Knot theory related to generalized and. cyclotomic Hecke algebras of type B. Soa Lambropoulou. Mathematisches Institut, Gottingen Universitat

Transcription

1 Knot theory related to generalized and cyclotomic Hecke algebras of type B Soa Lambropoulou Mathematisches Institut, Gottingen Universitat Introduction After Jones's construction of the classical by now Jones polynomial for knots in S 3 using Ocneanu's Markov trace on the associated Hecke algebras of type A, arised questions about similar constructions on other Hecke algebras as well as in other 3-manifolds. In [2] is established that knot isotopy in a 3-manifold may be interpreted in terms of Markov braid equivalence and, also, that the braids related to the 3-manifold form algebraic structures. Moreover, that the sets of braids related to the solid torus or to the lens spaces L(p; ) form groups, which are in fact the Artin braid groups of type B. As a consequence, in [2, 3] appeared the rst construction of a Jones-type invariant using Hecke algebras of type B following Jones approach, and this had a natural interpretation as an isotopy invariant for oriented knots in a solid torus. In a further `horizontal' development we constructed in [9], using a dierent technique, all such solid torus knot invariants derived from the Hecke algebras of type B. In this paper we consider all possible generalizations of the B-type Hecke algebras and we construct Markov traces on each of them, so as to obtain all possible dierent levels of homy-pt analogues in the solid torus related to the B-type Hecke algebras. Our strategy is based on the one in [3], which in turn followed []. So, in this sense, the construction in [2, 3] is incorporated as the most basic level. In more detail: It is well-understood from Jones's construction of the homy-pt (2-variable Jones) polynomial, P L, in [], that H n (q), the Iwahori- Hecke algebra of A n -type, is a quotient of the braid group algebra Z[q ]B n by factoring out the quadratic relations 2 i (q? ) i + q

2 2 S. Lambropoulou and that this relations reect precisely the skein property of P L : p q p P L+? p q p P L? ( p q? p q ) P L0 ; where L + is a regular projection of an oriented link containing a specied positive crossing, L? the same projection with a negative crossing instead, and L 0 the same projection with no crossing. We do now analogous considerations for the solid torus, which we denote by ST. Let us consider the following Dynkin diagram. (300,50) t 2 n (B n ) t t t p p p t n [h] The symbols t; ; : : : ; n labelling the nodes correspond to the generators of the Artin braid group of type B n, which we denote by B ;n. B ;n is dened therefore by the relations t t t t t i i t if i > i j j i if ji? jj > i i+ i i+ i i+ if i n? 2 Relations of these types will be called braid relations. B ;n may be seen as the subgroup of B n+, the classical braid group on n+ strands, the elements of which keep the rst strand xed (this is the reason for having chosen the symbol B ;n ). This allows for a geometric interpretation of the elements of B ;n as mixed braids in S 3. Below we illustrate the generators i ; t and the element t 0 i i : : : t : : : i in B ;n, which plays a crucial role in this work. i i+ n......, 2 n i+ n..., σ i Note that the inverses of i ; t are represented by the same geometric pictures, but with the opposite crossings. As shown in [2, 3], we can represent oriented knots and links inside ST by elements of the groups B ;n, where the xed strand represents the t t' i

3 Knot theory and B-type Hecke algebras 3 complementary solid torus in S 3, and the next n numbered strands represent the knot in ST. Also, that knot isotopy in ST can be translated in terms of equivalence classes in S n B ;n (Markov theorem), the equivalence being generated by the following two moves. (i) Conjugation: if ; 2 B ;n then. (ii) Markov moves: if 2 B ;n then n 2 B ;n+. Consider now the classical Iwahori-Hecke algebra of type B n, H n (q; Q), as a quotient of the group algebra Z[q ; Q ]B ;n by factoring out the ideal generated by the relations t 2 (Q? )t + Q and g 2 i (q? )g i + q for all i, where we denote the image of i in H n (q; Q) by g i. The idea in [2, 3, 9] was to construct invariants of knots in the solid torus by constructing trace functions on S n H n (q; Q) which support the Markov property: (hg n ) z(h); for z a xed parameter in Z[q ; Q ] and h 2 H n (q; Q). In other words, traces that respect the above braid equivalence on S n B ;n. The construction of such traces was only possible because we were able to nd an appropriate inductive basis on H n+ (q; Q), every element of which involves the generator g n or the element t 0 : g n n : : : g tg : : : g n at most once (see picture above for the lifting of t 0 i in B ;n ). In particular, the trace constructed in [2, 3] was well-dened inductively by the rules: ) tr(ab) tr(ba) a; b 2 H n (q; Q) 2) tr() for all H n (q; Q) 3) tr(ag n ) z tr(a) a 2 H n (q; Q) 4) tr(at 0 n) s tr(a) a 2 H n (q; Q) If we had not used the elements t 0 n in the above constructions we would have not been able to dene the trace with only four simple rules. The intrinsic reason for this is that B ;n splits as a semi-direct product of the classical braid group B n and of its free subgroup P ;n generated precisely by the elements t; t 0 ; : : : ; t0 n : B ;n P ;n B n : All Jones-type invariants in ST constructed from the above traces on S n H n(q; Q) satisfy the skein rule related to the quadratic relations g 2 i (q )g i +q plus another one reecting the quadratic relation t 2 (Q)t+Q (cf. [2, 3, 9] for an extensive treatment).

4 4 S. Lambropoulou During the work of S.L. and J. Przytycki in [5] on the problem of computing the 3rd skein module of the lens spaces L(p; ) following the above strategy, it turned out that the skein rule of the ST -invariants in [2, 3, 9] related to t was actually 'articial', so far that homy-pt type knot invariants were concerned, and that for analogous constructions in L(p; ) it was needed to have constructed rst the most generic 2-variable Jones analogue in ST, one that would not satisfy any skein relation involving t. We drop then the quadratic relation of t, and we consider the quotient of the group algebra Z[q ]B ;n by factoring out only the relations g 2 i (q? )g i + q for all i. This is now a new innite dimensional algebra, which we denote by H n (q; ) and we shall call it generalized Iwahori-Hecke algebra of type B. By g i above we denote the image of i in H n (q; ), whilst the symbol was chosen to indicate that the generator t satises no relation (since now any power t k ; for k 2 Z may appear, like in B ;n ). These algebras have been consequently studied in [6] in connection with the ane Hecke algebras of type A. But we would like to go one step back and, instead of removing from the dening relations of H n (q; Q) the quadratic relation for t, to require that t satises a relation given by a cyclotomic polynomial of degree d: (t? u )(t? u 2 ) (t? u d ) 0 Then we would obtain a nite-dimensional algebra known as cyclotomic Hecke algebra of type B, denoted here by H n (q; d). The corresponding cyclotomic Coxeter group of type B, which we denote by W n;d, is obtained as a quotient of B ;n over the relations g 2 i and t d. H n (q; d) may be seen as a `d-deformation' of W n;d : In order to obtain the group algebra we have to substitute the parameters of the cyclotomic polynomial by the dth routs of unity (and not by as in the classical case). These algebras and their corresponding groups have been introduced and studied independently by two groups of mathematicians in [, 2, 5, 4]. It follows from the discussion above that the cyclotomic Hecke algebras are also related to the knot theory of the solid torus and, in fact, they make the bridge between H n (q; Q) and H n (q; ). Like for the Hecke algebras of type B, in order to construct linear Markov traces on S H n n(q; ) or on S H n n(q; d), we need to nd appropriate inductive bases on both types of these algebras. This is the aim and the main result of Section 3. Note that, in the classical case of H n (q; Q), we could easily yield such an inductive basis using the results in [7]. This presumes at least

