1: 

2: \documentclass{article}

3: 

4: \setlength{\textwidth}{6.5in}

5: \setlength{\textheight}{9in}

6: \setlength{\evensidemargin}{0in}

7: \setlength{\oddsidemargin}{0in}

8: \setlength{\topmargin}{-.5in}

9: 

10: \usepackage{amsfonts,amsmath}

11: \usepackage{amssymb,latexsym}

12: \usepackage[dvips]{epsfig}

13: 

14: \title{\bf A Categorification of the Temperley-Lieb Algebra and 

15: Schur Quotients of $U(\mathfrak{sl}_2)$ via Projective and Zuckerman 

16: Functors\footnotetext{Published in 

17: Selecta Mathematica, New ser. 5 (1999), 199--241}}

18: \author{Joseph Bernstein, Igor Frenkel and Mikhail Khovanov}

19: \date{}

20: \newtheorem{prop}{Proposition}

21: \newtheorem{theorem}{Theorem}

22: \newtheorem{lemma}{Lemma}

23: \newtheorem{corollary}{Corollary}

24: \newtheorem{definition}{Definition}

25: \newtheorem{conjecture}{Conjecture}

26: \newcommand{\oplusop}[1]{{\mathop{\oplus}\limits_{#1}}}

27: 

28: 

29: \begin{document}

30: \maketitle

31: \baselineskip 12pt

32: 

33: \tableofcontents

34: 

35: 

36: \bigbreak

37: 

38: %\NoBlackBoxes

39: %\NoRunningHeads

40: %\NoPageNumbers

41: \def\C{\mathbb C}

42: \def\R{\mathbb R}

43: \def\N{\mathbb N}

44: \def\Z{\mathbb Z}

45: \def\Q{\mathbb Q}

46: \def\F{\mathbb F}

47: \def\Zq{\Z[q,q^{-1}]}

48: \def\l{\lbrace}

49: \def\r{\rbrace}

50: \def\o{\otimes}

51: \def\D{\Delta}

52: \def\O{\mathcal{O}}

53: \def\Ok{\mathcal{O}(k,n-k)}

54: %\def\Gr{\textrm{Gr}_n^k}

55: \def\slt{\mathfrak{sl}_2}

56: \def\U{U_q(\mathfrak{sl}_2)}

57: \def\Ug{U_q(\mathfrak g)}

58: \def\fsl{\mathfrak{sl}_n}

59: \def\gl{\mathfrak{gl}_n}

60: \def\td{\textrm{dim}}

61: \def\g{\mathfrak{g}}

62: \def\sln{\mathfrak{sl}_n}

63: \def\p{\mathfrak{p}}

64: \def\CU{\mathcal{U}}

65: \def\opsi{\overline{\psi}}

66: \def\ophi{\overline{\phi}}

67: 

68: \def\lra{\longrightarrow}

69: 

70: \def\E{\mathcal E} 

71: \def\F{\mathcal F}

72: 

73: \def\dU{\dot{U}(\mathfrak{sl}_2)}

74: \def\dUq{\dot{U}_q(\mathfrak{sl}_2)}

75: \def\dB{\dot{\mathbb B}}

76: 

77: \def\UL{\mathbf U}

78: \def\UZ{\mathbf U_{\Z}}

79: 

80: \def\M{\mathcal M} 

81: \def\Mi#1{\M_{Zf}(#1 )}

82: 

83: \def\binom#1#2{\left( \begin{array}{c} #1 \\ #2 \end{array}\right)}

84: 

85: \def\vn{V_1^{\otimes n}}

86: \def\vnQ{V_{1,\Q}^{\otimes n}}

87: \def\e{\epsilon}

88: 

89: \def\ch{\mathfrak h^*}

90: 

91: \def\OO#1{\mathcal O_{#1}}

92: \def\Og#1{\mathcal O(#1)}

93: \def\Ogc#1#2{\mathcal O_{#1}(#2)}

94: \def\Ogp#1#2{\mathcal O(#1,#2)}

95: \def\Ogcp#1#2#3{\mathcal O_{#1}(#2,#3)}

96: 

97: \def\Lo{L_n}

98: \def\Ld{L_n^{\ast}}

99: 

100: \def\Hom{\textrm{Hom}}

101: \def\mc{\mathcal} 

102: \def\mf{\mathfrak}

103: 

104: \def\gvm{$\mf{p}$-Verma module} 

105: 

106: \def\yesnocases#1#2#3#4{\left\{

107: \begin{array}{ll} #1 & \mbox{#2}\\ #3 & \mbox{#4}

108: \end{array} \right. }

109: 

110: \def\drawing#1{

111: \begin{center} \epsfig{file=#1} \end{center} }

112: 

113: 

114: \def\Ti{T_i}

115: \def\Tui{T^i} 

116: 

117: \def\Mp{\mc{M}(\mf{p})}

118: \def\Ind{\textrm{Ind}}

119: \def\Cfun{\mc{C}}

120: 

121: 

122: 

123: 

124: 

125: \section{Introduction}

126: 

127: One of the most important developments in the theory of quantum groups has 

128: been  the discovery of canonical bases which have remarkable integrality and 

129: positivity properties ([Lu]). Using these bases one expects to formulate 

130: the representation theory of quantum groups entirely over natural numbers. 

131: In the case of the simplest quantum group $\U,$ such a formulation can be 

132: achieved via the Penrose-Kauffman graphical calculus (see [FK]). 

133: The positive integral structure of representation 

134: theory suggests that it is itself a Grothendieck ring of a certain tensor 

135: 2-category. A strong support of this anticipation comes from 

136: identifying  coefficients of the transition matrix between the canonical 

137: and elementary bases in the $n$-th tensor power of the 

138: two-dimensional fundamental representation $V_1$ of $\U$ with the 

139: Kazhdan-Lusztig polynomials associated to $\gl$ for the maximal 

140: parabolic subalgebras ([FKK]). 

141: 

142: In this paper we will take one more step towards constructing a 

143: tensor 2-category with the Grothendieck ring isomorphic to the

144: representation category for $\U.$ The construction of tensor categories or 

145: 2-categories with given Grothendieck groups will be referred to as 

146: ``categorification''. We obtain a categorification of the $U(\mf{sl}_2)$ 

147: action in $\vn$ and the action of its commutant, the Temperley-Lieb algebra, 

148: using projective and Zuckerman functors between certain representation 

149: categories of $\gl.$ 

150: We extend this categorification to the comultiplication of $U(\mf{sl}_2).$ 

151: Our results are strongly motivated by the papers [BLM],[GrL] and 

152: [Gr], where the authors use the geometric rather than algebraic approach. 

153: In the geometric setting the categorification can be obtained via the 

154: categories of perverse sheaves. It is  expected that the algebraic 

155: and geometric languages will be equivalent, however, at the present 

156: moment the dictionary is still incomplete and the majority of our 

157: results do not allow a direct translation into the geometric 

158: language. Such a translation would require a nice geometric 

159: realization, so far unknown, 

160:  of singular blocks of the  highest weight categories

161: for $\gl$ and projective functors between these blocks. 

162: 

163: The main results of the paper are contained in Sections 3 and 4 and 

164: provide two categorifications of $U(\mf{sl}_2)$ and Temperley-Lieb algebra 

165: actions. Preliminary facts and 

166: definitions are collected in Section~\ref{sect-sl2}. 

167: The basis constituents of 

168: our construction are singular and parabolic categories of highest weight 

169: modules together with projective and Zuckerman functors acting on 

170: these categories. Projective functors in categories of $\gl$ modules, 

171: defined as direct summands of functors of tensoring with a 

172: finite-dimensional $\gl$-modules [J],[Zu], are extensively used in 

173: representation theory, (see  [BG] and [KV]). Being exact, projective 

174: functors induce linear maps in Grothendieck groups of categories 

175: of representations. Zuckerman functors are defined for any parabolic 

176: subalgebra $\mf{p}$ of $\gl$ by taking the maximal $U(\mf{p})$-locally 

177: finite submodule [KV]. Derived functors of Zuckerman functors are 

178: exact and also descend to Grothendieck groups. 

179: An important property of Zuckerman functors, namely their commutativity 

180: with the projective functors, yields in both categorifications what we 

181: consider the Schur-Weyl duality for $\U$ and the Temperley-Lieb algebra 

182: actions. 

183: 

184: In Section~\ref{sect-sing} we construct a categorification via 

185: singular blocks of the category $\O(\gl)$ of highest weight $\gl$-modules. 

186: More specifically, we realize $V_1^{\o n}$ as a Grothendieck group of 

187: the category 

188: \begin{equation} 

189: \O_n= \oplus_{k=0}^n \O_{k,n-k}, 

190: \end{equation} 

191: where $\O_{k,n-k}$ is a singular block of $\O(\gl)$ corresponding 

192: to the subgroup $\mathbb S_k\times \mathbb S_{n-k}$ of $\mathbb S_n.$ 

193: The simplest projective functors constructed by means of tensoring 

194: with the fundamental representation of $\gl$ and its dual descend on the 

195: Grothendieck group level to the action of generators $E$ and $F$ 

196: of $\mf{sl}_2$ (Section~\ref{functors-EF}). Various equalities between 

197: products of $E$ and $F$ result from functor isomorphisms 

198: (Section~\ref{sect-divp}). Moreover, we show that indecomposable 

199: projective functors in $\O_n$ correspond to elements of Lusztig 

200: canonical basis in the modified universal enveloping algebra 

201: $\dU$ (Section~\ref{canbas-indecomp}). Construction of the 

202: comultiplication for 

203: $U(\mf{sl}_2)$ requires studying the relation between categories 

204: $\O(\gl)\times \O(\mf{gl}_m)$ and $\O(\mf{gl}_{n+m}),$ given by 

205: the induction functor from the maximal parabolic subalgebra of 

206: $\mf{gl}_{n+m}$ that contains $\gl\oplus \mf{gl}_m.$ In particular, 

207: a categorification of the comultiplication formulas 

208: $\Delta E= E\o 1+ 1\o E$ and $\Delta F= F\o 1+ 1\o F$ for generators $E$ 

209: and $F$ is expressed by short exact sequences that employ certain properties 

210: of the induction functor (Section~\ref{sect-comult}). 

211: 

212: To categorify the action of the Temperley-Lieb algebra on $V_1^{\o n},$ 

213: we use derived functors of Zuckerman functors. 

214: We verify that defining relations 

215: for the Temperley-Lieb algebra result from appropriate functor 

216: isomorphisms. Projective and Zuckerman functors 

217: commute and that can be considered a ``functor'' version of 

218: the commutativity between the 

219: action of $\dU$ and the Temperley-Lieb algebra (Section~\ref{TL-and-Z}). 

220: 

221: In Section~\ref{parab-section} we construct another categorification,

222:  this time using parabolic subcategories of 

223: $\gl$ to realize $\vn$ as a Grothendieck group. There we consider the 

224: category 

225: \begin{equation} 

226: \O^n = \oplus_{k=0}^n \O^{k,n-k}, 

227: \end{equation} 

228: where $\O^{k,n-k}$ is a parabolic subcategory of a regular block, 

229: corresponding to the parabolic subalgebra of $\gl$ that contains 

230: $\mf{gl}_k\oplus \mf{gl}_{n-k}.$ In this picture the role of 

231: projective and Zuckerman functors is reversed, namely, the 

232: categorification of the Temperley-Lieb algebra action is given by 

233: projective functors, while the action of $U(\mf{sl}_2)$ is achieved via 

234: Zuckerman functors. We show that the composition of translation 

235: functors on and off the $i$-th wall at the Grothendieck group level 

236: yields the $i$-th generator of the Temperley-Lieb algebra by 

237: verifying that equivalences between these projective functors 

238: correspond to relations of the Temperley-Lieb algebra 

239: (Section~\ref{parab-tl}). This realization of the Temperley-Lieb algebra 

240: by functors was inspired and can be derived from the work [ES] of 

241: Enright and Shelton. 

242: 

243: In the second picture the action of $U(\mf{sl}_2)$ is 

244: categorified by Zuckerman functors (Section~\ref{sl2-and-Z}). 

245: This result can be extracted from the geometric approach of 

246: [BLM] and [Gr], which uses correspondences between flag varieties. 

247: The latter correspondences define functors between derived categories 

248: of sheaves, which are equivalent to the derived category of $\O^n.$ 

249: 

250: To summarize, we have two categorifications of the Temperley-Lieb algebra 

251: action on the $n$-th tensor power of the fundamental representation 

252: $\vn$: one by Zuckerman functors acting in singular blocks and the 

253: other by projective functors acting in parabolic categories. We also have 

254: two categorifications of the $U(\mf{sl}_2)$ algebra action on the same space: 

255: by projective functors between singular categories and by Zuckerman 

256: functors between parabolic categories. We conjecture that the 

257: Koszul duality functor of [BGS] exchanges these pairs of 

258: categorifications and that, more generally, the Koszul duality 

259: functor exchanges projective and Zuckerman functors. 

260: 

261: The categorification of the representation theory of $U(\mf{sl}_2)$ 

262: presented in our work explains the nature of integrality and 

263: positivity properties established in [FK] by a direct approach  

264: based on the Penrose-Kauffman calculus. 

265:  However, in this paper we did not reconsider 

266: some of the positivity and integrality results of [FK], e.g., positivity 

267: and integrality of the $6j$-symbol factorization coefficients. We expect 

268: that an extension of our approach will allow us to interpret these 

269: coefficients as dimensions of vector spaces of equivalences between 

270: appropriate functors. The problem of passing from 

271:  categorifying $U(\mf{sl}_2)$-representations to  $\U$-representations 

272: can most likely be solved by working with mixed versions of

273: projective and Zuckerman functors and  the 

274: category $\O(\gl).$ 

275: 

276: Moreover, many of our constructions 

277: admit a straightforward generalization from $U(\mf{sl}_2)$ to 

278: $U(\mf{sl}_m).$ In this case one should consider singular and parabolic 

279: categories corresponding to the subgroups $\mathbb{S}_{i_1}\times \dots 

280: \times \mathbb{S}_{i_m}, i_1+\dots + i_m=n,$ of $\mathbb{S}_n.$

281: The Temperley-Lieb algebra will be replaced by appropriate quotients 

282: of the Hecke algebra of $\mathbb{S}_n.$ A more difficult problem 

283: is to categorify the representation theory of $U(\mf{g})$ for an 

284: arbitrary simple Lie algebra $\mf{g}.$ Another interesting generalization 

285: of our results would be a categorification of the affine version of 

286: the Schur-Weyl duality. It is expected that in this case one should 

287: consider certain singular and parabolic categories of highest weight 

288: modules for affine Lie algebras $\widehat{\mf{gl}}_n.$

289: The functors of tensoring with a finite-dimensional $\mf{gl}_n$-module 

290: should be replaced by the Kazhdan-Lusztig tensoring with a tilting 

291: $\widehat{\mf{gl}}_n$-module (see [FM]). 

292: 

293: Finally we would like to discuss applications of our categorification 

294: results to a construction of topological invariants, which was the 

295: initial motivation for this work (see [CF]). It is well-known that 

296: the graphical calculus for representation theory of $\U$ and 

297: in particular for representations of the Temperley-Lieb algebra in 

298: tensor powers of $V_1$ is intimately related ([Ka]) to the Jones 

299: polynomial ([Jo]), which is a quantum invariant of links and can be 

300: extended to give an invariant of tangles. An arbitrary tangle in 

301: the three-dimensional space is a composition of elementary pieces such 

302: as braiding and local maximum and minimum tangles. To construct 

303: the Jones polynomial one attaches to these elementary tangles operators 

304: from $V_1^{\o m}$ to $\vn$ for suitable $m$ and $n$ and obtains an

305: isotopy invariant. 

306: 

307: Extending both categorifications of the Temperley-Lieb 

308: algebra at the end of Sections~\ref{sect-comult} and \ref{parab-tl} 

309: we define functors from derived categories of $\O_m$ to $\O_n$ and 

310: $\O^m$ to $\O^n$ corresponding to elementary tangles. Given a plane 

311: partition of a tangle, we can associate to it a functor, which is 

312: a composition of these basic functors. We conjecture that different 

313: plane projections produce isomorphic functors and we would get functor 

314: invariants of links and tangles. For links these invariants will take the 

315: form of $\Z$-graded homology groups. Given a diagram of a cobordism 

316: between two tangles, we can associate to it a natural transformation of 

317: functors. We expect that these natural transformations are isotopy invariants 

318: of tangle cobordisms, and, in the special case of a cobordism between 

319: empty tangles, invariants of 2-knots. To prove this conjecture 

320: one needs to present an arbitrary cobordism as a composition of 

321: elementary ones and verify all the relations between them. A complete

322: set of generators and relations has been found in [CS],[CRS] and was 

323: interpreted as a tensor 2-category in [Fi]. The match between 

324: tensor 2-categories arising from topology and representation theory will 

325: yield a graphical calculus for a categorification of the representation 

326: theory of $\U$ based on two-dimensional surfaces and as a consequence 

327: new topological invariants. 

328: 

329: {\it Acknowledgements: } J.B. and M.K. would like to thank the 

330: Institute for Advanced Study  and  the Institut des Hautes 

331: Etudes Scientifiques for their hospitality during 

332: our work on this paper. M.K. is indebted to Victor Protsak for 

333: interesting discussions. M.K. gratefully acknowledges NSF support in 

334: the form of grant DMS 9304580. The research of I.F.  was 

335: partially supported by NSF grant DMS 9700765. 

336: 

337: 

338: 

339: \section{Lie algebra $\slt$ and categories of highest weight modules}

340: \label{sect-sl2} 

341: 

342: \subsection{$\dU$ and its representations} 

343: 

344: 

345: \subsubsection{ Algebra $\dU$  }

346: 

347: 

348: The universal enveloping algebra of the Lie algebra $\slt$ is given by 

349: generators $E,F,H$ and defining relations 

350: 

351: $$ EF-FE =  H, \quad  HE-EH= 2E ,\quad HF-FH=-2F. $$

352: 

353: We will denote this algebra by $\UL.$ Throughout the paper we consider 

354: it as an algebra over the ring of integers $\Z.$ We will also need

355: two other  versions of this algebra, $\UZ$ and $\dU.$ 

356: 

357: Let $\UZ$ be the integral lattice in $\UL\o \Q$ spanned by 

358: 

359: $$E^{(a)}\binom{H}{b}F^{(c)}$$

360: 

361: for $a,b,c\ge 0.$ Here 

362: 

363: \begin{equation}

364: \label{divided-powers}

365: E^{(a)}=\frac{E^a}{a!},\quad F^{(c)}=\frac{F^c}{c!},\quad 

366: \binom{H}{b} = \frac{H(H-1)\dots(H-b+1)}{b!} 

367: \end{equation}

368: $E^{(a)}, F^{(a)}$ are known as \emph{divided powers} of $E$ and $F.$ 

369: The lattice $\UZ$ is closed under multiplication, and therefore inherits the 

370: algebra structure from that of $\UL\o \Q.$ Thus, $\UZ$ is an algebra over 

371: $\Z$ with  multiplicative generators 

372: $$1,E^{(a)},F^{(a)}, \binom{H}{a}, \quad a>0.$$ 

373: 

374: Some of the relations between the generators are written below 

375: 

376: \begin{eqnarray}

377: \label{divided-sl2-relations}

378: E^{(a)}E^{(b)} & = & 

379: \left( \begin{array}{c} a+b \\ a \end{array}\right)

380: E^{(a+b)} \\

381: F^{(a)}F^{(b)} & = & 

382: \left( \begin{array}{c} a+b \\ a \end{array}\right)

383: F^{(a+b)} \\

384: E^{(a)}F^{(b)} & = & 

385: \sum_{j=0}^{min(a,b)} F^{(b-j)}

386: \left( \begin{array}{c} H-a-b+2j \\ j \end{array}\right)

387: E^{(a-j)}. 

388: \end{eqnarray} 

389: 

390: It is easy to see that the comultiplication in $\UL\o \Q$ preserves 

391: the lattice $\UZ,$ i.e. $\Delta \UZ \subset \UZ \o \UZ,$ 

392: and the algebra $\UZ$ is actually a Hopf algebra. Note that on 

393: $E^{(a)},F^{(a)}$ the comultiplication is given by 

394: 

395: \begin{eqnarray}

396: \label{comult-div-E}

397: \Delta E^{(a)} & = & \sum_{b=0}^a E^{(b)}\o E^{(a-b)} \\

398: \label{comult-div-F}

399: \Delta F^{(a)} & = & \sum_{b=0}^a F^{(b)}\o F^{(a-b)}.  

