Anda di halaman 1dari 39

FUSION IN THE ENTWINED CATEGORY OF YETTERDRINFELD MODULES OF A RANK-1 NICHOLS ALGEBRA

A.M. SEMIKHATOV

arXiv:1109.5919v2 [math.QA] 23 Nov 2011

A BSTRACT. We rederive a popular nonsemisimple fusion algebra in the braided context, from a Nichols algebra. Together with the decomposition that we nd for the product of simple YetterDrinfeld modules, this strongly suggests that the relevant Nichols algebra furnishes an equivalence with the triplet W -algebra in the p, 1 logarithmic models of conformal eld theory. For this, the category of YetterDrinfeld modules is to be regarded as an entwined category (the one with monodromy, but not with braiding).

1. I NTRODUCTION The idea to construct purely algebraic counterparts of vertex-operator algebras (conformal eld theories) has a relatively long history [1, 2, 3, 4, 5, 6, 7]. In [8, 9, 10, 11, 12, 13, 14, 15], it was developed for logarithmic CFT models, which have been intensively studied recently (see [16, 17, 18, 19, 20, 21, 22, 23] and the references therein). In [24], further, a braided and arguably more fundamental algebraic counterpart of logarithmic CFT was proposed. It is given by Nichols algebras [25, 26, 27, 28, 29]; the impressive recent progress in their theory (see [30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40] and the references therein) is a remarkable spin-off of Andruskiewitsch and Schneiders program of classication of pointed Hopf algebras. Associating Nichols algebras with CFT models implies that certain CFT-related structures must be reproducible from (some) Nichols algebras. Here, we take the simplest, rank-1 Nichols algebra B p of dimension p 2 and, from the category of its Yetter Drinfeld modules, extract a commutative associative 2p-dimensional algebra on the xr , 1 r p, Z2 : (1.1) xr1 1 xr2 2
p 1 s

r2 p r1
r1 r2 1 step 2

xs1 2

p s 2p r1 r2 1 step 2


r p,

ps1 2 ,

with pr

x p , r p. This is the FHST fusion algebra [41] (also see [13]), which makes part of what we know from [12] (also see [42]) to be an equivalence of representation categoriesof the triplet algebra W p in the p, 1 logarithmic conformal models [43, 44, 45, 46, 41] and of a

2 xr 2 x p r 1 ,

SEMIKHATOV

small quantum s2 at the 2pth root of unity, proposed in this capacity in [8, 9] and then used and studied, in particular, in [47, 48, 49, 50, 51] (it had appeared before in [52, 53, 54]). The reoccurrence of the fusion algebra in the braided approach advocated in [24], together with some other observations, supports the idea that Nichols algebras are at least as good as the quantum groups proposed previously [8, 9, 10, 11, 55] for the logarithmic version of the KazhdanLusztig correspondence (the correspondence between vertex-operator algebras and quantum groups).1 Algebra (1.1) arises here as an algebra in the center of the category of YetterDrinfeld B p -modules; the xr are certain images of the simple YetterDrinfeld B p -modules Xr .2 More is actually true: from the study of the representation theory of B p , we obtain that the product of simple YetterDrinfeld B p modules decomposes as (1.2) Xr1 1 Xr2 2
p 1 s

r1 r2 p
r1 r2 1 step 2

Xs1 2

p s 2p r1 r2 1 step 2

Ps12 ,

where P p X p and Pr for 1 r module with the structure of subquotients (1.3) Pr Xr

p 1 is a reducible YetterDrinfeld B p

X p r 1 Xr 2 , X p r 1

Decompositions (1.2) were conjectured in [24] and are proved here. The Xr and Pr do not exhaust all the category of YetterDrinfeld B p -modules, but make up the most signicant part of it, and relations (1.2), together with the structure of Pr , already seem to imply that the category of YetterDrinfeld B p modules is equivalent to the W p representation category. This requires an important clarication, however. In the braided category of YetterDrinfeld B p modules, the simple objects are the Xr labeled by 1 r p and Z4 (and, accordingly, Z4 in Pr , and so on). There are twice as many objects as in the category of W p representations [41, 42, 12]. But the presumed equivalence is maintained for entwined categories [56]those endowed with only double braiding DY,Z cZ,Y cY,Z (the monodromy on the W p side). The properties of double braiding can be axiomatized without having to resort to the braid1But the actual motivation in [24], which is yet to be tested on more advanced examples, was that Nichols

algebras can actually do better than the old quantum groups. 2The notation is fully explained below, but here we note that the module comodule structure, e.g., of Xr depends only on r, whereas serves to distinguish isomorphic module comodules that nevertheless have different braiding.

FUSION FOR A RANK-1 NICHOLS ALGEBRA

ing itself [56]. This denes a twine structure and, accordingly, an entwined category. Remarkably, it was noted in [56] that many signicant notions apparently related to c actually depend only on D or [the twist] . The S-matrix, and the subcategory of transparent objects, which play an important role in the construction of invariants of 3-manifolds, are dened purely in terms of the double braiding. More surprisingly, the invariants of ribbon links . . . do not depend on the actual braiding, but only on D. In the entwined category of YetterDrinfeld B p modules, the objects with and 2 in their labels are isomorphic, which sets Z2 and resolves the representation doubling problem; everything else on the algebraic side appears to be already ne-tuned to ensure the equivalence. (We do not go as far as modular transformations in this paper, but the above quotation suggests that dealing with entwined categories is not an impediment to rederiving the W p modular properties at the Nichols algebra level, in a braided version of what was done in [8].) It may also be worth noting that we derive (1.1) and (1.2) independently (of course, from the same structural results on YetterDrinfeld B p -modules, but not from one another). In particular, (1.1) is obtained by directly composing the action of xr1 1 and xr2 2 on YetterDrinfeld modules, with xr : Y Y given by running Xr along the loop in the diagram (with the notation to be detailed in what follows)

B2

(1.4)

As such, the xr depend only on

Z2 there is no Z4 option for them.3

This paper is organized as follows. For the convenience of the reader, we summarize the relevant points from [24] in Sec. 2; a very brief summary is that for a Nichols algebra BX , a category of its YetterDrinfeld modules can be constructed using anloop, the braiding itself (and the ribbon map ). This does not affect the statement of the equivalence of entwined categories, but rather suggests exploring a further possibility, elaborating on the fact that the braiding of a YetterDrinfeld B p module with itself and with its dual also depends on Z2 , not Z4 (and the same for the ribbon map). An entwined category might allow these braidings in addition to twines. This is similar to the idea of twist equivalence in the theory of Nichols algebras [28] (the similarity is not necessarily supercial if we recall that the braiding of bare vertex operators is diagonal for B p ).
3Diagram (1.4) involves not only the squared braiding B2 of YetterDrinfeld modules but also, in the

SEMIKHATOV

other braided vector space Y (whose elements are here called vertices, and the Yetter Drinfeld modules the multivertex modules). In Sec. 3, we introduce duality and the related assumptions that make it possible to write diagrams (1.4). In Sec. 4, everything is specialized to a rank-1 Nichols algebra B p (depending on an integer p 2). First and foremost, everything includes multivertex YetterDrinfeld modules. We actually construct important classes of these modules quite explicitly (Appendix B), which allows proving (1.2) and also establishing duality relations among the modules. We also study their braiding, nd the ribbon structure, and nally use all this to derive (1.1) from (1.4) for B p . Basic properties of YetterDrinfeld modules over a braided Hopf algebra are recalled in Appendix A. Some technicalities used in calculations are collected in Appendix C. 2. T HE N ICHOLS
ALGEBRA OF SCREENINGS

We summarize the relevant points of [24] in this section. Screenings and BX . The underlying idea is that the nonlocalities associated with screening operatorsmultiple-integration contours, such as (2.1)

where s j z are the screening currents allow introducing a coproduct by contour cutting, called deconcatenation in what follows: (2.2) :

si1 z1 si2 z2 si3 z3 ,

z1 z2 z3

(with the line cutting symbol subsequently understood as ). A product of lines populated with crosses is also dened, as the quantum shufe product [57], which involves a braiding between any two screenings. It is well known that these three structures coproduct, product, and braidingsatisfy the braided bialgebra axioms [57]. The antipode is in addition given by contour reversal. The braided Hopf algebra axioms are then satised for quite a general braiding (by far more general than may be needed in CFT); it is rather amusing to see how the braided Hopf algebra axioms are satised by merging and cutting contour integrals [24]. The algebra generated by single crossesindividual screeningsis the Nichols algebra BX of the braided vector space X spanned by the different screening species (whose number is called the rank of the Nichols algebra). Nichols algebras. The Nichols algebrasbialgebras of type one in [25]are a crucial element in a classication program of ordinary Hopf algebras of a certain type (see [26, 28, 27, 33] and the references therein). Nichols algebras have several denitions, whose equivalence is due to [58] and [26]. The Nichols algebra BX of a braided linear space