5 Knot theory and B-type Hecke algebras 5 knowledge of some basis for the algebra. In Section 2, we study the structure of H n (q; ) and we construct a basis for it using the structure of the braid group B ;n and the known bases for H n (q; d). In Section 4 we construct Markov traces on S n H n(q; ) and on S n H n(q; d) using the inductive bases of Section 3. Finally in Section 5, we normalize the traces according to the Markov braid theorem in order to derive the corresponding knot invariants in ST, and we also give skein interpretations. The invariant related to H n (q; ) is the most interesting one for us, and in this sense, this work may be seen as the required fundament for extending such constructions to knots in the lens spaces. The knot invariant derived from H n (q; ) reproves the structure of the 3rd skein module of the solid torus, found before in [0, 7]. On the other hand, the knot invariants derived from H n (q; d) are related to submodules of the the 3rd skein module of ST. It may be worth noting that introducing and studying H n (q; ) has been independent of the studies on the cyclotomic analogues. It gives the author pleasure to acknowledge her thanks to V.F.R. Jones for his comments on this work and to T. tom Dieck for many discussions and his valuable suggestions. Many thanks are also due to. M. Geck for discussions and for pointing out the literature on the cyclotomic Hecke algebras of type B, and surely to J. Przytycki for our discussions and conjectures about the structure of the generalized Coxeter groups and Hecke algebras. Finally, nancial support by the SFB 70 in Gottingen and the European Union for parts of this work are gratefully acknowledged. 2 Finding a basis for H n (q; ) We start by introducing in more detail H n (q; ), H n (q; d) and their corresponding Coxeter-type groups W n;, W n;d. Denition The generalized Iwahori-Hecke algebra of type B n is dened as H n (q; ) : Z[q ] B ;n < i 2 (q? ) i + q for all i > : The underlying generalized Coxeter group of type B n is dened as W n; : B ;n < i 2 for all i > : It follows that if g i denotes the image of i in H n (q; ), then H n (q; ) is dened by the generators t; g ; g 2 ; : : : ; g n and their relations:

6 6 S. Lambropoulou tg tg g tg t tg i g i t for i > g i g i+ g i g i+ g i g i+ for i n? 2 g i g j g j g i for ji? jj > 2 g i (q? ) g i + q for all i H n (q; ) is an associative algebra with. Also, it is easily veried that, if S n is the symmetric group, then W n; Z n S n ( compare with the structure ofb ;n ): Denition 2 Let R : Z[q ; u ; : : : ; u d ; : : :], where q; u ; : : : ; u d ; : : : are indeterminates. The cyclotomic Iwahori-Hecke algebra of type B n and of degree d is dened as H n (q; d) : R B ;n < i 2 (q) i +q all i; (t?u )(t?u 2 ) (t?u d ) 0 > : The underlying cyclotomic Coxeter group of type B and of degree d is: W n;d : B ;n < i 2 for all i; t d ; d 2 IN > : The relation t d is derived by the cyclotomic polynomial by substituting the u i 's by the d'th roots of unity. Also, the Coxeter group of B n -type, in our notation W n;2, is the quotient of B ;n over the relations t 2 i2, for all i. H n (q; d) is an associative algebra with, and it is a free module over R of rank d n n!, which is precisely the order of W n;d (cf. [2],[5]). If d and u, then H n (q; ) is isomorphic to the Iwahori-Hecke algebra of type A (over Z[q ]). If d 2, u and u 2?Q, we recover the familiar relation of H n (q; Q), the Iwahori-Hecke algebra of type B (over Z[q ; Q ]). In H n (q; d) we have t d a d t d + + a 0 ; where a d?(u + +u d ); a d?2?(u u 2 + +u d u d ); : : : ; a 0 () d (u : : : u d ); from this we can derive easily a relation for t. W n;d may also be seen as the quotient W n; < t d ; d 2 IN > of W n;, and it is easily veried that W n;d Z d n S n Its order is 2 n n!; whilst W n;2 Z 2n S n (compare with the structure of B ;n ).

7 Knot theory and B-type Hecke algebras 7 Note W.l.o.g. we extend the ground ring of H n (q; ) to R. Then H n (q; d) may also be obtained from H n (q; ) by factoring out the cyclotomic relation. In this sense H n (q; d) is a `bridge' between H n (q; ) and H n (q; Q), the classical Hecke algebra. We shall now nd a basis for H n (q; ) as follows: We nd rst a canonical form for the braid group B ;n, which yields a basis for Z[q ]B ;n. The images of these basic elements in H n (q; ) through the canonical map spann H n (q; ). We then treat the spanning set in order to obtain a basis for H n (q; ). The advantage of this approach { apart from starting from the better-understood stucture of the braids { is that it shows clearly the relation among the structures of B ;n, H n (q; ), H n (q; d) and W n;, W n;. In order to proceed we need to recall the notion of the pure braid group and Artin's canonical form for pure braids: The classical pure braid group, P n, consists of all elements in B n that induce the identity permutation in S n ; P n B n and P n is generated by the elements A rs r r+ : : : s?2 s2 s?2 : : : r+ r s s?2 : : : r+ r2 r+ : : : s?2 s ; r < s n: Artin's canonical form says that `every element, A, of P n uniquely in the form: can be written A U U 2 U n where each U i is a uniquely determined product of powers of the A ij using only those with i < j'. Geometrically, this means that any pure braid can be `combed' i.e. can be written canonically as: the pure braiding of the rst string with the rest, then keep the rst string xed and uncrossed and have the pure braiding of the second string and so on (cf. [3] for a complete treatment). We nd now a canonical form for B ;n. An element w of B ;n induces a permutation 2 S n of the n numbered strands. We add at the bottom of the braid a standard braid in B n corresponding to, and then we add its inverse. Now, w is a pure braid on n + stands (including the rst xed one), and we apply to it Artin's canonical form. This separates the braiding of the xed strand from the rest:

8 8 S. Lambropoulou n... w... - p p... The above sketch in fact the proof that B ;n P ;n B n. >From the uniqueness of Artin's canonical form, it follows that any w 2 B ;n can be expressed uniquely as a product v (`vector-permutation'), where v is an element of the free group P ;n : v t 0 i k t 0 i 2 k 2 : : : t 0 ir kr ; k ; : : : ; k r 2 Z; where t 0 ik : i : : : t k : : : i ; and 2 B n is written in the induced by P n canonical form. Thus the set fvg forms a basis for the algebra Z[q ] B ;n, and, therefore, it spans the quotient H n (q; ). On the level of H n (q; ) we can already improve this spanning set, since on this level is a word in H n (q), the Iwahori-Hecke algebra of A n -type. So, can be written in terms of the canonical basis of H n (q) (cf. []): f(g i g i : : : g i?r )(g i2 g i2 : : : g i2?r 2 ) : : : (g i pg i p : : : g i p?rp)g; Therefore we showed for i < : : : < i p n? : Proposition The set ft 0 j k t 0 j 2 k 2 : : : t 0 jr kr g; where t 0 0 : t; t 0 k i : g i : : : g t k g : : : g i ; j ; : : : ; j r 2 f0; ; : : : ; ng, k ; : : : ; k r 2 Z and a basic element of H n (q), spans H n (q; ): Notice that the indices of the `vector' part are not ordered. Also, that the above canonical form for B ;n yields immediately the following canonical form fv g for W n; : fv g ft j k t j2 k 2 : : : t j r kr g; where t 0 : t; t ik : s i : : : s t k s : : : s i ; for 0 j < : : : < j r n?, k ; : : : ; k r 2 Z and 2 S n is an element of the canonical form of S n (where s i denotes the image of i in W n; ). Thus, this set also forms a basis for the group algebra Z[q ] W n;.

