$ under the bijection

2661: \eqref{eq:rowins}. Then $w^{-1}\mapsto (Q,P)$ under

2662: \eqref{eq:rowins}. A similar property holds for the bijection

2663: \eqref{eq:colins}.

2664: \end{prop}

2665: 

2666: \subsection{Skew insertion}

2667: \label{ss:skew} The following skew version of Robinson-Schensted row

2668: insertion is due to Sagan and Stanley \cite{SS90}. Let $T$ be a

2669: tableau of skew shape $\la/\mu$. In this algorithm the pair of

2670: shapes $\mu\subset\la$ is important, not just the set difference of

2671: their partition diagrams.

2672: 

2673: The \textbf{external} row insertion of a letter $x$ into the skew

2674: tableau $T$ is defined by inserting $x$ into the first row of $T$

2675: and propagating the insertion in the row insertion algorithm. One

2676: may imagine that the shape $\mu$ is occupied by a tableau having all

2677: letters smaller than those in $x$ or $T$. Let $s$ be the newly

2678: created cell. The result is a tableau of skew shape

2679: $(\la\cup\{s\})/\mu$. This process can be reversed given the cell

2680: $s$.

2681: 

2682: Let $s'$ be an addable corner of $\mu$. The \textbf{internal} row

2683: insertion on $T$ at $s'$ is defined by removing the letter of $T$ at

2684: $s'$ and inserting it into the next row, propagating the insertion

2685: as in the usual row insertion algorithm.

2686: 

2687: Let $s$ be the newly created cell. If $s'$ is not in $T$ then let

2688: $s=s'$ and declare the resulting tableau to be $T$. In either case

2689: the new pair of shapes is declared to be

2690: $(\la\cup\{s\})/(\mu\cup\{s'\}$. This process can also be reversed

2691: given $s$.

2692: 

2693: Note that if one is given a tableau $S$ of skew shape $\la/\mu$ and

2694: a corner $s$ for $\la$, the reverse skew row insertion on $S$ at $s$

2695: may be the reverse of either an internal or external insertion,

2696: depending on whether the reverse bumping path ends by hitting the

2697: shape $\mu$ or not.

2698: 

2699: A skew version of the row insertion Robinson Schensted

2700: correspondence is defined as follows. An \textbf{injective} taleau

2701: is one with no repeated entries. Fix two partitions $\beta$ and

2702: $\gamma$. Let $A$ (values) and $B$ (positions) be alphabets. The

2703: input for the algorithm is a triple $(T,U,w)$ where $w:B'\rightarrow

2704: A'$ is a bijection between subsets $B'\subset B$ and $A'\subset A$,

2705: and $T$ and $U$ are injective tableaux of respective skew shapes

2706: $\beta/\alpha$ and $\gamma/\alpha$ for some $\alpha$, such that the

2707: letters occurring in $T$ and $A'$ (resp. $U$ and $B'$) are disjoint

2708: subsets of $A$ (resp. $B$). The output is a pair $(P,Q)$ of

2709: injective tableaux of respective shapes $\delta/\gamma$ and

2710: $\delta/\beta$ for some $\delta$ with letters taken from the

2711: respective alphabets $A$ and $B$.

2712: 

2713: At the beginning, set the bumping tableau to be the skew tableau $T$

2714: and the recording tableau to be empty of skew shape $\beta/\beta$.

2715: Let $b$ be the smallest letter among $U$ and $B'$. If $b\in U$ then

2716: perform an internal insertion on the bumping tableau at the cell of

2717: $U$ containing $b$. If $b\in B'$ then externally insert $w(b)$ into

2718: the bumping tableau. Either way, adjoin the letter $b$ to the

2719: recording tableau at the newly created cell. Continue with the

2720: letters of $U$ and $B'$ in increasing order in this manner. Then $P$

2721: is the resulting bumping tableau and $Q$ the recording tableau.

2722: 

2723: \begin{example} \label{ex:skewins} Let $A=\{1,2,\dotsc,7\}$ and

2724: $B=A^*=\{7^*<6^*<\dotsm<1^*\}$. Let

2725: \begin{equation*}

2726: T=\young(36)\qquad U=\nn \qquad w=\begin{pmatrix} 7^* & 5^* & 4^* & 2^* & 1^*\\

2727: 1&5&2&4&7

2728: \end{pmatrix}.

2729: \end{equation*}

2730: Here $A'=\{1,2,4,5,7\}$ and $B'={A'}^*$.

2731: \begin{equation*}

2732: P=\young(1247,35,6) \qquad Q=\young(::\stb\sta,\stg\ste,\std)

2733: \end{equation*}

2734: \end{example}

2735: 

2736: \subsection{DDF bijection}

2737: Delest, Dulucq, and Favreau \cite{DDF88} defined a bijection from

2738: oscillating tableaux of shape $\la$ and length $L$, to pairs $(T,I)$

2739: where $I$ is a fixed-point-free involution on a subset

2740: $A\subset\{1,2,\dotsc, L\}$ and $T$ is an injective tableau of shape

2741: $\la$ consisting of exactly the letters in the subalphabet

2742: complementary to $A$ in $\{1,2,\dotsc,L\}$. A skew version of the

2743: DDF bijection was given by Dulucq and Sagan \cite{DS95} but we will

2744: not need the extra generality.

2745: 

2746: We now describe a straightforward extension of the DDF bijection to

2747: a bijection between Motzkin tableaux and pairs $(T,I)$ which are as

2748: above except that the involution $I$ may have fixed points. This

2749: extended DDF bijection also admits a skew version.

2750: 

2751: Let $b=b_L\dotsm b_2 b_1\in B_\cell(1^L,\la)$ and $\lad$ the

2752: associated Motzkin tableau. Start with $T_0$ the empty tableau and

2753: $I_0$ the empty involution. Suppose $T_{i-1}$ and $I_{i-1}$ have

2754: been constructed.

2755: \begin{enumerate}

2756: \item If $b_i=r$, let $T_i$ be obtained by adjoining the

2757: letter $i$ to $T_{i-1}$ in the $r$-th row and let $I_i=I_{i-1}$.

2758: \item If $b_i=\bar{r}$, let $T_{i-1}$ be the result of the reverse

2759: row insertion on $T_i$ at the corner cell in its $r$-th row. Let $a$

2760: be the ejected letter. Then let $I_i=I_{i-1}(a,i)$, that is, the map

2761: $I_i$ extends $I_{i-1}$ by the 2-cycle $(a, i)$.

