arXiv:1012.3687v2 [math.QA] 14 Dec 2012 elgop n[ In group. Weyl ipycnetdLegroup, Lie simply–connected ..Tepeetppri oiae y n astegroundwo the lays s and the by, by formulated motivated conjecture is monodromy paper trigonometric present the The 1.1. ojcue n[ in conjectured Let iesoa ersnaino h Yangian the of representation dimensional h oorm ftertoa aii oncinof connection Casimir rational the of monodromy the hc a oaihi iglrte ntero utr of subtori root the on singularities logarithmic has which ohatoswr upre yNFgat M–771 n DM and DMS–0707212 grants NSF by supported were authors Both phrases. and words Key 2010 relations Serre the of Proof References for A. solution Appendix Geometric type geometric of 7. Isomorphisms homomorphisms of 6. Uniqueness homomorphisms of 5. Existence type geometric of 4. Yangians Homomorphisms and algebras loop 3. Quantum 2. .Introduction 1. g eacmlx eiipeLealgebra, Lie semisimple complex, a be ahmtc ujc Classification. Subject Mathematics Abstract. Ginzburg–Vasserot. h unu opalgebra loop quantum the hwta titrwnsteatosof actions the intertwines it that show teghnti eutb osrciga xlctalgebra explicit an constructing U by result this strengthen vrta eoe nioopimwhen isomorphism an becomes Φ that over K t vlainiel ecntutasmlrhmmrhs f homomorphism similar a construct We ideal. evaluation its ~ ter n ooooyo h ait of variety the of cohomology and –theory ( L AGASADQATMLO ALGEBRAS LOOP QUANTUM AND YANGIANS g otecmlto of completion the to ) 32 AHNGUA N AEI OEAOLAREDO TOLEDANO VALERIO AND GAUTAM SACHIN httemndoyo h rgnmti aii connectio Casimir trigonometric the of monodromy the that ] 32 Let flat, a ] g ffieqatmgop,Ynin unu opalgebra. loop quantum Yangian, groups, quantum Affine eacmlx eiipeLeagba rnedsoe that showed Drinfeld algebra. Lie semisimple complex, a be W gl H U euvratconnection –equivariant n ~ 1. Y ( ⊂ ~ L ( g Introduction g Contents G of ) ihrsett t rdn.W hwmore- show We grading. its to respect with ) 73 (17B67,82B23,14F43). 17B37 aia ou and torus maximal a g U 1 eeeae oteYangian the to degenerates Y ~ ( ~ L ( U gl g ~ n G .B nlg ihtedsrpinof description the with analogy By ). n ( se asin flags –step L and ) g h orsodn once and connected corresponding the scmltdwt epc to respect with completed is ) Y ~ g ( gl ooopimΦfrom Φ homomorphism ∇ bandi [ in obtained b H n C nteequivariant the on ) n ausi n finite– any in values and C a osrce on constructed was S–0854792. d or W cn uhri [ in author econd osrce by constructed g h corresponding the = kfrapofof proof a for rk Y ~ ( gl 30 g n .We ). and , , 31 ,i was it ], n 32 ∇ b 54 50 34 30 24 19 14 H C ]. 7 1 2 S.GAUTAMANDV.TOLEDANOLAREDO is described by the action of the affine braid group of g arising from the quantum Weyl group operators of the quantum loop algebra U~(Lg). This raises in particular the problem of relating finite–dimensional representations of Y~(g) and U~(Lg).
1.2. Since their construction by Drinfeld [9, 10], these affine quantum groups have been extensively studied from several perspectives (see, e.g., [5, chap. 12], [2] and references therein), and are widely believed to share the same finite–dimensional representation theory. This belief is corroborated in part by the following facts
(1) The quantum loop algebra U~(Lg) degenerates to the Yangian Y~(g). Specif- ically, if U~(Lg) is filtered by the powers of the evaluation ideal at z = 1, its associated graded is isomorphic to Y~(g) [10, 17]. (2) Finite–dimensional simple modules over U~(Lg) are parametrised by I–tuples of (Drinfeld) polynomials Pi(u) i I satisfying Pi(0) = 1, where I is the set of vertices of the Dynkin{ diagram} ∈ of g [4]. Similarly, finite–dimensional simple modules over Y~(g) are classified by I–tuples of monic polynomials [11, 28, 29, 3]. (3) If g is simply laced, there exists, for every w NI, a Steinberg variety Z(w) ∈ endowed with an action of GL(w) C× (here GL(w) = i I GLwi ), and algebra homomorphisms × ∈ Q GL(w) C× Ψ : U~(Lg) K × (Z(w)) U → GL(w) C× Ψ : Y~(g) H × (Z(w)) Y → The variety Z(w) and the homomorphism ΨU were constructed by Nakajima [26], while ΨY was constructed by Varagnolo [33].
1.3. The above results go some way towards relating the categories of finite– dimensional representations of U~(Lg) and Y~(g). For example, exponentiating the roots of Drinfeld polynomials yields, via (2), a surjective map exp∗ between the set of isomorphism classes of irreducible finite–dimensional modules of Y~(g) and those of U~(Lg)–modules. If g is simply laced, the geometric realisations (3) imply further that exp∗ preserves the dimensions of these representations [33]. Despite these results however, and to the best of our knowledge, no natural re- lation between the categories of finite–dimensional representations of U~(Lg) and Y~(g) is known. Part of the difficulty in exploiting, say, the geometric realisations to pursue this question lies in the fact that the homomorphisms ΨU , ΨY are neither in- jective nor surjective [26]. Moreover, although these realisations yield all irreducible
representations, the categories Repfd(U~(Lg)) and Repfd(Y~(g)) are not semisimple.