9 Knot theory and B-type Hecke algebras 9 Notice here that the indices of the `vector' part are ordered. This suggests that it may be possible to order the indices j ; : : : ; j r of the words t 0 k j t 0 k 2 j 2 : : : t 0 kr jr in, so as to be left with a canonical basis for H n (q; ). To achieve this straight from is very dicult, because it is hard to get hold of an induction step, even though there are relations among the t 0 i ik 's. Instead, we change the t 0 ik 's to the elements t ik, where t 0 : t; and t i : g i : : : g tg : : : g i. These elements commute in H n (q; ). The following relations hold in H n (q; ) and in H n (q; d) and will be used repeatedly in the sequel. Lemma For 2 fg the following hold: (i) g i q g i? + (q? ); g i 2 (q? ) g i + q ; for q 6 0: (ii) g i (g k g k : : : g j ) (g k g k : : : g j )g i+ ; for k > i j; g i (g j g j+ : : : g k ) (g j g j+ : : : g k )g i ; for k i > j: (iii) g i g i : : : g j+ g j g j+ : : : g i g j g j+ : : : g i g i g i : : : g j+ g j ; g i g i : : : g j+g j g j+ : : : g i g j g j+ : : : g i g i g i : : : g j+g j : (iv) g i : : : g n g n g n g n : : : g i (q? ) P n?i r0 q r (g i : : : g n?r g n?r g n?r : : : g i ) + q (n?i+) P n?i+ r0 (q? ) r q r (g i : : : g n?r g n?r g n?r : : : g i ); where r if r n? i and n?i+ 0: Similarly, g i : : : g 2 g g g 2 : : : g i (q? ) P i r0 qr (g i : : : g r+2 g r+ g r+2 : : : g i ) + q i P i r0 (q? ) r q r (g i : : : g r+2 g r+ g r+2 : : : g i ), where r if r i? and i 0. (v) t g tg g tg t for 2 Z corr. Z d ; g i t k t k g i for k > i; k < i? ; g i t i q t i g i + (q? ) t i ; g i t i q t i g i + (q? ) t i ; g i t i q t i g i + (q? ) t i ; g i t i q t i g i + (q? ) t i.

10 0 S. Lambropoulou (vi) g i t 0 k t 0 k g i for k > i; k < i? ; g i t 0 i t 0 i g i + (q? ) t 0 i + (? q) t 0 i ; g i t 0 i t 0 i g i : (vii) t ik t j t j t i k for i 6 j and k; 2 Z corr. Z d : (viii) t 0 ik g i : : : g t k g : : : g i for k; 2 Z corr. Z d : Therefore we have in H n (q; d): (t 0 i? u )(t 0 i? u 2 ) : : : (t 0 i? u d ) 0, t 0d i a d t 0 d i + + a 0 ; where the a i 's are given in the relation () in Section 2. Proof. We point out rst that in the rest of the paper and in order to facilitate the reader we underline in the proofs the expressions which are crucial for the next step. We also use the symbol `P' instead of the phrase `linear combination of words of the type'. Except for (iv), all relations are easy consequenses of the dening relations of H n (q; ). Relation (vii) can be also checked using braid diagrams. We prove (iv) by induction on the length l n? i + of the word g n g n : : : g i. For l we have g n2 (q? )g n + q. Assume now (iv) holds up to l n? i. Then for l n? i + we have g i g i+ : : : g n g n : : : g i+ g i induction step g i [(q? ) P n?(i+) r0 q r (g i+ : : : g n?r g n?r g n?r : : : g i+ ) + q n?i ] g i (q? ) P n?(i+) r0 q r (g i : : : g n?r g n?r g n?r : : : g i ) + q n?i g i2 (q? ) P n?(i+) r0 q r (g i : : : g n?r g n?r g n?r : : : g i ) + (q? )q n?i g i + q n?i+ (q? ) P n?i r0 q r (g i : : : g n?r g n?r g n?r : : : g i ) + q n?i+ : Furthermore note that in the Relations (v) and (vi) a t i or a t 0 i will not change to a t i or a t 0 i respectively and, therefore, these relations preserve the total sum of the exponents of the t i 's and the t 0 i 's in a word. Note also that for j i? the relations (iii) boil down to the usual braid relation and its variations with inverses. 2 Theorem In H n (q; ) the set 2 ft i k t i2 k 2 : : : t i r kr g for i < : : : < i r n?, k ; : : : ; k r 2 Z and a basic element in H n (q), forms a basis for H n (q; ):

11 Knot theory and B-type Hecke algebras Notice that in 2 the indices of the `vector' part are ordered. Proof. To show that 2 spans H n (q; ) it suces, by Proposition, to show that an element of can be written as a linear combination of elements in 2. Indeed, let We do the proof by induction on w t 0 j k t 0 j 2 k 2 : : : t 0 jm km 2 : jk j + jk 2 j + + jk m j; the absolute number of t's in w. For either w t 0 i or w t 0 i : t 0 i g i : : : g tg : : : g i g i : : : g t(g : : : g i g i : : : g )g : : : g i t i ; where 2 H n (q), a linear combination of basic elements of H n (q). t 0 i g i : : : g t g : : : g i g i : : : g (g : : : g i g i : : : g )t g : : : g i Lemma;(iv) (q) P i r0 q r (g i : : : g r+2 g r+ g r+2 : : : g i )g i : : : g r+ g r : : : g t g : : : g i + q i t i (q? ) P i r0 q r (g i : : : g r+2 )g r : : : g t g : : : g i + q i t i (q? ) P i r0 q r t r (g i : : : g r+2 g r+ : : : g i ) + q i t i (q? ) P i r0 q r t r i + q i t i ; where i 2 H n (q): Suppose now the assumption holds for up to? t's in w. Then, the induction step holds in particular for all such words with. So, for jk j + jk 2 j + + jk m j we have: t 0 j k : : : t 0 jm km 8 < : 8 < : 8 >< >: t i : : : t i n n t 0 jm ; t n : : : t n 2 t 0 ; jm t 0 j k : : : t 0 jm km t 0 jm ; if k m > 0 t 0 k j : : : t 0 km+ t 0 ; if k jm jm m < 0 by induction for 2 H n (q); i < : : : < i n n? ; j j + + j n j? and for 2 2 H n (q); < : : : < n n? ; j j + + j n j?