2762: \item If $b_i=\nn$, let $T_i=T_{i-1}$ and add the fixed point $(i)$

2763: to the involution $I_{i-1}$ to obtain $I_i$.

2764: \end{enumerate}

2765: The bijection sends $b\rightarrow(T,I)$ where $T=T_L$ and $I=I_L$.

2766: 

2767: \begin{remark} \label{rem:DDFeven} $b\in B_{\hdom\,}(1^L,\la)$ (that is,

2768: $b$ has no symbols $\nn$) if and only if $I$ has no fixed points

2769: \cite{DDF88}.

2770: \end{remark}

2771: 

2772: \begin{example} \label{ex:DDF} Let $b$ be as in Example

2773: \ref{ex:hwv}. The computation of the DDF bijection for the Motzkin

2774: tableau associated with $b$ is given in Figure \ref{fig:mDDF} along

2775: with data that will be useful later.

2776: \end{example}

2777: \begin{figure}[ht]

2778: \begin{equation*}

2779: \begin{array}{|c|l|c|l|c|l|l|} \hline

2780: \scalebox{.7}{\vphantom{\young(1,1)}}i&\hphantom{xx}

2781: T&I&\hphantom{xxxx}S&I w_0&\hphantom{xxxx}P&\hphantom{xxxx}Q

2782: \\ \hline\hline

2783: 1&\vphantom{\young(1,1)}\young(1) &&  &  & \young(1) & \young(\bl) \\[2mm]

2784: 2&\young(12)&&  &  & \young(12) & \young(\bl\bl) \\[2mm]

2785: 3&\young(12,3)&&  &  & \young(12,3) & \young(\bl\bl,\bl) \\[4mm]

2786: 4&\young(1,3)&(24)& \young(24) & \pmt{\hd\hb}{24} & \young(124,3) & \young(\bl\hd\hb,\bl) \\[4mm]

2787: 5&\young(1,3)&(24)(5)& \young(24,5) & \pmt{\he\hd\hb}{524} & \young(124,35) & \young(\bl\he\hb,\bl\hd) \\[4mm]

2788: 6&\young(16,3)&(24)(5)& \young(24,5) &\pmt{\he\hd\hb}{524} &  \young(124,35,6) & \young(\bl\bl\hb,\bl\he,\hd) \\[5.5mm]

2789: 7&\young(36)&(24)(5)(17)&

2790: \young(1247,5)&\pmt{\hg\he\hd\hb\ha}{15247} &\young(1247,35,6) &

2791: \vphantom{\young(1,2,3,4)} \young(\bl\bl\hb\ha,\hg\he,\hd) \\

2792: \hline\hline

2793: \end{array}

2794: \end{equation*}

2795: \caption{DDF computation} \label{fig:mDDF}

2796: \end{figure}

2797: 

2798: \subsection{The Burge correspondence}

2799: Column insertion defines a bijection between involutions $I$ on a

2800: set $A$ and standard tableaux $S$ with alphabet $A$. Specifically,

2801: write $I$ as a biword with upper word given by the set $A$ written

2802: in decreasing order, and with $I(a)$ below $a$ for all $a\in A$.

2803: Define the tableau $S$ by

2804: \begin{equation} \label{eq:Sdef}

2805: \bilet{S}{S} = \left( I \rightarrow \bilet{\nn}{\nn} \right),

2806: \end{equation}

2807: which is well-defined by Proposition \ref{pp:inv}. Define

2808: \begin{equation} \label{eq:mudef}

2809: \mu=\shape(S).

2810: \end{equation}

2811: 

2812: \begin{example} \label{ex:involution}

2813: Let $A=\{1,2,4,5,7\}$ and $I=(2,4)(5)(1,7)$ an involution on $A$.

2814: Then

2815: \begin{equation*}

2816: I= \begin{pmatrix}

2817:   7&5&4&2&1 \\

2818:   1&5&2&4&7

2819: \end{pmatrix} \qquad

2820: S = \young(1247,5)

2821: \end{equation*}

2822: \end{example}

2823: 

2824: \begin{theorem} \label{th:Burge} \cite{B74} $I$ has no fixed points if and only if

2825: $S$ has even row lengths.

2826: \end{theorem}

2827: 

2828: \begin{remark} \label{rem:evenBurge} It follows that if $b\in P_\se(1^L,\la)$ for

2829: $\se\in \{\cell\,,\hdom\,\}$ then $\mu\in \Par^\se$.

2830: \end{remark}

2831: 

2832: Burge gave a direct bijection to compute $S$ from $I$ when $I$ is

2833: fixed-point-free \cite{B74}. Although not necessary for our proof,

2834: we include the following straightforward extension of Burge's

2835: algorithm, which gives the above bijection $I\mapsto S$ for any

2836: involution $I$. First, for each transposition $(a,i)$ in $I$, create

2837: the ordered pair $(a,i)$ where $a<i$. For each fixed point $i$ in

2838: $I$, create the ordered pair $(i,i)$. Sort the ordered pairs

2839: according to second component, giving the sequence

2840: $(a_1,i_1),(a_2,i_2),\dotsc,(a_k,i_k)$, say. Let $S_0$ be the empty

2841: tableau. Given $S_{j-1}$, form $S_j$ as follows.

2842: \begin{enumerate}

2843: \item If $a_j<i_j$ then column insert $a_j$ into $S_{j-1}$,

2844: forming the tableau $S'_j$ with new cell $s$. Let $S_j$ be obtained

2845: from $S'_j$ by adjoining the letter $i_j$ at the bottom of the

2846: column just to the right of that of $s$.

2847: \item If $a_j=i_j$ then $a_j$ is the largest letter so far; adjoin

2848: it at the bottom of the first column of $S_{j-1}$ to obtain $S_j$.

2849: \end{enumerate}

2850: 

2851: \begin{example} With $I$ as in Example \ref{ex:involution}, we have

2852: the sequence of $(a_j,i_j)$ given by $((2,4),(5,5),(1,7))$. Then

2853: \Yboxdim{10pt}%

2854: \begin{equation*}

2855: S'_1=\young(2)\quad S_1=\young(24)\quad S_2=\young(24,5)\quad

2856: S'_3=\young(124,5)\quad S_3=\young(1247,5)=S

2857: \end{equation*}

2858: \end{example}

2859: 

2860: \subsection{Insertion of DDF data}

2861: \label{ss:DDFins} Suppose $b\in P_\se(1^L,\la)$ for

2862: $\se\in\{\cell\,,\hdom\,\}$ with DDF data $(T,I)$ and $I\mapsto S$

2863: under the Burge correspondence. Let $A$ be the set on which $I$ is

2864: an involution. For $1\le i\le L$ let $\hat{i}=L+1-i$ and let

2865: $\hat{A}$ be the image of $A\subset \{1,2,\dotsc,L\}$ under

2866: $i\mapsto \hat{i}$. Then let $Iw_0$ denote the bijection from

2867: $\hat{A}$ to $A$ given by $\hat{a}\mapsto I(a)$. Its biword is

2868: obtained from that of $I$ by replacing each upper letter $x$ by

2869: $\hat{x}$.