400: \end{eqnarray} 

401: 

402: 

403: Algebra $\dU$ is obtained by adjoining a system of projectors, 

404: one for each element of the weight lattice, to the algebra $\UZ.$ 

405: Start out with a $\UZ$-bimodule, freely generated by 

406: the set \mbox{$1_n, n\in \Z.$} Quotient it out by relations

407: 

408:  \begin{eqnarray} 

409:  \binom{H}{a}1_i & = & \binom{i}{a}1_i \\

410:  1_i\binom{H}{a} & = & \binom{i}{a}1_i \\

411:  E^{(a)}1_i & = & 1_{i+2a} E^{(a)} \\

412:  F^{(a)}1_i & = & 1_{i-2a} F^{(a)}.  

413: \end{eqnarray} 

414: 

415: The quotient $\UZ$-bimodule has a unique algebra structure, compatible with 

416: the $\UZ$-bimodule structure and such that 

417: \begin{equation}

418: 1_n1_m=\delta_{n,m}1_n. 

419: \end{equation}

420: 

421: Denote the resulting $\Z$-algebra by $\dU.$ 

422: As a $\Z$-vector space, it is spanned by elements 

423: $$E^{(a)}1_n F^{(b)} \quad \mbox{ for }a,b\ge 0, n\in \Z.$$

424: As a left $\UZ$-module, $\dU$ decomposes into a direct sum

425: $$\dU =\oplus_{i\in \Z} \dU_i$$

426: where

427: $$\dU_i=\l x\in \dU|x1_i=x\r. $$ 

428: 

429: $\dU_i$ is spanned by $E^{(a)}1_{i-2b}F^{(b)}, a,b\ge 0.$ 

430: 

431: 

432: We will be using Lusztig's basis $\dB$ of $\dU,$ given by 

433: \begin{eqnarray} 

434: \label{canbasisexpl}

435: E^{(a)}1_{-i}F^{(b)} & \textrm{ for } & a,b,i\in \N, i\ge a+b \\

436: F^{(b)}1_i E^{(a)}   & \textrm{ for } & a,b,i\in \N, i > a+b . 

437: \end{eqnarray}

438:  \emph{Remark: }$ E^{(a)}1_{-a-b} F^{(b)} = F^{(b)}1_{a+b}

439: E^{(a)}.$

440: 

441: Define $\dB_i=\dB\cap\dU_i, i\in \Z.$ 

442: 

443: The important feature of Lusztig's basis is the positivity of the 

444: multiplication: for any  $x,y\in \dB$

445: 

446: $$xy=\sum_{z\in \dB}m_{x,y}^z z$$

447: with all structure constants $m_{x,y}^z$ being nonnegative integers. 

448: Although this positivity property 

449:  is very easy to verify, its generalization to 

450: quantum groups $U_q(\mathfrak g), \mathfrak g$ symmetrizable, conjectured 

451: by Lusztig, is, apparently, still unproved.  

452: The algebra $\dU$ is a special case ($q=1,\mathfrak g=\slt$) of the 

453: Lusztig's algebra $\dot{U}_q(\mathfrak g)$ (see [Lu]), obtained from  

454: the quantum group $U_q(\mathfrak g)$ by adding a system of projectors, one

455: for each element of the weight lattice. Lusztig defined a basis in 

456: $\dot{U}_q(\mathfrak g)$ and conjectured that the multiplication and 

457: comultiplication constants in this basis lie in $\N[q,q^{-1}].$ 

458: A proof of this conjecture, most likely, 

459: will require interpreting Lusztig's basis in terms of perverse sheaves on 

460: suitable varieties. 

461: 

462: Although in this paper we do not venture beyond $\slt,$ our results 

463: suggest a close link between the  Lusztig's basis of $\dot{U}_q(\mathfrak 

464: {sl}_N)$  and indecomposable projective functors for $\sln,$ 

465: $N$ and $n$ being independent parameters. 

466: 

467: \subsubsection{ Representations of $\dU$ } 

468: \label{JW-def}

469: Let $V_1$ be the two-dimensional representation over $\Z$ of $\UL$ spanned by 

470: $v_1$ and $v_{0}$ with the action of generators of $\UL$ given by 

471: \begin{eqnarray}

472: Hv_1=v_1 & Ev_1=0 & Fv_1=v_{0} \\

473: Hv_{0} = -v_{0} & Ev_{0}=v_1 & Fv_{0}=0 

474: \end{eqnarray}

475: 

476: In the obvious way, $V_1$ is also a representation of both $\UZ$ and $\dU.$ 

477: Using comultiplication, the tensor powers of $V_1$ become 

478: representations of $\UL, \UZ$ and $\dU.$ 

479: 

480: 

481: Denote by $V_0$ the one-dimensional representation of $\UL$ given 

482: by the augmentation homomorphism $\UL\to \Z$ of the universal enveloping 

483: algebra.  

484: Again, $V_0$ is a $\UZ$ and $\dU$ module in a natural way. 

485: 

486: Let $\delta$ be the module homomorphism $V_0\to V_1\o V_1$ given by 

487: \begin{equation}

488: \delta(1)=v_1\o v_{0}- v_{0}\o v_1   \label{eq:delta}

489: \end{equation}

490: 

491: 

492: For a sequence $I=a_1\dots a_n$ of ones and zeros, let 

493: $I_+$ be the number of ones in the sequence. We will denote the 

494: vector $v_{a_1}\o \dots \o v_{a_n}\in \vn$ by $v(I).$ 

495: 

496: 

497: 

498: We define a $\Q$-linear map (the symmetrization map) 

499: $p_n: \vn\o_{\Z} \Q \to \vn \o_{\Z} \Q$

500: by 

501: \begin{equation}

502: p_n(v(I))=

503: \binom{n}{I_+}^{-1}\sum_{J,J_+=I_+} v(J) \label{eq:JWprojector}

504: \end{equation}

505: where the sum on the right hand side is over all sequences 

506: $J$ of length $n$ with $J_+=I_+.$  

507: Then $p_n$ is a $\UL\o_{\Z} \Q$-module homomorphism;  in fact, it is 

508: the projection onto the unique $(n+1)$-dimensional irreducible 

509: $\UL\o_{\Z} \Q$ subrepresentation of $\vn\o_{\Z} \Q.$ 

510: 

511: \subsection{Temperley-Lieb algebra} 

512: \label{TL-algebra}

513: 

514: \begin{definition} 

515: {\rm The Temperley-Lieb algebra $TL_{n,q}$ is an algebra

516: over the ring $R=\Z[q,q^{-1}],$ where $q$ is a formal variable,  

517: with generators $U_1,\dots ,U_{n-1}$ and defining relations } 

518: \begin{eqnarray}

519: & U_iU_{i\pm 1}U_i &  = U_i \\

520: & U_iU_j & =U_jU_i \hspace{0.5in}|i-j|>1 \\

521: & U_i^2 & =-(q+q^{-1})U_i 

522: \end{eqnarray}

523: \end{definition} 

524: 

525: 

526: 

527: The Templerley-Lieb algebra admits a geometric interpretation via

528: systems of arcs on the plane. Namely, as a free $R$-module, it has 

529: a basis enumerated by isotopy classes of systems of simple, pairwise 

530: disjoint arcs that connect $n$ points on the bottom of a horizontal 

531: plane strip with $n$ points on the top. We only consider systems

532: without closed arcs. Two diagrams are multiplied by concatenating 

533: them. If simple closed loops appear as a result

534: of concatenation, we remove them, each time 

535: multiplying the diagram by $-q-q^{-1}.$ 

536: 

537: 

538: Example: let diagrams $A$ and $B$ be as depicted below. 

539: 

540: \begin{center} \epsfig{file=dr1.eps} \end{center}

541: 

542: 

543: Then their composition $BA$ can be depicted by

544: 

545: \begin{center} \epsfig{file=dr2.eps} \end{center}

546: 

547: The generator $U_i$ of $TL_{n,q}$ is given by the diagram  

548: 

549: \begin{center} \epsfig{file=dr3.eps} \end{center}

550: 

551: The defining relations have geometric interpretations. 

552: For instance, the 

553: first relation says that the diagrams below are isotopic 

554: 

555: \begin{center} \epsfig{file=dr4.eps} \end{center}

556: 

557: 

558: \begin{definition} 

559: {\rm The Temperley-Lieb algebra $TL_{n,1},$ 

560: respectively $TL_{n,-1},$ is an algebra

561: over the ring of integers, obtained from $TL_{n,q}$ by setting $q$ to

562: $1,$ respectively to $-1$,  

563: everywhere in the definition of the latter. } 

564: \end{definition} 

565: 

566: \vspace{0.15in} 

567: 

568: Thus, in $TL_{n,1}$ the value of a closed loop is $-2,$

569: in $TL_{n,-1}$ the value of a closed loop is $2,$  while in $TL_{n,q}$ 

570: a closed loop evaluates to $-q-q^{-1}.$ 

571: 

572: Recall that we denoted by $V_1$ the fundamental representation of 

573: $\UZ$. Let $u$ be an intertwiner $V_1^{\o 2}\to V_1^{\o 2}$ given 

574: by 

575: \begin{eqnarray*}

576: & u(v_1\o v_0)=-u(v_0\o v_1)& =v_0\o v_1-v_1\o v_0  \\

577: & u(v_1\o v_1)=u(v_0\o v_0) & = 0 

578: \end{eqnarray*}

579: Then $V_1^{\o n}$ is a representation of $TL_{n,1}$ with $U_i$ 

580: acting by $Id^{\o (i-1)}\o u \o Id^{\o (n-i-1)}.$ 

581: This action commutes with the Lie algebra $\slt$ action on the same 

582: space. 

583: 

584: The Temperley-Lieb algebra allows a generalization into the so-called 

585:  Temperley-Lieb category, as we now explain (for more details, see 

586:  [KaL],[Tu]). 

587: 

588: 

589: \begin{definition} {\rm The Temperley-Lieb category 

590: $TL$ has objects enumerated by nonnegative integers: $\mbox{Ob}(TL)=

591: \l \overline{0},\overline{1},\overline{2},\dots \r .$ 

592: The set of morphisms from $\overline{n}$ to $\overline{m}$ 

593: is a free $R$-module with a basis over $R$ given by the isotopy classes 

594: of systems of $\frac{n+m}{2}$ 

595: simple, pairwise disjoint arcs inside a horizontal 

596: strip on the plane that connect in pairs $n$ points on the bottom and $m$ 

597: points on the top in some order. 

598: 

599: Morphisms are composed by concatenating their diagrams. If closed loops 

600: appear after concatenation, we remove them, multiplying the diagram 

601: by $-q-q^{-1}$ to the power equal to the number of closed loops. } 

602: \end{definition} 

603: 

604: \vspace{0.15in} 

605: 

606: An example of a morphism from $\overline{5}$ to $\overline{3}$ is 

607: depicted below. 

608: 

609: \begin{center} \epsfig{file=dr5.eps} \end{center}

610: If $n+m$ is odd, there are no morphisms from 

611: $\overline n$ to $\overline m.$ 

612: Denote by $\cap_{i,n}$ for  $n\ge 2, 1\le i\le n-1$ the morphism of $TL$

613: from $\overline{n}$ to $\overline{n-2}$ given by the following 

614: diagram

615: 

616: \drawing{dr6.eps}

617: 

618: The diagram consists of $n-1$ arcs.  One of the arcs connects the 

619: $i$-th bottom point (counting from the left) with the $(i+1)$-th bottom 

620: point. The remaining arcs connect the $k$-th bottom point for $1\le k<i$ with 

621: the $k$-th top point and the $k$-th bottom point for $i+2\le k\le n$ with 

622: the $(k-2)$-th top point. 

623: 

624: Denote by $\cup_{i,n}, n\ge 0, 1\le i\le n+1$ the morphism 

625: in $TL$ from $\overline{n}$ to $\overline{n+2}$ given by the diagram

626: 

627: \drawing{dr7.eps}

628: 

629: Denote by $\mbox{Id}_{\overline{n}}$ the identity morphism from

630: $\overline{n}$ to $\overline{n}.$ This morphism  can be depicted by a diagram 

631: that is made of $n$ vertical lines: 

632: 

633: \drawing{dr8.eps}

634: 

635: The morphisms $\cap_{i,n}$ and $\cup_{i,n}$ will serve as generators 

636: of the set of morphisms in the Temperley-Lieb category. The following 

637: is a set of defining relations for $TL$  

638: 

639: \begin{eqnarray}

640: & \cap_{i+1,n+2}\circ \cup_{i,n} & = \mbox{Id}_{\overline{n}} 

641: \label{first-relation}\\

642: & \cap_{i,n+2}\circ \cup_{i+1,n} & = \mbox{Id}_{\overline{n}} 

643: \label{second-relation} \\

644: & \cap_{j,n}\circ \cap_{i,n+2}  &  = \cap_{i,n}\circ \cap_{j+2,n+2} 

645: \quad i\le j \label{third-relation} \\

646: & \cup_{j,n-2}\circ \cap_{i,n}  &  = \cap_{i,n+2}\circ \cup_{j+2,n} 

647: \quad i\le j \\

648: & \cup_{i,n-2}\circ \cap_{j,n}  &  = \cap_{j+2,n+2}\circ \cup_{i,n} 

649: \quad i\le j \\

650: & \cup_{i,n+2}\circ \cup_{j,n}  &  = \cup_{j+2,n+2}\circ \cup_{i,n} 

651: \quad i\le j 

652: \label{sixth-relation}\\

653: &  \cap_{i,n+2}\circ \cup_{i,n} & = - (q+q^{-1})\mbox{Id}_{\overline{n}}

654: \label{last-relation}

655: \end{eqnarray} 

656: 

657: The first $6$ types of relations come from isotopies of certain pairs 

658: of diagrams. For example, relations (\ref{first-relation}) and 

659: (\ref{third-relation}) correspond to the isotopies 

660: 

661: \begin{center} \epsfig{file=dr9.eps} \hspace{.5in} \end{center}

662: and 

663: \begin{center} \epsfig{file=dr10.eps} \hspace{.5in} \end{center}

664: respectively. 

665: Algebra $TL_{n,q}$ is the algebra of endomorphisms of the 

666: object  $\overline n$  of the Temperley-Lieb category. 

667: 

668: \subsection{The category of highest weight modules over a  reductive 

669: Lie algebra} 

670: 

671: \subsubsection{ Definitions } 

672: \label{definitions-LA}

673: 

674: In this section all Lie algebras and their representations are 

675: defined over the field $\C$ of complex numbers. 

676: Let $\mf{g}$ be a finite-dimensional reductive Lie algebra and 

677: $U(\mf{g})$ its universal enveloping algebra.  

678: Fix a 

679: triangular decomposition $\mf{g}=\mf{n}_+\oplus \mf{h}\oplus \mf{n}_-.$ 

680: Let $R_+$ be the set of positive roots and 

681: $\rho$ the half-sum of positive roots. 

682: For $\lambda\in \ch$ denote by $M_{\lambda}$ the Verma module with

683: highest weight $\lambda-\rho$ and by $L_{\lambda}$ the irreducible quotient

684: of $M_{\lambda}.$

685: The module $L_{\lambda}$ is finite-dimensional if and only if $\lambda-\rho$ 

686: is an integral dominant weight. 

687: 

688: Denote by $\Og{\mf{g}}$ the category of finitely generated $U(\mf{g})$-modules 

689: that are $\mf{h}$-diagonalizable and locally $U(\mf{n}_+)$-nilpotent. 

690: The category $\Og{\mf{g}}$ is called \textit{ the category of 

691: highest weight $\mf{g}$-modules}. 

692: Let $P_{\lambda}$ denote the projective 

693: cover of $L_{\lambda}$ (see [BGG] for the existence of projective covers). 

694: 

695: If $A$ is an additive category, denote by $K(A)$ the Grothendieck group 

696: of $A.$ Denote by $[M]$ the image of an object $M\in \textrm{Ob}(A)$ 

697: in the Grothendieck group of $A.$ Denote by $D^b(A)$ the bounded derived 

698: category of an abelian category $A.$ 

699: 

700: 

701: \subsubsection{ Projective functors }

702: 

703: 

704: This section is a brief introduction to projective 

705: functors. We refer the reader to Bernstein-Gelfand paper [BG] for a 

706: detailed treatment and further references. 

707: 

708: Denote by $\Theta$ the set of maximal ideals of the center $Z$ of 

709: $U(\mf{g}).$ 

710: We can naturally identify $\Theta$ with the quotient of the weight space

711: $\ch$ by the action by reflections 

712: of the Weyl group $W$ of $\gl.$ 

713: We will denote by $\eta$ the quotient map $\ch\to \Theta.$ 

714: For $\theta\in\Theta$ denote by $J_{\theta}$ the corresponding maximal 

715: ideal of $Z.$ Thus, the Verma module $M_{\lambda}$ with the highest weight 

716: $\lambda - \rho$ is annihilated by the maximal central 

717: ideal $J_{\eta(\lambda)}.$ 

718: 

719: For $\theta\in \Theta$ denote by $\Ogc{\theta}{\mf{g}}$ a full subcategory of 

720: $\Og{\mf{g}}$ consisting of modules that are annihilated by some power of 

721: the central ideal $J_{\theta}:$ 

722: 

723: \begin{equation} 

724: M\in \Ogc{\theta}{\mf{g}} \hspace{0.2in} \Longleftrightarrow 

725:  \hspace{0.2in} M\in \Og{\mf{g}} \mbox{ and }J_{\theta}^N M =0 

726: \textrm{ for sufficiently large }N

727: \end{equation}

728: A module $M\in \O(\mf{g})$ belongs to $\Ogc{\theta}{\mf{g}}$ if and only 

729: if all of the simple subquotients of $M$ are isomorphic to simple modules 

730: $L_{\lambda}, \lambda \in \eta^{-1}(\theta).$ 

731: We will call modules in $\Ogc{\theta}{\mf{g}}$ 

732: \textit{highest weight modules 

733: with the generalized central character} $\theta$. 

734: The category $\Og{\mf{g}}$ splits as a direct sum of categories 

735: $\Ogc{\theta}{\mf{g}}$ 

736: over all $\theta\in \Theta.$ 

737: 

738: Denote by $\textrm{proj}_{\theta}$ the functor from $\Og{\mf{g}}$ to 

739: $\Ogc{\theta}{\mf{g}}$ 

740: that, to a module $M,$ associates the largest submodule of $M$

741: with the generalized central character $\theta.$ 

742: Let $F_V$ be the functor of tensoring with a finite-dimensional 

743: $\mf{g}$-module $V.$ 

744: 

745: \begin{definition} {\rm $F:\Og{\mf{g}} \to \Og{\mf{g}}$ is a projective 

746: functor if it is isomorphic to a direct summand of the functor $F_V$ for some 

747: finite dimensional module $V.$ } 

748: \end{definition}

749: 

750: \vspace{0.15in} 

751: 

752: The functor $\textrm{proj}_{\theta}$ is an example of a 

753:  projective functor, since it is  

754: a direct summand of the functor of tensoring with the  

755: one-dimensional representation. We have an isomorphism of functors 

756: \begin{equation} 

757: F_V=\oplusop{\theta_1,\theta_2\in \Theta}

758: (\textrm{proj}_{\theta_1}\circ F_V\circ \textrm{proj}_{\theta_2})

759: \end{equation} 

760: 

761: 

762: Any projective functor takes projective objects 

763:  in $\Og{\mf{g}}$ to projective objects.  

764: The composition of projective functors is again a projective functor. 

765: Each projective functor splits as a direct sum of 

766: indecomposable projective functors. 

767: 

768: Projective functors are exact. 

769: Therefore, they induce endomorphisms of the Grothendieck 

770: group of the category $\Og{\mf{g}}.$ The following result is proved in [BG]: 

771: 

772: \begin{prop}

773: \label{prop:projective-functors-isomorphic}

774: Let $\lambda$ be a dominant integral weight, $\theta=\eta(\lambda)$ 

775:  and 

776: $F,G$ projective functors from $\Ogc{\theta}{\mf{g}}$ to 

777: $\Og{\mf{g}}.$ Then 

778: 

779: \begin{enumerate}

780: \item Functors 

781: $F$ and $G$ are isomorphic if and only if 

782: the endomorphisms of $K(\O(\mf{g}))$ induced by 

783: $F$ and $G$ are equal. 

784: \item Functors $F$ and $G$ are isomorphic if and only if 

785:  modules $FM_{\lambda}$ and $GM_{\lambda}$ are isomorphic.  