12 2 S. Lambropoulou 8 < : 8 < : t i : : : t i n n t j m(g j m : : : g g : : : g j m ) t : : : t n n 2 (g j m : : : g g : : : g j m)t j m t i : : : t i n n t e i 0 ; i 0 2 H n (q); t : : : t n n t e2 0 i; 0 i 2 H n (q); Lemma;(vii) Lemma;(iv);(v) where 0 (g j m : : : g g : : : g j m ) and (g j m : : : g g : : : g j m); 8 < : t i : : : t i r r t e t ir+ r+ : : : t i n n i 0 ; for i r < e < i r+ : t : : : t k k t e2 t k+ : : : t n n 0 i; for k < e 2 < k+ : I.e. in either case we obtained a linear combination of elements of 2 : We next show linear independency of the elements of 2 : Let P m i iw i 0 for w ; w 2 ; : : : ; w m 2 2. We assume rst that the exponents of the t j 's in the words w i are all positive for all i, and we choose d > k 2 IN, where k is the maximum of the exponents of the t j 's in P m i i w i. Then, the canonical epimorphism of H n (q; ) onto H n (q; d) applied on the equation P m i i w i 0 in H n (q; ) yields the equation P m i i w i 0 in H n (q; d). As shown in [2], Proposition 3.4 and Theorem 3.0, the elements of 2 with 0 < k ; : : : ; k r d? form a basis for H n (q; d); d 2 IN. (In [2] d is denoted by r, H n (q; d) is denoted by H n;r and is denoted by a w.) This implies that i 0; i ; : : : ; m in H n (q; d), and therefore, i 0; i ; : : : ; m also in H n (q; ). Assume nally that some w i 's contain t j 's with negative exponents. The idea is to resolve the negative exponents and then refer to the previous case. One way is to proceed as above, and after we have projected P m i i w i 0 on H n (q; d), to resolve the t j 's with negative exponents using the algebra relations; nally, to conclude that i 0; i ; : : : ; m in H n (q; d) (and therefore in H n (q; )), using induction and arguments from linear algebra. But we would rather give a more elegant argument, that was suggested by T. tom Dieck. Namely, let P be the product of all t k j ; k 2 IN for all j; k such that t?k j is in some w i. Since P is an invertible element of H n (q; ), we have P m i i w i 0, P P m i iw i 0: The last equation is eqivalent to P m i ip w i 0, where the elements P w i are pairwise dierent and the exponents of the t j 's contained in each P w i are positive for all i. We then refer to the previous case, and the proof of Theorem is now concluded. 2 Thus 2 is a basis of H n (q; ), and therefore H n (q; ) is a free module.

13 Knot theory and B-type Hecke algebras 3 3 Inductive bases for H n (q; ) and H n (q; d) The basis of H n (q; ) constructed in the previous section as well as the corresponding one for H n (q; d) yields an inductive basis for H n (q; ) and H n (q; d), which gives rise to two others, the last one being the appropriate for constructing Markov traces on these algebras. Here we give these three inductive bases and we conclude this section by giving another basic set for H n (q; ) and H n (q; d), which is analogous to the set 2, but using t 0 i's instead of t i 's. >From now on we shall denote by H n both H n (q; ) and H n (q; d) and by W n both W n; and W n;d. Also, whenever we refer to k 2 Z corr. k 2 Z d we shall always assume k 6 0. We now nd the rst inductive basis for H n+. This on the group level is an inductive canonical form, and it provides a set of right coset representatives of W n into W n+, which is completely analogous to [7], p. 456 for B-type Coxeter groups. Lemma 2 For k 2 Z (corr. k 2 Z d ) the following hold in H n+ (q; ) (corr. H n+ (q; d)): (i) t nk g n (q? ) P k j0 q j t nj t n k?j + q k g n t nk ; if k 2 IN and t nk g n (? q) P k j0 q j t nj t n k?j + q k g n t nk ; if k 2 Z? IN: (ii) t nk g n g n : : : g i (q? ) P k j0 q j (t nj g n g n?2 : : : g i )t n k?j + (q? )q k P k j0 q j (t n?2j g n?2 g n?3 : : : g i )g n t n k?j + (q? )q 2k P k j0 q j (t n?3j g n?3 : : : g i )g n g n t n?2 k?j + + (q? )q (n?i)k P k j0 q j (t ij )g n g n : : : g i+ t i k?j +q (n?i+)k g n g n : : : g i t ik ; if k 2 IN; whilst for k 2 Z? IN we have an analogous formula, only (q? ) is replaced by (? q); q k q?jkj and jk? jj + jjj jkj: Proof. We prove (i) for the case k > 0 by induction on k. (For k < 0 completely analogous.) For k we have t n g n (q? )t n + q g n t n : Suppose the assumption holds for k?. Then for k we have: t nk g n t n t n k g n by induction t n [(q? ) P k?2 j0 q j k?j t nj t n + q k k Lemma ;(vii) g n t n ]

14 4 S. Lambropoulou (q? ) P k?2 j0 qj k?j t nj t n + q k k Lemma ;(v) t n g n t n (q? ) P k?2 j0 qj k?j t nj t n + q k k (q? ) t n t n + q k k Lemma ;(vii) g n t n (q? ) P k j0 q j t nj t n k?j + q k g n t nk : We prove (ii) for the case k > 0 by decreasing induction on i. (For k < 0 completely analogous.) For i n we have (i). Assume it holds for i + < n (, i n? 2, n? i 2). Then for i we have: t nk g n : : : g i+ g i by induction [(q? ) P k j0 q j (t nj g n g n?2 : : : g i+ )t n k?j ]g i + + [(q? )q (n?(i+))k P k j0 q j (t ij )g n g n : : : g i+2 t i+ k?j ]g i + [q (n?i)k g n g n : : : g i+ t ik ]g i Lemma;(v)&Lemma2;(i) (q? ) P k j0 q j (t nj g n : : : g i+ g i )t n k?j + + (q? )q (n?(i+))k P k j0 q j (t ij g i )g n g n : : : g i+2 t i+ k?j + q (n?i)k (q? ) P k j0 q j g n g n : : : g i+ t ij t i k?j + q (n?i)k q k g n g n : : : g i t ik (q? ) P k j0 qj (t nj g n : : : g i+ g i )t n k?j + + (q? )q (n?(i+))k P k j0 qj (t ij g i )g n g n : : : g i+2 t i+ k?j + q (n?i)k (q? ) P k j0 qj (t ij )g n g n : : : g i+ t i k?j + q (n?i)k q k g n g n : : : g i t ik : 2 Theorem 2 Every element of H n+ (q; ) corr. H n (q; d) is a unique linear combination of words, each of one of the following types: ) w n 2) w n g n g n : : : g i 3) w n g n g n : : : g i t ik ; k 2 Z corr. k 2 Z d 4) w n t nk ; k 2 Z corr. k 2 Z d where w n is some word in H n (q; ) corr. H n (q; d). Thus, the above words furnish an inductive basis for H n+ (q; ) corr. H n (q; d).