2870: 

2871: \begin{example} With $A$ and $I$ as in Example \ref{ex:involution}, the biword for the

2872: bijection $Iw_0$ is given in the last row of Figure \ref{fig:mDDF}.

2873: \end{example}

2874: 

2875: Define

2876: \begin{equation} \label{eq:DDFPQ}

2877:   \bilet{P}{Q} = \left( \bilet{T}{\nn} \leftarrow (I w_0) \right),

2878: \end{equation}

2879: that is, row insert the biword $Iw_0$ into $T$, and let $P$ and $Q$

2880: be the bumping and recording tableaux. By \eqref{eq:Sdef} and Remark

2881: \ref{rem:rowcol} we have

2882: \begin{equation} \label{eq:DDFP}

2883:   P = [T \cdot S].

2884: \end{equation}

2885: Let $c=c_L\dotsm c_2c_1$ be the Yamanouchi word corresponding to

2886: $P$. Define

2887: \begin{equation}

2888:   \tau=\wt(c).

2889: \end{equation}

2890: Define the DDF map by

2891: \begin{equation} \label{eq:DDF}

2892: \begin{split}

2893:   b\mapsto (c,T\otimes S)

2894: \end{split}

2895: \end{equation}

2896: using Littlewood-Richardson data in the form of Theorem \ref{th:LR}

2897: part (3).

2898: 

2899: \begin{prop} \label{pp:DDFbij} The DDF map \eqref{eq:DDF} gives a bijection

2900: \eqref{eq:bij}.

2901: \end{prop}

2902: \begin{proof} The map is well-defined by construction and

2903: Remark \ref{rem:evenBurge}, which shows that the tricky condition

2904: $\mu\in \Par^\se$ is satisfied. Since the DDF map was comprised of

2905: bijections, given $(c,T\otimes S)$ one may reconstruct $I$ and then

2906: $b$ by reversing the extended Burge and DDF bijections.

2907: \end{proof}

2908: 

2909: \begin{example} \label{ex:PTS} Consult the last row of the table in

2910: Figure \ref{fig:mDDF}, which computes $Iw_0$, $P$, and $Q$ from $T$

2911: and $I$ corresponding to $b$ of Example \ref{ex:hwv}. The associated

2912: Yamanouchi word $c$ is given in Example \ref{ex:hwv}. For this data,

2913: $\la=(2)$, $\mu=(4,1)$, and $\tau=(4,2,1)$.

2914: \end{example}

2915: 

2916: \subsection{Incremental DDF} \label{ss:incDDF}

2917: We describe the change in the output data of the DDF map when one

2918: more letter is added to the path $b$. Let $b=x\otimes b'$ so that

2919: $x=b_L\in B_\se$. Denote by $T',I',S',P',Q'$ the DDF data associated

2920: with $b'$. Let $\la',\mu',\tau'$ be the shapes of $T',S',P'$

2921: respectively.

2922: 

2923: The biletter $\bilet{a}{i}$ will be written $(a,i)$. Restricting

2924: \eqref{eq:DDFP} to the alphabet of letters less than $L$, by

2925: Proposition \ref{pp:Knuthres}

2926: \begin{equation} \label{eq:Pres}

2927:   P|_{<L} = [(T|_{<L}) (S|_{<L})].

2928: \end{equation}

2929: We will show that in all cases

2930: \begin{equation} \label{eq:Presprime}

2931:   P|_{<L} = P'.

2932: \end{equation}

2933: Then $P$ is obtained by adjoining $L$ to $P'$ at the cell $s$, say.

2934: To determine this cell $s$ it is enough to compute $Q$ from $Q'$.

2935: This is done in each case as follows.

2936: \begin{enumerate}

2937: \item Suppose $x=r$. Let $s'$ be the addable cell in the $r$-th

2938: row of $\la'$. Then $T$ is obtained by adjoining $L$ to $T'$ at

2939: $s'$, $I=I'$, and $S=S'$. We have $P|_{<L} = [ T' S] =[T'S']=P'$,

2940: proving \eqref{eq:Presprime}. $Q'$ records the insertion of

2941: $I'w_0=Iw_0$ into $T'$. $Q$ records the insertion of the same word

2942: into $T'$ except that there is a large number $L$ present at the

2943: cell $s'$. Then $Q$ is given by the internal insertion $I_{s'}(Q')$.

2944: \item Suppose $x=\bar{r}$. Then $T$ is obtained by the reverse row

2945: insertion on $T'$ at the cell $s'$, ejecting the letter $i$, say. In

2946: particular $[T']=[T i]$. Also $I=I'(i,L)$, and $S$ is obtained from

2947: $S'$ by the column insertion of $i$ followed by the adjoining of $L$

2948: at the end of the column just right of the end of the bumping path

2949: of the column insertion. It follows that $S|_{<L}=[i S']$. We have

2950: $P|_{<L}=[T (i S')]=[(Ti)S']=[T'S']=P'$, proving

2951: \eqref{eq:Presprime}. The biword $Iw_0$ is obtained from $I'w_0$ by

2952: adding biletters $(i,L^*)$ and $(L,i^*)$. The biletter $(i,L^*)$ is

2953: the first in $Iw_0$ to be inserted into $T$. This produces the

2954: tableau $T'=[Ti]$, so that $L^*$ goes in the cell $s'$ of $Q$.

2955: Thereafter it is the same as inserting $I' w_0$ into $T'$, except

2956: that in the middle there is a biletter $(L,i^*)$ containing the

2957: largest value $L$. It follows that $Q$ is obtained from $Q'$ by the

2958: external insertion of $i^*$ and putting the smallest letter $L^*$ at

2959: the cell $s'$ (which is on the inside of the skew shape of $Q'$).

2960: \item Suppose $x=\nn$. Then $T=T'$, $I$ is obtained from $I'$ by

2961: adding the fixed point $L$, and  $S$ is obtained from $S'$ by

2962: adjoining $L$ at the end of the first column, that is, $S=[L S']$.

2963: Therefore $P|_{<L}=[T S']=[T'S']=P'$, proving \eqref{eq:Presprime}.

2964: $Iw_0$ is obtained from $I'w_0$ by putting the biletter $(L,L^*)$ at

2965: the beginning. Since $T=T'$ the only difference between the

2966: computations of $Q$ and $Q'$ is that one first row inserts $L$ and

2967: records by $L^*$. Therefore $Q$ is obtained from $Q'$ by the

2968: external insertion of $L^*$.

2969: \end{enumerate}

2970: 

2971: \subsection{$DDF=VXR$ and the proof of $X=K$}

2972: \begin{prop} \label{pp:PQV=D} $(P,Q,T)=(\Pt,\Qt,\Tt)$.

2973: \end{prop}

2974: \begin{proof}

2975: By induction $(P',Q',T')=(\Pt',\Qt',\Tt')$ where the primed elements

2976: correspond to the shorter path $b'$ where $b=x\otimes b'$. By the

2977: analysis in sections \ref{ss:incDDF} and \ref{ss:incVXR}, the

2978: changes $(P',Q',T')\rightarrow (P,Q,T)$ and

2979: $(\Pt',\Qt',\Tt')\rightarrow (\Pt,\Qt,\Tt)$ are the same.

2980: \end{proof}

2981: 

2982: We now finish the proof of Theorem \ref{th:KX}. It suffices to prove

2983: Theorem \ref{th:VXR}. By Proposition \ref{pp:nu} it suffices to

2984: prove Theorem \ref{th:VXR} for the case $\nu=1^L$. The VXR map

2985: $\nu_{1^L}$ is bijective since it agrees with the DDF map, which is

2986: known to be a bijection (Propositions \ref{pp:PQV=D} and

2987: \ref{pp:DDFbij}). Finally, \eqref{eq:bijstat} holds by Proposition

2988: \ref{pp:stat}. This concludes the proof of Theorem \ref{th:KX}.

2989: 

2990: \section{Schensted and DDF according to Fomin and Roby}

2991: \label{sec:Roby} Although not needed for the proofs, we include an

2992: interpretation of the DDF map using Fomin's growth diagrams,

2993: extending Roby's interpretation of the original DDF bijection

2994: \cite{R95}. This is insightful, as one sees the local coenergy

2995: calculations mirrored in this computation of the DDF map.

2996: 

2997: \subsection{Fomin's growth diagrams}

2998: Fomin observed that various Schensted algorithms may be understood

2999: solely in terms of chains of partitions \cite{F94,F95}. Indeed, many

3000: of the properties of these algorithms are most clearly seen from

3001: this perspective.

3002: 

3003: An injective tableau $T$ of shape $\la/\mu$ in the alphabet

3004: $\{1,2,\dotsc,N\}$ may be interpreted as the sequence of partitions

3005: $\mu=\la^{(0)}\subset\la^{(1)}\subset\dotsm\subset \la^{(N)}=\la$

3006: where each $\la^{(i)}=\la^{(i-1)}$ if $i$ does not appear in $T$ and

3007: otherwise $\la^{(i)}$ is obtained from $\la^{(i-1)}$ by adding the

3008: cell of $T$ containing $i$. Or, the outer shape of each $T|_{\le i}$

3009: is $\la^{(i)}$, where $T|_{\le i}$ is the skew tableau given by the

3010: restriction of $T$ to the letters of value at most $i$. Note that

3011: the letter $i$ is associated with the transition from the shape

3012: $\la^{(i-1)}$ to $\la^{(i)}$.

3013: 

3014: Consider the Sagan-Stanley skew RS map $(T,U,w)\rightarrow (P,Q)$

3015: given in subsection \ref{ss:skew}. There is an associated

3016: \textbf{shape array} $(\la_a^b)$, a two-dimensional array of

3017: partitions with row index set of size $|B|+1$ and column index set

3018: of size $|A|+1$. Let us add a smallest letter $a_0$ to $A$ and $b_0$

3019: to $B$. The array elements $\la_a^b$ will be indexed using these

3020: enlarged alphabets. The letters of $A$ (resp. $B$) naturally label

3021: the transitions from one row (resp. column) to the next. In other

3022: words, the pair $(a,b)\in A\times B$ naturally labels the point in

3023: the middle of a two-by-two submatrix, whose lower right entry is

3024: $\la_a^b$. In particular a pair $(w(b),b)$ for $b\in B'$, is

3025: naturally associated with such a two-by-two submatrix.

3026: 

3027: Let $w|_{\le a}^{\le b}$ be the restriction of $w$ to the positions

3028: at most $a$ and values at most $b$. In other words, viewing $w$ as a

3029: set of ordered pairs $(w(b),b)$, select only those pairs $(a',b')$

3030: such that $a'\le a$ and $b'\le b$.

3031: 

3032: The entry $\la_a^b$ is defined to be the shape of the bumping

3033: tableau associated with the data $(T|_{\le a},U|_{\le b},w|_{\le

3034: a}^{\le b})$ under skew RS. Then the left and top borders of the

3035: shape array give $T$ and $U$ respectively, and the right and bottom

3036: borders give $P$ and $Q$ respectively.

3037: 

3038: Fomin's observation is that given $(T,U,w)$, one may recover the

3039: entire shape array using a local rule. Set the left and top borders

3040: of the shape array to be $T$ and $U$ respectively, with all other

3041: entries to be determined.

3042: 

3043: Consider a two-by-two subarray of partitions

3044: \begin{equation} \label{eq:2by2}

3045: \young(\mu\nu,\rho\sigma)

3046: \end{equation}

3047: indexed by $(a,b)$, in which all but $\sigma$ have already been

3048: determined. Let $s$ and $t$ be the cells (possibly empty) that must

3049: be adjoined to $\mu$ to obtain $\nu$ and $\rho$ respectively. Then

3050: $\sigma$ is defined by the following rules.

3051: \begin{enumerate}

3052: \item If $s=t$ is not empty then $\sigma$ is obtained from

3053: $\nu=\rho$ by adjoining a cell to the $(r+1)$-st row where $s=t$ is

3054: in the $r$-th row of $\nu=\rho$.

3055: \item If $s=t$ is empty and $w(b)=a$ then $\sigma$ is

3056: obtained from $\mu=\nu=\rho$ by adjoining a cell to the first row.

