Lecture 6: The Variational Principle

Another powerful approximation to solve stationary Schrodinger equation

Simple derivation of the variational principle Simple examples Connections with stationary perturbation theory

The claim from the variational principle:

Given an arbitrary Hamiltonian that has a

ground state with an unknown eigenvalue

Then for any normalized trial state in the Hilbert space associated with that Hamiltonian:

We must have

Expectation value of the Hamiltonian on an arbitrary state can only approach, but never under-estimate, the ground state energy

Simple proof of the claim

A.

Choosing the basis states of the Hilbert space as the normalized and orthogonal eigenstates of the given Hamiltonian:

Ground state:

B.

Expanding the trial state

C.

Evaluating expectation value and it is done:

Searching for the bottom: What you need to know before you buy try

The trial state is rather arbitrary, but should be legitimate. For example, does it satisfy the boundary condition etc? is it normalized? If you are not able to predict a lower value, it doesnt mean there is no such lower value of the ground state energy. Try differently ! If your trial state is always orthogonal to the ground state, then the bottom we can reach is at most the first excited state energy. This observation is useful for predicting the eigenvalue of the first excited state.

A little bit calculus: Simple Gaussian integrals:

Such kind of Gaussian integrals will be useful because a Gaussian profile is frequently used in variational principle calculations

Example I: Harmonic oscillator

Hamiltonian: Trial wavefunction Normalization:

After two Gaussian integrals

Varying b to get a bottom of this trial: Exact answer!

Example II: delta-function potential

Hamiltonian: Same trial wavefunction as in example I

After one Gaussian integral and one trivial integral:

Varying b to get a bottom of this trial:

>
Higher than the true ground state, but enough to predict a bound state !

Example III: linear potential V(x)= |x|

Hamiltonian: Same trial wavefunction as in examples I,II

After two Gaussian Integrals:

Varying b to get a bottom of this trial:

Estimated ground state energy

Connections with non-degenerate perturbation theory:

In 1st order non-degenerate perturbation theory

Hence the approximate ground state energy is given by

According the variational principle, this guess of E using the zeroth order wavefunction is always higher than the true ground state energy (yes, regarding the zeroth-order wavefunction as one possible trial wavefunction)

Continuing the previous slide: As the 1st order correction over-estimates the ground state energy, the 2nd order is expected to be negative. This is indeed the case as seen in the 2nd order formula applied to the ground state case:

the denominator is always negative because is the lowest value of the unperturbed system

Summary:
Variational principle offers a powerful method for analytical or computational estimates of an upper bound of the ground state energy. There is probably no such thing as best strategy for the trial wavefunctions. As you use it more, you may develop a better sense of what kind of trial wavefunctions are good variational wavefunctions. The method can go well beyond the perturbative regime. Its importance for atomic and molecular physics will be illustrated by two examples: Hydrogen molecular ion and Helium.