The representation theory of cyclotomic Temperley-Lieb algebras

Transcription

1 The representation theory of cyclotomic Temperley-Lieb algebras Hebing Rui Department of Mathematics, East China Normal University, Shanghai, PRChina; Changchang Xi Department of Mathematics, Beijing Normal University, Beijing, PRChina; Abstract A class of associative algebras called cyclotomic Temperley-Lieb algebras is introduced in terms of generators and relations They are closely related to the group algebras of complex reflection groups on the one hand and generalizations of the usual Temperley-Lieb algebras on the other hand It is shown that the cyclotomic Temperley-Lieb algebras can be defined by means of labelled Temperley-Lieb diagrams and are cellular in the sense of Graham and Lehrer One thus obtains not only a description of the irreducible representations, but also a criterion for their quasi-heredity in the sense of Cline, Parshall and Scott The branching rule for cell modules and the determinants of Gram matrices for certain cell modules are calculated, they can be expressed in terms of generalized Tchebychev polynomials, which therefore play an important role for semisimplicity Introduction The Temperley-Lieb algebras were first introduced in 97 in the paper [5] where they were used to study the single bond transfer matrices for the Ising model Later they were independently found by Jones when he characterized the algebras arising from the tower construction of semisimple algebras in the study of subfactors Their relationship with knot theory comes from their role in the definition of the Jones polynomial The theory of quantum invariants of links nowadays involves many of research fields Thus, many important kinds of algebras related to the invariants of braids or links, such as Birman-Wenzl algebras[3], Hecke algebras and Brauer algebras, have been of great interest in mathematics and physics They are all deformations of certain group algebras or other well-known algebras Recently, several interesting type of such algebras have emerged: the cyclotomic Birman-Murakami-Wenzl algebras are introduced in [6] and cyclotomic Brauer algebras are investigated in [4] (see also [6]), while the cyclotomic Hecke algebras were already introduced by Broué and Malle in [4], and independently by Ariki and Koike for type G(m,, n) in [] They are deformations of the unitary reflection groups In the present paper, we focus our attention on the study of cyclotomic Temperley-Lieb algebras, which are generalizations of the classical Temperley-Lieb algebras They are also subalgebras of cyclotomic Brauer algebras, which are closely related to complex reflection groups We first present the ring theoretic definition of the cyclotomic Temperley-Lieb algebras in terms of generators and relations Then we show that our definition can be reformulated geometrically by means of labelled Temperley-Lieb diagrams Using this description we are able to prove that the cyclotomic Temperley-Lieb algebras are cellular, a notion introduced in [7] As a consequence, 2000 Mathematics Subject Classification: 6G0, 6K20, 7B0, 8G20, 20C05, 20G05, 57M25, 8R05 Key words: Temperley-Lieb algebra, cellular algebra, cyclotomic Temperley-Lieb algebra

2 we obtain both, the classification of the irreducible representations of the cyclotomic Temperley- Lieb algebras, and a criterion for a cyclotomic Temperley-Lieb algebra to be quasi-hereditary For cell modules, the branching rule is discussed, and also the discriminants of certain bilinear forms are calculated This leads us to introduce the n-th generalized Tchebychev polynomials It turns out that a necessary condition for a cyclotomic Temperley-Lieb algebra to be semisimple is that certain generalized Tchebychev polynomials do not vanish on its defining parameters 2 The ring theoretic definition of cyclotomic Temperley- Lieb algebras Throughout the paper, let R be a commutative ring containing an identity and elements δ 0, δ,, δ m Let n, m N be two positive integers In this section, we introduce the cyclotomic Temperley-Lieb algebra T L n,m (δ 0,, δ m ) of type G(m,, n) over R We shall prove that the R-rank of T L n,m (δ 0,, δ m ) is at most mn n+( 2n n ) Definition 2 The cyclotomic Temperley-Lieb algebra T L n,m (δ 0,, δ m ) (or T L n,m for simplicity) is the associative algebra over R with generators (the identity), e,, e n, t,, t n subject to the following conditions: () e i e j e i = e i if j i =, (2) e i e j = e j e i if j i >, (3) e 2 i = δ 0 e i for i n, (4) t m i = for i n, (5) t i t j = t j t i for i, j n, (6) e i t k i e i = δ k e i for k m, i n, (7) t i t i+ e i = e i, e i t i t i+ = e i for i n, (8) e i t j = t j e i if j {i, i + } If m =, then T L n,m is the usual Temperley-Lieb algebra, which is denoted by T L n (δ 0 ) or T L n for simplicity This algebra was first introduced in [5] to describe the transfer matrices for the Ising model and for the Potts model in statistical mechanics (see also [2]) It is known that dim R T L n = 2n ) n + ( n if R is a field The following lemma is due to Jones [8] Recall that an expression of a monomial w T L n (δ 0 ) (in the variables e, e 2,, e n ) is called reduced if the number of e i in the expression is minimal Lemma 22 () Any monomial w T L n (δ 0 ) has a reduced expression (e j e j e k )(e j2 e j2 e k2 ) (e jp e jp e kp ), where j i+ > j i k i, k i+ > k i for any i p (2) For any n, there is an isomorphism of T L n -modules T L n (δ 0 ) = T L n (δ 0 ) T L n (δ 0 )e n T L n (δ 0 ), where T L n (δ 0 ) is the subalgebra of T L n (δ 0 ) generated by, e,, e n 2 To obtain an upper bound on the rank of a cyclotomic Temperley-Lieb algebra, we need the following lemma Lemma 23 For any n, the cyclotomic Temperley-Lieb algebra T L n,m (δ 0,, δ m ) is spanned over R by the set M n = {t k tk 2 2 tk n n xt l t l 2 2 t l n n 0 k i, l i m, i n, x T L n (δ 0 )} 2