FUSION FOR A RANK-1 NICHOLS ALGEBRA

n such X can be characterized as a graded braided Hopf algebra BX n 0 BX that BX 1 X and this last space coincides with the space of all primitive elements PX x BX x x 1 1 x and it generates all of BX as an algebra.4
Nichols algebras occurred independently in [62], in constructing a quantum differential calculus, as fully braided generalizations of symmetric algebras, BX k X
r 2

ker Sr ,

where Sr is the total braided symmetrizer (braided factorial). The space of vertices Y . In addition to the braided linear space X spanned by the different screening species, we introduce the space of vertex operators taken at a xed point, (2.3) Y SpanV 0, where ranges over the different primary elds in a given CFT model. CFT also yields X X of any two screenings (which is always applied to two the braiding : X X Y X and screenings on the same line, as in (2.1)), as well as the braiding : X Y :Y X X Y of a screening and vertex (also on the same line, as in (2.4) below), and eventually the braiding : Y Y Y Y of any two vertices, but a large part of our construction can be formulated without this last. The two braided vector spaces X and Y are all that we need in this section; the braiding can be entirely general. Dressed vertex operators as BX -modules. We use the space Y to construct BX modules. Their elements are sometimes referred to in CFT as dressedscreened vertex operators, for example, (2.4)

It is understood that the and are decorated with the appropriate indices read off from the right-hand side; but it is in fact quite useful to suppress the indices altogether and let and respectively denote the entire spaces X and Y , and we assume this in what follows. Because the integrations can be taken both on the left and on the right of the vertex position, the resulting modules are actually BX bimodules. The left and right actions of BX are by pushing the new crosses into the different positions using braiding; the left action, for example, can be visualized as

si1 x1 si2 x2 V 0

si3 x3 .

x1 x2 0

0 x3

4An important technicality, noted in [59, 60], is a distinction between quantum symmetric algebras [61]

and Nichols algebras proper; the latter are selected by the condition that the braiding be rigid, which in particular guarantees that the duals X are objects in the same braided category with the X.

SEMIKHATOV


L :

Y
r r

where the arrows, somewhat conventionally, represent the braiding . Once again by deconcatenation, e.g.,

these bimodules are also bicomodules and, in fact, Hopf bimodules over BX (see [63, 64, 65, 66] for the general denitions). Braid group diagrams and quantum shufes. A standard graphical representation for the multiplication in BX and its action on its modules is in terms of braid group diagrams. For example, the above left action is represented as (to be read from top down)

(2.5)

id 1 21 X

X ,

where we use the leg notation, in the right-hand side, letting i denote the braiding of the ith and i 1th factors in a tensor product (our notation and conventions are the same as in [24]). The braid group algebra element X1,2 id 1 2 1 occurring here is an example of quantum shufes. The product in BX is in fact the shufe product (2.6) Xr,s : X
r

rs

on each graded subspace. The antipode restricted to each X r is up to a sign given by the half-twist the braid group element obtained via the Matsumoto section from the longest element in the symmetric group: (2.7)

Sr

1r 1 21 3 21 . . . r1r2 . . . 1 : X

(with the brackets inserted to highlight the structure, and the sign inherited from reversing the integrations); for example,

S5

r,s 0

The Hopf bimodules alluded to above are (some subspaces in)

s,

with

FUSION FOR A RANK-1 NICHOLS ALGEBRA

the left and right BX actions on these also expressed in terms of quantum shufes as Xr,s1t : X and Xs1t,r : X

s r

X
i 0

sri Y X t i sri Y X t i.

r i 0

Hopf-algebra diagrams. The four operations on bi(co)modules of a braided Hopf algebra B are standardly expressed as

which are respectively the left module structure B Z Z, the left comodule Z B Z, the right module structure Z B Z, and the right comodule structures Z Z B. The product and coproduct in the braided Hopf algebra itself are denoted as and The . braiding is still denoted as , but in contrast to the braid-group diagrams, each line now represents a copy of B or a B (co)module. Adjoint action and YetterDrinfeld modules. The left and right actions of a braided Hopf algebra B on its Hopf bimodule Z give rise to the left adjoint action B Z Z:

(2.8)

A fundamental fact is that the space of right coinvariants in a Hopf bimodule is invariant under the left adjoint action; this actually leads to an equivalence of categories, the category of Hopf bimodules and the category of YetterDrinfeld modules [63, 64, 67, 62]. We recall some relevant facts about YetterDrinfeld modules in Appendix A. In our case of modules spanned by dressed vertex operators, the right coinvariantsall those y that map as y y 1 under the right coactionare simply the vertex operators dressed by screenings only from the left, i.e., elements of X r Y , for example, . In terms of braid group diagrams (with the lines representing the X and Y spaces), an example of the left adjoint action on such spaces is given by

(2.9)

where a single new cross arrives to each of the three possible positions in two ways,

SEMIKHATOV

one with the plus and the other with the minus sign in front (which is something expected of an adjoint action). That the cross never stays to the right of is precisely a manifestation of the above invariance statement for the space of right coinvariants. This means that a number of terms that follow when expressing (2.8) in terms of braid group diagrams cancel. The left adjoint action (2.8) can in fact be expressed more economically as follows.

We dene a modied left action of BX on its Hopf bimodules spanned by dressed vertex operators by allowing the new crosses to arrive only to the left of , for example,

(2.10)

(more crosses might be initially placed to the right of the vertex ; the action does not see them). In general, is the map (2.11) Similarly, a modied right action the space of right coinvariants is dened by rst on letting the new cross to be braided with the vertex and then shufing into all possible positions relative to the old crosses:

Xr,s : X

rs Y.

which in general is (2.12)


where s,r is the braiding of an s-fold tensor product with an r-fold tensor product. The and actions preserve the spaces of right coinvariants and commute with each other. The economic expression for adjoint action (2.8) is [24]

Xs,r id

1,r : X

sr Y,

(2.13)

This diagram is the map (2.14)

r,s

i 0

s Xri,si Xs,i 1,i S i

s1

i,s 1

ri

:X

sr Y.

FUSION FOR A RANK-1 NICHOLS ALGEBRA

Multivertex YetterDrinfeld modules. More general, multivertex, YetterDrinfeld BX modules can be constructed by letting two or more vertices (the Y spaces) sit on the same line, e.g., (2.15)

or

These diagrams respectively represent X Y X 3 Y and X Y X 2 Y X Y (in general, different spaces could be taken instead of copies of the same Y , but in our setting they are all the same). By denition, the BX action and coaction on these are (2.16.1) the cumulative left adjoint action, and (2.16.2) deconcatenation up to the rst . The cumulative adjoint means that all the except the rightmost one are viewed on equal footing with the under this action: the adjoint action of X r on the space X s Y X t Y in a two-vertex module is given by r,s1t . For example, the left adjoint is given by the braid group diagrams that are action exactly those in the right-hand side of (2.9), with the corresponding strand representing not X but Y . The BX coaction by deconcatenation up to the rst vertex means, for example, that at most one can be deconcatenated in each diagram in (2.15). For multivertex YetterDrinfeld modules, the form (2.13) of the adjoint action is valid if is understood as the cumulative action preserving right coinvariants; for example,

is given just by the braid group diagrams in the right-hand side of (2.10) with the second strand representing not X but Y . Fusion product. The multivertex YetterDrinfeld modules are not exactly tensor products of single-vertex onesthey carry a different action, which is not Y Z , and the coaction is not diagonal either. They actually follow via a fusion product [24], which is dened on two single-vertex YetterDrinfeld modules (each of which is the space of right coinvariants in a Hopf bimodule) as (2.17) which is the map
t j 0

Xs, j 1, j : X
s

Y 2 and t

on each s,t component. For example, if s can be represented as

3, the top of the above diagram

10

SEMIKHATOV

and then in view of the denition of the meaning of (2.17) is that j 0 crosses , from the right factor are detached from their native module and sent to mix with the left crosses (the sum over j is taken in accordance with the denition of the coaction). The construction extends by taking the fusion product of multivertex modules: the coaction in (2.17) is then the one just described, by deconcatenation up to the rst vertex, action on a multivertex module is cumulative, i.e., each cross acting from and the the right, e.g., on , arrives at each of the ve possible positions. 3. D UALITY
IN THE CATEGORY OF

Y ETTER D RINFELD

MODULES

We now consider duality in a braided category of representations of a braided Hopf algebra B. We briey recall the standard denitions and basic properties, and then assume that duality exists in the setting of the preceding section; this then allows us to construct endomorphisms of the identity functor in Sec. 4. 3.1. For a B-module Z, we let Z denote the left dual module in the same (rigid) braided category. The duality means that there are coevaluation and evaluation maps

Z Z Z

and

which are morphisms in the category and satisfy the axioms


and

where the two straight lines are id

and idZ . It follows that


and similarly for the coevaluation.

The dual Z to a leftleft YetterDrinfeld B-module Z is a leftleft YetterDrinfeld B-module with the action and coaction, temporarily denoted by and , dened as [64]
S

(3.1)

and

S 1

FUSION FOR A RANK-1 NICHOLS ALGEBRA

11

The denitions are equivalent to the properties (which, inter alia, imply that the evaluation is a B module comodule morphism)
S

(3.2)

We prove the YetterDrinfeld property for for and for completeness. In view of (3.2), it is easiest to verify the YetterDrinfeld axiom by establishing that

and

S 1

(3.3)

Pushing the new action and then the coaction to the other side, we see that the left-hand side of (3.3), by the above properties, is equal to
S S S

S 1

In the rst diagram, we insert S at the position of the upper checkmark and S same line, at the lower checkmark, and use the properties of the antipode,
S S S
S 1S 1

S 1

S 1

S S 1

1 into the

and

This readily gives the second diagram above, where we further recognize the right-hand side of the YetterDrinfeld axiom assumed for the module. After using it (the third diagram), and after another application of the properties of S and S 1 , we obtain the fourth diagram, and it is immediate to see that it coincides with the right-hand side of (3.3) also rewritten by pushing and to the other side.

