Orthogonal decompositions and idempotent configurations in semisimple associative algebras. (Q941925)

The survey paper under review summarizes the author's research in the area for a couple of decades. All algebras considered in the paper are associative, semisimple, and finite-dimensional over the complex numbers. The trace of such an algebra is the trace \(Tr\) of its regular representation. Let \(A\) be an algebra and let \(B_1,\dots,B_r\) be subalgebras of \(A\) (containing the identity element of \(A\)). The family \(D=\{B_1,\dots,B_r\}\) is an \textit{orthogonal decomposition} of \(A\), OD for short, if all \(B_i\) are semisimple and the vector subspaces \(B_i^\circ\) are pairwise orthogonal with respect to the trace form on \(A\). Here \(B_i^\circ\) stands for the traceless elements of \(B_i\). In case all \(B_i\) are commutative then the OD is commutative. One may consider the primitive idempotents in each \(B_i\); the collection of these idempotents satisfies ``nice'' conditions involving the trace of \(A\). The author considers a generalization of the notion of combinatorial \(t\)-designs, called balanced \(t\)-systems. These are collections of idempotents \(e_1,\dots,e_v\) such that \(e_1+\cdots+e_v=k\) for some complex number \(k\), and for every choice of \(e_{i_1},\dots,e_{i_s}\), \(s\leq t\), one has \(Tr(e_{i_1}\cdots e_{i_s})=\lambda_s\) where \(\lambda_s\) depends only on \(s\) but neither on the idempotents nor on their order. The survey comprises three parts. The first of them (including, roughly speaking, Chapters 2, 3, and 4) is devoted to the orthogonal decomposition and the well known divisibility conjecture. Recall that the latter conjecture asks whether an algebra is a free module over the subalgebras defining the OD. The central result of the first part is Theorem 2.2.1 that answers the above conjecture in the affirmative when the OD is commutative. A variety of interesting applications of this theorem is collected as well. The second part of the survey studies systems of idempotents, and consists of Chapters 5 to 8. The author studies the automorphisms of balanced systems of idempotents and gives bounds for the number of the idempotents in such systems. Furthermore he describes the balanced systems in matrix algebras whose automorphism group is 2-transitive. A complete classification is obtained when the group is \(\text{PSL}_2(q)\). Later on he introduces the notion of a balanced basis, namely a balanced system that is a basis of the vector space of the algebra. Such algebras are called balanced. A strong relationship between such bases and symmetric designs in the commutative case is described. Moreover balanced bases for the matrix algebras of order \(p^n\) are constructed when \(p>2\) is a prime; and it is shown that there are no such balanced bases when \(p=2\). Part 3 of the survey studies the automorphisms of the OD for an algebra. Recall that if \(D\) is OD for the algebra \(A\) then the group \(\Aut(D)\) consists of the automorphisms that preserve the set \(\{B_1,\dots,B_r\}\). A well known conjecture asks whether \(\Aut(D)\) is finite. This is shown to be true for several classes of OD, and various generalizations are considered. The survey ends with a list of open problems in the area. The list of references consists of 41 titles and is quite complete. This survey may be quite interesting for specialists in Ring theory as well as Combinatorial algebra. Many of the topics discussed in it originate from the Lie theory, and admit direct applications to Lie algebras.