1.4. The aim of the present paper is to clarify the relation between U~(Lg) and Y~(g). We do so by constructing an explicit algebra homomorphism [ Φ : U~(Lg) Y~(g) −→ [ where Y~(g) is the completion of Y~(g) with respect to its N–grading, and show that it induces an isomorphism of completed algebras. We also show that Φ exponentiates YANGIANSANDQUANTUMLOOPALGEBRAS 3
the roots of Drinfeld polynomials, though we defer the study of the corresponding pull–back functor
F =Φ∗ : Rep (Y~(g)) Rep (U~(Lg)) fd → fd to the sequel of this paper [13]. To state out results more precisely, recall that U~(Lg) and Y~(g) are deformations 1 of the loop and current algebras U(g[z, z− ]) and U(g[s]) respectively. Denote by
U~(Lh),U~(Lb ) U~(Lg) and Y~(h),Y~(b ) Y~(g) ± ⊂ ± ⊂
1 1 the subalgebras deforming U(h[z, z− ]), U(b [z, z− ]) and U(h[s]), U(b [s]) respec- tively, where h g is the Lie algebra of H± and b g are the opposite± Borel ⊂ ± ⊂ subalgebras corresponding to a choice αi i I of simple roots of g. For any i I, { } ∈ ∈ let sli g be the corresponding 3–dimensional subalgebra and denote by 2 ⊂
i i U~(Lsl ) U~(Lg) and Y~(sl ) Y~(g) 2 ⊂ 2 ⊂
i 1 i the subalgebras which deform U(sl2[z, z− ]) and U(sl2[s]) respectively. Then, the main result of this paper is the following
[ Theorem. There exists an explicit algebra homomorphism Φ : U~(Lg) Y~(g) with the following properties → (1) Φ is defined over Q[[~]]. \ [ \ (2) Φ induces an isomorphism U~(Lg) Y~(g), where U~(Lg) is the completion → of U~(Lg) with respect to the ideal of z = 1. (3) Φ induces Drinfeld’s degeneration of U~(Lg) to Y~(g). [ (4) Φ restricts to a homomorphism U~(Lh) Y~(h) which induces the exponen- tiation of roots on Drinfeld polynomials.→ \ (5) Φ restricts to a homomorphism U~(Lb ) Y~(b ). ± → ± i \i (6) Φ restricts to a homomorphism U~(Lsl ) Y~(sl ) for any i I. 2 → 2 ∈ It is interesting to note that Theorem 1.4 stands in stark contrast with the analogous finite–dimensional situation. Indeed, if g ≇ sl2, no explicit isomorphisms are known between the quantum group U~g and the undeformed enveloping algebra Ug[[~]] [5, i 6.4]. Moreover, if g ≇ sl , no algebra isomorphism U~g Ug[[~]] maps U~sl to § 2 → 2 Usli [[~]] for every i I [31, Prop. 3.2]. 2 ∈
1.5. The homomorphism Φ has the following form. Let Ei,k, Fi,k,Hi,k i I,k Z be { } ∈ ∈ the loop generators of U~(Lg) and xi,m± ,ξi,m i I,m N those of Y~(g) (see [11] and { } ∈ ∈ 4 S.GAUTAMANDV.TOLEDANOLAREDO
Section 2 for definitions). Then, 1 Φ(Hi,0)= di− ti,0 ~ rm Φ(H )= t i,r 1 i,m m! qi qi− m 0 − X≥ + kσi + + Φ(Ei,k)= e gi,m xi,m m 0 ≥ − X kσi Φ(Fi,k)= e gi,m− xi,m− m 0 X≥ In the formulae above, r Z , k Z, q = e~/2 and q = qdi , where the d are ∈ ∗ ∈ i i the symmetrising integers for the Cartan matrix of g. The ti,m i I,m N are an { } ∈ ∈ alternative set of generators of the commutative subalgebra Y~(h) Y~(g) generated ⊂ by the elements ξi,m i I,m N. They are defined in [22] by equating the generating functions { } ∈ ∈ m 1 m 1 ~ ti,mu− − = log(1 + ~ ξi,mu− − ) m 0 m 0 X≥ X≥ The elements gi,m± i I,m N lie in the completion of Y~(h), and are constructed as { } ∈ ∈ follows. Consider the formal power series v G(v) = log vQ[[v]] ev/2 e v/2 ∈ − − and define γ (v) Y 0[[v]] by i ∈ t d r+1 c γ (v)= ~ i,r G(v) i r! −dv r 0 X≥ Then, 1/2 m ~ γi(v) g± v = exp (1.1) i,m q q 1 2 m 0 i i− X≥ − Finally, σi± are the homomorphisms of the subalgebras Y~(b ) Y~(g) generated ± ⊂ by ξj,r,xj,r± j I,r N, which fix the ξj,r and act on the remaining generators as the { } ∈ ∈ shifts x± x± . j,r → j,r+δij Note that the formulae connecting the generators Hi,k of U~(Lg) and ti,m of { } 1 { } Y~(h) essentially coincide with those connecting the generators of h[z, z− ] and h[s].
1.6. The above formulae apply equally well when g is a symmetrisable Kac–Moody algebra. Our proof of Theorem 1.4 shows that they define a homomorphism from the quantum affinization U~g of the quantum group U~g [20, 26], to the completion of the Yangian Y~(g), provided the following holds (1) the entries of the Cartand matrix of g satisfy a a 3 for i = j I. ij ji ≤ 6 ∈ (2) the PBW theorem holds for Y~(g). YANGIANSANDQUANTUMLOOPALGEBRAS 5
The first assumption is equivalent to requiring that all rank 2 subalgebras of g be finite–dimensional, and is needed in our proof of the q–Serre relations. The second is required for the construction of certain straightening homomorphisms on Y~(g) which are needed in the proof of Theorem 1.4. We note that, for the Yangians associated to affine Kac–Moody algebras, the PBW theorem was proved by Guay in type An, for n 4 [16] and, more recently, by Guay–Nakajima for all simply laced cases [18]. In≥ particular, the above formulae define a homomorphism from the tor quantum toro¨ıdal algebra U~ (g) associated to a simply laced, simple Lie algebra g ≇ sl2, to the completion of the affine Yangian Y~g.
1.7. We also construct in this paper a homomorphismb similar to the one described in 1.5 for g = gln, by relying on the geometric realisation of U~(Lgln) obtained by Ginzburg and Vasserot [15, 34]. More precisely, fix integers 1 n d, and let ≤ ≤ = 0= V V V = Cd F { 0 ⊆ 1 ⊆···⊆ n } be the variety of n–step flags in Cd. The cotangent bundle T may be realised as ∗F d T ∗ = (V ,x) End(C ) x(Vi) Vi 1 F { • ∈F× | ⊂ − } and therefore admits a morphism T ∗ via the second projection, where = x End(Cd) xn = 0 is the cone ofF→Nn–step nilpotent endomorphisms. DefineN the { ∈ | } Steinberg variety Z = T ∗ T ∗ . The group GLd C× acts on T ∗ and Z and there are surjective algebraF× homomorphismsN F × F
× GLd C ΨU : U~(Lgln) K × (Z) → × GLd C Ψ : Y~(gl ) H × (Z) Y n →
see [15, 34] for the definition of ΨU . To understand these more explicitly, one can use the convolution actions of × × × GLd C GLd C GLd C × K × (Z) on K × (T ∗ ) and of H × (Z) on HGLd C (T ∗ ), which are faithful. By using the equivariantF Chern character, we construct× anF algebra homomorphism \ Φ : U~(Lgl ) Y~(gl ) n → n which intertwines these two actions.
1.8. In the sequel to this paper [13], we shall prove that, for g semisimple, a modi- fication of the pull–back functor Φ∗ converges for numerical values of ~, and defines a functor Rep (Y (g)) Rep (U (Lg)) fd a → fd ǫ where Y (g) is the specialisation of Y~(g) at ~ = a C R and ǫ = exp(πia), and a ∈ \ defines an equivalence of an explicit subcategory of Repfd(Yag) with Repfd(Uǫ(Lg)). 6 S.GAUTAMANDV.TOLEDANOLAREDO
1.9. It is worth pointing out that most of our results relating U~(Lg) and Y~(g) have analogues for the affine and degenerate affine Hecke algebras and ′ associated to an affine Weyl group W [24], and were in fact inspired by theseH andH their further study in [6]. Indeed, in [24], Lusztig constructs an explicit isomorphism between appropriate completions of and ′. In this context, the isomorphism can be understood in terms of, andH in fact obtainedH from, the geometric realisations G C× Ξ : ∼ K × (Z) H −→ G C× Ξ′ : ′ ∼ H × (Z) H −→ where Z is the Steinberg variety corresponding to W [7, 14].
1.10. Outline of the paper. In Section 2, we review the definition of the quan- tum loop algebra U~(Lg) and Yangian Y~(g) of a semisimple Lie algebra g. We also introduce shift homomorphisms of the subalgebras Y~(b ), and straightening ± homomorphisms of the subalgebra Y~(h). [ In Section 3, we consider assignments mapping the generators of U~(Lg) to Y~(g). [ These have the form described in 1.5, where the elements g± Y~(h) are however i,m ∈ not necessarily given by formula (1.1). Our main result, Theorem 3.4, gives necessary and sufficient conditions for these elements to give rise to an algebra homomorphism. We call such homomorphisms of geometric type since, for g simply laced, they are related to the Chern character in the geometric realisation described in 1.2. The proof that the elements given by (1.1) satisfy the conditions of Theorem 3.4, and therefore give rise to an algebra homomorphism Φ, is given in Section 4 (Theorem 4.7). We also prove that the action of Φ on Drinfeld polynomials exponentiates their roots (Corollary 4.5). In Section 5, we prove the essential uniqueness of homomorphisms of geometric type by showing that any two differ by conjugation by an element of the torus H [ and an invertible element of Y~(h) (Theorem 5.11). In Section 6, we show that any homomorphism of geometric type Φ induces an \ [ \ isomorphism U~(Lg) Y~(g), where U~(Lg) is the completion with respect to the evaluation ideal at z →= 1 (Theorem 6.2). We show moreover that the associated graded map coincides with Drinfeld’s degeneration of U~(Lg) to Y~(g) (Proposition 6.5). Section 7 contains similar results for g = gln. In addition to constructing an \ explicit homomorphism Φ : U~(Lgln) Y~(gln) (Theorem 7.6), we review the geo- metric realisations of these algebras and→ show that Φ intertwines them (Theorem 7.19). Appendix A contains a proof of the Serre relations which is required to complete the proof of Theorem 3.4.