S 1

12

SEMIKHATOV

3.2. Assuming a rigid category. We further assume that the category of n-vertex Yetter Drinfeld B-modules is rigid; this means that the dual modules are modules in the same categoryin our case, multivertex YetterDrinfeld B-modules, and the action and coaction dened in (3.1) are just those in (2.16)and hence the evaluation map satises the properties
S

(3.4)

for any pair of YetterDrinfeld B-modules. Evidently, we then also have


S S 1

and

S 1

(3.5)

and

Our assumption. Dealing with multivertex YetterDrinfeld modules realized as right coinvariants in Hopf bimodules, we actually adopt a stronger assumption that implies the rst diagram in (3.4): we assume that
S2

(3.6)

hold for the left and right actions in (2.11) and (2.12). Then also
S2

and

and

3.3. If the category BYD of YetterDrinfeld B-modules is rigid, then for each Z B BYD, there is a morphism : Y Y for any Y BYD, dened as Z B B
Y

B2

(3.7)

B2

FUSION FOR A RANK-1 NICHOLS ALGEBRA

13

where B is dened in (A.3) and is any B module comodule morphism. In the second , diagram, the squared relative antipode dened in (C.5), occurs in view of (3.4).

That the map dened by (3.7) is a B module comodule morphism follows from the general argument that so are B, evaluation, and coevaluation (and ). It is also instructive
for for brevity):

to see this by diagram manipulation (temporarily writing



(3.7)

In the rst equality, we use only the YetterDrinfeld axiom, with B2 represented by the rst diagram for B2 in (A.3); the associativity of action was used in the second equality above; another use of the associativity in the lower part of the third diagram allows recognizing the left-hand side of (A.1); the YetterDrinfeld property is then applied in the third equality together with the rst property in (3.5), yielding the fourth diagram; there, we use the rst properties in (C.6) and (3.4) to obtain the last, fth diagram, where an antipode bubble is annihilated, showing that, indeed,

(3.7)


(3.7)

The commutativity of (3.7) with coaction can be veried similarly. 3.4. Ribbon structure. A ribbon structure is a morphism : Y such that (3.8)

B2

Y for every object Y

14

SEMIKHATOV

Whenever it exists, choosing in (3.7) makes Z multiplicative in Z. To show this, we calculate W Z Y by sliding one of the diagrams along the Y line into the middle of the other and then expanding:

B2

(3.9)
Y W B2

S B2

In the last diagram, we recognize the diagonal coaction (the two ) and action (two just below the respective checkmarks) on a tensor product of two YetterDrinfeld modules, as in (A.2). In the bottom right part of the diagram, we recall that
and

calculate

B2

where the rst three equalities are elementary (and well-known) rearrangements, the fourth involves (3.4), and the checked equality is veried by repeatedly applying the YetterDrinfeld axiom in its right-hand side. The sixth diagram involves B2 in the upper part and the diagonal action and coaction (A.2) in the lower part, which gives the last

FUSION FOR A RANK-1 NICHOLS ALGEBRA

15

equality. We therefore conclude that if (3.8) holds, then

B2

Substituting this in (3.9) shows that is indeed multiplicative: W Z Y 4. R ANK - ONE N ICHOLS
ALGEBRA

Y.

We specialize the preceding sections to the case of a rank-one Nichols algebra B p , whose relation to the p, 1 logarithmic CFT models was emphasized in [24]. An integer p 2 is xed throughout. Notation. We x the primitive 2pth root of unity q and introduce the q-binomial coefcients
r s

ep

r! , r! s! r s!

1 . . . r, r
q.

q2r 1 , q2 1

which are assumed to be specialized to q We sometimes use the notation aN

a mod N 0, 1, . . ., N 1.

4.1. The braided Hopf algebra B p . Taking the Nichols algebra to have rank 1 means that the braided linear space X is one-dimensional. We x an element F (a single screening in the CFT language) as a basis in X . We can then think of as just F, and write (r crosses). F r The braiding, taken from CFT, is (4.1) F r F s q2rs F s
rs r

F r,

and shufe product (2.6) then becomes F r F s and coproduct (2.2) is : F r


r s 0

F r s,

F s

F r s. The antipode dened in (2.7) acts

16

SEMIKHATOV

as S F r

generated by a single element F, such that F r write F F 1.

The algebra B p is the linear span of F r with 0

1r qrr1F r.

1 r r! F ,

p 1. It can also be viewed as r p 1, with F p 0. We

4.2. YetterDrinfeld B p -modules. We apply the construction in Sec. 2 of BX modules using another braided vector space Y , a linear span of vertex operators present in the relevant CFT model. 4.2.1. The vertices. For the p, 1 models corresponding to B p (see [41]), Y is a 2pdimensional space Y spanV a a Z4p with the diagonal braiding (4.2) and with (4.3) V a F r qar F r V a , F r V a qar V a F r. V a V b q 2 Vb
ab

V a.

This sufces for calculating the cumulative adjoint B p action on multivertex Yetter Drinfeld modules, as we describe next. In what follows, the integers a, b, . . . are tacitly considered modulo 4p. 4.2.2. Multivertex YetterDrinfeld B p modules. We saw in Sec. 2 that multivertex YetterDrinfeld modules (see (2.15) and (2.16)) can be represented as an essentially combinatorial construction for the crosses to populate, in accordance with the braiding rules, line segments that are separated from one another by vertex operators, e.g., , where X and Y (for a nite-dimensional Nichols algebra, each segment can carry only nitely many crosses). In the rank-1 case, each cross can be considered to represent the F element, and each segment is fully described just by the number of the Fs sitting there. For example, each two-vertex YetterDrinfeld module is a linear span of

V a and V b . Because the braiding is diagonal, there is a B p module comodule for each xed a and b (and c, . . . for multivertex modules). The simplest, one-vertex YetterDrinfeld B p modules are spanned by (4.5) Vsa

, where s and t must not exceed p 1 (s 2 and t 1 in the picture) and a and b indicate
(4.4)
a, Vs,t b a b

(s crosses),

FUSION FOR A RANK-1 NICHOLS ALGEBRA

17

where s ranges over a subset of 0, . . ., p 1. The B p coaction is by deconcatenation up to the rst vertex in all cases, i.e.,

Vsa
a, Vs,t b a, and similarly for Vs,t,b, c , and so on. u

s r 0 s r 0

F r Vsa r , F r Vsa, b , r,t

The B p action (which is the left adjoint action (2.8)) is then calculated as F Vsa

s as 1Vsa 1,

1 q2 ,

and the cumulative adjoint evaluates on multivertex spaces as


a, F Vs,t b a, F Vs,t,b, c u a, b s 2t a bs 1Vsa, b q2sa t bt 1Vs,t 1, 1,t a, b, c a, q2sa t 2u b ct 1Vs,t 1, u q2s2t ab u cu 1Vs,t,b,uc 1,

s 2t 2u a b cs 1Vsa, b, c u 1,t,

and so on. The braiding follows from (4.1), (4.2), and (4.3), for example, (4.6) Vsa Vtb q 2 a2sb2t Vtb
1

Vsa .

4.3. Module types and decomposition. We now study the category of YetterDrinfeld B p modules in some detail: we nd how the one-vertex and two-vertex spaces decompose into indecomposable YetterDrinfeld B p modules. We rst forget about braiding and study only the module comodule structure; the action and coaction are related by the YetterDrinfeld axiom, but we try to avoid speaking of YetterDrinfeld modules before we come to the braiding. 4.3.1. The relevant module comodules, which we construct explicitly in Appendix B, are as follows:

simple r-dimensional module comodules Xr, 1 r p; for r times use the special notation S p X p; the p-dimensional extensions (4.7) X p r Vr 1 r p 1, Xr where the arrow means that X p r the trivial piece B p B p Xr.

p, we some-

X p r

18

SEMIKHATOV

2p-dimensional indecomposable module comodules Pr with the structure of


subquotients (4.8) Pr Xr X p r 4.3.2. We also show in Appendix B that the p2 -dimensional one-vertex space V p 1 (4.9) SpanVsa 0 S p a, s p 1 Vr X p r Xr 1 r p 1.

decomposes into B p module comodules as V p 1


1 r p 1

and the p4 -dimensional two-vertex space V p 2 decomposes as (4.10) V p 2 p 2 S p


1 r p 1 a, SpanVs,t b 0

a, b, s,t

p 1

2r p rVr

1 r p 1

p r2Pr.

