Monoidal Ladder Categories
Mitchell Rowett
October 2019
A thesis submitted for the degree of Bachelor of Philosophy (Honours) of the Australian National University
Declaration
The work in this thesis is my own except where otherwise stated.
Mitchell Rowett
Acknowledgements
Firstly, I would like to thank my supervisor, Scott Morrison. Thank you for teach- ing me so much over the past few years, for your kindness and support, and for your incredible generosity with your time. I would also like to thank the lecturers at the MSI for their patience and enthusiasm, and their willingness to indulge all kinds of questions. Thanks should also go to the students and rest of the community at the MSI; it has been a privilege to have been a part of such a supportive and welcoming community. I would especially like to thank the other honours students, who have taken this journey with me. Last but not least, I would like to thank my friends and family for their continual support. Mum, Dad, Cooper, Brittany, and Nat, your encouragement is the reason I have made it this far. Chris and Huon, thank you for sticking by me this year, and for doing so much to help me when I needed it most. Natalie, I think you already know how much I have relied on your patience and positivity. I am so very proud of you.
v
Abstract
In this thesis, we construct a tensor product of module categories over a linear rigid monoidal category, which we call a ladder category. In the case of monoidal module categories over a braided category, we exhibit a monoidal structure on the ladder category. We then give two major examples. For the first, we show that given a fu- sion category C with a central functor from RepG, the de-equivariantisation of C can be realised as the idempotent completion of the ladder category of C with Vec over RepG. We also give a proof that the definition of de-equivariantisation by de-enrichment given in [MP19] is equivalent to the standard definition of de- equivariantisation. For the second example we give an explicit description of the ladder category of the two unital Ad E8 fusion categories over Fib, which appears to be a fusion cat- egory not previously studied. We also show that the ladder category construction is equivalent to the Deligne tensor product in the case of fusion categories.
vii
Contents
Acknowledgements v
Abstract vii
1 Introduction 1
2 Background 3 2.1 Monoidal Categories ...... 3 2.1.1 Definition of a Monoidal Category ...... 3 2.1.2 The Coherence Theorem ...... 4 2.1.3 String Diagrams ...... 6 2.1.4 Monoidal Functors and Monoidal Equivalences ...... 6 2.1.5 Braided Monoidal Categories ...... 8 2.1.6 Rigid Monoidal Categories ...... 10 2.2 Linear Categories ...... 11 2.2.1 Additive and Linear Categories ...... 11 2.2.2 Additive Completion ...... 13 2.2.3 Idempotent Complete Categories and the Karoubi Envelope 14 2.2.4 Semisimple Categories ...... 14 2.3 (Co)Algebras and (Co)Modules ...... 17 2.3.1 Monoids and Algebras ...... 17 2.3.2 Module Objects ...... 18 2.3.3 Coalgebras and Comodules ...... 18 2.3.4 Categories of Modules and Comodules ...... 19 2.3.5 Free Comodules ...... 21 2.4 Monoidal Module Categories ...... 22 2.4.1 Module Categories ...... 22 2.4.2 Monoidal Module Categories ...... 24
ix x CONTENTS
3 Ladder Categories 25 3.1 Ladder Categories ...... 25 3.1.1 Definition of Ladder Categories ...... 25 3.1.2 Ladder Categories Over Semisimple Categories ...... 28 3.2 Monoidal Ladder Categories ...... 28 3.2.1 Definition of Monoidal Ladder Categories ...... 28 3.2.2 Ladder Categories over Vec ...... 31 3.3 Properties of Ladder Categories ...... 32
4 De-equivariantisation 37 4.1 Definitions of de-equivariantisation ...... 37 4.1.1 Equivariantisation ...... 37 4.1.2 De-equivariantisation ...... 38 4.2 Frobenius Algebras ...... 39 4.2.1 Definition of Frobenius Algebras ...... 40
4.2.2 Equivalence Between Cpcomod-A and Ccomod-A ...... 42 4.3 Ladders and de-equivariantisation ...... 47 4.4 The Enrichment De-equivariantisation ...... 54 4.4.1 Enriched Categories ...... 54 4.4.2 Monoidally Enriched Categories ...... 56 4.4.3 Transporting Enrichment ...... 57 4.4.4 The Enrichment De-equivariantisation ...... 58 4.4.5 The Underlying Monoidal Categories ...... 60 4.4.6 Equivalence to De-equivariantisation ...... 63
5 Tensor Categories and the Deligne Product 69 5.1 Tensor Categories ...... 69 5.1.1 Fusion Categories as Tensor Categories ...... 71 5.2 The Deligne Product ...... 72 5.2.1 Monoidal Structure on the Deligne Product ...... 73 5.3 Ladder Categories and the Deligne Product ...... 74
− + 6 An Explicit Example: Ad E8 ⊗Fib Ad E8 77 6.1 Classification of “Small” Fusion Categories ...... 77 6.1.1 Frobenius-Perron Dimension ...... 77 6.1.2 Principal Graphs ...... 78 6.1.3 The ADE Fusion Categories ...... 79 6.1.4 Planar Algebras and a Classification ...... 80 CONTENTS xi
6.1.5 Fib and Ad E8 ...... 81 − + 6.2 Fusion Rules for Ad E8 ⊗Fib Ad E8 ...... 83 − + 6.2.1 Simple Objects of Ad E8 ⊗Fib Ad E8 ...... 83 − + 6.2.2 Calculating Fusion Rules of Ad E8 ⊗Fib Ad E8 ...... 87 − + 6.3 Ad E8 ⊗Fib Ad E8 as a Category of Modules ...... 90
A Adjunctions 93
Bibliography 94
1 Introduction
In this thesis, we give a construction of a tensor product of “monoidal module categories” M, N over a braided category C, which we call the ladder category
M ⊗C N . We give two major examples. The first is of the de-equivariantisation of a monoidal module category over RepG, where G is a finite group.
Theorem A. Given a braided linear functor RepG → Z(C) for a fusion category C, the de-equivariantisation can be realised as the idempotent completion of a monoidal ladder category
CG ' Kar(C ⊗RepG Vec).
A third definition of de-equivariantisation was given without proof in [MP19]; we provide a proof that it is indeed equivalent to the standard definition of de- equivariantisation. The second example is of unitary ADE fusion categories over a common sub- category.