3 Proof We claim that the R-module T L n,m spanned by M n is a left T L n,m -module This claim implies T L n,m = T L n,m (δ 0,, δ m ) since M n By the definition of M n, we see that T L n,m is stable under the left multiplication of t i, i n So we have to prove that for j n, ( ) e j t k tk 2 2 tkn n xt l t l 2 2 t ln n T L n,m Without loss of generality, we may assume that x is a monomial in e, e 2,, e n First, we consider the case j = n By Lemma 22, x = (e j e j e k )(e j2 e j2 e k2 ) (e jp e jp e kp ) By 2(8), xt n = t n x if j p n It follows from 2(7) that n n e n ( t ki i )x( n 2 t li i ) = ( n t ki i )e n x( t li i )tl n+k n k n n T L n,m i= i= i= Suppose j p = n If e n 2 does not occur in the expression e j e k e jp e kp, then ( ) follows from the following equality e n t k n k n n x = e n t k n k n n e n (e j e kp )(e n 2 e kp ) = δ k x, where k n k n k (mod m) If e n 2 occurs in the expression of e j e jp e kp, then e jp = e n 2 In this case, we have e n t k n k n n x =(e j e kp 2 )(e n t kn kn n e n 2 e n )(e n 3 e kp )(e n 2 e kp ) i= =(e j e kp 2 )t k n k n n 2 (e n 3 e kp )(e n e n 2 e kp ) If e n 3 does not occur in e j e kp 2, then (e j e kp 2 )t k n k n n 2 = t k n k n n 2 (e j e kp 2 ), and ( ) follows If e n 3 occurs in the expression of e j e jp 2 e kp 2, then e jp 2 = e n 3 In this case, ( ) follows from the argument similar to the case e jp = e n 2 together with an induction Thus we have proved ( ) in the case j = n For j n 2, we use induction on n In this case, e j t n = t n e j If e n does not occur in the expression of x, then ( ) follows from the induction assumption on n Now suppose that x = y(e n e n 2 e k ) for some y T L n and k N Note that e j+ e j t l j+2 = (e j+t l j+2 )e j = e j+ t l j+ e j = e j+ (t l j+ e j) = e j+ t l j e j = t l j e j+e j for all l and j By a direct computation, we have ( ) e n e k t l t ln n = t l t l k k tl k+2 k t l n n 2 e n e k t l k l k+ k Again by the induction hypothesis on n, we see that e j ( n i= tk i i )ytl t l k k tl k+2 k t l n n 2 can be expressed as a linear combination of the elements in M n Now, ( ) together with the 2(7)-(8) yields the desired form ( ) This completes the proof of the result The following lemma gives more explicit information on the elements in M n, which leads to an upper bound on the rank of T L n,m Lemma 24 For any x T L n, the element w = ( n i= tk i i )x( n j= tlj j ) M n with 0 k i, l j m can be written as ( p i= tk i j i )x( n i=p+ tl i ji ) with 0 k i, l j m Proof Without loss of generality, we may assume that x = (e j e j e k )(e j2 e j2 e k2 ) (e jp e jp e kp ) Suppose j p n Then x T L n,m and hence t n x = xt n Therefore, n n ( t k i i )x i= j= t l j j n = ( n t k i i )x( i= j= t l j j )tk n+l n n 3

4 By induction on n, the element ( n i= tki i )x n with 0 k i, l j m This proves the result Suppose j p = n By ( ), j= tl j j can be written as ( p i= tk i j i )x n i=p+ tl i ji w = = n i= n i= t ki i t k i i p (e ji e ji e ki ) i= p (e ji e ji e ki ) i= k i= k i= t li i t l i i n 2 i=k n 2 i=k t li+2 i (e n e k )t l k l k+ k t l i+2 i t m k n n (e n e j )t l k l k+ k Now the result follows immediately from the induction assumption, 2(8) and ( ) This completes the proof of Lemma 24 Let us remark that the proof of this lemma also shows that for a fixed x T L n, when we write w as the form ( p i= tk i j i )x( n i=p+ tl i ji ) with 0 k i, l j m, the lower index sets {j,, j p } and {j p+,, j n } depend only on x Corollary 25 If R is a field, then dim R T L n,m m n dim R T L n = mn 2n ) n + ( n In the next section, we shall show that over a commutative ring R the rank of T L n,m is equal to mn n+( 2n n ) Finally, let us point out that the notion of B-type Temperley-Lieb algebras was introduced by tom Dieck [6], whose approach was based on the knot theoretic point of view and root systems In fact, these algebras are completely different from our cyclotomic Temperley-Lieb algebras since the dimension of the B-type Temperley-Lieb algebra over a field is always of the form ( ) 2n n (see [6]) However, the algebra T Ln,m is closely related to the complex reflection groups W n,m of type G(m,, n) Recall that W n,m is generated by s 0, s,, s n satisfying the relations () s 2 i = for i and the braid relations for s,, s n ; (2) s m 0 =, and (3) s 0 s s 0 s = s s 0 s s 0, s 0 s i = s i s 0 for i 2 If we define t = s 0, t i = s i t i s i, then t m i = Thus, a deformation of the group algebra of W n,m is the cyclotomic Brauer algebra, which is clearly related to cyclotomic Birman-Wenzl algebra as mentioned in [6] T L n,m is a subalgebra of the cyclotomic Brauer algebra Thus it is related in this way to both, the complex reflection group W n,m, and the cyclotomic Brauer algebra As we know, Ariki-Koike algebras are deformations of the unitary reflection groups But these algebras can also be viewed as deformations of certain products of cyclic groups and Hecke algebras In this same way, the cyclotomic Temperley-Lieb algebras are deformations of certain products of cyclic groups and Temperley-Lieb algebras On the other hand, it is known that there are nice relationships between Temperley-Lieb algebra and the quantum group U q (sl 2 ) (see [2]) 3 The graphical definition of cyclotomic Temperley-Lieb algebras In this section, we shall redefine the cyclotomic Temperley-Lieb algebra in a geometrical way This is motivated by knot theory Let us denote by T L n,m the graphical cyclotomic Temperley-Lieb algebra The main result in this section is that the ring theoretic and the graphical definitions of cyclotomic Temperley-Lieb algebras are equivalent, namely, T L n,m = T Ln,m for any n and m First, we introduce labelled Temperley-Lieb diagrams These are special cases of dotted Brauer graphs introduced in [6]( see also [4]) Definition 3 A labelled Temperley-Lieb diagram D of type G(m,, n) is a Temperley- Lieb diagram with 2n vertices in which the arcs are labelled by the elements of Z m := Z/mZ 4

