Table 2 Comparison of DFT calculated datasets that can be used to train the neural network potential.
Dataset | Systems | Structures | # of | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Molecule | Bulk | Cluster | Slab | Adsorption | Disorder | Opt. | Vib. | MD | TS | Elements | Data | |
Materials Project18 | ✓ | ✓ | ✓ | Unlimited | > 1 × 105 | |||||||
OQMD53 | ✓ | ✓ | Unlimited | 8 × 105 | ||||||||
NOMAD54 | ✓ | ✓ | Unlimited | > 5 × 107 | ||||||||
Jarvis-DFT55 | ✓ | ✓ | Unlimited | > 4 × 105 | ||||||||
AFLOW56 | ✓ | ✓ | ✓ | Unlimited | > 3 × 106(*1) | |||||||
✓ | ✓ | 5 | 1 × 105 | |||||||||
PubChemQC57 | ✓ | ✓ | 30 | > 3 × 106(*2) | ||||||||
MD1758 | ✓ | ✓ | 4 | 9 × 106 | ||||||||
SN2 reactions59 | ✓ | ✓ | ✓ | ✓ | 6 | 4 × 105 | ||||||
ANI-114 | ✓ | ✓ | ✓ | ✓ | 5 | 2 × 107 | ||||||
ANI-2x15 | ✓ | ✓ | ✓ | ✓ | 7 | 9 × 106 | ||||||
COMP6v215 | ✓ | ✓ | ✓ | ✓ | 7 | 2 × 105 | ||||||
tensor-mol 0.1 water51 | ✓ | ✓ | 2 | 4 × 105 | ||||||||
tensor-mol 0.1 spider51 | ✓ | ✓ | 4 | 3 × 106 | ||||||||
TeaNet60 | ✓ | ✓ | ✓ | 18 | 3 × 105 | |||||||
✓ | ✓ | ✓ | ✓ | 56(*3) | 1 × 108 | |||||||
PFP molecular dataset (ours) | ✓ | ✓ | ✓ | ✓ | 9 | 6 × 106 | ||||||
PFP crystal dataset (ours) | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ | 45 | 3 × 106 | |