786: \end{enumerate}

787: \end{prop}

788: 

789: We will be computing the action of projective functors on Grothendieck 

790: groups of certain subcategories of the category of highest weight 

791: modules. The simplest basis in the Grothendieck group of $\O(\mf{g})$ 

792: is given by images of Verma modules. The following 

793: proposition shows that this basis is also handy for 

794: writing the action of projective functors on the Grothendieck group

795: of $\O(\mf{g}).$ 

796: 

797: \begin{prop}

798: \label{prop:Verma-tensor-finite}

799: Let $V$ be a finite-dimensional $\mf{g}$-module,

800: $\mu_1,\dots,\mu_m$ a multiset of weights of $V,$

801:  $M_{\chi}$ the Verma

802: module with the highest weight $\chi-\rho,$ then 

803: 

804: \begin{enumerate}

805: \item The module $V\o M_{\chi}$ admits a fitration with 

806: successive quotients isomorphic to Verma modules 

807: $M_{\chi+\mu_1},\dots ,M_{\chi+\mu_m}$ (in some order).

808: \item We have an equality in the Grothendieck group 

809: $K(\Og{\mf{g}})$:

810: $$[V\otimes M_{\chi}]=\sum_{i=1}^m [M_{\chi+\mu_i}]$$

811: \end{enumerate}

812: \end{prop}

813: 

814: 

815: 

816: \subsubsection{ Parabolic categories }

817: \label{parabolic-categories} 

818: 

819: Let $\mf{p}$ be a parabolic subalgebra of $\mf{g}$ that contains 

820: $\mf{n}_+\oplus \mf{h}.$ Denote by $\Ogp{\mf{g}}{\mf{p}}$ the full 

821: subcategory of $\Og{\mf{g}}$ that consists of $U(\mf{p})$ locally 

822: finite modules. Notice that projective functors preserve 

823: subcategories  $\Ogp{\mf{g}}{\mf{p}}.$

824: 

825: A generalized Verma module relative to a parabolic subalgebra 

826: $\mf{p}$ of $\mf{g}$ (see [Lp],[RC]) will be called 

827: a $\mf{p}$-Verma module. The Grothendieck group of $\Ogp{\mf{g}}{\mf{p}}$ 

828: is generated by images $[M]$ of generalized Verma modules. 

829: 

830: 

831: 

832: For a central character $\theta$ and a parabolic subalgebra $\mf{p}$ of 

833: $\mf{g},$ denote by $\O_{\theta}(\mf{g},\mf{p})$ the full subcategory of 

834: $\O(\mf{g})$ consisting of $U(\mf{p})$-locally finite modules 

835: annihilated by some power of the central ideal $J_{\theta}.$ The 

836: category  $\O_{\theta}(\mf{g},\mf{p})$  is the intersection of 

837: subcategories  $\O_{\theta}(\mf{g})$  and 

838: $\O(\mf{g},\mf{p})$ of $\O(\mf{g}).$ 

839: 

840: 

841: The following lemma is an obvious

842: generalization of a special case of Lemma 3.5 in [ES].

843: 

844: \begin{lemma}

845: \label{functor-equivalence}

846:  Let $T,S$ be covariant, exact functors from

847: $\Ogcp{\theta}{\mf{g}}{\mf{p}}$ to some abelian category $\mc A$ and let

848: $f$ be a natural transformation from

849: $T$ to $S.$ If $f_M: T(M)\to S(M)$ 

850: is an isomorphism for each generalized Verma

851: module $M\in \Ogcp{\theta}{\mf{g}}{\mf{p}},$ then $f$ is an isomorphism 

852: of functors.

853: \end{lemma}

854: 

855: Let $\mf{g}_1,\mf{g}_2$ be reductive Lie algebras, with fixed 

856: Cartan subalgebras 

857: $\mf{h}_j\subset \mf{g}_j,j=1,2.$ Suppose we have two parabolic 

858: subalgebras $\mf{p}_1,\mf{p}_2$ such that

859: $\mf{h}_j\subset \mf{p}_j\subset \mf{g}_j.$

860: Fix central characters $\theta_j$ of $\mf{g}_j.$ 

861: 

862: \begin{lemma} 

863: \label{category-equivalence}

864: Suppose that we have two exact functors

865: \begin{eqnarray*}

866: & f_{12}: \Ogcp{\theta_2}{\mf{g}_2}{\mf{p}_2} & \to 

867: \Ogcp{\theta_1}{\mf{g}_1}{\mf{p}_1} \\

868: & f_{21}: \Ogcp{\theta_1}{\mf{g}_1}{\mf{p}_1} & \to 

869: \Ogcp{\theta_2}{\mf{g}_2}{\mf{p}_2} 

870: \end{eqnarray*}

871: such that $f_{21}$ is isomorphic to both left and right adjoint 

872: functors of $f_{12},$ $f_{21}$ takes $\mf{p}_1$-Verma modules to 

873: $\mf{p}_2$-Verma modules, $f_{12}$ takes $\mf{p}_2$-Verma modules to 

874: $\mf{p}_1$-Verma modules, 

875: $f_{12}f_{21}(M)$ is isomorphic to $M$ for any $\mf{p}_1$-Verma module M and 

876: $f_{21}f_{12}(M)$ is isomorphic to $M$ for any $\mf{p}_2$-Verma module M. 

877: Then $f_{12}$ and $f_{21}$ are equivalences of categories and 

878: the natural transformations 

879: $$a:f_{12}f_{21}\to Id,\quad b: Id \to f_{12}f_{21}$$ 

880: coming from the adjointness are isomorphisms of functors. 

881: \end{lemma}

882: 

883: 

884: \emph{Proof: }By Lemma~\ref{functor-equivalence} it suffices to prove that, 

885: for any $\mf{p}_1$-Verma module $M,$ the module morphisms 

886: $$a_M: f_{12}f_{21}(M)\to M,\quad b_M: M\to f_{12}f_{21}(M)$$

887: are isomorphisms. Note that $f_{21}(M)$ is a $\mf{p}_2$-Verma 

888: module and $f_{12}f_{21}(M)$ is isomorphic to $M.$ 

889: The hom space $\mbox{Hom}_{\mf{g}_1}(M,M)$ is one-dimensional

890: and all morphisms are just scalings of the identity morphism. We have

891: a natural isomorphism

892: $$

893: \mbox{Hom}_{\mf{g}_1}(f_{12}f_{21}(M),M) = 

894: \mbox{Hom}_{\mf{g}_2}(f_{21}M,f_{21}M).$$

895: Under this isomorphism, the map $a_M: M \to M$ corresponds

896: to the identity map $f_{21}M\to f_{21}M.$ 

897: This identity map generates the space

898: $\mbox{Hom}_{\mf{g}_2}(f_{21}M,f_{21}M),$ therefore $a_M$ generates 

899: the hom space $\mbox{Hom}_{\mf{g}_1}(M,M),$ and 

900: thus, $a_M$ is an isomorphism of $M$,

901: being a non-zero multiple of the identity map. Therefore, $a$ is an 

902: isomorphism of functors. 

903: Similarly, $b$ is a functor isomorphism. $\square$

904: 

905: This lemma is used in Section~\ref{TL-and-Z} in the proof of 

906: Theorem~\ref{functors-and-isotopies}. 

907: 

908: 

909: \subsubsection{Zuckerman functors} 

910: 

911: 

912: 

913: 

914: Here we recall the basic properties of Zuckerman derived functors,

915: following [ES] and [EW]. Knapp and Vogan's book [KV] contains a 

916: complete treatment of Zuckerman functors, but here we will only need some 

917: basic facts. 

918: 

919: Throughout the paper we restrict Zuckerman functors to the category 

920: of highest weight modules. 

921: 

922: Let $\mf{g},\mf{p}$ be as in Section~\ref{parabolic-categories}. 

923: The parabolic Lie algebra 

924: $\mf{p}$ decomposes as a direct sum $\mf{m}\oplus \mf{u}$ where $\mf{m}$ 

925: is the maximal reductive subalgebra of $\mf{p}$  

926:  and $\mf{u}$ is the nilpotent radical of $\mf{p}.$  The reductive subalgebra 

927: $\mf{m}$ contains the Cartan subalgebra $\mathfrak h$ of $\mf{g}.$ 

928: Let $d=\mbox{dim}(\mf{m})-\mbox{dim}(\mf{\mathfrak h}).$

929: We denote by $\ast$ the contravariant duality functor  in $\O.$

930: 

931: Let $\Gamma_{\mf{p}}$ be the functor from $\Og{\mf{g}}$ to 

932: $\Ogp{\mf{g}}{\mf{p}}$ that to a module $M\in \Og{\mf{g}}$ associates 

933: its maximal locally $U(\mf{p})$-finite submodule. 

934: $\Gamma_{\mf{p}}$ is called the Zuckerman functor. Functor 

935: $\Gamma_{\mf{p}}$ is left exact and the category $\Og{\mf{g}}$ has 

936: enough injectives, so we can define the derived functor 

937: $$\mc{R}\Gamma_{\mf{p}} : D^b(\Og{\mf{g}}) \lra D^b(\Ogp{\mf{g}}{\mf{p}})$$

938: and its cohomology functors 

939: $$\mc{R}^i\Gamma_{\mf{p}}: \Og{\mf{g}} \lra \Ogp{\mf{g}}{\mf{p}}.$$

940: 

941: 

942: 

943: \begin{prop}

944: \label{Zproperties}

945: \begin{enumerate}

946: \item For $i>d , \mc{R}^i\Gamma_{\mf{p}}=0.$

947: \item Projective functors commute with Zuckerman functors. More precisely,

948: if $F$ is a projective functor, then there are natural isomorphisms

949:  of functors 

950: \begin{eqnarray*}

951: F\circ \Gamma_{\mf{p}} & \cong & \Gamma_{\mf{p}}\circ F \\

952: F\circ \mc{R}\Gamma_{\mf{p}} & \cong & \mc{R}\Gamma_{\mf{p}}\circ F

953: \end{eqnarray*}

954: \item The functors $M\longmapsto \mc{R}^i\Gamma_{\mf{p}}(M)$ and

955: $M\longmapsto \mc{R}^{d-i}\Gamma_{\mf{p}}

956: (M^{\ast})^{\ast}, M\in \O(\mf{g})$ are 

957: naturally equivalent.

958: \item $\mc{R}^d\Gamma_{\mf{p}}$ is isomorphic to the functor that to a 

959: module $M\in \O(\mf{g})$ associates the maximal locally 

960: $\mf{p}$-finite quotient of $M.$ 

961: 

962: 

963: \end{enumerate}

964: \end{prop}

965: 

966: \emph{Proof: } See [EW].  Zuckerman functors commute with functors 

967: of tensor product by a finite-dimensional module. A projective functor 

968: is a direct summand of a tensor product functor. Part 2 of the 

969: proposition follows. $\square$ 

970: 

971: 

972: \subsection{ Singular blocks of the highest weight category 

973: for $\gl$}

974: 

975: \subsubsection{ Notations }

976: \label{section-def-gln}

977: 

978: We fix once and for all a triangular decomposition

979: $\mathfrak{n}_+\oplus \mathfrak{h}\oplus \mathfrak{n}_-$ of 

980: the Lie algebra $\gl.$ The Weyl group 

981: of $\gl$ is isomorphic to the symmetric group $S_n.$  

982: Choose an orthonormal basis $e_1,\dots,e_n$ in the Euclidean space 

983: $\R^n$ and identify the complexification $\C\o_{\R}\R^n$ with 

984: the dual $\ch$ of Cartan subalgebra so that 

985: $R_+=\l e_i-e_j, i < j\r$ is  the set of positive roots and 

986: $\alpha_i=e_i-e_{i+1},1\le i\le n-1$ are simple roots.

987: The generator $s_i$ of the Weyl group  $W=S_n$ acts on $\ch$ by permuting

988:  $e_i$ and $e_{i+1}.$

989: Denote by $\rho$ the half-sum of positive roots 

990: $$\rho=\frac{n-1}{2}e_1+\frac{n-3}{2}e_2+\dots+\frac{1-n}{2}e_n$$

991: Sometimes we will use the notation $\rho_n$ instead of $\rho.$ 

992: %For $i=0,1,\dots n$ denote by $\lambda_i$ the weight $e_1+\dots +e_i.$ 

993: 

994: For a sequence $a_1,\dots,a_n$ of zeros and ones, denote by 

995: $M(a_1\dots a_n)$ the Verma module with the 

996: highest weight $a_1e_1+\dots+a_ne_n-\rho.$ Similarly, $L(a_1\dots a_n)$

997: will denote the simple quotient of $M(a_1\dots a_n)$ and $P(a_1\dots a_n)$ 

998: the minimal projective cover of $L(a_1\dots a_n)$. 

999: 

1000: The sequence of $n$ zeros, respectively ones, will be denoted by $0^n$, 

1001: respectively $1^n.$ If $I_1,I_2$ are two sequences of $0$'s and $1$'s, 

1002: we denote their concatenation by $I_1I_2.$ 

1003: 

1004: Recall that $\lambda_i=e_1+\dots+e_i$ is a fundamental 

1005: weight of $\gl$. Denote by $\theta_i=\eta(\lambda_i)$ the 

1006: corresponding central 

1007: character of $\gl.$ We denote the category $\O_{\theta_i}(\gl)$ by

1008:  $\O_{i,n-i}.$ 

1009: A module $M\in \O(\gl)$ lies in $\O_{i,n-i}$ if and only if all of its simple 

1010: subquotients are isomorphic to $L(a_1\dots a_n)$ for sequences 

1011: $a_1\dots a_n$ of zeros and ones with exactly $i$ ones. 

1012: 

1013: Denote by $\O_n$ the direct sum of categories $\O_{i,n-i}$ as 

1014: $i$ ranges over all integers from $0$ to $n$: 

1015: \begin{equation} 

1016: \O_n= {\mathop{\oplus}\limits_{i=0}^n}\O_{i,n-i}. 

1017: \end{equation}   

1018: When $i<0$ or $i> n,$ denote by $\O_{i,n-i}$ the subcategory of $\O(\gl)$ 

1019: consisting of the zero module. 

1020: We have an isomorphism of Grothendieck groups 

1021: \begin{equation} 

1022: K(\O_n)= {\mathop{\oplus}\limits_{i=0}^n}K(\O_{i,n-i}).

1023: \end{equation} 

1024: 

1025: Let $\Upsilon_n$ be 

1026: the isomorphism of abelian groups $\Upsilon_n: K(\O_n)\to \vn$

1027: given by 

1028: \begin{equation}

1029: \label{iso-upsilon}  

1030: \Upsilon_n [M(a_1\dots a_n)]=v_{a_1}\o \dots \o v_{a_n}.

1031: \end{equation} 

1032: 

1033: 

1034: \subsubsection{ Simple and projective module bases in the Grothendieck 

1035: group $K(\O_n)$} 

1036: \label{sec:canbasis-formulas}

1037: 

1038: Abelian group isomorphism $\Upsilon_n$ identifies the Grothendieck 

1039: group of $\O_n$ and the $\dU$-module $\vn.$ In this section we describe 

1040: the images under $\Upsilon_n$ 

1041: of simple modules and indecomposable projectives in $\O_n.$ 

1042: The only result of this section that we use later is the 

1043: formula (\ref{fla:0101})

1044: for the image of the indecomposable projective $P(0^j1^k0^l1^m).$ 

1045: This formula is 

1046: used in the proof of Theorem~\ref{thrm:indecomposable-proj-functors}. 

1047: 

1048: 

1049: 

1050: Let us define $3$ bases in $\vn.$ We are working over $\Z$ and thus 

1051: $\vn$ is a free abelian group of rank $2^n.$ 

1052: We will parametrize basis elements 

1053: by sequences of ones and zeros of length $n.$ 

1054: 

1055: First, the basis $\l v(a_1\dots a_n),a_i\in \l 0,1\r \r$ 

1056: will be given by 

1057: $$v(a_1\dots a_n)=v_{a_1}\o \dots \o v_{a_n}$$

1058: so that, for a sequence $I$ of length $n$ of zeros and ones, we 

1059: have $\Upsilon_n[M(I)]=v(I).$ We will call this basis \textit{ the product 

1060: basis of } $\vn.$  

1061: 

1062: Next we introduce the 

1063: basis $\l l(a_1\dots a_n),a_i\in \l 0,1\r  \r$ 

1064: by induction of $n$ as follows: 

1065: 

1066: (i) $l(1)=v_1,l(0)=v_{0},$ 

1067: 

1068: (ii) $l(0a_2\dots a_n)=v_{0}\o l(a_2\dots a_n),$

1069: 

1070: (iii) $l(a_1 \dots a_{n-1}1)=l(a_1,\dots,a_{n-1})\o v_1,$

1071: 

1072: (iv) 

1073: \begin{eqnarray*}

1074: &  & l(a_1\dots a_{i-1}10a_{i+2}\dots a_n) = \\

1075: & & (\mbox{Id}^{\o (i-1)}\o \delta \o \mbox{Id}^{\o (n-i-1)})l(a_1\dots a_{i-1}

1076: a_{i+2}\dots a_n),

1077: \end{eqnarray*}

1078: where $\delta$ is the intertwiner $V_0\to V_1\o V_1$ defined by 

1079: the formula (\ref{eq:delta}) and $\mbox{Id}$ denotes the identity homomorphism 

1080: of $V_1.$  

1081: These rules are consistent and uniquely define $l(a_1\dots a_n)$ 

1082: for all $a_1\dots a_n.$ 

1083: 

1084: Let us now define the basis 

1085: $\l p(a_1\dots a_n),a_i\in \l 0,1\r \r.$ 

1086: Let $I$ be a sequence of zeros and ones. Then define the basis inductively by 

1087: the rules 

1088: 

1089: (i) $p(1I)=v_1\o 

1090: p(I),$

1091: 

1092: (ii) $p(I0)=p(I)\o v_0,$

1093: 

1094: 

1095: (iii) If a sequence $I_1$ is either empty or ends with at least 

1096: $j$ zeros, and a sequence $I_2$ is either empty or starts with 

1097: at least $k$ ones, then 

1098: $$p(I_10^j1^kI_2)=\binom{j+k}{j}

1099: (\mbox{Id}^{\o |I_1|}\o p_{j+k}\o \mbox{Id}^{\o |I_2|} ) 

1100: p(I_11^{k}0^{j}I_2)$$

1101: where $|I|$ stands for the length of $I.$ 

1102: 

1103: \begin{prop} These rules are consistent, and for each sequence $I$ define an 

1104: element of $V_1^{\o |I|}.$ 

1105: \end{prop}

1106: 

1107: Proposition follows from the results of [K], Section 3,

1108:  setting $q$ to $1.$ 

1109: Note that it is not even obvious that $p(I)$ lies in 

1110: $\vn$ because the projector $p_i$ is defined (see 

1111: Section~\ref{JW-def}) 

1112: as an operator in  $V_1^{\o i}\o_{\Z} \Q$, rather than in $V_1^{\o i}.$  

1113: 

1114: 

1115: 

1116: 

1117: 

1118: \begin{prop}

1119: \label{prop:simple-projective-bases}

1120:  The isomorphism $\Upsilon_n:K(\O_n)\to \vn$ takes the images of simple, 

1121: resp. indecomposable projective modules 

1122: to elements of the basis $\l l(I) \r_I $, resp. 

1123: $\l p(I)\r_I$ of $\vn:$

1124: 

1125: \begin{eqnarray*}

1126: \Upsilon_n[L(I)] & = & l(I) \\

1127: \Upsilon_n[P(I)] & = & p(I)

1128: \end{eqnarray*}

1129: \end{prop}

1130: 

1131: \emph{Proof: } The rules (i)-(iii) for $p(I)$ 

1132: can be used to write down the relations 

1133: between the coefficients of the transformation matrix 

1134: from the basis $\l v(I)\r_I$ to the basis $\l p(I) \r_I$ of 

1135: $\vn.$ These relations are equivalent to the $q=1$ 

1136: specialization of Lascoux-Sch\"{u}tzenberger's 

1137: recursive formulas (see [LS] and [Z]) 

1138: for the Kazhdan-Lusztig polynomials in the Grassmannian case, as follows 

1139: from the computation at the end of [FKK] (again, setting $q$ to $1$). 

1140: 

1141: Kazhdan-Lusztig polynomials 

1142: in the Grassmannian case for $q=1$ are 

1143: coefficients of the transformation matrix from the Verma module to

1144: the simple module basis of Grothendieck groups of certain parabolic 

1145: subcategories of a regular block of $\O(\gl).$ These parabolic subcategories 

1146: are Koszul dual (see [BGS], Theorem 3.11.1 for the general statement) 

1147: to the singular blocks $\O_{i,n-i}, 0\le i\le n$ of 

1148: $\O(\gl).$ Koszul duality functor descends to the isomorphism 

1149: of Grothendieck groups 

1150: that exchanges simple and projective modules in corresponding categories. 

