nLMF training data for LDH

These datasets were generated using a modified version of Turbomole, employing the def2-QZVPPD basis set, RI-DFT approximation, and the ωDH25 functional with gridsize 3. They were utilized for training neural-network local mixing function (LMF) models, specifically in the development of the first double local hybrid functionals.

Each data entry has the following structure:

ra, gax, gay, gaz, la, ta, haxx, haxy, haxz, hayy, hayz, hazz, rb, gbx, gby, gbz, lb, tb, hbxx, hbxy, hbxz, hbyy, hbyz, hbzz, exx_sra+exx_lra, exx_srb+exx_lrb, exx_sra, exx_srb, XXX, XXX, XXX, XXX

Each feature is represented in scientific notation (E24.17).

XXX columns represent part of trained functional and were not used in training of n-LMF.

Data were used as input for training with: https://github.com/awodynski/nLMFs/