5 In the following a labelled Temperley-Lieb diagram D will simply be called a labelled TLdiagram; if i and j are the endpoints of an arc in D, we shall simply write {i, j} D Graphically, we may represent a labelled TL-diagram D of type G(m,, n) in a rectangle of the plane, where there are n numbers {, 2,, n} on the top row from left to right, and there are another n numbers {, 2,, n} on the bottom row again from left to right To indicate the label i Z m on an arc, we mark the arc with a dot and write the label i in parentheses above or below the dot Sometimes we draw i dots directly on the arc For example, the following is a labelled TL-diagram of type G(m,, 6) with m (m ) An arc in a labelled TL-diagram is said to be horizontal if its endpoints both lie in the top row or in the bottom row; and otherwise it said to be vertical In order to have a graphical version of cyclotomic Temperley-Lieb algebras, we need to define the multiplication of two labelled TL-diagrams Here we follow the definition in [6] (see also [4]) From here onward, we make the following convention : Given a horizontal arc {i, j} with i < j, we call i (resp j) the left (resp right) endpoint of the arc {i, j}, and always assume that all dots in a labelled TL-diagram are marked at the left endpoints of the arcs A dot marked at the left (or right) endpoint of an arc will be called a left (or right) dot of the arc For a vertical arc we do not define its left endpoint and its right endpoint The rule for movements of dots We allow dots to move along an arc from left to right They may also move to another arc () A left dot of a horizontal arc {i, j} is equal to m right dots of the arc {i, j}, and conversely, a right dot of an horizontal arc is equal to m left dots (2) A dot on a vertical arc can move freely to the endpoints of the arc (3) Given two distinct arcs {i, j} and {j, k}, we allow that a dot at the endpoint j of the arc {i, j} can be replaced by a dot at the endpoint j of the arc {j, k} The rule for compositions Given two labelled TL-diagrams D and D 2 of type G(m,, n), we define a new labelled TL- diagram D D 2, called the composition of D and D 2, in the following way: First, we compose D and D 2 in the same way as was done for Temperley-Lieb algebras Thus we have a new Temperley-Lieb diagram P (which is possibly not a labelled TL-diagram) Second, we apply the rule for movements to relabel each arc in P, and thus obtain a labelled TL-diagram graph, denoted by D D 2 The rule for counting closed cycles For each closed cycle appearing in the above natural concatenation of D and D 2 we apply the rule for movements of dots to relabel the cycle Note that the number of dots in each cycle lies in Z/(m) We denote by n(ī, D, D 2 ) the number of relabelled closed cycles in which there are i dots The following lemma can be proved easily Lemma 32 Given two labelled TL-diagrams D and D 2, we define D D 2 = m i=0 δn(ī,d,d 2 ) i D D 2 Then (D D 2 ) D 3 = D (D 2 D 3 ) for arbitrary labelled TL-diagrams D, D 2 and D 3 Definition 33 Let R be a commutative ring containing and δ 0,, δ m A graphical cyclotomic Temperley-Lieb algebra ( T L n,m, ) is an associative algebra over R with a basis consisting of all labelled TL-diagrams of type G(m,, n) The multiplication is given by D D 2 = m,d 2 ) i=0 δn(ī,d i D D 2 5

6 It is easy to see that T L n,m is the usual Temperley-Lieb algebra if m = and that T L n,m is a subalgebra of the cyclotomic Brauer algebra of type G(m,, n) (see [6]) Now let us illustrate this definition by an example If =, D 2 = D, then we have a diagram D = Thus the composition D D 2 of D and D 2 is as follows: D D 2 = (m 3) Now we relabel the closed cycles in D By definition, = = In this case, n( 0, D, D 2 ) = n(, D, D 2 ) = 0 and n( 2, D, D 2 ) = n( 3, D, D 2 ) = for m 4 Thus D D 2 = δ δ 3D D 2 for m 4 Now let us prove that the graphical definition and the ring theoretic definition of cyclotomic Temperley-Lieb algebras coincide Theorem 34 Suppose that R is a commutative ring containing, δ 0,, δ m Then T L n,m = T L n,m for any m and n Therefore, T L n,m is a free R-module of rank n+( mn 2n ) n In particular, if R is a field, then dim R T L m,n = mn 2n ) n + ( n Proof Let E i be the labelled TL-diagram in which {i, i + } is an arc in the top row and also an arc in the bottom row, and the other vertex j ( i, i + ) in the top row connects with the vertex j in the bottom row Let T i be the labelled TL-diagram in which the j-th vertex in the top row connects with the j-th vertex in the bottom row for j =, 2,, n, and the i-th vertical arc carries one dot If we replace e i with E i and t i with T i and apply the three rules above, then we know that all E i and T i satisfy the relations in Definition 2 This induces an algebra homomorphism φ : T L n,m T L n,m with φ(t i ) = T i and φ(e i ) = E i Since T L n,m is generated as an R-algebra by E i and T j with i n, j n, the map φ is surjective We show that T L n,m is a free R-module Put r = n+( mn 2n ) n By Lemma 24, there is a surjective R-module homomorphism f : R r T L n,m Thus, we have a surjective R-module homomorphism φf from the free R-module R r to the free R-module T L n,m of rank r Let K be the kernel of φf Then we have a split exact sequence of R-modules: 0 K R r R r 0 6

7 Here we identify the R-module T L n,m with R r This sequence also shows that K is a finitely generated projective R-module We claim K = 0 Let p be a maximal ideal in R Since localization preserves (split) exact sequences, we have a split exact sequence 0 K p (R p ) r (R p ) r 0, where M p stands for the localization of an R-module M at p Thus (R p ) r (R p ) r K p as R p -modules Since R p is a local ring and every finitely generated projective module over a local ring is free, we see that the R p -module K p is free Note that any commutative ring with identity has the invariant dimension property It follows from (R p ) r (R p ) r K p that K p = 0, and therefore K = 0 (All facts on localization used in the above argument can be found in standard text books on commutative rings, for example [2]) If K = 0, then φf is an isomorphism of R-modules and f must be injective Thus T L n,m is a free R-module of rank r and φ is an isomorphism of R-modules This also implies that φ is an isomorphism of R-algebras The proof is complete Finally, let us remark that in [3] the so called blob algebras are considered, but those algebras have different defining relations and are therefore completely different from our cyclotomic Temperley-Lieb algebras 4 Cellular algebras Now let us recall the definition of cellular algebras due to Graham and Lehrer Definition 4 (Graham and Lehrer [7]) An associative R algebra A is called a cellular algebra with cell datum (I, M, C, i) if the following conditions are satisfied: (C) The finite set I is partially ordered Associated with each λ I there is a finite set M(λ) The algebra A has an R basis CS,T λ where (S, T ) runs through all elements of M(λ) M(λ) for all λ I (C2) The map i is an R linear anti automorphism of A with i 2 = id which sends CS,T λ to CT,S λ (C3) For each λ I and S, T M(λ) and each a A, the product acs,t λ can be written as ac λ S,T = U M(λ) r a (U, S)C λ U,T + r, where r belongs to A <λ consisting of all R- linear combination of basis elements with upper index µ strictly smaller than λ, and the coefficients r a (U, S) R do not depend on T In this paper, we call an R linear anti automorphism i of A with i 2 = id an involution of A The following is a basis-free definition of cellular algebras in [9] which is equivalent to that given by Graham and Lehrer Definition 42 Let A be an R algebra Assume there is an anti-automorphism i on A with i 2 = id A two sided ideal J in A is called a cell ideal if and only if i(j) = J and there exists a left ideal J such that is finitely generated and free over R and that there is an isomorphism of A-bimodules α : J R i( ) (where i( ) J is the i image of ) making the following diagram commutative: J i J α α R i( ) x y i(y) i(x) R i( ) The algebra A (with the involution i) is called cellular if and only if there is an R-module decomposition A = J J 2 J n (for some n) with i(j j ) = J j for each j and such that 7