1151: Therefore, Lascoux-Sch\"{u}tzenberger's formulas also describe 

1152: coefficients of decomposition of projective modules in the Verma module 

1153: basis of $K(\O_n).$ It now follows that $\Upsilon_n[P(I)]=p(I).$ 

1154: 

1155: Introduce a bilinear form on $K(\O_n)$ by $<[M(I)],[M(J)]>=\delta_I^J.$ 

1156: The BGG reciprocity implies $<[P(I)],[L(J)]> =\delta_I^J,$ i.e., 

1157: the basis $\l [L(I)] \r $  of $K(\O_n)$ is dual to the basis 

1158: $\l [P(I)] \r$ of $K(\O_n).$ Abelian group isomorphism $\Upsilon_n$ 

1159: transform this bilinear form to a 

1160: bilinear form on $\vn$  such that $<v(I),v(J)>=\delta_I^J.$ 

1161: From the main computation of Chapter 3 of [K], specializing $q$ to $1$, 

1162: it follows 

1163: that bases $\l l(I) \r $ and 

1164: $\l p(I) \r $ are orthogonal relative to this form. 

1165: We have  $\Upsilon_n[M(I)]=v(I)$ by definition of $\Upsilon_n$ and 

1166: we have already established that $\Upsilon_n[P(I)]= p(I)$. We conclude that 

1167: $\Upsilon_n[L(I)]= l(I).$ 

1168: 

1169: $\square$  

1170: 

1171: 

1172: To prove Theorem~\ref{thrm:indecomposable-proj-functors} 

1173: we will need an explicit formula for $p(0^j1^k0^l1^m):$

1174: 

1175: \begin{equation}

1176: \label{fla:0101}

1177: \begin{array}{lll}

1178: & & p(0^j1^k0^l1^m) = \\

1179: & & =  \binom{j+k}{k}\binom{j+l+m}{m} 

1180: (p_{j+k}\o \mbox{Id}^{\o (l+m)})(\mbox{Id}^{\o k}\o p_{j+l+m}) \\

1181: & & (v_1^{\o (k+m)}\o v_0^{\o (j+l)}) \textrm{ if }k\le l \\

1182: & & = \binom{l+m}{m}\binom{j+k+m}{j}

1183: (\mbox{Id}^{\o (j+k)}\o p_{l+m})(p_{j+k+m}\o \mbox{Id}^{\o l}) \\

1184: & & (v_1^{\o (k+m)}\o 

1185: v_0^{\o (j+l)}) \textrm{ if }k\ge l. 

1186: \end{array}

1187: \end{equation}

1188: 

1189: This formula follows from recurrent relations (i)-(iii) 

1190: for $p(I)$ that we gave earlier in this section. 

1191: 

1192: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1193: %%%

1194: %%%         SINGULAR CATEGORIES 

1195: %%%

1196: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1197: 

1198: \section{Singular categories} 

1199: \label{sect-sing} 

1200: 

1201: \subsection{ Projective functors and 

1202: $\mathfrak{sl}_2$ }

1203: \label{proj-sl2} 

1204: 

1205: 

1206: \subsubsection {Projective functors $\E$ and $\F$} 

1207: \label{functors-EF}

1208:  

1209: Let $\Lo$ be the $n$-dimensional representation of $\gl$ with 

1210: vweights $e_1,e_2,\dots ,e_n.$ The dual representation $\Ld$ has 

1211: weights $-e_1,-e_2,\dots, -e_n.$ 

1212: 

1213: Recall that $\O_{i,n-i}, i=0,1,\dots n$ 

1214: is the singular block of $\O(\gl)$ consisting 

1215: of modules with generalized central character $\theta_i=\eta(\lambda_i).$ 

1216: For $i<0$ and $i>n$ we defined $\O_{i,n-i}$ to be the trivial subcategory 

1217: of $\O(\gl).$ 

1218: 

1219: Denote by $\E_i$ the projective functor 

1220: $$(\textrm{proj}_{\theta_{i+1}})

1221: \circ F_{\Lo}: \quad \O_{i,n-i}\to \O_{i+1,n-i-1}$$

1222: given by tensoring with the $n$-dimensional representation

1223: $\Lo$ and then taking the 

1224: largest submodule of this tensor product that lies in $\O_{i+1,n-i-1}.$ 

1225: 

1226: Similarly,  

1227: denote by $\F_i$ the projective functor 

1228: from $\O_{i.n-i}$ to $\O_{i-1,n-i+1}$

1229: given by tensoring with $\Ld$ and then taking the largest  

1230: submodule that belongs to $\O_{i-1,n-i+1}$

1231: 

1232: \begin{theorem}

1233: \label{thrm:EF-relation}

1234: For $i=0,1,\dots n$ there are isomorphisms of projective functors 

1235: \begin{equation} 

1236: (\E_{i-1}\circ \F_i) \oplus

1237: \textrm{Id}^{\oplus(n-i)}

1238: \cong

1239: (\F_{i+1}\circ \E_i) \oplus

1240: \textrm{Id}^{\oplus i}  

1241: \end{equation} 

1242: where $Id$ denotes the identity functor $Id:\O_{i,n-i}\to \O_{i,n-i}.$ 

1243: \end{theorem}

1244: 

1245: \emph{Proof:} By Proposition~\ref{prop:projective-functors-isomorphic}

1246:  it suffices to check the equality

1247: of endomorphisms of the Grothendieck group of $\O_{i,n-i}$ 

1248: induced by these projective functors.  

1249: 

1250: 

1251: Denote by $[\E_i]$ and $[\F_i]$ 

1252: the homomorphisms of the Grothendieck groups induced by functors 

1253: $\E_i$ and $\F_i:$

1254: \begin{eqnarray*}

1255: & [\E_i]: &  

1256: K(\O_{i,n-i})\to K(\O_{i+1,n-i-1}) \\

1257: & [\F_i]: & 

1258: K(\O_{i,n-i})\to K(\O_{i-1,n-i+1})

1259: \end{eqnarray*}

1260: The Grothendieck group $K(\O_{i,n-i})$ is

1261: free abelian of rank $\binom{n}{i}$ and 

1262: is spanned by images $[M(a_1\dots a_n)]$ 

1263: of Verma modules $M(a_1\dots a_n)$ for all possible sequences

1264: $a_1\dots a_n$ of zeros and ones with $i$ ones. 

1265: 

1266: 

1267: 

1268: By Proposition~\ref{prop:Verma-tensor-finite}

1269: \begin{equation}

1270: [M(a_1\dots a_n)\otimes \Lo]=

1271: \sum_{j=1}^n[M(a_1\dots a_j'\dots a_n)]

1272: \end{equation}

1273: where $a_j'=a_j+1.$ 

1274: 

1275: The functor $\E_i$ is a composition of tensoring with $\Lo$

1276:  and a projection onto $\O_{i+1,n-i-1},$ hence we get 

1277: 

1278: \begin{prop} Let $a_1\dots a_n$ be a sequence of zeros and ones that 

1279: contains $i$ ones. Then 

1280: 

1281: \begin{equation}

1282:  [\E_iM(a_1 \dots a_n)] =

1283: \sum_{j=1,a_j=0}^{n}[M(a_1\dots a_{j-1}1 a_{j+1}\dots a_n)]

1284: \end{equation}

1285: 

1286: \end{prop}

1287: 

1288: In the same fashion, we obtain 

1289: 

1290: \begin{prop} Let $a_1\dots a_n$ be a sequence of zeros and ones that 

1291: contains $i$ ones. Then 

1292: 

1293: \begin{equation}

1294: [\F_iM(a_1 \dots a_n)] =

1295: \sum_{j=1,a_j=1}^{n}[M(a_1\dots a_{j-1}0 a_{j+1}\dots a_n]

1296: \end{equation}

1297: 

1298: \end{prop}

1299: 

1300: 

1301: Therefore, after identifying the Grothendieck group $K(\O_n)$ with 

1302: $\vn$ via the isomorphism $\Upsilon_n$ (formula (\ref{iso-upsilon}), 

1303: we see that 

1304: maps $[\E_i],[\F_i]$  from $K(\O_{i,n-i})$ to $K(\O_{i+ 1, n-i-1})$

1305: and $K(\O_{i-1,n-i+1})$    

1306: coincide with maps induced by the 

1307: Lie algebra $\slt$ generators $E$ and $F$ on 

1308: the weight $2i-n$ subspace of the module $\vn.$  

1309: This immediately gives 

1310: \begin{prop} We 

1311: have the following equality of endomorphisms of the abelian group 

1312: $K(\OO{i,n-i})$ : 

1313: $$[\E_{i-1}] [\F_i]+(n-i)\textrm{Id} = [\F_{i+1}] [\E_i]+ 

1314:  i \cdot \textrm{Id}$$  

1315: \end{prop} 

1316: 

1317: Theorem~\ref{thrm:EF-relation} follows. $\square$ 

1318: 

1319: This proof also implies

1320: 

1321: \begin{corollary} \label{tensor-power}

1322:  Considered as  an 

1323: $\slt$-module with $E$, resp. $F$ acting as $\sum_i[\E_i],$ resp. 

1324: $\sum_i[\F_i],$ the Grothendieck group $K(\O_n)$ 

1325: is isomorphic to the $n$-th  tensor power of the fundamental

1326: two-dimensional representation of $\slt.$

1327: \end{corollary}

1328: 

1329: 

1330: \subsubsection {Realization of $\dU$ by projective functors} 

1331: \label{sect-divp} 

1332: 

1333: Next we provide a realization of the divided powers of 

1334: $E$ and $F$  by projective functors.

1335: 

1336: 

1337: Let 

1338: $\E^{(k)}_i$ be the functor from $\O_{i,n-i}$ to $\O_{i+k,n-i-k}$ given 

1339: by tensoring with the $k$-th exterior power of $\Lo$ and then projecting 

1340: onto the submodule with the  generalized central character $\theta_{i+k}:$ 

1341: \begin{equation}

1342: \E^{(k)}_i(M)=\textrm{proj}_{\theta_{i+k}}(\Lambda^k \Lo \o M)

1343: \end{equation}

1344: Similarly, 

1345: \begin{equation}

1346: \F^{(k)}_i:\O_{i,n-i}\to \O_{i-k,n-i+k}

1347: \end{equation} 

1348: is given by 

1349: \begin{equation}

1350: \F^{(k)}_i(M)=\textrm{proj}_{\theta_{i+k}}(\Lambda^k\Ld\o M)

1351: \end{equation}

1352: 

1353: Denote by $[\E_i^{(k)}], [\F_i^{(k)}]$ the induced homomorphisms of 

1354: the Grothendieck group $K(\O_n)=\oplus_{j\in \Z}K(\O_{j,n-j}).$

1355: Note that $[\E_i^{(k)}],[\F_i^{(k)}]$ map $K(\O_{j,n-j})$ to $0$ unless 

1356: $i=j.$ The following theorem 

1357: is proved in the same way as Theorem~\ref{thrm:EF-relation}. 

1358: 

1359: 

1360: \begin{theorem} 

1361: \label{thrm:EF-divided-relation} Under the abelian group isomorphism

1362: $$\Upsilon_n: K(\O_n)\to \vn$$ 

1363: the endomorphism $[\E^{(k)}_i],$ resp. $[\F^{(k)}_i]$ 

1364: of $K(\O_n)$ coincides with the endomorphism of $\vn$ given by the action of   

1365: $E^{(k)}1_{2i-n}\in \dU,$ resp. $F^{(k)}1_{2i-n}.$

1366:  In other words, the following 

1367: diagrams are commutative 

1368: 

1369: \[

1370: \begin{array}{ccc}

1371: K(\O_{n}) & \stackrel{\Upsilon_n}{\longrightarrow} & \vn  \\

1372: \Big\downarrow\vcenter{ \rlap{$[\E^{(k)}_i]$}}     & & 

1373: \Big\downarrow\vcenter{ \rlap{$E^{(k)}1_{2i-n}$}}  \\

1374: K(\O_n) & \stackrel{\Upsilon_n}{\longrightarrow} &  \vn  

1375: \end{array}

1376: \]

1377: 

1378: \[

1379: \begin{array}{ccc}

1380: K(\O_n) & \stackrel{\Upsilon_n}{\longrightarrow} & \vn  \\

1381: \Big\downarrow\vcenter{ \rlap{$[\F^{(k)}_i]$}}     & & 

1382: \Big\downarrow\vcenter{ \rlap{$F^{(k)}1_{2i-n}$}}  \\

1383: K(\O_n) & \stackrel{\Upsilon_n}{\longrightarrow} &  \vn  

1384: \end{array}

1385: \]

1386: 

1387: \end{theorem}

1388: 

1389: Let $S$ be the set $\l E^{(a)}1_j,F^{(a)}1_j | a,j\in \Z\r.$ It 

1390: is a subset of $\dU.$  To each element of $S$ 

1391: we can now associate a projective functor from the category $\O_n$ to 

1392: itself as follows. The functor associated to an element  $x\in S $ will be 

1393: denoted $f_n(x).$ 

1394: 

1395: \begin{eqnarray*}

1396: & f_n( E^{(a)}1_{j})=& \left\{ 

1397: \begin{array}{ll} 

1398:      \E^{(a)}_{\frac{j+n}{2}} & \mbox{ if $j+n=0(\mbox{mod }2)$} \\

1399:       0                        & \mbox{ otherwise } 

1400: \end{array}

1401: \right.          \\

1402: & f_n( F^{(a)}1_{j})=& \left\{ 

1403: \begin{array}{ll} 

1404:      \F^{(a)}_{\frac{j+n}{2}} & \mbox{ if $j+n=0(\mbox{mod }2)$} \\

1405:       0                        & \mbox{ otherwise } 

1406: \end{array}

1407: \right. 

1408: \end{eqnarray*}

1409: 

1410: Let $c$ be an arbitrary  product $x_1\dots x_m$ of elements of $S.$ 

1411: To $c$ we associate a functor, denoted 

1412: $f_n(c),$ from $\O_n$ to $\O_n$ by 

1413: $$f_n(c) = f_n(x_1) \circ \dots \circ f_n(x_m).$$

1414: 

1415: Proposition~\ref{prop:projective-functors-isomorphic} implies 

1416: 

1417: \begin{theorem}

1418: Let $c_1,\dots, c_s,d_1,\dots, d_t$ be arbitrary products of 

1419: elements of $S.$ 

1420: The endomorphisms of $\vn$ induced by the elements 

1421:  $c_1+\dots +c_s$ and 

1422: $d_1+\dots +d_t$ of $\dU$ coincide  if and only if 

1423: the functors $\oplus_{i=1}^s f_n(c_i)$ and $\oplus_{j=1}^t f_n(d_i)$ 

1424: are isomorphic. 

1425: \end{theorem}

1426: 

1427: 

1428: 

1429: \begin{corollary} There exist isomorphisms of projective functors

1430: 

1431: \begin{eqnarray*}

1432: \E^{(b)}_{i+a}\circ\E^{(a)}_i  \cong  

1433: (\E^{(a+b)}_i)^{\oplus \binom{a+b}{a}}  & & \\

1434: \F^{(b)}_{i-a}\circ\F^{(a)}_i  \cong 

1435: (\F^{(a+b)}_i)^{\oplus \binom{a+b}{a}} \\ 

1436: \oplus_{k=0}^{min(a,b)}(\E^{(a-k)}_{i-b+k}\circ \F^{(b-k)}_{i}

1437: )^{\oplus\binom{n-i-a+b}{k}} \cong   & & \\ 

1438: \cong \oplus_{l=  0}^{min(a,b)}  

1439:  (\F^{(b-l)}_{i+a-l}\circ \E^{(a-l)}_{i})^{\oplus

1440: \binom{i}{l}} & & 

1441: \end{eqnarray*}

1442: \end{corollary} 

1443: 

1444: 

1445: \emph{Remark: } We expect that functor isomorphisms in the above 

1446: theorem can be made canonical so that they satisfy certain relations, 

1447: including associativity relations. 

1448: 

1449: 

1450: 

1451: \subsubsection{ Canonical basis of $\dU$ and indecomposable 

1452:   projective functors } 

1453: \label{canbas-indecomp}

1454: 

1455: 

1456: Recall that the canonical basis $\dB$ of $\dU$ is given by 

1457: \begin{eqnarray*} 

1458: E^{(a)}1_{-i}F^{(b)} & \textrm{ for } & a,b,i\in \N, i\ge a+b \\

1459: F^{(b)}1_i E^{(a)}   & \textrm{ for } & a,b,i\in \N, i > a+b 

1460: \end{eqnarray*}

1461: 

1462: Conveniently, each element of $\dB$ is a product of 

1463: $E^{(a)}1_i, F^{(a)}1_i$ for various $a$ and $i,$ specifically, 

1464: \begin{eqnarray*} 

1465: E^{(a)}1_{-i}F^{(b)}=E^{(a)}1_{-i}F^{(b)}1_{-i+2b} & 

1466: \textrm{ for } & a,b,i\in \N, i\ge a+b \\

1467: F^{(b)}1_i E^{(a)}=F^{(b)}1_i E^{(a)}1_{i-2a}

1468:    & \textrm{ for } & a,b,i\in \N, i > a+b 

1469: \end{eqnarray*}

1470: 

1471: In the previous section to each such product and each $n=1,2,\dots $ 

1472: we associated a 

1473: projective functor. Therefore, we can associate a projective functor 

1474: to each element of the canonical basis $\dB$ of $\dU.$ 

1475: To $x\in\dB$ we associate a projective functor $f_n(x)$ 

1476: from $\O_n$ to $\O_n$ by the following rule. 

1477: 

1478: \begin{eqnarray*} 

1479: f_n(E^{(a)}1_{-i}F^{(b)})& = & \E^{(a)}_{\frac{-i+n}{2}}

1480: \F^{(b)}_{\frac{-i+n}{2}+b}   \\

1481:  & \textrm{ for } & a,b,i\in \N, i\ge a+b, i+n=0(\mbox{mod }2), \\

1482: f_n(F^{(b)}1_i E^{(a)}) &  = & \F^{(b)}_{\frac{i+n}{2}} 

1483: \E^{(a)}_{\frac{i+n}{2}-a} \\

1484:    & \textrm{ for } & a,b,i\in \N, i > a+b, i+n=0(\mbox{mod }2), \\

1485: f_n(E^{(a)}1_{-i}F^{(b)}) & = & f_n(F^{(b)}1_i E^{(a)}) = 0  \\

1486:    & \textrm{ for } & i+n = 1(\mbox{mod }2). 

1487: \end{eqnarray*}

1488: 

1489:  

1490: 

1491: In this way to each canonical basis element $b\in \dB$ there is associated 

1492: an exact functor 

1493: $$f_n(b): \O_n \to \O_n  $$

1494: 

1495: The multiplication in $\dU$ correspond to composition of projective 

1496: functors: for $x,y\in \dB$ the product $xy$ is 

1497: a linear combination of elements of $\dB$ with 

1498: integral nonnegative coefficients 

1499:  $xy=\sum_{z\in \dU}m_{x,y}^{z}z.$

1500: In turn, the  functor $f_n(x)\circ f_n(y)$ decomposes as a direct sum of 

1501: functors $f_n(z)$ with multiplicities $m_{x,y}^z$: 

1502: 

1503: $$f_n(x)f_n(y)=\oplus_{z\in \dB}f_n(z)^{\oplus m_{x,y}^z}$$

1504: 

1505: \begin{theorem}

1506: \label{thrm:indecomposable-proj-functors} Fix $n\in \N.$ 

1507: Let $x\in \dB.$ Then the projective functor $f_n(x)$ is either $0$ or 

1508: isomorphic to an indecomposable projective functor. Moreover, for 

1509: each indecomposable projective functor $A:\O_n\to \O_n$ 

1510: there exists exactly one $x\in \dB$ such that $f_n(x)$ is isomorphic to 

1511: $A.$ 

1512: \end{theorem}

1513: 

1514: \emph{Proof: }Let $x\in \dB_{2j-n}.$

1515: Recall that in our notations the dominant Verma module in $\O_{j,n-j}$

1516: is $M(1^j0^{n-j}).$  

1517: From the properties of projective functors we know that 

1518: $f_n(x)M(1^j0^{n-j})$ is a projective module in $\O_n,$

1519: $f_n(x)$ is the 

1520: trivial functor if and only if $f_n(x)M(1^j0^{n-j})$ is the trivial 

1521: module, and

1522: $f_n(x)$ is an indecomposable projective functor if and only if 

1523: $f_n(x)M(1^j0^{n-j})$ is an indecomposable projective module. 

1524: Isomorphism classes of projective 

1525: modules in the category of highest weight modules are determined by 

1526: their images in the Grothendieck group. Thus, all computations to 

1527: check whether $f_n(x)M(1^j0^{n-j})$ is indecomposable, trivial, etc. 