3057: \item Otherwise $\sigma=\nu\cup\rho$.

3058: \end{enumerate}

3059: Applying this local rule one may uniquely recover the entire array

3060: and read off $P$ and $Q$ at the end.

3061: 

3062: Or, starting with $P$ and $Q$ at the right and bottom borders with

3063: the rest of the shape array to be determined, one may apply the

3064: reverse of the above local rule. Suppose in \eqref{eq:2by2} (indexed

3065: by the pair $(a,b)$) one knows all of the partitions but $\mu$. Then

3066: $\mu$ can be recovered, as well as whether or not $w(b)=a$.

3067: 

3068: Let $s'$ and $t'$ be the cells (possibly empty) which must be

3069: removed from $\sigma$ to obtain $\rho$ and $\nu$ respectively.

3070: \begin{enumerate}

3071: \item If $s'=t'$ is nonempty and $s'=t'$ is in the first row of

3072: $\sigma$, then add the pair $(a,b)$ to $w$ (that is, extend the

3073: bijection $w$ by declaring that $w(b)=a$) and set $\mu$ equal to

3074: $\nu=\rho$. We call such a two-by-two array \textbf{special}.

3075: \item If $s'=t'$ is nonempty and $s'=t'$ is in the $(r+1)$-st row of

3076: $\sigma$ with $r\ge1$, then $\mu$ is obtained from $\nu=\rho$ by

3077: removing a cell in the $r$-th row.

3078: \item Otherwise $\mu=\nu\cap\rho$.

3079: \end{enumerate}

3080: Using the reverse local rule the rest of the shape array may be

3081: determined, yielding the right and top borders $T$ and $U$ and the

3082: bijection $w$.

3083: 

3084: \begin{theorem} \cite{SS90} \label{th:RSinv} The skew RS map

3085: defined in subsection \ref{ss:skew} has the involution property,

3086: that is, $(T,U,w)\mapsto (P,Q)$ if and only if $(U,T,w^{-1})\mapsto

3087: (Q,P)$.

3088: \end{theorem}

3089: \begin{proof} Fomin's local rule is invariant under the transposition of

3090: the two-by-two matrix \eqref{eq:2by2}.

3091: \end{proof}

3092: 

3093: \subsection{Roby on DDF}

3094: Roby \cite{R95} observed that a skew extension \cite{DS95} of the

3095: DDF bijection could be computed using Fomin's local rule. We only

3096: require the nonskew case.

3097: 

3098: Let $\lad$ be an oscillating tableau of shape $\la$ and length $L$

3099: and $(T,I)$ its image under the DDF bijection. The pair $(T,I)$ may

3100: be recovered as follows. In the skew RS algorithm, set both

3101: alphabets $A$ and $B$ to be $\{1,2,\dotsc,L\}$. Start with an

3102: unfilled $(L+1)\times (L+1)$ shape array. Put the oscillating

3103: tableau on the \textit{antidiagonal}, starting from the upper right

3104: and proceeding to the lower left. Use the reverse of Fomin's local

3105: rule to fill in the half of the diagram above the antidiagonal. To

3106: start the process one must be able to compute $\mu$ in the reverse

3107: local rule for two-by-two subarrays \eqref{eq:2by2} such that $\nu$

3108: and $\rho$ are consecutive shapes in the oscillating tableau,

3109: without knowledge of $\sigma$. But consecutive shapes in oscillating

3110: tableaux are never equal, so one does not need $\sigma$ to determine

3111: $\mu$. Filling in the triangle of the shape array above the

3112: antidiagonal, the left border recovers the injective tableau $T$.

3113: The top border consists of empty partitions since the oscillating

3114: tableau starts with the empty partition. One also obtains a set of

3115: special positions $(a,i)$ above the antidiagonal, which yield the

3116: permutation data: $I$ is the product of the transpositions of the

3117: form $(a,\hat{i})=(a,L+1-i)$.

3118: 

3119: This is readily extended for Motzkin tableaux. In this case there is

3120: ambiguity on how to compute the reverse local rule when consecutive

3121: shapes in the Motzkin tableau are equal. We resolve this ambiguity

3122: by declaring that if $\la^{(i-1)}=\la^{(i)}$ then the corresponding

3123: position $(i,\hat{i})$ is special. In other words, for a two-by-two

3124: subarray \eqref{eq:2by2} on the antidiagonal, if $\nu=\rho$ we

3125: declare that $\sigma$ is obtained from $\nu=\rho$ by adding a cell

3126: to the first row, and set $\mu$ equal to $\nu=\rho$. Then there is a

3127: unique way to fill in the triangular part of the shape array above

3128: the antidiagonal. Again $T$ is the left border and the top border

3129: consists of empty partitions. As before, the special positions

3130: strictly above the antidiagonal give the transpositions in the

3131: desired involution $I$, while the fixed points $i$ of $I$ are given

3132: by the special positions $(i,\hat{i})$ on the antidiagonal.

3133: 

3134: See Figure \ref{fig:Mshape} for the shape array associated with the

3135: Motzkin tableau given in Figure \ref{fig:mDDF}, in which the special

3136: positions are indicated by the symbol $\otimes$.

3137: 

3138: \begin{figure}

3139: \Yboxdim{3pt}\Yinterspace{1pt}

3140: \begin{equation*}

3141: \begin{array}{|c||ccccccccccccccc|} \hline

3142: \raisebox{5pt}{\vphantom{1}} && \hg && \hf && \he && \hd && \hc && \hb && \ha &\\

3143:  \hline \hline

3144:  & \dt && \dt && \dt && \dt && \dt && \dt && \dt && \dt \\ 1 &&\otimes&&&&&&&&&&&&& \\

3145:  & \dt &&  \yng(1) &&\yng(1) &&\yng(1) &&\yng(1) &&\yng(1) &&\yng(1)

3146: &&\yng(1) \\ 2 &&&&&&&&\otimes&&&&&&& \\

3147:  & \dt && \yng(1) &&\yng(1) &&\yng(1) && \yng(2) &&\yng(2)

3148: &&\yng(2) &&\yng(2) \\ 3 &&&&&&&&&&&&&&&\\

3149:  & \yng(1) && \yng(1,1) &&\yng(1,1) &&\yng(1,1) &&\yng(2,1) &&\yng(2,1)

3150: &&\yng(2,1) &&\yng(2,1) \\ 4 &&&&&&&&&&&&\otimes &&&\\

3151:  & \yng(1) && \yng(1,1) &&\yng(1,1) &&\yng(1,1) &&\yng(2,1) &&\yng(2,1)

