volcorr with your own DFT results
Predicts how far a PBE / PBE+U relaxation done the Materials Project way will miss the experimental volume of a crystal. Enter a formula, optionally the space group, or give the structure. If you have run the relaxation yourself, you can also give its relaxed structure, band gap and magnetization (section below the volume field). Everything runs in your browser; nothing you enter is sent anywhere. The version without this option is at utksi/volcorr.
Results of your own PBE(+U) relaxation (optional, most accurate)
For a relaxation done with Materials Project settings. Give the relaxed structure, its band gap and its total magnetization; this replaces the volume field above.
The predicted volume error appears here. Positive means the calculation overestimates the experimental (room-temperature) volume.
What the prediction refers to
The model learned from relaxations in the Materials Project database, so it predicts the error of a calculation done the same way: PBE with PAW potentials, 520 eV cutoff, spin polarization, and a force-based ionic relaxation. Oxides and fluorides of Co, Cr, Fe, Mn, Mo, Ni, V and W get PBE+U with the Materials Project U values (Co 3.32, Cr 3.7, Fe 5.3, Mn 3.9, Mo 4.38, Ni 6.2, V 3.25, W 6.2 eV); you can override this choice above. It does not cover other functionals (for example r²SCAN) or other U values.
Only compounds whose PBE(+U) volume lies within 10 % of experiment were used for training, so materials where PBE fails badly (strongly van der Waals bound or some f-electron systems) are outside its range. Experimental volumes are at room temperature, so the prediction includes thermal expansion.
Method and accuracy
Gradient-boosted trees (LightGBM, absolute-error loss) predict y = ln(Vexp/VDFT) from composition, shell-filling and symmetry descriptors and the calculation settings. They were trained on 40,629 Materials Project entries matched to ICSD structures; five models are averaged.
| Input | mean absolute error on held-out compounds |
|---|---|
| no correction (plain PBE(+U)) | 3.79 % |
| formula | 2.07 % |
| formula + space group | 2.06 % |
| experimental structure | 1.82 % |
| + your PBE(+U) volume | 2.05 % |
| your finished relaxation (structure, band gap, magnetization) | 1.98 % |
| + formation energy and energy above hull | 1.98 % |
The full correction in the accompanying paper reaches 1.59 % for these calculations, but it also uses outputs of the finished DFT calculation (relaxed geometry, band gap, magnetization), which are not asked for here. A MACE-MPA-0 route needs PyTorch and is available in the command-line tool. Models: utksi/volcorr-models.