1528: can be done in the Grothendieck group of the category $\O_n.$

1529: We claim that 

1530: $[f_n(x)M(1^j0^{n-j})]=0$ or $[f_n(x)M(1^j0^{n-j})]=[P(0^k1^l0^m1^s)]$ 

1531: for some $k,l,m,s$  where, we recall, 

1532: $P(0^k1^l0^m1^s)$ denotes the indecomposable projective cover of 

1533: the simple module with highest weight 

1534: $e_{l+1}+\dots +e_{k+l}+e_{k+l+m+1}+\dots +e_{k+l+m+s}-\rho$ 

1535: (see Section~\ref{section-def-gln}).   

1536: 

1537: Due to the isomorphism $\Upsilon_n$ between the Grothendieck group 

1538: of $\O_n$ and the abelian group $\vn$ 

1539: and Proposition~\ref{prop:simple-projective-bases} we can

1540:  work with the latest group instead. 

1541: The notations of Section~\ref{sec:canbasis-formulas}

1542:  are used below. 

1543: 

1544: \begin{prop} 

1545: \begin{eqnarray*}

1546: E^{(a)}1_{i} v(1^b0^c) & = &  \\

1547:  & = &  

1548: \left\{

1549: \begin{array}{ll}  \binom{c}{a} 

1550: (\mbox{Id}^{\otimes b}\o p_c) v(1^{b+a}0^{c-a}) & \mbox{ if }

1551: i=b-c\mbox{ and }c\ge a \\ 

1552: 0  & \mbox{otherwise}

1553: \end{array} \right.     \\ 

1554: F^{(a)}1_{i} v(1^b0^c)  & =  &  \\ 

1555:  & =  & \left\{

1556: \begin{array}{ll}  \binom{b}{a} 

1557: (p_b\o \mbox{Id}^{\otimes c}) v(1^{b-a}0^{c+a}) & \mbox{ if }

1558: i=b-c\mbox{ and }b\ge a\\ 

1559: 0  & \mbox{otherwise}

1560: \end{array} \right.  

1561: \end{eqnarray*}

1562: \end{prop} 

1563: 

1564: \emph{Proof:} Clearly, $1_i v(1^b0^c)= v(1^b0^c)$ if $i=b-c$ and 

1565:   $1^iv(1^b0^c)=0$ otherwise. $E^{(a)}$ acts on $v^(1^b0^c)$ in 

1566:   the following way 

1567:  $$ E^{(a)} v(1^b0^c)= \sum_{I} v(1^b I)$$

1568:   where the sum is over all sequences $I$ that contain $a$ ones and 

1569:   $c-a$ zeros. Proposition follows.  $\square$ 

1570: 

1571: Let $a,b,i$ be non-negative integers with $i\ge a+b.$ 

1572: Then $E^{(a)}1_{-i}F^{(b)}$ is 

1573: an element of the canonical basis $\dB.$ We compute its action on the 

1574: element $v(1^c0^d)$ of $V_1^{\o (c+d)}.$ We restrict to the 

1575: case  $i=2b+d-c$ and $c\ge b$ 

1576:  as otherwise the result is $0.$ The condition $i\ge a+b$ 

1577: can now be written as $c-b\le d-a.$ 

1578: 

1579: \begin{eqnarray*}

1580: & & E^{(a)}1_{-i}F^{(b)} v(1^c0^d) = \\

1581: & & E^{(a)}\binom{c}{b}(p_c\o Id^{\o d})v(1^{c-b}0^{d+b}) = \\

1582: & & \binom{c}{b}(p_c\o Id^{\o d})E^{(a)}v(1^{c-b}0^{d+b}) = \\

1583: & & \binom{c}{b}(p_c\o Id^{\o d})

1584:  \binom{d+b}{a}(Id^{\o (c-b)}\o p_{d+b})v(1^{c-b+a}0^{d+b-a}) = \\

1585:  & & \binom{c}{b}\binom{d+b}{a}(p_c\o Id^{\o d})

1586:  (Id^{\o (c-b)}\o p_{d+b})v(1^{c-b+a}0^{d+b-a}) = \\

1587: & &  p(0^{b}1^{c-b}0^{d-a}1^{a}) 

1588: \end{eqnarray*}

1589: The last equality follows from formula (\ref{fla:0101}) (substituting 

1590: $j=b, k=c-b, l=d-a, m=a$) and the condition $c-b\le d-a.$ 

1591: 

1592: In the same fashion, for $a,b,c,d,i\in \Z_+$ such that 

1593: $i=c-d+2a, i\ge a+b$ (which implies $c-b\ge d-a$) 

1594: we get 

1595: \begin{equation}

1596: F^{(b)}1_iE^{(a)} v(1^c0^d) = p(0^b1^{c-b}0^{d-a}1^a)

1597: \end{equation}

1598: 

1599: Thus, for any $x\in \dB$ and $c,d\in \Z_+$ the element 

1600: $x v(1^c0^d)\in \vn$ is either $0$ or equal to the image under $\Upsilon_n$ 

1601: of the Grothendieck class of the indecomposable projective 

1602: module $P(0^k1^l0^m1^s)$ for some quadruple $(k,l,m,s).$ 

1603: 

1604: Therefore, for any $x\in \dB_{2j-n}$ and 

1605: a dominant Verma module $M(1^j0^{n-j})$ the projective module 

1606: $f_n(x)M(1^j0^{n-j})$ is 

1607: either the trivial or an indecomposable projective module, i.e. 

1608: the projective functor $f_n(x)$ is either trivial or indecomposable 

1609: projective. All other statements of 

1610: Theorem~\ref{thrm:indecomposable-proj-functors}

1611:  follow easily from the above analysis and the

1612: classification of projective functors (see [BG]). $\square$

1613: 

1614: 

1615: 

1616: 

1617: \subsubsection{Comultiplication}

1618: 

1619: \label{sect-comult} 

1620: 

1621: In previous sections we studied projective functors in categories $\O_n.$ 

1622: We established that on the Grothendieck group level these functors descend 

1623: to the action of generators of $\dU$ on the $n$-th tensor power of 

1624: the fundamental representation $V_1$ and that the composition of functors

1625: descends to the multiplication in $\dU.$ Yet, we do not have a functor

1626:  realization of the whole algebra $\dU$, 

1627: only of its finite-dimensional quotients, also called Schur quotients, that 

1628: are the homomorphic images or $\dU$ in the endomorphisms of $\vn.$  

1629: It is unconvenient to think about comultiplication in $\dU$ if 

1630: only some of its finite quotients are available. We notice, although, that 

1631: comultiplication allows one to introduce a module structure in a  

1632: tensor product and, with a categorification of $\vn$ at hand, we can 

1633: try  to construct functors between  

1634: $\O_n \times \O_m$ and $\O_{n+m}$  corresponding to 

1635: module isomorphism 

1636: \begin{equation}

1637: \label{comultv}

1638: V_1^{\o n}\o V_1^{\o m} \cong V_1^{\o (n+m)}

1639: \end{equation}

1640: Before we considered projective 

1641: projective functors 

1642: $\E_i^{(a)}, \F^{(a)}_i: \O_n \lra \O_n$ for a fixed $n$ and in the notations 

1643: for these functors we suppressed the dependence on $n$. In this section

1644: the rank of $\mf{gl}$ will vary (we'll have functors between 

1645: categories $\O(\gl)\times \O(\mf{gl}_m)$ and $\O(\mf{gl}_{n+m})$ and we 

1646: redenote the functors $\E_i^{(a)}, \F^{(a)}_i:\O_n \lra \O_n $ by 

1647: $\E_{i,n}^{(a)}, \F_{i,n}^{(a)}.$ 

1648: 

1649: 

1650: For the rest of the section we fix positive integers $n$ and $m.$ 

1651: Let $\mf{p}$ be the 

1652: maximal parabolic subalgebra of 

1653: $\mf{gl}_{n+m}$ that contains $\mf{gl}_n\oplus \mf{gl}_m$ and 

1654: $\mf{n}_+,$ i.e. $\mf{p}$ is the standard subalgebra of block uppertriangular 

1655: matrices. Denote by $\mf{u}$ the nilpotent radical of $\mf{p}.$ 

1656: Denote by $\Mp$ the category of finitely-generated 

1657: $U(\mf{p})$-modules that are $\mf{h}$-diagonalizable and 

1658: $U(\mf{n}_+)$-locally nilpotent. 

1659: 

1660: Let $\Ind$ be the induction functor 

1661: $\Ind:\Mp \to \O(\mf{gl}_{n+m}).$

1662: To a $U(\mf{p})$-module 

1663: $N\in \Mp$ it associates 

1664: the $U(\mf{gl}_{n+m})$-module 

1665: $U(\mf{gl}_{n+m})\otimes_{U(\mf{p})}N.$ 

1666: Recall that we denoted by $F_V$ the functor of tensoring with a finite 

1667:  dimensional $\mf{g}$-module $V.$ 

1668: 

1669: \begin{lemma} \label{product-and-induction}

1670: Let $V$ be a finite dimensional  $\mf{gl}_{n+m}$-module. Denote by

1671: $F'_V$ the functor from $\Mp$ to $\Mp$ 

1672: given by tensoring with $V,$ considered as a $U(\mf{p})$-module.

1673: Then there is  a canonical isomorphism of functors

1674: \begin{equation} 

1675: F_V\circ \Ind \cong \Ind \circ F'_V

1676: \end{equation} 

1677: \end{lemma}

1678: 

1679: \emph{Proof: }

1680: For $N\in \Mp$ we have natural isomorphisms

1681: \begin{eqnarray*}

1682: & & \textrm{Hom}_{U(\mf{gl}_{n+m})}(\Ind \circ F'_V(N),F_V\circ \Ind (N)) \\

1683: & = & \textrm{Hom}_{U(\mf{p})}(F'_V(N),F_V\circ \Ind (N)) = \\

1684: & = & \textrm{Hom}_{U(\mf{p})}(V\o N, V\o   (U(\mf{gl}_{n+m})

1685: \otimes_{U(\mf{p})} N )),

1686: \end{eqnarray*}

1687: the first isomorphism coming from the adjointness of induction 

1688: and restriction functors.  

1689: The third  hom-space has a distinguished element coming from the 

1690: $U(\mf{p})$-module map $V \otimes N \lra  V\otimes (U(\mf{gl}_{n+m})

1691: \otimes_{U(\mf{p})} N )$ given  by $v\o n \longmapsto v \o ( 1 \o n).$ 

1692: It is easy to see that the corresponding map of $U(\mf{gl}_{n+m})$ modules 

1693: 

1694: $$\Ind \circ F'_V(N) \lra F_V \circ \Ind (N)$$ 

1695: is an isomorphism. $\square$ 

1696: 

1697: Let $Y$ be the one-dimensional representation of $\mf{gl}_n\oplus 

1698: \mf{gl}_m$ which has weight $\frac{m}{2}(e_1+\dots + e_n)$ as 

1699: a representation of $\mf{gl}_n$ and weight $\frac{-n}{2}(e_1+\dots + e_m)$ as 

1700: a representation of $\mf{gl}_m.$ 

1701: 

1702: Let $\Cfun_0$ be the functor $\O_n\times \O_m\lra \Mp$ defined as 

1703: follows. Tensor a product $M\times N \in \O_n\times \O_m$

1704: with $Y,$ and then make $Y\o M \o N$ 

1705: into a $\mf{p}$-module with the trivial action of the nilpotent radical 

1706: $\mf{u}$ of $\mf{p}.$  

1707: 

1708: Define $\Cfun: \O_n \times \O_m \lra \O_{n+m}$ to be the composition 

1709: of $\Cfun_0$ and $\Ind:$ 

1710: 

1711: $$\Cfun = \Ind \circ \Cfun_0.$$

1712: 

1713: Bifunctor $\Cfun$ is exact and when  

1714: applied to a product of Verma modules 

1715: $M(a_1\dots a_n)\times M(b_1\dots b_m)$ produces the Verma module 

1716: $M(a_1\dots a_n b_1\dots b_m).$ Therefore, we have a commutative diagram 

1717: of isomorphisms of abelian groups:  

1718: 

1719: $$

1720: \begin{array}{ccc}

1721:   K(\O_n)\times K(\O_m) & \stackrel{\Upsilon_n \times \Upsilon_m}

1722:   {\lra}  &  V_1^{\o n}\times V_1^{\o m} \\

1723:   \downarrow [\Cfun]  &   &    \|    \\

1724:   K(\O_{n+m})  & \stackrel{\Upsilon_{n+m}}{\lra} & V_1^{\o (n+m)} 

1725: \end{array} 

1726: $$

1727: 

1728: Let us present more evidence that  

1729:  $\Cfun$ is the bifunctor that categorifies the 

1730: identity (\ref{comultv}) by showing how to use $\Cfun$ to 

1731: categorify the comultiplication 

1732: formula $\Delta E= E\o 1+ 1\o E.$

1733: 

1734: Recall from Section~\ref{functors-EF} that $\Lo$ denotes the $n$-dimensional 

1735: fundamental representation of $\mf{gl}_n.$ We can make $\Lo$ into 

1736: a $U(\mf{p})$-module by making $\mf{u}\oplus \mf{gl}_m\subset \mf{p}$ 

1737: act by $0.$ Similarly, $L_m$ and $L_{n+m}$ are $U(\mf{p})$-modules in 

1738: the natural way and we have a short

1739:  exact sequence of $U(\mf{p})$-modules

1740: \begin{equation}

1741: \label{modules-exact}

1742: 0 \to \Lo \to L_{n+m} \to L_m \to 0, 

1743: \end{equation}

1744: which gives rise to a short exact sequence of functors 

1745: from $\Mp$  to $\Mp$: 

1746: 

1747: $$0\to F_{\Lo}'\to F_{L_{n+m}}'\to F_{L_{m}}'\to 0,$$

1748: where $F_{\Lo}'$ is the functor of tensoring with $\Lo,$ considered 

1749: as a $U(\mf{p})$-module, and so on. 

1750: 

1751: Functors $\Ind$ and $\Cfun_0$ are exact; and 

1752: composing with them we obtain an  

1753: exact sequence

1754: $$0\to \Ind \circ F_{\Lo}' \circ \Cfun_0 \to \Ind \circ 

1755: F_{L_{n+m}}'\circ \Cfun_0\to 

1756: \Ind \circ F_{L_{m}}'\circ \Cfun_0\to 0$$

1757: of functors. 

1758: We have isomorphisms of functors from $\O_n\times \O_m$ to 

1759: $\Mp:$  

1760: \begin{eqnarray*}

1761: F_{\Lo}' \circ \Cfun_0 & \cong & 

1762: \Cfun_0 \circ (F_{\Lo}\times Id) \\

1763: F_{L_m}' \circ \Cfun_0 & \cong & 

1764: \Cfun_0 \circ (Id \times F_m) 

1765: \end{eqnarray*}

1766: and the isomorphism (see Lemma~\ref{product-and-induction}) 

1767: $$\Ind \circ F_{L_{n+m}}' \cong F_{L_{n+m}} \circ \Ind $$

1768: 

1769: 

1770: We thus get an exact sequence of functors 

1771: \begin{equation}

1772: 0\to 

1773: \Ind \circ \Cfun_0 \circ (F_{\Lo}\times Id)\to 

1774: F_{L_{n+m}}\circ \Ind \circ \Cfun_0\to 

1775: \Ind \circ \Cfun_0 \circ (Id \times F_{L_m}) 

1776: \to 0,

1777: \end{equation}

1778: and recalling that $\Cfun=\Ind \circ \Cfun_0$ we obtain an exact sequence 

1779: \begin{equation}

1780: \label{exact-seq} 

1781: 0\to 

1782: \Cfun \circ (F_{\Lo}\times Id)\to 

1783: F_{L_{n+m}}\circ \Cfun \to 

1784: \Cfun \circ (Id \times F_{L_m}) 

1785: \to 0

1786: \end{equation}

1787: $\E_{i,n}$ is a direct summand of the functor $F_{\Lo}$ 

1788: and from this we derive 

1789: 

1790: \begin{prop} Exact sequence (\ref{exact-seq}) contains the following exact 

1791: sequence of functors as a direct summand

1792: \begin{equation}

1793: 0\to 

1794: \Cfun \circ (\E_{i,n}\times Id)\to 

1795: \E_{i+j,n+m}\circ  \Cfun \to 

1796: \Cfun \circ (Id \times \E_{j,m}) 

1797: \to 0. 

1798: \end{equation}

1799: \end{prop} 

1800: This exact sequence can be considered as a categorification of 

1801: the comultiplication formula 

1802: \begin{equation}

1803: \Delta(E)=E\o 1 +1 \o E

1804: \end{equation}  

1805: In the same fashion, using the exact sequence dual to (\ref{modules-exact})

1806: we obtain an exact sequence of 

1807: functors 

1808: \begin{equation} 

1809: 0\to 

1810: \Cfun \circ (Id \times \F_{j,m})\to 

1811: \F_{i+j,n+m}\circ \Cfun \to 

1812: \Cfun \circ (\F_{i,n}\times Id) \to 0

1813: \end{equation} 

1814: 

1815: We proceed to ``categorify'' the comultiplication rules for 

1816: the divided powers $\E^{(a)}_{i,n}, \F^{(a)}_{i,n}.$ 

1817: 

1818: The $\mf{gl}_{n+m}$-module $\Lambda^a L_{n+m},$ 

1819: considered as a $\mf{p}$-module, 

1820: admits a filtration 

1821: $\Lambda^a L_{n+m}=G_{a+1}\supset G_{a} \supset \dots \supset G_{0}=0$ 

1822: such that the module $G_{k+1}/G_{k}, k=0,\dots ,a$ is 

1823: isomorphic to $\Lambda^{a-k}\Lo\otimes \Lambda^{k} L_m.$

1824: Therefore, we obtain 

1825: 

1826: \begin{prop} 

1827:  The exact functor $\E^{(a)}_{i+j,n+m}\circ \Cfun$ has a 

1828: filtration by exact functors 

1829: $$\E^{(a)}_{i+j,n+m}\circ \Cfun=\mc G_{a+1}\supset \mc G_{a} \supset \dots 

1830: \supset \mc G_{0}=0$$

1831: together with short exact sequences of functors

1832: $$0\to \mc G_{k}\to \mc G_{k+1}\to \Cfun \circ 

1833: (\E^{(a-k)}_{i,n}\times \E^{(k)}_{j,m}) \to 0.$$ 

1834: 

1835:  The exact functor $\F^{(a)}_{i+j,n+m}\circ \Cfun$ has a 

1836: filtration by exact functors 

1837: $$\F^{(a)}_{i+j,n+m}\circ \Cfun=\mc G_{a+1}\supset \mc G_a \supset \dots 

1838: \supset \mc G_{0}=0$$

1839: such that the sequences below are exact

1840: $$0\to \mc G_k\to \mc G_{k+1}\to \Cfun \circ 

1841: (\F^{(k)}_{i,n}\times \F^{(a-k)}_{j,m}) \to 0.$$ 

1842: \end{prop}

1843: 

1844: This can be considered as a categorification of the comultiplication 

1845: formulas (\ref{comult-div-E}) and (\ref{comult-div-F}). 

1846: 

1847: \emph{Remark:} 

1848: In [Gr] Grojnowsky uses perverse sheaves to categorify the comultiplication 

1849: rules for $U_q(\mf{sl}_k).$ His approach appears to be Koszul dual to 

1850: ours. 

1851: 

1852: 

1853: 

1854: 

1855: 

1856: \subsection{Zuckerman functors and Temperley-Lieb algebra} 

1857: \label{TL-and-Z}

1858: 

1859: \subsubsection{Computations with Zuckerman functors} 

1860: \label{comp-Z}

1861: 

1862: 

1863: Let $\mf{g}_i, 1\le i\le n-1$ be a subalgebra of $\mf{gl}_n$ 

1864: consisting of matrices that 

1865: can have non-zero entries only on intersections of $i$-th or $(i+1)$-th 

1866: rows with $i$-th or $(i+1)$-th columns. $\mf{g}_i$ is isomorphic to 

1867: $\mf{gl}_2.$ Denote by $\O_{k,n-k}^i$ the subcategory of $\O_{k,n-k}$ 

1868: consisting of locally $U(\mf{g}_i)$-finite modules. 

1869: Denote by $\Gamma_i: \O_{k,n-k}\lra \O^i_{k,n-k}$ the Zuckerman functor 

1870: of taking the maximal locally $U(\mf{g}_i)$-finite submodule, and by 

1871: $\mc{R}\Gamma_i$ the derived functor of $\Gamma_i.$ 

1872:  The derived functor goes from the bounded derived category 

1873:  $D^b(\O_{k,n-k})$ to $D^b(\O_{k,n-k}^i).$ 

1874: Denote by $\Gamma_i^j$ the cohomology functor 

1875: $\mc{R}^j\Gamma_i: \O_{k,n-k}\to \O_{k,n-k}^j.$ 

1876: This functor is zero if $j>2$ by Proposition~\ref{Zproperties}. 