1.11. Acknowledgments. We are very grateful to Ian Grojnowski from whom we learned that the quantum loop algebra and Yangian should be isomorphic after ap- propriate completions. His explanations and friendly insistence helped us overcome our initial doubts. We are also grateful to V. Drinfeld for showing us a proof that YANGIANSANDQUANTUMLOOPALGEBRAS 7
finite–dimensional representations separate elements of the Yangian and allowing us to reproduce it in Appendix A. We would also like to thank N. Guay for sharing a preliminary version of the preprint [17] and E. Vasserot for useful discussions.
2. Quantum loop algebras and Yangians 2.1. Let g be a complex, semisimple Lie algebra and ( , ) a non–degenerate, invari- · · ant bilinear form on g. Let h g be a Cartan subalgebra of g, αi i I h∗ a basis ⊂ { } ∈ ⊂ of simple roots of g relative to h and aij = 2(αi, αj)/(αi, αi) the entries of the corre- sponding Cartan matrix A. Set d = (α , α )/2, so that d a = d a for any i, j I. i i i i ij j ji ∈ Let ν : h h∗ be the isomorphism determined by the inner product ( , ) and set 1 → · · hi = ν− (αi)/di. Choose root vectors ei gαi ,fi g αi such that [ei,fi] = hi. Recall that g is presented on generators e∈,f , h subject∈ − to the relations { i i i}
[hi, hj ] = 0 [h , e ]= a e [h ,f ]= a f i j ij j i j − ij j [ei,fj]= δijhi for any i, j I and, for any i = j I ∈ 6 ∈
1 aij ad(ei) − ej = 0
1 aij ad(fi) − fj = 0
A closely related, but slightly less standard presentation may be obtained by 1 setting ti = ν− (αi)= dihi and choosing, for any i I, root vectors xi± g αi such + ∈ ∈ ± that [xi ,xi−]= ti. Then g is presented on xi±,ti i I subject to the relations { } ∈
[ti,tj] = 0
[t ,x±]= d a x± i j ± i ij j + [xi ,xj−]= δijti 1 aij ad(xi±) − xj± = 0
2.2. Throughout this paper, q and ~ are formal variables related by q2 = e~. For di ~di/2 any i I, we set qi = q = e . We use the standard notation for Gaussian integers∈
qn q n [n] = − − q q q 1 − − n [n] ! [n] ! = [n] [n 1] [1] = q q q − q · · · q k [k] ![n k] ! q q − q 8 S.GAUTAMANDV.TOLEDANOLAREDO
2.3. The quantum loop algebra [11]. Let U~(Lg) be the unital, associative alge- bra over C[[~]] topologically generated by elements Ei,k, Fi,k,Hi,k i I,k Z subject to the following relations { } ∈ ∈ (QL1) For i, j I and r, s Z ∈ ∈
[Hi,r,Hj,s] = 0
(QL2) For any i, j I and k Z, ∈ ∈ [H , E ]= a E [H , F ]= a F i,0 j,k ij j,k i,0 j,k − ij j,k (QL3) For any i, j I and r Z , ∈ ∈ × [ra ] [ra ] [H , E ]= ij qi E [H , F ]= ij qi F i,r j,k r j,r+k i,r j,k − r j,r+k (QL4) For i, j I and k, l Z ∈ ∈ E E qaij E E = qaij E E E E i,k+1 j,l − i j,l i,k+1 i i,k j,l+1 − j,l+1 i,k aij aij F F q− F F = q− F F F F i,k+1 j,l − i j,l i,k+1 i i,k j,l+1 − j,l+1 i,k (QL5) For i, j I and k, l Z ∈ ∈ ψi,k+l φi,k+l [Ei,k, Fj,l]= δij − 1 q q− i − i (QL6) Let i = j I and set m = 1 a . For every k ,...,k Z and l Z 6 ∈ − ij 1 m ∈ ∈ m s m ( 1) Ei,kπ(1) Ei,kπ(s) Ej,lEi,kπ(s+1) Ei,kπ(m) = 0 − s q · · · · · · π Sm s=0 i X∈ X m s m ( 1) Fi,kπ(1) Fi,kπ(s) Fj,lFi,kπ(s+1) Fi,kπ(m) = 0 − s q · · · · · · π Sm s=0 i X∈ X where the elements ψi,r, φi,r are defined by
r ~di 1 s ψ (z)= ψ z− = exp H exp (q q− ) H z− i i,r 2 i,0 i − i i,s r 0 s 1 X≥ X≥ r ~di 1 s φi(z)= φi, rz = exp Hi,0 exp (qi qi− ) Hi, sz − − 2 − − − r 0 s 1 X≥ X≥ with ψi, k = φi,k = 0 for every k 1. − 0 ≥ We shall denote by U U~(Lg) the commutative subalgebra generated by the ⊂ elements Hi,r i I,r Z. { } ∈ ∈ YANGIANSANDQUANTUMLOOPALGEBRAS 9
2.4. The Yangian [11]. Let Y~(g) be the unital, associative C[~]–algebra generated by elements x± ,ξi,r i I,r N, subject to the following relations { i,r } ∈ ∈ (Y1) For any i, j I and r, s N ∈ ∈ [ξi,r,ξj,s] = 0 (Y2) For i, j I and s N ∈ ∈ [ξ ,x± ]= d a x± i,0 j,s ± i ij j,s (Y3) For i, j I and r, s N ∈ ∈ diaij~ [ξ ,x± ] [ξ ,x± ]= (ξ x± + x± ξ ) i,r+1 j,s − i,r j,s+1 ± 2 i,r j,s j,s i,r (Y4) For i, j I and r, s N ∈ ∈ diaij~ [x± ,x± ] [x± ,x± ]= (x± x± + x± x± ) i,r+1 j,s − i,r j,s+1 ± 2 i,r j,s j,s i,r (Y5) For i, j I and r, s N ∈ ∈ + [xi,r,xj,s− ]= δijξi,r+s (Y6) Let i = j I and set m = 1 a . For any r , , r N and s N 6 ∈ − ij 1 · · · m ∈ ∈ x± , x± , , x± ,x± = 0 i,rπ(1) i,rπ(2) · · · i,rπ(m) j,s · · · π Sm X∈ h h h i ii Y~(g) is an N–graded algebra by deg(ξi,r) = deg(xi,r± )= r and deg(~) = 1.
