Identification of an algebraic domain in two dimensions from a finite number of its generalized polarization tensors (Q2332932): Difference between revisions

For every bounded and open domain \(G\subset \mathbb{R}^2\), the set of \textit{generalized polarization tensors} correspond to the tensors \[ M_{\alpha,\beta}(G,\lambda) := \int_{\partial G}\left(\lambda - \mathcal{K}^*_G\right)^{-1}[\nu(x)\cdot \nabla x^{\alpha}]y^{\beta}\,d\sigma(y), \] where \(\lambda\in\mathbb{C}\), \(\mathcal{K}_G^*\) denotes the Neumann-Poincare operator and \(\nu\) denotes the outward normal to \(\partial G\). Consider on the other hand the set of polynomials of order \(d\), with domain in \(\mathbb{R}^2\), vanishing on \(\partial G\): \[ P_d(G) := \left\{p(x) = \sum_{|\alpha|\leq d} p_{\alpha}x^\alpha;\quad p(x) = 0\quad \forall x \in\partial G\right\}, \] This paper clarifies a close connection between these two sets. In particular, it is proved that given a domain \(G\) with Lipschitz boundary of degree \(d\) and belonging to the set of algebraic domains, then a suitable subset of \(P_d(G)\) can be identified with the set of real coefficients \(p_{\alpha}\), \(|\alpha|\leq d\), satisfying \[ \sum_{|\alpha|<d}M_{\beta,\alpha}(G,\lambda)p_{\alpha} = 0, \] with \(|\beta|\leq 2d\) and where \(\lambda\) belongs to a discrete subset of \(\mathbb{C}\). The proof of this result relies on a careful study of the holomorphic matrix-valued function \(\mu\rightarrow \left(\mu M_{\alpha,\beta}(G,\mu)\right)_{\alpha,\beta}\). Extensions to non-algebraic domains and the use of this result to numerically reconstruct algebraic domains from their low order polarization tensors are also discussed.