Theorem B. The adjoint subcategories of the two unitary E8 fusion categories are (right and left) monoidal module categories for Fib. Hence
− + Ad E8 ⊗Fib Ad E8 is a monoidal category, and if semisimple has principal graph
XL
2 XR 5 3 . 1 X 2 Y τ
YR 2
YL
We believe this to be a fusion category which has not previously been studied.
1 2 CHAPTER 1. INTRODUCTION
We also provide a comparison between monoidal ladder categories and the Deligne tensor product of tensor categories, and show that they are equivalent for fusion categories. One can think of the monoidal ladder category construction as an alternative to the Deligne tensor product which can be applied to a strictly larger class of categories, though unfortunately it does not always agree with the Deligne tensor product. However, monoidal ladder categories have an advantage as they give a way of taking tensor products of categories which are not necessarily abelian, which can be inaccessible for some constructions of linear monoidal categories. We do not assume any familiarity with most of the notions described above, but rather assume only basic knowledge of categories, functors, natural transforma- tions, and adjunctions. Most of the definitions used above are defined in Chapter 2. The remainder of this thesis is structured as follows. In Chapter 2 we re- call definitions from category theory which will be useful as we develop later theory. In Chapter 3 we define monoidal ladder categories and show some ba- sic results about them. In Chapter 4 we give two alternative definitions of de- equivariantisation, firstly as a monoidal ladder category and secondly as a “de- enrichment” [MP19], and show that these are both equivalent to the standard definition of de-equivariantisation. In Chapter 5 we give an overview of the basic definitions of tensor categories and the Deligne tensor product, and show that the Deligne tensor product is equivalent to the monoidal ladder category in the case of fusion categories. Finally, in Chapter 6 we further describe the theory of the unitary − + ADE fusion categories, and given an explicit construction of Ad E8 ⊗Fib Ad E8 . 2 Background
In this chapter, we recall aspects of the theory of monoidal, linear, and semisimple categories which will be important to the developments in later chapters. We also give a brief overview of the theory of algebra and module objects, which will be especially useful in Chapter 4, and then progress to the categorical analogues by defining module and monoidal module categories. We assume the reader is familiar with the definitions of categories, functors, natural transformations, and adjunctions (a brief summary of the last is given in the appendix). A discussion of this theory can be found in [Mac98]. Where examples are given, they are chosen primarily for use later in this thesis.
2.1 Monoidal Categories
2.1.1 Definition of a Monoidal Category
A monoidal category is a category with a functor ⊗ which acts as a kind of “tensor product”. For example, in the category of finite-dimensional vector spaces Vec we can take tensor products of both objects and morphisms using the standard tensor product. A category can often be made into a monoidal category in more than one way. For example, Vec can also be made into a monoidal category with the direct sum as the tensor product. Definition 2.1. A monoidal category is a category C with • a functor ⊗ : C × C → C called the tensor product, • a distinguished object 1 ∈ C called the unit object, • a natural isomorphism a with components of the form
∼ ax,y,z :(x ⊗ y) ⊗ z −→ x ⊗ (y ⊗ z) called the associator, • natural isomorphisms l and r with components of the form
∼ lx : 1 ⊗ x −→ x ∼ rx : x ⊗ 1 −→ x
3 4 CHAPTER 2. BACKGROUND
called the left unitor and right unitor respectively, such that the following two diagrams commute for all w, x, y, z ∈ C: • the triangle identity:
ax,1,y (x ⊗ 1) ⊗ y x ⊗ (1 ⊗ idy)
rx⊗idy idx ⊗ly x ⊗ y
• the pentagon identity:
((w ⊗ x) ⊗ y) ⊗ z
aw,x,y⊗idz aw⊗x,y,z
(w ⊗ (x ⊗ y)) ⊗ z (w ⊗ x) ⊗ (y ⊗ z)
aw,x⊗y,z aw,x,y⊗z w ⊗ ((x ⊗ y) ⊗ z) w ⊗ (x ⊗ (y ⊗ z)) idw ⊗ax,y,z
We will often suppress the ⊗ symbol, instead simply concatenating both objects and morphisms as xy and fg. We will therefore always write composition explicitly as g ◦ f.
Examples 2.2. i) As discussed above, the category Vec of finite-dimensional vector spaces with ⊗ given by the tensor product of vector spaces.
ii) The category RepG of finite-dimensional representations over a finite group
G with ⊗ given by the tensor product of representations; that is, if (V, ρV :
G → GL(V )) and (W, ρW : G → GL(W )) are representations, then their
tensor product is (V ⊗ W, ρV ⊗W ) where
ρV ⊗W (g) = ρV (g) ⊗ ρW (g).
The unit object is the trivial representation.
2.1.2 The Coherence Theorem
For any ordered sequence of objects x1, . . . , xn ∈ C, we can take the tensor product x1 . . . xn. However, we can take different ways of parenthesising this product, and a priori these may be different objects. For n = 3, the associator gives an ∼ isomorphism (x1x2)x3 −→ x1(x2x3), while for n ≥ 4 it is fairly simple to see that 2.1. MONOIDAL CATEGORIES 5
for any two ways P1,P2 of parenthesising there is a chain of associators (tensored with identity morphisms) which gives an isomorphism from P1 to P2. However, we run into a problem in that there may well be multiple ways of chaining these associators to obtain an isomorphism, giving potentially different isomorphisms. If these isomorphisms are truly different, then we cannot say that
P1 and P2 are canonically isomorphic, as there is no reason to prefer one particular chain of associators over another. The purpose of the pentagon diagram, then, is to assert that for n = 4 all sequences of associators from one parenthesisation to another are equal. The coherence theorem makes an even stronger claim for all n, which is that every (well-behaved1) diagram comprised of associators and unitors commutes. In other words, all isomorphisms formed from associators and unitors (potentially tensored with identities) between two parenthesisations are equal.
Theorem 2.3 (Coherence Theorem). Let x1, . . . , xn ∈ C, and let P1,P2 be two parenthesisations of the product x1 . . . xn with arbitrary insertions of the unit 1C.
Let f, g : P1 → P2 be two isomorphisms formed from the composition of associators, unitors and their inverses (potentially tensored with identity morphisms). Then f = g.
We will omit the proof of this theorem here. It can be found as Theorem 2.9.2 in [EGNO15]. This is one of the most fundamental results in monoidal category theory, and importantly it allows us to suppress associators and unitors from our morphisms. For example, if we have a morphism f : xx → x, we can write a diagram
f⊗idx xxx xx
idx ⊗f
The reader may insert their favourite isomorphism (xx)x −→∼ x(xx) where re- quired to make this formally correct, and any possible composition of associators and unitors will in fact be the same isomorphism.
