B-type coefficient polynomial

B-TYPE COEFFICIENT POL YNOMIAL NOBORU ITO AND MA YUKO KON Abstract. An A-type co efficient polynomial introduced by Kaw auchi [4] re- cov ers the HOMFL Y–PT polynomial as a formal p ow er series within sk ein theory . A notable feature of this construction is that each co efficient defines a link in v arian t, yielding an infinite sequence of inv arian ts, while the low-degree coefficients are relativ ely easy to compute. In this paper, we extend this view- point to the B-t ype setting. Unlik e the A-t yp e case, the B-t ype setting requires a genuinely new inductiv e scheme due to the four-term skein relation. More precisely , we introduce coefficient p olynomials asso ciated with the B-type sk ein relation and show that their generating series recov ers the Kauffman p olyno- mial. W e further pro ve that these co efficient p olynomials are well-defined and that the resulting generating series is in v arian t under the corresp onding Rei- demeister mov es. 1. Introduction Sk ein theory has play ed a fundamental role in low-dimensional top ology as a framew ork for treating link inv arian ts in a unified manner. The Jones p olynomial and the HOMFL Y–PT polynomial [1, 5] are represen tativ e examples, b oth of whic h can be c haracterized b y sk ein relations together with normalization conditions. The Kauffman polynomial [3] is c haracterized b y a four-term skein relation together with regular isotop y conditions. The A-type coefficient p olynomial introduced b y Kaw auc hi [4] pro vides a sk ein- theoretic construction that reco v ers the HOMFL Y–PT p olynomial as a formal p o w er series. In this theory , eac h co efficien t can b e in terpreted as a link in v ariant, yielding an infinite sequence of inv ariants, and in particular the lo w-degree co effi- cien ts are relatively easy to compute. F rom this viewp oin t, the A-type co efficien t p olynomial ma y b e regarded as a co efficien t-level refinement of the HOMFL Y–PT p olynomial. The purp ose of this pap er is to sho w that this co efficient-lev el construction can b e extended to the B-type setting. More precisely , following the same formal pow er se- ries framew ork as in Kaw auchi’s construction, we introduce co efficient p olynomials asso ciated with the B-t ype skein relation and show that their generating series re- co vers the Kauffman p olynomial. Th us, the Kauffman polynomial is reconstructed from a family of co efficien t in v ariants arising from skein theory . A cen tral p oin t is that the B-t yp e case is not a formal repetition of the A-type one. In the A-t yp e case, the w ell-definedness is based on a three-term sk ein relation together with an inductiv e argument on monotone diagrams. In contrast, in the B-t yp e case, one first needs a new co efficient-lev el formulation of a four-term skein relation, and the same inductive mechanism do es not apply directly . Date : March 26, 2026. 1 2 NOBORU ITO AND MA YUKO KON W e ov ercome this difficult y by refining the initial conditions for the inductive construction and by introducing a bo okk eeping of the c hange in the n umber of comp onen ts under the tw o splice op erations. This reflects a structural difference b et w een three-term and four-term skein relations: while the inductive construction in the A-t yp e case closes within the class of monotone diagrams, the four-term rela- tion naturally produces connected sums of monotone diagrams. As a consequence, Ka wauc hi’s metho d extends to the B-type setting only after incorp orating this new phenomenon in to the induction scheme. As a result, w e obtain a co efficien t-level extension of Kaw auchi’s metho d to the B-t yp e setting, pro viding a foundation for further systematic studies of co efficien t in v ariants asso ciated with four-term sk ein relations. The organization of this pap er is as follows. In Section 2 we in tro duce the notation and diagrammatic con v entions used throughout the paper. In Section 3 w e state the main theorem describing the coefficient polynomials α n ( D ; y ). Section 4 is dev oted to the construction of α n ( D ; y ) and the proof of the main theorem using an inductiv e argumen t based on w arping crossings. In Section 5 we prov e uniqueness for the co efficient p olynomials and for the resulting generating series, and show that the latter recov ers the Kauffman p olynomial. Finally , Section 6 contains remarks on the structure of the coefficient p olynomials and directions for further study . 2. Preliminaries and not a tion Throughout this pap er, diagrams are considered up to planar isotop y . W e fix the con v entions and notation used in the construction of the coefficient polynomials asso ciated with the B-type sk ein relation. 2.1. Link diagrams and crossings. Definition 2.1 (link diagrams) . An unoriente d link diagr am D is a generic im- mersion of a disjoin t union of circles in the plane with o v er/under information at eac h transverse double p oin t; an oriente d link diagr am is such a diagram with an orien tation on each comp onen t. Notation 2.2 ( c ( D ), r ( D )) . Let c ( D ) be the crossing num ber of D and r ( D ) the n umber of comp onen ts of D . Definition 2.3 (writhe) . Giv en an oriented diagram D , each crossing p has a sign ϵ ( p ) ∈ { +1 , − 1 } . The writhe is defined by w ( D ) = X p : a crossing of D ϵ ( p ) .    @ @ D p D ∞ D 0 Figure 1. A crossing D p and its tw o smoothings D ∞ and D 0 . 