Hodge Star as Braided Fourier Transform

We study super-braided Hopf algebras $\Lambda$ primitively generated by finite-dimensional right crossed (or Drinfeld-Radford-Yetter) modules $\Lambda^1$ over a Hopf algebra $A$ which are quotients of the augmentation ideal $A^+$ under right multiplication and the adjoint coaction. Here super-bosonisation $\Omega=A\ltimes\Lambda$ provides a bicovariant differential graded algebra on $A$. We introduce $\Lambda_{max}$ providing the maximal prolongation, while the canonical braided-exterior algebra $\Lambda_{min}=B_-(\Lambda^1)$ provides the Woronowicz exterior calculus. In this context we introduce a Hodge star operator $\sharp$ by super-braided Fourier transform on $B_-(\Lambda^1)$ and left and right interior products by braided partial derivatives. Our new approach to the Hodge star (a) differs from previous approaches in that it is canonically determined by the differential calculus and (b) differs on key examples, having order 3 in middle degree on $k[S_3]$ with its 3D calculus and obeying the $q$-Hecke relation $\sharp^2=1+(q-q^{-1})\sharp$ in middle degree on $k_q[SL_2]$ with its 4D calculus. Our work also provided a Hodge map on quantum plane calculi and a new starting point for calculi on coquasitriangular Hopf algebras $A$ whereby any subcoalgebra $L\subseteq A$ defines a sub braided-Lie algebra and $\Lambda^1\subseteq L^*$ provides the required data $A^+\to \Lambda^1$.


Introduction
Differential exterior algebras Ω on quantum groups were extensively studied in the 1990s starting with [39] and have a critical role as examples of noncommutative geometry more generally. However, one problem which has remained open since that era is the general construction of a Hodge star operator in noncommutative geometry, even in the quantum group case. Until now the Hodge operator has been treated mainly in an ad-hoc manner in particular examples, motivated typically by 2 = ±id as a requirement, e.g. [33,7]. We also note a framework [9] based on a pair of differential structures and contraction with a generalised metric, and [12] in another q-deformation framework. By contrast our new approach depends canonically on the braided-Hopf algebra structure of the exterior algebra which applies at least for bicovariant calculi on quantum groups and covariant calculi on quantum-braided planes. Moreover, this new approach gives very different, and we think more interesting, answers than the previous approaches. Specifically, Section 3.2 includes the example of k(S 3 ), the function algebra on the permutation group S 3 , with its 2-cycle calculus, where in middle degree obeys 3 = id so that as a whole has order 6. We also cover electromagnetism on this finite group as in [33] but using the new Hodge star and again achieving a full diagonalisation of the Laplace-Beltrami operator dδ + δd. Similarly, Section 4.2 compute our Hodge operator for k q [SL 2 ] with its standard 4D calculus [39] and in middle degree we find, rather unexpectedly, that it obeys the well-known q-Hecke relation 2 = id + (q − q −1 ) when suitably normalised. Both of these examples are very different from requiring 2 = ±id. While the nicest version of the theory assumes a central bi-invariant metric and volume form, our Fourier approach is more general as illustrated in Section 3.1 on the quantum plane A 2 q . Conceptually, we adopt a novel point of view [29,17] on what the Hodge star is even classically. Namely, at every point of a manifold M of dimension n the exterior algebra of differential forms has fibre the exterior algebra Λ(R n ) with generators θ i = dx i in local coordinates and the usual 'Grassmann algebra' relations θ i θ j + θ j θ i = 0. This is a finite-dimensional super-Hopf algebra and we can apply a super version of usual Hopf algebra Fourier transform F : Λ → Λ * using the Berezin integral Vol = 1 where Vol = θ 1 · · · θ n . This then extends to the whole manifold and in the presence of a metric gives the classical Hodge operator : Ω m → Ω n−m . This point of view was also recently used for the Hodge star on supermanifolds [4]. The same approach applies to bicovariant differential exterior algebras on Hopf algebras, where we recall that these are parallelizable via an exterior algebra of left-invariant differential forms Λ forming a super-braided-Hopf algebra, so we can do super-Fourier transform on this.
In algebraic terms, Ω in the bicovariant case is a super-Hopf algebra equipped with a split super-Hopf algebra projection Ω H. Hence by a super-version of the Radford-Majid theorem [38,23] one knows that Ω ∼ =A· <Λ is the super-bosonisation of Λ as a super-braided Hopf algebra in the braided category of crossed (or Drinfeld-Radford-Yetter) modules [40,38,5,41], where we assume that A has invertible antipode. Moreover, in the standard setting Ω is generated by degrees 0,1 and hence Λ has primitive generators Λ 1 . As a result we can focus on such primitivelygenerated super-braided Hopf algebras Λ and translate elements of noncommutative geometry in these terms. Particularly, the exterior derivative restricts to d : Λ → Λ to make Λ into a differential graded algebra (but not within the category, d is not a morphism) and in Section 2 we give an explicit construction of this in the case Λ = Λ max corresponding to the maximal prolongation of Ω 1 . Any other exterior algebra corresponds to a quotient of this and at the other extreme we revisit the more well-known case Λ = Λ min = B − (Λ 1 ) given by the canonical (super) braided-linear space associated to an object of an Abelian braided category. This B ± (Λ 1 ) construction appeared in the case where Λ 1 is rigid in [27] while quadratic primitively generated braided-Hopf algebras appeared in [24,18]. The B + construction as an algebra is often called a 'Nichols algebra' cf [37] or 'Nichols-Woronowicz algebra' cf [39] but we note that neither of these works considered B ± as braided-Hopf algebras, that structure being introduced following the development of braided-Hopf algebras as above. The super-braided Hopf algebra interpretation of the Woronowicz construction of bicovariant differential exterior algebras was in [29,28] among other works. In this context the universal property of B − corresponds in some sense to the minimal relations needed to ensure Poincaré duality, a remark that will be reflected in our approach to the Hodge star. It was also observed recently [35] that the exterior derivative d on a bicovariant exterior algebra is not only a super-derivation but also a super-coderivation where | | is the degree operator and ∆ is the super-coproduct. This turns out to be key to going the other way of building (Ω, d) from data d on Λ. Although these matters are somewhat familiar, the braided approach to the differential structure requires proofs which we provide as part of a necessary systematic treatment. Section 2.4 similarly provides a canonical construction for the differential exterior algebra on B + (Λ 1 ) as a quantum-braided plane.
As part of this, and critical for Fourier transform on B ± (Λ 1 ), is the notion of braided-exponential in our approach to these algebras (being used notably in [27] to inductivively build up the quasitriangular structure of quantum groups U q (g) as a succession of q-exponentials). Here [27] are defined in terms of braided factorials [m, Ψ]! as in [24,18]. We recall this theory in Section 2.2. In the B + case we have previously proposed braided Fourier theory on the Fomin-Kirillov algebra and its super-version as Hodge star in [29], but without a systematic treatment.
Our central results appear in Section 3. If we have a unique bi-invariant top degree (say of degree n) then super-braided-Hopf algebra Fourier transform gives us a map F : Λ m → Λ * (n−m) and in the presence of a quantum metric a Hodge star map : Λ m → Λ n−m . This extends in the context above to the geometric : Ω m → Ω n−m . Proposition 3.8 establishes in some generality that commutes with the braided antipode S and is involutive in degrees 0, 1, n − 1, n. This in turn follows from some general results about super-braided Fourier transform in Section 3.1 which builds on our previous diagrammatic work, particularly [11]. Section 3.1 also covers the Hodge operator on the well-known quantum plane A 2 q , where Λ = B − (Λ 1 ) = A 0|2 q , the fermionic quantum plane.
In Section 4 we focus on the case of coquastriangular Hopf algebras (A, R) [5,21,18]. In line with the braided-Hopf algebra methods of the present paper, we first present a starting point for the construction of Λ 1 itself, namely as the dual of a braided-Lie algebra. We show that every subcoalgebra L ⊆ A is a braided-Lie algebra in the sense introduced in [26]. This gives a significantly cleaner result than previous attempts such as [8] and builds on our recent work [35]. Everything is worked out in detail for C q [SL 2 ] recovering previously 'R-matrix' formulae when we take the standard matrix subcoalgebra, including the 4D braided-Lie algebra [26] and the Woronowicz 4D calculus [39] from Λ = L * . We then compute the canonical braided-Fourier transform on the latter with results as described above.