Multivertex spaces give rise to zigzag YetterDrinfeld modules, which we do not consider here. 4.3.3. Notation. Compared with representation theory of Lie algebras, the role of highesta, weight vectors is here played by left coinvariants V0a and V0,tb . When a module comodule of one of the above types A X, V, or P is constructed starting with a left coinvariant, a a, b we use the notation A0 or A0,t to indicate the coinvariant, and sometimes also use the

notation such as X0,t r to indicate the dimension (although it is uniquely dened by a, t, b, and the module type).

a, b

4.3.4. The module comodules that can be constructed starting with one-vertex coinvariants V0a are classied immediately, as we show in B.1. The module comodule generated from V0a under the B p action is isomorphic to Xr whenever a p r 1 (1 r p). If r p 1, then extension (4.7) follows immediately. 4.3.5. The strategy to classify two-vertex B p module comodules according to their chara, acteristic left coinvariant V0,tb is to consider the following cases that can occur under the action of F s on the left coinvariant.
a, (1) F s V0,tb is nonvanishing and not a coinvariant for all s, 1 case, there are the possibilities that

p 1. In this

FUSION FOR A RANK-1 NICHOLS ALGEBRA a, a, a, (a) F s V0,tb is a coinvariant, i.e., F s V0,tb const V0,tb s , for some s and a, (b) F s V0,tb is not a coinvariant for any s p 1.

19

p 1,

a, (2) F s V0,tb

a, (a) For some s s, F s V0,tb is a coinvariant, and (b) F s V a, b is not a coinvariant for any s s. We then distinguish the cases 0,t

0 for some s

p 1. In this case, further possibilities are

where a, (i) V0,tb is in the image of F, and


a, (ii) V0,tb is not in the image of F.

We show in Appendix B that these cases are resolved as follows in terms of the parameters a, t, and r a b 2t p 1:

1a: 1 r p 1 and either t a p r or a p 1 t p r 1. Then the left coinvariant is the leftmost coinvariant in (1.3), and the YetterDrinfeld module generated from it is the leftbottom half Lr of Pr (see B.2.3). 1b: r p. Then X p S p is generated from the left coinvariant. 2a: is not realized. 2(b)i: 1 r p 1 and either t p r a p 1 or p r t a p . Then the bottom YetterDrinfeld submodule Br in P p r is generated from the left coinvariant. 2(b)ii: 1 r p 1 and either t a p and a p r 1 t p 1 r or t a p 1 and p r t p r a p. Then Xr is generated from the left coinvariant.

4.4. Braiding sectors. The Xr and the other module comodules appearing above satisfy the YetterDrinfeld axiom. Considering them as YetterDrinfeld B p -modules means that isomorphic module comodules may be distinguished by the braiding. This is indeed a a, b the case: for example, shifting a a p in X0 or X0,t does not affect the module comodule structure described in Appendix B, but changes the braiding with elements of B p by a sign in accordance with (4.3). We thus have pairs A , A 1 , Z2 , of isomorphic module comodules distinguished by a sign occurring in their braiding. In particular, there are 2p nonisomorphic simple YetterDrinfeld modules. Further, these YetterDrinfeld modules can be viewed as elements of a braided category, whose braiding (see (A.3)) involves (4.6). The dependence on a in (4.6) is modulo 4p, and hence we have not pairs but quadruples A Z4 , with the different A distinguished by their braiding with other such modules. In particular, there are 4p nonisomorphic simple objects in this braided category of YetterDrinfeld B p -modules [24].

20

SEMIKHATOV

It is convenient to write a for simple modules, with X0

a p p, Z4 [24], and introduce the notation Xr


Xr whenever a r 1 p.

As before, r is the dimension, and we sometimes refer to as the braiding sector or braiding index. For Z4 , the isomorphisms are in the braided category of YetterDrinfeld B p -modules. The quadruple structure occurs totally similarly for other modules, including those realized in multivertex spaces; for example, for any a, b Z, we have the isomorphisms among the simple YetterDrinfeld modules realized in the two-vertex space (cf. (B.10)!): X0,t

a, b

Xr

whenever a b 2t

r 1 p and (B.8)

(B.9) holds.

For the reducible extensions as in (4.7), the two subquotients have adjacent braiding indices, and we conventionally use one of them in the notation for the reducible module: (4.11) Vr X p r 1 , Xr r 1 p and (B.8) or (B.9) holds.

and V0,t

a, b r

Vr whenever a b 2t

In the above formulas and diagrams, Z4 if the modules are viewed as objects of the braided category of YetterDrinfeld B p -modules. But if the YetterDrinfeld B p -modules are considered as an entwined category, then the braiding sectors and 2 become indistinguishable, and hence Z2 . In particular, there are 2p nonisomorphic simple objects in the entwined category of YetterDrinfeld B p -modules. 4.5. Proof of decomposition (1.2). Decomposition (1.2) can be derived from the list in 4.3.5 as follows. The fusion product (2.17) of two one-vertex modules is the map (assuming that a, b p 1 to avoid writing a p and b p ) (4.12) Vsa Vtb
b i 0

In (4.8), the relevant braiding indices range an interval of three values, and we use the leftmost value in the notation for the entire reducible YetterDrinfeld module, which a, b yields (1.3), with P0,t r Pr whenever a b 2t r 1 p and (B.19) holds.

qai

si s

Vsa, b i . i,t

In evaluating X0 s X0 t , this formula is applied for 0 s a and 0 t b. Then a, b the left coinvariants produced in the right-hand side are V0, u , where 0 u b and u a. But the conditions dening the different items in the list in 4.3.5 have the remarkable

FUSION FOR A RANK-1 NICHOLS ALGEBRA

21

property that the module A0,u generated from each such coinvariant is as follows:
Xa, b , 0, u a, b X 0, u , 0, u a, b S 0, u , a, b

a,b

ab ab ab

p 1, p and u

(4.13)

a,b A
0,u

a, b , a b

p and a b 2u p p and a b 2u p p and a b 2u p

a b p 2, 0,

This is established (only for 0 u a, b p 1) by direct inspection of each case in the a, b a, b a, b list at the end of 4.3.5. The module L0, u is the leftbottom half of P0, u , and B0, u is the bottom sub(co)module in another P module; the details are given in B.2.3. The a, b a, b crucial point is that each L0, u can be extended to P0, u (while the B, on the other hand, are not interesting in that they are sub(co)modules in the L that are already present). We a b next claim that each of the Ls occurring in X0 s X0 t indeed occurs there together with the entire P module; this follows from counting the dimensions and from the fact a, b that there are no more left coinvariants among the Vv, w appearing in the right-hand side of (4.12) (and, of course, from the structure of the modules described in Appendix B). Once it is established that each L occurs in (4.13) as a sub(co)module of the corresponding P, it is immediate to see that (4.13) is equivalent to (1.2). 4.6. Duality. We now recall Sec. 3. The structures postulated there are indeed realized for the n-vertex YetterDrinfeld B p -modules. 4.6.1. One-vertex modules: coev and ev maps. For the irreducible YetterDrinfeld a Xr as in B.1.1, the coevaluation map coev : k Xa Xa is given module X0 0 0 in terms of dual bases as

B0, u , a b

1, 2.

a X
0

r1

a X
0

Vsa

Usa,

a p 1,

s 0

and the evaluation map ev : X0


X0

a Xa 0
X0

k, accordingly, as Vtb 1s,t a,b .

We then use (3.1) to nd the B p module comodule structure on the Usa . Simple calculation shows that F r Usa q ra2rs
r r 1

: Usa

s r

s1

t s r

t a Usar ,

22

SEMIKHATOV

Usa

p s 1 r 0

1r qra2srrr1F r

Usa r .

It follows that we can identify Usa 1as qs1sa2Vp1s (the action and coactionand in fact the braidingare identical for both sides). The coevaluation and evaluation maps can therefore be expressed as
a 2 2p

a X
0

r1

a X
0

Vsa

s 0 a

Vp1s
 a b Vs , Vt

2p a 2

1asqs1sa2,
2

a p 1,

1sqs sa1 st,p1ab,2p2 . For a p 1 mod p, evidently, a r 1 p implies that 2p a 2 p r 1 1 p, and therefore the module left dual to Vr in (4.11), with r a p 1, can be
: Vs Vtb identied as

and

Vr

. X p r 1

Xr

where Xr is dual to Xr in (4.11).

The properties postulated in (3.6) (and hence, of course, the rst relation in (3.4)) and in the second relation in (3.4) now hold, as is immediate to verify.
a, 4.6.2. Two-vertex modules. Similarly to 4.6.1, for the Us,tb that are dual to the twovertex basis,  a, b c, d Us,t , Vu, v ac,0 bd,0 s,u t,v ,

it follows from (3.1) that F r


a, Us,tb r u 0

1r qrr1rb2s2t csa,rb ur, uUsa, bru,t u. u,t


p 1 t and noting that the coefcients ca, b r, u s,t r u,

Replacing here s p 1 s and t in (B.4) have the symmetry ca, b r, u s,t


a, Us,tb

a 2, q2rr2t 2sab csb2 p1t ur, u, p 1 ru,

we arrive at the identication

1t sqt s22abt s3Vpa12, bs, p21t .

