Planned Features

[marvinfriede]

A list of planned features and (long term) goals for dxtb.

• GFN2-xTB
• higher-order derivatives with functorch and implicit differentiation (replace xitorch with parts of torchopt)
• improved solvation
• Python 3.12 support (Support Python 3.12 #172)
• macOS support (Make libcint interface optional #166)
• Windows support (Make libcint interface optional #166)
• spin polarization (spGFN)

[yuanqidu]

Thanks for proposing the features! Great work! I am wondering when GFN2-xTB is planned to be supported.

[marvinfriede]

Thanks for proposing the features! Great work! I am wondering when GFN2-xTB is planned to be supported.

Thanks! We will tackle GFN2 probably around the beginning of next year, maybe we can draft a rough version a bit earlier. But currently, we are focusing on applying dxtb.

[yuanqidu]

Thank you so much! Looking forward to using GFN2-xTB!

[rdguerrerom]

Hi,

I would like to inquire whether there are any plans to include support for calculating excited state energies, gradients, and non-adiabatic couplings using an efficient method, such as hole–hole Tamm–Dancoff approximated density functional theory (hh-TDA-DFT).

Thank you for your time and consideration.

[marvinfriede]

Hi,

I would like to inquire whether there are any plans to include support for calculating excited state energies, gradients, and non-adiabatic couplings using an efficient method, such as hole–hole Tamm–Dancoff approximated density functional theory (hh-TDA-DFT).

Thank you for your time and consideration.

Hi, unfortunately, we are not planning on implementing excited state methods. I'm sorry.