Indexed by
Publication date: Nov 06, 2017
Interatomic potentials are often necessary to describe complex realistic systems that would be too costly to study from first-principles. Commonly, interatomic potentials are designed using functional forms driven by physical intuition and fitted to experimental or computational data. The moderate flexibility of these functional forms limits their ability to be systematically improved by increasing the fitting datasets; on the other hand, their qualitative description of the essential physical interactions ensures a modicum degree of transferability. Recently, a novel trend has emerged where potential-energy surfaces are represented by neural networks fitted on large numbers of first-principles calculations, thus maximizing flexibility but requiring extensive datasets to ensure transferability. Gaussian Approximation Potentials in particular are a novel class of potentials based on non-linear, non-parametric Gaussian-process regression. Here we generate a Gaussian Approximation model for the α-phase of iron training on energies, stresses and forces taken from first-principles molecular dynamics simulations of pristine and defected bulk systems, of surfaces and γ-surfaces with different crystallographic orientations.
No Explore or Discover sections associated with this archive record.
| File name | Size | Description |
|---|---|---|
|
DB_bccFe_Dragoni.tar.gz
MD5md5:64e7a4f996e56f90e80ef9b0e10f1a72
|
5.4 MiB | Training database for α-iron: the database contains atomic positions and cell vectors in Cartesian coordinates along with the corresponding energies (plus forces or stresses when available) obtained from first-principles calculations. Data are reported in a XYZ format. The database can be partitioned into 8 sub-databases covering different local atomic environments as discussed in the README.txt included in the download file. Approximately 100,000 environments are included. All ab-initio calculations are performed with the Quantum Espresso package using an ultrasoft pseudopotential from the PSlibrary which best reproduces all-electron equilibrium properties of the crystal. The choice of k-points sampling and cutoffs is crucial in order to generate a homogeneous and accurate database. These are selected as to ensure, across the entire database, a convergence of energy differences (per atom), forces and stresses within 1 meV, 0.01 eV/Ang and 0.1 GPa respectively. * DB1 - distorted primitive cells * DB2 - bulk vibrations * DB3 - monovacancy * DB4 - double vacancies * DB5 - tri-vacancies and "small" cluster-vacancies * DB6 - Self-interstitials * DB7 - Bulk-terminated surfaces (100, 110, 111, 211) * DB8 - gamma surfaces (110, 211) |
|
gp2016.tar.gz
MD5md5:c9ed568ed9795ee7ab3482d2ccaa962e
|
59.0 MiB | GAP potential for α-iron trained on the first-principles database described above. The tar.gz consists of two files, gp33b.xml and gp33b.xml.sparseX.GAP_2016_10_3_60_19_29_10_8911. The .xml file contains all the technical details of the potential, including the configurations actually used for training. The "sparse" file contains the basis set which is used to construct the potential. Both files are necessary for running calculations with the QUIP+GAP code (http://www.github.com/libAtoms/QUIP) and the LAMMPS package. |
No external references available for this Materials Cloud Archive record.
| 2017.0006/v2 (version v2) [This version] | Nov 06, 2017 | DOI10.24435/materialscloud:2017.0006/v2 |
| 2017.0006/v1 (version v1) | Jun 21, 2017 | DOI10.24435/materialscloud:2017.0006/v1 |