2.5. PBW theorem for Y~(g). For any positive root β of g, choose a sequence of simple roots α ,...,α such that β = α + + α and i1 ik i1 · · · ik
[xi±, [xi± , , [xi±− ,xi± ] ]] g β 1 2 · · · k 1 k · · · ∈ ± are non–zero vectors. For any r N, define x± Y~(g) by choosing a partition ∈ β,r ∈ r = r + + r of length k and setting 1 · · · k x± = [x± , [x± , , [x± ,x± ] ]] β,r i1,r1 i2,r2 · · · ik−1,rk−1 ik,rk · · ·
Theorem ([23]). Fix a total order on the set = ξi,r,x± i I,r N,β Σ+ . Then, the G { β,r} ∈ ∈ ∈ ordered monomials in the elements of form a basis of Y~(g). G 0 Let Y ,Y ± Y~(g) be the subalgebras generated by the elements ξi,r i I,r N ⊂ 0 0 { } ∈0 ∈+ (resp. xi,r± i I,r N) and Y ≥ ,Y ≤ Y~(g) the subalgebras generated by Y ,Y 0{ } ∈ ∈ ⊂ and Y ,Y − respectively. The following is a direct consequence of Theorem 2.5. Corollary. 0 (1) Y is a polynomial algebra in the generators ξi,r i I,r N. { } ∈ ∈ (2) Y ± is the algebra generated by elements xi,r± i I,r N subject to the relations { } ∈ ∈ (Y4) and (Y6). 0 0 (3) Y ≥ (resp. Y ≤ ) is the algebra generated by elements ξi,r,xi,r± i I,r N sub- { } ∈ ∈ ject to the relations (Y1)–(Y4) and (Y6). 10 S.GAUTAMANDV.TOLEDANOLAREDO
(4) Multiplication induces an isomorphism of vector spaces 0 + Y − Y Y Y~(g) ⊗ ⊗ → 2.6. The shift operators σ±. Fix i I. By Corollary 2.5 (3), the assignment i ∈ x± x± ξj,r ξj,r j,r → j,r+δij → extends to an algebra homomorphism Y 0 Y 0 (resp. Y 0 Y 0) which we ≥ → ≥ ≤ → ≤ shall denote by σi±. 2.7. The relations (Y2)–(Y3). We rewrite below the defining relations (Y2)– (Y3) of Y~(g) in terms of the shift operators σj± and the generating series
r 1 1 ξ (u)=1+ ~ ξ u− − Y~(g)[[u− ]] (2.1) i i,r ∈ r 0 X≥ Lemma. The relations (Y2)–(Y3) are equivalent to
~diaij [ξi(u),xj,s± ]= ± ξi(u)xj,s± u σ± ~d a /2 − j ± i ij 1 where the rational function on the right–hand side is expanded in powers of u− . r 1 Proof. Set a = diaij/2. Multiplying (Y3) by ~u− − and summing over r 0 yields ≥ 1 u[ξ (u) 1 ~u− ξ ,x± ] [ξ (u) 1,x± ]= ~a x± ,ξ (u) 1 i − − i,0 j,s − i − j,s+1 ± { j,s i − } where x,ξ = xξ + ξx. Using (Y2) and x,ξ = [x,ξ] + 2ξx, yields { } { } (u σ± ~a)[ξ (u),x± ]= 2~aξ (u)x± (2.2) − j ± i j,s ± i j,s 0 r 1 as claimed. Conversely, taking the coefficients of u and u− − in (2.2) yields (Y2) and (Y3) respectively.
2.8. The relations (Y4) and (Y6). We shall use the following notation for an operator T End(V ), T End(V m) is defined as • ∈ (i) ∈ ⊗ i 1 m i T = 1⊗ − T 1⊗ − (i) ⊗ ⊗ for an algebra A, ad(m) : A m End(A) is defined as • ⊗ → ad(m) (a a )= ad(a ) ad(a ) 1 ⊗···⊗ m 1 ◦···◦ m Proposition. (1) The relation (Y4) for i = j is equivalent to the requirement that the following holds for any A(v , v ) 6 C[[v , v ]] 1 2 ∈ 1 2 A(σ±,σ±)(σ± σ± a~)x± x± = A(σ±,σ±)(σ± σ± a~)x± x± i j i − j ∓ i,0 j,0 i j i − j ± j,0 i,0 where a = diaij/2. YANGIANSANDQUANTUMLOOPALGEBRAS 11
(2) The relation (Y4) for i = j is equivalent to the requirement that the following holds for any B(v , v ) C[[v , v ]] such that B(v , v )= B(v , v ) 1 2 ∈ 1 2 1 2 2 1 µ B(σ± ,σ± )(σ± σ± d ~)x± x± = 0 (2.3) i,(1) i,(2) i,(1) − i,(2) ∓ i i,0 ⊗ i,0 2 where µ : Y~(g)⊗ Y~(g) is the multiplication. (3) The relation (Y6)→ is equivalent to the requirement that the following holds for any i = j and A C[v , . . . , v ]Sm with m = 1 a 6 ∈ 1 m − ij m (m) ⊗ ad A(σi,±(1),...,σi,±(m)) xi,±0 xj,l± = 0 Proof. (1) The relation (Y4)
[x± ,x± ] [x± ,x± ]= a~(x± x± + x± x± ) i,r+1 j,s − i,r j,s+1 ± i,r j,s j,s i,r may be rewritten as
r s r s σ± σ± σ± σ± a~ x± x± = σ± σ± σ± σ± a~ x± x± i j i − j ∓ i,0 j,0 i j i − j ± j,0 i,0 (2) If i = j, a = di and the above may be written as
r s µ σ± σ± σ± σ± d ~ x± x± = i,(1) i,(2) i,(1) − i,(2) ∓ i i,0 ⊗ i,0 s r µ σ± σ± σ± σ± d ~ x± x± i,(1) i,(2) i,(2) − i,(1) ± i i,0 ⊗ i,0 which is equivalent to
r s s r µ σ± σ± + σ± σ± σ± σ± d ~ x± x± = 0 i,(1) i,(2) i,(1) i,(2) i,(1) − i,(2) ∓ i i,0 ⊗ i,0 (3) is just the reformulation of (Y6).
Corollary. If (2.3) holds for some B C[[v , v ]], then B(v , v )= B(v , v ). ∈ 1 2 1 2 2 1 Proof. By (2) of Proposition 2.8, we may assume that B(v , v )= B(v , v ) and 1 2 − 2 1 therefore that B = (v1 v2)B where B is symmetric in v1 v2. Using the grading − ↔ r s s r on Y~(g), we may further assume that B is proportional to v1v2 + v1v2 for some r s N. An application of (Y4) then yields ≥ ∈ r s s r µ (σ± σ± )(σ± σ± + σ± σ± )(σ± σ± d ~)x± x± i,(1) − i,(2) i,(1) i,(2) i,(1) i,(2) i,(1) − i,(2) ∓ i i,0 ⊗ i,0 = 2 (x± x± x± x± d ~x± x± ) (x± x± x± x± d ~x± x± ) i,r+2 i,s− i,r+1 i,s+1∓ i i,r+1 i,s − i,r+1 i,s+1− i,r i,s+2∓ i i,r i,s+1 If r s + 2, the above is not zero by the PBW Theorem 2.5 and B = 0. If r = s≥+ 1, a further application of (Y4) shows that the second of the above two parenthesized summands is zero and again B = 0 by Theorem 2.5. Finally, if r = s, (Y4) implies that the two parenthesized summands are opposites of each other and again B = 0. 12 S.GAUTAMANDV.TOLEDANOLAREDO
2.9. An alternative system of generators for Y 0. The following generators of Y 0 were introduced in [22]. For any i I, define the formal power series ∈ r 1 0 1 t (u)= ~ t u− − Y [[u− ]] i i,r ∈ r 0 X≥ by
r 1 t (u) = log(ξ (u)) = log 1+ ~ ξ u− − (2.4) i i i,r r 0 X≥ 0 Since (2.4) can be inverted, ti,r i I,r N is another system of generators of Y . These { } ∈ ∈ are homogeneous, with deg(t )= r, since ζ ξ (u)= ξ (ζ 1u), where is the action i,r · i i − · of ζ C∗ on Y~(g) given by the grading. Moreover, ti,0 = ξi,0 and ti,r = ξi,r mod ~ for any∈ r 1 since ≥ r 1 2 ti(u)= ~ ξi,ru− − mod ~ r 0 X≥ To compute the commutation relations between ti,r and xj,s± , we introduce the following formal power series (inverse Borel transform of ti(u)) vr B (v)= B(t (u)) = ~ t Y 0[[v]] (2.5) i i i,r r! ∈ r 0 X≥ Lemma. For any i, j I we have ∈ aij v aij v ± qi qi− σ v B (v),x± = − e j x± (2.6) i j,s ± v j,s h i aij Proof. To simplify notations set a = ~d a /2, so that ea = q± . By Lemma 2.7 ± i ij i
1 u σj± + a ξi(u)xj,s± ξi(u)− = − xj,s± u σ± a − j − so that 1 1 (σ± a)u− [t (u),x ] = log − j − x i j,s± 1 j,s± 1 (σ± + a)u − j − ! Using pv 1 1 e B log(1 pu− ) = − (2.7) − v this yields
± ± (σ a)v (σ +a)v av av 1 e j − 1 e j e e ± − σj v [Bi(v),xj,s± ]= − − xj,s± = − e xj,s± v − v ! v as claimed. YANGIANSANDQUANTUMLOOPALGEBRAS 13
Remark. Expanding the right–hand side of (2.6) as a power series in v yields the commutation relations r/2 2l ⌊ ⌋ r (~diaij/2) [ti,r,xj,s± ]= diaij xj,r± +s 2l ± 2l 2l + 1 − Xl=0 These relations were obtained in this form in [22, Lemma 1.4].
