On an \(n\)-Lie algebra of Jacobians (Q1972206)

A vector space \(L\) over a field \(F\) is called an \(n\)-Lie algebra if it is endowed with an \(n\)-ary multilinear operation \([\cdot,\dots,\cdot]: L^n\to L\) which satisfies the generalized anticommutativity law and the Jacobi identities: (i) \([x_1,\dots,x_n]=\text{sign} (\sigma)[x_{\sigma(1)},\dots,x_{\sigma(n)}],\;\) (ii) \([x_1,\dots,x_n]R(y)=\sum_{i=1}^n[x_1,\dots,x_iR(y),\dots,x_n]\), where \(\sigma\in S_n,\;R(y)=R(y_1,\dots,y_{n-1})\) is the right-hand multiplication operator on \(L\) defined by \(R(y)(x)=[x,y_1,\dots,y_n]\). This notion was introduced by the author in [Sib. Math. J. 26, 879-891 (1985); translation from Sib. Mat. Zh. 26, No. 6, 126-140 (1985; Zbl 0585.17002)]. Let \(A\) be a commutative and associative algebra and let \(D_1,\dots,D_n\) be fixed commuting derivations of \(A\); then, for arbitrary \(x_1,\dots,x_n\in A\), we set \([x_1,\dots,x_n]=\det \|D_i(x_j)\|\). The vector space \(A\) is proven to be an \(n\)-Lie algebra with respect to this operation. The author denotes this algebra by \(A^*(D_1,\dots,D_n)\) and calls it an algebra of Jacobians of the algebra \(A\). Further, let \(R_{n,m}\) be a polynomial algebra over \(F\) in \(nm\) variables \(p_{11},\dots,p_{m1},\dots,p_{1n},\dots,p_{mn}\). Define on \(R_{n,m}\) the \(n\)-ary operation \( [f_1,\dots,f_n]=\sum_{i=1}^m\|\frac{\partial {f_j}}{\partial {p_{ik}}}\|. \) It is well known that the algebra \(R_{2,m}\) is a Lie algebra (an algebra of Poisson bracket). The algebra \(R_{n,1}\) is an algebra of Jacobians, with \(D_k=\frac{\partial} {\partial p_{1k}}\), and is therefore an \(n\)-Lie algebra. On the other hand, the author shows that, for arbitrary \(n>2\) and \(m>1\), the algebra \(R_{n,m}\) is not an \(n\)-Lie algebra. Finally, suppose that \(G\) is an abelian group, \(F[G]\) is the group algebra of \(G\) over a field \(F\), \(t=(t_1,\dots,t_n)\in G^n\), and \(h=(h_1,\dots,h_n)\in (G^*)^n\), where \(G^*=\text{Hom}(G,\langle F,+\rangle)\). Define the linear mappings \(D_i\) on \(F[G], i=1,\dots,n\), by setting \(D_i(g)=h_i(g)(g+t_i)\) for \(g\in G\) and then extending \(D_i\) to \(F[G]\) by linearity. Then the algebra \(A(G,h,t)=F[G]^*(D_1,\dots,D_n)\) is an algebra of Jacobians if and only if the equalities \(h_i(t_j)=0\) hold for all \(i\neq j\); moreover, the product in \(A(G,h,t)\) is nontrivial if and only if the mappings \(h_i\) are linearly independent. The author proves that if the field \(F\) is infinite, \(G=F^n\), \(t_i=0\), and \(h_i\) is the projection of \(F^n\) onto the \(i\)th component, then the \(n\)-Lie quotient algebra of \(A(G,h,0)\) by its center \(F\cdot (0,\dots,0)\) is simple. He conjectures that, in the general case, the \(n\)-Lie algebra \(\widetilde A(G,h,0)\) over a field \(F\) of characteristic \(p>n\) or \(p=0\) is simple if and only if the mappings \(h_i\), \(i=1,\dots,n\), are linearly independent and have no common nonzero elements in their kernels.