Linear Algebra and Quantum Chemistry

From MaRDI portal





Let \(\{f_ 1,\dots,f_ N\}\) be a basis of unit vectors in \({\mathcal H}=C^ N\). Orthogonalization means an operator \(B\in{\mathcal B}({\mathcal H})\) such that \(Bf_ n=e_ n\) satisfies \(\langle e_ n,e_ m\rangle=\delta_{mn}\), \(1\leq m\), \(n\leq N\). The authors study here the Löwdin problem (L.Pr.) and its extensions associated with the understanding of properties in terms of electron pair transferability. The L.Pr. is to choose \(B\) to minimize the least squares deviation \(\sum_{n=1}^ N\| f_ n-e_ n\|^ 2\). Let \(B=UL\) (\(L=| B|=(B^*B)^{1/2}\)). The L.Pr. becomes to find a unitary matrix \(U\) to attain \(\min_{U\in{\mathcal U}(N)}\| L^{-1}-U\|_{{\mathcal G}_ 2}^ 2\), where \(\| A\|_{{\mathcal G}_ 2}^ 2=\text{tr}(A^*A)\). They solve L.Pr. by finding the \(U\). Next they show the following two extensions of the L.Pr.; Let \(\{g_ 1,\dots,g_ N\}\) be a given orthonormal basis in \({\mathcal H}\). The first extension is to choose \(B\) to minimize \[ F_ b(B)=b\sum_{n=1}^ N\| f_ n-e_ n\|^ 2+(1-b)\sum_{n=1}^ N\| g_ n-e_ n\| \qquad (0\leq b\leq 1). \] The second extension is to choose \(U\in{\mathcal U}(N)\) to minimize \[ G_ b(U)=\sum_{n=1}^ M b_ n\| A_ n-U\|_{{\mathcal G}_ 2}^ 2. \] Here \(A_ 1,\dots,A_ M\in{\mathcal B}({\mathcal H})\), \(b_ 1,\dots,b_ M\in R\), \(b_ n\geq 0\) (\(n=1,\dots,M\)) and \(\sum_{n=1}^ M b_ n=1\). They solve their two extensions. Finally they show precisely the following contribution of Prof. J. Arazy to their problems: That is, by using Mirsky's inequality they show that the solutions of the Löwdin problems in the various norms \(\|\cdot\|_{{\mathcal G}_ E}\) are the same \(U\in{\mathcal U}(N)\).











This page was built for publication: Linear Algebra and Quantum Chemistry

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3986353)