# Schrödinger equation approximation – continuity of eigenvalues with respect to potential

The question has been crossposted from Stackexchange after receiving no answers.

Setup: the time-independent Schrödinger equation (eigenvalue problem):

$$(-\frac{\hbar^2}{2m}\Delta +V)\psi = E\psi$$

(On any 'reasonable' domain $$D \subset \mathbb{R^3}$$ with 'reasonable' boundary conditions ($$\psi \in L^2(D)$$ atleast) with $$V: D \rightarrow \mathbb{R}$$ a real valued potential function, again can roughly assume it decays at large distances, but can have isolated singularities; can reduce down to a $$1$$D problem to simplify it as well)

When dealing with computing bound states/bound state energies of say an electron in complicated potentials, one trick I have seen informally is that we can replace the given complicated potential with something that has 'similar behaviour' but has known eigenstates/eigenvalues to 'estimate' the eigenstates/eigenvalues. What I wish to know is, under what conditions is something like this scheme faithful/what conditions need to be imposed to achieve this kind of approximation? Roughly speaking, If $$V$$ is the given potential, and $$V_{\epsilon}$$ is another potential whose eigenvalues and eigenstates are known, such that $$\|V_\epsilon-V\|<\epsilon$$ (in some norm) then if $$E_{\epsilon}, E$$ denote the ground state energies (the lowest energy eigenvalues) in $$V_{\epsilon}, V$$ respectively, then what is an estimate on $$|E_{\epsilon}-E|$$? Is it of the order $$\epsilon$$?

I do not know what this class of problems is called so I could not look for it by just googling for example. Any references on it are requested.

• isn't this just what we try to achieve with perturbation theory? assuming the ground state is not degenerate, the correction $\delta E=E-E_\epsilon$ equals $\int (V-V_\epsilon)|\Psi_\epsilon|^2 d^3 r$ plus terms of order $\epsilon^2$ (with $E_\epsilon$ and $\Psi_\epsilon$ the ground state eigenvalue and eigenfunction in the potential $V_\epsilon$) Mar 18 at 11:08
• Thank you for your comment; could you also provide a source for such an analysis? With respect to how the error terms could be controlled Mar 18 at 11:11
• Well in this case there is no parameter valued expansion I am assuming a-priori; is that the only case in which some scheme like this can be justified? @CarloBeenakker Mar 18 at 11:13
• Denoting the unperturbed potential by $V_{\epsilon }$ is maximally confusing here. Write $V=V_0 + \epsilon V'$ with the $V_0$ problem solvable and develop the perturbation series in $\epsilon$. Unless there are degeneracies in the spectrum of the $V_0$ problem, the spectrum of the $V$ problem can indeed be written as a power series in $\epsilon$. Mar 18 at 13:40
• I see, thank you for your inputs. Mar 18 at 14:11