Add more open-source DFT Kernels for various pore types/materials/probes

[celalutku]

I know this code mainly a framework for processing adsorption data and isotherm fitting but, it is sad that it contains only 1 example kernel for dft calculations, although it is very well written code.

[pauliacomi]

Hi,

I'm aware that this is a shortcoming of the program in the current state. The kernel provided was calculated by a collaborator at my former institution (MADIREL) and therefore I managed to obtain permission for its use.

Unfortunately, calculation of DFT (or NLDFT/QSDFT/GCMC) kernels is not easy and requires quite a bit of computation. There are other kernels available from established companies such as Micromeritics although I believe they are proprietary and limited to use with their software.

I am going to leave this issue open for now, as it is a good avenue for making the program better if someone has the capability of calculating or donating non-proprietary kernels.

Best, Paul

[sblanky]

I make no promises, however once I complete my PhD I would like to look into facile bespoke DFT kernel development. If I'm able to figure this out the software and any generated kernels will of course be FOSS.

[a11enh]

Hi everyone!

I definitely want to second this here for everyone, even though it may seem obvious. I dug into this approach a bit when seeking out a kernel for slit/cylinder pores in carbon materials, and unfortunately it seems like the kernels mentioned in literature are often not disclosed (other by name/type) and become proprietary kernels.

For interested researchers- do we have an idea of what a kernel needs to look like? Are all kernels just like the example kernel - i.e. a table of pressures and pore-sizes? I'm attempting to import a kernel I found through supplementary materials on an article I've read, and I may have successfully imported it.. but haven't tested it out yet, so will have to come back and update any success or failure.

Big cheers for PyGaps! Thanks so much! If I successfully get a public kernel imported, I'll try and share as well! -A

[sblanky]

My understanding is that, yes the kernel should consist of pressures and pore sizes as columns and indeces. The data in each cell is a density I believe.

I look forward to seeing a new kernel! It would be interesting to publish a comparison of this open source kernel to proprietary alternatives.

Thanks, Scott