Codimension growth of simple Jordan superalgebras (Q2130515)

Let \(F\) be a field of characteristic 0, \(F\{X\}\) the free (nonassociative) algebra on the infinite free generating set \(X\), and \(A\) an \(F\)-algebra. Denote by \(Id(A)\) the ideal of all polynomial identities of \(A\) in \(F\{X\}\), and by \(P_n\) the vector space in \(F\{X\}\) of all multilinear elements of degree \(n\) in the variables \(x_1\), \dots, \(x_n\). Since \(F\) is of characteristic 0, it is well known that \(Id(A)\) is determined by its multilinear elements, that is by \(Id(A)\cap P_n\) for \(n\ge 1\). One studies the quotient \(P_n(A)=P_n/(P_n\cap Id(A))\) since, in a sense, it is much ``smaller'' than \(P_n\cap Id(A)\). One of the most important numerical invariants of an algebra \(A\) is its codimension sequence \(c_n(A)=\dim P_n(A)\). It is well known that for \(A\) finite dimensional one has \(c_n(A)\le d^n\) for every \(n\) where \(d=\dim A\). Hence one defines the lower and the upper PI exponents of \(A\) as \(\underline{\exp}(A) = \liminf\sqrt[n]{c_n(A)}\) and \(\overline{\exp}(A) = \limsup\sqrt[n]{c_n(A)}\), respectively; if they coincide, the limit \(\lim_{n\to\infty} \sqrt[n]{c_n(A)}\) is the PI exponent of \(A\). It was proved by \textit{A. Giambruno} and \textit{M. Zaicev} [Adv. Math. 140, No. 2, 145--155 (1998; Zbl 0920.16012); Adv. Math. 142, No. 2, 221--243 (1999; Zbl 0920.16013)] that for every associative PI algebra the PI exponent exists and is an integer. It is also known that the same statement holds for finite dimensional Lie algebras. If \(A\) is simple finite dimensional algebra then \(\exp(A)\) does exist. In the case of simple finite dimensional associative, Lie, alternative, or Jordan algebra and \(F\) is algebraically closed then \(\exp(A)=\dim A\). If one considers superalgebras then the latter equality is not always true. For example, in the case of simple Lie superalgebras the PI exponent may be less than the dimension, see [\textit{A. Giambruno} and \textit{M. Zaicev}, J. Lond. Math. Soc., II. Ser. 85, No. 2, 534--548 (2012; Zbl 1271.17003)]. Since superalgebras are graded by the cyclic group of order two, \(\mathbb{Z}_2\) one is led naturally to study the behaviour of the codimension sequence \(c_n^{gr}(A)\) of a \(\mathbb{Z}_2\)-graded algebra \(A\). As in the ungraded case one defines the graded versions of the PI exponent, and the lower and the upper PI exponent. If \(A\) is finite dimensional then \(c_n^{gr}(A)\le d^n\) where \(d=\dim A\). The inequality \(c_n(A)\le c_n^{gr}(A)\) was obtained by \textit{Y. Bahturin} and \textit{V. Drensky} [Linear Algebra Appl. 357, No. 1--3, 15--34 (2002; Zbl 1019.16011)], thus one gets \(\overline{\exp(A)}\le \overline{\exp^{gr}(A)}\). The authors consider the simple finite dimensional Jordan superalgebra they denote by \(P(t)\). It is a vector subspace of the matrices of order \(2t\), and \(P(t)=P_0\oplus P_1\) where \(P_0\) is formed by all matrices having blocks \(a\) and \(a^*\) of size \(t\times t\) on the main diagonal, and 0 elsewhere; \(P_1\) consists of the block matrices with blocks \(b\) and \(c\) at positions \((1,2)\) and \((2,1)\), respectively, such that \(b^*=b\), \(c^*=-c\). Here \(*\) stands for the usual transpose involution on \(t\times t\) matrices. The authors prove that both the ordinary and the graded PI exponent of \(P(t)\) exist, and moreover these are less than \(\dim P(t)=2t^2\). This fact sharply contrasts the behaviour of the PI exponent in the known cases. Furthermore they prove that the PI exponent of \(P(2)\) is not an integer. They also compute the graded PI exponent, which is not an integer either: \(\exp^{gr}(P(2))=4+2\sqrt{3}\).