1We cannot say that every diagram commutes, as two formally different vertices may be equal in a certain category in such a way that breaks the commutativity. However, so long as every vertex contains the same ordered sequence of objects (plus units) this issue does not arise. 6 CHAPTER 2. BACKGROUND
2.1.3 String Diagrams
String diagrams are a useful graphical tool for describing monoidal categories, and can help make certain arguments more intuitive. In this section, we let C be a monoidal category, and we will suppress associators and unitors as permitted by the coherence theorem. We denote a morphism f : x → y by a string with a box labelled f:
y
f
x
Note that we read the morphism from bottom to top. We then concatenate vertically to denote composition, concatenate horizontally to denote the tensor product, and use a plain line to denote the identity morphism:
x g
= g ◦ f f g = f ⊗ g idx = f x
We will often omit labels of objects (as above) when the object is either unam- biguous or arbitrary.
A general function f : x1 . . . xn → y1 . . . ym from and to a tensor product can be denoted by multiple lines going into the morphism:
y1 ...ym
f
x1 ... xn
We will denote the identity on 1C by the omission of a line, and so a morphism f : 1 → x is denoted as x
f .
2.1.4 Monoidal Functors and Monoidal Equivalences
Definition 2.4. Let C and D be monoidal categories. A monoidal functor from C to D is • a functor F : C → D, 2.1. MONOIDAL CATEGORIES 7
• an isomorphism ∼ : 1D −→ F (1C) called the unitor, and • a natural isomorphism µ with components of the form
∼ µx,y : F (x)F (y) −→ F (xy)
called the tensorator, satisfying the following conditions: • (Associativity) The following diagram commutes for all x, y, z ∈ C, where aC and aD are the associators in C and D:
aD (F (x)F (y))F (z) F (x),F (y),F (z) F (x)(F (y)F (z))
µx,y idF (z) idF (x) µx,y F (xy)F (z) F (x)F (yz)
µxy,z µx,yz
F ((xy)z) C F (x(yz)) F (ax,y,z)
• (Unitality) The following diagrams commutes for all x ∈ C, where lC, lD, rC, and rD are the left and right unitors in C and D:
D lF (x) 1DF (x) F (x)
C −1 idF (x) F (lx)
F (1C)F (x) F (1Cx) µ1C,x and D rF (x) F (x)1D F (x)
C −1 idF (x) F (rx )
F (x)F (1C) F (x1C) µx,1C We can weaken this definition slightly. A lax monoidal functor has the same definition as for a monoidal functor, except we do not require and µx,y to be isomorphisms. An oplax monoidal functor has instead a morphism : F (1C) → 1D and a natural transformation µx,y : F (xy) → F (x)F (y), satisfying appropriately modified associativity and unitality conditions. Mostly in this thesis we will discuss monoidal functors (with isomorphisms for the constraints), the exception being §4.4. 8 CHAPTER 2. BACKGROUND
Definition 2.5. A monoidal functor F is called a monoidal equivalence if it is an equivalence of ordinary categories.
2.1.5 Braided Monoidal Categories
A braided monoidal category can be thought of as a monoidal category where the tensor product is commutative; as always, we mean this only up to isomorphism.
Definition 2.6. A braided monoidal category is a monoidal category C with a natural isomorphism ∼ βx,y : xy −→ yx called the braiding, such that the following two hexagon diagrams commute for all x, y, z ∈ C:
β x(yz) x,yz (yz)x
ax,y,z ay,z,x
(xy)z y(zx)
βx,y idz idy βx,z (yx)z y(xz) ay,x,z and β (xy)z xy,z z(xy) −1 −1 ax,y,z az,x,y
x(yz) (zx)y
xβy,z βx,zy
x(zy) −1 (xz)y ax,z,y
As a string diagram, the braiding is denoted by crossed strings:
β = β−1 =
The use of the opposite crossing as the inverse braiding is intuitive, as it amounts to the assertion that
= = 2.1. MONOIDAL CATEGORIES 9
We can now describe the hexagon diagrams as string diagrams:
= =
Definition 2.7. Let C and D be braided categories with braidings βC and βD respectively. A monoidal functor (F, µ) is called braided if the following diagram commutes for all x, y ∈ C:
βD F (x)F (y) F (x),F (y) F (y)F (x)
µx,y µy,x
F (xy) C F (yx) F (βx,y)
Given any monoidal category C, we can construct a braided category from the objects in C which can be braided past any other object.
Definition 2.8. Let C be a monoidal category. The (Drinfeld) centre of C is the category Z(C) with objects (z, γ), where z ∈ C and
∼ γx : xz −→ zx is a natural isomorphism called a half-braiding, denoted in string diagrams by
γx = , x z such that the following holds for all x, y ∈ C:
= .
x y z x y z
0 0 0 A morphism (z, γ) → (z , γ ) is a morphism f ∈ C(z, z ) such that (f idx) ◦ γx = 0 γx ◦ (idx f) for all x ∈ C, which as string diagrams is
z0 x z0 x
f = . f
x z x z 10 CHAPTER 2. BACKGROUND
It is straightforward to verify that the Drinfeld centre of a monoidal category C is a braided monoidal category with the associativity morphism inherited from C, and braiding given by 0 β(z,γ),(z0,γ0) := γz. For any monoidal category D, there is a forgetful functor R : Z(D) → D which takes (z, γ) to z.
2.1.6 Rigid Monoidal Categories
Let C be a monoidal category, and let x ∈ C.
Definition 2.9. An object x∗ ∈ C is said to be a left dual of x if there exist ∗ ∗ morphisms evx : x ⊗ x → 1 and coevx : 1 → x ⊗ x , called evaluation and coevaluation respectively, such that the compositions
x coevx ⊗ idx x ⊗ x∗ ⊗ x idx ⊗ evx x
id ∗ ⊗ coev ev ⊗ id ∗ x∗ x x x∗ ⊗ x ⊗ x∗ x x x∗ are the identity morphisms.
Definition 2.10. An object ∗x ∈ C is said to be a right dual of x if there exist ∗ ∗ morphisms evx : x ⊗ x → 1 and coevx : 1 → x ⊗ x, called evaluation and coevaluation respectively, such that the compositions
x idx ⊗ coevx x ⊗ ∗x ⊗ x evx ⊗ idx x
coevx ⊗ id∗ id∗ ⊗ evx ∗x x x∗ ⊗ x ⊗ x∗ x ∗x are the identity morphisms.