3152: &&\yng(3,1) &&\yng(3,1) \\ 5 &&&&&&\otimes &&&&&&&&&\\

3153:  & \yng(1) && \yng(1,1) &&\yng(1,1) &&\yng(2,1) &&\yng(2,2) &&\yng(2,2) &&\yng(3,2) &&\yng(3,2)

3154:  \\ 6 &&&&&&&&&&&&&&&\\

3155:  & \yng(2) && \yng(2,1) &&\yng(2,1) &&\yng(2,2) &&\yng(2,2,1) &&\yng(2,2,1)

3156: &&\yng(3,2,1) &&\yng(3,2,1) \\ 7 &&&&&&&&&&&&&&\otimes&\\

3157:  & \yng(2) && \yng(2,1) &&\yng(2,1) &&\yng(2,2) &&\yng(2,2,1) &&\yng(2,2,1)

3158: &&\yng(3,2,1) &&\yng(4,2,1) \\ \hline

3159: \end{array}

3160: \end{equation*}

3161: \Yboxdim{13pt}\Yinterspace{1pt} \caption{Shape array for a Motzkin

3162: path} \label{fig:Mshape}

3163: \end{figure}

3164: 

3165: \subsection{The lower triangle of the shape array}

3166: For our application the lower triangle of the shape array is

3167: crucial. Starting with the Motzkin tableau of shape $\la$, the upper

3168: triangle and special positions (and therefore $T$ and $I$) are

3169: determined as before. To fill out the positions in the lower

3170: triangle below the antidiagonal it is necessary to know which of

3171: those positions are special. We define the special positions below

3172: the antidiagonal to be those whose mirror images across the

3173: antidiagonal, are special. This is equivalent to defining the total

3174: set of special positions to be given by the bijection $Iw_0$ defined

3175: in section \ref{ss:DDFins}.

3176: 

3177: The triangle below the antidiagonal may be filled using Fomin's

3178: local rule. At the end, let $P$ and $Q$ be the right and bottom

3179: borders. $P$ is a standard tableau of some shape, say $\tau$, and

3180: $Q$ is an injective skew tableau of shape $\tau/\la$. Then the skew

3181: RS algorithm gives

3182: \begin{equation} \label{eq:DDFSS}

3183: (T,\nn,Iw_0)\mapsto (P,Q).

3184: \end{equation}

3185: This coincides with the $P$ tableau defined in \eqref{eq:DDFP}.

3186: 

3187: \begin{example} \label{ex:RobyDDF} Figure \ref{fig:mDDF} contains

3188: the data $I w_0,P,Q$ for each of the Motzkin tableaux associated

3189: with right factors $b_i\dotsm b_2 b_1$ of $b$ in Example

3190: \ref{ex:hwv}.

3191: \end{example}

3192: 

3193: \subsection{Ghost of coenergy in DDF}

3194: Let $b\in P_\se(1^L,\la)$ and $\dck\otimes c$ be as in

3195: \eqref{eq:dcdef}. Remarkably, the computation of the coenergy

3196: $\Db(\dck\otimes c)$ in section \ref{ss:grading} can be seen clearly

3197: in the Fomin shape array of the DDF map applied to $b$. Compare the

3198: underlined positions in the computation of $R_+$ given in Example

3199: \ref{ex:dckc}, with the special positions in the shape array for $b$

3200: in Figure \ref{fig:Mshape} that are on or southeast of the main

3201: antidiagonal. They coincide if the shape array is rotated 45 degrees

3202: clockwise.

3203: 

3204: \section{The missing cases}

3205: \label{sec:end} Our methods don't apply when $\se=\vdom\,$. Although

3206: the paths for types $\se=\vdom$ and $\se=\hdom$ for tensor powers of

3207: $B_\se$ are in an easy bijection given by ``transposing shapes", the

3208: coenergy functions are quite different in nature. This case requires

3209: a new idea.

3210: 

3211: We also expect that the VXR map will succeed for any tensor product

3212: of KR crystals for types $\se\in\onetwoset$. To do this one must

3213: construct the KR modules $W^{(r)}_s$ and crystals $B^{r,s}$ in

3214: general and show, as conjectured in \cite{OSS03} that the virtual

3215: crystal methods work for these crystals. For the case of tensor

3216: products of KR crystals of the form $B^{r,1}$ the virtual crystals

3217: have already been constructed \cite{OSS03}. Our methods should work

3218: in this case, using intrinsic energy rather than coenergy and

3219: compatible splitting maps that remove a box from a column. Note that

3220: this case is not the transpose of the case proven in Theorem

3221: \ref{th:KX}; the asymmetry is seen in the classical decomposition of

3222: the KR modules (the $U'_q(C_n^{(1)})$-crystal $B^{r,1}$ is

3223: isomorphic to $B(\omega_r)$ as $U_q(C_n)$-crystals) and in the $K$

3224: formula in the case $\hdom\,$ where the horizontal dominoes do not

3225: get transposed.

3226: 

3227: \begin{thebibliography}{xxx}

3228: 

3229: \bibitem{AK97} T. Akasaka and M. Kashiwara,

3230: Finite-dimensional representations of quantum affine algebras, Publ.

3231: Res. Inst. Math. Sci. \textbf{33} (1997) 839--867.

3232: 

3233: \bibitem{B74} W. Burge, Four correspondences between graphs and

3234: generalized Young tableaux, J. Combinatorial Theory Ser. A

3235: \textbf{17} (1974), 12--30.

3236: 

3237: \bibitem{Ch01} V. Chari, On the fermionic formula and the

3238: Kirillov-Reshetikhin conjecture, Internat. Math. Res. Notices

3239: 2001, 629--654.

3240: 

3241: \bibitem{CP95} V. Chari and A. Pressley, Minimal affinizations of

3242: representations of quantum groups: the nonsimply-laced case. Lett.

3243: Math. Phys. \textbf{35} (1995) 99--114.

3244: 

3245: \bibitem{CK00} V. Chari and M. Kleber, Symmetric functions and

3246: representations of quantum affine algebras, Recent developments in

3247: infinite-dimensional Lie algebras and conformal field theory

3248: (Charlottesville, VA, 2000), 27--45, Contemp. Math., \textbf{297},

3249: Amer. Math. Soc., Providence, RI, 2002.

3250: 

3251: \bibitem{DDF88} M.-P.~Delest, S.~Dulucq and L.~Favreau,

3252: An analogue to Robinson-Schensted correspondence for oscillating

3253: tableaux, S\'eminaire Lotharingien de Combinatoire, B20b (1988).