2.2. Lo cal splices. In Figure 1, w e denote by D ∞ and D 0 the t wo smo othings of D at a crossing p shown there. W e call the smo othing pro ducing D ∞ (resp. D 0 ) the A -splice (resp. B -splice). Accordingly , w e set D p 1 = D ∞ , D p 2 = D 0 . B-TYPE COEFFICIENT POL YNOMIAL 3 Notation 2.4 (F our lo cal diagrams) . W e use the notation ( D + , D − , D ∞ , D 0 ) for the four standard lo cal diagrams at a crossing. The skein relation is formulated for unorien ted diagrams. F or the purp ose of pro ofs, we ma y temporarily equip the strands near p with an auxiliary orientation. This determines a sign ϵ ( p ) and fixes an ordered pair ( D ϵ ( p ) , D − ϵ ( p ) ) whic h is either ( D + , D − ) or ( D − , D + ). This auxiliary choice is local and fixed within eac h argument. R emark 1 . F or notational conv enience, we use the sym b ols p 1 and p 2 to distinguish the t wo splices; th us D p 1 = D ∞ and D p 2 = D 0 . 2.3. Comp onen t-c hange quan tities. Definition 2.5 (Comp onen t indicator) . Let p b e a crossing of D . Define δ ( p ) = ( 0 if the strands at p belong to the same component , 1 otherwise . Definition 2.6 (Signed comp onen t change) . F or the spliced diagrams D p 1 and D p 2 , define ∆( p 1) = r ( D p 1 ) − r ( D ) , ∆( p 2) = r ( D p 2 ) − r ( D ) . Th us ∆( pi ) records the signed change in the num b er of comp onents under the corresp onding splice. R emark 2 . In the inductiv e construction, the coefficient index shifts according to n 7→ n + ∆( pi ) − 1 , and successiv e splices produce additiv e shifts suc h as n + ∆( p 1) + ∆( q 1) − 2 . This shift is essen tial in form ulating the coefficient-lev el skein relation. 2.4. Algebraic con v entions. Notation 2.7. Fix a commutativ e ring R containing Z [ y ± 1 ] as a subring. W e write R [[ z ]] for the ring of formal pow er series in z with coefficients in R . 2.5. Co efficien t p olynomials. Definition 2.8. F or an unoriented diagram D , we define elemen ts α n ( D ; y ) ∈ R ( n ∈ Z ) . Definition 2.9 (Generating series of the co efficient p olynomials) . Let D be an unorien ted diagram. Define L D ( y , z ) = z 1 − r ( D ) X n ≥ 0 α n ( D ; y ) z n ∈ R [[ z ]] . Definition 2.10 (Normalization) . F or an oriented diagram D , define F D ( y , z ) = y − w ( D ) L D ( y , z ) . 4 NOBORU ITO AND MA YUKO KON 2.6. Base p oin ts and w arping degree. Definition 2.11 (Sequences of base p oin ts and induced directions) . Let D b e an unorien ted diagram with r ( D ) = r comp onen ts. A se quenc e of b ase p oints is an ordered r -tuple a = ( a 1 , . . . , a r ) with one p oin t a i on each comp onent. F or each comp onen t, we additionally choose a direction of trav el starting from a i . (Equiv- alen tly , w e choose an orien tation on each comp onent; this choice is used only to define w arping data.) Definition 2.12 (Connected sequences of base points) . Tw o sequences of base p oin ts a = ( a 1 , . . . , a r ) and a ′ = ( a ′ 1 , . . . , a ′ r ) on a diagram D are said to b e c onne cte d if a i and a ′ i b elong to the same connected comp onen t of D for eac h i = 1 , . . . , r . Definition 2.13 (First-encounter rule at a crossing) . Fix a based-and-directed diagram ( D , a ) as in Definition 2.11. F or a crossing p of D , tra v erse eac h component once starting from its base p oin t in the chosen direction. Among the t wo strands meeting at p , exactly one of them is encountered first in this trav ersal. W e call it the first-enc ounter e d str and at p (with respect to ( D , a )). Definition 2.14 (W arping crossing and w arping degree) . A crossing p of D is called a warping cr ossing p oint of ( D, a ) if the first-encoun tered strand at p is the under -strand of the crossing. The warping de gr e e d a ( D ) is the n um b er of warping crossing points of ( D, a ). Definition 2.15 (Monotone diagram) . A based-and-directed diagram ( D, a ) is called monotone if d a ( D ) = 0. Notation 2.16 (Complexit y) . W e write cd ( D ) =  c ( D ) , d a ( D )  for the pair consisting of the crossing n umber and the w arping degree, which will b e used as a complexit y measure in the inductive arguments. 2.7. Disjoin t union and connected sum. Notation 2.17. Let O denote the zero-crossing knot diagram. W e denote disjoin t union b y ⊔ and may write O + D for O ⊔ D . Notation 2.18. W e write D # D ′ for a connected sum of unoriented diagrams. When the orientation of D # D ′ is needed, it will b e specified after summing. 3. Main Resul t Theorem 3.1. F or every unoriente d link diagr am D , ther e exists a se quenc e of c o efficient p olynomials α n ( D ; y ) ∈ R ( n ∈ Z ) such that the fol lowing pr op erties hold. (1) α n ( D ; y ) is invariant under the R eidemeister moves of typ e I I and typ e I I I , and tr ansforms under the R eidemeister move of typ e I as in (1) : α n ( ; y ) = y α n ( ; y ) , (1) α n ( ; y ) = y − 1 α n ( ; y ) . B-TYPE COEFFICIENT POL YNOMIAL 5 (2) F or the zer o-cr ossing knot diagr am O , α n ( O ; y ) = δ n, 0 , (2) wher e δ n, 0 denotes the Kr one cker delta. (3) F or ( D + , D − , D p 1 , D p 2 ) as in Figur e 1, α n ( D + ; y ) + α n ( D − ; y ) (3) = α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) . Her e, ∆( p 1) and ∆( p 2) ar e as in Definition 2.6. 