Tensors with eigenvectors in a given subspace (Q2120244)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Tensors with eigenvectors in a given subspace
scientific article

    Statements

    Tensors with eigenvectors in a given subspace (English)
    0 references
    0 references
    0 references
    31 March 2022
    0 references
    The authors study the Kalman varieties for symmetric tensors and for certain more general tensors, extending previous results obtained for \(n\times m\) matrices. In particular, they prove that these varieties are irreducible and they compute their codimensions and degrees. In the article a \(d\)-dimensional linear subspace \(L\subset \mathbb{C}^n\) is fixed. For a symmetric tensor \(T \in \mathrm{Sym}^k\mathbb{C}^n\), a vector \(v \in \mathbb{C}^n\) is an eigenvector of \(T\) if \(\exists \lambda \in \mathbb{C}\) such that \[ Tv^{k-1}:= \left( \sum_{i_2,\dots, i_k=1}^n T_{i,i_2,\dots, i_k}v_{i_2}\cdots v_{i_k}\right)_i=\lambda v \ . \] Similarly, for a tensor \(T \in \mathbb{C}^{n_1}\otimes \ldots \otimes \mathbb{C}^{n_k}\), a \(k\)-tuple \((v_1,\ldots, v_k)\in \mathbb{C}^{n_1}\setminus \{0\}\times \ldots \times \mathbb{C}^{n_k}\setminus \{0\}\) is a singular \(k\)-tuple for \(T\) if for any \(i=1,\ldots,k\) exists \(\lambda_i \in \mathbb{C}\) such that \[ T(v_1,\dots, \widehat{v_i}, \dots, v_k)=\lambda_i v_i \ . \] Then one defines the Kalman varieties \[ \kappa_{d,n,k}^s(L):=\left\{ f \in \mathrm{Sym}^k\mathbb{C}^n \ | \ \exists v \in L \ \text{eigenvector for} \ f \right\} \ , \] \[ \kappa_{d,n_1,\ldots,n_k}(L):=\left\{T \in \mathbb{C}^{n_1}\otimes \ldots \otimes \mathbb{C}^{n_k} \ | \ \exists (v_1,\ldots, v_k) \text{ singular } k\text{-tuple for } T, \ v_1 \in L\right\} \ . \] The authors prove that \(\kappa_{d,n,k}^s(L)\) (Proposition 2.1) and \(\kappa_{d,n_1,\dots,n_k}(L)\) (Proposition 3.2) are irreducible and have codimensions \(n-d\) and \(n_1-d\) respectively: the images of the projections from the incidence varieties \[ \{(f,v) \ | \ v \ \text{eigenvector of} \ f\}\rightarrow \mathbb{P}(\mathrm{Sym}^k\mathbb{C}^n) \ , \] \[ \{(T,(v_1,\dots, v_k)) \ | \ (v_1,\dots, v_k) \text{ singular } k\text{-tuple for } T \} \rightarrow \mathbb{P}(\mathbb{C}^{n_1}\otimes \ldots \otimes \mathbb{C}^{n_k}) \] restricted to the conditions \(\{v\in L\}\) and \(\{v_1 \in L\}\) are exactly the Kalman varieties. Moreover, the degrees of \(\kappa_{d,n,k}^s(L)\) and \(\kappa_{d,n_1,\ldots, n_k}(L)\) are computed (Theorem 2.3; Theorem 3.4) by exhibiting a vector bundle on \(\mathbb{P}(\mathrm{Sym}^k\mathbb{C}^n)\times \mathbb{P}(\mathbb{C}^n)\) and \(\mathbb{P}(\mathbb{C}^{n_1}\otimes \ldots \otimes \mathbb{C}^{n_k})\times \mathbb{P}(\mathbb{C}^{n_1}\times \ldots \times \mathbb{C}^{n_k})\) respectively with a section vanishing on the incidence varieties above.\\ Finally, the results are applied to the Kalman variety of \(n\times m\) matrices for \(n\leq m\): in Proposition 3.6 its degree is computed, and in Lemma 3.8 a sufficient and necessary condition for the existence of singular pairs is given.
    0 references
    eigenvectors
    0 references
    tensors
    0 references
    singular tuples
    0 references
    vector bundles
    0 references
    Chern classes
    0 references
    Kalman varieties
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references