# Switching to a Matrix Hamiltonian - Conceptual Issues

Switching to a Matrix Hamiltonian -- Conceptual Issues

It's probably very clear and well-established for those who rigorously studied Quantum Mechanics but I don't think what I am going to ask is easily 'google'-able but if it is so - please send me to the correct source before spending time.

But don't recommend the whole Landau-Lifgarbagez QM text please.

The thing is, I am routinely performing numerical simulations that involve a discretized single particle, one-band effective mass Hamiltonian almost everyday. I discretize free space (note that I am using an effective mass, so that's okay even for an electron moving in a solid
which is okay.

Which is not okay is that if I don't use an effective mass approach and decide to go to an ATOMISTIC hamiltonian which could be derived from first principles the lattice I am going to work on will be discrete by itself! The point is, everything is made up of atoms and I will have to work on a discrete lattice anyway.

And this Hamiltonian will be exact if I am not mistaken. Now the question:

How do you start from an analytical Hamiltonian and obtain an exact matrix representation??

I kind of know everything (numerical discretization, real lattice discretization, eigenspace discretization,) boils down to the concept of basis functions but I just can't connect the dots.

I hope there'll be some interest in this,

## Answers and Replies

For me the grammar in some parts of your post is hard to understand, you may want to rephrase your question.

If $|\psi_i\rangle$ form a basis then the matrix elements $h_{i j}$ of the Hamiltonian in that basis are given by $h_{i j} = \langle\psi_i|H|\psi_j\rangle$. I doubt this answers your question, but it seems related.

How do you discretize the Hamiltonian?

Is it exact?

What is the real space representation?

What is analytical and what is purely numerical?

Hans de Vries
Science Advisor
Gold Member

How do you discretize the Hamiltonian?

Working on a discrete lattice you might typically use:

$$\frac\tau2~\Big\{~-1~,~~0~~,~+1~\Big\}$$ for first order differentiation and

$$\tau^2\Big\{~~~1~,~-2~,~~~1~~~\Big\}$$ for second order differentiation.

The rest should stay the same. Tau determines the time steps.

Regards, Hans