SoLiD26: A first principles solid-liquid interface dataset for machine-learned interatomic potentials

SoLiD26 is a curated solid–liquid interface dataset, containing approximately 15.4 million first-principles atomic structures with up to 576 atoms and 15 chemical elements for training and evaluating machine-learned interatomic potentials (MLIPs). The structures were compiled from density functional theory (DFT) calculations performed in studies of solid–liquid interfaces, with most configurations originating from ab initio molecular dynamics (AIMD) simulations. Each record contains atomic species, positions, simulation cell, periodic boundary condition flags, total energy and atomic forces. The dataset includes aqueous metal interfaces, electrode–electrolyte systems and selected bulk reference structures, calculated with VASP using the PBE functional and D3 dispersion corrections.

The dataset is provided in ASE format as a collection of sequentially numbered .aselmdb files. The data files are compressed with xz and can be decompressed with $ xz -d *.xz. The train_val_test_split_indices.json file contains the indices of the splits used for technical validation of the dataset.

The .aselmdb files can be read using Python by installing the ase and ase-db-backends packages. With both packages installed, the standard ase.db.connect() method will recognize the .aselmdb file format and the data can be used as described in the ASE database documentation. Some atomistic machine learning packages already support the .aselmdb file format and have dataset classes that can be used to load the dataset, such as fairchem-core and mace-torch. The code used to prepare the dataset is available on GitHub.

Users of SoLiD26 should be aware that the dataset contains structures (ase.Atoms objects) that have specified constraints, such as fixed atoms, which were defined in the original calculations. This means that using the standard atoms.get_forces()method in ASE may inadvertently set some forces to zero. Instead use atoms.get_forces(apply_constraint=False), or get the forces directly from the underlying data structure, to access the raw, unmodified forces, as using the altered forces in a machine learning context can lead to erroneous results.

MD5 checksums of the compressed files:

$ md5sum ./*.aselmdb.xz
5181c7563aa6f57248f863cf9b8db485  ./data_part_01_of_16.aselmdb.xz
b4737f94f0bdfb92b03c7528dd8e73a7  ./data_part_02_of_16.aselmdb.xz
b71ca983d5d6d663f24463348d0e35e7  ./data_part_03_of_16.aselmdb.xz
f3616f564bcdc558bf1b14164356f763  ./data_part_04_of_16.aselmdb.xz
78ef2a3b30d9853656be6a888151129f  ./data_part_05_of_16.aselmdb.xz
be6b874079b42bc1d4277286a3b33bb6  ./data_part_06_of_16.aselmdb.xz
f738a28f2b6e13a2938d8e4c0525320f  ./data_part_07_of_16.aselmdb.xz
fe7bb69fe16b4e7192e1394ed7783dcf  ./data_part_08_of_16.aselmdb.xz
17e9201985ad592e8364d681e0590c1f  ./data_part_09_of_16.aselmdb.xz
be2696a072fd48c4fc9bf3caffb53cab  ./data_part_10_of_16.aselmdb.xz
0444eeef95b36b177bf8241066fcd00f  ./data_part_11_of_16.aselmdb.xz
1d9c4561c6eddc5823dcc4e14a61f30d  ./data_part_12_of_16.aselmdb.xz
84f426ad94ccacdbc49a2cf3a6547742  ./data_part_13_of_16.aselmdb.xz
aa091f0ccf5a11600dc67abc69fc247b  ./data_part_14_of_16.aselmdb.xz
772a87be9a96cd7acc51c00cd69fba55  ./data_part_15_of_16.aselmdb.xz
75f683dd6dde4134f58640a99c245a9b  ./data_part_16_of_16.aselmdb.xz

MD5 checksums of the decompressed files:

$ md5sum ./*.aselmdb
d56a851954823755ecd16e656dfbea9a  ./data_part_01_of_16.aselmdb
0649760519b57e81af49efd8f3cc6935  ./data_part_02_of_16.aselmdb
22a363d4123bd5b2679122d92b4314d0  ./data_part_03_of_16.aselmdb
682d3f926b4c402d3c3ea9296cb36951  ./data_part_04_of_16.aselmdb
60b33e9f29da8dabb6ff126336362f7f  ./data_part_05_of_16.aselmdb
f9c3a7c0fa71377f0853c029a2773afa  ./data_part_06_of_16.aselmdb
2271c3d6db013472572a7851a7fd62da  ./data_part_07_of_16.aselmdb
45c92d8708543d0e58142f526b88edcf  ./data_part_08_of_16.aselmdb
4eaccac9025bb09d6c2fba994b532d9e  ./data_part_09_of_16.aselmdb
d0f446a599a549aacf4250ffbbb17ba1  ./data_part_10_of_16.aselmdb
b48312b2728382420e254a13792f8bac  ./data_part_11_of_16.aselmdb
93fd55ddca00099155c82f96d2e5cde6  ./data_part_12_of_16.aselmdb
55499806a452ba3303e348b9f241925a  ./data_part_13_of_16.aselmdb
8b6ce3d4659e3f16d63259b9d36a2b44  ./data_part_14_of_16.aselmdb
07682efb6aed3d19c2b32f497a12184d  ./data_part_15_of_16.aselmdb
f9160daf8f0cb8a38169ad1c608a9f1e  ./data_part_16_of_16.aselmdb