It is clear that if x∗ is a left dual of x, then x is a right dual of x∗, and vice versa. It is fairly simple to show that if x has a left (resp. right) dual, then it is unique up to unique isomorphism. When working with duals, string diagrams are directed, with an upward arrow indicating an object and a downward arrow indicating its dual. The evaluation and coevaluation maps are written as directed cups and caps:
Left duals: ev = coev =
Right duals: ev = coev = 2.2. LINEAR CATEGORIES 11
Then the conditions on compositions amount to the assertion that we can “straighten”:
Left duals: = =
Right duals: = =
Definition 2.11. A monoidal category C is said to be rigid if every object in C has both left and right duals.
Definition 2.12. Since left duals are unique up to unique isomorphism, when C is rigid we can define a contravariant left duality functor (−)∗ : C → C which sends x to x∗ and sends f : x → y to
x∗
f ∗ := f .
y∗
This gives a monoidal functor (−)∗ : Cop → Cmop, where Cmop is the monoidal category with the same underlying category as C, but opposite tensor product.
That is, x ⊗Cmop y = y ⊗C x.
Definition 2.13. A pivotal category is a rigid category C together with a natural ∼ ∗∗ isomorphism ψ : idC −→ (−) .
2.2 Linear Categories
2.2.1 Additive and Linear Categories
Definition 2.14. A category C is called additive if • Every set C(x → y) has the structure of an abelian group, and composition is a bi-homomorphism with respect to this structure, • There exists a zero object 0 such that C(0 → 0) = 0, and
• All finite biproducts exist. That is, for all x1, x2 ∈ C there exists a p ∈ C and
morphisms p1 : p → x1, p2 : p → x2, i1 : x1 → p and i2 : x2 → p such that 12 CHAPTER 2. BACKGROUND
· p1 ◦ i1 = idx1 ,
· p2 ◦ i2 = idx2 ,
· p2 ◦ i1 = 0,
· p1 ◦ i2 = 0, and
· i1 ◦ p1 + i2 ◦ p2 = idp.
Such a biproduct is called a direct sum and denoted by x1 ⊕ x2. It is unique up to unique isomorphism. If C is monoidal, then we also require that ⊗ is a bi-homomorphism with respect to the abelian group structure on the sets C(x → y).
Let k be a field. Definition 2.15. An additive category C is called k-linear if every set C(x → y) has the structure of a vector space over k, and composition is bilinear with respect to this structure. If C is monoidal, then we also require that ⊗ is bilinear with respect to the vector space structure on the sets C(x → y). Definition 2.16. Let F : C → D be a functor between two additive categories. We say F is additive if the associated maps C(x → y) → D(F (x) → F (y)) are homomorphisms of abelian groups. If C and D are linear categories, then we say F is linear if the homomorphisms above are linear. Since the direct sums in an additive category are biproducts, the hom-sets in an additive category are distributive with respect to direct sums. That is, there is a canonical isomorphism C(x ⊕ y → z) −→C∼ (x → z) ⊕ C(y → z), and similarly if the direct sum is the target. It follows that if F : C → D is an additive functor, then there is a natural isomorphism F (x ⊕ y) −→∼ F (x) ⊕ F (y), corresponding to the image of idx⊕y in C(x ⊕ y → x ⊕ y) ∼= C(x ⊕ y → x) ⊕ C(x ⊕ y → y).
Example 2.17. Both Vec and RepG are linear braided monoidal categories, and are rigid with duals given by the usual duals of vector spaces and representations respectively. They also have a canonical pivotal structure coming from the natural isomorphism of vector spaces V −→∼ V ∗∗. 2.2. LINEAR CATEGORIES 13
2.2.2 Additive Completion
A category C is called pre-additive if every set C(x → y) has the structure of an abelian group, and composition is a bi-homomorphism with respect to this structure. Given such a category, we can formally add direct sums in order to construct an additive category Mat(C) such that there is a fully faithful functor C → Mat(C).
Definition 2.18. The additive completion Mat(C) of a pre-additive category C is the category with Ln • objects given by formal sums i=1 ci (including the empty sum), Ln Lm • for objects c = i=1 ci and d = j=1 dj, a morphism c → d is given by a m × n matrix M, where each entry Mji is a morphism Mji : ci → dj, • composition of morphism is given by “multiplication” of matrices, except using composition in C instead of multiplication of scalars: X (M ◦ N)ki = Mkj ◦ Nji, j • addition of morphisms is given by addition of matrices.
Remark 2.19. It is worth checking that taking the additive completion of a cate- gory in which some direct sums exist respects those direct sums. Indeed, if x1 ⊕ x2 exists in a pre-additive category C, then it is isomorphic to the formal sum of x1 and x by the maps 2 " # h i p1 i1 i2 and . p2 Remark 2.20. Let C be a pre-additive category, let D be an additive category, and let F : C → D be a functor which is a homomorphism on hom-groups (this is precisely the definition of an additive functor, except we do not require C to be additive). Then F extends to an additive functor Mat(C) → D, which takes the formal direct sum x⊕y to F (x)⊕F (y). The analogous result holds for constructing linear functors Mat(C) → D.
Let C and D be linear categories. We have two obvious (though somewhat naive) constructions.
Definition 2.21. The direct sum C ⊕ D is the additive completion of the category with objects given by pairs (c, d) for c ∈ C and d ∈ D, and hom-spaces
C ⊕ D((c1, d1) → (c2, d2)) := C(c1 → c2) ⊕ D(d1 → d2). 14 CHAPTER 2. BACKGROUND
Definition 2.22. The naive tensor product C ⊗D is the additive completion of the category with objects given by pairs (c, d) for c ∈ C and d ∈ D, and hom-spaces
C ⊗ D ((c, d) → (c0, d0)) := C(c → c0)D(d → d0).
2.2.3 Idempotent Complete Categories and the Karoubi Envelope
An idempotent in some category C is an endomorphism p : c → c such that p◦p = p. We would often like an idempotent to have something analogous to an image, on which the idempotent is the identity. Such an idempotent is called split.
Definition 2.23. An idempotent p : c → c is split if there exists r : c → b and s : b → c such that s ◦ r = p and r ◦ s = idb. The object b is denoted im(p). A category is said to be idempotent complete if every idempotent is a split idempotent.
For an arbitrary category C we can construct an idempotent complete category Kar(C) called the Karoubi envelope of C, for which there is a fully faithful functor I : C → Kar(C). This is also called the idempotent completion of C.
