Bilinear forms on Frobenius algebras. (Q2575684)

Recall that a finite dimensional algebra \(R\) over a field \(k\) is Frobenius (resp. symmetric) if there exists a nondegenerate associative (resp. symmetric) bilinear form \(B\colon R\times R\to k\). Two such forms \(B\) and \(B'\) are called homothetic if there exist \(V\in\Aut_k(R)\) and \(\alpha\in k^*\) such that \(B'(r,s)=\alpha B(Vr,Vs)\), for all \(r,s\in R\). The author is interested in the following question: when is the (symmetric) form unique up to homothety? This is studied in the case where \(R\) has an ideal \(m\) such that \(R/m\cong k\); this property holds if \(R\) is a finite dimensional Hopf algebra (take \(m=\text{Ker}(\varepsilon)\)). If \((R,m)\) is a local symmetric algebra over a field of characteristic different from 2, and \(R/m\cong k\), then the following results are shown: there exists a unique symmetric form up to homothety; there exists a unique form up to homothety if and only if \(R\) is commutative. In the case where \(R\) is Frobenius, the author proceeds as follows. If \(B\) and \(B'\) are two nondegenerate associative bilinear forms, then there exists \(u\in U(R)\) such that \(B'(r,s)=B(r,su)\). Then he considers the Nakayama automorphism \(\sigma\) of \(R\), measuring how far \(R\) is from being symmetric, and assumes that \(\sigma\) is of finite inner order \(n\). He introduces the norm \(N_\sigma(r)=r\sigma(r)\sigma^2(r)\cdots\sigma^{n-1}(r)\). Then he can show that \(B\) and \(B'\) are homothetic if and only if \(N_\sigma(u)\in Z(R)\); the converse property holds in the case where \(R\) is the so-called extended Nakayama-Nesbitt algebra, and the author conjectures that it holds in general. Finally, in the case where \(R\) is even dimensional, the author shows that the value of the determinant of a form in \(k^*/(k^*)^2\) is an invariant of the algebra.