4. Construction and Proof of Theorem 3.1 Before pro ceeding to the pro of, w e briefly explain the main difference from the A- t yp e case. In the A-type construction, the sk ein relation consists of three terms, and the inductiv e argumen t closes within the class of monotone diagrams. In con trast, the B-type sk ein relation in volv es four terms. When comparing the expansions at t wo distinct warping crossings, the splice terms naturally pro duce intermediate diagrams whic h may decomp ose as connected sums of monotone diagrams. F or this reason, in pro ofs of this section, w e enlarge the initial class to include connected sums of monotone diagrams. Apart from this p oint, the inductiv e structure follows the same strategy as in the A-t yp e case. The necessit y of this enlargement is in trinsic to the four-term nature of the B-t ype skein relation and do es not arise in the three-term A-type case. T o pro v e the theorem, w e use induction on the crossing num ber c ( D ) = m . When m = 0, w e set α n ( D , a ; y ) = ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 , where a is a sequence of base p oints and r is the num ber of connected comp onen ts. Then an y diagram D with m = 0 satisfies prop erties (1)–(3) in Theorem 3.1. W e supp ose that for any diagram D with c ( D ) < m , there exists α n ( D ; y ) which satisfies prop erties (1)–(3) in Theorem 3.1. W e sho w that we can construct α n ( D ; y ) for D with c ( D ) = m b y the follo wing steps. (i) F or a pair ( D, a ) of a diagram and a sequence of base points with c ( D ) = m , w e define α n ( D , a ; y ). (ii) W e show that α n ( D , a ; y ) do es not dep end on the choice of the sequence of base points a . W e denote this b y α n ( D ; y ). (iii) F or an y diagram D with c ( D ) ≤ m , α n ( D ; y ) satisfies prop erties (1)–(3). W e define α n ( D , a ; y ) as follows. Definition 4.1. When D is an unorien ted monotone diagram or a connected sum of monotone diagrams, w e set α n ( D , a ; y ) = y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 . When s = d a ( D ) > 0, we define α n ( D , a ; y ) b y induction: α n ( D , a ; y ) = α n ( D ϵ ( p ) , a ; y ) (4) = − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) . Here ∆( p 1) and ∆( p 2) are as in Definition 2.6. 6 NOBORU ITO AND MA YUKO KON Lemma 4.2. L et D b e a diagr am with c ( D ) = m . When s > 0 , α n ( D , a ; y ) do es not dep end on the choic e of the warping cr ossing p oint p of ( D , a ) . Pr o of. Let q  = p b e another w arping crossing point of ( D , a ). W e will sho w the follo wing equalit y: − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) = − α n ( D − ϵ ( q ) , a ; y ) + α n +∆( q 1) − 1 ( D q 1 ; y ) + α n +∆( q 2) − 1 ( D q 2 ; y ) . Since d a ( D − ϵ ( p ) ) < s and q is a warping crossing point of ( D − ϵ ( p ) , a ), by the induction h yp othesis on s , w e ha ve α n ( D − ϵ ( p ) , a ; y ) = − α n (( D − ϵ ( p ) ) − ϵ ( q ) , a ; y ) + α n +∆( q 1) − 1 (( D − ϵ ( p ) ) q 1 ; y ) + α n +∆( q 2) − 1 (( D − ϵ ( p ) ) q 2 ; y ) . Next, since c ( D p 1 ) < m and c (( D p 1 ) − ϵ ( q ) ) < m , by the induction hypothesis on the n umber of crossings, w e ha ve α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 1) − 1 (( D p 1 ) − ϵ ( q ) ; y ) = α n +∆( p 1)+∆ ′ ( q 1) − 2 (( D p 1 ) q 1 ; y ) + α n +∆( p 1)+∆ ′ ( q 2) − 2 (( D p 1 ) q 2 ; y ) . Here ∆ ′ ( q 1) (resp. ∆ ′ ( q 2)) denotes the signed change in the num b er of comp onen ts under the corresp onding splice at q in the diagram D p 1 ; namely , ∆ ′ ( q i ) = r  ( D p 1 ) q i  − r ( D p 1 ) ( i = 1 , 2) . Similarly , w e ha ve α n +∆( p 2) − 1 ( D p 2 ; y ) + α n +∆( p 2) − 1 (( D p 2 ) − ϵ ( q ) ; y ) = α n +∆( p 2)+∆ ′ ( q 1) − 2 (( D p 2 ) q 1 ; y ) + α n +∆( p 2)+∆ ′ ( q 2) − 2 (( D p 2 ) q 2 ; y ) . F rom these equations, w e obtain − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) = α n (( D − ϵ ( p ) ) − ϵ ( q ) , a ; y ) − α n +∆( q 1) − 1 (( D − ϵ ( p ) ) q 1 ; y ) − α n +∆( q 2) − 1 (( D − ϵ ( p ) ) q 2 ; y ) − α n +∆( p 1) − 1 (( D p 1 ) − ϵ ( q ) ; y ) + α n +∆( p 1)+∆ ′ ( q 1) − 2 (( D p 1 ) q 1 ; y ) + α n +∆( p 1)+∆ ′ ( q 2) − 2 (( D p 1 ) q 2 ; y ) − α n +∆( p 2) − 1 (( D p 2 ) − ϵ ( q ) ; y ) + α n +∆( p 2)+∆ ′ ( q 1) − 2 (( D p 2 ) q 1 ; y ) + α n +∆( p 2)+∆ ′ ( q 2) − 2 (( D p 2 ) q 2 ; y ) . Similarly , w e ha ve − α n ( D − ϵ ( q ) , a ; y ) + α n +∆( q 1) − 1 ( D q 1 ; y ) + α n +∆( q 2) − 1 ( D q 2 ; y ) = α n (( D − ϵ ( q ) ) − ϵ ( p ) , a ; y ) − α n +∆( p 1) − 1 (( D − ϵ ( q ) ) p 1 ; y ) − α n +∆( p 2) − 1 (( D − ϵ ( q ) ) p 2 ; y ) − α n +∆( q 1) − 1 (( D q 1 ) − ϵ ( p ) ; y ) + α n +∆( q 1)+∆ ′ ( p 1) − 2 (( D q 1 ) p 1 ; y ) + α n +∆( q 1)+∆ ′ ( p 2) − 2 (( D q 1 ) p 2 ; y ) − α n +∆( q 2) − 1 (( D q 2 ) − ϵ ( p ) ; y ) + α n +∆( q 2)+∆ ′ ( p 1) − 2 (( D q 2 ) p 1 ; y ) + α n +∆( q 2)+∆ ′ ( p 2) − 2 (( D q 2 ) p 2 ; y ) . Note that the total change in the num b er of comp onen ts is indep enden t of the order of the splices. Th us, w e obtain the desired equality . □ B-TYPE COEFFICIENT POL YNOMIAL 7 Lemma 4.3. F or any diagr am D with c ( D ) = m and any se quenc e of b ase p oints a , α n ( D , a ; y ) do es not dep end on the choic e of a se quenc e of b ase p oints that is c onne cte d to a . Pr o of. It is sufficient to show α n ( D , a ; y ) = α n ( D , a ′ ; y ) for the sequences of base p oin ts a = ( a 1 , . . . , a i − 1 , a i , a i +1 , . . . , a r ) , a ′ = ( a 1 , . . . , a i − 1 , a ′ i , a i +1 , . . . , a r ) as sho wn in Figure 2. p a’ a i i p a’ a i i Figure 2. A connected c hange of the sequence of base points. The base point a i mo ves along the component across the crossing p to a ′ i . Supp ose that q  = p is another w arping crossing p oin t. Then w e ha v e α n ( D , a ; y ) + α n ( D − ϵ ( q ) , a ; y ) = α n +∆( q 1) − 1 ( D q 1 ; y ) + α n +∆( q 2) − 1 ( D q 2 ; y ) , α n ( D , a ′ ; y ) + α n ( D − ϵ ( q ) , a ′ ; y ) = α n +∆( q 1) − 1 ( D q 1 ; y ) + α n +∆( q 2) − 1 ( D q 2 ; y ) . By the induction h yp othesis on the n umber of crossing points, we obtain α n ( D , a ; y ) − α n ( D , a ′ ; y ) = α n ( D − ϵ ( q ) , a ′ ; y ) − α n ( D − ϵ ( q ) , a ; y ) . By applying crossing changes aw a y from p if necessary , the diagram D b ecomes one of the follo wing: • δ ( p ) = 1 and d a ( D ) = d a ′ ( D ) = 1, or δ ( p ) = 1 and d a ( D ) = d a ′ ( D ) = 0; • δ ( p ) = 0 and d a ( D ) = 1 , d a ′ ( D ) = 0; • δ ( p ) = 0 and d a ( D ) = 0 , d a ′ ( D ) = 1. When d a ( D ) = d a ′ ( D ) = 0, we hav e α n ( D , a ; y ) = y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 = α n ( D , a ′ ; y ) . If d a ( D ) = d a ′ ( D ) = 1, then the diagram D − ϵ ( p ) is a monotone diagram. Thus, w e obtain α n ( D − ϵ ( p ) , a ; y ) = α n ( D − ϵ ( p ) , a ′ ; y ) . Therefore, w e ha ve α n ( D , a ; y ) = − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) = − α n ( D − ϵ ( p ) , a ′ ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) = α n ( D , a ′ ; y ) . 8 NOBORU ITO AND MA YUKO KON No w, let us consider the case where δ ( p ) = 0, d a ( D ) = 1, and d a ′ ( D ) = 0. Note that the n umber of crossing points of D p 1 and D p 2 is m − 1. By taking a suitable sequence of base p oin ts and directions, D p 1 and D p 2 b ecome monotone diagrams or connected sums of monotone diagrams. Using (4), w e hav e α n ( D , a ; y ) = − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) . Since δ ( p ) = 0, either ∆( p 1) = 1 and ∆( p 2) = 0, or ∆( p 1) = 0 and ∆( p 2) = 1 holds. In either case, w e obtain − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) = − y w ( D ) − 2 ϵ ( p ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 + y w ( D ) − ϵ ( p ) ( − 1) n  r n  ( y + y − 1 ) r − n + y w ( D ) − ϵ ( p ) ( − 1) n − 1  r − 1 n − 1  ( y + y − 1 ) r − n = y w ( D ) ( − 1) n − 1 ( y + y − 1 ) r − n − 1 ·  y − 2 ϵ ( p )  r − 1 n  − y − ϵ ( p )  r n  ( y + y − 1 ) + y − ϵ ( p )  r − 1 n − 1  ( y + y − 1 )  . Since ϵ ( p ) = ± 1, by a straigh tforward computation, w e ha v e y − 2 ϵ ( p )  r − 1 n  − y − ϵ ( p )  r n  ( y + y − 1 ) + y − ϵ ( p )  r − 1 n − 1  ( y + y − 1 ) = −  r − 1 n  . Th us, w e see that α n ( D , a ; y ) = y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 . □ Lemma 4.4. When c ( D ) ≤ m , the L aur ent p olynomials α n ( D ; y ) satisfy α n ( ; y ) = y α n ( ; y ) , α n ( ; y ) = y − 1 α n ( ; y ) . Mor e over, they ar e invariant under R eidemeister moves of typ es I I and I I I . Pr o of. Let D ′ b e a diagram obtained from D by a Reidemeister mo v e. First, we sho w that by choosing suitable base p oin ts, the w arping crossing p oints of ( D , a ) and those of ( D ′ , a ) are in one-to-one corresp ondence. If D ′ is obtained by a Reidemeister mo v e of t yp e I, b y choosing a base p oin t a i as sho wn in Figure 3, w e obtain the desired one-to-one corresp ondence. Next, we show that when t w o strands b elong to differen t comp onen ts of D , α n ( D , a ; y ) is inv arian t under the lo cal transformation shown in Figure 4; that is, α n ( D , a ; y ) = α n ( D ′ , a ; y ). By (4), w e hav e α n ( D ϵ ( p ) , a ; y ) = − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) . B-TYPE COEFFICIENT POL YNOMIAL 9 a i a i a i a i a i a i a i a i Figure 3. Correspondence of the sequence of base p oints under a Reidemeister mo v e of t yp e I. The four p ossible orien tation cases are sho wn. q p q p Figure 4. A lo cal mo ve relating diagrams D (left) and D ′ (righ t). The crossings p and q corresp ond under this mov e. Again using (4), w e obtain − α n ( D − ϵ ( p ) , a ; y ) = α n ( D ′ , a ; y ) − α n +∆( q 1) − 1 (( D − ϵ ( p ) ) q 1 ; y ) − α n +∆( q 2) − 1 (( D − ϵ ( p ) ) q 2 ; y ) . F rom these equations, w e ha ve α n ( D , a ; y ) = α n ( D ′ , a ; y ) + α n − 1 ( D p 1 ; y ) + α n − 1 ( D p 2 ; y ) − α n − 1 (( D − ϵ ( p ) ) q 1 ; y ) − α n − 1 (( D − ϵ ( p ) ) q 2 ; y ) . (5) F or an y orien tation of the strands, b y the induction hypothesis on the n umber of crossing p oin ts, w e see that α n ( D , a ; y ) = α n ( D ′ , a ; y ) . Figure 5 illustrates one case of the ab ov e calculation. F or the other orien tations, the same argument works. The details are left to the reader. Using this, w e can adjust the upper and low er positions of the strands. By Lemma 4.2, setting the base p oint a i as shown in Figure 6 yields a one-to-one corresp ondence. Moreo ver, for a Reidemeister mov e of t yp e I I I, w e hav e a one-to-one corresp on- dence as shown in Figure 7. 10 NOBORU ITO AND MA YUKO KON q p q p q p D D p1 q p D p2 D ε - (p ) q p D’ q p q p ( ) q1 ( ) q2 D ε - (p ) D ε - (p ) Figure 5. Corresp ondence of the six diagrams app earing in equa- tion (5), obtained b y applying the skein relation twice. The dia- gram D − ϵ ( p ) is further expanded at the crossing q . The opp osite orien tation case, corresp onding to a non-braid type bigon, is omit- ted. In the follo wing, we pro ceed by induction on s = d a ( D ) = d a ( D ′ ). When D is a monotone diagram or a connected sum of monotone diagrams, w e hav e α n ( D , a ; y ) = y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 , α n ( D ′ , a ; y ) = y w ( D ′ ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 . B-TYPE COEFFICIENT POL YNOMIAL 11 a i a i a i a j a j a j a i a j i ≧ j i ＜ j Figure 6. Correspondence of the sequence of base p oints under a Reidemeister mov e of t yp e I I. The t w o cases i ≥ j and i < j are sho wn. q p q p A A a i a i Figure 7. Corresp ondence of the sequence of base p oin ts under a Reidemeister mov e of type I I I. The crossings p and q corresp ond under the mov e. F rom this, we see that α n ( ; y ) = y α n ( ; y ) , α n ( ; y ) = y − 1 α n ( ; y ) . Since Reidemeister mov es of types I I and I I I do not change the writhe, w e hav e α n ( D , a ; y ) = α n ( D ′ , a ; y ). Next, we assume that the result holds for d a 1 ( D 1 ) = d a 1 ( D ′ 1 ) < s , where the w arping crossing points of D and D ′ are assumed to b e in one-to-one corresp on- dence, as discussed abov e. Let p be a corresp onding w arping crossing p oin t of ( D , a ) and ( D ′ , a ). Then w e hav e α n ( D , a ; y ) = − α n ( D − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) , α n ( D ′ , a ; y ) = − α n ( D ′ − ϵ ( p ) , a ; y ) + α n +∆( p 1) − 1 ( D ′ p 1 ; y ) + α n +∆( p 2) − 1 ( D ′ p 2 ; y ) . If D ′ is the diagram obtained from D by a Reidemeister mo v e of t yp e I, b y the induction h yp othesis on s , w e ha ve α n ( D − ϵ ( p ) , a ; y ) = y τ α n ( D ′ − ϵ ( p ) , a ; y ) . Here, τ ∈ { +1 , − 1 } denotes the writhe change asso ciated with the type I mo v e. If D ′ is obtained by a Reidemeister mov e of t yp es I I or I I I, w e hav e α n ( D − ϵ ( p ) , a ; y ) = α n ( D ′ − ϵ ( p ) , a ; y ) . W e distinguish tw o cases: first, the case where the chosen warping crossing is not one of the three crossings app earing in Figure 7; second, the case where it is one of those three crossings. In the former case, ( D p 1 , D ′ p 1 ) and ( D p 2 , D ′ p 2 ) are pairs of diagrams with at most m − 1 crossings related by a Reidemeister mov e, resp ectiv ely . 12 NOBORU ITO AND MA YUKO KON By the induction hypothesis on the crossing num ber m , for a Reidemeister mov e of t yp e I, we hav e α n +∆( p 1) − 1 ( D p 1 ; y ) = y τ α n +∆( p 1) − 1 ( D ′ p 1 ; y ) , α n +∆( p 2) − 1 ( D p 2 ; y ) = y τ α n +∆( p 2) − 1 ( D ′ p 2 ; y ) . F or Reidemeister mov es of types I I and I I I, we obtain α n +∆( p 1) − 1 ( D p 1 ; y ) = α n +∆( p 1) − 1 ( D ′ p 1 ; y ) , α n +∆( p 2) − 1 ( D p 2 ; y ) = α n +∆( p 2) − 1 ( D ′ p 2 ; y ) . In the latter case, let u denote the chosen crossing in Figure 7. Note that in Figure 7, the crossings lab eled A and q can b e warping crossing p oints, whereas p cannot. W e see that ( D u 1 , D ′ u 1 ) and ( D u 2 , D ′ u 2 ) are pairs of diagrams that are either identical or related by t w o Reidemeister mo ves of t yp e I I, ha ving at most m − 1 crossings. Thus, by the induction hypothesis on the crossing n umber m , w e ha ve α n +∆( u 1) − 1 ( D u 1 ; y ) = α n +∆( u 1) − 1 ( D ′ u 1 ; y ) , α n +∆( u 2) − 1 ( D u 2 ; y ) = α n +∆( u 2) − 1 ( D ′ u 2 ; y ) . This completes the proof. □ Lemma 4.5. F or any diagr am D with c ( D ) = m , α n ( D , a ; y ) do es not dep end on the choic e of the se quenc e of b ase p oints a . Pr o of. If the n umber of components is r = 1, w e ha v e a = ( a 1 ), and an y sequence of base p oin ts is connected to a . Thus, the result holds b y Lemma 4.3. Next, assume that r ≥ 2, and suppose that the result holds for diagrams with r − 1 comp onen ts. W e pro ceed b y induction on s = d a ( D ). If s = 0, meaning D is a monotone diagram, we can transform D in to O + D ′ b y Reidemeister mov es of types I, I I, and I I I, where D ′ is a monotone diagram. By Lemma 4.4 and the induction h yp othesis on m , w e ha ve α n ( D , a ; y ) = y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 , α n ( D ′ , a ; y ) = y w ( D ′ ) ( − 1) n  r − 2 n  ( y + y − 1 ) r − n − 2 . Hence, w e obtain α n ( D , a ; y ) = y w ( D ) − w ( D ′ ) ( y + y − 1 ) r − 1 r − n − 1 α n ( D ′ , a ′ ; y ) . By the induction h yp othesis on r , w e see that α n ( D ′ , a ′ ; y ) do es not dep end on a ′ . Th us, α n ( D , a ; y ) does not dep end on a . F or the case s > 0, by rep eatedly applying the skein relation, we can reduce α n ( D , a ; y ) to diagrams with fewer crossings, and ultimately to the monotone case s = 0. Hence the result follows from the monotone case. □ Next, w e sho w the following. Lemma 4.6. We c an c ompute α n ( D ; y ) ( n = 0 , ± 1 , . . . ) using (1)–(3) in The- or em 3.1. In p articular, α n ( D ; y ) = 0 exc ept for finitely many n , and we have α n ( D ; y ) = 0 for n < 0 . B-TYPE COEFFICIENT POL YNOMIAL 13 Pr o of. First, we show that for an r -comp onen t zero-crossing trivial link diagram O r , α n ( O r ; y ) = ( ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 , if n ≥ 0, 0 , if n < 0. (6) Let ( D + , D − , D p 1 , D p 2 ) satisfy c ( D + ) = c ( D − ) = 1. If δ ( p ) = 1, there exists another crossing, so c ( D + ) = c ( D − ) ≥ 2. Thus, w e hav e δ ( p ) = 0. Moreo ver, either ∆( p 1) = 1 and ∆( p 2) = 0, or ∆( p 1) = 0 and ∆( p 2) = 1. Using (1) and (3), we obtain ( y + y − 1 ) α n ( O r ; y ) = α n ( O r +1 ; y ) + α n − 1 ( O r ; y ) . W e can compute α n ( O r +1 ; y ) inductively using this equation. If r = 1, by (2), we ha ve α n ( O 2 ; y ) = ( y + y − 1 ) α n ( O ; y ) − α n − 1 ( O ; y ) = ( y + y − 1 ) δ n, 0 − δ n − 1 , 0 . This satisfies (6). Next, assume that (6) holds for O r , and consider the case of O r +1 . If n ≥ 0, w e ha ve α n ( O r +1 ; y ) = ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − ( − 1) n − 1  r − 1 n − 1  ( y + y − 1 ) r − ( n − 1) − 1 = ( − 1) n  r n  ( y + y − 1 ) r − n . On the other hand, if n < 0, w e hav e α n ( O r ; y ) = 0 and α n − 1 ( O r ; y ) = 0, and hence α n ( O r +1 ; y ) = 0 . Th us, w e see that equation (6) holds for the