The problems of classifying pairs of forms and local algebras with zero cube radical are wild. (Q556885): Difference between revisions

The authors consider, over an algebraically closed field \(F\) of characteristic different from 2, the problem of classifying: (i) pairs of sesquilinear forms in which the second is Hermitian, (ii) pairs of bilinear forms in which the second is symmetric (skew-symmetric), (iii) local algebras with zero cube radical and square radical of dimension 2. They use that classification problems that contain the problem of classifying pairs of matrices up to simultaneous similarity are wild. They show that the three problems are wild. They consider congruent matrices with respect to an involution defined on the field \(F\), that is, two matrices \(A\) and \(B\) are *-congruent if and only if \(B=S^*AS\) for a nonsingular matrix \(S\), where \(S^*\) is the transpose of the matrix whose elements are the images of the elements of \(S\) under the involution. They consider \(t\)-tuples \((A_1,A_2,\dots,A_t)\) of matrices, all the matrices of a \(t\)-tuple will have the same size and this will be called the size of the \(t\)-tuple. Pairs of \(t\)-tuples \((A_1,A_2,\dots,A_t)\) and \((B_1,B_2,\dots,B_t)\) are called equivalent if and only if there exist nonsingular matrices \(R\) and \(S\) such that \((B_1,B_2,\dots,B_t)=R(A_1,A_2,\dots,A_t)S\), these \(t\)-tuples will be called congruent if \(R=S^*\). For each pair of matrices \(A\) and \(B\) of order \(n\), and each \(\varepsilon\) in \(F\) they define a pair \(T_\varepsilon(A,B)\) and they show first that for each \(\varepsilon\) in \(F\), matrix pairs \((A,B)\) and \((C,D)\) are similar over \(F\) if and only if \(T_\varepsilon(A,B)\) and \(T_\varepsilon(C,D)\) are *-congruent. Then they show that the problem of classifying (i) and (ii) is wild. They reduce problem (iii) to the problem of classifying algebras (without identity) with \(R_3=0\) and \(\dim R_2=2\), proving that each of these last algebras is the radical of some local algebra \(\Lambda\) with zero cube radical and square radical of dimension 2 and that \(\Lambda\) is fully determined by \(R\). Thus the new problem is also reduced to a matrix problem, which is proved to be wild and as a consequence problem (iii) is a wild problem.