2.10. The operators λi±(v). We introduce below operators which straighten mono- 0 0 mials of the form x± ξ, ξ Y , into elements of Y Y . i,m ∈ · ± 0 Proposition. There are operators λi±;s i I,s N on Y such that the following holds { } ∈ ∈ 0 0 (1) For any ξ Y , the elements λ± (ξ) Y are uniquely determined by the ∈ i;s ∈ requirement that, for any m N, ∈ xi,m± ξ = λi±;s(ξ)xi,m± +s (2.8) s 0 X≥ (2) For any ξ, η Y 0, ∈ λi±;s(ξη)= λi±;k(ξ)λi±;l(η) (2.9) kX+l=s 0 0 (3) The operator λi;s : Y Y is homogeneous of degree s. 0 0 → − (4) Let λ±(v) : Y Y [v] be given by i → s λi±(v)(ξ)= λi±;s(ξ) v s 0 X≥ 0 0 and extend the N–grading on Y to Y [v] by deg(v) = 1. Then λi±(v) is an algebra homomorphism of degree 0. ǫ1 ǫ2 (5) λ (v1) and λ (v2) commute for any i1, i2 I and ǫ1,ǫ2 . i1 i2 ∈ ∈ {±} (6) For any i I, ∈ + + λi (v)λi−(v)=id= λi−(v)λi (v) (7) For any i, j I, ∈ aij v2 aij v2 − qi qi v1v2 λj±(v1) (Bi(v2)) = Bi(v2) − e ∓ v2 (8) For any i I and r N, ∈ ∈ r λ±(v)(t )= t d a v mod ~ j i,r i,r ∓ i ij Proof. (1)–(2) by Lemma 2.7, (2.8) holds when ξ is one of the generators ξj,r of 0 0 Y . Since (2.8) holds for ξη if it holds for ξ, η Y , with λi;s(ξη) given by (2.9), the 0 ∈ λi;s can be defined as operators on Y . The fact they are uniquely characterised by (2.8) and satisfy (2.9) follows from Corollary 2.5. (3) the linear independence of the elements on the right–hand side of (2.8) implies 0 that deg(λi;s(ξ)) = deg(ξ) s for any homogeneous ξ Y . (4) is a rephrasing of − ∈ 0 (2) and (3). (5) and (6) follow from (7) since the elements ti,n generate Y . (7) follows from Lemma 2.9. (8) is a direct consequence of (7). { } 14 S.GAUTAMANDV.TOLEDANOLAREDO
Remark. Using the shift operators, the relation (2.8) can be rewritten as
xi,m± ξ = λi±(σi±)(ξ) xi,m±
3. Homomorphisms of geometric type
Let Y~g be the completion of Y~(g) with respect to its N–grading. In this section, we define an assignment d Φ : Hi,r, Ei,r, Fi,r i I,r Z Y~g { } ∈ ∈ −→ and find necessary and sufficient conditions for Φ to extend to a homomorphism d U~(Lg) Y~g. → 3.1. Definitiond of Φ. Define 1 Φ(Hi,0)= di− ti,0 and, for r Z ∈ × k ~ r Bi(r) Φ(Hi,r)= ti,k = q q 1 k! q q 1 i i− k 0 i i− − X≥ − 0 where Bi(v) is the formal power series (2.5). Let U be the polynomial ring on 1 the generators Hi,r i I,r Z . The above assignment extends to an homomorphism { } ∈ ∈ Φ0 : U 0 Y 0. e → 0 Let now g± i I,m N be elements of Y , and define further { i,m} ∈ ∈ e c + + Φ(Ei,0)= c gi,mxi,m m 0 X≥ Φ(Fi,0)= gi,m− xi,m− m 0 X≥
In terms of the shift operators σi±, the above may be written as + + + Φ(Ei,0)= gi (σi )xi,0 (3.1)
Φ(Fi,0)= gi−(σi−)xi,−0 (3.2) where m 0 g±(v)= g± v Y[[v] i i,m ∈ m 0 X≥ with the completion of Y 0[v] taken with respect to the N–grading which extends that on Y 0 by deg(v) = 1.
1 e0 0 U is isomorphic to the subalgebra U of U~(Lg) generated by {Hi,r}i∈I,r∈Z by the PBW Theorem for U~(Lg) [1], but we shall not need this fact. YANGIANSANDQUANTUMLOOPALGEBRAS 15
[ 3.2. Homomorphisms of geometric type. If Φ : U~(Lg) Y~(g) is a homo- morphism of the above form, we shall say that it is of geometric→ type since its form is related to the Chern character in the geometric realisations of U~(Lg) and Y~(g) discussed in 1.2. Such a homomorphism has the following properties i \i i (1) It restricts to a homomorphism U~(Lsl2) Y~(sl2) for any I, where U~(Lsl2) i → U~(Lg) and Y~(sl2) Y~(g) are the subalgebras generated by Ei,r, Fi,r,Hi,r r Z ⊂ ⊂ { } ∈ and x± ,ξi,k k N respectively. { i,k } ∈ \ (2) It restricts to a homomorphism U~(Lb ) Y~(b ), where U~(Lb+),U~(Lb ) ± → ± − U~(Lg) and Y~(b ) Y~(g) are the subalgebras generated respectively by ⊂ ± ⊂ Ei,r,Hi,r i I,r Z, Fi,r,Hi,r i I,r Z and x± ,ξi,k i I,k N . { } ∈ ∈ { } ∈ ∈ { i,k } ∈ ∈ Note however that, unless g± = 0 for any i I and m 1, Φ does not map the i,m ∈ ≥ [ quantum group U~g = Ei,0, Fi,0,Hi,0 i I U~(Lg) to Ug[[~]] Y~(g). h i ∈ ⊂ ⊂ 3.3. The following result shows that the requirement that Φ extends to an algebra homomorphism determines its value on generators Ei,k, Fi,k. Proposition. The assignment Φ is compatible with relations (QL2)–(QL3) if, and only if
+ kσi + + + Φ(Ei,k)= e gi (σi )xi,0 (3.3) − kσi Φ(Fi,k)= e gi−(σi−)xi,−0 (3.4) Proof. We only consider the case of the E s. Let i, j I and k Z. By (Y2), ′ ∈ ∈ + 1 kσj + + + [Φ(Hi,0), Φ(Ej,k)] = [di− ξi,0, e gj (σj )xj,0] + kσj + + 1 + = e gj (σj )[di− ξi,0,xj,0] = aijΦ(Ej,k) so that Φ is compatible with (QL2). Next, if r Z , Lemma 2.9 yields ∈ × 1 + kσj + + + [Φ(Hi,r), Φ(Ej,k)] = 1 [Bi(r), e gj (σj )xj,0] q q− i − i raij raij q q− + + i i rσj kσj + + + = − 1 e e gj (σj )xj,0 r(q q− ) i − i [ra ] = ij qi Φ(E ) r j,r+k and Φ is compatible with (QL3).
