Machine learning the screening factor in the soft bond valence approach for rapid crystal structure estimation
Bibliographic record
Abstract
Development of new functional ceramics is important for several applications, including electrochemical batteries and fuel cells. Computational prescreening and selection of such materials can help discover novel materials but is challenging due to the high cost of electronic structure calculations which would be needed to compute the structures and properties of interest such as the material's stability and ion diffusion properties. The soft bond valence (SoftBV) approach is attractive for rapid prescreening among multiple compositions and structures, but the simplicity of the approximation can make the results inaccurate. We explore the possibility of enhancing the accuracy of the SoftBV approach when estimating crystal structures by adapting the parameters of the approximation to the chemical composition. Specifically, on the examples of perovskite- and spinel-type oxides that have been proposed as promising solid-state ionic conductors, the screening factor, an independent parameter of the SoftBV approximation, is modeled using linear and non-linear methods as a function of descriptors of chemical composition. We find that making the screening factor a function of composition can noticeably improve the ability of SoftBV to correctly model structures, in particular new, putative crystal structures whose structural parameters are yet unknown. We also analyze the relative importance of nonlinearity and coupling in improving the model and find that while the quality of the model is improved by including nonlinearity, coupling is relatively unimportant. While using a neural network showed no improvement over linear regression, the recently proposed GPR-NN method that is a hybrid between a single hidden layer neural network and kernel regression showed substantial improvement, enabling the prediction of structural parameters of new ceramics with accuracy on the order of 1%.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".