We work over a general ground field k and q ∈ k × . In examples, we will assume characteristic zero for our calculations. We use the Sweedler notation ∆a = a (1) ⊗a (2) for coproducts and ∆ R v = v (0) ⊗v (1) for right coactions (summations understood). We denote the kernel of the counit by A + and π : A → A + defined by π (a) = a − (a) is the counit projection. We will make extensive use of the theory of braided-Hopf algebras [21] including the diagrammatic notation used for this kind of algebra in [22,19].

Braided construction of exterior algebras on Hopf algebras
In this preliminary section we give a self-contained braided-Hopf algebra approach to bicovariant exterior algebras on Hopf algebras building on our recent work [35]. We recall first that a differential graded algebra means a graded algebra Ω = ⊕ n Ω n equipped with a super-derivation d increasing degree by 1 and squaring to 0. The standard setting is where Ω 1 is spanned by elements of the form adb where a, b ∈ A = Ω 0 , and Ω is generated by degrees 0,1; in this case we say that we have an exterior algebra over A. Any first order differential structure (Ω 1 , d) over A can be extended to a maximal prolongation.
2.1. Maximal prolongation on a Hopf algebra. When A is a Hopf algebra we can ask that left and right comultiplication extends to a bicomodule structure with coactions ∆ L , ∆ R on Ω 1 and d a bicomodule map. In this case it is known that Ω becomes a super-Hopf algebra with coproduct that of A on degree zero and ∆ L + ∆ R on degree 1, see [3,35]. In this case π : Ω → A which sends all degrees > 0 to zero and is the identity on degree 0, forms a Hopf algebra projection split by the inclusion of A. As a result, assuming that the antipode of A is invertible and by a super version of [38,23], we have Ω ∼ =A· <Λ where Λ is a super-braided Hopf algebra in the braided category of right A-crossed (or Drinfeld-Yetter) modules. We recall that a right A-crossed module means a vector space Λ 1 which is both a right module and a right comodule such that In this case there is an associated map Ψ : (1) . A similar map for any pair of crossed modules makes the category of these braided when the antipode S is invertible. Here A + = ker is itself a right crossed module by right multiplication and Ad R (a) = a (2) ⊗ (Sa (1) )a (3) and the result in [39] that first order calculi (Ω 1 , d) are classified by ad-stable rightideals can be viewed as saying that they are classified by surjective morphisms : This point of view was recently used in [35] to generalise beyond the surjective case, where we do not assume that is surjective. The exterior algebra is similarly given as bosonisation of a pair (Λ, d) consisting of a primitively generated (by degree 1) super-braided Hopf algebra Λ in the crossed module category equipped with a super-derivation (the restriction of d) which is a right A-comodule map and obeys d 2 = 0. This is required to have the further characteristic properties [35] (2.1) d (a) + ( π a (1) )( π a (2) ) = 0 for all a ∈ A + and η ∈ Λ, where we underlined the braided-coproduct. It is shown in [35] that given such (Λ, d), we obtain a bicovariant calculus (Ω, d) with da = a (1) π a (2) on degree 0 and that d is also a supercoderivation. Here (2.1) is called the Maurer-Cartan equation cf [39]. These results also clarify the surjective case: We use (2.3) on degree 1 in the form (1) in the start of the induction and so as to be able to similarly compute ∆(ωdη). We omit further details of the induction but we still need to establish both properties on degree 1. Thus for all a ∈ A + , b ∈ A. We use π (ab) = π (a)b + (a)π b for all a, b ∈ A. For (2.3) we check the degree 1 version as ∆d (a) = −∆(( π a (1) )( π a (2) )) for all a ∈ A. The latter part of the calculation here amounts to the identity for the crossed module braiding.
Hence in order to construct (Λ, d) in the surjective case it suffices to take (2.1) as a definition d (a) := −( π a (1) )( π a (2) ) and show that this is well-defined and extends as a super-derivation of square zero. This will then make (Λ, d) itself into a DGA over k.

Proposition 2.2. Let
: A + Λ 1 be a surjective morphism in the category of right crossed modules. Then together with d defined by the Maurer-Cartan equation gives a super-braided Hopf algebra in the category which is also a differential graded algebra obeying (2.2)-(2.3). Its bosonisation Ω max = A· <Λ max is the maximal prolongation differential calculus extending (Ω 1 , d).
Proof. We quotient by the minimal subspace in degree 2 for which d : Λ 1 → Λ 2 is well-defined by the Maurer-Cartan equation. Let be the usual cobar coboundary, where ∆ j denotes the coproduct in the i'th position, and define d on degree m by d.( π ) m = − · ( π ) ⊗(m+1) ∂. This is well-defined for the same reason as before because ∂ is the sum of terms each acting via the the coproduct. Clearly ∂ is a super-derivation and squares to 0, so d has the same features. We also need to check that we have a super-braided Hopf algebra. Since the algebra is quadratic, the main relation to check is ∆(( π a (1) )( π a (2) )) := ( π a (1) )( π a (2) ) ⊗ 1 + 1 ⊗ ( π a (1) )( π a (2) ) + π a (1) ⊗ π a (2) − Ψ( π a (1) ⊗ π a (2) ) vanlshes whenever a ∈ I = ker π . This is clear for the first two terms and for the remaining two we use (2.4) to obtain ( π ⊗ π )Ad R (a) (much as in the proof of Lemma 2.1) which indeed vanishes as I is Ad-stable because was a morphism. Hence by the lemma we have the required data (Λ max , d) and obtain a bicovariant calculus after bosonisation, something one can also check directly from I an Ad-stable right ideal and the structure of A· <Λ max . It is also clear from the construction, since we imposed the minimal relations compatible with the Maurer-Cartan equation, that our calculus is isomorphic to the maximal prolongation.

Braided linear spaces.
Here we take an aside to recall the theory of braidedlinear spaces introduced in [24,18] but in a cleaner form as braided operators rather than braided matrices. Braided linear spaces was our term for primitively generated graded braided Hopf algebras, with particular emphasis in [27] on what have later been called 'Nichols-Woronowicz algebras' [1]. If V is an object in an Abelian braided category then it inherits a morphism Ψ = Ψ V,V : V ⊗ V → V ⊗ V obeying the braid relations. Our setting is categorical but we use only the pair (V, Ψ) and tensor powers of V in the following definition.
Definition 2.3. [24,18] Let (V, Ψ) be an object in an Abelian monoidal category and a braiding on it. The braided binomials here are defined recursively by where 0 < r < n and Ψ i denotes Ψ acting in the i, i + 1 tensor factors. We also define 'braided integers' These are operator versions of binomial coefficients and generalise q-binomials when applied to the category of Z-graded vector spaces with braiding given by powers of q. Relevant to us, the braided factorials also generalise symmetrizers and antisymmetrizers. We need the following main theorem about them: The proof in [18] is written in matrix terms but immediately translates as operators in our setting. In fact, these results amount to identities in the group algebra of the braid group and are best done diagrammatically. The key observations are that since the first r − 1 terms in [r, Ψ] commute with Ψ r · · · Ψ n−1 and by functoriality (since the braided-binomial is a morphism) or directly by induction using Definition 2.3 and repeated use of the braid relations. Using these properties, [24,18] then proves by induction on n that from which the theorem follows by repeated application. Also, by writing the above definitions as diagrams and turning the diagrams up-side down, we have co-binomial maps and co-integers defined by [n, Ψ] = 1 + Ψ n−1 + Ψ n−2 Ψ n−1 + · · · + Ψ 1 · · · Ψ n−1 Moreover, where the factorials coincide by repeated use of the braid relations or because both cases can be written as σ∈Sn Ψ i1 · · · Ψ i l(σ) where σ = s i1 · · · s i l(σ) is a reduced expression in terms of simple transpositions s i = (i, i + 1).
Next we consider the tensor algebra T V in an Abelian braided tensor category as a direct sum of different degrees T n V := V ⊗n and product given by concatenation of ⊗. Here T 0 V = k or more precisely the unit object of the category. The unit η of the algebra T V is the identity map from k → T 0 V . Thus is the identity map with suitable rebracketing (with Φ as necessary in the general case). We also consider the identity maps η n : V ⊗n → T n V with η 0 = η. Although all these maps are the identity, we are viewing them in different ways. We will consider two different braided Hopf algebra structures T ± V on T V , as a braided-Hopf algebra or as a super-braided Hopf algebra in the category. The tensor algebra has a braided Hopf algebra/super Hopf algebra structure T ± V with coproduct for the two cases. The counit is TnV = 0 for all n > 0.