Conversely, if Φ is compatible with (QL3) then Φ(Ei,r)= r/[2r]qi [Φ(Hi,r), Φ(Ei,0)] + for r = 0 and the computation above shows that this is equal to erσi Φ(E ). 6 i,0 16 S.GAUTAMANDV.TOLEDANOLAREDO
0 0 3.4. Necessary and sufficient conditions. Let λi±(v) : Y Y [v] be the ho- momorphism defined in Proposition 2.10. → [ Theorem. The assignment Φ extends to an algebra homomorphism U~(Lg) Y~(g) if, and only if the following conditions hold → (A) For any i, j I ∈ + + + gi (u)λi (u)(gj−(v)) = gj−(v)λj−(v)(gi (u)) (B) For any i I and k Z ∈ ∈ ku + + 0 ψi,k φi,k e g (u)λ (u)(g−(u)) =Φ − i i i m 1 u =ξi,m qi q− − i
(C) For any i, j I and a = d a /2 ∈ i ij u v a~ v u a~ e e ± e e ± g±(u)λ±(u)(g±(v)) − = g±(v)λ±(v)(g±(u)) − i i j u v a~ j j i v u a~ − ∓ − ∓ Proof. By construction and Proposition 3.3, Φ is compatible with the relations (QL1)–(QL3). The result then follows from Lemmas 3.5 and 3.6 below and the proof of the q–Serre relations (Proposition A.1 in the Appendix).
3.5.
Lemma. Φ is compatible with the relation (QL5) if, and only if (A) and (B) hold.
Proof. Compatibility with (QL5) reads
0 ψi,k+l φi,k+l [Φ(Ei,k), Φ(Fj,l)] = δijΦ − 1 q q− i − i for i, j I and k, l Z. We begin by computing the left–hand side. To this end, it will be∈ convenient to∈ write formulae (3.3)–(3.4) as
+,(k) + ,(k) Φ(Ei,k)= gi,m xi,m and Φ(Fi,k)= gi,m− xi,m− m 0 m 0 X≥ X≥ ,(k) 0 ,(k) m kv where gi,m± Y are defined by m 0 gi,m± v = e gi±(v). This yields ∈ ≥ +,(Pk) + ,(l) Φ(Ei,k)Φ(cFj,l)= gi,m xi,mgj,n− xj,n− m,n 0 X≥ +,(k) + ,(l) + = gi,m λi,s gj,n− xi,m+sxj,n− m,n,s 0 X≥ +,(k) + ,(l) + = gi,m λi,s gj,n− xj,n− xi,m+s + δijξi,m+n+s m,n,s 0 X≥ YANGIANSANDQUANTUMLOOPALGEBRAS 17
,(l) +,(k) + where we used (Y5). Similarly, Φ(Fj,l)Φ(Ei,k)= m,n,s gj,m− λj,s− gi,n xj,m− +sxi,n. Define R(k,l),L(k,l) Y 0[[u, v]] by P ∈
(k,l) +,(kc) m + s ,(l) n ku lv + + R = g u λ u g− v = e e g (u)λ (u)(g−(v)) i,m i,s j,n i i j m 0 s 0 n 0 X≥ X≥ X≥ (k,l) ,(l) m s +,(k) n ku lv + L = g− v λ− v g u = e e g−(v)λ−(v)(g (u)) j,m j,s i,n j j i m 0 s 0 n 0 X≥ X≥ X≥ By the PBW Theorem 2.5, Φ is compatible with (QL5) if, and only if R(k,l) = L(k,l) and, for i = j, ψ φ R(k,l) =Φ0 i,k+l − i,k+l m n 1 u v =ξi,m+n qi q− − i The first equation is clearly equivalent to (A) and the second to (B).
3.6. Lemma. Φ is compatible with the relation (QL4) if, and only if (C) holds. Proof. We prove the claim for the E’s only. Compatibility with (QL4) reads Φ(E )Φ(E ) qaij Φ(E )Φ(E )= qaij Φ(E )Φ(E ) Φ(E )Φ(E ) i,k+1 j,l − i i,k j,l+1 i j,l i,k+1 − j,l+1 i,k for any i, j I and k, l Z. Assume first i = j and set a = diaij/2, so that aij a~ ∈ ∈ 6 qi = e . Since + + rσi sσj + + + + + + + + Φ(Ei,r)Φ(Ej,s)= e e gi (σi )λi (σi ) gj (σj ) xi,0xj,0 the above reduces to
+ lσ+ + σ++a~ ekσi e j g+(σ+)λ+(σ+)(g+(σ+)) eσi e j x+ x+ i i i i j j − i,0 j,0 + lσ+ + ~ σ+ = ekσi e j g+(σ+)λ+(σ+)(g+(σ+)) eσi +a e j x+ x+ j j j j i i − j,0 i,0 Using (1) of Proposition 2.8, we get
+ + σ σj +a~ + σ++a~ e i e eσi e j x+ x+ = − σ+ σ+ a~ x+ x+ − i,0 j,0 σ+ σ+ a~ i − j − i,0 j,0 i − j − + σ++a~ eσi e j = − σ+ σ+ + a~ x+ x+ σ+ σ+ a~ i − j j,0 i,0 i − j − The PBW Theorem 2.5 then shows that the above is equivalent to (C). Assume now that i = j, then + + rσi + + + sσi + + + Φ(Ei,r)Φ(Ei,s)= e gi (σi )xi,0 e gi (σi )xi,0 rσ+ sσ+ = µ e i,(1) e i,(2) g+(σ+ )λ+(σ+ ) g+(σ+ ) x+ x+ i i,(1) i i,(1) i i,(2) i,0 ⊗ i,0 18 S.GAUTAMANDV.TOLEDANOLAREDO
The compatibility with (QL4) therefore reduces to
kσ+ lσ+ σ+ σ+ +d ~ µ e i,(1) e i,(2) g+(σ+ )λ+(σ+ )(g+(σ+ )) e i,(1) e i,(2) i x+ x+ i i,(1) i i,(1) i i,(2) − i,0 ⊗ i,0 lσ+ kσ+ σ+ +d ~ σ+ = µ e i,(1) e i,(2) g+(σ+ )λ+(σ+ )(g+(σ+ )) e i,(2) i e i,(1) x+ x+ i i,(1) i i,(1) i i,(2) − i,0 ⊗ i,0 that is, to
kσ+ lσ+ lσ+ kσ+ µ e i,(1) e i,(2) + e i,(1) e i,(2)
σ+ σ+ +d ~ g+(σ+ )λ+(σ+ )(g+(σ+ )) e i,(1) e i,(2) i x+ x+ = 0 i i,(1) i i,(1) i i,(2) − i,0 ⊗ i,0 By (2) of Proposition 2.8 and Corollary 2.8, this equation is equivalent to the re- quirement that eu ev+di~ g+(u)λ+(u)(g+(v)) − i i i u v d ~ − − i be symmetric under u v, which is precisely condition (C) for i = j. ↔
3.7. For later use, we shall need the following
\0 Lemma. Let gi±(u) i I Y [u] be elements satisfying condition (B) of Theorem 3.4. Then, { } ∈ ⊂
1 \0 gi±(u)= mod Y [u]+ di± + where di± i I C× satisfy di di− = di for each i I. In particular, each gi±(u) is invertible.{ } ∈ ⊂ ∈ Proof. Condition (B) for k = 0 yields
