1: \begin{abstract}

2: Standard perturbation theory of eigenvalue problems consists of obtaining approximations of eigenmodes in the neighborhood of a Hamiltonian where the corresponding eigenmode is known.

3:     Nevertheless, if the corresponding eigenmodes of several nearby Hamiltonians are known, standard perturbation theory cannot simultaneously use all this knowledge to provide a better approximation. We derive a formula enabling such an approximation result, and provide numerical examples for which this method is more competitive than standard perturbation theory.

4: % 	In eigenvalue problems, standard perturbation theory consists in obtaining approximations of eigenmodes at a point $G$ of the parameter space when the eigenmodes at another point $G_1$ of the parameter space are known. Nevertheless, if we know eigenmodes at $n \in \N$, $n \ge 2$ points $G_j$, $j \in \{1,\dots,n\}$, standard perturbation theory cannot simultaneously use all this knowledge to provide a better approximation. We derive a formula enabling to do so, and provide numerical examples for which this method is more competitive than standard perturbation theory.

5: % Consider a parameter space $\cG$, a familly of operators $H(G)$ where $G \in \cG$, and $\cF(G)$ their solution, being the eigenprojector at a given level, the lowest one for instance. Considering $G$ in the affine space spanned by the $G_i$'s, we study a Multipoint formula for resolvents enabling to approximate $\cF(G)$ by using all the $\cF(G_i)$'s simultaneously. We can expand it in $\max_{i,j} \norm{G_i - G_j}^2$, significantly accelerating the convergence with respect to standard methods as perturbation theory. % We present numerical applications of our formula to Schrödinger operators and to graph Laplacians.

6: % \cgd{mention the symmetry groups as well}

7: \end{abstract}

8: