A new classification of algebraic identities for linear operators on associative algebras (Q2075241)

The paper under review examines the classification of algebraic identities satisfied by linear operators on an associative algebra, \(A\), over a characteristic zero field, \(\mathbb{F}\), a problem broached in [\textit{G.-C. Rota}, Gian-Carlo Rota on combinatorics: Introductory papers and commentaries. Edited by Joseph P. S. Kung. Boston, MA: Birkhäuser. 504--512 (1995; Zbl 0841.01031)]. Let \(B: A \times A \to A\), \(B(x, y) := xy\), denote the associative bilinear product on \(A\), and let \(L : A \to A\) be any linear map. A monomial of degree \(p\) in \(A\) is a given product \(a_1 a_2 \ldots a_p\) of indeterminants. An operator monomial of degree \(p\) and multiplicity \(q\) is a monomial of degree \(p\) that contains \(q\) insertions of \(L\). Let \({\mathcal{M}}(p, q)\) denote the set of all operator monomials of degree \(p\) and multiplicity \(q\), and let \({\mathcal{O}}(p, q)\) denote the vector space over \(\mathbb{F}\) with basis \({\mathcal{M}}(p, q)\). The authors place a lexicographic order on the elements of \({\mathcal{M}}(p, q)\). Let \(\mathcal{O}\) denote the (non-symmetric) operad generated by one unary operation \(L\) and one associative binary operation \(B\). See [\textit{M. R. Bremner} and \textit{V. Dotsenko}, Algebraic operads. An algorithmic companion. Boca Raton, FL: CRC Press (2016; Zbl 1350.18001)] for more on operads in this context. Then \({\mathcal{O}}(p, q)\) is a homogeneous subspace of \(\mathcal{O}\). Any (non-zero) element \( R \in {\mathcal{O}}(p, q)\) is mathcalled an operator identity. A consequence of \(R\) is defined as a new operator identity of higher degree or multiplicity (or both) obtained from any sequence of partial compositions with \(B\) or \(L\) applied to \(R\). The authors study consequences that arise by increasing both \(p\) and \(q\) by one in either order (\(p\) first, \(q\) second or \(q\) first, \(p\) second). For \(R \in {\mathcal{O}}(p, q)\), the matrix of consequences of \(R\), written \(M(R)\), has \((i, j)\) entry as the coefficient of the \(j\)-th monomial in \(M(p+1, q+1)\) of the \(i\)-th consequence. If \(R\) has an indeterminant, then \(M(R)\) is a matrix over a polynomial ring. Studied in the paper is the rank of \(M(R)\) for the cases \(p = 2\), \(q = 1\), and \(p = 2\), \(q = 2\). Consider \(R \in {\mathcal{O}}(2, 1)\) with \[ R = a L(xy) + b L(x)y + cxL(y), \] where \(a\), \(b\), \(c\) are indeterminants in \(\mathbb{F}\). Proven is that \(M(R)\) has rank 14 or 17, with explicit conditions on \(L\) given in the rank 14 case. Now, let \(R \in {\mathcal{O}}(2, 2)\) with \[ R = aL^2(xy) + bL(L(x)y) + cL^2(x)y + dL(xL(y)) + eL(x)L(y) + fxL^2(y), \] where \(a\), \(b\), \(c\), \(d\), \(e\), \(f\) are indeterminants in \(\mathbb{F}\). Then the (transpose) matrix \(M(R)^t\) has rank 16, 19 or 20, with the conditions determining these ranks stated explicitly.