Proof. This is the content of [18, Ma:fre,Propn 10.4.9] in the free case where we impose no relations, but we rework the proof in the current more formal notations and we state the super case explicitly. We start with the linear coproduct and for T + V we extend this as a Hopf algebra in the braided category, while for T − V we extend as a super-Hopf algebra in the braided category. We do the first case; the other is exactly the same by replacing Ψ by −Ψ. The proof is by induction assuming the formula for ∆ Tn−1V , where in the first line we split off the r = 0 part of ∆ Tn−1V and · is the braided tensor product. We then compute out the latter and for the 4th equality we renumber r + 1 → r in the second sum and absorb the otherwise missing r = 1. For the 5th equality we use Definition 2.3 and finally combine terms to obtain the desired expression for ∆ Tn V . Again, this is really a result at the level of the braid group algebra and can be done with diagrams.
In this situation we are now ready to define the (super)Hopf algebra quotients as the braided-symmetric algebra and braided exterior algebra on V respectively. That the coproduct descends to B ± (V ) follows immediately from Proposition 2.5 and Theorem 2.4. That ⊕ m [m, ±Ψ]! is an ideal or equivalently that the product in T V descends to the quotient follows from the arrow-reversed version of Theorem 2.4 where the factorials are on the right. It is easy to see that when our construction is in a braided category and φ : V → W is a morphism then φ ⊗ (the relevant power) in each degree is a morphism B ± (V ) → B ± (W ) of (super)braided-Hopf algebras. This is because, by functoriality of the braidings, the braidings and braided factorials are intertwined by φ on each strand in the diagrammatic picture. The B + (V ) case is also called the Nichols-Woronowicz algebra of V due to the structure of the algebra, but the above description and the fact that it is a (super) braided-Hopf algebra is due to the author. The earliest examples were the braided-line and braided quantum-plane (see [24,18]) while other early examples were U q (n + ) in the work of Lusztig [14].
Our own motivation to consider (2.7) to all degrees of relations was in the case when V has a right dual V * . Recall that V * ⊗n is right-dual to V ⊗n by the nested use of ev V and we use the same nesting convention for a duality pairing , on tensor products. In this case [24,18] the tensor algebras T ± V * and T ± V are dually paired by and B ± (V * ), B ± (V ) are clearly the quotients by the kernel of the pairing. That the product on one side is the coproduct on the other follows immediately from Proposition 2.5 and Theorem 2.4. This means that B ± (V * ), B ± (V ) are nondegenerately paired (super) Hopf algebras in the braided category and the relations of B ± (V ) are the minimal relations compatible with this duality. The merit of this approach is that we also have the immediate result, which will need later: Corollary 2.6. [27] If V has a right dual and B ± (V ) has a finite top degree then it has a right dual via ev = , and coevaluation map coev : where the construction is independent of the choice of inverse image of [m, ±Ψ]!.
This makes more precise the notion of braided-exponentials in [24,18] without formally assuming that the braided factorials are invertible. It was used explicitly in [27] to construct quasitriangular structures. If we take the well-known case Λ 1 = kx in the braided category of Z/(n + 1)-graded vector spaces with braiding

2.3.
Minimal prolongation on a Hopf algebra. We now return to our setting of a Hopf algebra A with invertible antipode and a surjective morphism : A + → Λ 1 in the braided category of right A-crossed modules. The following braided-Hopf algebra version of Woronowicz's construction [39] is largely known eg [28] but we provide a new direct construction for d on B − (Λ 1 ) going through (2.1)-(2.3) from [35]. This is a rather different from the approach in [39], which was to formally adjoin an inner element θ and define d = [θ, }.
Proposition 2.7. Let : A + → Λ 1 be a surjective morphism is the category of right crossed modules and Λ min = B − (Λ 1 ). This is a quotient of Λ max and inherits d obeying (2.1)-(2.3). Its super-bosonisation Ω min = A· <Λ min recovers the Woronowicz bicovariant calculus [39] on A associated to the Ad-stable right ideal ker .
Proof. We start with the identity (2.4) and Ad-invariance of I implies now that ( π ⊗ π )∆(I) ⊂ ker[2, −Ψ] = id − Ψ, meaning that the relations ( π ⊗ π )∆(I) = 0 of the maximal prolongation in Proposition 2.2 already hold among the quadratic relations in Λ min = B − (Λ 1 ). The latter is therefore a quotient of the Λ max . Next we consider the coboundary ∂ m : A ⊗m → A ⊗(m+1) as the proof of Proposition 2.2 and regard A as a right crossed module by Ad R and a b = π (a)b. Then π becomes a morphism. Lemma 2.8 below shows that ∂ descends to B − (A) → B − (A) in each degree as it respects the kernels of the relevant braided-factorials. Next, π being a crossed module morphism induces by π ⊗m in degree m a map given by −dπ ⊗m = π ⊗m ∂ m because the kernel of π ⊗m is spanned by elements where at least one of the tensor factors is 1. When we apply ∂ m then every term has at least one tensor factor 1 which is then killed by the final π ⊗m . This is the first cell of Similarly, : A + → Λ 1 being a morphism of crossed modules induces B − (A + ) → B − (Λ 1 ) given by ⊗n in degree n, andd descends to a map d : defined by d ⊗m = ⊗md . This is because the kernel of ⊗m consists of terms where at least one of the tensor factors is in I. When we compute thed of such terms using ∂ m , either a ∆ j does not act on this tensor factor, in which case this tensor factor is present in the output of ∆ j and the whole term is killed by the action of ( π ) ⊗m , or ∆ j does act on this element. But then · · · ⊗ ( π ⊗ )π (∆I) ⊗ · · · is in the kernel of id − Ψ in the relevant place as seen above, hence vanishes in B − (Λ). We are using the fact that the kernel in each degree contains the degree 2 relations between adjacent tensor factors. In this way, d equips Λ min = B − (Λ 1 ) with a differential as a quotient of the construction for Λ max in Proposition 2.2. One can show that if Ω 1 is inner by θ ∈ Λ 1 then the same applies to Ω, but we are not assuming this. The algebra structure of the bosonosation Ω = A· <Λ min is more well-known to be isomorphic to the one in [39].
The following lemma was needed to complete the proof. Here Ψ is the braiding for the crossed-module structure on A whereby π becomes a morphism.
Lemma 2.8. Let A be a Hopf algebra and Ψ i = Ψ the induced braiding acting in the i, i + 1 position of a tensor power, Here Ad ⊗j denotes the tensor product right coaction on j copies (acting here in the first position).
Proof. Clearly since the operators act on different tensor factors, just the numbering changes in the 2nd case. We also find by direct computation in the Hopf algebra that where Ad = Ad R and ∆ ⊗ Ad is the tensor product right coaction. As a warm-up, using these relations, we show Starting with this, we next prove by induction that Assuming this for m − 1 in the role of m, for the 2nd equality, where we picked out and computed the Ψ 1 · · · Ψ m−1 ∆ m−1 term from the sum in ∂ m−1 . Looking now at the last expression, we compute where the Ad terms cancel between the sum and the displaced sum except for the top term of one sum and the bottom term of the other. In the ∆ sum all the indices of Ψ are two or more smaller than the index of ∆ so commute to the right. Combining with our previous calculation, we have which proves (2.9).
Next we use this result as initial base for induction on i in a formula where [m] ≡ [m, Ψ] for brevity and the nesting is rightmost as for braided factorials The case i = 1 is (2.9) which we have already proven while the case i = m − 1 or i = m, suitably interpreted in the sense of absent sums or products when out of range, proves the lemma. We use identities where the commutation relation is due to acting in different spaces, with renumbering due to the notation. The first equation is a direct computation. One also has we do not need right now, in the first case due to different tensor products and in the second case because Ψ is a morphism in the crossed module category and hence commutes with Ad applied to tensor powers that include those on which Ψ acts. Assuming (2.10) for i − 1 in the role of i, what we need to show to prove (2.10) for i is Now, the first sum commutes with [m−i+1] since on the left this is 1−Ψ i +Ψ i Ψ i+1 + · · · + (−1) m−i Ψ i · · · Ψ m−1 due to the right-most embedding. These commute past the Ad ⊗j 1 getting changed to [m−i+1] embedded on the right (where the numbering is reduced by one). Hence the first term on the left is as the Ψ i+1 and higher commute, reducing index by 1, while Ψ i computes as shown. Hence the middle terms gives , the first term of which completes our previous sum to give the first desired term. Accordingly we need only show for the remaining term that But this is just the same identity (2.9) already proven but for [r]∂ r−1 , i.e. r = m − i + 1 in the role of m, for the m tensor factors numbered i, · · · , m. This completes our proof of (2.10) for all i and proves the lemma.