8 setting J j = j l= J l gives a chain of two sided ideals of A: 0 = J 0 J J 2 J n = A (each of them fixed by i) and for each j (j =,, n) the quotient J j = J j/j j is a cell ideal (with respect to the involution induced by i on the quotient) of A/J j (We call this chain a cell chain for the cellular algebra A) Cellular algebras include a large variety of important algebras related to links in knot theory such as cyclotomic Hecke algebras, Temperley-Lieb algebras [7] and cyclotomic Brauer algebras [4] as well as Birman-Wenzl algebras [8] Given a cellular algebra A with the cell datum (I, M, C, i), for each λ I, one can define a cell module (λ) and a symmetric, associative bilinear form Φ λ : (λ) R (λ) R in the following way (see [7]): As an R-module, (λ) has an R-basis {CS λ S M(λ)} and the module structure is given by acs λ = r a (U, S)CU λ, U M(λ) where the coefficients r a (U, S) are determined by (C3) in Definition 4 The bilinear form Φ λ is defined by Φ λ (C λ S, C λ T )C λ U,V C λ U,SC λ T,V (mod A <λ ), where U and V are arbitrary elements in M(λ) Let rad (λ)= {c (λ) Φ λ (c, c ) = 0 for all c (λ)} Then rad (λ) is a submodule of (λ) Put L(λ) = (λ)/rad (λ) Then a complete set of irreducible representations of A can be described as follows: Lemma 43 (Graham and Lehrer [7]) Suppose R is a field Then () {L(λ) Φ λ 0} is a complete set of non-isomorphic irreducible A-modules (2) The algebra A is semisimple if and only if all cell modules are simple and pairwise nonisomorphic In the following, we shall see an easy example of cellular algebras, which will be used later on Let G m,n be the R-subalgebra of T L n,m generated by t, t 2,, t n Note that G m,n is isomorphic to the group algebra of the abelian group n i= Z/(m) t m i Suppose that R is a splitting field of x m Therefore the relation t m i = implies that = m j= (t i u j ) = 0 for some u,, u m R Let Λ(m, n) = {(i, i 2,, i n ) i j m} We assume that in case n = 0 the set Λ(m, n) consists of only one element Now we define (i, i 2,, i n ) (j, j 2,, j n ) if and only if i k j k for all k n For each i = (i, i 2,, i n ), define C i, = n m j= l=i j + (t j u l ) (Here the product over the empty set is ) Note that {C, i i Λ(m, n)} is a cellular basis for the algebra G m,n with respect to the identity involution Let us remark that in this case each cell G m,n -module (i) is one-dimensional In fact, this cell G m,n -module corresponds to the subquotient G i m,n/g <i m,n The simple G m,n -modules are parametrized by the following set Lemma 44 Suppose R is a splitting field of x m with characteristic p () If p divides m, say m = p t s with (p, s) =, then a complete set of non-isomorphic simple G m,n -modules can be chosen as {L(i) i = (i, i 2,, i n ) with p t divides i j for all j} its cardinality is s n (2) If p does not divide m (for example, p = 0), then a complete set of non-isomorphic simple G m,n -modules is {L(i) i Λ(m, n)} In this case, the algebra G m,n is semisimple 8

9 Proof It is easy to check that m m (t j u l ) (t j u k ) = (u i u l ) (t j u k ) + k>i It follows from the above equality that n C,C i, i ( j= k>i m k>i m (u ij u k ))C, i (mod G <i m,n) k>i j (t j u k ) If p divides m, then we see that each root of the polynomial x s is a root of x m with multiplicity p t But all roots of x s are simple roots Hence we may assume that (u, u 2,, u m ) = (,,, ξ,, ξ,, ξ s,, ξ s ), where ξ is a primitive s-th root of x s Thus () follows If p does not divide m, then the algebra G m,n is semisimple, and therefore (2) follows 5 Irreducible representations of T L n,m In this section, we assume that R is a splitting field of x m We shall prove that T L n,m is a cellular algebra in the sense of [7] Using the standard results on cellular algebras, we classify the irreducible representations of T L n,m over the field R Let us first introduce some auxiliary notions An (n, k)-labelled parenthesized graph is a graph consisting of n vertices {, 2,, n} and k horizontal arcs (hence 2k n and there are n 2k free vertices which do not belong to any arc), and satisfying the following conditions: () there are at most m dots on an arc, (2) there are no arcs {i, j} and {q, l} satisfying i < q < j < l, and (3) there is no arc {i, j} and free vertex q such that i < q < j (Given an (n, k)-labelled parenthesized graph, the vertices which do not belong to any arc are called free vertices) Let P (n, k) be the set of all (n, k)-labelled parenthesized graphs and let V (n, k) be the free R-module with P (n, k) as its basis Recall that G m,n is the R-subalgebra of T L n,m generated by t, t 2,, t n Lemma 5 There is an R-module isomorphism V (n, k) R V (n, k) R G m,n 2k = Mn,k, where M n,k is the free R-module spanned by all labelled TL-diagrams with 2n vertices and 2k horizontal arcs Proof Given a labelled TL-diagram D, we can write it uniquely as D D 2 x, where D i is obtained from D in the following manner: After cutting all vertical arcs and forgetting all dots on the vertical arcs, the top row is defined to be the D and the bottom is the D 2 Suppose that in D the free vertices are {i, i 2,, i n 2k } and that in D 2 the free vertices are {j, j 2,, j n 2k } Then in D the vertical arcs are {i, j },, {i n 2k, j n 2k } Suppose there are m s dots on the arc {i s, j s } Then we define x = t m tm 2 2 tm n 2k n 2k G m,n 2k Conversely, given such an expression D D 2 x, we have a unique labelled TL-diagram D in M n,k Hence the result follows Thus we have the following equivalent description of the graphical basis of T L n,m Usually, this basis is not a cellular basis Corollary 52 The set {v v 2 x 0 k [n/2], v, v 2 P (n, k), x G m,n 2k } is a basis of T L n,m In the following, we shall construct a cellular basis for T L n,m Here we keep the notation introduced in the previous section Let Λ m,n = {(k, i) 0 k [n/2], i Λ(m, n 2k)} We define a partial order on Λ m,n by saying that (k, i) (l, j) if k > l; or if k = l and i j Then (Λ m,n, ) is a finite poset For each 9