zero-crossing trivial diagram O r +1 . Consequen tly , (6) holds for any n and r . F rom this and (1) in Theorem 3.1, if D is monotone, then we hav e α n ( D ; y ) = ( ( − 1) n y w ( D )  r − 1 n  ( y + y − 1 ) r − n − 1 , if n ≥ 0, 0 , if n < 0, (7) where w ( D ) is the writhe. Next, w e show that α n ( D ; y ) can b e computed for any diagram D with cd ( D ) = ( k , s ), assuming that α n ( D ′ ; y ) can b e computed for any diagram D ′ with cd ( D ′ ) < ( k , s ). If s = 0, D is a monotone diagram, and α n ( D ; y ) satisfies equation (7). Therefore, w e consider the case where s > 0. W e choose a sequence of base p oin ts a = ( a 1 , . . . , a i , . . . , a r ) with d a ( D ) = s and a w arping crossing p oin t p . Then there exist in tegers s 1 , s 2 ≥ 0 such that cd ( D − ϵ ( p ) ) ≤ ( k , s − 1) < ( k , s ) , cd ( D p 1 ) ≤ ( k − 1 , s 1 ) < ( k , s ) , (8) cd ( D p 2 ) ≤ ( k − 1 , s 2 ) < ( k , s ) . By the induction hypothesis, we can compute α n ( D − ϵ ( p ) ; y ), α n +∆( p 1) − 1 ( D p 1 ; y ), and α n +∆( p 2) − 1 ( D p 2 ; y ). Thus, we can calculate α n ( D ; y ) using (3). Next, w e show that α n ( D ; y ) = 0 for n < 0. Assume that α n ( D ′ ; y ) = 0 for n < 0 for an y diagram D ′ with c ( D ′ ) < k . Let D b e a diagram with c ( D ) = k . F or 14 NOBORU ITO AND MA YUKO KON an y crossing point p , w e see that c ( D p 1 ) < k , c ( D p 2 ) < k , n + ∆( p 1) − 1 < 0, and n + ∆( p 2) − 1 < 0. Th us, b y (3), we hav e α n ( D + ; y ) + α n ( D − ; y ) = 0 , where D = D + or D = D − . By applying crossing c hanges rep eatedly , the diagram D can b e transformed into a monotone diagram. Therefore, we conclude that α n ( D ; y ) = 0 for n < 0. Finally , using (3) and (8), and by induction on cd ( D ) = ( k , s ), we see that α n ( D ; y ) = 0 except for finitely many n . □ F rom Theorem 3.1, the generating series L associated with the co efficient p oly- nomials α n and its writhe-normalized v ersion F satisfy the follo wing prop erties. Prop osition 4.7. If D and D ′ ar e r e gularly isotopic diagr ams, then L D = L D ′ . Mor e over, the fol lowing identities hold for al l skein quadruples of diagr ams that ar e identic al everywher e exc ept in the lo c al pictur es indic ate d b elow: L D + + L D − = z ( L D ∞ + L D 0 ) , L = y L , L = y − 1 L , L  = 1 . Prop osition 4.8. The L aur ent p olynomial F D ( y , z ) ∈ R [ z , z − 1 ] is an ambient isotopy invariant of oriente d links. Lemma 4.9 verifies that the p olynomial L b eha v es multiplicativ ely under con- nected sum and satisfies the expected form ula for disjoint union. Lemma 4.9. The fol lowing formulas hold for the p olynomial L with r esp e ct to c onne cte d sum and disjoint union: L D # D ′ = L D L D ′ , L D ⊔ D ′ = dL D L D ′ , wher e d = z − 1 ( y + y − 1 ) − 1 . Pr o of. First, we consider the case where D and D ′ are monotone diagrams or connected sums of monotone diagrams with r and r ′ comp onen ts, resp ectively . Then w e ha ve L D = z 1 − r ∞ X n =0 α n ( D ; y ) z n = z 1 − r ∞ X n =0 y w ( D ) ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 z n , L D ′ = z 1 − r ′ ∞ X n =0 α n ( D ′ ; y ) z n = z 1 − r ′ ∞ X n =0 y w ( D ′ ) ( − 1) n  r ′ − 1 n  ( y + y − 1 ) r ′ − n − 1 z n . B-TYPE COEFFICIENT POL YNOMIAL 15 Hence, w e obtain L D L D ′ = z 2 − r − r ′ ∞ X n =0 n X k =0 α k ( D ; y ) α n − k ( D ′ ; y ) ! z n = z 2 − r − r ′ ∞ X n =0 n X k =0 y w ( D ) ( − 1) k  r − 1 k  ( y + y − 1 ) r − k − 1 · y w ( D ′ ) ( − 1) n − k  r ′ − 1 n − k  ( y + y − 1 ) r ′ − ( n − k ) − 1 ! z n = z 2 − r − r ′ ∞ X n =0 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 2 · n X k =0  r − 1 k  r ′ − 1 n − k  ! z n = z 2 − r − r ′ ∞ X n =0 y w ( D # D ′ ) ( − 1) n  r + r ′ − 2 n  ( y + y − 1 ) r + r ′ − n − 2 z n = L D # D ′ , b y Definition 4.1. Next, we assume that L D # D ′ = L D L D ′ whenev er cd ( D ) < ( k , s ). Let p b e a w arping crossing p oin t of D . F rom Proposition 4.7, w e ha ve L D # D ′ = − L D − ϵ ( p ) # D ′ + z L D ∞ # D ′ + z L D 0 # D ′ . Since cd ( D − ϵ ( p ) ) < ( k , s ), cd ( D ∞ ) < ( k , s ), and cd ( D 0 ) < ( k , s ), by the induction h yp othesis, we hav e L D # D ′ = ( − L D − ϵ ( p ) + z L D ∞ + z L D 0 ) L D ′ = L D L D ′ . W e next show L D ⊔ D ′ = dL D L D ′ when D and D ′ are monotone diagrams or connected sums of monotone diagrams with r and r ′ comp onen ts, resp ectively . 16 NOBORU ITO AND MA YUKO KON Using the computation abov e, we obtain dL D L D ′ = z 2 − r − r ′ ∞ X n =0 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 2  r + r ′ − 2 n  z n ! ·  z − 1 ( y + y − 1 ) − 1  = z 1 − r − r ′ ∞ X n =0 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 1  r + r ′ − 2 n  z n ! + z 1 − r − r ′ ∞ X n =0 y w ( D )+ w ( D ′ ) ( − 1) n +1 ( y + y − 1 ) r + r ′ − n − 2  r + r ′ − 2 n  z n +1 ! = z 1 − r − r ′ ∞ X n =1 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 1  r + r ′ − 2 n  z n + y w ( D )+ w ( D ′ ) ( y + y − 1 ) r + r ′ − 1  r + r ′ − 2 0  z 0 ! + z 1 − r − r ′ ∞ X n =1 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 1  r + r ′ − 2 n − 1  z n ! = z 1 − r − r ′ ∞ X n =1 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 1  r + r ′ − 1 n  z n + y w ( D )+ w ( D ′ ) ( y + y − 1 ) r + r ′ − 1 ! = z 1 − r − r ′ ∞ X n =0 y w ( D )+ w ( D ′ ) ( − 1) n ( y + y − 1 ) r + r ′ − n − 1  r + r ′ − 1 n  z n = L D ⊔ D ′ . The general case of the disjoin t union form ula is pro ved by the same induction on cd ( D ) as in the connected-sum case ab o ve. □ 5. Uniqueness 5.1. Uniqueness under regular isotopy . The uniqueness for three-term skein relations is w ell known (see, e.g., [2]). In contrast, the case of four-term skein relations, esp ecially for in v ariants under regular isotopy , requires a more careful analysis. F or am bient isotopy inv ariants, the argumen t proceeds in parallel with the classical case. On the other hand, for regular isotopy in v ariants, the b eha v- ior under Reidemeister mov es of type I in tro duces additional complications, and the inductiv e argument must be handled more carefully . In this section, we study the uniqueness in the four-term skein setting, b oth for regular and ambien t iso- top y . While our argument follows the general strategy of the three-term case, it incorp orates essential mo difications reflecting the four-term structure. As a conse- quence, w e sho w that the construction dev elop ed abov e, in particular the coefficient p olynomials α n , directly leads to a reconstruction of the Kauffman p olynomial. B-TYPE COEFFICIENT POL YNOMIAL 17 Lemma 5.1. L et R b e a c ommutative ring c ontaining Z [ y ± 1 ] and in which a, b, c, d ∈ R ar e invertible. L et f and f ′ b e two link invariants with values in R satisfying the same four-term skein r elation a f ( L + ) + b f ( L − ) + c f ( L ∞ ) + d f ( L 0 ) = 0 (9) for every skein quadruple ( L + , L − , L ∞ , L 0 ) , to gether with the normalization f ( O ) = f ′ ( O ) (10) for the zer o-cr ossing trivial knot diagr am O , and the kink r elations f ( ) = y f ( ) , f ( ) = y − 1 f ( ) (11) (and the same for f ′ ). Then f ≡ f ′ as invariants under r e gular isotopy. Pr o of. Let F = f − f ′ . It suffices to pro ve F ( L ) = 0 for every link L . Step 1: zer o-cr ossing unlink diagr ams. F rom (10) we ha ve F ( O ) = 0. Using (9) together with the kink relations (11), one can express F ( O r +1 ) linearly in terms of F ( O r ). (Concretely , apply the skein relation to a one-crossing diagram whose smo othings are O r and O r +1 ; since c and d are inv ertible in R , one can solve for the v alue on one smo othing in terms of the other.) Hence, by induction on r ≥ 1, w e obtain F ( O r ) = 0 for all zero-crossing unlink diagrams O r . If a diagram D of a trivial link is connected to O r b y a finite sequence of Reidemeis- ter mov es, then the kink relations together with regular isotopy inv ariance imply F ( D ) = 0. Hence F ( L ) = 0 for every trivial link L . Step 2: minimal c ounter example. Assume for contradiction that there exists a non trivial link L with F ( L )  = 0. Cho ose suc h an L with minimal crossing n umber c ( L ), and fix a diagram D realizing c ( D ) = c ( L ). It is standard that by changing crossings of D finitely many times one obtains a diagram of a trivial link. Let D = D (0) , D (1) , . . . , D ( r ) b e a sequence of diagrams obtained from D by successive crossing changes, so that D ( r ) represen ts a trivial link. Let L ( i ) b e the link represen ted b y D ( i ) . By construction, c ( D ( i ) ) = c ( D ) = c ( L ) for all i. Fix i and apply the skein relation (9) at the crossing that is changed from D ( i ) to D ( i +1) . The tw o smoothing diagrams hav e strictly few er crossings: c  ( D ( i ) ) ∞  , c  ( D ( i ) ) 0  ≤ c ( D ( i ) ) − 1 = c ( L ) − 1 . Therefore the corresp onding links hav e crossing n umber < c ( L ), and by the mini- malit y of c ( L ) w e m ust ha ve F  ( L ( i ) ) ∞  = F  ( L ( i ) ) 0  = 0 . Since a and b are in v ertible in R , the sk ein relation reduces to an equiv alence F ( L ( i ) ) = 0 ⇐ ⇒ F ( L ( i +1) ) = 0 , and hence F ( L ( i ) )  = 0 ⇐ ⇒ F ( L ( i +1) )  = 0 . 18 NOBORU ITO AND MA YUKO KON Consequen tly F ( L ( r ) )  = 0. But L ( r ) is a trivial link, so Step 1 gives F ( L ( r ) ) = 0. This con tradiction pro ves F ≡ 0, hence f ≡ f ′ . □ R emark 3 . Lemma 5.1 applies to regular isotopy in v ariants L D ( y , z ) ∈ R [[ z ]] (in particular R = Z [ y ± 1 ][[ z ]]). Moreov er, coefficient extraction [ z n ] is an R -linear map on R [[ z ]], so the same uniqueness applies co efficien t wise: if L D = L ′ D , then α n ( D ; y ) = α ′ n ( D ; y ) for all n ≥ 0. 5.2. Uniqueness under am bien t isotopy. Prop osition 5.2. L et L D ( y , z ) b e a r e gular isotopy invariant taking values in R [[ z ]] and satisfying the kink r elations (11) . Define the writhe-normalize d invariant F D ( y , z ) := y − w ( D ) L D ( y , z ) for oriente d diagr ams D . Then F D is invariant under ambient isotopy. Mor e over, if L and L ′ satisfy the same skein r elation and the same normalization on the unknot, then their writhe-normalizations c oincide under ambient isotopy. Pr o of. Reidemeister mo ves of t yp e I I and type I I I do not c hange the writhe, so y − w ( D ) is unchanged and the regular isotop y inv ariance of L D implies inv ariance of F D under t yp e I I and type I I I. Under a t yp e I mo ve, w ( D ) c hanges b y ± 1 and L D c hanges by a factor of y ± 1 b y (11), so the pro duct y − w ( D ) L D is unchanged. Hence F D is an ambien t isotop y in v arian t. F or the uniqueness statemen t, Lemma 5.1 implies L ≡ L ′ under regular isotop y . Applying the same writhe normalization to both yields F ≡ F ′ under ambien t isotop y . □ Assume that y = 1, w e ha ve Corollary 5.3. Corollary 5.3. L et f and f ′ b e two link invariants with values in a c ommutative ring R such that a f ( L + ) + b f ( L − ) + c f ( L ∞ ) + d f ( L 0 ) = 0 for every skein quadruple ( L + , L − , L ∞ , L 0 ) , wher e a, b, c, d ∈ R ar e invertible. If f ( U ) = f ′ ( U ) for the unknot U , then f ≡ f ′ . Pr o of. Let F = f − f ′ . It suffices to sho w that F ( L ) = 0 for every link L . Since f ( U ) = f ′ ( U ), we ha ve F ( U ) = 0. Let L + and L − b e the knots repre- sen ted b y the p ositive and negativ e one-crossing knot diagrams, let L 0 b e the link represen ted b y the zero-crossing 2-component diagram, and let L ∞ b e the knot rep- resen ted by the zero-crossing knot diagram. Since L + , L − , and L ∞ all represent the unknot U , w e obtain F ( L 0 ) = − c − 1 ( a + b + d ) F ( U ) = 0 . Rep eating the same argument inductively , w e obtain F ( U ℓ ) = 0 for every trivial link U ℓ . Assume for contradiction that there exists a link L such that F ( L )  = 0. Cho ose suc h an L with minimal crossing num b er c ( L ), and fix a diagram D realizing c ( D ) = c ( L ). B-TYPE COEFFICIENT POL YNOMIAL 19 By changing crossings of D finitely many times, we obtain a diagram of a trivial link. Let D = D (0) , D (1) , . . . , D ( r ) b e the sequence of diagrams obtained by successive crossing changes, and let L ( i ) b e the link represented by D ( i ) . Then L ( r ) is a trivial link. F or eac h i , applying the sk ein relation at the crossing changed from D ( i ) to D ( i +1) , w e obtain a F ( L ( i ) ) + b F ( L ( i +1) ) + c F (( L ( i ) ) ∞ ) + d F (( L ( i ) ) 0 ) = 0 . Since smoothing reduces the crossing num b er, c (( L ( i ) ) ∞ ) < c ( L ) , c (( L ( i ) ) 0 ) < c ( L ) , and b y minimalit y of c ( L ) w e hav e F (( L ( i ) ) ∞ ) = F (( L ( i ) ) 0 ) = 0 . Hence F ( L ( i +1) ) = − b − 1 a F ( L ( i ) ) , so in particular F ( L ( i ) )  = 0 ⇐ ⇒ F ( L ( i +1) )  = 0 . Therefore F ( L ( r ) )  = 0. But L ( r ) is a trivial link, so F ( L ( r ) ) = 0, a con tradiction. Th us F ≡ 0, and hence f ≡ f ′ . □ 5.3. A direct uniqueness pro of for the co efficien t p olynomials α n . R emark 4 . The uniqueness of the coefficient polynomials α n ( D ; y ) can also be sho wn directly from the inductiv e construction based on the complexit y cd ( D ) = ( c ( D ) , d a ( D )), without inv oking Lemma 5.1. W e include a pro of sketc h for com- pleteness. Sketch of dir e ct pr o of. First determine α n ( O r ; y ) for trivial link diagrams O r . Us- ing the skein relation together with the normalization α n ( O ; y ) = δ n, 0 , one obtains the recurrence ( y + y − 1 ) α n ( O r ; y ) = α n ( O r +1 ; y ) + α n − 1 ( O r ; y ) . This uniquely determines α n ( O r ; y ) = ( ( − 1) n  r − 1 n  ( y + y − 1 ) r − n − 1 , n ≥ 0 , 0 , n < 0 . Next pro ceed by induction on cd ( D ) = ( k , s ). If s = 0, then D is monotone and the ab o ve form ula determines α n ( D ; y ) uniquely . Assume s ≥ 1 and that α n is uniquely determined for diagrams with smaller cd . Let p b e a w arping crossing of ( D, a ). Then α n ( D ; y ) = − α n ( D − ϵ ( p ) ; y ) + α n +∆( p 1) − 1 ( D p 1 ; y ) + α n +∆( p 2) − 1 ( D p 2 ; y ) , and each term on the righ t-hand side has strictly smaller cd . By induction, each term is uniquely determined, hence so is α n ( D ; y ). □ Com bining Remark 4 with Lemma 4.6, w e obtain the follo wing uniqueness state- men t. 20 NOBORU ITO AND MA YUKO KON Prop osition 5.4. We c an uniquely c ompute α n ( D ; y ) ( n = 0 , ± 1 , . . . ) using (1)– (3) in The or em 3.1. In p articular, α n ( D ; y ) = 0 exc ept for finitely many n , and α n ( D ; y ) = 0 for n < 0 . 6. Discussion W e ha ve established the existence of B-t yp e co efficient p olynomials α n ( D ; y ) within the framework of formal p o wer series, extending Kaw auchi’s A-type con- struction to the intrinsically four-term skein setting. The esse n tial new phenome- non in the B-t yp e case is that the inductiv e closure cannot be ac hiev ed within the class of monotone diagrams alone. The necessary enlargement to connected sums of monotone diagrams reflects a genuine structural difference betw een three-term and four-term skein theories. F rom this viewp oint, the generating series of the B-type coefficient p olynomials should b e regarded not merely as a reconstruction of the Kauffman p olynomial [3], but as a co efficien t-lev el refinement of the four-term skein structure. Eac h coeffi- cien t p olynomial defines a link in v arian t, yielding an infinite sequence of in v arian ts asso ciated naturally with the B-t yp e skein relation. The present pap er focuses on the construction and well-definedness of this theory . F urther inv estigations, including structural comparisons with mutation phenomena and in teractions with random knot models, will b e pursued elsewhere. References 1. P . F reyd, D. Y etter, J. Hoste, W. B. R. Lick orish, K. Millett, and A. Ocnean u, A new p olynomial invariant of knots and links , Bulletin of the American Mathematical So ciet y 12 (1985), no. 2, 239–246. 2. Nob oru Ito, Jun Y oshida, and Keita Nak agane, Quantum Knot Homology , CRC Press, to appear. 3. Louis H. Kauffman, An invariant of r egular isotopy , T rans. Amer. Math. So c. 318 (1990), no. 2, 417–471. MR 958895 4. Akio Kaw auchi, On c o efficient p olynomials of the skein p olynomial of an oriente d link , Kob e J. Math. 11 (1994), no. 1, 49–68. MR 1309992 5. J´ ozef H. Przytyc ki and Pa we lT raczyk, Invariants of links of c onway typ e , Kobe Journal of Mathematics 4 (1987), 115–139. Dep ar tment of Ma thema tics, F acul ty of Engineering, Shinshu University, W akasa to 4-17-1 Nagano, Nagano, 380-8553, Jap an Email addr ess : nito@shinshu-u.ac.jp F acul ty of Educa tion, Shinshu University, 6-Ro, Nishinagano, Nagano City 380-8544, Jap an Email addr ess : mayuko k@shinshu-u.ac.jp