nakib / elphbolt Goto Github PK

We need Wannier space data from EPW for the various materials we are going to calculate.

Calculate the spectral versions of all the transport coefficients.

Calculate g(q,k) for all IBZ q.

Do we need to generalize the electron velocity beyond what we are doing now?

Need to test on intel compiler, especially using coarrays.

Calculation of the ibz is time consuming for dense phonon wave vector meshes. The computed results for a given mesh/Fermi window thickness should be reused for subsequent temperature and concentration sweeps.

Currently the code runs on spglib 1.6.0. Need to test using the newer versions of this library. Alternatively, we can replace it with something else.

Need symmetrizers for the electron wave vector dependent vector quantities.

Calculate the phonon-electron transition probabilities for all IBZ q.

Implement the necessary procedures for ph-iso scattering.

Implement parallel e BTE.

Need to automate the calculation of the transport of the n- and p-type carriers. How to handle systems with overlapping electron and hole pockets like bismuth?

For the coupled BTEs, is there a better iteration scheme than advancing each BTE one iteration at a time, sequentially?

Need a standalone tool that generates the chemical potentials for n- and p-doped large band gap materials for a give range of temperatures and carrier concentrations. This will then be used for temperature and carrier concentration sweeps.

Calculate ph-e RTA scattering rates and include ph-e term in the ph BTE.

This flag is useful when we want to sweep over chemical potentials for a given temperature. (W depends on temperature, but not on chemical potential.)

Add functionality to plot along high symmetry directions phonons, electrons, and the e-ph matrix elements.

Calculate the e-ph transition probabilities, X+ and X-, for all IBZ k.

Should implement transport coefficient accumulation with respect to mean free path. These are often useful for understanding the physics.

• phonon thermal conductivity
• charge conductivity
• electron thermal conductivity
• electron and phonon thermopowers

Need interpolation routine to do W(q-mesh) -> W(k-mesh), where, in general, the k mesh is finer than/equal to the q mesh. This is needed in the drag enabled eBTE iterator only and should be used on-the-fly.

Implement an easily manageable method for disk io.

Certain Fourier transforms can, in principle, be accelerated on gpus. However, large amounts of data might have to be moved to the gpus, which might be very slow. Regardless, worth trying this out.

Implement drag terms in the ph and e BTEs.

Use collective intrinsics from the Fortran 2018 standard.

Would be nice to have a fast way to do this. Because life is short.

Generate the Ford documentation and call maps.

Set up internal looper for T and/or chemical potential sweeps.

Add (charged) e-imp scattering.

Calculate g(k,q) for all IBZ k. Here, q belongs to the k-mesh.

Generalize code to be able to handle 2d systems.

Make output pretty and greppable. While at it, print a nice banner also, because why the hell not.

We need to bake in the imposition of the KO relations with the iteration process.

Calculate coarse k, q mesh, wannierization, and wannierized quantities by directly interfacing with QE and Wannier90.

Quadrupole corrections might be very important for certain polar and non-polar insulators and semi-conductors. To include this we need to modify both EPW and elphbolt.

Can the BTE iterators be sped up further?

Allow user to choose either IFC or Wannier phonons.

Check where we can safely release memory.

I don't think this is well defined without providing the band edge. Apart for a gapped system, I don't see how this can be easily calculated internally.

It would be useful to time the various expensive parts of the program.

The energy conserving delta functions can be used to eliminate large portions of V-

The fact that the number of the transport active bands is smaller than the potentially large number of Wannier bands can be used to lower the memory (RAM and disk) requirement of the electronic quantities.

Need to optimize the bottleneck - V- calculation.

Print out quantities such as response functions, transport coefficient tensors, spectral quantities, etc. at the RTA, partial decoupling, and fully converged levels.