Developing a machine learning interatomic potential for advancedceramics
Bibliographic record
Abstract
Advanced ceramics, such as boron carbide, exhibit high strength, abrasion resistance, chemical and thermal stability, and low density making them candidate material for extreme condition applications like body armour, wear-resistant components, and cutting tools.However, the microstructures that initiate these properties are often complex and not fully understood, since they have features at multiple length scales.A major challenge in understanding the relationship between microstructure and properties is the availability and deficiencies of current interatomic potentials used in molecular dynamics (MD) simulations.Additionally, the complex lattice structure of boron carbide consisting of a 12-atom icosahedral cage surrounding a 3-atom linear chain along the (111) rhombohedral axis is challenging to equilibrate due to the covalent bonds between the atoms.Moreover, the small difference between the atomic number of boron and carbon causes the precise composition and distribution of carbon atoms difficult to measure, and this chemical similarity induces substitutional disorder.Improving the accuracy of potentials will facilitate more efficient studies of the effects of different microstructures on material properties and accelerate the development of new ceramics with enhanced properties.Building on these considerations, this research employs machine-learning-based (ML) methods to develop more accurate potentials for advanced ceramics that can capture complex microstructural dynamics relevant to material failure.ML potentials can provide more precise results compared to the traditional potentials since they benefit from high-dimensional representation of underlying potential energy surfaces (PES).ML potentials are composed of three parts: 1. an initial dataset including the energy, force and stress of structures, 2. the descriptor vector defining the atomistic configurations, and 3. numerical regression that maps the initial dataset to the PES.In this study, Monte-Carlo simulation is employed to randomly create structures.Afterwards, the energy and force of the sampled structures are calculated by DFT.The local environments of the structures are then defined with the Atom Centered Symmetry Function.Lastly, the initial dataset, alongside the feature vectors, are used as the input of the regression.When completed, the results of MD simulations with the new potential(s) can be used in multi-scale modelling of the dynamic failure of ceramics, and to make novel contributions in the discovery of new mechanisms and designing new nanomaterials.This talk will discuss ongoing structure prediction and sampling, calculation of initial inputs for ML using DFT, and evaluation of feature vectors and employing them in a neural network regression.
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 | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| 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".