The product of the eigenvalues of a symmetric tensor (Q725522): Difference between revisions

Given a finite-dimensional Euclidean vector space \(V\) and a symmetric tensor \(f\in \text{Sym}^d V,\) a non-zero vector \(x\in V^\mathbb{C}\) such that \(||x||=1\) is called a \(E\)-eigenvector of \(f\) if there exists \(\lambda \in \mathbb{C}\) such that \(x\) is solution of the equation \[ \dfrac{1}{d}\nabla f(x)=\lambda x. \] The scalar \(\lambda\) is called an \(E\)-eigenvalue of \(f\). The author gives a formula for the product of all \(E\)-eigenvalues of \(f\) when \(f\) admits the maximum number of \(E\)-eigenvalues. The formula \[ \lambda_1\cdots\lambda_N=\pm \frac{\text{Res}\left(\frac{1}{d}\nabla (f)\right)} {{\Delta_{\widetilde{Q}}(f)}^{\frac{d-2}{2}}} \] is given in terms of the resultant of \(\frac{1}{d}\nabla(f)\) and of the \(\widetilde{Q}\)-discriminant of \(f\). In order to do this, the author studies the \(E\)-characteristic polynomial of \(f\), which generalizes the characteristic polynomial \(\psi_A(\lambda)=\text{det}(A-\lambda I)\) of a symmetric matrix \(A.\)