15 Knot theory and B-type Hecke algebras 5 Proof. By Theorem it suces to show that every element v n 2 2, where v is a product of t i 's and n 2 H n+ (q), can be expressed uniquely in terms of ), 2), 3) and 4). We prove this by induction on n: For n 0 there are no g i 's in the word, so v 0 t k, a word of type ). Suppose the assertion holds for all basic words in 2 with indices up to n?, and let w 2 2 such that w contains elements of index n. We examine the dierent cases: w t i k t i2 k 2 : : : t i r kr t nk n ; i < : : : < i r < n and n 2 H n (q). Then, by Lemma,(v), w t i k : : : t i r kr n t nk w n t nk ; a word of type 4). w t i k t i2 k 2 : : : t i r kr n ; where i r < n and n n (g n g n : : : g i ) 2 H n+ (q). Then w t i k t i2 k 2 : : : t i r kr n (g n g n : : : g i ) w n g n g n : : : g i ; a word of type 2). Finally, let w t i k t i2 k 2 : : : t i r kr t nk n ; where n n (g n g n : : : g i ) 2 k Lemma ;(v) H n+ (q). Then w t kr i : : : t i r t nk n g n g n : : : g i k Lemma 2;(ii) t kr i : : : t i r n t nk g n g n : : : g i w n t n k?j + w n g n g n : : : g s t s k?j ; for j 0; : : : ; k? : I.e. w is a sum of words of type 4) and type 3). The uniqueness of these expressions follows from Lemma and Lemma 2. 2 Theorem 2 rephrased weaker says that the elements of the inductive basis contain either g n or t nk at most once. But, as explained in the beginning, our aim is to nd an inductive basis for H n+ using the elements t 0 i g i g i : : : g tg : : : g i g i ; as these are the right ones for constructing Markov traces on S H n n: We go from the t i 's to the t 0 i 's via the `intermediate' elements T k i : g i g i : : : g t k g : : : g i g i ; k 2 Z: Theorem 3 Every element of H n+ (q; ) corr. H n (q; d) is a unique linear combination of words, each of one of the following types: 0 ) w n 2 0 ) w n g n g n : : : g i 3 0 ) w n g n g n : : : g i T k i; k 2 Z corr. k 2 Z d 4 0 ) w n T k n; k 2 Z corr. k 2 Z d

16 6 S. Lambropoulou where w n is some word in H n (q; ) corr. H n (q; d). Proof. It suces to show that elements of the inductive basis given in Theorem 2 can be expressed uniquely as sums of the above words. For this we need the following three lemmas. Lemma 3 For k 2 IN (corr. k 2 Z d ) and 2 fg the following hold in H n+ (q; ) (corr. H n+ (q; d)): t n (k+) P n r ;:::;r k 0 (q? ) r ++r k q (r ++r k) g n g n : : : g t (g : : : g n?r : : : g )t : : : t (g : : : g n?rk : : : g )t g : : : g n g n ; where ri if r i 0; : : : ; n? ; n 0 and g 0 : : Proof. We show the case + by induction on k. The proof for is completely analogous. For k we have: t n2 g n g n : : : g tg : : : g n g n g n g n : : : g tg : : : g n g n Lemma ;(iv) P n r0 (q? ) r q r g n g n : : : g t(g : : : g n?r : : : g )tg : : : g n g n : Assume that the statement holds for any k 2 IN. Then for k + we have: k+ by induction P t n t nk t n n r ;:::;r k0 (q? ) r ++r k q r ++r k gn : : : g t(g : : : g n?r : : : g )t : : : t(g : : : g n?rk : : : g )tg : : : g n (g n : : : g tg : : : g n ) Lemma ;(iv) P n (q? r ;:::;r k 0 )r ++r k q r ++r k g n : : : g t(g : : : g n?r : : : g )t : : : t(g : : : g n?rk : : : g )tg : : : g n : 2 Lemma 4 For k 2 IN (corr. k 2 Z d ) and 2 fg the following hold in H n (q; ) (corr. H n (q; d)): (i) t g t k g g t k g t + (q? )t g t k + (? q )t k g t and (ii) t? g t k g g t k g t? + (q? )t (k) g + (? q )g t (k) : Proof. We only prove (i) for the case +, by induction on k. All other statements are proved similarly. For k we have tg tg g tg t. Assume the assertion is correct for k. Then for k + we have: tg t k+ g tg t k g g Lemma ;(i) tg q tg t k g g tg + (q? ) tg t k g tg induction step

17 Knot theory and B-type Hecke algebras 7 q g t k g tg tg + q (q? ) tg t k g tg + q (? q) t k g tg tg + (q? ) g t k g t 2 g + (q? )(q? ) tg t k+ g + (q? )(? q) t k g t 2 g rels:; induction step q g t k+ g tg 2 + (? q ) g t k g t 2 g + (? q )(q? ) tg t k+ g + (? q )(? q) t k g t 2 g + (q? ) t k+ g tg 2 + (q? ) g t k g t 2 g + (q? )(q? ) tg t k+ g + (q? )(? q) t k g t 2 g Lemma ;(i) q (q? ) g t k+ g tg + g t k+ g t + (q? )(q? ) t k+ g tg + (q? )q t k+ g t Lemma ;(v) (?q ) g 2 tg t k+ +g t k+ g t+(q )(q ) t k+ g tg +(?q) t k+ g t (? q )(q? ) g tg t k+ + (? q )q tg t k+ + g t k+ g t+ (q? )(q? ) t k+ g tg + (? q) t k+ g t Lemma ;(v) g t k+ g t + (q? ) tg t k+ + (? q) t k+ g t: 2 Lemma 5 (Fundamental Lemma (F.L.)) For i; k 2 IN (corr. i; k 2 Z d ) and for 2 fg the following hold in H n (q; ) (corr. H n (q; d)): (i) t i g t k g g t k g t i + (q? ) [t g t (k+i) + t 2 g t (k+i?2) + + t i g t k ]+ (? q ) [t k g t i + t (k+) g t (i) + + t (k+i) g t ] and (ii) t?i g t k g g t k g t?i + (q? ) [t (k) g t?(i) + t (k?2) g t?(i?2) + + t (k?i) g ]+ (? q ) [t?(i) g t (k) + t?(i?2) g t (k?2) + + g t (k?i) ]: Proof. We prove (i) for the case +, by induction on i. The proof for is completely analogous. For i the assertion is true by Lemma 4,(i). Assume it holds for i. Then for i + we have: t i+ g t k g tt i g t k g induction step tg t k g t i + (q? ) [t 2 g t k+i + t 3 g t k+i?2 + + t i+ g t k ]+ (? q) [t k+ g t i + t k+2 g t i + + t k+i g t] Lemma 4;(i)

18 8 S. Lambropoulou g t k g t i+ + (q? ) tg t k+i + (? q) t k g t i+ + (q? ) [t 2 g t k+i + t 3 g t k+i?2 + + t i+ g t k ]+ (? q) [t k+ g t i + t k+2 g t i + + t k+i g t]: We go back now to the proof of Theorem 3. By Lemma 3, a typical summand of t n (k+) 2 H n+ (q; ) (corr. H n+ (q; d)) is: g n : : : g t (g : : : g n?l : : : g )t 2 : : : t N(g : : : g n?ln : : : g )t N + g : : : g n ; where ; 2 ; : : : ; N + 2 IN such that + + N + k + and l i < n for i ; : : : ; N (since the cases l i n are incorporated in t i ). In order to prove the theorem we want to show that such a word can be expressed in terms of words of the form 0 ), 2 0 ), 3 0 ) and 4 0 ). This is a very slow process as we shall readily see. In order to obtain an inductive argument on the number N + of the intermediate powers of t, we show rst the following, seemingly more general result, where an unsymmetric expression appears also in the word. It is 'seemingly more general' because this unsymmetry of the word appears anyhow in a later stage of the calculations. Proposition 2 Let k 2 IN (corr. k 2 Z d ), 2 fg; l; m; l 2 ; : : : ; l N n and let ; 2 ; : : : ; N + 2 IN such that + + N + k +. Then it holds in H n+ (q; ) (corr. H n+ (q; d)) that words of the form: w n g n : : : g t (g : : : g l )(g : : : g m : : : g )t 2 (g : : : g l2 : : : g )t 3 : : : t N + g : : : g n where only between the rst two powers of t appears the unsymmetric expression (g : : : g l )(g : : : g m : : : g ); can be expressed as sums of words of the form 0 ), 2 0 ), 3 0 ) and 4 0 ). Note that, if l 0, we obtain the generic summand of t n (k+). Proof. We prove the statement for + by induction on the number N + of intermediate powers of t. The proof for is completely analogous. For N 0 we have w n g n : : : g t g : : : g n ; where k + i.e. w n T k+ n : Suppose the assertion holds for N. Then for N + we have: A w n g n : : : g t (g : : : g l )(g : : : g m : : : g )t (g : : : g l2 : : : g )t 3 : : : t N + g : : : g n w n g n : : : g t (g : : : g l )(g m : : : g : : : g m )t (g l2 : : : g : : : g l2 )t 3 : : : t N + g : : : g n : Here we also use the symbol `P' to mean `linear combination of words of the type', the symbol `w n ' for not always the same word in H n, and, in order to shorten the words, we substitute the expression g l2 : : : g : : : g l2 t 3 : : : t N + g : : : g n by S. 2