1877: 

1878: 

1879: 

1880: Recall that for a sequence $a_1\dots a_n$ of zeros and ones with exactly 

1881: $k$ ones, the Verma module 

1882: $M(a_1\dots a_n)$ belongs to  $\O_{k,n-k}.$

1883: 

1884: 

1885: \begin{prop}

1886: \label{gamma-computed} 

1887: If $a_i=1,a_{i+1}=0,$ 

1888: $$\Gamma^2_i M(a_1\dots a_n)=M(a_1\dots a_n)/

1889: M(a_1\dots a_{i-1}01a_{i+2}\dots a_n)$$

1890: If $a_i=0,a_{i+1}=1,$ 

1891: $$\Gamma^1_i M(a_1\dots a_n)=M(a_1\dots a_{i-1}10a_{i+2}\dots a_n)/

1892: M(a_1\dots a_n)$$

1893: For all other values of $(i,j,a_1,\dots, a_n)$ with  

1894: $i\in {1,\dots , n-1},j\in \Z,a_1,\dots ,a_n\in \l 0,1\r $

1895: we have

1896: $$\Gamma^j_iM(a_1\dots a_n)=0.$$

1897: \end{prop} 

1898: 

1899: \emph{Proof:} Functors $\Gamma^0_i$ and $\Gamma^2_i$ have  a

1900: simple description as functors of taking the maximal $U(\mf{g}_i)$-locally 

1901: finite submodule/quotient. So the proposition is easy to check for $j\not= 1.$ 

1902: For $j=1$ it is a special case of Proposition 5.5 of [ES]. 

1903: $\square$ 

1904: 

1905:  The simple module 

1906: $L(a_1\dots a_n)$ with $a_1,\dots, a_n\in \l 0,1\r $ 

1907: and $a_1+\dots +a_n=k$ belongs to $\O^i_{k,n-k}$

1908: if and only if $a_i=1, a_{i+1}=0.$ 

1909: Therefore, a simple object of $\O_{k,n-k}$ cannot belong simultaneously 

1910: to $\O^i_{k,n-k}$ and $\O^{i+1}_{k,n-k}.$ 

1911: 

1912: \begin{corollary} If a module $M\in \O_{k,n-k}$ lies in both 

1913: $\O^i_{k,n-k}$ and $\O^{i+1}_{k,n-k},$ it is trivial.

1914: \end{corollary}

1915: 

1916: \begin{prop} For any $M\in \O_{k,n-k}^i, $   

1917: $$\Gamma^j_{i\pm 1}M=0 \mbox{ if }j\not=1$$

1918: \end{prop}

1919: 

1920: \emph{Proof: } We know that $\Gamma^j_{i\pm 1} 

1921: M$ can be nontrivial only for $0\le j\le 2.$ But $\Gamma^0_{i\pm 1}M$ 

1922: is the maximal locally $\mf{g}_{i\pm 1}$-finite submodule of $M$, 

1923: i.e. the zero module (by the last corollary). 

1924: Similarly, $\Gamma^2_{i\pm 1}M=0.$ 

1925: $\square$

1926: 

1927: 

1928: 

1929: \begin{corollary} 

1930: \begin{enumerate}

1931: 

1932: \item  Functors $\mc{R}\Gamma_{i\pm 1}[1]$ and 

1933: $\Gamma^1_{i\pm 1}$, restricted to the subcategory 

1934: $\O^i_{k,n-k},$ are naturally equivalent (more precisely, 

1935: the composition of 

1936: $\Gamma^1_{i\pm 1},$ restricted to $\O^i_{k,n-k},$ 

1937: with the embedding functor $\O^{i\pm 1}_{k,n-k}\lra 

1938: D^b(\O_{k,n-k}^{i\pm 1})$ 

1939: is equivalent to the restriction of the functor 

1940: $\mc{R}\Gamma_{i\pm 1}[1]$ to the subcategory $\O_{k,n-k}^i$ of 

1941: $D^b(\O_{k,n-k}^i).$)

1942: 

1943: 

1944: \item  The functor 

1945: $\Gamma^1_{i\pm 1}: \O_{k,n-k}^i \to \O_{k,n-k}^{i\pm 1}$

1946: is exact.

1947: \end{enumerate} 

1948: \end{corollary}

1949: 

1950: 

1951: Generalized Verma modules for $\mf{g}_i$ are isomorphic to 

1952: the quotients 

1953: $$M(a_1\dots a_{i-1}10a_{i+2}\dots a_n)/

1954:   M(a_1\dots a_{i-1}10a_{i+2}\dots a_n).$$

1955: Denote this quotient module by $M_i(a_1\dots a_{n-2}).$

1956: To simplify notations, if $a_1\dots a_{i-1}a_{i+2}\dots a_n$ is fixed, 

1957: we will denote $M(a_1\dots a_{i-1}10a_{i+2}\dots a_n)$ by $M_{10}$ and 

1958: $M(a_1\dots a_{i-1}01a_{i+2}\dots a_n)$ by $M_{01}.$ Then 

1959: $M_i(a_1\dots a_{n-2})=M_{10}/M_{01}.$ 

1960: 

1961: 

1962: \begin{prop} 

1963: $$\Gamma_{i\pm 1}^1

1964: M_i(a_1\dots a_{n-2})=M_{i\pm 1}(a_1\dots a_{n-2}).$$

1965: \end{prop} 

1966: 

1967: \emph{Proof: }We have an exact sequence

1968: $$0\to M_{01}\to M_{10}\to M_i\to 0$$

1969: which induces a long exact sequence of Zuckerman cohomology functors

1970: $$\dots \to \Gamma_{i-1}^jM_{01}\to \Gamma_{i-1}^jM_{10}\to 

1971: \Gamma_{i-1}^j M_i\to \Gamma_{i-1}^{j+1}M_{01}\to \dots $$

1972: Consider the following segment of this sequence

1973: \begin{equation}

1974: \label{five-terms}

1975: \Gamma_{i-1}^1M_{01}\to \Gamma_{i-1}^1M_{10}

1976: \to \Gamma_{i-1}^1M_i\to \Gamma_{i-1}^2 M_{01}

1977: \to \Gamma_{i-1}^2 M_{10} 

1978: \end{equation}

1979: From Proposition~\ref{gamma-computed}, we find that 

1980: $$\Gamma_{i-1}^1M_{01}=\Gamma_{i-1}^2M_{10}=0$$

1981: and (\ref{five-terms}) becomes a short exact sequence

1982: $$0 \to \Gamma_{i-1}^1M_{10}

1983: \to \Gamma_{i-1}^1M_i\to \Gamma_{i-1}^2 M_{01}

1984: \to 0 $$

1985: We proceed by considering two different cases: 

1986: 

1987: (i) If $a_{i-1}=0,$ then by Proposition~\ref{gamma-computed}

1988: we have $\Gamma_{i-1}^1M_{10}=M_{i-1}(a_1\dots a_{n-2}) $ and 

1989: $\Gamma_{i-1}^2 M_{01}=0.$ 

1990: 

1991: (ii) If $a_{i-1}=1,$ then 

1992: $\Gamma_{i-1}^2M_{01}=M_{i-1}(a_1\dots a_{n-2}) $ and 

1993: $\Gamma_{i-1}^1 M_{10}=0.$ 

1994: 

1995: In both cases we get $\Gamma_{i-1}^1M_i=

1996: M_{i-1}(a_1\dots a_{n-2}).$ The proof for $\Gamma_{i+1}$ is 

1997: the same. 

1998: $\square$

1999: 

2000: 

2001: \subsubsection{Equivalences of categories} 

2002: \label{equiv-of-cat} 

2003: 

2004: 

2005: Let $\varepsilon_i$ be the inclusion functor 

2006: $\O^i_{k,n-k}\to \O_{k,n-k}.$

2007: Consider a pair of functors 

2008: 

2009: \begin{eqnarray*}

2010: \Gamma_{i-1}^1\varepsilon_i & : 

2011: \O^i_{k,n-k}  & \to \O^{i-1}_{k,n-k} \\

2012: \Gamma_{i}^1\varepsilon_{i-1} & : 

2013: \O^{i-1}_{k,n-k}  & \to \O^{i}_{k,n-k} 

2014: \end{eqnarray*}

2015: 

2016: These two functors are exact, take generalized Verma modules to 

2017: generalized Verma modules and the compositions 

2018: $\Gamma_{i-1}^1 \varepsilon_i

2019: \Gamma_{i}^1 \varepsilon_{i-1}$ and 

2020: $\Gamma_{i}^1\varepsilon_{i-1}

2021: \Gamma_{i-1}^1\varepsilon_i$ are identities on generalized Verma 

2022: modules.  

2023: 

2024: \begin{theorem} 

2025: \label{functors-and-isotopies}

2026: Functors $\Gamma_{i - 1}^1\varepsilon_i$ 

2027: and $\Gamma_i^1\varepsilon_{i-1}$ 

2028: are equivalences of 

2029: categories $\O_{k,n-k}^i$ and $\O_{k,n-k}^{i- 1}.$

2030: The composition $\Gamma_{i}^1\varepsilon_{i- 1}

2031: \Gamma_{i- 1}^1\varepsilon_{i}$ is  isomorphic

2032:  to the identity functor from the category $\O_{k,n-k}^i$ to itself. 

2033: \end{theorem} 

2034: 

2035: \emph{Proof:} By Lemma~\ref{category-equivalence} 

2036: it remains to show that functors 

2037: $\Gamma_{i-1}^1\varepsilon_i$ and 

2038: $\Gamma_{i}^1\varepsilon_{i-1}$ are two-sided adjoint. 

2039: We recall that $\Gamma_i:\O_{k,n-k}\to \O^i_{k,n-k}$ is 

2040: right adjoint to the inclusion functor $\varepsilon_i.$ 

2041: Besides, in the derived category, 

2042: the derived functor $R\Gamma_i$ is isomorphic to the left adjoint of 

2043: the shifted inclusion functor $\varepsilon_i[2].$ 

2044: 

2045: 

2046: 

2047: Let $M,N\in \O^i_{k,n-k}.$ 

2048: We have natural vector space isomorphisms 

2049: \begin{eqnarray*}

2050: \mbox{Hom}((\Gamma_{i-1}^1\varepsilon_i )M , N ) & \cong & 

2051: \mbox{Hom}((R\Gamma_{i-1}[1]\circ \varepsilon_i) M , N ) 

2052: \\

2053: & \cong & 

2054: \mbox{Hom}(\varepsilon_i M , \varepsilon_{i-1}N[1] ) 

2055: \\

2056: & \cong & 

2057: \mbox{Hom}(M , (R\Gamma_{i})\varepsilon_{i-1}N[1] ) 

2058: \\

2059: & \cong & 

2060: \mbox{Hom}(M , (R\Gamma_{i})[1]\varepsilon_{i-1}N) 

2061: \\

2062: & \cong & 

2063: \mbox{Hom}(M , \Gamma_{i}^1\varepsilon_{i-1} N) 

2064: \end{eqnarray*}

2065: which imply that $\Gamma_{i-1}^1\varepsilon_i$  

2066:  is left adjoint to $\Gamma_i^1\varepsilon_{i-1}.$ 

2067: A similar computation tells us that  $\Gamma_{i-1}^1\varepsilon_i$  

2068:  is also right  adjoint to $\Gamma_i^1\varepsilon_{i-1}.$ 

2069: 

2070: \emph{Remark:} For an abelian category $\mc{A}$ the embedding functor 

2071: $\mc{A}\lra D^b(\mc{A})$ is fully faithful and for $M,N\in \mc{A}$ 

2072: we have canonical isomorphisms of hom-spaces 

2073: $$\mbox{Hom}_{\mc{A}}(M,N) = \mbox{Hom}_{D^b(\mc{A})}(M,N).$$ 

2074: That permits us in the chain of identities above to go freely between 

2075: hom spaces in $\O^i_{k,n-k}$ and $D^b(\O^i_{k,n-k}).$ $\square$

2076: 

2077: 

2078: 

2079: \begin{corollary} The categories $\O_{k,n-k}^i$ and $\O_{k,n-k}^j$ 

2080: are equivalent for all $i,j, 1\le i,j \le n-1.$

2081: \end{corollary} 

2082: 

2083: 

2084: \begin{prop} 

2085: \label{opiat-dvoika}

2086: Let $M\in \O_{k,n-k}^i.$ Then 

2087: $$\Gamma_i^j\varepsilon_iM=

2088: \left\{ 

2089: \begin{array}{ll} 

2090: M & \mbox{ if }j=0\mbox{ or }j=2 \\

2091: 0 & \mbox{ otherwise }

2092: \end{array} \right. 

2093: $$

2094: \end{prop}

2095: 

2096: \emph{Proof: } $\Gamma_i^0\varepsilon_iM$ 

2097: is the maximal $\mf{g}_i$-locally finite 

2098: submodule of $M$ and $\Gamma_i^2\varepsilon_iM$ 

2099: is the maximal $\mf{g}_i$-locally 

2100: finite quotient of $M.$ Thus, $\Gamma_i^0\varepsilon_iM=

2101: \Gamma_i^2\varepsilon_iM=M.$ 

2102: It remains to check that $\Gamma_i^1\varepsilon_iM=0$ 

2103: for any $M\in \O_{k,n-k}^i.$ 

2104: 

2105: The derived functor $\mc{R}\Gamma_i$ is exact and induces a map of 

2106: Grothendieck groups 

2107: $$[\mc{R}\Gamma_i]: K(\O_{k,n-k})\to K(\O_{k,n-k}^i).$$

2108: Using Proposition~\ref{gamma-computed} 

2109: we can easily compute $[\mc{R}\Gamma_i].$ 

2110: The bases of $K(\O_{k,n-k})$ and 

2111: $K(\O_{k,n-k}^i)$ consisting of images of Verma modules, respectively

2112: generalized Verma modules, are convenient for writing down $[R\Gamma_i]$ 

2113: explicitly: 

2114: 

2115: \begin{eqnarray*} 

2116: & [\mc{R}\Gamma_i(M(a_1\dots a_{i-1}00 a_i\dots a_{n-2}))]= & 

2117: 0\\

2118: & [\mc{R}\Gamma_i(M(a_1\dots a_{i-1}11 a_i\dots a_{n-2}))]= & 

2119: 0 \\

2120: & [\mc{R}\Gamma_i(M(a_1\dots a_{i-1}10 a_i\dots a_{n-2}))]= & 

2121: [M_i(a_1\dots a_{n-2})] \\

2122: & [\mc{R}\Gamma_i(M(a_1\dots a_{i-1}01 a_i\dots a_{n-2}))]= & 

2123: -[M_i(a_1\dots a_{n-2})] 

2124: \end{eqnarray*}

2125: where, we recall

2126: $$M_i(a_1\dots a_{n-2})=M(a_1\dots a_{i-1}10a_i\dots a_{n-2})/

2127: M(a_1\dots a_{i-1}01a_i\dots a_{n-2}).$$

2128: 

2129: Therefore, 

2130: $$[\mc{R}\Gamma_i\circ \varepsilon_i (M_i(a_1\dots a_{n-2}))]=

2131: [M_i(a_1\dots a_{n-2})]\oplus

2132: [M_i(a_1\dots a_{n-2})]$$

2133: and 

2134: $$[(\mc{R}\Gamma_i\circ\varepsilon_i) M]=[M]\oplus [M]$$

2135: for any $M\in \O^i_{k,n-k}.$ On the other hand, 

2136: $$[(\mc{R}\Gamma_i\circ\varepsilon_i)M]=

2137: [\Gamma_i^0\varepsilon_iM]-[\Gamma_i^1\varepsilon_iM]+

2138: [\Gamma_i^2\varepsilon_iM]=[M]-[\Gamma_i^1\varepsilon_iM]+[M].$$

2139: Thus, $[\Gamma_i^1\varepsilon_iM]=0$ for any $M\in \O^i_{k,n-k}$

2140:  and, hence, $\Gamma_i^1\varepsilon_iM=0$ for any $M\in \O^i_{k,n-k}.$ 

2141: $\square$ 

2142: 

2143: 

2144: 

2145: 

2146: \begin{prop} 

2147: \label{id-and-id}

2148: Restricting to the subcategory 

2149: $D^b(\O_{k,n-k}^i),$ 

2150: we have an equivalence of functors 

2151: \begin{equation}

2152: \label{double-id}

2153: \mc{R}\Gamma_i\circ\varepsilon_i\cong Id\oplus Id[-2].

2154: \end{equation}

2155: \end{prop} 

2156: 

2157: \emph{Proof: }Consider the natural transformation 

2158: $$e:Id \to \mc{R}\Gamma_i\circ \varepsilon_i.$$

2159: coming from the adjointness of $\mc{R}\Gamma_i$ and $\varepsilon_i.$ 

2160: Let $C_i$ be the functor which is the cone of $e.$ Then 

2161: by [MP], Lemma 3.2,  we have an isomorphism 

2162: of functors 

2163: $$\mc{R}\Gamma_i\circ \varepsilon_i=Id\oplus C_i.$$

2164: From Proposition~\ref{opiat-dvoika}

2165: $$C_i^jM=\left\{

2166: \begin{array}{ll} M  & \mbox{ if }j=2 \\

2167:                   0  & \mbox{ if }j\not= 2 

2168: \end{array}  \right. 

2169: $$

2170: Therefore, $C_i=Id[-2]$ and we have the isomorphism (\ref{double-id}).

2171: 

2172: $\square$ 

2173: 

2174: \subsubsection{A realization of the Temperley-Lieb algebra by functors} 

2175: \label{TL-by-functors} 

2176: 

2177: Define functors $\mc{V}_i, 1\le i\le n-1$ from 

2178: $D^b(\O_{k,n-k})$ to $D^b(\O_{k,n-k})$ by 

2179: \begin{equation}

2180: \mc{V}_i = \varepsilon_i \circ \mc{R}\Gamma_i [1]

2181: \end{equation} 

2182: 

2183: \begin{theorem}\label{TL-via-derived} 

2184: There are natural equivalences of functors 

2185: \begin{eqnarray}

2186: \label{dva-i-dva}

2187:   (\mc{V}_i )^2 & \cong & \mc{V}_i [-1] \oplus \mc{V}_i [1] \\

2188: \label{commutativity-from-afar}

2189:    \mc{V}_i \mc{V}_j & \cong & \mc{V}_j\mc{V}_i \mbox{ for }|i-j|>1 \\

2190: \label{little-braiding}

2191:    \mc{V}_i \mc{V}_{i\pm 1}\mc{V}_i & \cong & \mc{V}_i 

2192: \end{eqnarray}

2193: \end{theorem}  

2194: 

2195: \emph{Proof:} Isomorphism (\ref{dva-i-dva}) follows from 

2196: Proposition~\ref{id-and-id}. Isomorphism (\ref{commutativity-from-afar})

2197: is implied by a commutativity isomorphism $\Gamma_i\Gamma_j \cong 

2198: \Gamma_j \Gamma_i$ for $|i-j|>1.$ 

2199: The last isomorphism is a corollary of Theorem~\ref{functors-and-isotopies}. 

2200: $\square$ 

2201: 

2202: Summing over all $k$ from $0$ to $n$ we obtain functors 

2203: $$\mc{V}_i: D^b(\O_n) \lra D^b(\O_n)$$ 

2204: together with isomorphisms 

2205: (\ref{dva-i-dva})-(\ref{little-braiding}). On the Grothendieck 

2206: group level these functors descend to the action of the Temperley-Lieb 

2207: algebra $TL_{n,1}$ on $\vn.$ We thus have a categorification of 

2208: the action of the Temperley-Lieb algebra on the tensor product $\vn.$ 

2209: This action is faithful, so we can loosely say that we have 

2210: a categorification of the Temperley-Lieb algebra $TL_{n,1}$ itself.  

2211: In fact if we consider the shift by $1$ in the derived category as 

2212: the analogue of the multiplication by $q$, then we have categorified 

2213: the Temperley-Lieb algebra $TL_{n,q}.$ 

2214: 

2215: 

2216: Recall from Section~\ref{TL-algebra} that the element $U_i$ of the 

2217: Temperley-Lieb algebra is the product of morphisms $\cup_{i,n-2}$ and 

2218:  $\cap_{i,n}$ of the Temperley-Lieb category.  Morphisms 

2219:  $\cap_{i,n}$ and $\cup_{i,n-2}$  go  between objects $\overline{n}$ 

2220:  and $\overline{n-2}$ of the Temperley-Lieb category.  

2221:   On the other hand, 

2222:  the functor $\mc{V}_i,$ which categorifies $U_i,$ is the composition 

2223:   of functors $\varepsilon_i$  and $ \mc{R}\Gamma_i [1]$. 