This fleshes out the braided-Hopf algebra interpretation of the Woronowicz exterior algebra on a Hopf algebra [39] using [35] for the direct treatment of d.

Differential calculi on braided linear spaces.
For completeness, we give another braided construction namely the exterior algebra Ω(B) on a Hopf algebras B in a braided Abelian category. This includes braided symmetric algebras B = B + (Λ 1 ) as above generated canonically by an object Λ 1 . The further data we will need is a surjective morphism : B → Λ 1 in the category such that (2.11) • · = ⊗ + ⊗ This data arises naturally as follows: suppose B is a (possibly degenerately) dually paired braided-Hopf algebra from the right (so the pairing is ev : B ⊗ B → 1) and L a rigid primitive sub-object L ⊂ B (so that the coproduct restricted to L is the additive one). We view the duality pairing restricted to a map B ⊗ L → 1 as a map which then obeys (2.11). This is surjective if there does not exist η ∈ L which pairs to zero with all of B. In the case of B = B + (Λ 1 ) or any other graded braided Hopf algebra of the form B = 1 ⊕ Λ 1 ⊕ B >1 generated in degree 1 by an object Λ 1 , we simply take : B → Λ 1 as the projection to degree 1.
Proposition 2.9. Let B be a Hopf algebra in an Abelian braided category and : B → Λ 1 a surjective morphism obeying (2.11). Then is a differential exterior algebra on B in the category (one in which all structure maps are morphisms).
Proof. The proof is done diagrammatically in Figure 2.4 and applies generally but for convenience of exposition we also refer to concrete elements. The braided tensor product B⊗Λ 1 in concrete terms means (b ⊗ v)(c ⊗ w) = bΨ(v ⊗ c)w and featured already in the definitiion of a braided-Hopf algebra. Part (a) computes d(bc) using the braided coproduct homomorphism property and (2.11). Using the counit axioms Figure 1. Diagrams in the proof of Proposition 2.9 for quantum differentials on braided planes and a morphism we obtain bdc for the first term and (db)c for the second when we remember the braided tensor product. Part (b) checks that d extends as a graded derivation with respect the braided tensor product. We compute This gives a differential structure on our braided-symmetric algebras B + (V ) regarded as noncommutative spaces. If the category is the comodules of a coquastriangular Hopf algebra, for examples, our construction is covariant in that all structure maps are comodule maps. Also note that if {e i } is a basis of V , we have explicitly where ∂ i are the (right handed) braided partial derivatives defined by They are given explicitly at the level of the tensor algebra by where the last tensor factor of the result is viewed in Λ 1 .
where ! denotes the Koszul dual. If a quadratic algebra on a vector space W has relations R ⊂ W ⊗ W as the subspace being set to to zero then its Koszul dual is the quadratic algebra on W with relations R ⊥ ⊂ W ⊗ W . This is normally done in the category of vector spaces but we do it here in a braided category using the right dual so that W = V . Example 2.11. We let B = A 2 q = B + (V ) be the quantum plane associated to the standard corepresentation V = span{x, y} in the braided category of right k q [GL 2 ]-comodules with q 2 = 1. Here is given by a particular non-standard normalisation of the usual R on k q [GL 2 ] (one that does not descend to k q [SL 2 ]). The kernel of id + Ψ gives us the relations yx = qxy of the quantum plane since (id where this time the same basis is denoted {dx, dy} as a basis of Λ 1 = V and one can check for example that (id + Ψ)(dx ⊗ dy) = dx ⊗ dx + qdy ⊗ dx from the stated braiding. Indeed, it known that A 2 q and A 0|2 q as Koszul dual as first pointed out by Manin [36]. The differential on . This again comes from the same braiding as above but viewed now as defining the relations One can check for example that d(yx − qxy) = 0 as it should. By construction, this exterior algebra on the quantum plane is k q [GL 2 ]-covariant. The associated partial derivatives are using the braided coproduct on general monomials computed in [18] from the braided-integers [n, Ψ]. The partial derivatives here were first found by Wess and Zumino in another approach. They are naturally 'braided right derivations' with an extra q n in the first expression, in order that d is a left derivation, acting as braided q 2 -derivatives in each variable. One can check that ∂ 2 ∂ 1 = q∂ 1 ∂ 2 as operators, also as per the general theory in [18].
This reworks the treatment of quantum-braided planes and their differentials in [18,Sec. 10.4] [24] now as an example of our above canonical construction based on B + (Λ 1 ), as opposed to a compatible pair of R-matrices R, R as previously.

Braided fourier transform and application to Hodge theory
Fourier transform on Hopf algebras is part of their classical literature. It was extended to braided-Hopf algebras in [15] and related works and applied to braided linear spaces in [11], though not the ones we consider here. We used diagram proofs and will do so again, while another work from that era is [16]. We first explain the general (super) formulation and then apply it to the Hodge operator, including k(S 3 ) as an example.
3.1. Super-braided Fourier theory. In any braided category C and B ∈ C a braided Hopf algebra dually paired with a braided Hopf algebra B , we have three actions which we will consider and which we collect in Figure 1 in a diagrammatic notation [22,19]. Diagrams are read as operations flowing down the page, with tensor products and the unit object 1 suppressed. Two strands flowing own and merging denotes the product and one strand flowing down and splitting denotes the coproduct. As in Section 2 when discussing duals, we assume a pairing ev : B ⊗ B → 1 which we can write diagrammatically as ∪ and with respect to which In a concrete k-linear setting we can suppose that 1 = k and η(1) = 1 to simplify the above. Reg makes B a right B module algebra in the braided category. The principal ingredient of Reg here is actually a left action making B a left Bmodule algebra in the braided category. Similarly, we have a straightforward right action under which B is a right B-module algebra [19].
We also need the notion of a left integral and the simplest thing is to require a morphism : B → 1 in the sense (id ⊗ )∆ = η ⊗ . However, we do not want to be too strict about this. For example, for the finite anyonic braided line B = k[x]/(x n+1 ) in the braided category of Z/(n + 1)-graded spaces with braiding given by an n + 1-th root of 1 and |x| = 1, the obvious x m = δ m,n is not a morphism to 1. Our approach is to live with this and not necessarily assume any morphism properties; we can still use the diagrammatic notation but be careful not to pull the map through any braid crossings. A more formal approach is to view it as a morphism B → K where K = k taken with degree n in the case of the anyonic braided line. The uniqueness of the integral when it exists is similar to the Hopf algebra case (see [2] for a formal proof).
For Fourier transform we need not only that B is dually paired but that B is actually rigid with dual object B * . Again, this is a very strong assumption, analogous to finite-dimensionality of B and amounting to this in the typical k-linear case. It means that there is a coevaluation map exp = coev : 1 → B ⊗ B * , denoted by ∩ in the diagrammatic notion, which obeys the well-known 'bend-straightenning axioms' with respect to ∪. We similarly require a right integral which is not necessarily a morphism : B → 1. We can live with this or suppose formally that : B → K * where K * ⊗ K = 1 = K ⊗ K * as objects. In our k-linear setting this will be by the identification with k. The theory below could be generalised to include some infinite-dimensional cases or else these could be treated formally eg in a graded case with B * a graded dual, each component rigid and the result a formal power series in a grading parameter. These maps are no longer morphisms if the integrals are not, or one can say more formally that F : B → K ⊗ B and F : B → K * ⊗ B . The following extends and completes [11].
If the integrals are both unimodular and morphisms then FF = µS and [F, S] = 0 when µ is invertible (see Figure 2 for the general case).
