Data for Mn HEO analysis Vinca

Published: 12 October 2026| Version 1 | DOI: 10.17632/6mvff3hrpt.1
Contributors:
Tanja Asanovic Antonic,

Description

This dataset contains the data behind a study of phase stability and oxygen-vacancy physics in high-entropy oxides (MgCoNiCuZn)O and their variants that contain Mn. The single-phase stability and the oxygen-vacancy behaviour of these oxides are controlled by individual cations, and the machine-learned interatomic potential is helping to solve these effects. All disordered structures were generated as special quasirandom structures (SQS) with icet, in a 480-atom supercell (240 cations, 240 O), and were relaxed with the UMA-S-1.p.1 potential. The data is organized like this: 01 SQS structures and their composition 02 SQS convergence tests: norms versus Monte Carlo steps and random seeds. 03 Relaxed structures, relaxation logs, and final energies. 04 Reference binary oxides: relaxed wurtzite ZnO and tenorite CuO, and the energies of all six references (rock salt MgO, CoO, NiO, MnO; wurtzite ZnO; tenorite CuO). 05 Mixing enthalpy and bond-length spread, where the fraction of Mn was changed; then for the series with Mn fixed at 1/6, one cation varied, while the other four are equimolar. Four-cation compositions with known experimental phase behavior are used for calculation of the thresholds. 06 Oxygen vacancy formation energies for all 240 oxygen sites in the five-metal and six-metal oxides. The first-shell cation composition of each site was considered, also relaxed defect structures. 07 Nudged elastic band calculations of 72 vacancy migration paths at six selected sites, with forward and reverse barriers and reaction energies. The conclusion is that Mg, Co, and Ni lower both the mixing enthalpy and the bond-length spread and stabilise the system. Cu raises both, with the biggest destabilising effect. Zn raises the mixing enthalpy without increasing distortion. Then: if copper is in the first-neighbour shell, it will lower the vacancy formation energy. Mg raises it. Mn, too, with a small effect. Sites that have low formation energy have high escape barriers. Cu increases the number of vacancies near it, but reduces their mobility. The data can be used to produce figures correlated to this project.

Files

Steps to reproduce

The calculations were run with Python 3.10.12, on an NVIDIA GPU with CUDA 12.2. All relaxations, vacancy and NEB calculations use the UMA-S-1p1 model, omat, through fairchem-core 2.12.0, without fine-tuning. 01sqs_generation: Special quasirandom structures were generated with icet on a conventional rocksalt cell, where the lattice parameter is 4.20 Å, pair and triplet cutoffs with values of 6.0 Å and 4.0 Å and with 20000 Monte Carlo steps. All structures use a 3x4x5 supercell which has 480 atoms, 240 O and 240 cations. sqs_manifest.csv lists each structure, composition and number of atoms. 02sqs_convergence_tests: The norms folder contains the L1, L2 and Linf distances between the cluster vector of the generated structure and the random target (as a function of the number of the Monte Carlo steps for equimolar oxide with six metals, with number of atoms 480). Folder seeds contains the same norms for 25 random seeds at 2000, 20000 and 60000 steps. 03mlip_relaxation: Each structure from step1 was relaxed with UMA, with cell parameters and atomic positions being optimized. There are two steps: FIRE and BFGS - they are running until the largest force fall below 0.02 eV/Å. Relaxed CIFs and logs are in cifs_and_logs. relaxation_results.csv gives the final energies and bond-length statistics. 04reference_binary_oxides: MgO, CoO, NiO and MnO in rocksalt, ZnO in wurtzite and CuO in tenorite were relaxed with the same way as in step 3. binary_references.csv gives their energies per atom. 05mixing_enthalpy: The mixing enthalpy was calculated as ΔHmix = E(HEO) - sum_i x_i E(M_iO), per atom, using the references explained in step4. The σ_bonds is the length spread, the standard deviation of all metal-oxygen bond lengths. mixing_energies_groundstate.csv has the (MgCoNiCuZn)_(1-x)/5 Mn_x O series. On the other side, hmix_results_480.csv contains the series with manganese fixed at 1/6 and one cation M varied (other four are equimolar). calib_manifest.csv and calib_mixing.csv contain the four-cation compositions with known experimental outcome, used to calibrate the single-phase thresholds. 06oxygen_vacancies: In the relaxed five- and six-metal structures, each of the 240 oxygen atoms was removed one after another, and the cell was relaxed again, as in step 3. The formation energy was calculated as Evf = E_defect + 1/2 E_O2 - E_perfect. Each CSV gives, for every oxygen site, its atom index (o_index) in the relaxed CIF, its Evf and the number of each cation in the set of its six nearest neighbours. 07NEB_migration_barrier: Six vacancy sites were selected, two in the five- and four in the six-metal oxide. Every oxygen has twelve nearest oxygens. All twelve hop directions at each site were computed with the nudged elastic band method, giving 72 paths. The bands folder contains one subfolder per site. neb_full_table.csv gives, for each path, the forward and reverse barriers, the reaction energy ΔE, the two gate cations and the initial and final shells.

Institutions

Categories

Materials Science, Computational Materials Science

Licence