Recommended by

Indexed by

Structure database of glass-ceramic lithium thiophosphate electrolytes

Haoyue Guo1*, Nongnuch Artrith1,2,3*

1 Department of Chemical Engineering, Columbia University, New York, NY 10027, USA

2 Columbia Center for Computational Electrochemistry, Columbia University, New York, NY 10027, USA

3 Materials Chemistry and Catalysis, Debye Institute for Nanomaterials Science, Utrecht University, 3584 CG Utrecht, The Netherlands

* Corresponding authors emails: haoyue1619@gmail.com, n.artrith@uu.nl
DOI10.24435/materialscloud:j5-tz [version v1]

Publication date: Feb 01, 2022

How to cite this record

Haoyue Guo, Nongnuch Artrith, Structure database of glass-ceramic lithium thiophosphate electrolytes, Materials Cloud Archive 2022.17 (2022), doi: 10.24435/materialscloud:j5-tz.


This database contains computationally generated atomic structures of glass-ceramics lithium thiophosphates (gc-LPS) with the general composition (Li2S)x(P2S5)1-x in the XCrySDen structure format (XSF). Total energies and interatomic forces from density-functional theory (DFT) calculations are included as additional meta information. The extended XSF format is compatible with the atomic energy network (ænet) package for artificial neural network (ANN) potential construction and application. The DFT calculations used projector-augmented-wave (PAW) pseudopotentials and the Perdew−Burke−Ernzerhof (PBE) exchange-correlation functional as implemented in the Vienna Ab Initio Simulation Package (VASP) and a kinetic energy cutoff of 520 eV. The first Brillouin zone was sampled using VASP’s fully automatic k-point scheme with a length parameter Rk = 25Å. The gc-LPS structures were generated using a combination of different sampling methods. Initial amorphous structure models were generated with ab initio molecular dynamics (AIMD) simulations of supercells at 1200 K using a Nose-Hoover thermostat with a time step of 1 fs. To obtain near-ground-state structures as reference for the machine-learning potential, 150 evenly spaced snapshots were extracted from the AIMD trajectories that were reoptimized with DFT geometry optimizations at zero Kelvin. Additional structures were generated by scaling the lattice parameters of the crystalline LPS structures (see below) by ±15% and perturbing atomic positions in AIMD simulations as described above. The resulting database was used to train a specialized ANN potential for the sampling of structures along the Li2S–P2S5 composition line with a genetic-algorithm (GA) as implemented in the atomistic evolution (ævo) package, following a previously reported protocol. Starting from supercells of the ideal crystal structures, either Li and S atoms were removed with a ratio of 2:1, or P and S atoms were removed with a ratio of 2:5, and low-energy configurations were determined with GA sampling. A population size of 32 trials and a mutation rate of 10% were employed. The ANN potential was iteratively refined by including additional sampled structures in the training. For each composition, at least 10 lowest energy structure models identified with the ANN-GA approach were selected and fully relaxed with DFT. Also included in the present database are the XSF files of the previously reported crystalline phases LiPS3, Li2PS3, Li4P2S7, Li7P3S11, α-Li3PS4, β-Li3PS4, γ-Li3PS4, and Li48P16S61. The crystal structures were obtained from the Inorganic Crystal Structure Database (ICSD), the Materials Project (MP) database, the Open Quantum Materials Database (OQMD), and the AFLOW database. The directory names indicate the journal reference and the database. Further details can be found in the associated research article.

Materials Cloud sections using this data

No Explore or Discover sections associated with this archive record.


File name Size Description
173.2 KiB README.txt - Description of the data set and the contents of all other files
16.1 KiB Structures and DFT energies of the crystalline LPS phases in an extended XSF format
6.5 MiB Structures, DFT energies, and atomic forces of glass-ceramic LPS in an extended XSF format used for ANN potential training with aenet


Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Preprint (Preprint of the article that makes use of this data set.)


VASP DFT ænet lithium thiophosphates solid electrolytes

Version history:

2022.17 (version v1) [This version] Feb 01, 2022 DOI10.24435/materialscloud:j5-tz