Agenda should store index of terms in the result as well. This required to form the differential equation system.

4 commutations with an agenda on real space (100 sites) yields (term count):

• default build
  • GCC: 1354
  • Clang: 1354
  • Intel: 1354
• release build
  • GCC: 1340
  • Clang: 1767
  • Intel: 1354

Why is that? What is correct?

At many places dx=dy=1 is assumed for the lattice (e.g. ExpectationValue2D). This could be changed so that no dx can be different from dy.

Use a template

template <typename T>
std::ostream& operator<<(std::ostream& strm, const T& t) {
    return (strm << hubbard::format(t));
}

and implement format() functions for all types in the library.

Currently most time is spent in the matrix initialization and not the integrator. Building of the matrix should benefit very much from multiple cores as the elements are calculated independently.

While implementing more routines for the momentum space calculations I noticed, that it might be better to have a folder structure like this:

include
 |-- ieompp
      |-- models
           |-- hubbard_real_space
           |-- hubbard_momentum_space
           |-- hubbard_real_space.hpp
           |-- hubbard_momentum_space.hpp

Most functionality is not shared between momentum space and real space. Otherwise a third folder hubbard_common might make sense.

The regex to parse commutators matches the string --help.

This can be solved by inserting non-vanishing expectation values directly into the target vector.

See TODO note in unittest https://github.com/f-koehler/hubbard/blob/commutation/tests/algebra/test_hamiltonian_fourier.cpp

It seems that momentum is not conserved in all terms

./fermion_cas "[c_{k_1,↑}^†c_{k_1,↑};c_{q,s}] gives the expected output while ./fermion_cas "[c_{k_1,↑}^†c_{k_1,↑};c_{q,σ}] does not.

It would be nice to have values (see commutation with interaction term). So the following will be possible:
δ_{σ_1,↑}

The commutation in momentum space is not yet implemented. Two way are possible and should be tried:

• using the Fourier transformed Hamiltonian
• by Fourier transforming the result of the real space commutation

Title says it all, dispersion is a functor like object.

I already tried to implement it but today I am too stupid.

In the Hamilton<Term>::commutate functions it would be nice to use a reference to a term in the output container for convenient and efficient usage of these methods. I am too tired tonight to fix this.

I have to give more thought about the license to chose for this project.