3254: 

3255: \bibitem{DS95} S. Dulucq and B. Sagan, La correspondance

3256: de Robinson-Schensted pour les tableaux oscillants gauches, Formal

3257: power series and algebraic combinatorics (Montreal, PQ, 1992).

3258: Discrete Math. \textbf{139} (1995), no. 1-3, 129--142.

3259: 

3260: \bibitem{F94} S. Fomin, Duality of graded graphs, J. Algebraic Combin.

3261: \textbf{3} (1994),  no. 4, 357--404.

3262: 

3263: \bibitem{F95} S. Fomin, Schensted algorithms for dual graded graphs, J.

3264: Algebraic Combin. \textbf{4} (1995),  no. 1, 5--45.

3265: 

3266: \bibitem{FG93} S. Fomin and C. Greene, A Littlewood-Richardson miscellany,

3267: European J. Combin. \textbf{14} (1993), no. 3, 191--212.

3268: 

3269: \bibitem{Ful97} W. Fulton, Young tableaux, with applications to representation

3270: theory and geometry. London Mathematical Society Student Texts,

3271: \textbf{35}, Cambridge University Press, Cambridge, 1997.

3272: 

3273: \bibitem{Her05} D. Hernandez, The Kirillov-Reshetikhin conjecture and solutions of

3274: T-systems, arXiv:math.QA/0501202.

3275: 

3276: \bibitem{HKOT00} G.~Hatayama, A.~Kuniba, M.~Okado,

3277: T.~Takagi, Combinatorial $R$ matrices for a family of crystals:

3278: $C\sb n\sp {(1)}$ and $A\sb {2n-1}\sp {(2)}$ cases, Physical

3279: combinatorics (Kyoto, 1999), 105--139, Progr. Math., 191, Birkhäuser

3280: Boston, Boston, MA, 2000.

3281: 

3282: \bibitem{HKOT02} G.~Hatayama, A.~Kuniba, M.~Okado,

3283: T.~Takagi, Combinatorial $R$ matrices for a family of crystals:

3284: $B\sp {(1)}\sb n,D\sp {(1)}\sb n,A\sp {(2)}\sb {2n}$, and $D\sp

3285: {(2)}\sb {n+1}$ cases, J. Algebra 247 (2002) 577--615.

3286: 

3287: \bibitem{HKOTT01} G.~Hatayama, A.~Kuniba, M.~Okado, T.~Takagi, and Z.~Tsuboi,

3288: Paths, Crystals and Fermionic Formulae, arXiv:math.QA/0102113.

3289: 

3290: \bibitem{HKOTY99} G.~Hatayama, A.~Kuniba, M.~Okado, T.~Takagi, and

3291: Y.~Yamada, Remarks on fermionic formula, in Recent developments in

3292: quantum affine algebras and related topics (Raleigh, NC, 1998),

3293: 243--291, Contemp. Math., \textbf{248}, Amer. Math. Soc.,

3294: Providence, RI, 1999.

3295: 

3296: \bibitem{JMO00} N.~Jing, K.~C.~Misra, and M.~Okado,

3297: $q$-wedge modules for quantized enveloping algebras of classical

3298: type, J. Algebra \textbf{230} (2000), no. 2, 518--539.

3299: 

3300: \bibitem{Kac90} V. Kac, Infinite-dimensional Lie algebras, third

3301: edition, Cambridge University Press, Cambridge, 1990.

3302: 

3303: \bibitem{Kas95} M. Kashiwara, On crystal bases. Representations of groups (Banff, AB, 1994),

3304: 155--197, CMS Conf. Proc., \textbf{16}, Amer. Math. Soc.,

3305: Providence, RI, 1995.

3306: 

3307: \bibitem{Kas03} M. Kashiwara,

3308: Level zero fundamental representations over quantized affine

3309: algebras and Demazure modules, arXiv:math.QA/0309142.

3310: 

3311: \bibitem{KKM94} S.~J.~Kang, M.~Kashiwara, and K.~C.~Misra,

3312: Crystal bases of Verma modules for quantum affine Lie algebras.

3313: Compositio Math. \textbf{92} (1994),  no. 3, 299--325.

3314: 

3315: \bibitem{KKR88} S.~.V.~Kerov, A.~N.~Kirillov, and

3316: N.~Yu.~Reshetikhin, Combinatorics, the Bethe ansatz and

3317: representations of the symmetric group, J. Soviet Math.

3318: \textbf{41} (1988), no. 2, 916--924.

3319: 

3320: \bibitem{Kl02} M. Kleber, Embeddings of Schur functions into types

3321: $B/C/D$. J. Algebra \textbf{247} (2002), no. 2, 452--466.

3322: 

3323: \bibitem{KMN92} S.-J.~Kang, M.~Kashiwara, K.~C.~Misra, T.~Miwa, T.~Nakashima, and

3324: A.~Nakayashiki, Affine crystals and vertex models, Infinite

3325: analysis, Part A, B (Kyoto, 1991), 449--484, Adv. Ser. Math.

3326: Phys., \textbf{16}, World Sci. Publishing, River Edge, NJ, 1992.

3327: 

3328: \bibitem{KMN92a} S.-J.~Kang, M.~Kashiwara, K.~C.~Misra, T.~Miwa, T.~Nakashima, and

3329: A.~Nakayashiki, Perfect crystals of quantum affine Lie algebras.

3330: Duke Math. J.  \textbf{68} (1992),  no. 3, 499--607.

3331: 

3332: \bibitem{KN94} M. Kashiwara and T. Nakashima,

3333: Crystal graphs for representations of the $q$-analogue of classical

3334: Lie algebras, J. Algebra \textbf{165} (1994) 295--345.

3335: 

3336: \bibitem{Kn70} D. E. Knuth, Permutations, matrices, and generalized Young

3337: tableaux, Pacific J. Math. \textbf{34} (1970) 709--727.

3338: 

3339: \bibitem{Ko99} Y. Koga, Level one perfect crystals for $B_n^{(1)}$, $C_n^{(1)}$,

3340: and $D_n^{(1)}$, J. Algebra \textbf{217} (1999), no. 1, 312--334.

3341: 

3342: \bibitem{KR88} A.~N.~Kirillov, and

3343: N.~Yu.~Reshetikhin, The Bethe ansatz and the combinatorics of

3344: Young tableaux, J. Soviet Math. \textbf{41} (1988), no. 2,

3345: 925--955.