10 (k, i) Λ n,m, let I(k, i) = {(v, i) v P (n, k)} In the following, we shall show that this datum defines a cellular algebra Theorem 53 Let R be a splitting field of x m Then T L n,m is a cellular algebra with respect to the involution σ which sends v v 2 x to v 2 v x for all v, v 2 P (n, k) and x G m,n 2k, 0 k [n/2] Proof For any (k, i) Λ m,n and v, v 2 P (n, k), we define C v (k,i),v 2 = v v 2 C, i By 52, the set {C v (k,i),v 2 (k, i) Λ n,m, v, v 2 P (n, k)} is a basis of T L n,m We show that it is a cellular basis Let us verify the conditions in Definition 4 By definition, 4(C)-(C2) follow It remains to check the condition 4(C3) Take a labelled TL-diagram D D 2 x with D, D 2 P (n, k) and x = t m tm 2 2 tm n 2k n 2k G m,n 2k Suppose that i, i 2,, i n 2k are the free vertices in D and that j, j 2,, j n 2k are the free vertices of D 2, where i s n and j s n for all s =, 2,, n 2k Then D D 2 x = X (D D 2 id n 2k ) = (D D 2 id n 2k ) Y, where X = T m i T m 2 i 2 T m n 2k i n 2k and Y = T m j T m 2 j 2 T m n 2k j n 2k (see 34 for the definition of T i ) Thus, for any labelled TL-diagram D D 2 x, (D D 2 x) C v (k,i),v 2 T L (k,i) n,m where T L (k,i) n,m is the free R-submodule spanned by C (k,i ) v,v 2 P (n, k ) Suppose that (D D 2 x) C v (k,i),v 2 spanned by C v (k,i),v 2 with v, v 2 P (n, k) Then T L n,m (k,i), where T L n,m (k,i) (D D 2 x) C (k,i) v,v 2 = D v 2 x C i, with (k, i ) (k, i) and v, v 2 is the free R-submodule for some D in P (n, k) and some x G m,n 2k, here x does not depend on v 2 Write x = for some 0 k j m, j n 2k By an easy calculation, we know that n 2k j= tk j j x C i, n 2k j= u k j i j C i, (mod G <i m,n 2k ), where G <i m,n 2k is the free R-submodule spanned by Cj, with j < i Note also that the coefficient is independent of v 2 This implies that 4 (C3) is true n 2k j= uk j i j As a corollary of Theorem 53, we classify the irreducible representations of cycolotomic Temperley-Lieb algebras Corollary 54 Suppose R is a splitting field of x m Let p be the characteristic of R Then : (i) suppose n is odd If m = p t s with (p, s) = and t 0, then the set {L(k, i) 0 k [n/2], i = (i, i 2,, i n 2k ) Λ(m, n 2k) with all i j divisible by p t } is a complete set of pairwise non-isomorphic simple T L n,m -modules (ii) Suppose n is even ) If not all δ i are zero and if m = p t s with (p, s) = and t 0, then the set {L(k, i) 0 k [n/2], i = (i, i 2,, i n 2k ) Λ(m, n 2k) with all i j divisible by p t } is a complete set of pairwise non-isomorphic simple T L n,m -modules 2) Suppose all δ i are zero If m = p t s with (p, s) = and t 0, then a complete set of pairwise non-isomorphic simple T L n,m -modules can be parametrized by {(k, i) 0 k < [n/2], i = (i, i 2,, i n 2k ) with all i j divisible by p t }} 0

11 Proof For any D, D 2 P (n, k) and i Λ(m, n 2k), we have (D D 2 C i,)(d D 2 C i,) = D D 2 xc i,c i,, x G m,n 2k If this product is not equal to zero, then C i,c i, 0 Now suppose that n is odd If D = E E 3 E 2k and D 2 = E 2 E 4 E 2k, then x = id Hence statement (i) follows from 44 Assume that n is even First case: there is some δ j 0 and p does not divide m Then for k = n/2 and i =, the bilinear form Φ (k,i) 0 For the other (k, i), we take D and D 2 as above This implies Φ (k,i) 0 Hence the index set of non-isomorphic simple modules is Λ m,n Second case: there is some δ j 0 and p divides m By arguments similar to the above, we have that a complete set of non-isomorphic simple modules is {(k, i) 0 k [n/2], i = (i, i 2,, i n 2k ) with all i j divisible by p t } Assume δ j = 0, 0 j m In this case, Φ (k,i) = 0 for k = n/2 and i = For k n/2, our discussion will be the same as above, namely, if m = p t s, then index set of simple modules is {(k, i) 0 k < [n/2], i = (i, i 2,, i n 2k ) with all i j divisible by p t }; if p does not divide m, then the index set of simple modules is Λ m,n \ {(n/2, )} The following result follows from the proof of Theorem 53 Corollary 55 Let (k, i) be the cell module corresponding to (k, i) Λ n,m Then dim R (k, i) = m k [ ( ) ( n k n k ) ] 6 Quasi-heredity of T L n,m In this section, we shall characterize the parameters for which the cyclotomic Temperley-Lieb algebras are quasi-hereditary in the sense of [5] First, we recall the definition of quasi-hereditary algebras Definition 6 (Cline, Parshall and Scott [5]) Let R be a field and let A be an R-algebra An ideal J in A is called a heredity ideal if J is idempotent, J(rad(A))J = 0 and J is a projective left (or, right) A-module, where rad(a) is the Jacobson radical of A The algebra A is called quasi-hereditary provided there is a finite chain 0 = J 0 J J 2 J n = A of ideals in A such that J j /J j is a heredity ideal in A/J j for all j Such a chain is then called a heredity chain of the quasi-hereditary algebra A From the ring theoretic definition of cellular algebras, we see immediately that there is a large intersection of the class of cellular algebras with that of quasi-hereditary algebras Typical examples of quasi-hereditary cellular algebras include Temperley-Lieb algebras with non-zero parameters [7] and Birman-Wenzl algebras for most choices of parameters [8] as well as certain cyclotomic Brauer algebras [4] The main result in this section is the following theorem Theorem 62 Suppose R is a splitting field of the polynomial x m Then the cyclotomic Temperley-Lieb algebra T L n,m is quasi-hereditary if and only if the characteristic of R does not divide m and one of the following is true: () n is odd; (2) n is even and δ j 0 for some 0 j m Proof In [7, Remark 30] it is shown that A is quasi-hereditary if the index set of the nonisomorphic simple modules over a cellular algebra A with cell datum (I, M, C, i) is I Conversely, A is not quasi-hereditary if there is a cell datum (I, M, C, i) of A such that the index set of the non-isomorphic simple modules is not I [0, Theorem 3] In other words, every chain of ideals in A is not a heredity chain Thus Theorem 62 follows immediately from Corollary 54 For the cases which are not included in Theorem 62, we can get a quasi-hereditary quotient of T L n,m In order to make T L n,m quasi-hereditary, we need first to ensure that the group algebra G m,n is semisimple The following result follows from the above fact and the definition 4