19 Knot theory and B-type Hecke algebras 9 We proceed by examining the cases l < m; l > m and l m: For l < m we have: A w n g n : : : g t g m : : : g 2 g : : : g m (g : : : g )g n : : : g t g : : : g m g g 2 : : : g l t S m> w n (g m : : : g g )g n : : : g t g : : : g m g 2 : : : g l t S w n (g m : : : g)g 2 n : : : g t g t g 2 : : : g m g 2 : : : g l S F:L: w n g n : : : g 2 t g t g g 2 : : : g m g 2 : : : g l S+ P i+j+ w n g n : : : g 2 t i g t j g 2 : : : g m g 2 : : : g l S w n t g n : : : g t g : : : g m g 2 : : : g l S+ Pi+j+ w n t i g n : : : g g 2 : : : g m g 2 : : : g l t j S Lemma ;(ii); l<m (w n t g 2 : : : g l )g n : : : g t g : : : g m S+ P i+j+ (w nt i g : : : g m g : : : g l )g n : : : g t j S (w n g n : : : g t g : : : g m S + P i+j+ w n g n : : : g t j S and the number of intermediate powers of t has reduced to N in all summands of t n k+. For l > m we have: A w n g n : : : g t (g : : : g l )g m : : : g : : : g m t S (w n g m : : : g : : : g m )g n : : : g t g : : : g l t S m<l; Lemma ;(ii) w n g n : : : g t g t g 2 : : : g l S F:L: w n g n : : : g 2 t g t g : : : g l S+ P w i+j+ ng n : : : g 2 t i g t j g 2 : : : g l S w n t g n : : : g t g : : : g l S+ P i+j+ w n t i g n : : : g (g 2 : : : g l )t j S Lemma ;(ii) w n g n : : : g t g : : : g l S + P i+j+ (w n t i g : : : g l )g n : : : g t j S w n g n : : : g t g : : : g l + P i+j+ w n g n : : : g t j and the number of intermediate powers of t has reduced to N in all summands of t n k+. Finally if l m we have: A w n g n : : : g t (g : : : g m )g m : : : g : : : g m t S Lemma; (iv) w n g n : : : g t g 2 : : : g m t S+

20 20 S. Lambropoulou P m w r0 ng n : : : g t (g m?r : : : g 2 g : : : g m?r )g 2 : : : g m t Lemma; (ii) S (w n g : : : g m )g n : : : g t + S+ P m r0 (w n g m?r : : : g )g n : : : g t g t (g 2 : : : g m?r )g 2 : : : g m S F:L: w n g n : : : g t + S + P m r0 w n g n : : : g 2 t g t (g : : : g m?r )g 2 : : : g m S+ P P m i+j+ r0 w n g n : : : g 2 t i g t j (g 2 : : : g m?r )g 2 : : : g m S w n g n : : : g t + S+ P m r0 w n t g n : : : g t (g : : : g m?r )g 2 : : : g m?r : : : g m S + P P m i+j+ r0 w n t i g n : : : g (g 2 : : : g m?r )g 2 : : : g m t j Lemma; (ii) S w n g n : : : g t + S+ P m r0 (w n t g 2 : : : g m?r )g n : : : g t (g : : : g m?r gm?rg 2 m?r+ : : : g m ) S+ P i+j+ P m r0 (w nt i g : : : g m?r g : : : g m )g n : : : g t j S w n g n : : : g t + S + P m w r0 ng n : : : g t (g : : : g m?r g m?r+ : : : g m ) S+ P m r0 w n g n : : : g t j S P m r0 w n g n : : : g t (g : : : g m ) S + P i+j+ w n g n : : : g t + S + P m r0 w n g m?r+ : : : g m g n : : : g t (g : : : g m?r ) S+ P m w r0 ng n : : : g t (g : : : g m ) S + P P m w i+j+ r0 ng n : : : g t j S w n g n : : : g t + S + P m r0 w n g n : : : g t (g : : : g m?r ) S+ P m r0 w n g n : : : g t (g : : : g m ) S + P i+j+ P m r0 w n g n : : : g t j S and the number of the intermediate powers of t has reduced to N in all summands of t n k+. We can now conclude the proof of Theorem 3, since for the dierent possibilities of a word w 2 H n+ we have: Case. Case 2. If w w n or w w n g n : : : g i for i 0; : : : ; n there is nothing to show. If w w n t k n; k 2 Z corr. Z d, then by Proposition 2, w is a unique linear combination of words of type 0 ), 2 0 ), 3 0 ) and 4 0 ). 2 Case 3. Finally, if w w n g n : : : g i+ t k; k 2 Z corr. Z i d, by Proposition 2, t k is i written in terms of words w i ; w i g i : : : g r for r i; w i g i : : : g r+ T k r and w i T k i : Therefore w can be written uniquely in terms of the words

21 Knot theory and B-type Hecke algebras 2 w n g n : : : g i+ w i g i : : : g r for r 0; : : : ; i, w n g n : : : g i+ w i g i : : : g r+ T k r and w n g n : : : g i+ w i T k i : w i commutes with g n : : : g i+, unless i 0, where the word is already arranged in a trivial manner. So the above words reduce to the types w n g n : : : g r or w n g n : : : g j+ T k j : Theorem 3 rephrased weaker says that the elements of the inductive basis contain either g n or T k n at most once. We can now pass easily to the inductive basis that we need for constructing Markov traces on S n H n. Indeed we have the following: Theorem 4 Every element of H n+ (q; ) corr. H n+ (q; d) can be written uniquely as a linear combination of words, each of one of the following types: 00 ) w n 2 00 ) w n g n g n : : : g i 3 00 ) w n g n g n : : : g i+ t 0 ik ; k 2 Z corr. Z d 4 00 ) w n t 0 nk ; k 2 Z corr. Z d where w n is some word in H n (q; ) corr. H n (q; d). Proof. By Theorem 3 it suces to show that expressions of the forms 3 0 ) and 4 0 ) can be written (uniquely) in terms of 00 ), 2 00 ), 3 00 ) and 4 00 ). Indeed, for k 2 Z corr. Z d, let w w n g n g n : : : g i+ T k i w n g n g n : : : g i+ g i : : : g t k g : : : g i : We apply the relation g r q g r + (q? ) to all letters of the word g : : : g i to get: w w n g n : : : g i+ g i : : : g t k g : : : g i + P w n g n : : : g t k g j : : : g j k ; where in the words g j : : : g j k there are possible gaps of indices. Let the closest to t k gap occur at the index ; then w w n g n : : : g i+ t 0k i + P w n g n : : : g t k g : : : g g + : : : g j k w n g n : : : g i+ t 0 k i + P (w n g : : : g j k )g n : : : g t k g : : : g 2

