Abstract

The materials-analayzer.info service is designed for the rapid assessment of the electrical conductivity of materials with the composition AnBnOn (where O - oxygen, A and B - metals, and n - any integer), thereby determining the applicability of a potential conductive material in electrochemical devices, such as: solid oxide fuel cells and all-solid-state metal-ion batteries. The assessment of electrical conductivity is based on predicting the band gap using developed machine learning models. The service also offers an OxyHopper tool for analyzing the oxygen ion migration pathways in crystal structures.

1. Description

The band gap (Eg) is a fundamental property that is directly related to a material's potential use in optical, electronic, and energy applications. By controlling a material's composition or structure, its Eg can be tuned, enabling the creation of materials suited for specific applications. Selecting such materials and evaluating their conductive properties is a complex challenge, which has recently been addressed through computational modeling. However, these methods are unsuitable for the large-scale screening of compound databases due to their high computational cost. An optimal solution to this problem is the use of machine learning methods, which can predict physical properties in cases where the target property cannot be determined directly without significant experimental or computational resources.

The developed service predicts Eg values with high accuracy, thereby determining the suitability of a potential conductive material for the aforementioned electrochemical devices. For the rapid assessment of Eg, a machine learning model was used, trained on data from The Materials Project, AFLOW, and OQMD databases for structures with the composition AnBnOn (where O - oxygen, A and B - metals, and n - any integer). The training utilized over 12000 compounds, with a test set comprising more than 1300 compounds.

2. Tools

2.1 Band Gap Estimator

The prediction of Eg is based on computable features and reference data. The calculation of geometric features for crystal structures utilizes Voronoi partitioning. The ionicity degrees of chemical bonds are determined based on the electronegativity values of the elements. Using data on the number of outer-shell electrons, atomic masses of the elements, and the unit cell volume of the crystal, the electron density values for the structures are calculated.

2.2 OxyHopper

In this section, the search for possible interstitial sites for oxygen during its diffusion is performed. This search is also based on partitioning the crystal space into Voronoi polyhedra.

3. Input files

The service can only process a CIF file as input, which must contain information about the crystal structure, including the number of formula units (_cell_formula_units_Z). The Drag-and-Drop technology implemented in the service allows users to easily upload files to the site by dragging and dropping them into a dedicated form.

4. Output files

The Band Gap Estimator tool generates a report in *.pdf format. This report presents the values of descriptors calculated from the user-uploaded structure and the predicted band gap value.

OxyHopper generates a report in *.pdf format, which provides the concentration of interstitial oxygen atoms and the ionicity degrees of the oxygen bond with each metal, as well as a *.cif file. This CIF file contains the original crystal structure information along with additional atoms marked as H, representing the interstitial sites.

5. Errors, limitations, and troubleshooting recommendations

• Unsupported file format. The file containing crystal structure data must be in *.cif format with the number of formula units specified.
• Research Objects. The study focuses on ionic structures containing only "cation-anion" bonds.
• User perceives the Eg estimation as incorrect. Please verify that your structure is not in the primitive P1 symmetry, as symmetry was one of the significant descriptors used in training the model. To increase symmetry, the user can utilize the AFLOW service or Python libraries such as pymatgen or ASE.
• The structure is disordered. In this case, the service may generate an error and fail to process the compound. It is necessary to create a pseudo-ordered supercell that corresponds to the given composition. This can be done using, for example, the SUPERCELL program.

6. Copyright

Certificate of State Registration of the Computer Program No. 2025680575 dated August 7, 2025. "Materials Analyzer" - assessment of ionic and electronic conductivity in solid-state materials.

Rightsholder: LIMITED LIABILITY COMPANY "ANALIZ MATERIALOV" (RU).

Authors: Morkhova Yelizaveta Aleksandrovna (RU), Smolkov Mikhail Igorevich (RU), Kabanov Artem Anatolyevich (RU), Osipov Vladislav Timofeevich (RU), Blatov Vladislav Anatolyevich (RU).