|This article needs additional citations for verification. (December 2009)|
In mathematics, a bialgebra over a field K is a vector space over K which is both a unital associative algebra and a coalgebra, such that the algebraic- and coalgebraic structure satisfy certain compatibility relations. Specifically, the comultiplication and the counit are both unital algebra homomorphisms, or equivalently, that the multiplication and the unit of the algebra both be coalgebra morphisms. These statements are equivalent in that they are expressed by the same commutative diagrams. A bialgebra homomorphism is a linear map that is both an algebra and a coalgebra homomorphism.
As reflected in the symmetry of the commutative diagrams, the definition of bialgebra is self-dual, so if one can define a dual of B (which is always possible if B is finite-dimensional), then it is automatically a bialgebra.
(B, ∇, η, Δ, ε) is a bialgebra over K if it has the following properties:
- B is a vector space over K;
- there are K-linear maps (multiplication) ∇: B ⊗ B → B (equivalent to K-multilinear map ∇: B × B → B) and (unit) η: K → B, such that (B, ∇, η) is a unital associative algebra;
- there are K-linear maps (comultiplication) Δ: B → B ⊗ B and (counit) ε: B → K, such that (B, Δ, ε) is a (counital coassociative) coalgebra;
- compatibility conditions expressed by the following commutative diagrams:
- Multiplication ∇ and comultiplication Δ
- where τ: B ⊗ B → B ⊗ B is the linear map defined by τ(x ⊗ y) = y ⊗ x for all x and y in B,
- Multiplication ∇ and counit ε
- Comultiplication Δ and unit η
- Unit η and counit ε
Coassociativity and counit
The K-linear map ε: B → K is a counit if .
Coassociativy and counit are expressed by the commutativity of the following two diagrams with B in place of C (they are the duals of the diagrams expressing associativity and unit of an algebra):
These statements are meaningful once we explain the natural structures of algebra and coalgebra in all the vector spaces involved besides B: (K, ∇0, η0) is a unital associative algebra in an obvious way and (B ⊗ B, ∇2, η2) is a unital associative algebra with unit and multiplication
so that or, omitting ∇ and writing multiplication as juxtaposition, ;
similarly, (K, Δ0, ε0) is a coalgebra in an obvious way and B ⊗ B is a coalgebra with counit and comultiplication
Then, diagrams 1 and 3 say that Δ: B → B ⊗ B is a homomorphism of unital (associative) algebras (B, ∇, η) and (B ⊗ B, ∇2, η2)
- , or simply Δ(xy) = Δ(x) Δ(y),
- , or simply Δ(1B) = 1B ⊗ B;
diagrams 2 and 4 say that ε: B → K is a homomorphism of unital (associative) algebras (B, ∇, η) and (K, ∇0, η0):
- , or simply ε(xy) = ε(x) ε(y)
- , or simply ε(1B) = 1K.
Equivalently, diagrams 1 and 2 say that ∇: B ⊗ B → B is a homomorphism of (counital coassociative) coalgebras (B ⊗ B, Δ2, ε2) and (B, Δ, ε):
diagrams 3 and 4 say that η: K → B is a homomorphism of (counital coassociative) coalgebras (K, Δ0, ε0) and (B, Δ, ε):
A simple example of a bialgebra is the set of functions from a group G to , which we may represent as a vector space consisting of linear combinations of standard basis vectors eg for each g ∈ G, which may represent a probability distribution over G in the case of vectors whose coefficients are all non-negative and sum to 1. An example of suitable comultiplication operators and counits which yield a counital coalgebra are
which represents making a copy of a random variable (which we extend to all by linearity), and
(again extended linearly to all of ) which represents "tracing out" a random variable — i.e., forgetting the value of a random variable (represented by a single tensor factor) to obtain a marginal distribution on the remaining variables (the remaining tensor factors). Given the interpretation of (Δ,ε) in terms of probability distributions as above, the bialgebra consistency conditions amount to constraints on (∇,η) as follows:
- η is an operator preparing a normalized probability distribution which is independent of all other random variables;
- The product ∇ maps a probability distribution on two variables to a probability distribution on one variable;
- Copying a random variable in the distribution given by η is equivalent to having two independent random variables in the distribution η;
- Taking the product of two random variables, and preparing a copy of the resulting random variable, has the same distribution as preparing copies of each random variable independently of one another, and multiplying them together in pairs.
A pair (∇,η) which satisfy these constraints are the convolution operator
again extended to all by linearity; this produces a normalized probability distribution from a distribution on two random variables, and has as a unit the delta-distribution where i ∈ G denotes the identity element of the group G.
Other examples of bialgebras include the Hopf algebras. Similar structures with different compatibility between the product and comultiplication, or different types of multiplication and comultiplication, include Lie bialgebras and Frobenius algebras. Additional examples are given in the article on coalgebras.
- Dăscălescu, Năstăsescu & Raianu (2001). Hopf Algebras: An introduction. pp. 147 & 148.
- Dăscălescu, Năstăsescu & Raianu (2001). Hopf Algebras: An introduction. p. 148.
- Dăscălescu, Năstăsescu & Raianu (2001). Hopf Algebras: An introduction. p. 151.