12 Proposition 63 Suppose that R is a splitting field of x m and p m, 2 n and δ j = 0 for all 0 j m Suppose J is the two-sided ideal of T L n,m generated by all (n, n/2)-labelled TL-diagrams Then the quotient T L n,m /J is quasi-hereditary 7 Restriction and Induction of the cell modules In this section, we assume that R is a splitting field of x m The main result of this section is the branching rule for the cell modules of T L n,m Recall that V (n, k) is the R-space spanned by all labelled parenthesized graphs with k arcs Let J i := [n/2] j=i V (n, j) R V (n, j) R G m,n 2j Then we have a chain 0 J [n/2] J i+ J i J ɛ = T L n,m of ideals in T L m,n, where ɛ is zero if n is even, and if n is odd For any (k, i) Λ n,m, the cell module (k, i) = V (n, k) R v 0 R (i), where v 0 P (n, k) is a fixed diagram and (i) is the cell module of the algebra G m,n with respect to i In the sequel, we choose v 0 to be the (n, k)-labelled parenthesized graph with arcs {, 2},, {2k, 2k} and free vertices 2k +, 2k + 2,, n Note that the subquotient V (n, j) R V (n, j) R G m,n 2j is a T L n,m -module and the cell module structure on V (n, k) R v 0 R (i) is induced from this subquotient We make the following convention: Throughout this section we fix an m and the parameters δ 0, δ,, δ m and consider the algebra T L n,m canonically as a subalgebra of T L n,m by adding the vertical arc {n, n } to the right side of each labelled TL-diagram in T L n,m This embedding can be visualized as follows: n T L n,m T L n,m Note that the identity in T L n,m is sent to the identity of T L n,m Thus every T L n,m -module is also a T L n,m -module via this embedding The cell modules V (n, k) R v 0 R (i) over T L n,m will be denoted by (n, k; i, i 2,, i n 2k ) Then we have Proposition 7 (a) For all n and 0 k [n/2], there is an exact sequence α 0 (n, k; i, i 2,, i n 2k ) (n, k; i, i 2,, i n 2k ) β m j=0 V (n, k ) R v 0 R (i, i 2,, i n 2k )t j n 2k+ 0, where M is the restriction of a T L n,m -module M to a T L n,m -module, and (i, i 2,, i n 2k )t j n 2k+ stands for (i, i 2,, i n 2k ) R Rt j n 2k+ (b) If I 0 = 0 I I m = R t n 2k+ is a cell chain of the group algebra R t n 2k+ = G m,, that is, I j is the free R-module generated by { m l>s (t n 2k+ u l ) s j}, then there are m short exact sequences 0 V (n, k ) R v 0 R (i, i 2, i n 2k ) I j γ V (n, k ) R v 0 R (i, i 2, i n 2k ) I j δ (n, k ; i, i 2,, i n 2k, j) 0 (c) If T L n,m is semisimple, then n (n, k; i, i 2,, i n 2k ) = (n, k; i, i 2,, i n 2k ) m (n, k ; i, i 2,, i n 2k, j) j= 2