2224:  We now explain how to modify the latter functors into functors 

2225:   between derived categories $D^b(\O_n)$ and $D^b(\O_{n-2})$ that 

2226:   cat be viewed as categorifications of morphisms $\cap_{i,n}$ and 

2227:   $\cup_{i,n-2}.$ 

2228: 

2229: Let $\varsigma_{n} $ be the functor from 

2230: $\O_{k-1,n-k-1}$ to $\O_{k,n-k}^1$ given as follows. 

2231: First we tensor an $M\in \O_{k-1,n-k-1}$ with the fundamental 

2232: representation $L_2$ of $\mf{gl}_2$ to get a 

2233: $\mf{gl}_2\oplus \mf{gl}_{n-2}$-module $L_2\otimes M.$ 

2234: Let $Y$ be the one-dimensional $\mf{gl}_2\oplus \mf{gl}_{n-2}$-module 

2235: with weight $\frac{n-3}{2}(e_1+e_2)$ relative to $\mf{gl}_2$ and 

2236: $-(e_1+\dots +e_{n-2})$ relative to $\mf{gl}_{n-2}.$ 

2237: Let $\mf{p}$ be the maximal parabolic subalgebra of $\mf{gl}_n$ that 

2238: contains $\mf{gl}_2\oplus \mf{gl}_{n-2}$ and the subalgebra of 

2239: upper triangular matrices. Then $Y\o (L_2\otimes M)$ is naturally a 

2240: $\mf{p}$-module with the nilradical of $\mf{p}$ 

2241: acting trivially. 

2242: Now we parabolically induce from $\mf{p}$ to $\mf{gl}_n$ and define

2243: $$\varsigma_n(M)=U(\mf{gl}_n)\otimes_{U(\mf{p})}(Y\o L_2\o M).$$

2244: 

2245: Let $\nu_n$ be the functor 

2246: from $\O_{k,n-k}^1$ to $\O_{n-1,n-k-1}$ defined as follows. 

2247: For an $M\in \O_{k,n-k}^1,$ take the sum of the weight subspaces

2248: of $M$ of weights $e_1+x_3e_3+\dots +x_ne_n-\rho_n,$ where 

2249: $x_3,\dots ,x_n\in \Z$ and $\rho_n$ is the half-sum of the positive roots 

2250: of $\mf{gl}_n.$ This direct sum is a $\mf{gl}_{n-2}$-module in a natural way. 

2251: Define $\nu_n(M)$ as the tensor product of this module with 

2252: the one-dimensional $\mf{gl}_{n-2}$-module of weight $e_1+\dots +e_{n-2}.$  

2253: 

2254: \begin{prop} Functors $\varsigma_n$ and $\nu_n$ are mutually inverse 

2255: equivalences of categories $\O_{k-1,n-k-1}$ and $\O_{k,n-k}^1.$ 

2256: \end{prop} 

2257: 

2258: We omit the proof as it is quite standard. 

2259: 

2260: \begin{corollary} The categories $\O_{k,n-k}^i,1\le i\le n-1$ and 

2261: $\O_{k-1,n-k-1}$ are equivalent. 

2262: \end{corollary} 

2263: 

2264: Denote by $\Xi_{n,i}$ the equivalence of categories 

2265: $$\Xi_{n,i}: \O_{k,n-k}^i \lra \O_{k-1,n-k-1}$$

2266: given by the composition 

2267: $$\Xi_{n,i}= \nu_n \circ \Gamma^1_1 \circ \varepsilon_2 

2268: \circ \Gamma^1_2 \circ \dots \varepsilon_{i-1}

2269: \circ \Gamma_{i-1}^1\circ \varepsilon_i,$$

2270:  i.e., $\Xi_{n,i}$ is the composition of equivalences

2271: of categories 

2272: $$ \O_{k,n-k}^i \stackrel{\cong}{\lra} \O_{k,n-k}^{i-1} 

2273: \stackrel{\cong}{\lra} \dots \stackrel{\cong}{\lra} 

2274:  \O_{k,n-k}^1 \stackrel{\cong}{\lra} \O_{k-1,n-k-1}. $$

2275: 

2276: Denote by $\Pi_{n,i}$ the equivalence of categories 

2277: $$\Pi_{n,i}: \O_{k,n-k} \lra \O_{k+1,n+1-k}^i$$

2278: given by 

2279: $$\Pi_{n,i}= \Gamma^1_i\circ \varepsilon_{i-1} 

2280: \circ \Gamma^1_{i-1} \circ \dots \circ \varepsilon_2\circ \Gamma^1_2

2281: \circ \varepsilon_1\circ\varsigma_n .$$

2282: 

2283: Denote the derived functors

2284: of these functors  by  $\mc{R}\Xi_{n,i}$ and $\mc{R}\Pi_{n,i}$.

2285: Define functors 

2286: \begin{eqnarray*} 

2287: \cap_{i,n}: &  D^b(\O_{k,n-k}) & \lra D^b(\O_{k-1,n-k-1}) \\

2288: \cup_{i,n}: &  D^b(\O_{k,n-k}) & \lra D^b(\O_{k+1,n+1-k})  

2289: \end{eqnarray*} 

2290: by 

2291: \begin{eqnarray}

2292: \label{here-is-cap}

2293: \cap_{i,n} & = & \mc{R}\Xi_{n,i}\circ \mc{R}\Gamma_i[1] \\

2294: \label{here-is-cup}

2295: \cup_{i,n} & = & \varepsilon_i \circ \mc{R}\Pi_{n,i}  

2296: \end{eqnarray} 

2297: Recall from Section~\ref{TL-algebra} defining relations

2298: (\ref{first-relation})-(\ref{last-relation}) for the Temperley-Lieb 

2299: category $TL.$ 

2300: 

2301: \begin{conjecture} There are natural equivalences 

2302: (\ref{first-relation})-(\ref{sixth-relation})

2303:  with functors $\cap_{i,n}$ and $\cup_{i,n}$ defined by 

2304: (\ref{here-is-cap}) and (\ref{here-is-cup}). 

2305: \end{conjecture} 

2306: 

2307: 

2308: The first two of these equivalences follow from the results of this section. 

2309: The relation (\ref{last-relation}) will become 

2310: $\cap_{i,n+2}\circ \cup_{i,n}\cong Id[1]\oplus Id[-1].$

2311: 

2312: We next state a conjecture on a functor realization of the category of 

2313:  tangles in $\R^3.$ Consider the following 3 elementary tangles 

2314: 

2315: \begin{center} \epsfig{file=dr11.eps} \end{center}

2316: 

2317: Every tangle in 3-space can be presented as a concatenation of these 

2318: elementary tangles. We associate to these 3 types of tangles the 

2319: following functors: 

2320: 

2321: To the 

2322: tangle (a) associate functor $\cap_{i,n}$ given by the formula (\ref{here-is-cap}). 

2323: 

2324: To the tangle (b) associate functor $\cup_{i,n}$ given by the formula 

2325: (\ref{here-is-cup}). 

2326: 

2327: To the tangle (c) associate functor $R_{i,n}$ from $D^b(\O_n)$ to $D^b(\O_n)$ 

2328: which is the cone of the adjointness morphism of functors 

2329: $$ \varepsilon_i \circ \mc{R}\Gamma_i \lra Id. $$

2330: Given a presentation $\alpha$ of a tangle $t$ as a composition 

2331: of elementary tangles of types (a)-(c), to $\alpha$ we associate the 

2332: functor $f(\alpha)$ which is the corresponding composition of 

2333: functors $\cap_{i,n}, \cup_{i,n}, R_{i,n}.$ 

2334: 

2335: \begin{conjecture}

2336: \label{tangle-invariants-sing}

2337:  Given two such presentations $\alpha, \beta$ of 

2338: a tangle $t$, functors $f(\alpha)$ and $f(\beta)$ are 

2339: isomorphic, up to shifts in the derived category. 

2340: \end{conjecture}

2341: 

2342: This conjecture, if true, will give us functor invariants of tangles. 

2343: Certain natural  transformations between these 

2344: functors, corresponding to adjointness morphisms between 

2345: the (shifted) identity functor and compositions 

2346: of $\varepsilon_i $ and $\mc{R}\Gamma_i,$ are expected to produce 

2347: invariants of 2-tangles and 2-knots (to be discussed elsewhere). 

2348: When we have a link rather than a tangle, the associated functor 

2349: goes between categories of complexes of vector spaces up to homotopies, 

2350: and the resulting invariants of links will be $\Z$-graded homology groups. 

2351: 

2352: 

2353: 

2354: 

2355: 

2356: 

2357: 

2358: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2359: %%%

2360: %%%        PARABOLIC CATEGORIES 

2361: %%%

2362: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2363: 

2364: 

2365: \section{Parabolic categories} 

2366: \label{parab-section}

2367: \subsection{Temperley-Lieb algebra and projective functors}

2368: \label{parab-tl} 

2369: \subsubsection{On and off the wall translation functors in parabolic 

2370:  categories} 

2371: \label{on-and-off} 

2372: 

2373: Let $\mu$ be an integral dominant regular weight and $\mu_i, i=1,\dots , 

2374: n-1$ an integral dominant subregular weight on the $i$-th wall. 

2375: Let $\O_{\mu}$ and $ \O_{\mu_i}$

2376: be the subcategories of $\O(\gl)$ of modules with generalized 

2377: central characters $\eta(\mu)$ and $\eta(\mu_i).$ 

2378: Then $\O_{\mu}$ is a regular block of $\O(\gl)$ and $\O_{\mu_i}$ is a 

2379: subregular block of $\O(\gl).$ 

2380: Verma modules $M_{\mu}$ and $M_{\mu_i}$ with highest weights 

2381: $\mu-\rho$ and $\mu_i-\rho$ are dominant Verma 

2382: modules in the corresponding categories. 

2383: 

2384: Let $\Tui, \Ti$ be translation functors 

2385: on and off the $i$-th wall

2386: 

2387: \begin{eqnarray*}

2388: \Tui : & \O_{\mu} \lra & \O_{\mu_i} \\

2389: \Ti : & \O_{\mu_i} \lra & \O_{\mu} 

2390: \end{eqnarray*}

2391: These functors are defined up to an isomorphism by the condition that 

2392: they are projective functors between $\O_{\mu}$ and 

2393: $\O_{\mu_i}$ and 

2394: \begin{enumerate}

2395: \item Functor $\Tui$ takes the Verma module $M_{\mu}$ to the Verma module 

2396: $M_{\mu_i}.$ 

2397: \item Functor $\Ti$ takes the Verma module $M_{\mu_i}$ to the projective 

2398: module $P_{s_i\mu}$ where $s_i$ is 

2399: the transposition $(i,i+1).$ On the Grothendieck group level, 

2400: $$[\Ti M_{\mu_i}] = [M_{\mu}] +  [M_{s_i\mu}].$$

2401: \end{enumerate} 

2402: 

2403: Let $\mf{p}_k$ be the maximal parabolic subalgebra of $\gl$ 

2404: such that $\mf{p}_k\supset \mf{n}_+\oplus \mf{h}$ and 

2405: the reductive subalgebra of $\mf{p}_k$ is $\mf{gl}_k\oplus \mf{gl}_{n-k}.$

2406: Let 

2407: $\O^{k,n-k}$, resp $\O^{k,n-k}_i$ 

2408: be the full subcategory of $\O_{\mu}$, resp. $\O_{\mu_i}$ 

2409: consisting 

2410: of modules that are locally $U(\mathfrak{p}_k)$-finite. 

2411: 

2412: From now on we fix $k$ between $0$ and $n$. Let $\tau_i^{i + 1}$ 

2413: be the composition of $\Ti$ and $T^{i+1}$: 

2414: $$\tau_i^{i+1}= T^{i+1}\circ \Ti.$$

2415: This is a functor from $\O_{\mu_i}$ to $\O_{\mu_{i+1}}.$ 

2416: Similarly, let $\tau_i^{i-1}$ be the functor from 

2417: $\O_{\mu_i}$ to $\O_{\mu_{i-1}}$ given by 

2418: $$\tau_{i}^{i-1}= T^{i-1} \circ \Ti.$$

2419:  Projective functors preserve subcategories of $U(\mf{p}_k)$-locally 

2420: finite modules and thus functors $\tau_i^{i \pm 1}$ restrict to 

2421: functors from $\O_i^{k,n-k}$ to $ \O_{i \pm 1}^{k,n-k}.$ 

2422: Our categorification of the Temperley-Lieb algebra by projective functors 

2423: is based on the following beautiful result of Enright and Shelton:  

2424: 

2425: \begin{theorem}\label{project-equivalence} 

2426: Functors $\tau_i^{i \pm 1}$ establish equivalences 

2427: of categories $\O_i^{k,n-k}$ and $\O_{i\pm 1}^{k,n-k}.$ 

2428: \end{theorem}

2429: 

2430: \emph{Proof:} See [ES], Lemma 10.1 for a proof of a 

2431: slightly more general statement. 

2432: An alternative proof that we give below uses  

2433: 

2434: \begin{lemma} The functor $\tau_{i+1}^i \tau_i^{i+1}$ restricted to 

2435: $\O_i^{k,n-k}$ is isomorphic to the identity functor.  

2436: \end{lemma} 

2437: 

2438: \emph{Proof} We first study this functor as a projective functor from 

2439: the subregular block $\O_{\mu_i}$ to itself. We will  show that this 

2440: functor is a direct sum of the identity functor and another projective 

2441: functor that vanishes when restricted to $\O_i^{k,n-k}.$ 

2442:  

2443: An isomorphism class of a projective 

2444: functor is determined by its action on the dominant Verma module. 

2445: So let us compute the action of $\tau_{i+1}^i\tau_i^{i+1}$ 

2446: on $M_{\mu_i}$ on the Grothendieck group level. 

2447: 

2448: \begin{eqnarray*}

2449:  [\tau_{i+1}^i \tau_i^{i+1} M_{\mu_i}] & = & 

2450:   [\Tui T_{i+1} T^{i+1} \Ti M_{\mu_i}] \\ 

2451:                                      & = & 

2452:   [\Tui T_{i+1} T^{i+1} (M_{\mu}\oplus M_{s_i \mu}) ] \\

2453:                                      & = & 

2454:   [\Tui T_{i+1} (M_{\mu_{i+1}} \oplus M_{s_i\mu_{i+1}}) ] \\

2455:                                      & = & 

2456:   [\Tui ( M_{\mu} \oplus M_{s_{i+1}\mu} \oplus M_{s_i \mu} 

2457:    \oplus M_{s_is_{i+1}\mu}]  \\

2458:                                      & = & 

2459:   [M_{\mu_i} \oplus M_{s_{i+1}\mu_i} \oplus M_{s_i \mu_i} 

2460:    \oplus M_{s_is_{i+1}\mu_i}]  \\

2461:                                      & = & 

2462:   [M_{\mu_i}] +  [M_{s_{i+1}\mu_i}] +  [M_{s_i \mu_i}] + 

2463:    [M_{s_is_{i+1}\mu_i}]  \\

2464:                                      & = & 

2465:   [M_{\mu_i}] +  [M_{s_{i+1}\mu_i}] +  [M_{\mu_i}] + 

2466:    [M_{s_is_{i+1}\mu_i}] 

2467: \end{eqnarray*} 

2468:                                       

2469: The projective module $P_{s_is_{i+1}\mu_i}$ 

2470: decomposes in the Grothendieck group as the following sum 

2471: $$[P_{s_is_{i+1}\mu_i}]=  

2472: [M_{\mu_i}] + [M_{s_{i+1}\mu_i}] + [M_{s_is_{i+1}\mu_i}].$$

2473: Therefore,  

2474: $$[\tau_{i+1}^i \tau_i^{i+1} M_{\mu_i}] = [M_{\mu_i}] + 

2475: [P_{s_is_{i+1}\mu_i}].$$

2476: From the classification of projective functors (see [BG]), we 

2477: derive that $\tau_{i+1}^i \tau_i^{i+1}$ is isomorphic to 

2478: the direct sum of the identity functor and an indecomposable 

2479:  projective functor that 

2480: takes $M_{\mu_i}$ to the indecomposable projective module 

2481: $P_{s_is_{i+1}\mu_i}.$ 

2482: Denote this functor by $\wp.$ Then 

2483: $$\tau_{i+1}^i \tau_i^{i+1}\cong Id \oplus \wp.$$ 

2484: 

2485: 

2486: Let $\xi_i$ be an integral dominant weight on the intersection of 

2487: the $i$-th and $(i+1)$-th walls. We require that $\xi_i$ be 

2488: a generic weight with these conditions, i.e. $\xi_i$ does not 

2489: lie on any other walls. 

2490: Let $\O_{\xi_i}$ be the subcategory of $\O(\gl)$ consisting 

2491: of modules with generalized central character $\eta(\xi_i).$ 

2492: 

2493: Let $T_{\mu_i}^{\xi_i}$ and $T_{\xi_i}^{\mu_i}$ be translation 

2494: functors from $\O_{\mu_i}$ to $\O_{\xi_i}$ and back. Then 

2495: $T_{\mu_i}^{\xi_i}$ takes the Verma module $M_{\mu_i}$ to the 

2496: Verma module $M_{\xi_i}$ while $T_{\xi_i}^{\mu_i}$ takes 

2497: $M_{\xi_i}$ to $P_{s_is_{i+1}\mu_i}.$ 

2498: Therefore, functor $\wp $ is isomorphic to the composition 

2499: $T_{\xi_i}^{\mu_i}T_{\mu_i}^{\xi_i}$ and 

2500: $$\tau_{i+1}^i \tau_i^{i+1}\cong Id \oplus 

2501: T_{\xi_i}^{\mu_i}T_{\mu_i}^{\xi_i}.$$ 

2502:  The category $\O_{\xi_i}$ contains no $U(\mf{p}_k)$-locally finite 

2503: modules other than the zero module. Hence, the functor $T_{\mu_i}^{\xi_i},$ 

2504: restricted to the subcategory $\O_i^{k,n-k}$ is the zero functor. 

2505: Therefore, $\tau_{i+1}^i\tau_i^{i+1}$ , restricted to $\O_i^{k,n-k},$ 

2506: is isomorphic to the identity functor. This proves the lemma. 

2507: $\square$

2508: 

2509: In exactly the same fashion we establish that $\tau_i^{i+1}\tau_{i+1}^i$ 

2510: is isomorphic to the identity functor from $\O_{i+1}^{k,n-k}$ to itself. 

2511: Therefore functors  $\tau_i^{i+1}$ and $\tau_{i+1}^i$ are mutually 

2512: inverse and provide an equivalence of categories $\O_{i}^{k,n-k}$ and 

2513: $\O_{i+1}^{k,n-k}.$ 

2514: $\square$ 

2515: 

2516: \subsubsection{Projective functor realization of the Temperley-Lieb  

2517:   algebra} 

2518: \label{pr-and-TL} 

2519: 

2520: Define the  functor $\CU_i, i=1,\dots n-1$ from $\O^{k,n-k}$ to $\O^{k,n-k}$ 

2521: as the composition of functors $\Ti$ and $\Tui$: 

2522: $$\CU_i=\Ti \circ \Tui$$

2523: 

2524: \begin{prop}

2525: \label{TL-projective}

2526: There are equivalences of functors

2527: \begin{eqnarray*}

2528: \CU_i\circ \CU_j   & \cong & \CU_j\circ \CU_i\mbox{ for }|i-j|>1 \\

2529: \CU_i\circ \CU_i   & \cong & \CU_i\oplus \CU_i \\

2530: \CU_i\circ \CU_{i\pm 1}\circ \CU_i & \cong &  \CU_i 

2531: \end{eqnarray*}

2532: \end{prop}

2533: 

2534: \emph{Proof: }The first two equivalences hold even if we consider 

2535: $\CU_i$ as the functor in the bigger category $\O_{\mu}.$ For example, 

2536: $$\CU_i\circ \CU_i= \Ti\circ \Tui\circ \Ti \circ \Tui= 

2537: (\Ti \circ \Tui) \oplus (\Ti\circ \Tui) =

2538: \CU_i \oplus \CU_i$$

2539: where the second equality follows from the result that the 

2540: composition $\Tui\circ \Ti$ of projective functors off and on the wall 

2541: is the direct sum of two copies of the identity functor. 

