Support for toml input files and updated hubbard example notebook

[adsrichards]

Rick and I have discussed the possibility of supporting input files in toml format. I have just created a basic script implementing toml input support called toml_to_moha.py within a new directory called scripts, and an example toml input file called hubbard.toml in the examples/toml directory to demonstrate its usage.

The script is called by using
py toml_to_moha.py /path/to/input_file.toml
The input_file.toml is then read and a model hamiltonian object is created according to the defined options within input_file.toml.

For example, using
py toml_to_moha.py ../examples/toml/hubbard.toml
from within the scripts directory calculates a HamHub object in the same way as in the hubbard example ipynb notebook.

In this example, the control parameters defined in hubbard.toml are

[control]
prefix = "hubbard"
outdir = "./fcidump"
save_intergrals = true
integral_format = "fcidump"

so that the script creates a new directory specified by outdir = "./fcidump", and the integrals are saved to fcidump format as "hubbard.out" within the outdir.

Additionally, I have cleaned up the hubbard example notebook which previously had some incorrectly calculated FCI energies.

[RichRick1]

That looks great to me! If I understand it correctly, it can support right now only Hubbard model hamiltonian? Can we add support for Huckel, PPP, and Heisenberg models?

[adsrichards]

I have just included support for all of the implemented hamiltonians in the toml_to_ham.py file including Huckel, PPP, Hubbard, Heisenberg, Ising, and RG. I have also made an example toml file for each in examples/toml. This should be up to date and working with everything now.

[RichRick1]

Great job!

Overall there are few minor changes that needs to be done:

1. There is a typo in the toml file to save integrals
2. We need to support saving .npz file through the toml file
3. When generating output file it's a good idea to use os.path.join() function, so it will take care of format of the path, that may vary for different os

[RichRick1]

we need to clearly document that input of PPP hamiltonian supports only input of the diagonal term $\gamma_{pq} = \gamma_0\cdot \delta_{pq}$

[adsrichards]

I have added the line:
"Only diagonal form of gamma matrix for PPP model is supported:
gamma_pq = gamma0 * delta_pq."

to the description of build_moha_moltype_1d to address this.

[RichRick1]

I've just realized, that 2 body term in our definition is

$$\frac{1}{2} \sum_{p \neq q} \gamma_{p q}\left(\hat{n}{p \alpha}+\hat{n}{p \beta}-Q_p\right)\left(\hat{n}{q \alpha}+\hat{n}{q \beta}-Q_q\right)$$

so that p!= q. It seems like $\gamma_0$ is inside $U_p$
I don't think we even need this parameter to be specified.

[RichRick1]

It might be better to use os.path.join() function to generate the out_file. Join symbol / can vary between different os: windows use "", while MacOS and Linux use "/"

[adsrichards]

I have replaced this line with:
out_file = join(data["control"]["outdir"], data["control"]["prefix"])
and added 'from os.path import join' to the imports

[RichRick1]

I recently added saving hamiltonian as .npz output file. It would be great if we can support it too

[adsrichards]

I have made two changes to the save output. As you suggested, I include the option to specify save as npz as an option in the toml input, which calls ham.savez(out_file). I have also changed the file extension for fcidump files to be .fcidump, so that not every output file has a .out file extension.

[adsrichards]

Thank you for noticing the typos and comment formatting. Those issues should be fixed now. I have also added the option to support saving npz, and I use os.path.join() instead of + "/" +.

[RichRick1]

Also, do you think we can/should add default symmetry here? Make it 4 as default

[adsrichards]

symmetry has been included in the defaults.toml file and is set to 4

[RichRick1]

I've just realized, that 2 body term in our definition is

$$\frac{1}{2} \sum_{p \neq q} \gamma_{p q}\left(\hat{n}{p \alpha}+\hat{n}{p \beta}-Q_p\right)\left(\hat{n}{q \alpha}+\hat{n}{q \beta}-Q_q\right)$$

so that p!= q. It seems like $\gamma_0$ is inside $U_p$
I don't think we even need this parameter to be specified.

[RichRick1]

Add defaults for the Huckel

[adsrichards]

defaults specific to Huckel are now set in the set_defaults function

[adsrichards]

The parameters for the PPP model should be set correctly now, two-electron integral symmetry is set to 4 by default and included in defaults.toml, and parameters specific to the Huckel model are set in the set_defaults function.

[adsrichards]

I have changed the implementation slightly. The toml_to_ham is changed from taking a string for the name of toml_file to a dict containing model data. This will be much more useful because we can generate the dict using other methods (e.g. ChatGPT), and the dict used in toml_to_ham.py is just loaded in one line. All of the methods in toml_to_ham.py can now be imported and used in the upcoming chatgpt_to_moha.py script.

[RichRick1]

Looks great to me!