Proof. Here FReg and F F are already covered in [11, so we do not repeat all the details here. We recall only the diagram proof for F * F using the lemma in [11, Fig. 2(b)] at the first equality in Figure 2 and note that we did not need to assume that , * are morphisms to 1 as in [11] as long as we keep the integrals to the left. The second line now uses the same lemma but this time on B to compute FF * as shown provided * is also a left integral so that the lemma applies and is also a right integral. If , * are morphisms to 1 so we can take them through braid crossings to obtain µS and then µFS = FF F = µSF. The general result (F ⊗ id) = F • · follows more simply from the duality pairing and associativity of the product of B.
The map F here is a right-integral version of the theory which is being used to define the adjoint Fourier transform and converted to a left version via Ψ. The braided antipode S plays the role of the minus sign familiar in classical Fourier theory and µ plays the role of 2π. If µ and S are invertible then the stated results imply that F is invertible at least in the k-linear setting (with F −1 = S −1 F in the unimodular trivial morphism case). Also, if we compose F with S then the first property above becomes  and ditto with x replaced by y. One can see that F F = µS using q n(n+1) 2 = (−1) n . Due to the nontrivial braidings of the integrals, however, the right hand side in Figure 2 gives where D is the monomial degree operator. The same method as in the proof above now gives us FS = q 2D+1 SF or equivalently SF = FSq 2D+1 , which one may verify from the stated F, S.
There is up to scale a unique top degree in each case, namely Vol = e 1 e 2 and Vol * = f 1 f 2 and we find Vol * , Vol = ev(f 1 ⊗ f 2 , [2, −Ψ](e 1 ⊗ e 2 ) = −q −1 , so that We define integrals via Vol = 1 and * Vol * = 1 but note that these are not morphisms. Rather we use braidings Ψ(f 1 ⊗e 1 ) = e 1 ⊗f 1 +(1−q 2 )e 2 ⊗f 2 , Ψ(f 2 ⊗e 2 ) = e 2 ⊗f 2 , Ψ(f i ⊗e j ) = qe j ⊗f i for i = j (these are obtained from the 2nd inverseR as in [18,Propn. 10.3.6] for R normalised to our case) to find Ψ(f i ⊗ Vol) = qVol ⊗ f i and hence Ψ(Vol * ⊗ Vol) = q 2 Vol⊗Vol * . We similarly have Ψ(f i ⊗Vol * ) = qVol * ⊗f i . From these it is clear that and * are not morphisms in the underlying comodule category. Again, there can be further signs according to the super degrees for the actual super-braiding Ψ sup when we read diagrams in the super-braided case. In particular, we find needed in the computation of F . We now read off from the diagrammatic definitions in Figure 1, One can verify that F F = µS as it must by Proposition 3.2. We also have where D is the monomial degree and one can check that this agrees with the lower line in Figure 2 where we use the above computations to read off the right hand side. In this case FS = q 2(D−1) SF or SF = FSq 2(D−1) as one can verify from the stated form of F, S.
A similar approach can be used for other quantum planes to express their differential exterior algebras as super-braided Hopf algebras with possibly a different underlying coquastriangular structure from one used for the coordinate algebra as a braided-Hopf algebra. In the presence of an invariant quantum metric we reproduce the otherwise ad-hoc approach to q-epsilon tensors and Hodge theory on braided-quantum planes in [18]. The f i generate antisymmetric vectors and , define an interior product connected to the exterior algebra product via F. This example should be seen as a warm-up to Section 3.2 where we look at bicovariant differentials on Hopf algebras themselves. As illustrated here, the actual theory is read off the diagrams with the appropriate braiding including signs. We could indeed shift all constructions to this new super-braided category and say that the above example is an ordinary braided-Hopf algebra there, but we not do so since there will normally be other (bosonic) objects also of interest in the original category. In our context the nicest case is where , * are morphisms to 1 when viewed in the original category but do not necessarily respect the super-degree, for example they could be odd maps in the super-sense in which case they are not morphisms in the super version with extra signs (so we need the slightly more general picture as above). We assume they have the same parity of support (both odd or both even maps). Then F, F also have this party. Proof. For FF we have to compute the right hand side of the lower diagram in Figure 2, which now has extra signs. We can still bring out µ = ( * ⊗ )Ψ −1 sup exp since any signs from crossing the leg cancel with signs from crossing the leg by our assumptions. Next, we can lift through the crossing at the price of (−1) p in computing µ . We already have F F = µS from Proposition 3.2 and can then conclude the rest.
The behaviour of F with respect to Reg, has an unchanged form as these statements do not involve additional transpositions, except that the actions themselves are computed for the super-braided Hopf algebra eg with the super-braided coproduct and hence the super-braided Leibniz rule expressed in super-braided module algebra structures. The property (3.1) becomes due to the crossing of the first input on the right hand side with the integral in F.

3.2.
Hodge theory on Hopf algebras. We are now going to compute our super-Fourier theory for B = Λ min = B − (Λ 1 ) where Λ 1 is a rigid object in the braided category of right A-crossed modules and A is a Hopf algebra with invertible antipode. Here B is a super-braided Hopf algebra in the category and we assume it has a top degree component K of dimension 1, i.e. up to scale a unique top form Vol ∈ B − (Λ 1 ). This gives us a unimodular integral B → k by Vol = 1 and zero for lower degrees. To see this, note that the formula in Proposition 2.5 ensures that ∆Vol = Vol ⊗ 1 + 1 ⊗ Vol plus terms of intermediate degree, and we never reach the top degree on applying ∆ to lower degree. One can think of this more formally as a morphism B → K with some possibly non-trivial generator. We also have an identification B = B − (Λ 1 * ) by extending the duality pairing Λ 1 * ⊗Λ 1 as a braided-Hopf algebra pairing, given that this is now non-degenerate after quotienting by the relations of B − as explained in Section 2.2. Hence we obtain a unimodular integral on this too. In the nicest case, the top forms Vol, Vol * of degree n (say) span the trivial object 1 so that , * are morphisms to 1 but of parity n mod 2, so we are in the setting of Corollary 3.5.
Next, in non-commutative geometry a metric is g ∈ Ω 1 ⊗ A Ω 1 with an inverse ( , ) : Ω 1 ⊗ A Ω 1 → A. One can show that in this case g must be central. Normally, one also requires the metric to be 'quantum symmetric' in the sense of the product ∧(g) = 0 in Ω 2 . We are interested in left-invariant metrics where g ∈ Λ 1 ⊗ Λ 1 .
Lemma 3.6. A bi-invariant metric on a Hopf algebra A with bicovariant calculus is equivalent to an A-crossed module isomorphism g : Λ 1 * ∼ =Λ 1 . The metric is quantum symmetric if and only if Ψ(g) = g.
Proof. The metric being bi-invariant means that it is an element g ∈ Λ 1 ⊗ Λ 1 which is invariant under the coaction ∆ R on the tensor product. The existence of a bimodule map ( , ) requires g to be central which in turn requires that g is invariant under the crossed module right action (since this determines the cross product of A· <Λ). So a metric is equivalent to a morphism 1 → Λ 1 ⊗ Λ 1 in the crossed module category. Evaluation from the left makes this equivalent a morphism as stated, which we also denote g. Here Λ * is again a right crossed module in the usual way (via the antipode). Clearly ∧(g) = 0 if and only if g ∈ ker[2, −Ψ] = ker(id − Ψ) according to the relations of B − (Λ 1 ).
Given a bi-invariant metric we therefore have B ± (Λ 1 * ) ∼ =B±(Λ 1 ) hence combined with the above remarks in the finite-dimensional case, an isomorphism which we also denote g : B ± (Λ 1 ) * → B ± (Λ 1 ). We are now ready to define the Hodge operator, using the B − version. We do it in the nicest case but the same ideas can be used more generally as we have seen in Section 3.1. By construction our is a morphism in the crossed-module category. In geometric terms this means that it extends as a bimodule map and is bicovariant under the quantum group action on Ω. We also define = (−1) D • F • g −1 where D is the degree operator.