3346: 

3347: \bibitem{KR90} A. N. Kirillov and N.~Yu.~Reshetikhin, Representations of Yangians

3348: and multiplicities of the inclusion of the irreducible components

3349: of the tensor product of representations of simple Lie algebras,

3350: J. Soviet Math. \textbf{52} (1990),  no. 3, 3156--3164.

3351: 

3352: \bibitem{KSS02} A.~N.~Kirillov, A. Schilling, and M. Shimozono,

3353: A bijection between Littlewood-Richardson tableaux and rigged

3354: configurations, Selecta Math. (N.S.) \textbf{8} (2002),  no. 1,

3355: 67--135.

3356: 

3357: \bibitem{L91} A. Lascoux, Cyclic permutations on words, tableaux

3358: and harmonic polynomials.  Proceedings of the Hyderabad Conference

3359: on Algebraic Groups (Hyderabad, 1989), 323--347, Manoj Prakashan,

3360: Madras, 1991.

3361: 

3362: \bibitem{Le04} C. Lecouvey, A duality between $q$-multiplicities in tensor

3363: products and $q$-multiplicities of weights for the root systems $B$,

3364: $C$ or $D$, preprint, arXiv:math.RT/0407522.

3365: 

3366: \bibitem{Le04a} C. Lecouvey, Branching rules, Kostka-Foulkes polynomials and

3367: $q$-multiplicities in tensor product for the root systems $B_n$,

3368: $C_n$ and $D_{n}$, preprint arXiv:math.CO/0412548.

3369: 

3370: \bibitem{LR34} D. E. Littlewood and A. R. Richardson, Group

3371: characters and algebra, Philos. Trans. R. Soc., A, \textbf{233}

3372: (1934) 99-141.

3373: 

3374: \bibitem{LS81} A. Lascoux and M. P. Sch\"utzenberger, Le Mono\"ide

3375: Plaxique, Noncommutative structures in algebra and geometric

3376: combinatorics (Naples, 1978), pp. 129--156, Quad. Ricerca Sci.,

3377: \textbf{109}, CNR, Rome, 1981.

3378: 

3379: \bibitem{Lu83} G. Lusztig, Singularities, character formulas, and a

3380: $q$-analog of weight multiplicities, Analysis and topology on

3381: singular spaces, II, III (Luminy, 1981), 208--229, Ast\'erisque,

3382: 101-102, Soc. Math. France, Paris, 1983.

3383: 

3384: \bibitem{Mac95} I. G. Macdonald,

3385: Symmetric functions and Hall polynomials, Second edition, With

3386: contributions by A. Zelevinsky, Oxford Mathematical Monographs,

3387: Oxford Science Publications, The Clarendon Press, Oxford

3388: University Press, New York, 1995.

3389: 

3390: \bibitem{NY97} A. Nakayashiki and Y. Yamada, Kostka polynomials and

3391: energy functions in solvable lattice models, Selecta Math. (N.S.)

3392: \textbf{3} (1997), no. 4, 547--599.

3393: 

3394: \bibitem{OSS03} M. Okado, A. Schilling, and M. Shimozono, Virtual

3395: crystals and fermionic formulas of type $D^{(2)}_{n+1}$,

3396: $A^{(2)}_{2n}$, and $C^{(1)}_n$, Represent. Theory  \textbf{7}  (2003),

3397: 101--163 (electronic).

3398: 

3399: \bibitem{OSS03a} M. Okado, A. Schilling, and M. Shimozono,

3400: A crystal to rigged configuration bijection for nonexceptional

3401: affine algebras, ``Algebraic Combinatorics and Quantum Groups",

3402: Edited by N. Jing, World Scientific (2003), 85--124.

3403: 

3404: \bibitem{OSS03b} M. Okado, A. Schilling, and M. Shimozono,

3405: Virtual crystals and Kleber's algorithm, Commun. Math. Phys.

3406: \textbf{238} (2003) 187-209.

3407: 

3408: \bibitem{R95} T. Roby, The connection between the Robinson-Schensted

3409: correspondence for skew oscillating tableaux and graded graphs,

3410: Formal power series and algebraic combinatorics (Montreal, PQ,

3411: 1992), Discrete Math. \textbf{139}  (1995),  no. 1-3, 481--485.

3412: 

3413: \bibitem{Sc61} C. Schensted, Longest increasing and decreasing

3414: subsequences, Canad. J. Math. \textbf{13} (1961) 179--191.

3415: 

3416: \bibitem{Sci04} A. Schilling,

3417: A bijection between type $D_n^{(1)}$ crystals and rigged

3418: configurations, preprint arXiv:math.QA/0406248.

3419: 

3420: \bibitem{ScS04} A. Schilling and M. Shimozono, $X=M$ for

3421: symmetric powers, preprint arXiv:math.QA/0412376.

3422: 

3423: \bibitem{ScWa99} A. Schilling and S. Ole Warnaar,

3424: Inhomogeneous lattice paths, generalized Kostka polynomials and

3425: $A_{n-1}$ supernomials, Comm. Math. Phys. \textbf{202} (1999)

3426: 359--401.

3427: 

3428: \bibitem{S02} M. Shimozono, Affine type A crystal structure

3429: on tensor products of rectangles, Demazure characters, and nilpotent

3430: varieties.  J. Algebraic Combin. \textbf{15} (2002) 151--187.

3431: 

3432: \bibitem{S02a} M. Shimozono, Multi-atoms and monotonicity of

3433: generalized Kostka polynomials, European J. Combin. \textbf{22}

3434: (2001) 395--414.

3435: 

3436: \bibitem{SS90} B. Sagan and R. P. Stanley, Robinson-Schensted algorithms

3437: for skew tableaux, J. Combin. Theory Ser. A \textbf{55}  (1990),

3438: no. 2, 161--193.

3439: 

3440: \bibitem{St01} J. Stembridge, Multiplicity-free products of Schur

3441: functions, Ann. Comb. \textbf{5} (2001), 113--121.

3442: 

3443: \bibitem{SZ04} M. Shimozono and M. Zabrocki, Deformed universal characters

3444: for classical and affine algebras, preprint arXiv:math.CO/0404288.

3445: 

3446: \bibitem{Y98} S. Yamane, Perfect crystals of $U_q(G_2^{(1)})$,

3447: J. Algebra \textbf{210} (1998),  no. 2, 440--486.

3448: 

3449: \end{thebibliography}

3450: 

3451: \end{document}

3452: