Abstract. We introduce a class of Banach algebras of generalized matrices and study the existence of approximate units, ideal structure, and derivations of them.
IntroductionLet X be a compact metrizable space and m be a Borel probability measure on X. In this note we study some aspects of the algebraic structure of a Banach algebra M of generalized complex matrices whose their arrays are indexed by elements of X 2 and vary continuously. The multiplication of M is defined similar to the ordinary matrix multiplication and uses m as the weight for arrays. See Section 2 for exact definition. In the case that m has full support, M is isometric isomorphic to a subalgebra of compact operators acting on the Banach space of continuous functions on X. Indeed any element of M defines an integral operator in a canonical way. Thus M can be interpreted as a Banach algebra of integral operators or kernels ([2]). In Section 3 we investigate the existence of approximate units of M. In Section 4 we show that if X is infinite then the center of M is zero. In Section 5 we study ideal structure of M. In Section 6 we consider some classes of representations of M. In Section 7 we show that under some mild conditions bounded derivations on M are approximately inner.Notations. For a compact space X and a Banach space E we denote by C(X; E) the Banach space of continuous E-valued functions on X with supremum norm. We also let C(X) := C(X; C). There is a canonical isometric isomorphism C(X; E) ∼ = C(X)⊗E wherě ⊗ denotes the completed injective tensor product. The phrase "point-wise convergence topology" is abbreviated to "pct". By pct on C(X; E) we mean the vector topology under which a net (f λ ) λ ∈ C(X; E) converges to f if and only if f λ (x) → f (x) in the norm of E for every x ∈ X. If f and f ′ are complex functions on spaces X andThe support of a Borel measure m is denoted by Spm. B x,δ denotes the open ball with center at x and radius δ.
The main definitionsLet X be a compact metrizable space and m be a Borel probability measure on X. By analogy with matrix multiplication we let the convolution of f, g ∈ C(X 2 ) be defined by f ⋆ g(x, y) = X f (x, z)g(z, y)dm(z). Also by analogy with matrix adjoint we let f * ∈ C(X 2 ) be defined by f * (x, y) =f (y, x). It is easily verified that ⋆ is an associative multiplication, * is an involution, and also, f ⋆ g ∞ ≤ f ∞ g ∞ and f * ∞ = f ∞ ; thus C(X 2 ) becomes 2010 Mathematics Subject Classification. 46H05; 46H35; 46H10; 47B48; 46H25.