Hence, under the pairing


 a, b
c, d Vs,t , Vu, v

1st qst 2ab1st ac,2bd,2su,p1 t v,p1,

FUSION FOR A RANK-1 NICHOLS ALGEBRA

23

the module left dual to P0,t can be identied with P0, prt 1 (as before, a b 2t r 1 p, 1 r p 1). The module dual to (1.3) has the structure X p r 1

a, b

a2, b2

P0,t

a, b r

P0, prt 1

a2, b2 r

Xr2 X p r 1

Xr

4.7. Ribbon structure. We set

Vsa
q 2 ab2t 1
1

q 2 a1
1

1V a ,
s i1

which obviously commutes with the B p action and coaction, and (4.14)
a, Vs,t b
2

1 qia i t i
s i 0

j 0

t j b Vsa, b i i,t

(we recall that

1 q2 ).

4.8. Algebra (1.1) from (1.4). With the above ribbon structure, we now calculate diagram (3.7) in some cases. To maintain association with the diagram, we write Z Y as Z (the reasons for choosing the right action are purely notationalgraphical). The Y calculations in what follows are based on a formula for the double braiding: for two one-vertex modules, the last diagram in (A.3) evaluates as (4.15) B
2

Vsa

Vtb

s t s t min i,t n 0i n j 0 i j

qab2 j j1in1in2b jan2it


i j st j s st n in
i 1 j 0

j b Vsa, bn, n . t

4.8.1. If Y is irreducible, Y Xr , then Xr by a number; indeed, we nd that for all x X0 with a p where

Z can only amount to multiplication X0

p 1, x

a, bx,

a, b

qa1b1 qa1b1 . qa1 qa1

It is instructive to reexpress this eigenvalue by indicating the representation labels rather than the relevant coinvariants: for a r 1 p and b r 1 p, we nd that Xr Xr amounts to multiplication by r r rr , ; r, 1 r1 r p q q r qr qr

24

SEMIKHATOV

r1 r p

r i 1

qr r12i .
Xr amounts to

The last form is also applicable in the case where r p, and S p multiplication by p, ; r, 1 1r1 pr.

For Y Vr in (4.11), it may be worth noting that the identity r , ; r, p r , 1; r, , 1 r p 1, explicitly shows that the action is the same on both subquotients. 4.8.2. Next, the action Pr Xr has a diagonal piece, given again by multiplica , ; r, , and a nondiagonal piece, mapping the top subquotient in tion by r X p r 1 Pr

Xr X p r 1

Xr 2

into the bottom subquotient. Specically, in terms of the top and bottom elements dened in (B.16) and (B.17), we have uta,b 1 where Xr

r , ; r, uta,b 1 r , ; r, vta,b r 1,

r , ; r,

q qrr qr qr rqrr qrr qr qr . a, b Xr commutes with the B p action and coaction, and because P0,t is genBecause c erated by the B p action and coaction from uta,b 1, the action of X0 is thus dened on a, b . all of P
0,t

11 r r p qrq qr 3 q

r r

4.8.3. Let xr and pr be the respective operations have relations (1.1), which are the fusion algebra in [41]. We see explicitly from the above formulas that A only modulo 2.

Xr and

Pr . We then

Xr depends on both and

FUSION FOR A RANK-1 NICHOLS ALGEBRA

25

5. C ONCLUSION The construction of multivertex YetterDrinfeld BX -modules has a nice combinatorial avor: elements of the braided space X populate line intervals separated by vertex . operatorselements of another braided space Y , as This construction and the BX action on such objects are universal in that they are formulated at the level of the braid group algebra and work for any braiding. However, even for diagonal braiding, extracting information such as fusion from Nichols algebras by direct calculation is problematic, except for rank 1 (and maybe 2). Much greater promise is held by the program of nding the modular group representation and then extracting the fusion from a generalized Verlinde formula like the one in [13]. Importantly, those Nichols algebras that are related to CFT (and some certainly are, cf. [68]) presumably carry an SL2, Z representation on the center of their YetterDrinfeld category.

Going beyond Nichols algebras BX may also be interesting, and is meaningful from the CFT standpoint: adding the divided powers such as F p in our B p case, which are not in BX but do act on B p modules, would yield a braided (and, in a sense, onesided) analogue of the innite-dimensional quantum group that is KazhdanLusztig-dual to logarithmic CFT models viewed as Virasoro-symmetric theories [14, 15]. Acknowledgments. The content of Sec. 2 and Secs. 4.14.2 is the joint work with I. Tipunin [24]. I am grateful to N. Andruskiewitsch, T. Creutzig, J. Fjelstad, J. Fuchs, A. Gainutdinov, I. Heckenberger, D. Ridout, I. Runkel, C. Schweigert, I. Tipunin, A. Virelizier, and S. Wood for the very useful discussions. A. Virelizier also brought paper [56] to my attention. This paper was supported in part by the RFBR grant 10-01-00408 and the RFBRCNRS grant 09-01-93105.

A PPENDIX A. Y ETTER D RINFELD

MODULES

In the category of leftleft module comodules over a braided Hopf algebra B, a Yetter Drinfeld (also called crossed) module [67, 63, 64] is a left module under an action :B Y Y and left comodule under a coaction : Y B Y such that the axiom

B Y

(A.1)

26

SEMIKHATOV

holds. The category BYD of YetterDrinfeld B-modules is monoidal and braided. The B action and coaction on a tensor product of YetterDrinfeld modules are diagonal, respectively given by

(A.2)

and

For two YetterDrinfeld modules, their braiding and its inverse and square are given by

(A.3)

B1

, S 1

and

B2

S
MODULES

A PPENDIX B. C ONSTRUCTION

OF

Y ETTER D RINFELD B p

B.1. One-vertex modules. One-vertex YetterDrinfeld B p -modules [24] are spanned by Vsa (see (4.5)) for a xed a Z with s ranging over a subset of 0, . . ., p 1, under the B p action and coaction given in 4.2.
a B.1.1. Simple modules X0 . From each left coinvariant V0 , the action of B p generates a simple module comodule of dimension a p 1:

X0

SpanVsa 0

a p

(simply because F Va p 0 in accordance with the above formulas). The module coa module structure (in particular, the matrix of F r in the basis of Vsa , Eq. (B.2)) depends on a only modulo p, and hence there are just p nonisomorphic simple one-vertex module comodules, for which we choose the notation Xr indicating the dimension 1 r p; then there are the B p module comodule isomorphisms X0

Xr

whenever

a p

r 1. X p.

B.1.2. As noted above, we sometimes use a special notation S p

FUSION FOR A RANK-1 NICHOLS ALGEBRA

27

B.1.3. For each 1 r p 1, X0 extends to a reducible module comodule Vr with a Xr as a submodule and with the quotient isomorphic to X p r, as shown X0 in (4.7). In terms of basis, this is (B.1) va 1 where
F

va r 1 ...
F

...

va p

va r F i1 V0a , i r,

is a basis in X0 r, with the last, rth element given by va r CrVra 1 with a nonzero Cr, and hence the upper oor starts with the element va r 1 CrVra . The downward x1 whenever X arrow in (B.1) can be understood to mean X s F s xs ; this convention is a reasonable alternative to representing the same diagram as
F r

va i

va r

F 1

va r 1

...

va p

to express the idea that va r 1 B p F r

va 1

...

Spanva j 1
s 1 r i s

r 1 .

The general form of the adjoint action on the one-vertex space is (B.2)

Vsa

rs r

i aVsar,

B.1.4. We verify that (4.9) holds by counting the total dimension of the modules just constructed: dim S p
p 1 r 1

dim Vr

p p 1 p

p2 .

With the braiding (4.3), each of the above module comodules satises the Yetter Drinfeld axiom. B.2. Two-vertex modules. A two-vertex YetterDrinfeld module is a linear span of a, some Vs,t b , 0 s,t p 1, for xed integers a and b (see (4.4)). The left adjoint action of B p on these is given by (B.3) where (B.4)
a, cs,tb a, F r Vs,t b r u 0

ca, b r, uVsa, bu,t u, s,t r


r 1 i u

r, u

sru q ru
r u 2s a

t u u

s i 2t a b

u 1 j 0

t j b.

28

SEMIKHATOV a p, b

The dependence on b in (B.4) is modulo p, and on a, modulo 2p. However, cs,t r, u a 1uca, br, u and the matrix of F r in the basis Vs,t p, b0 s,t p1 is the same as in s,t a, the basis 1t Vs,t b 0 s,t p1 ; moreover, the coaction is unaffected by this extra sign. Hence, the module comodule structure depends on both a and b modulo p. We arrive at decomposition (4.10) by rst listing all module comodules generated from left coinvariants
a, V0,tb

(t crosses),

and then studying their extensions. In accordance with (B.3), the algebra acts on left coinvariants as F r with the coefcients (B.5) cta,b r, s
a,b c0,t r, s a, V0,tb r s 0

cta,b r, sVra, b s, s,t


s1

r qsa

r1 t s s i s

i 2t a b t j b.
j 0

In practical terms, the cases in 4.3.5 can be conveniently studies as follows.


a, a, (2) V0,tb is not in the image of F and F r V0,tb vanishes for some r a, a, (3) F r V0,tb is a coinvariant, i.e., F r V0,tb which is equivalent to a, (1) F r V0,tb is nonvanishing and not a coinvariant for all r, 1

p 1.

p 1, i.e., p 1,

cta,b r, s

0,

r.
a, const V0,tb r , for some r

cta,b r, s cta,b Let a b 2t p 1. For 0 b In fact, every triple a, b,t , 0 a, b,t (B.6)

r, r

0, 0.

r 1,

a, b,t a, 2t 1 a p, t ,
t s qsa s
r r

p 1, this is equivalent to b 2t 1 a p. p 1, can be uniquely represented as 1

p.

In this parameterization, coefcients (B.5) become cta,b

r, s

i s 1

j a t .

j 1

and the analysis of the above cases becomes relatively straightforward. The results are as follows.

FUSION FOR A RANK-1 NICHOLS ALGEBRA

29

B.2.1. Irreducible dimension-p modules S0,t p (case (1)). A simple module comodule of dimension p, isomorphic to S p in B.1.2, is generated under the B p action a, from a coinvariant V0,tb if and only if (B.7)

a, b

a b 2t p 1
a, b a, b

p.

When this condition is satised, we write S0,t , or even S0,t module isomorphic to S p.5

a, b p, for this module co-

B.2.2. Reducible dimension-p modules V0,t r (case (2)). A simple module comodule isomorphic to Xr for some 1 r p 1 is generated under the action of B p from a, a coinvariant V0,tb that is not itself in the image of F if and only if r a b 2t p 1 and either of the two conditions holds:6

a p and a p r 1 t p 1 r, t a p 1 and p r t p r a p. (B.9) a, b a, b In this case, we write X0,t or X0,t r for the corresponding module comodule: a, b Xr whenever r a b 2t 1 and (B.8) or (B.9) holds. (B.10) X0,t p
(B.8) t Every X0,t

a, b is further extended as in (4.7), which in terms of basis is now realized as


r 1 s 0

r 1! cta,br 1, sVra, bs,t sF

...

a, V0,tb

...

Fr 1 s 0

r 1! cta,br 1, sVra, b1s,t s

with the south-west arrow meaning the same as in B.1.3; the quotient is isomorphic to a, b X p r. The notation V0,t r for this dimension-p module comodule explicitly indicates the relevant left coinvariant and the dimension of the sub(co)module; the module a, b comodule structure depends only on r: V0,t r Vr.

a b 1 p mod p with a b even, or t a b 1 with a b odd. For even p, it holds if and only if either t 1 p 1 a b mod p or t 1 a b 1 (which selects only odd a b). 2 2 6The logic of the presentation is that we assume that 0 a, b p 1, and hence a a; but we do not p omit the operator of taking the residue modulo p because we refer to formulas given here also in the case where a Z.
5Condition (B.7) is actually worked out as follows: For odd p, it holds if and only if either t
1 2 1 2

30

SEMIKHATOV

B.2.3. Three-oor modules P0,t r (case (3)). We next assume that none of the above conditions (B.7), (B.8), (B.9) is satised. An exemplary exercise shows that the negation of (B.7) (B.8) (B.9) is the or of the four conditions (B.11) (B.12) (B.13) (B.14) t t pr p r a p 1, t

a, b

where again r a, variant V0,tb is then a sub(co)module in an indecomposable module comodule with the structure of subquotients X p r Xr X p r Xr

a p r, a p 1 t p r 1, a b 2t p 1, 1 r p 1. The module generated from the coin-

a p,

where r is either r or p r, as we now describe.

i. If t r p (which means that either (B.11) or (B.12) holds), then the submodule a, b a, generated from V0,tb is isomorphic to Xr. We let it be denoted by B0,t r. (B is for bottom, and L is for left.)

ii. If t r p 1 (which means that either (B.13) or (B.14) holds), then the suba, b a, module generated from V0,tb , denoted by L0,t r, is a p-dimensional reducible module comodule with B0,t r p r quotient isomorphic to Xr:

a, b

X p r as a sub(co)module and with the

a, b r L
0,t

Xr X p r.

In terms of basis, this diagram is


a, V0,tb F

...

r 1 s 0

r 1! cta,br 1, sVra, b1s,t s


F a, r!V0,tbr F

...

The set of all diagrams of this type actually describes both cases i and ii: according a, b to whether a given coinvariant V0, u is or is not in the image of F, it occurs either in the bottom line (case i) or in the upper line (case ii) of the last diagram.

FUSION FOR A RANK-1 NICHOLS ALGEBRA

31

Every such diagram is extended further, again simply because of the cofree nature of the coaction: (B.15)
a, V0,tb F

Tta,b r r ...
F r 1 s 0

r 1! cta,br 1, sVra, bs1,t s

F a, r! cta,br, rV0,tbr F

...

where, evidently, Tta,b r r Setting (B.16) (B.17)

r 1 s 0

r 1! cta,br 1, sVra, bs,t s.


F i1 Tta,b r, r
a, F i1 V0,tb ,

uta,b i vta,b i

we have the full picture extending (B.15) as (omitting the ta,b labels for brevity) u1 v1
F F

...

u p r

...

vr

u p r 1
F

...

u p

vr 1 Here,7 and, similarly,

...

v p vr . . . F 1 v p . . ..

u1 u p r 1

u1

r 1! F 1
1 q2r r 1!

u p r 1

is not generated from this element), we write P0,t

a, b to indicate both the module type and a, b the characteristic coinvariant. An even more redundant notation is P0,t r, indicating
the length r of the left wing (which of course is r comodule structure depends only on r: (B.18) P0,t

a, To label such modules by the leftmost coinvariant V0,tb (even though the entire module

a b 2t p 1).

The module

a, b r Pr.

7The closure of the rhombus in the above diagram is a good illustration of the use of the Yetter

Drinfeld axiom, which is also used in several other derivations without special notice. The relative factor q2r in the next two formulas, in particular, is an immediate consequence of the YetterDrinfeld condition.

32

SEMIKHATOV

3 X1, 0 22 S0,1,00 510 0, 1 1, 0 X0, 2 31 1, 0 L 11 0, 3 1, 0 4 B


B0, 4

X0, 0 1 X0,0,00 40 1 1 0,0, 0 L 2 21 0,0, 0 S 5


0, 3 0, 0 2

X0, 1 2 S0,0,01 510 0, 1 0, 1 X0, 2 31 0, 1 L 1


0, 1 4 X1, 1 32 X0,1,01 10 1 0 0,1, 1 X 2 41 0,1, 1 X 21 0, 3 1, 1 5 S
B0, 4
0, 4 2 0, 3 1

X0, 2 3 L0,0,02 100 1 0,0, 2 B0, 2 41 0, 2 X 2


0, 2 X1, 252 4 X0,1,02 20 1 0 0,1, 2 S0, 2 51 1, 2 X 31 0, 3 1, 2 1 X
S0, 4
0, 4 1 0, 3 1

X0, 3 4 L0,0,03 200 1 0,0, 3 S0, 2 51 0, 3 B 3


0, 3 1 S1, 3 51 0 X0,1, 3 300 1 0,1, 3 L 2 10 0,1, 3 B 4 1 0, 3 1, 3 2 X
X0, 4
0, 4 1 0, 3 1

L4, 0 3 0, 1 0 4, 0 L0, 2 10 4, 0 B 4
B0, 4
0, 3 4, 0

0, 4

................................................................................ 4, 0 4, 1 4, 2 4, 3 4, 4 S0, 0 50 L0, 0 11 L0, 0 21 L0, 0 31 L0, 0 41

2 1

B4, 1 4 0, 1 0 4, 1 L0, 2 20 4, 1 S 5
B0, 4
0, 3 4, 1

3 1

S4, 2 5 0, 1 0 4, 2 B0, 2 30 4, 2 L 1
B0, 4

4, 2 4

0, 3

L4, 3 1 0, 1 1 4, 3 B0, 2 40 4, 3 B 2
S0, 4

4, 3 5

0, 3

L4, 4 2 0, 1 1 4, 4 S0, 2 50 4, 4 B 3
B0, 4
0, 3 4, 4

L1, 4 121 0 B0,1, 4 401 0, 1 1, 4 L0, 2 20 1, 4 S 51 0, 3 1, 4 3 B


B0, 4
0, 4 1

S0, 4 5 0 L0,0, 4 300 1 0, 0, 4 L0, 2 10 0, 4 B 4


0, 3 0, 4 1

1 0

t p 1, 0 a p 1, 0 b p 1 (where a,b p 5), the module comodule generated from V0,t r is indicated as Aa,b r , 0,t where r is the dimension of the relevant subquotient, is the braiding index, and A indicates the module type. Only a 0, 1, 4 are shown for compactness. a,b a,b Whenever an L0,t r occurs in a column of height 5, the B0,t r p r module is present in the same column. We do not replace negative braiding indices 1 with the canonical representative 3 in Z4 for continuity.

F IGURE B.1. For each 0

a, To summarize, given a coinvariant V0,tb , (B.18) holds if and only if (for r 1)

a b 2t p

1 (B.19) t

a p r

p 1 and

or a p 1

p r 1.

B.2.4. Completeness. We verify (4.10) by counting the total dimension of the modules constructed. This gives p4 , the dimension of V p 2, as follows. There are p2 modules S p constructed in B.2.1, 2r p r modules Xr in B.2.2 for each 1 r p 1, making 1 the total of 3 p p2 1, and, nally, p r2 modules Lr in B.2.3 for each 1 r p 1, making the total of 1 p p 12p 1. Each S p is p-dimensional, each Xr extends 6 to a p-dimensional module, and each L0,t total dimension is
1 3

a, b r extends to a 2p-dimensional module. The


1 6

p2 p p p2 1 p p p 12p 1 2p

p4 .

FUSION FOR A RANK-1 NICHOLS ALGEBRA

33

B.2.5. Example. Decomposition (4.10) is illustrated in Fig. B.1 for p 5. The gure a, lists all the modules generated from the V0,tb with a 0, 1, 4, b 0, 1, 2, 3, 4, and t 0, 1, 2, 3, 4 (two values of a are omitted for compactness). Each B0, u YetterDrinfeld sumbodule in the L0, ur p r module in the same column of p modules. The subscript additionally indicates the braiding sectors (see 4.4). A PPENDIX C. S OME
TECHNICALITIES

a, b

a, b r module is a
5

C.1. Braiding of the fusion product. For any two (single-vertex or multivertex) Yetter Drinfeld BX modules Y and Z, the braiding of their fusion product (see (2.17)) B : Y Z Z Y is given by the diagram
S

(C.1)

B:

where : Z Z is an idempotent dened in C.2 below. This braiding operation commutes with the BX action and coaction (2.16). Direct calculation with diagrams shows that the braiding squared is
S

(C.2)

B2 :

where is the squared relative antipode dened below. The use of this somewhat unprepossessing diagram in this paper is that it encodes a rather simple analytic expression in Eq. (4.15).

34

SEMIKHATOV

C.2. Idempotents and 2 . For each Hopf bimodule Z over BX , there are idempotents
S

(C.3)

and

S 1

which map Z

Z and project onto left coinvariants:


and

(where is the unit of BX ). It also follows that


and

(where is the counit of BX ). Moreover, the two idempotents preserve the spaces of right coinvariants, and are therefore dened on YetterDrinfeld BX -modules realized as right coinvariants. On single-vertex YetterDrinfeld modules, in particular, we have X

0,
s

1, 0

id, s

and on two-vertex (and similarly on multivertex) YetterDrinfeld modules, (C.4) X


s

Y X

id

Xs,t s,1 S s .

Bespalovs squared relative antipode is dened as [63]

(C.5)

S2

(see [64, 69] for its further properties and use). Two useful properties are

(C.6)

S2

and

S2

FUSION FOR A RANK-1 NICHOLS ALGEBRA

35

C.3. Specializing to B p . We now specialize to the B p algebra in Sec. 4. Formula (C.4) for the idempotent acting on the two-vector space becomes
a, Vs,t b

1sqss1a s t t

a, V0, sb t .

For the idempotent , similarly,


a, b Vs,t

1sqss1a2t s t t

a, V0, sb t .

The squared relative antipode in (C.5) evaluates on elements of one-vertex and twovertex modules as

2Vsa
a, 2Vs,t b

q2ss1aVsa ,
t i q qia2t i i
s 2s s 1 a b 2t i 0

i1

j 0

b t j Vsa, b i. i,t

Using this to evaluate diagram (C.2) on two-vertex modules, we express the squared braiding on the two-vertex space explicitly as
a, B 2Vs,t b

qab2t t 1ab

s i 0

qia i

i 1 t i i

j 0

t j b Vsa, bi,t i.

This formula underlies formula (4.14) for the ribbon map. It is also easily recalculated into (4.15). C.4. Equivalent formulas for coevaluation maps. C.4.1. The coevaluation and evaluation maps in 4.6.1 can be conveniently expressed in terms of the va i introduced in B.1.3. If a r 1 p with 1 r p 1 (the module Xr), then
 2pa2 v j , va i
p

ji,p11ir1 p i
i

where q2r r1 i module S p), then

is q2r if r 1

p2 i2 r4i 2q r

p and 1 otherwise.

q2r r1 i p, If a p 1 p (the
2

 2pa2 v j , va i

ji,p11 p p i p p 1! qi .

C.4.2. For any a, b,t satisfying (B.19), the coevaluation map in 4.6.2 for the correa, b sponding module P0,t is expressed in terms of basis (B.16), (B.17) as
p P0,t

P0,t

a, b

a, b

i 1

ta,b

vta,b i

a u2,b2 p 1 i p rt 1

36

SEMIKHATOV

where

p i 1

ta,b

uta,b i

a v2,b2 p 1 i p rt 1

ta,b i

qt 2t ra2ari1ira11it i1t p,t, r



a v2,b2 p 1 i, uta,b i p rt 1

a u2,b2 p 1 i, vta,b i p rt 1

and

ta,b i

qt 2t ra22ri1ira11irt i1t r p,t, r


r 1! r1t j b. r j 0

with

p,t, r

p2 r p i

t r r

We note that for r 1

a2,b p, prt 1 2 p 1 i
R EFERENCES

q2b4a6 ta,b i.

[1] D. Kazhdan and G. Lusztig, Tensor structures arising from afne Lie algebras, I, J. Amer. Math. Soc. 6 (1993) 905947; II, J. Amer. Math. Soc. 6 (1993) 9491011; III, J. Amer. Math. Soc. 7 (1994) 335381; IV, J. Amer. Math. Soc. 7 (1994) 383453. [2] M. Finkelberg, An equivalence of fusion categories, Geometric and Functional Analysis (GAFA) 6 (1996) 249267. [3] V.G. Turaev, Quantum Invariants of Knots and 3-Manifolds, Walter de Gruyter, BerlinNew York (1994). [4] G. Moore and N. Seiberg, Lectures on RCFT, in: Physics, Geometry, and Topology (Trieste spring school 1989), p. 263; Plenum (1990). [5] J. Fuchs, I. Runkel, and C. Schweigert, TFT construction of RCFT correlators I: Partition functions, Nucl. Phys. B 646 (2002) 353 [hep-th0204148]; TFT construction of RCFT correlators II: Unoriented surfaces, Nucl. Phys. B 678 (2004) 511 [hep-th0306164]; TFT construction of RCFT correlators III: Simple currents, Nucl. Phys. B 694 (2004) 277 [hep-th0403158]; TFT construction of RCFT correlators IV: Structure constants and correlation functions, Nucl. Phys. B 715 (2005) 539 [hep-th0412290]. [6] J. Frhlich, J. Fuchs, I. Runkel, and C. Schweigert, Correspondences of ribbon categories, Adv. Math. 199 (2006) 192 [math.CT/0309465]. [7] J. Fuchs and C. Schweigert, Hopf algebras and nite tensor categories in conformal eld theory, Revista de la Union Matematica Argentina 51 (2010) 4390 [arXiv:1004.3405 [hep-th]]. [8] B.L. Feigin, A.M. Gainutdinov, A.M. Semikhatov, and I.Yu. Tipunin, Modular group representations and fusion in logarithmic conformal eld theories and in the quantum group center, Commun. Math. Phys. 265 (2006) 4793 [arXiv:hep-th/0504093].

FUSION FOR A RANK-1 NICHOLS ALGEBRA

37

[9] B.L. Feigin, A.M. Gainutdinov, A.M. Semikhatov, and I.Yu. Tipunin, KazhdanLusztig correspondence for the representation category of the triplet W -algebra in logarithmic CFT, Theor. Math. Phys. 148 (2006) 12101235 [arXiv:math/0512621 [math.QA]]. [10] B.L. Feigin, A.M. Gainutdinov, A.M. Semikhatov, and I.Yu. Tipunin, Logarithmic extensions of minimal models: characters and modular transformations, Nucl. Phys. B757 (2006) 303343 [arXiv: hep-th/0606196]. [11] B.L. Feigin, A.M. Gainutdinov, A.M. Semikhatov, and I.Yu. Tipunin, KazhdanLusztig-dual quantum group for logarithmic extensions of Virasoro minimal models, J. Math. Phys. 48 (2007) 032303 [arXiv:math/0606506 [math.QA]]. [12] K. Nagatomo and A. Tsuchiya, The triplet vertex operator algebra W p and the restricted quantum group at root of unity, arXiv:0902.4607 [math.QA]. [13] A.M. Gainutdinov and I.Yu. Tipunin, Radford, Drinfeld, and Cardy boundary states in 1, p logarithmic conformal eld models, J. Phys. A42 (2009) 315207 [arXiv:0711.3430]. [14] P.V. Bushlanov, B.L. Feigin, A.M. Gainutdinov, and I.Yu. Tipunin, Lusztig limit of quantum sl 2 at root of unity and fusion of 1, p Virasoro logarithmic minimal models, Nucl. Phys. B818 (2009) 179195 [arXiv:0901.1602]. [15] P.V. Bushlanov, A.M. Gainutdinov, and I.Yu. Tipunin, Kazhdan-Lusztig equivalence and fusion of Kac modules in Virasoro logarithmic models, arXiv:1102.0271. [16] T. Abe and Y. Arike, Intertwining operators and fusion rules for vertex operator algebras arising from symplectic fermions, arXiv:1108.1823. [17] T. Creutzig and D. Ridout, Relating the archetypes of logarithmic conformal eld theory arXiv:1107. 2135. [18] Y.-Z. Huang and J. Yang, Logarithmic intertwining operators and associative algebras arXiv:1104. 4679. [19] Y. Arike and K. Nagatomo, Some remarks on pseudo-trace functions for orbifold models associated with symplectic fermions, arXiv:1104.0068. [20] R. Vasseur, J.L. Jacobsen, and H. Saleur, Indecomposability parameters in chiral logarithmic conformal eld theory, Nucl. Phys. B851 (2011) 314345 [arXiv:1103.3134]. [21] D. Adamovic and A. Milas, On W -algebra extensions of 2, p minimal models: p 3, arXiv:1101. 0803. [22] Y.-Z. Huang, J. Lepowsky, and L. Zhang, Logarithmic tensor category theory for generalized modules for a conformal vertex algebra, I: Introduction and strongly graded algebras and their generalized modules, arXiv:1012.4193; Logarithmic tensor category theory, II: Logarithmic formal calculus and properties of logarithmic intertwining operators, arXiv:1012.4196; Logarithmic tensor category theory, III: Intertwining maps and tensor product bifunctors, arXiv:1012.4197; Logarithmic tensor category theory, IV: Constructions of tensor product bifunctors and the compatibility conditions; arXiv: 1012.4198; Logarithmic tensor category theory, V: Convergence condition for intertwining maps and the corresponding compatibility condition, arXiv:1012.4199; Logarithmic tensor category theory, VI: Expansion condition, associativity of logarithmic intertwining operators, and the associativity isomorphisms, arXiv:1012.4202. [23] M.R. Gaberdiel, I. Runkel, and S. Wood, A modular invariant bulk theory for the c 0 triplet model J. Phys. A44 (2011) 015204 [arXiv:1008.0082]. [24] A.M. Semikhatov and I.Yu. Tipunin, The Nichols algebra of screenings, arXiv:1101.5810. [25] W. D. Nichols, Bialgebras of type one, Commun. Algebra 6 (1978) 15211552. [26] N. Andruskiewitsch and M. Graa, Braided Hopf algebras over non abelian nite groups, Bol. Acad. Nacional de Ciencias (Cordoba) 63 (1999) 4578 [arXiv:math9802074 [math.QA]].