22 22 S. Lambropoulou w n g n : : : g i+ t 0k i + P w n g n : : : g t 0 k : Hence w is a sum of words of type In the case where w w n T k n ; k 2 Z corr. Z d, we apply the same reasoning as above. Theorem 4 rephrased weaker says that the elements of the inductive basis contain either g n or t 0 nk at most once. Notice also that if we were working on the level of the Iwahori-Hecke algebra H n (q; Q), we would omit Theorem 3. Remark All three inductive bases of H n+ (q; ) corr. H n+ (q; d) given in Theorems 2, 3 and 4 induce the same complete set of right coset representatives, S n+, of W n; corr. W n;d in W n+; corr. W n+;d, namely: S n+ : fs n s n : : : s i j i ; : : : ; ng S fs n s n : : : s t k s : : : s i j i ; : : : ; n? ; k 2 Z corr. k 2 Z d ; k 6 0g S ft nk j k 2 Z corr. k 2 Z d g: 2 We now give the nal result of this section, namely, a basic set of H n+ which is a proper subset of. Theorem 5 The set ft 0 i k t 0 i 2 k 2 : : : t 0 ir kr g for i < i 2 < : : : < i r n; k ; : : : ; k r 2 Z corr. Z d and 2 H n+ (q) forms a basis in H n+ (q; ) corr. H n+ (q; d). Proof. By Theorem 4 it suces to show that words in the inductive basis 00 ), 2 00 ), 3 00 ) and 4 00 ) can be written in terms of elements of. Indeed, by induction on n we have: if n 0 the only non-empty words are powers of t, which are of type 4 00 ) and which are elements of trivially. Assume the result holds for n?. Then for n we have: Case. If w w n there is nothing to show (by induction). Case 2. If w w n g n : : : g i, then, by induction w n t 0 i k : : : t 0 ir kr, a word of restricted on H n : Thus w t 0 i k : : : t 0 ir kr g n : : : g i 2 ; since g n : : : g i is an element of the canonical basis of H n+ (q). Case 3. If w w n g n : : : g i+ t 0 k i, then, by induction step w n t 0 k i : : : t 0 kr, ir a word of restricted on H n ; so w t 0 k i : : : t 0 kr g ir n : : : g i+ t 0 Lemma; (vi) ik

23 Knot theory and B-type Hecke algebras 23 t 0 k i : : : t 0 kr t 0 nk Lemma; (vi) g ir n : : : g i+ t 0 k i : : : t 0 kr t 0 k ir n g n : : : g i. Now g n : : : g i is a basic element of H n+ (q), thus w 2. Case 4. Finally, if w w n t 0 k n ; by induction step we have w n t 0 k i : : : t 0 kr, ir a word of restricted on H n. Then w t 0 k i : : : t 0 kr t 0 Lemma; (vi) ir nk t 0 k i : : : t 0 kr t 0 k ir n 2 : 2 4 Construction of Markov traces The aim of this section is to construct Markov linear traces on the generalized and on each level of the cyclotomic Iwahori-Hecke algebras of B-type. As these algebras factor through the braid groups, the constructed traces will actually attach to each braid a Laurent polynomial. The traces as well as the strategy of their construction are based on and include as special case the one constructed on the classical B-type Hecke algebras in [2], [3] (Theorem 5), which in turn was based on Ocneanu's trace on Hecke algebras of A-type, cf. [] (Theorem 5.). In the next section we combine these results with the Markov braid equivalence for knots in a solid torus, so as to obtain analogues of the homy-pt polynomial for the solid torus. Let R Z[q ; u ; : : : ; u d ; : : :] and let H n denote either H n (q; ) or H n (q; d). Note that the natural inclusion of the group B ;n into B ;n+ (geometrically, by adding one more strand at the end of the braid) induces a natural inclusion of H n into H n+. Therefore it makes sense to consider B : S n B ;n and H : S n H n. Then we have the following result: Theorem 6 Given z; s k, specied elements in R with k 2 Z corr. Z d and k 6 0, there exists a unique linear trace function tr : H : determined by the rules: [ n H n?! R(z; s k ) k2 Zcorr. Z d ) tr(ab) tr(ba) a; b 2 H n 2) tr() for all H n 3) tr(ag n ) z tr(a) a 2 H n 4) tr(at 0 nk ) s k tr(a) a 2 H n ; k 2 Z corr. Z d

24 24 S. Lambropoulou Proof. The idea of the proof of Theorem 6 is to construct tr on S n H n inductively using Theorem 4 and the two last rules of the statement above. For this we need the following: Lemma 6 The map c n : (H n NH n H n ) L k2 Zcorr. Z d H n?! H n+ given by c n (a b k e k ) : ag n b + P k2 Zcorr. Z d e k t 0 k n is an isomorphism of (H n ; H n )-bimodules. Proof. It follows from Theorem 4 that the set L n below provides a basis of H n as a free H n -module (compare with Remark for W n+ ): L n : fg n g n?2 : : : g i j i ; : : : ; n? g S ft 0 k n j k 2 Z corr. k 2 Z d g S fg n g n?2 : : : g t k g : : : g i j i ; : : : ; n? 2; k 2 Z corr. k 2 Z d ; k 6 0g: Then we have: H n L b2l n H n b; and using the universal property of tensor product we obtain: H n N H n H n H n NH n ( L b2l n H n b) L b2l n (H n NH n H n b) L b2l n H n b: Therefore: H n N H n H n Lk2 Zcorr. Z d H n L b2l n H n b L k2 Zcorr. Z d H n : Applying now the same reasoning as above, the set L n+ below provides a basis of H n+ as a free H n -module: L n+ : fg n g n : : : g i j i ; : : : ; ng S ft 0 k n j k 2 Z corr. k 2 Z d g S fg n g n : : : g t k g : : : g i j i ; : : : ; n? ; k 2 Z corr. k 2 Z d ; k 6 0g: The latter isomorphism then proves that c n is indeed an isomorphism of (H n ; H n )-bimodules, since it corresponds bijectively basic elements to elements of the set L n+. We can now dene inductively a trace, tr, on H S n H n as follows: assume tr is dened on H n and let x 2 H n+ be an arbitrary element. By Lemma 6 there exist a; b; e k 2 H n ; k 2 Z corr. Z d, such that x : c n (b e k a k ): 2

25 Knot theory and B-type Hecke algebras 25 Dene now: tr(x) : z tr(ab) + tr(e 0 ) + X k2 Zcorr. Z d s k tr(e k ): Then tr is well-dened. Furthermore, it satises the rules 2), 3) and 4) of the statement of Theorem 6. Rule 3) reects the Markov property (recall the discussion in Introduction), and therefore, if the function tr is a trace then it is in particular a Markov trace. In fact one can check easily using induction and linearity, that tr satises the following seemingly stronger condition: (3 0 ) tr(ag n b) z tr(ab); for any a; b 2 H n : In order to prove the existence of tr, it remain to prove the conjugation property, i.e. that tr is indeed a trace. We show this by examining case by case the dierent possibilities. Before continuing with the proof, we note that having proved the existence, the uniqueness of tr follows immediately, since for any x 2 H n+ ; tr(x) can be clearly computed inductively using rules ), 2), 3), 4) and linearity. We now proceed with checking that tr(ax) tr(xa) for all a; x 2 H. Since tr is dened inductively the assumption holds for all a; x 2 H n, and we'll show that tr(ax) tr(xa) for a; x 2 H n+. For this it suces to consider a 2 H n+ arbitrary and x one of the generators of H n+. I.e. it suces to show: tr(ag i ) tr(g i a) a 2 H n+ ; i ; : : : ; n tr(at) tr(ta) a 2 H n+ : By Theorem 4, a is of one of the following types: i) a w n ii) a w n g n g n : : : g i iii) a w n g n g n : : : g i+ t 0 ik ; k 2 Z corr. Z d iv) a w n t 0 nk ; k 2 Z corr. Z d, where w n is some word in H n. If a w n and x t or x g i for i ; : : : ; n? the assumption holds from the induction step, whilst for x g n it follows from (3 0 ) that tr(w n g n ) z tr(a) tr(g n w n ).