Proof. Here Vol, Vol is non-zero since otherwise Vol would be zero in B − (Λ 1 ), and its inverse supplies the coefficient of the top component of exp, which is µ. Since µ = 0 we can apply Corollary 3.5 to see in particular that , S graded-commute. That S| 0,1,n−1,n = (−1) D i.e. on the outer degrees is clear on degrees 0,1 and then holds on degrees n, n − 1 due to , S graded-commuting. Next, in terms of with the metric identification, the result in Corollary 3.5 becomes = µ(−1) D S and (−1) D = µ(−1) n S since the parity of the integral is n mod 2. Taking the (−1) D to the left in the latter equation makes it (−1) n−D so that = µ(−1) D S −1 on all degrees, giving | 0,1,n−1,n = µ −1 on the outer degrees. On the other hand, we have exp = 1 ⊗ 1 + g + · · · + g (n−1) + µVol ⊗ Vol (for some element g (n−1) ∈ Λ n−1 ⊗ Λ n−1 ), while the definition of is such that it is given by integration agains Ψ −1 exp without any signs. Since g (by the quantum symmetry assumption) and 1 ⊗ 1 are invariant under Ψ, these terms are the same, and hence = on degrees n − 1, n and hence 2 = µ on these degrees. In that case ( ω) = µω = 2 ω on all ω of degree n − 1, n tells us that = on degrees 0,1 also, and hence that 2 = µ on these degrees also. This means that Ψ −1 sup exp = 1 ⊗ 1 − g + · · · + (−1) n−1 g (n−1) + (−1) n µVol ⊗ Vol for the computation of F and similarly without the signs for .
We similarly define left and right interior products by restricting the left and right actions in Section 3.1 (these are the left and right braided-partial derivatives in the sense of [24,18]). We then extend these to bicovariant bimodule maps : given by (aη) (bω) = (aη, bω (1) )ω (2) , (bω) (aη) = bω (1) (ω (2) , aη), ∀a, b ∈ A, η ∈ Λ 1 , ω ∈ Λ where we underline the braided-coproduct of Λ. In other words, we extend the braided coproduct as a bimodule map Ω → Ω ⊗ A Ω (not to be confused with the super-coproduct of Ω as a super-Hopf algebra) and then use the quantum metric pairing to evaluate, taken as zero when degrees do not match.
We can now interpret our Fourier theory in Section 3.1 as The use of (S ) −1 here is adapted to the left handed and left-handed partial derivatives defined by df = a (∂ a f )e a for any choice of basis {e a } of Λ 1 . One could equally well use but this would be adapted to and right-handed partial derivatives. We also define the Leibnizator as in [32].
Corollary 3.9. The codifferential and Hodge Laplacian in (3.4) obey for all f ∈ A and ω ∈ Ω. Moreover, where α = α a e a in a basis and g ab = (e a , e b ) (summation understood). If δα = 0 Proof. The formula for δ(f ω) follows immediately from the derivation property of d and the first interior product property in (3.3). The formula for (f ω) then follows from this and the Leibniz rule for d as in [32]. These results then give the explicit formulae for α = α a e a .
Note concerning α that (e a (e b e c )) + g bc e a − g ab e c = (e a e b ) e c in the classical case, which is antisymmetric in a, b, while L δ (df, ω)+L df ω = 2∇ df ω in the classical case as shown in [32]. Here ∇ is the classical Levi-Civita connection referred back to a derivative along 1-forms via the metric. The special case shown in Corollary 3.9 is relevant to 'Maxwell theory' where F = dα and Maxwell's equation δF = J has a degree of freedom to change α by an exact form, which freedom can be reduced by fixing δα = 0. Maxwell's equation then becomes α = J where J is required to be a coexact 'source'. Example 3.10. The standard 3D calculus on the permuation group S 3 on 3 elements is given by the conjugacy class of 2-cycles. We recall that Λ 1 = k − span{e u , e v , e w } is a k(S 3 )-crossed module as above, where a = u, v, w are the 2-cycles u = (12), v = (23), w = uvu = vuv = (13). The minimal exterior algebra in this case is known to be a super version of the Fomin-Kirillov algebra [6,29] with relations and exterior derivative e u e v +e v e w +e w e u = 0, e v e u +e w e v +e u e w = 0, e 2 u = 0, de u +e v e w +e w e v = 0 and the two cylic rotations of these where u → v → w → u. The dimensions in the different degrees are dim(Λ) = 1 : 3 : 4 : 3 : 1 so there is a unique top form up to scale, which we take as Vol = e u e v e u e w . This is clearly central and one can check that it is also bi-invariant. This can be done noting that Vol = e u e v e u θ and computing ∆ R (e u e v e u ) = e u e v e u ⊗ (δ e + δ w ) + e w e u e w ⊗ (δ u + δ vu ) + e v e w e v ⊗ (δ uv + δ v ).
Hence we have a canonical Hodge star.
The coevaluation element exp is a computation from Proposition 2.6 which in a basis reads exp = n m=0 ·[m, −Ψ]! −1 (e a1 ⊗ · · · ⊗ e am ) ⊗ f am · · · f a1 where {e a } is our basis of Λ 1 and {f a } is a dual one and we sum repeated indices. This is in general but in our case and using the metric identification comes out as We see that on the different degrees, 2 | 0,1,3,4 = −id, 3 | 2 = id so that has order 6. The first of these illustrates Proposition 3.8 while we note that on degree 2 coincides with minus the braided-antipode S of B − (Λ 2 ) (because this is braided-multiplicative along with an extra sign for the super case, and S| 1 = −id). The cohomology for this calculus in characteristic zero is known to be H 0 = k, H 1 = k, H 2 = 0, H 3 = k, H 4 = k and one can see that is an isomorphism H m∼ =H 4−m as expected, the cohomologies being spanned by 1, θ = e u + e v + e w in degrees 0,1 and their in degrees 3,4.
which is (-2 times) the standard graph Laplacian for the corresponding Cayley graph on S 3 . It is fully diagonalised as usual by the matrix elements of irreducible representations (the eigenvalues are 0, 6, 12 with eigenspaces of dimensions 1,4,1 respectively). We also have δα = g ab ∂ a α b and if this vanishes then We note that the last term here only has contributions from a = b. The above expression is a short computation from Corollary 3.9 using δ(e a e b ) = e b − e aba , e a (e b e c ) = δ a,b e c − δ aba,c e b from which we see that δde a = 3e a − θ, δ(e a e b ) + e a de b = e b − 2e aba + δ a,b θ.
Solving Maxwell theory in the form α = J, the source J has to be coexact. From the remarks above and H 3 (S 3 ) = kθ , this is equivalent to It is a useful check of our formula for on Ω 1 to see directly that when restricted to coexact forms its image indeed is again coexact. Moreover, by computer one finds the same eigenspaces (each 4-dimensional) as in [33] with eigenvalues 3,6,9, so that up to an overall constant the Laplacian restricted to coexact 1-forms is the same in spite of the Hodge operators being rather different. Explicitly, and their cyclic rotations under u → v → w → u have eigenvalue 3 (and along with their cyclic rotations add up to zero). Multiplying these by the sign function on S 3 gives eigenvectors of eigenvalue 9 while for the eigenvectors of eigenvalue 6 we can use the 'point sources' in [33], where three points that share a common node in the graph have a zero sum of their sources. These are related to the matrix elements ρ ij of the 2-dimensional representation. For example, if we work over R and ρ(u) = diag(1, −1), ρ(v) = This solves the 'Maxwell theory' on S 3 for this calculus by diagonalising on coexact 1-forms.
In [33] and all other such models until now it has been assumed that the Hodge operator should be designed to square to ±1, whereas our canonical Hodge operator in this example is order 6 and looks very different, but nevertheless gives the same reasonable Laplacians in degrees 0,1. Our construction also works for S 4 and S 5 with their 2-cycles calculus and can be analysed similarly, while higher S n , n > 5 are conjectured [29] to have infinite-dimensional B − (Λ 1 ).

Calculus and Hodge operator on coquasitriangular Hopf algebras
Here we start with a new, braided-Lie algebra, approach to the construction of bicovariant (Ω 1 , d) on quantum groups such as k q [G] for G a complex semisimple Lie group. We recall that these are all coquasitriangular in that they come with a convolution-invertible map R : This is just dual to Drinfeld's theory in [5], see [21,18]. We will need the 'quantum Killing form' The construction of differential calculi on a coquasitriangular Hopf algebra A has its roots in R-matrix constructions from the 1990s but the following general construction builds on our recent treatment in [35]. It is shown there that A is a left A-crossed module by Then any subcoalgebra L becomes a left A-crossed module by restriction and its dualisation L * in the finite-dimensional subcoalgebra case becomes a right A-crossed module. It is shown that the quantum Killing form regarded by evaluation on its first input as a map Q : A + → L * is a morphism of crossed modules. This gives:  We also recall the associated braidingΨ and braided-Killing form, and the adjoint action of a braided-Hopf algebra on itself.