13 Proof If T L n,m is semisimple, then every T L n,m -module is projective Therefore, each short exact sequence in (a) and (b) splits Now the statement (c) follows immediately from (a) and (b) The map γ in (b) is the canonical injective map and the map δ in (b) comes from the canonical projection I i I i /I i One can easily prove that (b) is a short exact sequence of vector spaces Obviously, both γ and δ in (b) are T L n,m -module homomorphisms Now let us prove the statement (a) Since we may consider (n, k; i, i 2,, i n 2k ) as a subset of T L n,m, the map α is just the restriction of the above embedding It is obvious that α is an injective map Note that T L n,m is generated as an algebra by {e i, t j i n, j n} To show that α is a T L n,m - module homomorphism, it suffices to prove that for D {e i, t j i n 2, j n }, α(dv v 0 C (i,i2,,i n 2k) ) = Dα(v v 0 C (i,i2,,i n 2k) ) However, a vertex ( n) in v is free if and only if it is free in v, and {i, j} is an arc in v if and only if it is an arc in v, where v is the (n, k )-labelled parenthesis obtained from v by adding the vertex n By the multiplication of labelled TL-diagrams in 32, we can see immediately that the above equation holds Hence α is a T L n,m -module homomorphism Now let us define the map β Given an (n, k)-labelled parenthesis v P (n, k), we denote by v the labelled parenthesis obtained from v by deleting the vertex n and removing the arc connected with n if it exists Let v be in P (n, k) If the vertex n in v is a free vertex, then β sends v v 0 C (i,i2,,i n 2k), to zero If the vertex n in v is connected by an arc in which there are l dots, then β sends v v 0 C (i,i2,,i n 2k), to v v 0 C (i,i2,,i n 2k), t l n 2k+, where v 0 is the (n, k )-labelled parenthesis with arcs {, 2}, {3, 4},, {2k 3, 2k 2} and n 2k + free vertices In fact, we can extend β to a map from V (n, k) v 0 G m,n 2k to V (n, k ) v 0 G m,n 2k+ This map β can be illustrated as follows: j n j β (l) (l) 2k n 2k 2 n (The image of an (n, k)-labelled TL-diagram under the map β is obtained from the given (n, k)-labelled TL-diagram by deleting both the arc {2k, 2k} and its endpoints from the bottom row, and then shifting the vertex n from the top row to the bottom row, and finally renaming the vertices at the bottom from left to right) It is trivial that the sequence is an exact sequence of vector spaces To finish the proof, it remains to show that β is also a T L n,m -module homomorphism Since β restricted to the image of α preserves the module structure, we need only to prove that β preserves the T L n,m -module structure on the elements of the form a(v v 0 C (i,i 2,,i n 2k ), ), where a {e i, t j i n 2, j n } and the vertex n in v is not free In the following, we show more generally that the extended map β is a T L n,m -module homomorphism Let v v 0 x with v P (n, k) such that n is connected to j by an arc in v Suppose a = t s or a = e r with r {j, j} In this case, by inspecting the picture, it is easy to see that β preserves the module structure on the element a(v v 0 x) Now suppose a = e s with s = j or s = j In the latter case, since there are no free vertices between j and n in v, the labelled TL-diagram β(e j (v v 0 x)) is just the graph e j β(v v 0 x) This is what we wanted to prove In the former case, if j is a free vertex in v, then e j (v v 0 x) lies in the image of α, which is mapped to zero under β Moreover, the element e j β(v v 0 x) is also zero since it contains one more arc Now assume that j is adjacent to a vertex s in v Then s < j In this case, there are no free vertices between s and n in v Again by inspecting picture we see that β(e j (v v 0 x)) = e j β(v v 0 x) This completes the proof The following result follows from Proposition 7 and Frobenius reciprocity 3

14 Proposition 72 If T L n,m is semisimple, then (n, k; i,, i n 2k ) (n, k + ; i, i 2,, i n 2k 2 ) m j= (n, k; i,, i n 2k, j), where (n, k; i,, i n 2k ) stands for the T L n,m -module induced from the T L n,m -module (n, k; i,, i n 2k ) 8 Gram matrices and their determinants In this section, we assume that the field R contains a primitive m-th root of unity (for example, if R is algebraically closed and of characteristic p which does not divide m, then our assumptions are satisfied) The goal in this section is to calculate the discriminant of the bilinear form Φ (k,i) for certain (k, i), where 0 k [ n 2 ] and i = (i, i 2,, i n 2k ) Recall that Φ (k,i) is defined on the cell module (k, i) in the following way (v v C i,)(v 2 v 2 C i,) Φ (k,i) (v, v 2 )v v 2 C i, (mod T L <(k,i) n,m ), where T L <(k,i) n,m is a free R-submodule spanned by C (k,i ) v,v 2 with (k, i ) < (k, i) and v, v 2 P (n, k ) According to a general construction in [], there is a bilinear form φ (n,k) from V (n, k) R V (n, k) to G m,n 2k such that the product can be written as (v v C i,)(v 2 v 2 C i,) v v 2 φ (n,k) v,v 2 (t, t 2,, t n 2k )(C i,) 2, (mod T L <(k,i) n,m ), where φ (n,k) v,v 2 (t, t 2,, t n 2k ) is an element in G m,n 2k Define a(k, i) = n 2k m j= l>i j (u ij u l ) It follows that Φ (k,i) (v, v 2 ) = a(k, i)φ (n,k) v,v 2 (u i, u i2,, u in 2k ), where u, u 2,, u m are the roots of x m Now let us compute the matrix Ψ(n, k) := (φ (n,k) v,v 2 ) for the case k = Let v i be the element in P (n, ) whose unique arc is {i, i + } and let v (j) i be the (n, )-labelled parenthesis in which there are j dots on the arc {i, i + } The elements in P (n, ) can be ordered as follows: v (0) := v, v (),, v(m ), v (0) 2 := v 2, v () 2,, v(m ) 2,, v (0) n := v n, v () n,, v(m ) n Thus: A B B T A B 2 B2 T A B 3 Ψ(n, ) =, A B n 2 A B T n 2 where B i is the matrix with the (s, t)-entry t s t i for s, t m, the matrix Bi T stands for the transpose of B i, and δ 0 δ δ m δ δ 2 δ 0 A = δ m δ 0 δ m 2 Let us look at a simple example We consider the algebra T L (4,3) (δ 0, δ, δ 2 ), that is, n = 4 4

15 and m = 3 In this case we have t 3 i = for i =, 2 Thus δ 0 δ δ 2 t 2 t δ δ 2 δ 0 t t δ 2 δ 0 δ t 2 t t t 2 δ 0 δ δ 2 t 2 2 t 2 Ψ(n, ) = t 2 t δ δ 2 δ 0 t 2 t 2 2 t t 2 δ 2 δ 0 δ t 2 2 t t 2 t 2 2 δ 0 δ δ t 2 2 t 2 δ δ 2 δ t 2 t 2 2 δ 2 δ 0 δ Suppose u is a primitive m-th root of unity Define u j = u j for j = 2,, m and u m = u 0 = Then u k = u m k Let V = V m (, u,, u m ) be the Vandermonde matrix of order m: V m (, u,, u m ) = u u 2 u m u 2 u 2 2 u 2 m u m u m 2 u m m Since we shall evaluate each t j as some u ij, when we calculate the value of Φ (k,i) (v, v 2 ), we may suppose that t j = u ij for all j n 2 Thus the matrix B j is of the form B j = u ij u m i j (, u i j,, u (m ) i j ) = u ij u m i j (, u m i j,, u m m i j ) Now we define Y j to be the matrix of order m with at the (i j, m i j )-position and 0 otherwise For i j = m, we define Y j to be the matrix with (m, m)-entry and all other entries 0 Then B j = V Y j V T Let p(x) = δ 0 x m + δ x m δ m R[x] We write p(x) x m = δ + δ δ m x u x u 2 x u m Since u is a primitive m-th root of unity and u i u j for i j, we have δ j = p(u j )/ i j (u j u i ) for all j =, 2,, m Now we can rewrite δ k = m δ j= j u k j Note that the index k can be an arbitrary natural numbers and that δ l = δ k if l k (mod m) Thus the matrix A can be written as (δ k+l ) 0 k,l m Furthermore, we have A = V ĀV T, where Ā = diag( δ, δ 2,, δ m ) is the diagonal matrix Since x m = (x u )(x u 2 ) (x u m ), we know that the k-th elementary symmetric polynomial in u, u 2,, u m is zero for k m Hence Newton s identities imply that σ k (u, u 2,, u m ) := m u k j = j= { m, if k = m, 0, if k m Thus we have V V T = m m 0 0 m 0 0 m 0 0 5

16 This implies that (det(v )) 2 = ( ) (m )(m 2)/2 m m Thus det(ψ(n, )) = ( ) 2 (m )(m 2)(n ) m m(n ) det(ψ(n, )), where Ψ(n, ) = Ā Y Ā Y 2 Y T Y T 2 Ā Y 3 Ā Y n 2 Ā Now let us calculate the determinant of Ψ(n, ) Since each matrix Y j is of special form, we partition (i, i 2,, i n 2 ) into (i,, i,2,, i,j, i 2,, i 2,2,, i 2,j2,, i r,jr ) with j + j j r = n 2 such that m divides i p,q + i p,q+ for all p with q < j p and that m does not divide i p,jp + i p+, for all p < r Thus Y T n 2 det(ψ(n, )) = ( δ δ2 δ m ) n r p= ( δ m ip,jp jp q= δ ip,q ) r det p= δ ip, δip,2 δip,3 δip,jp δip,jp Let P (x,, x n ) = det x x 2 x 3 xn x n Then det(ψ(n, )) = ( ) 2 m(m )(n ) m m(n ) ( δ δ2 δ m ) n r r p= ( δ jp δ P ( δ ip,, δ ip,2,, δ ip,j p m ip,jp q= ip,q ) ) We have proved the following proposition Proposition 8 Let R be a field containing a primitive m-th root of unity Then the determinant of the Gram matrix of the bilinear form Φ (,i) is detφ (,i) = ( ) 2 m(m )(n ) a(, i) m(n ) m m(n ) ( δ δ2 δ m ) n r r p= ( δ jp δ P ( δ ip,, δ ip,2,, δ ip,jp ) m ip,jp q= ip,q ) p= p= As a consequence of 8, we know that under the above assumption, a necessary condition for T L n,m to be semisimple is that all the polynomials P ( δ ip,, δ ip,2,, δ ip,j ) are non zero p The following is a description of the polynomial P (x,, x n ) Let I(n) := {n, n 2, n 4, } {0} N and define Γ(n, r) := {(i, i 2,, i r ) i < i 2 < < i r n, i r n (mod 2); i j+ i j + (mod 2) for all j r } for all r I(n) If α = (i, i 2,, i r ) Γ(n, r) we write x α for x i x i2 x ir Then P (x, x 2,, x n ) = ( ) (n r)/2 x α r I(n) α Γ(n,r) 6

17 This can be proved by induction on n and the recursive formula P (x, x 2,, x n ) = x n P (x, x 2,, x n ) P (x, x 2,, x n 2 ) In fact, the set Γ(n, r) is a disjoint union of {(i, i 2,, i r, n) i < i 2 < < i r n, i r n (mod 2); i j+ i j + (mod 2) for all j r 2} and Γ(n 2, r) Thus this decomposition of Γ(n, r) corresponds just to the two summands in the recursive formula of P (x, x 2,, x n ) Note that if m = or if x = x 2 = = x n, then both detφ (,i) and P (x, x,, x) are Tchebychev-type polynomials which play an important role in the study of Temperley-Lieb algebras (see [7] and [7]) Hence we call P (x, x 2,, x n ) the n-th generalized Tchebychev polynomial It follows from the recursive formula that P (x, x 2,, x n ) is irreducible in the polynomial ring R[x, x 2,, x n ] with n variables x, x 2,, x n Acknowledgements The authors are grateful to Yongjian Hu at BNU for useful discussions on matrices; to Dieter Vossieck at BNU for useful conversations which let to a vast improvement of the original statement in 34, and to BKeller for his help with English language The research work of CCXi was supported by TCTPF of the Education Ministry of China while H Rui is supported by the Foundation for University Key Teachers by the Ministry of Education of China References [] SAriki and KKoike, A Hecke algebra of (Z/rZ) S n and the construction of its irreducible representations Adv Math 06, (994) [2] MFAtiyah and IGMacdonald, Introduction to commutative algebra Addison-Wesley Publishing Company, 969 [3] JBirman and HWenzl, Braids, link polynomials and a new algebra Trans Amer Math Soc 33, (989) [4] MBroué and GMalle, Zyklotomische Heckealgebren Astérisque 22, 9-89 (993) [5] ECline, BParshall and LScott, Finite dimensional algebras and highest weight categories J reine angew Math 39, (988) [6] RHäring-Oldenburg, Cyclotomic Birman-Murakami-Wenzl algebras J Pure Appl Alg 6, 3-44 (200) [7] JGraham and GLehrer, Cellular algebras Invent Math 34, -34 (996) [8] VFRJones, Index for subfactors Invent Math 72, -25 (983) [9] SKönig and CCXi, On the structure of cellular algebras In: IReiten, SSmalø and ØSolberg (Eds): Algebras and Modules II Canadian Mathematical Society Conference Proceedings 24, (998) [0] SKönig and CCXi, When is a cellular algebra quasi-hereditary? Math Ann 35, (999) [] SKönig and CCXi, Cellular algebras: inflations and Morita equivalences Journal of the London Math Soc (2) 60, (999) [2] PMartin, Potts models and related problems in statistical mechanics World scientific, Singapore, 99 [3] PMartin and HSaleur, The blob algebra and the periodic Temperley-Lieb algebra Lett Math Phys 30, No 3, (994) [4] HRui, Representations of cyclotomic Brauer algebras Preprint, 2000 [5] HNVTemperley and EHLieb,Relations between percolation and colouring problems and other graph theoretical problems associated with regular planar lattices: some exact results for the percolation problem Proc Roy Soc London (Ser A) 322, (97) [6] T tom Dieck, Knotentheorie und Wurzelsysteme II Math Göttingensis 44 (993) [7] BWWestbury, The representation theory of the Temperley-Lieb algebra Math Z 29, no4, (995) [8] CCXi, On the quasi-heredity of Birman-Wenzl algebras Adv Math 54, (2000) 7