26 26 S. Lambropoulou If a is of type ii) or of type iii) and x t or x g i for i ; : : : ; n we apply the same reasoning as above using rule (3 0 ). So we have to check still the cases where a w n g n g n : : : g i or a w n g n g n : : : g i+ t 0 ik and x g n, i.e. tr(w n g n : : : g i g n ) tr(g n w n g n : : : g i ) tr(w n g n : : : g i+ t 0 k i g n ) tr(g n w n g n : : : g i+ t 0k i ) () If a is of type iv) and x t or x g i for i ; : : : ; n? we have to check: tr(w n t 0 k n t) tr(tw n t 0 k n ) tr(w n t 0 k n g i ) tr(g i w n t 0 k n ) () Finally, if a is of type iv) and x g n we have to check: tr(w n t 0 k n g n ) tr(g n w n t 0 k n ) ( ) Before checking (); () and ( ) we need the following: Lemma 7 The function tr satises the following stronger version of rule 4): for any x; y 2 H n ; k 2 Z corr. Z d : (4 0 ) tr(xt 0 nk y) s k tr(xy); Proof. It suces to prove (4 0 ) for the case that y is of the form y y t y 2, where y is a product of the g i 's for i ; : : : ; n? ; 2 Z corr. Z d and y 2 an arbitrary word in H n. Indeed we have: tr(xt 0 k n y) tr(xt 0 nk y t y 2 ) Lemma;(vi) tr(xy t 0 nk t y 2 ) tr(xy g n : : : g t k g g 2 : : : g n t y 2 ) Lemma;(vi) tr(xy g n : : : g t k g t g 2 : : : g n ) A The latter underlined expression says that we have to consider four possibilities depending on k; being positive or negative. We show here the case where both k; are positive. The rest are proved completely analogously. For k; positive, Lemma 5,(i) says: g t k g t t g t k g + (q? ) [t g t k+ + + g t k+ ] We substitute then in A to obtain: + (? q ) [t k g t + + t k+ g t]:

27 Knot theory and B-type Hecke algebras 27 A tr(xy g n : : : g 2 t g t k g : : : g n y 2 ) +(q? ) [tr(xy g n : : : g 2 t g t k+ g 2 : : : g n y 2 ) + +tr(xy g n : : : g t k+ g 2 : : : g n y 2 )] +(? q ) [tr(xy g n : : : g 2 t k g t g 2 : : : g n y 2 ) + +tr(xy g n : : : g 2 t k+ g tg 2 : : : g n y 2 )] tr(xy t t 0 nk y 2 ) Lemma ;(vi) +(q? ) [tr(xy t g n : : : g g 2 : : : g n t k+ y 2 ) + +tr(xy g n : : : g g 2 : : : g n t k+ y 2 )] +(? q ) [tr(xy t k g n : : : g g 2 : : : g n t y 2 ) + +tr(xy t k+ g n : : : g g 2 : : : g n ty 2 )] tr(xy t t 0 nk y 2 ) Lemma ;(iii) +(q? ) [tr(xy t g : : : g n g n : : : g t k+ y 2 ) + +tr(xy g : : : g n g n : : : g t k+ y 2 )] +(? q ) [tr(xy t k g : : : g n g n : : : g t y 2 ) + +tr(xy t k+ g : : : g n g n : : : g ty 2 )] (30 ) tr(xy t t 0 nk y 2 ) + (q? )z [tr(xy t +k y 2 ) + (? q )z [tr(xy t k+ y 2 ) tr(xy t t 0 nk y 2 ): 2 The relations () follow now immediately from Lemma 7, since: tr(w n t 0 k n g i ) (40 ) induction step s k tr(w n g i ) s k tr(g i w n ) tr(g i w n t 0 nk ); for all i < n, and similarly for x t. We next show () for a w n g n : : : g i. The case a w n g n : : : g i+ t 0 ik is shown similarly. On the one hand we have: tr(w n g n g n : : : g i g n ) tr(w n g n g n g n g n?2 : : : g i ) tr(w n g n g n g n g n?2 : : : g i ) (30 ) z tr(w n g n2 g n?2 : : : g i ) (q? )z tr(w n g n : : : g i ) + qz tr(w n g n?2 : : : g i ):

28 28 S. Lambropoulou On the other hand in order to calculate tr(g n w n g n : : : g i ) we examine the dierent possibilities for w n : { If w n 2 H n ; then tr(g n w n g n : : : g i ) tr(w n g n2 g n : : : g i ) (q? )z tr(w n g n : : : g i ) + qz tr(w n g n?2 : : : g i ): { If w n bg n c, where b; c 2 H n, then tr(g n bg n cg n g n : : : g i ) tr(bg n g n g n cg n : : : g i ) (30 ) z tr(bg n2 cg n : : : g i ) (q? )z tr(bg n cg n : : : g i ) + qz tr(bcg n : : : g i ) (q? )z tr(bg n cg n : : : g i ) + qz 2 tr(bcg n?2 : : : g i ) (q? )z tr(bg n cg n : : : g i ) + qz tr(bg n cg n?2 : : : g i ) (q? )z tr(w n g n : : : g i ) + qz tr(w n g n?2 : : : g i ): { Finally, if w n bt 0 nk, where b; 2 H n, then tr(g n bt 0 k n g n : : : g i ) tr(bg n t 0 k n g n : : : g i ) q tr(bt 0 k n g n : : : g i ) + (q? ) tr(bg n t 0 nk g n : : : g i ) (40 );(3 0 ) qz tr(bt 0 nk g n?2 : : : g i ) + (q? )z tr(bt 0 nk g n : : : g i ) qz tr(w n g n?2 : : : g i ) + (q? )z tr(w n g n : : : g i ): Note 2 The relations () imply that tr(xg n yg n ) tr(g n xg n y) for any x; y 2 H n. It remains now to show ( ). On the one hand we have: tr(w n t 0 k Lemma ;(vi) n g n ) tr(w n g n t 0 k n ) (30 ) z tr(w n t 0 nk ): On the other hand in order to calculate tr(g n w n t 0 nk ) we examine the dierent possibilities for w n : { If w n 2 H n ; then tr(g n w n t 0 k n ) tr(w n g n2 t 0 k n g n ) (q? ) tr(w n t 0 k n ) + q tr(w n t 0 k n g n ) (q? ) tr(w n t 0 k n ) +z tr(w n t 0 k n ) + (? q) tr(w n t 0 k n ) z tr(w n t 0 k n ): { If w n ag n b; where a; b 2 H n ; then tr(g n ag n bt 0 k n ) tr(ag n g n g n bt 0 nk g n q tr(ag n g n g n bt 0 k n g n ) + (q? ) tr(ag n g n g n bt 0 k n )