Definition 2.24. The Karoubi envelope Kar(C) of a category C has as objects pairs (c, p) where c ∈ C and p : c → c is an idempotent. The morphisms (c, p) → (c0, p0) are given by morphisms f : c → c0 such that f ◦ p = f = p0 ◦ f. Composition is given simply by the composition of morphisms in C, and the identity morphism on (c, p) is given by p. If C is monoidal, then Kar(C) is also monoidal, with
(c, p) ⊗ (c, p0) := (cc0, pp0).
The fully faithful functor I : C → Kar(C) is given by sending c to (c, idc), which is monoidal if C is monoidal. Thus we can consider C to be a full subcategory of Kar(C).
2.2.4 Semisimple Categories
The following is not the conventional definition of semisimplicity, which is defined only for abelian categories, though it is equivalent in that case. We discuss this further in Chapter 5.
Definition 2.25. A k-linear monoidal category C is called (M¨uger)semisimple if: 2.2. LINEAR CATEGORIES 15
• all hom-spaces C(x → y) are finite-dimensional; •C is idempotent complete;
• there exist objects xi ∈ C indexed by a set I such that k if i = j C(xi → xj) = 0 if i 6= j and such that for any pair a, b ∈ C, the composition law M C(xi → b) ⊗ C(a → xi) → C(a → b) i∈I
is an isomorphism. The objects xi are called simple objects. A semisimple category C is said to be finitely semisimple if it has finitely many simple objects.
We will often also call an object simple if it is isomorphic to an object xi, in which case the definition of finitely semisimple should say that it requires finitely many isomorphism classes of simple objects. Remark 2.26. This definition implies that every object in a semisimple category
C is a direct sum of simple objects. To see this, we note that C(xi → c) and
C(c → xi) are dual vector spaces for every simple xi and c ∈ C; the coevaluation map
C(c → xi)C(xi → c) → k is given by sending f ⊗ g to f ◦ g, noting that C(xi → xi) = k (that is to say, all morphisms xi → xi are scalar multiples of the identity idxi ). P Under the composition law isomorphism, idc corresponds to i∈I ei for ei ∈ C(xi → c)C(c → xi). The evaluation map
k → C(xi → c)C(c → xi) is given by sending idxi = 1 ∈ k to ei. This indeed satisfies the straightening relations, and so we see that C(c → xi) and C(xi → c) are dual vector spaces. ∗ Hence if we choose a basis {αi,j} of C(c → xi), we have a dual basis {αi,j} of
C(xi → c). When we consider the meaning of the coevaluation and evaluation maps, this means that id if i = i and j = j ∗ xi1 1 2 1 2 αi1,j1 ◦ αi2,j2 = 0 otherwise
X ∗ αi,jαi,j = idc . i,j 16 CHAPTER 2. BACKGROUND
If we define
[c : xi] := dim(C(c → xi)) = dim(C(xi → c)), then this exactly means that . . M ⊕[c:xi] αi,j : c → xi . . i∈I is an isomorphism with inverse h i ∗ M ⊕[c:xi] . . . αi,j ... : xi → c. i∈I
We can therefore think of [c : xi] as the number of copies of xi in c. Note that ∼ if c = d, then [c : xi] = [d : xi], and so this decomposition into a direct sum of simple objects is unique.
Remark 2.27. In particular, we can consider a semisimple category to be the additive completion of its full subcategory of simple objects, and so by Remark 2.20 to define a linear functor from a semisimple category it suffices to define it on the full subcategory of simple objects.
Proposition 2.28. A finitely semisimple category C is equivalent to Vec⊕n, where n is the number of simple objects in C.
⊕n Proof. Observe that Vec is clearly finitely semisimple, with simple objects ki which are a copy of k in the ith copy of Vec. We can define a linear functor ⊕n C → Vec which sends a simple object xi to ki. This is essentially surjective since every simple object in Vec⊕n is in the image of the functor. The hom- spaces between simple objects are either k or 0, and so it is fully faithful on simple objects since it must take idxi to idki . Hence it is fully faithful, since all hom-spaces decompose into the direct sum of hom-spaces between simple objects.
Definition 2.29. Let k be an algebraically closed field. A fusion category is a finitely semisimple rigid monoidal k-linear category C in which 1C is a simple object.
Examples 2.30. Let k be an algebraically closed field. i) The category Vec is a fusion category a single simple object k. ii) For any finite group G, RepG is a fusion category with simple objects the irreducible representations. 2.3. (CO)ALGEBRAS AND (CO)MODULES 17 2.3 (Co)Algebras and (Co)Modules
2.3.1 Monoids and Algebras
We can generalise the notion of a monoid to describe a type of object in any monoidal category. Recall that a monoid is defined to be a set with an associative multiplication and a unit for that multiplication. The following is a generalisation of this definition.
Definition 2.31. A monoid object (or simply monoid) in a monoidal category C is an object a together with • a morphism µ : aa → a called multiplication, and
• a morphism η : 1C → a called the unit such that the following associativity and unitality axioms hold:
a a a a a
µ µ µ µ = and = = µ µ η η
a a a a a a a a a
Examples 2.32. a) Monoid objects in Set with the Cartesian product are precisely monoids in the usual sense.
b) In Ab, the category of abelian groups with ⊗Z as the tensor product, the monoid objects are rings.
c) The monoid objects in Veck are the (finite-dimensional) k-algebras. For this reason, monoid objects in k-linear categories are more commonly referred to as algebra objects (or simply algebras).
d) Let G be a finite group. The object Fun G ∈ RepG of functions G → k (where the G-action is given by g · f(h) = f(gh)) is an algebra in RepG, with multiplication given by pointwise multiplication of functions and unit given by the map k → Fun G which sends c ∈ k to the constant function c(g) = c for all g.
We will develop this last example further over the next sections, as it is used extensively in Chapter 4.
Definition 2.33. An algebra (A, µ) in a braided category C is called commutative 18 CHAPTER 2. BACKGROUND
if µ = µ ◦ βA,A, where β is the braiding in C. That is, if
A A
µ = µ .