Proposition 4.1. cf [35] Let A be a coquasitriangular Hopf algebra and L ⊆ A a nonzero finite-dimensional subcoalgebra. Then Λ 1 = image(Q) and = Q defines a bicovariant differential calculus Ω 1 on A.
In [35] we used a version of this to naturally construct possibly non-surjective differential calculi with Λ 1 = L * , but we also see from this result that L itself is the more fundamental object as starting point.

Braided-Lie algebras.
Our new approach is to start with a Hopf algebra B in a braided category C and find a 'Lie algebra' for it. We then take its dual to define a calculus. A principal result in the case of an Abelian braided category is the construction of the braided-enveloping algebra U (L) as a bialgebra. This is defined by the relations of commutativity with respect to the associated braidingΨ. In the category of sets a braided-Lie algebra reduces to a quandle and this was used recently to prove the cohomology theorem for finite group bicovariant calculi [34]. Nondegeneracy of the Killing form also turns out to be an interesting characteristic related at one extreme to the Roth property of a finite group [13]. The axioms themselves, however, were inspired by the properties of the braided adjoint action of a braided-Hopf algebra on itself as also recalled in Figure 3. The first part was done in [26]. Braided cocommutativity with respect to a Bmodule is just the axiom (L2) when specialised to [ , ] = Ad and the proof that the adjoint action then obeys (L3) appeared in [25,Prop A.2] in dual form (turn the diagrams there up-side-down). Clearly: We next recall that if A is coquasitriangular then there is a braided Hopf algebra version B(A) of A called its transmutation. This is also denoted A and has the same coalgebra as A but a modified product [21] and lives in M A by Ad R . Its product is braided-commutative, which can be written equivalently as while its braided-antipode is (2) ).
Because we have a mix of both types of structure on the same vector space, we will be more careful to underline the braided versions where they are different. Proof. We start by computing the left braided-adjoint action by applying S to a (2) and using the braiding to commute this past b before multiplying up with respect to •: where we use the definitions, the coaction properties and the multiplicativity property of R. We next unpack the adjoint coactions on a, and use multiplicativity of (11) )R(Sa (5) ⊗ a (12) ) where the 2nd equality is by quasicommutativity of A, the 3rd uses multiplicativity of R to recognise Ad R on b (0) . We then expand out by multiplicativity to recognise ν −1 (a) = R(S 2 a (1) ⊗ a (2) ). This is known [21,18] to be convolution inverse to ν(a) = R(a (1) ⊗ Sa (2) ) and to obey ν −1 (a (1) )a (2) = S 2 a (1) ν −1 (a (2) ), which we use to move to the right. The seventh equality uses multiplicativity of R so that we can use quasicommutativity on S 2 a (11) Sa (6) and the braid or Yang-Baxter equations on the last three factors to give the 8th equality. On this we use multiplicativity to cancel a (4) Sa (5) and obtain the 9th equality and two mutually inverse copies of R for the 10th. We finally cancel a (2) Sa (3) and move ν −1 to the left to cancel ν. We then recognise the answer in terms of Q and take this for our braided-Lie bracket.
The Lemma tells us that we have (L1) for free. Next, we verify (L2) for [ , ] = Ad noting that (L2) can be written in the form where in our case Ψ is the braiding of M A and using our result for Ad. We compute as required, where we used the coaction properties of Ad R and the multiplicativity property of R to make a cancellation. The above Lemma then tells us that we get (L3) for free. These results then apply for an subcoalgebra L ⊆ A since, due to the form of Ad, we see that Ad(L ⊗ L) ⊆ L, since Ad R (L) ⊆ L ⊗ A because L is a subcoalgebra (in other words a sub-coalgebra of A is also a subobject and hence a braided sub-coalgebra L ⊆ B(A). For the last part, we can equivalently writẽ  The braided-Killing form is where {e i } is a basis of L and {f i } a dual basis and u(a) = R(a (2) ⊗ Sa (1) ).
Proof. The braided-Killing form is obtained by reading down the diagram, as (summation understood) as stated. For the remark about the dualisation we note that A has a right crossedmodule structure given by Ad R and (2) ), ∀a, b ∈ A and its crossed module braiding turns out, by similarly using the properties of R as above, to coincide with the fundamentalΨ for the braided-Lie algebra (as computed in the proof of Theorem 4.5). On the other hand this crossed module is the right handed version of (4.1) which dualized to give the crossed module structure defining the calculus in Proposition 4.1. This means that U (L) is the Koszul or quadratic algebra dual of Λ quad (where we impose only the degree 2 relations of B − (Λ 1 )). The braided-Lie bracket and exterior derivative can also be related as part of a general theory of 'quantum Lie algebras' in [39] when 1 / ∈ L. Here every bicovariant calculus gives a quantum Lie algebra in the sense of [39] and meanwhile (one can show that) every non-unital braided-Lie algebra L gives a quantum Lie algebra by extending by 1 and then taking the kernel of the counit.
The above theorem is a new result and is needed to complete the picture. In the special case where L has a matrix coalgebra form on a basis {t i j } (such data defines a matrix corepresentation of A) we recover the R-matrix braided-Lie algebra construction introduced in [26] but now as a corollary of the above. Corollary 4.7. cf [26] Let A be a coquasitriangular Hopf algebra and t ∈ M n (A) a matrix corepresentation. Then the matrix subcoalgebra L = {t i j } has braided-Lie bracket, categorical braiding and braided Killing form The braided enveloping algebra U (L) is generated by the {t i j } with new relations . We sum over repeated indices in these expressions.
Proof. We expand out Q using the properties of R, then the above bracket can also be written explicitly as (1) ) and the categorical braiding in M A is (2) )R(Sa (2) ⊗b (4) )R(a (4) ⊗b (5) )R(a (5) ⊗Sb (1) ) From these we immediately read off the expressions stated, where R i j k l = R(t i j ⊗ t k l ) andR i j k l = R(t i j ⊗ St k l ) is the 'second inverse'. We likewise read off the relations of U (L) fromΨ or from (4.3) to give the result stated. In all cases we can moveR and another R to the left hand side, for example the relations can be written compactly as R 21 t 1 • Rt 2 = t 2 • R 21 t 1 R where the suffices refer to the position in M n ⊗ M n with values in U (L), also clear from (4.4). These are the relations of B(R) [20,18], the braided analog of the more familiar FRT bialgebra A(R).
This derives the explicit R-matrix formulae needed to compute examples. This in turn recovers the 4D braided-Lie algebra of k q [SL 2 ] found in [26]: Example 4.8. [26] For A = k q [SL 2 ] with q 2 = ±1, its standard matrix coalgebra and rescaled generators where λ = 1 − q −2 (we use different symbols for the entries of t i j to avoid confusion with the quantum group), the nonzero braided-Lie brackets are . Here (2) q = q + q −1 and we used Corollary 4.7 and the standard R-matrix for SL 2 with nonzero entries R 1 2 2 1 = q − q −1 , R 1 1 2 2 = R 2 2 1 1 = 1, R 1 1 1 1 = R 2 2 2 2 = q. The braided Killing form is [4, q −2 ]/q 10 times the nonzero values The enveloping algebra U (L) = B q [M 2 ] is generated by α, β, γ, δ with relations βα = q 2 αβ, γα = q −2 αγ, δα = αδ at this point denotes commutator not Lie bracket. This is the algebra of q-deformed 2 × 2 braided hermitian matrices which means that geometrically it should be thought of as q-Minkowski space [20,18]. There are two natural central elements, the braided determinant det q = αδ − q 2 γβ which should be thought of as the q-Lorentzian distance from the origin and q-trace tr q = q −1 α + qδ = t which should be thought of as the 'time' direction. In these variables (as opposed to the rescaled 'Lie algebra' variables) the classical limit is commutative allowing us to think of this as a noncommutative geometry. Over C our braided-Lie algebra has a natural real form or * -involution α * = α, β * = γ, δ * = δ for real q, which fits with the mentioned geometric picture.

4.2.
