Positive semi-definiteness in a group algebra (Q1124866): Difference between revisions

Let \(U\) be a finite dimensional unitary space and \(L(U)\) be the ring of all linear operators \(U\to U\). For a finite group \(G\) consider the group ring \(L(U)G\) of \(G\) with coefficients in \(L(U)\). Given an element \(M=\sum_{\sigma\in G} M(\sigma)\sigma\in L(U)G\), define \(M^*\in L(U)G\) by \(M^*(\sigma)=M(\sigma^{-1})^*\), where \(\sigma\in G\). Then the mapping \(M\to M^*\) is an involution on \(L(U)G\), i.e., a ring antiautomorphism of order \(2\) such that \((aM)^*=\overline aM^*\), for all \(a\in\mathbb{C}\), \(M\in L(U)G\). An element \(M\in L(U)G\) is called positive semi-definite (p.s.d.) if \(M=N^*N\) for some \(N\in L(U)G\). In Section 2 the author proves several necessary and sufficient conditions for an element in \(L(U)G\) to be p.s.d. Denote by \(Q\) the regular complex representation of \(G\) and by \(A_1,\dots,A_h\) the complete list of pairwise inequivalent irreducible unitary representations of \(G\). Then for \(M\in L(U)G\) the following are equivalent (see Theorem 2.2, Theorem 2.3): (a) \(M\) is p.s.d. (b) \(\sum_{\sigma\in G} M(\sigma)\otimes A(\sigma)\) is p.s.d. for every unitary representation \(A\) of \(G\); (c) \(\bigoplus_{j=1}^h(\sum_{\sigma\in G} M(\sigma)\otimes A_j(\sigma))\) is p.s.d.; (d) \(\sum_{\sigma\in G} M(\sigma)\otimes Q(\sigma)\) is p.s.d.; (e) \(\sum_{\sigma\in G} Q(\sigma)\otimes M(\sigma)\) is p.s.d. Here \(M(\sigma)\) is considered as the matrix of the operator \(M(\sigma)\) with respect to a fixed orthonormal basis of \(U\). Define an inner product \((-,-)\) on \(L(U)G\) by \((M,N)=\text{tr}\sum_{\sigma\in G} M(\sigma)N(\sigma)^*\), where \(M,N\in L(U)G\). Theorem 2.5 says that \(M\in L(U)G\) is p.s.d. if and only if \((MN,N)\geq 0\) for all \(N\in L(U)G\). An application to symmetric operators is given in Section 3 (Theorem 3.1).