2542: 

2543: For the third equivalence the restriction to $\O^{k,n-k}$ is 

2544: absolutely necessary. Then 

2545: \begin{eqnarray*} 

2546: \CU_i \CU_{i+1}\CU_i & = & \Ti\Tui T_{i+1}T^{i+1}\Ti \Tui \\

2547:                      & = & \Ti \tau_{i+1}^i \tau_i^{i+1}\Tui  \\

2548:                      & = & \Ti \Tui  \\

2549:                      & = & \CU_i 

2550: \end{eqnarray*} 

2551: $\square$

2552: 

2553:   This proposition gives a functor realization of the Temperley-Lieb 

2554:   algebra by projective functors in parabolic categories $\O^{k,n-k}.$ 

2555: Suitable products of generators $U_i$ produce a basis of the Temperley-Lieb 

2556:  algebra $TL_{n,q},$ which admits a graphical interpretation: elements 

2557:   of this basis correspond to isotopy classes of systems of $n$ simple 

2558:  disjoint arcs in the plane connecting $n$ bottom and $n$ top points. 

2559:  Moreover, this basis is related (see [FG]) 

2560:  to the Kazhdan-Lusztig basis in the Hecke algebra as well as to Lusztig's 

2561:   bases in tensor products of $U_q(\mf{sl}_2)$-representations (see [FK]). 

2562:  Proposition~\ref{TL-projective} implies that to an element of this basis, 

2563:   we can associate a projective functor from $\O^{k,n-k}$ to $\O^{k,n-k},$

2564:   which is defined as a suitable composition of functors 

2565:  $\CU_i, 1\le i\le n-1.$  

2566:   We conjecture that these compositions of $\CU_i$'s are 

2567:   indecomposable, and, in turn, an indecomposable projective functor 

2568:   from $\O^{k,n-k}$ to $\O^{k,n-k}$ 

2569:   is isomorphic to one of these compositions

2570:  (compare with Theorem~\ref{thrm:indecomposable-proj-functors}).

2571:  

2572:   

2573:   In [ES] Enright and Shelton, among other things, constructed an 

2574: equivalence of categories $\O_1^{k,n-k}$ and $\O^{k-1,n-k-1}$ 

2575: (see [ES], $\S 11$). This equivalence allows us to factorize 

2576: $\CU_i$ as a composition of a functor from $\O_{k,n-k}$ to 

2577: $\O_{k-1,n-k-1}$ and a functor from $\O_{k-1,n-k-1}$ to $\O_{k,n-k}$. 

2578: This is very much in line with the factorization of the element $U_i$ 

2579: of the Temperley-Lieb algebra as the composition  

2580: $\cup_{i,n-2}\circ \cap_{i,n}$ 

2581: of morphisms $\cup_{i,n-2}$ and $\cap_{i,n}$ of the Temperley-Lieb category. 

2582: 

2583: We now offer the reader a conjecture on 

2584: realizing the Temperley-Lieb category via functors between parabolic 

2585: categories $\O^{k,n-k}.$ Let 

2586: \begin{equation}

2587: \zeta_n:\O^{k,n-k}_1 \stackrel{\cong}{\lra} \O^{k-1,n-k-1}

2588: \end{equation}

2589: be the Enright-Shelton equivalence of categories. 

2590: Introduce functors 

2591: \begin{eqnarray*} 

2592: \cap_{i,n}: &  \O^{k,n-k} & \lra \O^{k-1,n-k-1} \\

2593: \cup_{i,n}: &  \O^{k,n-k} & \lra \O^{k+1,n+1-k} 

2594: \end{eqnarray*} 

2595: given by 

2596: \begin{eqnarray}

2597: \label{another-cap}

2598: \cap_{i,n} & = & \zeta_n \circ \tau_2^1 \circ \tau_3^2

2599: \circ \dots \circ \tau_{i-1}^{i-2} 

2600: \circ \tau_i^{i-1} \circ T^i  \\

2601: \label{another-cup}

2602: \cup_{i,n} & = & T_i \circ \tau_{i-1}^i \circ \tau_{i-2}^{i-1}

2603: \dots \circ \tau_2^3 \circ \tau_1^2 \circ \zeta_{n+2}^{-1} 

2604: \end{eqnarray} 

2605: 

2606: \begin{conjecture} There are natural isomorphisms

2607: (\ref{first-relation})-(\ref{last-relation}) 

2608: ( with  $q$ set to $-1$ in (\ref{last-relation}))

2609: of functors 

2610: where $\cap_{i,n}$ and $\cup_{i,n}$ are defined by

2611: (\ref{another-cap}) and (\ref{another-cup}). 

2612: \end{conjecture}

2613: 

2614: Equivalences (\ref{first-relation}) and (\ref{second-relation}) 

2615:  are immediate from [ES] and the results of this section.

2616: Relation (\ref{last-relation}) is implied by the fact

2617:  that the composition of translation functors from and to a wall is 

2618: equivalent to two copies of the identity functor. 

2619:  To prove the remaining four equivalences, 

2620: a thorough understanding of the functor $\zeta_n$, 

2621: defined in [ES] in quite a tricky way, will be required. 

2622: 

2623: To categorify the Temperley-Lieb algebra $TL_{n,q}$ for arbitrary $q$, 

2624: rather than just $q=-1,$ one needs to work with the mixed version of 

2625: parabolic categories and projective functors. The conjecture of 

2626: Irving [Ir] that projective functors admit a  mixed structure 

2627: can probably be approached via a recent work [BGi] of  Beilinson and Ginzburg 

2628: where wall-crossing functors are realized geometrically. 

2629: 

2630: We next state the parabolic analogue of  

2631: Conjecture~\ref{tangle-invariants-sing}.

2632:  It is convenient to suppress parameter $k$ in the definition 

2633:  of $\cap_{i,n}$ and $\cup_{i,n}$ by summing over $k$ and 

2634:  passing to categories $\O^n= \oplus_k \O^{k,n-k}.$ 

2635:   We switch to derived categories and extend functors $\cap_{i,n}, 

2636:  \cup_{i,n}$ and $\CU_i$ to derived functors 

2637:  \begin{eqnarray*} 

2638:   \cap_{i,n} & : & D^b(\O^n) \lra D^b(O^{n-2}) \\

2639:   \cup_{i,n} & : & D^b(\O^n) \lra D^b(\O^{n+2}) \\

2640:   \CU_i      & : & D^b(\O^n) \lra D^b(\O^n) 

2641:   \end{eqnarray*} 

2642: 

2643: Recall elementary tangles (a)-(c) described at the end of 

2644: Section~\ref{TL-and-Z}. 

2645: 

2646: To the 

2647: tangle (a) associate functor $\cap_{i,n}$ given by the formula 

2648: (\ref{another-cap}). 

2649: 

2650: To the tangle (b) associate functor $\cup_{i,n}$ given by the formula 

2651: (\ref{another-cup}). 

2652: 

2653: To the tangle (c) associate functor $R_{i,n}$ from $D^b(\O^n)$ to $D^b(\O^n)$ 

2654: which is the cone of the adjointness morphism of functors 

2655: $$ \CU_i  \lra Id. $$

2656: Given a presentation $\alpha$ of a tangle $t$ as a composition 

2657: of elementary tangles of types (a)-(c), to $\alpha$ we associate the 

2658: functor $g(\alpha)$ which is the corresponding composition of 

2659: functors $\cap_{i,n}, \cup_{i,n}, R_{i,n}.$ 

2660: 

2661: \begin{conjecture}

2662: \label{tangle-invariants-parab} 

2663:  Given two presentations $\alpha, \beta$ of 

2664: a tangle $t$ as products of elementary tangles (a)-(c), 

2665: functors $g(\alpha)$ and $g(\beta)$ are  

2666: isomorphic, up to shifts in the derived category. 

2667: \end{conjecture}

2668: 

2669: 

2670: 

2671: 

2672: 

2673: \subsection{$\mf{sl}_2$ and Zuckerman functors} 

2674: \label{sl2-and-Z} 

2675: 

2676: The Grothendieck group of the category  $\O^n=\oplus_{k=0}^n \O^{k,n-k}$ 

2677: has rank $2^n$. In the previous section we showed that

2678: the projective functors, restricted to $\O^n$, ``categorify'' 

2679:  the Temperley-Lieb 

2680: algebra action on $\vn.$

2681: The Lie algebra $\mf{sl}_2$ action on $\vn$ commutes with the Temperley-Lieb 

2682: algebra action, while Zuckerman functors commute with projective functors. 

2683: It is an obvious guess now that Zuckerman functors between different 

2684: blocks of $\O^n$ provide a ``categorification'' of this $\mf{sl}_2$

2685: action. This fact is well-known and dates back to [BLM] and [GrL], where 

2686: it is presented in a different language

2687: and in the more general case of $\mf{sl}_k$ rather than $\mf{sl}_2.$ 

2688:  Beilinson, Lusztig and 

2689: MacPherson in [BLM] count points over finite fields 

2690: in certain correspondences between flag varieties. These correspondences 

2691: define functors between derived categories of sheaves on these 

2692: flag varieties, smooth along Schubert stratifications. 

2693: Counting points is equivalent to computing the 

2694: action of the corresponding functors on Grothendieck groups. 

2695: 

2696: These derived categories are equivalent to derived categories of

2697: parabolic subcategories of a  regular block of the highest weight 

2698: category for $\mf{gl}_n.$ Pullback and pushforward functors are 

2699: then isomorphic (up to shifts) to the embedding functor from smaller

2700: to bigger parabolic subcategories and its adjoint 

2701: functors which are Zuckerman functors (see [BGS], Remark (2) on page 504). 

2702: Combining this observation with  the computation 

2703: of [GrL] for the special case of the Grassmannian rather than an arbitrary 

2704: partial flag variety, one can check that Zuckerman functors

2705: between various pieces of $\O^n$ categorify the action of

2706: $E^{(a)}$ and $F^{(a)}$ on $\vn.$ 

2707: This fact is stated more accurately below. 

2708: 

2709: 

2710: Fix $n\in \N.$ Recall that we denoted by $\mf{p}_k$ the parabolic subalgebra 

2711: of $\mf{gl}_n$ consisting of block upper-triangular matrices with 

2712: the reductive part $\mf{gl}_k\oplus \mf{gl}_{n-k}.$ 

2713: Denote by $\mf{p}_{k,l}$ the parabolic subalgebra of $\mf{gl}_n$ which 

2714: is the intersection of $\mf{p}_k$ and $\mf{p}_{k+l}.$ The maximal 

2715: reductive subalgebra of $\mf{p}_{k,l}$ is isomorphic to 

2716: $\mf{gl}_k\oplus \mf{gl}_{l}\oplus \mf{gl}_{n-k-l}.$ 

2717: 

2718: Denote by $\O^{k,l,n-k-l}$ the complete subcategory of $\O_{\mu}$ consisting

2719: of $U(\mf{p}_{k,l})$-locally finite modules. We have embeddings of 

2720: categories 

2721: \begin{eqnarray}

2722: \mc{I}_{k,l}: & \O^{k,n-k} & \lra \O^{k,l,n-k-l} \\

2723: \mc{J}_{k,l}: & \O^{k+l,n-k-l} & \lra \O^{k,l,n-k-l} 

2724: \end{eqnarray} 

2725: Denote by $\mc{K}_{k,l}$ and $\mc{L}_{k,l}$ derived functors 

2726: of the right adjoint functors 

2727: of $\mc{I}_{k,l}$ and $\mc{J}_{k,l}$: 

2728: 

2729: \begin{eqnarray}

2730: \mc{K}_{k,l}: & D^b(\O^{k,l,n-k-l}) & \lra D^b(\O^{k,n-k}) \\

2731: \mc{L}_{k,l}: & D^b(\O^{k,l,n-k-l}) & \lra D^b(\O^{k+l,n-k-l}) 

2732: \end{eqnarray} 

2733: These are derived functors of Zuckerman functors. Compositions 

2734: $\mc{L}_{k,l}\circ \mc{I}_{k,l}$ and $\mc{K}_{k,l}\circ 

2735: \mc{J}_{k,l}$ are exact functors between derived categories 

2736: $D^b(\O^{k,n-k})$ and $D^b(\O^{k+l,n-k-l}),$ and  in the Grothendieck 

2737: group of $\O^n=\oplus_{k=0}^n \O^{k,n-k}$ 

2738: descend to the action of $E^{(l)}$ and $F^{(l)}$ on 

2739: the module $\vn.$ 

2740: 

2741: \hspace{0.5in} 

2742: 

2743: \begin{thebibliography}{abcd 99}

2744: 

2745: \bibitem[BLM]{BLM} A. A. Beilinson, G. Lusztig and R. MacPherson,

2746:  A geometric setting for the quantum deformation of $GL_n,$

2747: \emph{ Duke Math J.} {\bf 61}, 2 (1990), 655-677.

2748: 

2749: \bibitem[BGi]{BGi} A. Beilinson, V. Ginzburg, Wall-Crossing 

2750: functors and D-modules, preprint alg-geom/9709022. 

2751: 

2752: \bibitem[BGS]{BGS} A. Beilinson, V. Ginzburg and W. Soergel, 

2753: Koszul duality patterns in representation theory, {\it Journal of 

2754: the AMS} {\bf 9}, 2 (1996), 473-527. 

2755: 

2756: \bibitem[BGG]{BGG} J. N. Bernstein, I. M. Gelfand and S. I. Gelfand, 

2757: Category of $\mf{g}$-modules, \emph{Functional Anal. and Appl.}

2758: {\bf 10}, 2 (1976), 87-92.  

2759: 

2760: \bibitem[BG]{BG} J. N. Bernstein, S. I. Gelfand, Tensor products of finite and 

2761: infinite dimensional representations of semisimple Lie algebras, 

2762: \emph{Compositio Math.} {\bf 41}, (1980), 245-285. 

2763: 

2764: \bibitem[CS]{CS} J. S. Carter, M. Saito, 

2765: Reidemeister moves for surface isotopies and 

2766: their interpretation as moves to movies, \emph{J. Knot Theory 

2767: Ramifications} {\bf 2}, 3 (1993), 251-284. 

2768: 

2769: \bibitem[CRS]{CRS} J. S. Carter, J. H. Rieger and M. Saito, 

2770:  A combinatorial description of knotted 

2771: surfaces and their isotopies, \emph{Adv. Math.} {\bf 127}, 1 (1997), 1-51. 

2772: 

2773: \bibitem[CF]{CF} L. Crane, I. B. Frenkel, Four-dimensional topological 

2774: quantum field theory, Hopf categories, and the canonical bases, 

2775: \emph{J. Math. Phys.} {\bf 35}, (1994), 5136-5154. 

2776: 

2777: \bibitem[ES]{ES} T. J. Enright, B. Shelton, Categories of 

2778: highest weight modules: applications to classical Hermitian symmetric pairs, 

2779: \emph{Memoirs of the AMS} {\bf 367}, (1987). 

2780:           

2781: \bibitem[EW]{EW} T. J. Enright, N. R. Wallach, Notes on homological algebra 

2782: and representations of Lie algebras, \emph{Duke Math. Jour.} {\bf 47}, 1 

2783: (1980), 1-15. 

2784: 

2785: \bibitem [Fi]{Fi} J. Fischer, 2-categories and 2-knots, 

2786: \emph{Duke Math J.} {\bf 75}, 2 (1994), 493-526. 

2787: 

2788: \bibitem [FK]{FK} I. B. Frenkel, M. Khovanov, 

2789: Canonical bases in tensor products and graphical calculus for

2790: $U_q(\frak{sl}_2),$ \emph{Duke Math J.} {\bf 87}, 3 (1997), 409-480. 

2791: 

2792: \bibitem [FKK]{FKK}  I. B. Frenkel, M. Khovanov, A. Kirillov, Jr., 

2793: Kazhdan-Lusztig polynomials and canonical basis, 

2794: \emph{Transformation Groups} {\bf 3}, 4 (1998), 321-336.   

2795: 

2796: \bibitem [FG]{FG} C. K. Fan, R. M. Green, Monomials and Temperley-Lieb 

2797:  algebras, \emph{J. Algebra} {\bf 190}, 2 (1997), 498-517. 

2798: 

2799: \bibitem [FM]{FM} I. B. Frenkel, F. Malikov, Annihilating ideals and 

2800: tilting functors, preprint q-alg/9801065. 

2801: 

2802: \bibitem [Gr]{Gr} I. Grojnowski,  The coproduct for quantum $GL_n$, preprint, 

2803:  (1992).  

2804: 

2805: \bibitem [GrL]{GrL} I. Grojnowski, G. Lusztig, On bases of irreducible 

2806: representations of quantum $GL_n$, in \emph{Kazhdan-Lusztig theory 

2807: and related topics, Chicago, IL, 1989,} Contemp. Math. {\bf 139}, 167-174. 

2808: 

2809: \bibitem [Ir]{Ir} R. S. Irving, Singular blocks of the category $\O,$ 

2810: \emph{Math. Z.} {\bf 204}, (1990), 209-224. 

2811: 

2812: \bibitem [J]{J} J. C. Jantzen, Moduln mit einem h\"{o}chsten Gewicht, 

2813: \emph{Lect. Notes in Math.} {\bf 750}, 1979. 

2814: 

2815: \bibitem [Jo]{Jo} V. F. R. Jones, A polynomial invariant for knots 

2816: via von Neumann algebras, \emph{Bull. Amer. Math. Soc.} {\bf 12}, 

2817: 1 (1985), 103-111. 

2818: 

2819: \bibitem [Ka]{Ka} L. H. Kauffman, State models and the Jones polynomial, 

2820: \emph{Topology} {\bf 26}, 3 (1987), 395-407. 

2821: 

2822: \bibitem [KaL]{KaL} L. H. Kauffman, S. Lins, Temperley-Lieb Recoupling 

2823: Theory and Invariants of 3-manifolds, \emph{Ann. of Math. Studies}

2824: {\bf 134}, Princeton U. Press, Princeton, 1994. 

2825: 

2826: \bibitem [K]{K} M. Khovanov, Graphical calculus, canonical bases and 

2827: Kazhdan-Lusztig theory, \emph{Thesis}, Yale University, (1997). 

2828: 

2829: \bibitem [KV]{KV} A. W. Knapp, D. A. Vogan, Jr., Cohomological Induction

2830: and Unitary Representations,  Princeton University Press, 

2831: 1995.

2832: 

2833: \bibitem [LS]{LS} A. Lascoux, M.-P. Sch\"{u}tzenberger,

2834:  Polynomes de Kazhdan \& Lusztig pour les grassmaniennes, 

2835: \emph{ Ast\'{e}risque, } {\bf 87-88}, (1981) 249-266. 

2836: 

2837: \bibitem [Lp]{Lp} J. Lepowsky, A generalization of the 

2838: Bernstein-Gelfand-Gelfand resolution, \emph{J. Algebra} {\bf 49}, 2 (1977), 

2839: 496-511. 

2840: 

2841: \bibitem[Lu]{Lu} G. Lusztig, Introduction to Quantum Groups, 

2842: Birkh\"{a}user Boston,  1993. 

2843: 

2844: \bibitem [MP]{MP} D. Mili\u{c}i\'{c}, 

2845: P. Pand\u{z}i\'{c}, On degeneration of the spectral 

2846: sequence for the composition of Zuckerman functors, 

2847: \emph{ Glas. Mat. Ser. III} {\bf 32(52)}, 2 (1997), 179-199.  

2848: 

2849: \bibitem [RC]{RC} A. Rocha-Caridi, Splitting criteria for $\mf{g}$-modules 

2850: induced from a parabolic and the Bernstein-Gelfand-Gelfand resolution 

2851: of a finite dimensional, irreducible $\mf{g}$-module, 

2852: \emph{Trans. AMS} {\bf 262}, 2 (1980), 335-366. 

2853: 

2854: \bibitem [RT]{RT} N. Yu. Reshetikhin, V. G. Turaev, 

2855: Ribbon graphs and their invariants derived from quantum groups, 

2856: \emph{Comm. Math. Phys.} {\bf 127}, (1990), 1-26. 

2857: 

2858: \bibitem[Tu]{Tu} V. G. Turaev, Quantum Invariants of Knots and 3-Manifolds, 

2859: \emph{De Gruyter Studies in Mathematics} {\bf 18}, 1994. 

2860: 

2861: \bibitem[Z]{Z} A. V. Zelevinski, Small resolutions of singularities of 

2862: Schubert varieties, \emph{Funktional. Anal. and Appl.} {\bf 17}, (1983), 

2863: 75-77. 

2864: 

2865: \bibitem[Zu]{Zu} G. Zuckerman, Tensor products of finite and infinite 

2866: dimensional representations of semisimple Lie groups, 

2867: \emph{Ann. of Math.} {\bf 106}, (1977), 295-308. 

2868: 

2869: \end{thebibliography}

2870: 

2871: \end{document}

2872: 

2873: 

2874: 

2875: 

2876: 

2877: 

2878: 

2879: 

2880: 

2881: 

2882: 

2883: 

2884: 

2885: 

2886: 

2887: 

2888: 

2889: 

2890: 

2891: 

2892: 

2893: 

2894: 

2895: 

2896: