Star-polynomial identities: computing the exponential growth of the codimensions (Q330168)

Let \(A\) be a PI algebra over a field \(K\) of characteristic 0, then the polynomial identities of \(A\) are determined by the multilinear ones. A powerful tool in studying the multilinear polynomial identities is the usage of the representation theory of symmetric groups. Denote by \(P_n\) the vector subspace of the free associative algebra \(K\langle X\rangle\) spanned by all monomials \(x_{\sigma(1)}\cdots x_{\sigma(n)}\) where \(\sigma\in S_n\), the symmetric group permuting 1, \dots, \(n\). Clearly \(P_n\) is isomorphic to the left regular \(S_n\)-module. If \(T(A)\) is the T-ideal of \(A\) then \(P_n\cap T(A)\) is a submodule of \(P_n\), and the quotient \(P_n(A)=P_n/(P_n\cap T(A))\) is an \(S_n\)-module as well. The structure of the modules \(P_n\cap T(A)\) and \(P_n(A)\) is known for very few ``easy'' cases. One may wish to know then the codimension sequence of \(A\), \(c_n=c_n(A)=\dim P_n(A)\). But the exact values of \(c_n(A)\) are known, too, in very few cases. Therefore one studies the asymptotic behaviour of \(c_n(A)\). A celebrated result, see [the first author and \textit{M. Zaicev}, Adv. Math. 140, No. 2, 145--155 (1998; Zbl 0920.16012); ibid. 142, No. 2, 221--243 (1999; Zbl 0920.16013)], states that \(\lim (c_n(A)^{1/n})\) exists for any PI algebra \(A\) and is a non-negative integer. This limit is called the PI exponent of \(A\), \(\exp(A)\). The latter result has been transferred to large classes of nonassociative algebras as well (but in general it does not hold). One may ask a similar question for graded algebras, for algebras with involution and so on. The present paper establishes the result for algebras with involution. The authors prove the following theorem. Let \(A\) be an associative PI algebra with involution \(*\), and let \(c_n^*(A)\) stand for the \(n\)-th \(*\)-codimension of \(A\). Then there exist constants \(C_1>0\), \(C_2\), \(t_1\), \(t_2\) and \(d\) depending on \(A\), where \(d\) is an integer, such that \(C_1n^{t_1} d^n\leq c_n^*G(A)\leq C_2 n^{t_2} d^n\) for all \(n\). Here \(G(A)\) is the Grassmann envelope of \(A\); it is a superalgebra with an appropriate superinvolution. These inequalities imply that the \(*\)-\(\exp(A)\) is always an integer. Moreover, the authors provide an algorithm to compute the \(*\)-exponent. It should be mentioned that the authors prove, in fact, that the above inequalities hold as well whenever \(A\) is a superalgebra with a superinvolution.