A A A A
2.3.2 Module Objects
Just as we can have modules over rings in the usual sense, we can also have modules over monoid objects in monoidal categories (recalling that a ring is a monoid object in Ab). We can think of a right module over a ring R as an abelian group M with a “right multiplication by R” M ⊗Z R → M which respects multiplication and addition in R. We formalise the generalisation of this in the definition below. Let C be a monoidal category and let a ∈ C be a monoid. Definition 2.34. A (right) module over a in C is an object m ∈ C, together with a morphism · : ma → m (called the action of a on m), such that the following associativity and unitality axioms hold: m m m m
· · · = and = µ · η
m a a m a a m m There is a corresponding notion of a left module, where the monoid acts from the left. From here we will simply refer to modules over a monoid, which should be taken to mean right modules unless otherwise specified. Example 2.35. In Ab, the monoid objects are rings. Correspondingly, the module objects over a ring R are precisely the modules over R.
2.3.3 Coalgebras and Comodules
There is also a notion dual to that of algebras and modules. We will assume C is linear to justify the use of the word coalgebra, though the definitions in this section do not require that assumption. Let C be a linear monoidal category. Definition 2.36. A coalgebra in C is an algebra in Cop (the category with the same objects as C but formally reversed morphisms). That is, a coalgebra is an object A ∈ C together with 2.3. (CO)ALGEBRAS AND (CO)MODULES 19
• a morphism ∆ : A → AA called comultiplication,
• a morphism ε : A → 1C called the counit, satisfying appropriate coassociativity and counitality conditions (which are dual to the associativity and unitality axioms for algebras).
A coalgebra (A, ∆) in a braided category C is called commutative if ∆ = βA,A ◦ ∆, where β is the braiding in C.
Let A be a coalgebra in C.
Definition 2.37. A (right) comodule for A is a right module for A as an algebra in Cop. That is, a right comodule for A is an object c ∈ C together with a morphism α : c → cA (called the coaction of A on c) satisfying appropriate coassociativity and counitality conditions (which are dual to the associativity and unitality axioms for right modules).
Example 2.38. For G a finite group, Fun G ∈ RepG is a coalgebra with comulti- P plication which sends f to g f(g)δg ⊗ δg, where δg is the indicator function on g. P The counit sends f to g f(g). This is not the usual coalgebra structure given to Fun G, but it gives Fun G the structure of a separable Frobenius algebra, which we will define and use in Chapter 4. The usual coalgebra structure on Fun G has comultiplication X δg 7→ δgk ⊗ δk−1 , k∈G and counit given by evaluation at the group identity; this gives Fun G the structure of a Hopf algebra (which we do not discuss in this thesis).
2.3.4 Categories of Modules and Comodules
Let C be a linear monoidal category.
Definition 2.39. Given an algebra A ∈ C, we define the category of right modules
Cmod-A to be the category with objects (m, p) (where p : mA → m is the action of A on m) which are right modules for A, and morphisms (m1, p1) → (m2, p2) are given by f ∈ C(m1 → m2) such that
f ◦ p1 = p2 ◦ (f idA).
We can similarly define the category A-modC of left modules for A. Given a coalgebra A ∈ C, we can define the category of right (respectively, left) comodules Ccomod-A (respectively, A-comodC) analogously. 20 CHAPTER 2. BACKGROUND
Remark 2.40. Suppose (A, µ, η) is an algebra in C. Then A∗ is a coalgebra, with comultiplication given by µ∗ and counit given by η∗ as defined in Definition 2.12. Since the left duality functor is a monoidal functor Cop → Cmop, it induces a monoidal functor ∗ Cmod-A → A -comodC, which sends (m, p) to (m∗, p∗). This in fact defines an equivalence, where the inverse functor is the right duality functor. Example 2.41. Recall that when G is a finite group, Fun G ∈ RepG is an algebra and a coalgebra with
• multiplication µ(f1 ⊗ f2)(g) = f1(g)f2(g), • unit η which sends c ∈ k to the constant function with value c, P • comultiplication ∆(f) = g f(g)δg ⊗ δg where δg is the indicator function on g, and P • counit ε(f) = g f(g). We note that Fun G is self-dual, and that if we dualise the algebra structure on Fun G we get the coalgebra structure on Fun G. Hence we have ∼ Cmod- Fun G = Fun G-comodC. We can also note that Fun G is commutative both as an algebra and as a coalgebra. Definition 2.42. An central algebra (respectively, coalgebra) A with half-braiding β is the image under the forgetful functor R : Z(C) → C of an algebra (A, β) ∈ Z(C). A central algebra (respectively, coalgebra) is commutative if it is commuta- tive in Z(C). Given a central algebra A ∈ C with half-braiding β and a right A-module (m, p), we can view it as a left module with action
β p Am −→m mA −→ m.
Given a right A-module (m, p) and a left A-module (n, q), the tensor product of m and n over A is defined to be the coequaliser of the diagram
p idn mAn mn m ⊗A n. idm q
When A is a central algebra, the category of right modules Cmod-A is monoidal, with the tensor product given by the tensor product over A, considering the right modules as left modules where needed. Once again, we can make analogous definitions for the categories of left modules, and of left/right comodules. 2.3. (CO)ALGEBRAS AND (CO)MODULES 21
Proposition 2.43. Let A ∈ C be an algebra in an idempotent complete category
C. Then Cmod-A is idempotent complete.
Proof. We need to exhibit a splitting of an arbitrary idempotent. An idempotent e ∈ Cmod-A((m, p) → (m, p)) is given by a idempotent e ∈ C(m → m) such that e ◦ p = p ◦ (e idA). Since C is idempotent complete, we can take im(e) and define the right action of A by im(e)
r p s
im(e) A where s : im(e) → m and r : m → im(e) are the maps which exhibit im(e) as a splitting of e. It is then straightforward to check that this defines a right action on im(e), and so e is a split idempotent.
The analogous result holds for Ccomod-A when A is a coalgebra in an idempo- tent complete category C.
2.3.5 Free Comodules
Definition 2.44. Let A ∈ C be an algebra. The category Cfcomod-A of free comodules has the same objects as C, and hom-spaces
Cfcomod-A(x → y) := C(xA → y).
For f ∈ Cfcomod-A(x → y) and g ∈ Cfcomod-A(y → z), their composition is given by
z g z y f g f ◦ = . ∆
y A x A x A
When A is a central algebra, Cfcomod-A is monoidal with tensor product on 22 CHAPTER 2. BACKGROUND objects inherited from the tensor product in C, and on morphisms given by
y1 y2
y1 y2 f1 f2
f1 f2 ⊗ = . ∆
x1 A x2 A x1 x2 A
Definition 2.45. We will write Cpcomod-A for the idempotent completion
Kar(Cfcomod-A).
We can make analogous definitions for the categories of left free comodules and left/right modules.
2.4 Monoidal Module Categories
2.4.1 Module Categories
Let C be a monoidal category.
Definition 2.46. A (right) module category over C is • a category M, • a functor · : M × C → M called the action of C on M, • a natural isomorphism α with components
∼ αm,x,y : m · (xy) −→ (m · x) · y
for all x, y ∈ C and m ∈ M, called the associator, • a natural isomorphism e with components
∼ em : m · 1C −→ m
for all m ∈ M, called the unitor, such that the following two diagrams commute: • the triangle identity:
αm,1,x m · (1Cx) (m · 1) · x
idm ·lx em·idx x · m 2.4. MONOIDAL MODULE CATEGORIES 23
• the pentagon identity: m · ((xy)z)
idm ·ax,y,z αm,xy,z
m · (x(yz)) (m · (xy)) · z
αm,x,yz αm,x,y·idz (m · x) · (yz) ((m · x) · y) · z αm·x,y,z Example 2.47. If C and M are monoidal categories, then a monoidal functor F : C → M gives M the structure of a right module over C where m · c = mF (c), and using the tensorator as the associator. It also gives M the structure of a left module over C where c · m = F (c)m. Example 2.48 (Linear categories as modules over Vec). Given any linear category
C, there is a monoidal linear functor Vec → C determined by sending k to 1C. Hence we have both left and right actions of Vec on C, and so can write V · c and c · V for c ∈ C,V ∈ Vec. Observe that since the functor Vec → C is linear, we have V · c ∼= c⊕ dim V ∼= c · V. Indeed, this is an example of a monoidal module category, as we will see in the next section. We can use the example above to define both a left and right adjoint to any linear functor between two fusion categories. We can state this in slightly more generality.
Lemma 2.49. Let k be an algebraically closed field, let C be a finitely semisimple category, and let D be a semisimple category. Then for any linear functor F : C → D, F has a right adjoint R defined on simple objects by M R(yj) := xi ·D(F (xi) → yj),
xi∈C simple and a left adjoint L defined on simple objects by M L(yj) := xi ·D(yj → F (xi)).
xi∈C simple Proof. To see that this is a right adjoint, it suffices to check for simple objects in
C and D. Let xi ∈ C and yj ∈ D be simple objects. Then we have
C(xi → R(yj)) = C(xi → xi ·D(F (xi) → yj)) ∼ = C(xi → xi) ⊗ D(F (xi) → yj) ∼ = D(F (xi) → yj), 24 CHAPTER 2. BACKGROUND
and these are natural in xi and yj. The proof for the left adjoint is analogous.
2.4.2 Monoidal Module Categories
If C and M are monoidal categories, then a monoidal functor F : C → M gives M the structure of a right module over C where m · c = mF (c), and using the tensorator as the associator. However, if C is braided, then we can define a different right module structure on M, where m · c = F (c)m, and the associator is given by first using the braiding in C and then the tensorator. For these two structures to be equivalent, we need F (c)m and mF (c) to be isomorphic in a way which is compatible with the braiding in C. In other words, we need that F have the structure of a central functor.
Definition 2.50. A central functor is a functor F : C → D with a factorisation
Z F : C −−→ZF (D) −→DR where F Z is braided and R is the forgetful functor.
Definition 2.51. Let V be a braided monoidal category. A (right) monoidal module category over V is a monoidal category C with a central functor F : V → C. A left monoidal module category over V is a monoidal category C with a central functor F : Vrev → C. 3 Ladder Categories
In this chapter, all categories and functors are assumed to be linear. For C a rigid monoidal category, M a right C-module category, and N a left C-module category, we build the ladder category M ⊗C N . In the case of monoidal module categories over a braided rigid monoidal category, we describe an explicit construction of a monoidal structure on the ladder category. We then develop some basic results about monoidal ladder categories. Ladder categories can be considered to be a special case of the tensor product of n-categories construction as described in section 6 of [MW12]. An explicit definition in the case where the modules are over a fusion category is given in [BBJ19]. In this thesis we generalise this definition to simply require the category be rigid (and linear), and define a monoidal structure on the ladder category in the case of monoidal modules over a braided rigid monoidal category.
3.1 Ladder Categories
3.1.1 Definition of Ladder Categories
Let C be a rigid monoidal category, let M be a right C-module category, and let N be a left C-module category.
Definition 3.1. The ladder category M ⊗C N is the additive completion of the category with objects given by pairs (m, n) for m ∈ M and n ∈ N , and
!, 0 0 M 0 0 M ⊗C N ((m, n) → (m , n )) := M(m · c → m )N (n → c · n ) K, c∈C where K is a subspace we will describe shortly. For f ∈ M(m · c → m0) and g ∈ N (n, c · n0), we will write (f, c, g) for their tensor product in M(m · c → m0)N (n, c · n0), and call such a morphism pure (so that an arbitrary morphism is a sum of pure morphisms).
25 26 CHAPTER 3. LADDER CATEGORIES
We can depict a pure morphism (f, c, g):(m, n) → (m0, n0) as
m0 n0
f c , g
m n where we consider the left side of the ladder to be in M and the right side of the ladder to be in N . For morphisms, we require
m0 n0 m0 n0
f h f = h , g g
m n m n by letting K be the span of morphisms of the form
0 (f ◦ (idm ·h), c, g) − (f, c , (h · idn0 ) ◦ g) for all h ∈ C(c → c0), f ∈ M(m · c0 → m0), and g ∈ N (n → c · n0). Hence we can treat a morphism h ∈ C(c → c0) as not belonging to either side of the ladder, but rather to the middle:
m0 n0
f h . g
m n The reader can choose to interpret this diagram as having h associate to the left or to the right as preferred, and this requirement ensures that both interpretations are equal.
