Central polynomials of graded algebras: capturing their exponential growth (Q2118936)

Let \(G\) be a finite abelian group and \(F\) a fixed field of characteristic 0. Assume \(A\) is a \(G\)-graded associative algebra, and denote by \(F\langle X,G\rangle\) the free \(G\)-graded algebra freely generated by \(X\) over the field \(F\). A polynomial \(f\in F\langle X,G\rangle\) is a (graded) central polynomial for \(A\) if it takes central values in \(A\) for each substitution of the variables with elements of \(A\) respecting the grading. If \(f\) takes only zero value on \(A\) then \(f\) is a graded identity for \(A\). The central polynomials without any gradings are extremely important in the structure theory of PI algebras. Their existence for the full matrix algebras was conjectured by Kaplansky, and established by [\textit{E. Formanek}, J. Algebra 23, 129--132 (1972; Zbl 0242.15004)], and independently by [\textit{Ju. P. Razmyslov}, Math. USSR, Izv. 7, 479--496 (1974; Zbl 0314.16016); translation from Izv. Akad. Nauk SSSR, Ser. Mat. 37, 483--501 (1973)]. The set of all central polynomials is a vector subspace (in fact a subalgebra) of the free algebra which is closed under endomorphisms. Such vector spaces are called T-spaces. Studying the central polynomials for a given algebra is difficult and their complete description seems to be at the same level of difficulty as that of describing all polynomial identities for the algebra. That is why one is led to study the numerical invariants of T-spaces. As it turns out in the case of the identities for an algebra, introducing a grading on the algebra makes the things somewhat easier. The additional structure, namely the grading, offers some tools unavailable in the case of no grading. A similar phenomenon happens in the case of central polynomials. As the base field is of characteristic 0, instead of studying all identities of an algebra one can consider only the multilinear ones. The same holds for central polynomials. The multilinear polynomials of degree \(n\) form a module over the symmetric group \(S_n\), hence one can apply the theory of representations of \(S_n\) in order to study polynomial identities and central polynomials. One of the most important numerical invariants of a T-ideal (or a T-space) is its codimension sequence. In analogy to the ordinary case one can attach to a \(G\)-graded algebra \(A\) the following codimension sequences. First \(c_n^G(A)\) is the usual graded codimension, that is the dimension of the vector space \(P_n^G\) of the graded multilinear polynomials of degree \(n\) modulo the graded identities of \(A\). Then \(c_n^{G,z}(A)\) is the dimension of \(P_n^G\) modulo its graded central polynomials, and \(\delta_n^G(A)\) is the dimension of the vector space of the graded central polynomials in \(P_n^G\) modulo the graded identities of \(A\). Call these the graded central and the graded proper central codimensions, respectively. It is clear that \(c_n^G(A)=c_n^{G,z}(A)+ \delta_n^G(A)\) for each \(n\). One of the principal directions in PI theory in the last decades has been the study of the asymptotic behaviour of the codimension sequences of PI algebras, both in the ordinary and in the graded case. In [\textit{A. Giambruno} and \textit{M. Zaicev}, Adv. Math. 140, No. 2, 145--155 (1998; Zbl 0920.16012)], and [\textit{A. Giambruno} and \textit{M. Zaicev}, Adv. Math. 142, No. 2, 221--243 (1999; Zbl 0920.16013)] it was proved that for the ordinary codimension sequence \(c_n(A)\) of a PI algebra \(A\) the limit \(\exp(A)=\lim_{n\to\infty} \sqrt[n]{c_n(A)}\) always exists and is a non-negative integer. It is the PI exponent of \(A\). Later on analogues were obtained in the group graded case. The main results of the paper under review consist in proving that similar results hold for graded central polynomials. The authors prove that if \(A\) is an algebra graded by a group \(G\) and admitting graded central polynomials then the graded central as well as the graded proper central exponents exist and are non-negative integers. The authors give a method to compute these graded exponents. Furthermore they relate the three graded exponents. Namely they prove that \(\exp(c_n^{G,z}(A)) = \exp(C_n^G(A))\) or \(\exp(c_n^{G,z}(A)) =0\). The paper is well written and combines deep methods in PI theory in order to obtain interesting new results.