Calculus and Hodge operator on k q [SL 2 ]. In the case of a coquasitriangular Hopf algebra A with a generating matrix subscoalgebra {t i j }, Proposition 4.1 or dualization of Corollary 4.7 recovers a version of a known R-matrix construction [10] of quantum group covariant calculi. We let {E α β } be the standard basis of M n (k) and dual to the {t i j } basis of L. This then becomes a right A-crossed module with defines the possibly non-surjective bicovariant calculus, which is, however typically surjective for the standard quantum groups and the above right covariance. The calculus has an inner form with The associated right crossed-module braiding on Λ 1 will be denotedΨ also (it is adjoint to the one for the braided-Lie algebra) and is computed from the right crossed module structure as St j3 j4 ⊗ t k3 q ) and expands out as from which we see thatΨ(E α β ⊗ θ) = θ ⊗ E α β so that, in particular, θ 2 = 0 in Λ min = B − (Λ 1 ).
We now focus on A = k q [SL 2 ] where the smallest nontrivial irreducible is 2dimensional, giving us Λ 1 = M 2 (C). We write basis E 1 1 = e a , E 1 2 = e b , E 2 1 = e c , E 2 2 = e d and use the standard SL 2 R-matrix as in Example 4.8 to give the bimodule relations of the well-known 4D calculus first found in [39], where [x, y] q ≡ xy −qyx and λ = 1−q −2 = 0. The exterior differential is necessarily inner with θ = e a + e d which implies that Note that we should scale d or θ by λ −1 in order to have the right classical limit but we have not done this in order to follow the general construction. The right coaction on left-invariant 1-forms is where e z := q −2 e a − e d and the calculation is from ∆ R e α β with the relevant Rmatrix. We use the symmetric q-integers so that (2) q = q + q −1 . The crossed module braiding comes out as As in degree 0, we note that λ −1 d has the right classical limit not d itself. The dimensions here in each degree are dim(Λ) = 1 : 4 : 6 : 4 : 1. The following is mostly known e.g. [30] but we give a short proof as it is critical for us. Proposition 4.9. For the above 4D calculus on k q [SL 2 ] with q 2 = ±1 there is a unique bi-invariant central metric g = e c ⊗ e b + q 2 e b ⊗ e c + q 3 (2) q (e z ⊗ e z − θ ⊗ θ).
The inverse metric is Proof. Using the quantum group relations one has   a 2 (2) q ab b 2 ca 1 + (2) q bc db c 2 (2) q cd d 2 a 2 (2) q ab b 2 ca 1 + (2) q bc db c 2 (2) q cd d 2 which gives us the unique generically q-invariant element of the tensor square of the space spanned by {−e b , e z , q −1 e c } (the quadratic elements here generate k q [SO 3 ]). We can add to this a multiple of θ ⊗ θ since this is also invariant [30], which we have now fixed so that g is central. Thus for example, (e c ⊗ e b + q 2 e b ⊗ e c )a = e c ⊗ ae b + q 3 λe b ⊗ be a + q 2 e b ⊗ ae c = a(e c ⊗ e b + q 2 e b ⊗ e c ) + qλbe a ⊗ e b + q 3 λbe b ⊗ e a + q 4 λ 2 ae a ⊗ e a (e z ⊗ e z − θ ⊗ θ)a = −(1 + q −2 )(λe a ⊗ e a + e a ⊗ e d + e d ⊗ e a )a = a(e z ⊗ e z − θ ⊗ θ) − (1 − q −4 )(q 2 λae a ⊗ e a + qbe b ⊗ e a + q −1 be b ⊗ e a ) using the above commutation relations. Comparing these we see that [g, a] = 0. Similarly for the other generators of k q [SL 2 ]. It is also clear that ∧(g) = 0. The inverse is immediate. For Vol the element e b e c e z is invariant again for reasons coming from the representation theory of k q [SO 3 ]. As θ is also invariant, we know that e b e c e z θ is invariant and hence so is Vol being a multiple of this. For centrality, we check for example ae a e b e c e d = q −1 e a ae b e c e d = e a e b e c q −1 ae d = e a e b e c e d a discarding unwanted terms using the wedge product relations.
We now use g to identify Λ 1 * ∼ =Λ 1 and compute using Vol = 1. Proof. We first explicitly compute the exp element in the form exp = 1⊗1+g+e i1 e i2 ( 2 B) −1 IJ ⊗e j1 e j2 +e i1 e i2 e i3 ( 3 B) −1 IJ ⊗e j1 e j2 e j3 +e 1 e 2 e 3 e 4 ( 4 B) −1 ⊗e 1 e 2 e 3 e 4 where e i , 1 ≤ i ≤ 4 refer in order to e a , e b , e c , e d and I = (i 1 , j 2 , · · · , i m ) with i 1 < i 2 · · · < i m labels of a basis of Λ m and m B IJ = e i1 · · · e im , e j1 · · · e jm = ev(e i1 ⊗ e i2 · · · ⊗ e im , [m, −Ψ]!(e j1 ⊗ e j2 · · · e jm )) = g i1p1 · · · g impm [m, −Ψ]! pm···p2p1 j1j2···jm In the last line refer operators to matrices, for example [2, −Ψ](e m ⊗ e n ) = e p ⊗ e q [2, −Ψ] pq mn and we remember the metric identification where g ij = (e i , e j ). This is the general picture but with bases labelled in the classical way in the present example for generic q. We obtain in the basis enumerations 12, 13, 14, 23, 24, 34 and 123, 124, 134, 234 respectively for the middle cases here. In particular, we see that µ = q 6 . The matrix 1 B −1 here is the inverse of the matrix g ij in our basis and necessarily gives the coefficients of the metric g ∈ Λ 1 ⊗ Λ 1 , and we note also that µ = 1/ det(g) in this basis. We then carefully integrate against this exp, for example where we read from the 2nd row of 2 B −1 for the terms in exp of the form e a e c ⊗ · · · and from the last row for terms of the form e c e d ⊗· · · . The other possibilities in our basis for the first tensor factor of exp have zero integral. We then evaluate the first displayed integral as −q 2 and the second integral as λ on using the relations of the exterior algebra, to give q 4 e b e d +(1−q 4 )e a e b as stated. Integrating against g is easier and gives on degree 3. We could now deduce on degree 1 using Proposition 3.8 that 2 = µ on degrees D = 2 but one can also compute it similarly by integrating against 3 B −1 for a direct calculation and then verify 2 . The polynomial identify for on degree 2 is a direct calculation.
We see that on degree 2 is not of finite order for generic q but is a deformation of order 2. Indeed, µ −1 = det(g) (see the proof above) suggests a geometric normalisation to = µ − 1 2 = q −3 in our case, then is involutive on degree D = 2 and obeys the standard q-Hecke relation 2 = id + (q − q −1 ) on degree 2. This is the same relation as obeyed by the braiding in the defining representation of the quantum group, which is also the braiding on the generators of the associated quantum plane. One can also compute directly and verify that it is given on degree 2 by q 6 S −1 as in Corollary 3.5 (and otherwise equals ). Here the braided antipode on degree 2 is obtained from S(e i e j ) = − ·Ψ(e i ⊗ e j ) as One may similarly compute S on degrees 3,4 to find S = (−1) D id on all degrees D = 2 as must be the case by Proposition 3.8. Note that this feature of the antipode is not true for the outer degrees of all braided exterior algebras, see Example 3.4. Finally, one can check that [ , S] = 0 on all degrees as it must by Corollary 3.5, which in degree 2 provides a very good cross-check of both the displayed S and computations.
where k, n, m are now the degree operators for the powers of c, b, d respectively when acting on a monomial. Then ∂ z = λ (2)q (A + T ) and ∂ 0 = λ (2)q (B + T ) and (2)q λ (1−q m+n−k ) commutes with T (since the latter changes both k, n equally and does not change m). Putting in these results and the value of A + B we obtain ∂ b ∂ c + q −2 ∂ c ∂ b + q −3 (2) q (∂ z 2 − ∂ 02 ) = 2q −1 λ 2 ∆ q . This can also be used to expresses ∂ 0 or ∆ q in terms of ∂ b , ∂ c , ∂ z .
It remains to check that δe i = 0 so that Corollary 3.9 applies and this is the Hodge Laplacian. Note that we have been working algebraically but, over C, our constructions are compatible with the * -involution corresponding to the compact real form SU 2 .