~ ~ di di Hi,0 Hi,0 + + 0 e 2 e− 2 g (u)λ (u) g−(u) m =Φ − i i i u =ξi,m 1 qi qi− ! − \0 Computing mod ~, and a fortiori mod Y [u]+, yields ~ ~ di di Hi,0 Hi,0 0 e 2 e− 2 0 1 Φ − 1 =Φ (Hi,0)= di− ti,0 q q− i − i ! \0 \0 Write gi±(u)= pi± mod Y [u]+, where pi± C[tj,0]j I. Computing mod Y [u]+, we get ∈ ∈ + + + + + g (u)λ (u) g−(u) m = p (tj,0)λ (p−(tj,0))ξi,0 = p (tj,0)p−(tj,0 diaij)ξi,0 i i i u =ξi,m i i;0 i i i − where we used (8) of Proposition 2.10. Comparing both sides and using ξi,0 = ti,0 yields the claim. YANGIANSANDQUANTUMLOOPALGEBRAS 19
4. Existence of homomorphisms [ In this section, we construct an explicit homomorphism U~(Lg) Y~(g) by ex- hibiting a joint solution to equations (A)–(C) of Theorem 3.4. We→ begin by giving an intrinsic expression for the right–hand side of equation (B) (Proposition 4.2). Until 4.7, we fix i I, and consider the subalgebras U 0 U 0, Y 0 Y 0 generated ∈ i ⊂ i ⊂ by Hi,r r Z and ξi,r r N respectively. { } ∈ { } ∈ e e 4.1. The functions G(v) and γi(v). Consider the formal power series v G(v) = log vQ[[v]] (4.1) ev/2 e v/2 ∈ − − Define γ (v) Y[0[v] by i ∈ i + t γ (v)= ~ i,r ( ∂ )r+1 G(v) i r! − v r 0 X≥ vr Recall that Bi(v)= ~ r 0 ti,r r! is the inverse Borel transform of ti(u). This allows us to write γ (u) more compactly≥ as i P γ (v)= B ( ∂ )G′(v) (4.2) i − i − v 4.2. Proposition. The following holds in Y 0 for any k Z i ∈ ψ φ ~ Φ0 i,k − i,k = c ekv exp (γ (v)) 1 1 i n qi q− qi q− v =ξi,n − i − i
The above identity will be proved in 4.3–4.6 by injectively mapping both sides to a family of polynomial rings, and verifying their equality there.
4.3. Universal Drinfeld polynomials. Fix an integer m 1. Following [26, 33], consider the rings ≥
1 1 1 Sm S(m)= C[q± , A1± ,...,Am± ]
Sm R(m)= C[~, a1, . . . , am] Define a homomorphism U : U 0 S(m) by D i → m 1 q z q− A U (ψ (z))e = i − i p = U (φ (z)) (4.3) D i z A D i p=1 p Y − where the first and second equalities are obtained by expanding the middle term in powers of z 1 and z respectively. Similarly, define Y : Y 0 R(m) by − D i → m Y u + di~ ap (ξi(u)) = − (4.4) D u ap pY=1 − 20 S.GAUTAMANDV.TOLEDANOLAREDO
The homomorphism U (resp. Y ) gives the action of ψ (z), φ (z) (resp. ξ (u)) on D D i i i the highest weight vector of the indecomposable U~(Lsl2) (resp. Y~(sl2)) module 1 1 with Drinfeld polynomial (1 A− z) (1 A z) (resp. (u a ) (u a )). − 1 · · · − m− − 1 · · · − m 0 0 4.4. The following result spells out the image of the generators of Ui and Yi under U and Y respectively. D D e Proposition. (1) The following holds: U (ψ )= qm = U (φ ) 1 and, for any r N D i,0 i D i,0 − ∈ ∗ m 1 ′ U 1 r qiAp qi− Ap (ψi,r) = (qi qi− ) Ap − (4.5) D − Ap Ap′ p=1 p′=p X Y6 − m 1 ′ U 1 r qiAp qi− Ap (φi, r)= (qi qi− ) Ap− − (4.6) D − − − Ap Ap′ p=1 p′=p X Y6 − Moreover, U maps H to m and, for any r Z , D i,0 ∈ ∗ 2r m U 1 1 qi− r ((q q− )H )= − A (4.7) D i − i i,r r p p=1 X (2) The homomorphism Y maps ξ to d m and, for any r N D i,0 i ∈ m Y r ap ap′ + di~ (ξr)= di ap − (4.8) D ap ap′ p=1 p′=p X Y6 − Moreover, for any r N, ∈ m r+1 r+1 1 a (ap di~) Y (t )= p − − (4.9) D r r + 1 ~ p=1 X (3) If B (v) Y 0[[v]] is the series defined by (2.5), then i ∈ i m 1 e di~v Y (B (v)) = − − eapv (4.10) D i v p=1 X U m U 1 Proof. (1) The fact that (ψi,0)= qi = (φi,0)− follows by taking the values of the middle term P (z) inD (4.3) at z = andD z = 0 respectively. Next, the partial fraction decomposition of P (z) is readily∞ seen to be
m m 1 1 ′ qiz qi− Ap m 1 qiAp qi− Ap 1 − = qi + (qi qi− ) Ap − (4.11) z Ap − Ap Ap′ z Ap p=1 p=1 p′=p Y − X Y6 − − 1 The relations (4.5)–(4.6) follow by expanding this into powers of z− and z respec- tively. For later use, note that the evaluation of (4.11) at z = 0 yields the identity m 1 ′ U m m 1 qiAp qi− Ap (ψi,0 φi,0)= qi qi− = (qi qi− ) − (4.12) D − − − Ap Ap′ p=1 p′=p X Y6 − YANGIANSANDQUANTUMLOOPALGEBRAS 21
Since U (ψ )= qm, it follows that U (H )= m, and D i,0 i D i,0 m 2 U 1 s U 1 z qi− Ap exp (q q− ) H z− = (ψ− ψ (z)) = − D i − i i,s D i,0 i z A s 1 p=1 p X≥ Y − 1 taking the log of both sides and expanding in powers of z− (resp. z) yields (4.7) for r> 0 (resp. r< 0). (2) The fact that Y (ξ )= d m follows by taking the coefficient of u 1 in (4.4). D i,0 i − The partial fraction decomposition of Y (ξ (u)) is D i m m u + di~ ap ap ap′ + di~ 1 − =1+ di~ − u ap ap ap′ u ap p=1 p=1 p′=p Y − X Y6 − − r 1 and (4.8) follows by taking the coefficient of u− − . Taking the log of both sides of (4.4) yields Y 1 1 (t (u)) = log 1 a u− + log 1 (a d ~)u− (4.13) D i − − p − p − i p X and therefore (4.9). (3) Follows by applying (2.7) to (4.4).
Y 0 Corollary. The homomorphism : Yi m 1 R(m) is injective. D → ≥ r Proof. This follows from (4.9) and the factL that the power sums p ap are alge- braically independent. P 4.5. Let R\(m) be the completion of R(m) with respect to the N–grading defined by deg(~) = deg(a ) = 1. Since the map Y : Y 0 R(m) preserves the grading, it p D i → extends to a homomorphism Y 0 R\(m). i → Corollary. Let ch : S(m) R(m) be the homomorphism defined by → c q e~/2 and A eap 7−→ p 7−→ Then, the following diagram commutes
U 0 / Ui D S(m)
Φ0 ch e Y 0 / \ Yi D R(m) 0 where Φ is defined in Section 3.1c. 0 Proof. It suffices to check the commutativity on the generators Hi,r r Z of Ui . { } ∈0 The statement now follows from (4.7), (4.10) and the fact that, for r =0,Φ (Hi,r)= 1 6 B (v)/(q q− ) . e i i − i v=r
22 S.GAUTAMANDV.TOLEDANOLAREDO
4.6. Proof of Proposition 4.2. By Corollaries 4.4 and 4.5, it suffices to prove that, for any k Z, ∈ ψ φ ~ ch U i,k − i,k = Y ekv exp (γ (v)) 1 1 i n D qi q− qi q− D v =ξn − i − i
By (4.5)–(4.6) and (4.12), the left–hand side is equal to
1 ′ k qiAp qi− Ap ch Ap − Ap Ap′ p p′=p X Y6 − We now compute the right–hand side. By (4.2) and (4.10),
di~∂v Y ap∂v 1 e (γ (v)) = e− − ∂ G(v)= G(v a ) G(v a + d ~) D i ∂ v − p − − p i p v p X X where we used ea∂v G(v)= G(v + a). Thus,
v 1 ap Y kv kv v ap qie qi− e e exp (γi(v)) = e − − D v a + d ~ ev eap p p i Y − − By (4.8), the substitution vn = ξ in a formal power series F Y[0[v] gives i,n ∈ i ′ Y Y ap ap + di~ n (F (v) v =ξn )= di (F (ap)) − D | D ap ap′ p p′=p X Y6 − v ap ap v 1 ap Since (v ap)/(e e ) = e− and (qie q− e )/(v ap + di~) = − − |v=ap − i − v=ap ap 1 e (qi qi− )/di~, this implies that − Y kv e exp (γ (v)) n D i |v =ξi,n ′ 1 ′ ap 1 ap ′ qi qi− kap ap ap qie qi− e ap ap + di~ = − e − a − a ′ − ~ ap ap′ + di~ e p e p ap ap′ p p′=p X Y6 − − − 1 a 1 a ′ q q− q e p q− e p = i − i ekap i − i ~ eap eap′ p p′=p X Y6 − as claimed.
4.7. A joint solution to equations (A)–(C). Proposition 4.2 suggests replacing equation (B) of Theorem 3.4 by the stronger requirement that, for any i I ∈ + + ~ gi (v)λi (v)(gi−(v)) = 1 exp (γi(v)) (B) q q− i − i Using equation (A) for j = i and u = v then shows that e
+ ~ gi−(v)λi−(v)(gi (v)) = 1 exp (γi(v)) q q− i − i YANGIANSANDQUANTUMLOOPALGEBRAS 23
+ Applying the operator λi−(v) to the first of these equations, and using λi (v)λi−(v)= id (Proposition 2.10) and the second equation, yields λi−(v)(γi(v)) = γi(v). Simi- + + larly, applying λi (v) to the second of these equations yields λi (v)(γi(v)) = γi(v). This suggests in turn that a solution to the above equations may be given by gi±(v)= gi(v), where 1/2 ~ γi(v) gi(v)= 1 exp (4.14) q q− 2 i − i We now show that this is indeed the case.
Theorem. The series gi±(v) = gi(v) satisfy the conditions (A),(B) and (C) of Theorem 3.4, and therefore give rise to a homomorphism Φ : U~(Lg) Y~g. e→ 4.8. We shall need the following d Lemma. Let i, j I, and set a = d a /2. Then, ∈ i ij G(v u + a~) G(v u a~) λ±(u) (g (v)) = g (v)exp − − − − i j j ± 2 where G(v) is given by (4.1).
Proof. By Proposition 2.10,
a~v a~v e e− uv λ±(u)(B (v)) = B (v) − e i j j ∓ v Since γ (v)= B ( ∂ )∂ G(v), we get j − j − v v a~∂v a~∂v e e− u∂v λi±(u)γj(v)= γj(v) − e− ∂vG(v) ± ∂v = γ (v) (G(v u + a~) G(v u a~)) j ± − − − − The claim follows by exponentiating.
4.9. Proof of condition (A). We need to prove that for every i, j I, we have ∈ + gi(u)λi (u)(gj (v)) = gj(v)λj−(v)(gi(u)) By Lemma 4.8, this is equivalent to
G(v u + a~) G(v u a~) g (u)g (v)exp − − − − i j 2 G(u v a~) G(u v + a~) = g (u)g (v)exp − − − − i j 2 The result now follows since G is an even function. 24 S.GAUTAMANDV.TOLEDANOLAREDO
4.10. Proof of condition (B). Lemma 4.8 implies that G(d ~) G( d ~) g (u)λ+(u)(g (eu)) = g (u)2 exp i − − i i i i i 2 2 = gi(u) ~ = 1 exp (γi(u)) q q− i − i where the second equality holds because G is even.
4.11. Proof of condition (C). Let i, j I and set a = diaij/2. We need to prove that ∈ u v a~ v u a~ e e ± e e ± g (u)λ±(u)(g (v)) − = g (v)λ±(v)(g (u)) − i i j u v a~ j j i v u a~ − ∓ − ∓ By Lemma 4.8 and the fact that G is even, we get the following equivalent assertion eu ev a~ ev eu a~ exp (G(v u a~) G(v u a~)) − ± = − ± − ± − − ∓ u v a~ v u a~ − ∓ − ∓ Using the definition of G, the above becomes the equality v u a~ ev eu a~ eu ev a~ ev eu a~ − ± − ± − ± = − ± ev a~ eu v u a~ u v a~ v u a~ ± − − ∓ − ∓ − ∓
5. Uniqueness of homomorphisms The aim of this section is to prove that homomorphisms of geometric type are unique up to conjugation and scaling.
5.1. Let be the set of solutions g = gi±(u) i I of equations (A)–(C) of Theorem G { } ∈ \0 3.4. Given a collection r = ri±(u) i I of invertible elements of Y [u], set { } ∈
r g = ri±(u) gi±(u) i I · { · } ∈ Lemma. Let g . Then, r g if and only if the following holds ∈ G · ∈ G (A ) For any i, j I, 0 ∈ + + + ri (u)λi (u)(rj−(v)) = rj−(v)λj−(v)(ri (u)) (B ) For any i I, 0 ∈ + + + ri (u)λi (u)(ri−(u))=1= ri−(u)λi−(u)(ri (u))
(C±) For any i, j I, 0 ∈
ri±(u)λi±(u)(rj±(v)) = rj±(v)λj±(v)(ri±(u)) YANGIANSANDQUANTUMLOOPALGEBRAS 25
Proof. Let h = r g. The following assertions are straightforward to check · h satisfies (A) if and only if r satisfies (A0). • h satisfies (B) if r satisfies (B ). • 0 h satisfies (C) if and only if r satisfies (C±). • 0 There remains to prove that if h lies in , then r satisfies (B0). G + + We claim that (A0) and (C0±) imply that ci(u) = ri (u)λi (u)(ri−(u)) lies in C[[~, u]]. Assuming this, write c (u)= c(n)un, where c(n) C[[~]]. Then, i n i i ∈ + + + + h (u)λ (u)(h−(u)) m = Pci(u)g (u)λ (u)(g−(u)) m i i i u =ξi,m i i i u =ξi,m