38

SEMIKHATOV

[27] N. Andruskiewitsch and H.-J. Schneider, On the classication of nite-dimensional pointed Hopf algebras, Ann. Math. 171 (2010) 375417 [arXiv:math0502157 [math.QA]]. [28] N. Andruskiewitsch and H.-J. Schneider, Pointed Hopf algebras, in: New directions in Hopf algebras, MSRI Publications 43, pages 168. Cambridge University Press, 2002. [29] N. Andruskiewitsch, Some remarks on Nichols algebras, in: Hopf algebras, Bergen, Catoiu and Chin (eds.) 2545. M. Dekker (2004). [30] I. Heckenberger, Classication of arithmetic root systems, Adv. Math. 220 (2009) 59124 [math.QA 0605795]. [31] I. Heckenberger, The Weyl groupoid of a Nichols algebra of diagonal type, Invent. Math. 164 (2006) 175188. [32] N. Andruskiewitsch, I. Heckenberger, and H.-J. Schneider, The Nichols algebra of a semisimple YetterDrinfeld module, Amer. J. Math., 132 (2010) 14931547 [arXiv:0803.2430 [math.QA]]. [33] N. Andruskiewitsch, D Radford, and H.-J. Schneider, Complete reducibility theorems for modules over pointed Hopf algebras, J. Algebra, 324 (2010) 29322970 [arXiv:1001.3977]. [34] M. Graa, I. Heckenberger, and L. Vendramin, Nichols algebras of group type with many quadratic relations, arXiv:1004.3723. [35] M. Graa and I. Heckenberger, On a factorization of graded Hopf algebras using Lyndon words, J. Algebra 314 (2007) 324343. [36] N. Andruskiewitsch, F. Fantino, G.A. Garcia, and L. Vendramin, On Nichols algebras associated to simple racks, Contemp. Math. 537, Amer. Math. Soc., Providence, RI, 2011, pp. 31-56. [arXiv:1006. 5727]. [37] N. Andruskiewitsch, I. Angiono, and H. Yamane, On pointed Hopf superalgebras, Aparecer en Contemp. Math. 544, Amer. Math. Soc., Providence, RI, 2011 [arXiv:1009.5148]. [38] I. Angiono, On Nichols algebras with standard braiding, Algebra & Number Theory 3 (2009) 35106 [arXiv:0804.0816]. [39] I.E. Angiono, A presentation by generators and relations of Nichols algebras of diagonal type and convex orders on root systems, arXiv:1008.4144. [40] I. Angiono, On Nichols algebras of diagonal type, arXiv:1104.0268. [41] J. Fuchs, S. Hwang, A.M. Semikhatov, and I.Yu. Tipunin, Nonsemisimple fusion algebras and the Verlinde formula, Commun. Math. Phys. 247 (2004) 713742 [arXiv:hep-th/0306274]. c [42] D. Adamovi and A. Milas, Lattice construction of logarithmic modules for certain vertex algebras, Selecta Math. New Ser. 15 (2009) 535561 [arXiv:0902.3417 [math.QA]]. [43] H.G. Kausch, Extended conformal algebras generated by a multiplet of primary elds, Phys. Lett. B 259 (1991) 448. [44] M.R. Gaberdiel and H.G. Kausch, Indecomposable fusion products, Nucl. Phys. B477 (1996) 293 318 [hep-th9604026]; [45] M.R. Gaberdiel and H.G. Kausch, A rational logarithmic conformal eld theory, Phys. Lett. B 386 (1996) 131137 [hep-th9606050]. [46] M.R. Gaberdiel and H.G. Kausch, A local logarithmic conformal eld theory, Nucl. Phys. B538 (1999) 631658 [hep-th9807091]. [47] J. Murakami, K. Nagatomo, Logarithmic knot invariants arising from restricted quantum groups, Internat. J. Math 19 (2008) 12031213 [arXiv:0705.3702 [math.GT]]. [48] P. Furlan, L. Hadjiivanov, and I. Todorov, Zero modes fusion ring and braid group representations for the extended chiral WZNW model, Lett. Math. Phys. 82 (2007) 117151 [arXiv:0710.1063]. [49] Y. Arike, Symmetric linear functions of the restricted quantum group Uq sl2 C, arXiv:0706.1113; Symmetric linear functions on the quantum group g p,q, arXiv:0904.0331 [math.QA].

FUSION FOR A RANK-1 NICHOLS ALGEBRA

39

[50] H. Kondo and Y. Saito, Indecomposable decomposition of tensor products of modules over the restricted quantum universal enveloping algebra associated to sl2 , arXiv:0901.4221 [math.QA]. [51] A.M. Semikhatov, Heisenberg double HB as a braided commutative YetterDrinfeld module algebra over the Drinfeld double, Commun. Algebra 39 (2011) 18831906 [arXiv:1001.0733]. [52] A. Alekseev, D. Gluschenkov, and A. Lyakhovskaya, Regular representation of the quantum group slq 2 (q is a root of unity), St. Petersburg Math. J. 6 (1994) 88. [53] R. Suter, Modules over Uq sl2 , Commun. Math. Phys. 163 (1994) 359393. [54] J. Xiao, Finite dimensional representations of Ut sl 2 at roots of unity, Can. J. Math. 49 (1997) 772787. [55] A.M. Semikhatov, Factorizable ribbon quantum groups in logarithmic conformal eld theories, Theor. Math. Phys. 154 (2008) 433453 [arXiv:0705.4267 [hep-th]]. [56] A. Bruguires, Double braidings, twists and tangle invariants, J. Pure Appl. Algebra 204 (2006) 170 194. [57] M. Rosso, Groupes quantiques et algbres de battage quantiques, C.R.A.S. Paris, 320 (Srie I) (1995) 145148. [58] P. Schauenburg, A characterization of the Borel-like subalgebras of quantum enveloping algebras, Commun. Algebra 24 (1996) 28112823. [59] N. Andruskiewitsch, About nite dimensional Hopf algebras, Contemp. Math 294 (2002) 157. [60] M. Graa, A freeness theorem for Nichols algebras, J. Alg. 231 (2000) 235257. [61] M. Rosso Quantum groups and quantum shufes, Invent. math. 133 (1998) 399416. [62] S.L. Woronowicz, Differential calculus on compact matrix pseudogroups (quantum groups), Commun. Math. Phys. 122 (1989) 125170. [63] Yu.N. Bespalov, Crossed modules, quantum braided groups, and ribbon structures, Theor. Math. Phys. 103 (1995) 621637. [64] Yu.N. Bespalov, Crossed modules and quantum groups in braided categories, arXiv:q-alg/9510013. [65] Yu. Bespalov and B. Drabant, Hopf (bi-)modules and crossed modules in braided monoidal categories, J. Pure and Applied Algebra 123 (1998) 105129. [66] S. Majid, Foundations of Quantum Group Theory, Cambridge University Press, 1995. [67] P. Schauenburg, Hopf modules and YetterDrinfeld modules, J. Algebra 169 (1994) 874890. [68] A.M. Semikhatov, Virasoro central charges for Nichols algebras, arXiv:1109.1767. [69] Yu. Bespalov, T. Kerler, V. Lyubashenko, and V. Turaev, Integrals for braided Hopf algebras, q-alg/ 9709020. L EBEDEV P HYSICS I NSTITUTE , RUSSIAN ACADEMY
OF

S CIENCES

Anda mungkin juga menyukai