The identity on (m, n) is given by (idm, id1C , idn), which in string diagrams is m n
.
m n Composition of pure morphisms is given by
0 0 0 0 0 0 (f , c , g ) ◦ (f, c, g) = (f ◦ (f · idc0 ), cc , (idc ·g ) ◦ g), 3.1. LADDER CATEGORIES 27 which in string diagrams is simply vertical concatenation:
m00 n00 m0 n0 m00 n00
0 0 0 f c f c0 f c f . ◦ = 0 0 g g g c g
m0 n0 m n m n ∼ Remark 3.2. We can observe that (m1 ⊕ m2, n) = (m1, n) ⊕ (m2, n) (where this second direct sum is the formal sum introduced by taking the additive completion) by the maps " # h i (p1, 1C, idn) (i1, 1C, idn)(i2, 1C, idn) and , (p2, 1C, idn) where i1, i2, p1, p2 are the inclusion and projection maps as described in Definition ∼ 2.14. Similarly we have (m, n1 ⊕ n2) = (m, n1) ⊕ (m, n2). Proposition 3.3. Suppose f ∈ M(m · c → m0) and g ∈ N (n → c · n0) are 0 0 isomorphisms. Then (f, c, g):(m, n) → (m , n ) is an isomorphism in M ⊗C N , with inverse m n
f −1 (f,˜ c∗, g˜) := c , g−1
m0 n0 where we are using the existence of a right dual for c (as C is rigid). Proof. The composition (f,˜ c∗, g˜) ◦ (f, c, g) is given by m n m n m n m n
f −1 f c c g−1 = g−1 = g−1 = . c c g g g m n m n m n m n −1 The first equality is simply that f ◦f = idm·c. Next, we use one of the “straighten-
= −1 ing” identities for right duals: . The final equality is simply that g ◦g = idn. ˜ ∗ Hence (f, c , g˜) ◦ (f, c, g) = id(m,n). ˜ ∗ The proof that (f, c, g) ◦ (f, c , g˜) = id(m0,n0) is analogous. Corollary 3.4. There is a canonical isomorphism (m, c · n) −→∼ (m · c, n) for all m ∈ M, n ∈ N , and c ∈ C.
Proof. Take (idm·c, c, idc·n), which is an isomorphism by Proposition 3.3. 28 CHAPTER 3. LADDER CATEGORIES
3.1.2 Ladder Categories Over Semisimple Categories
Let V be a semisimple rigid category with simple objects {xi}i∈I , let M be a right C-module category, and let N be a left C-module category. 0 0 For any pure morphism (f, c, g): M ⊗C N ((m, n) → (m , n )), we have from Remark 2.26 X ∗ idc = αi,j ◦ αi,j, i,j ∗ where {αi,j}j are a basis for C(c → xi) and {αi,j}j are a dual basis for C(xi → c) under the evaluation and coevaluation maps defined in Remark 2.26. Hence we can write
m0 n0 m0 n0
f c X f xk ∈ L M(m · x → m0)N (n, x · n0). = α∗ k k g i,j g k∈I i,j αi,j m n m n
Therefore in the definition of morphisms in M ⊗C N , taking the direct sum over simple objects is the same as taking the direct sum over all objects (after taking respective quotients), giving !, 0 0 M 0 0 ˜ M ⊗C N ((m, n) → (m , n )) = M(m · xk → m )N (n, xk · n ) K. k∈I
Here K˜ is the restriction of K to the direct sum over just simple objects, since all of the relations can also be split into maps with simple objects in the middle, and this decomposition gives morphisms in K˜ .
3.2 Monoidal Ladder Categories
3.2.1 Definition of Monoidal Ladder Categories
Let V be a braided rigid monoidal category, let F : V → C be the structure of a right monoidal V-module C, and let G : Vrev → D be the structure of a left monoidal V-module D. Recall that F is by definition a central functor, and so comes equipped with a braided monoidal functor F Z : V → Z(C). For any v ∈ V, we define σv to be the half-braiding given by F Z (v) = (F (v), σv). We do likewise for the central functor G, defining γv by GZ (v) = (G(v), γv). 3.2. MONOIDAL LADDER CATEGORIES 29
We will often write v rather than F (v) as an object in C, and likewise will write v ∈ D rather than G(v).
Proposition 3.5. The ladder category C ⊗V D has the structure of a monoidal category, with tensor product on objects given by (c, d) ⊗ (m, n) = (cm, dn), and on pure morphisms given by
c0 d0 m0 n0 c0 m0 d0 n0
f v h w w ⊗ = , g i v
c d m n c m d n where the crossings are defined as
v w := σm := γd0 . m v d0 w
This is to say that
v w (f, v, g) ⊗ (h, w, i) := (fh ◦ idc σm idw, vw, idv γd0 idn0 ◦gi).
Proof. We need to show that the tensor product is functorial, which is to say that
((f 0, v0, g0) ◦ (f, v, g)) ⊗ ((h0, w0, i0) ◦ (h, w, i)) = ((f 0, v0, g0) ⊗ (h0, w0, i0)) ◦ ((f, v, g) ⊗ (h, w, i)).
Evaluating the left side, we obtain
c00 d00 m00 n00
f 0 v0 h0 w0 ⊗ c00 d00 m00 n00 g0 i0 f 0 h0 c0 d0 m0 n0 f h v0 w0 ◦ ◦ = ⊗ v w c0 d0 m0 n0 g0 i0 g i f v h w ⊗ c d m n g i
c d m n 30 CHAPTER 3. LADDER CATEGORIES
c00 m00 d00 n00
= .
c m d n
Evaluating the right side, we obtain
c00 d00 m00 n00 c00 m00 d00 n00
f 0 v0 h0 w0 w0 ⊗ g0 i0 v0
c0 d0 m0 n0 c0 m0 d0 n0
◦ ◦ = ◦
c0 d0 m0 n0 c0 m0 d0 n0
f v h w w ⊗ g i v
c d m n c m d n c00 m00 d00 n00
= .
c m d n
To show that these morphisms are equal, we need to show that we can swap the middle two rungs of the ladder:
c00 m0 d00 n0 c00 m0 d00 n0
↔ .
c0 m d0 n c0 m d0 n 3.2. MONOIDAL LADDER CATEGORIES 31
Indeed, we have
c00 m0 d00 n0 c00 m0 d00 n0 c00 m0 d00 n0 c00 m0 d00 n0
0 f h
w = = = . v0
0 g i c0 m d0 n c0 m d0 n c0 m d0 n c0 m d0 n
−1 The first equality is a substitution idv0 idw = βw,v0 ◦ βw,v0 , where β is the braiding in V. Next, we consider v acting on C, and by the naturality of the half-braiding σv on F (v) we can swap the order of the braiding and h, which gives the second equality. Likewise, the third equality comes from considering the half-braiding γw0 on G(w0).
The unit object is given by (1C, 1D), while the associator
α(c1,d1),(c2,d2),(c3,d3) : ((c1, d1) ⊗ (c2, d2)) ⊗ (c3, d3) → (c1, d1) ⊗ ((c2, d2) ⊗ (c3, d3)) is given by taking the associator on both sides: