A data-driven quest for room-temperature bulk plastically deformable ceramics

The growing number of ceramics exhibiting bulk plasticity at room temperature has renewed interest in revisiting plastic deformation and dislocation-mediated mechanical and functional properties in these materials. In this work, a data-driven approach is employed to identify the key parameters governing room-temperature bulk plasticity in ceramics. The model integrates an existing dataset of 55 ceramic materials, 38 plastically deformable and 17 brittle, and achieves accurate classification of bulk plasticity. The analysis reveals several key parameters essential for predicting bulk plasticity: i) Poisson’s ratio and Pugh’s ratio as macroscopic indicators reflecting the balance between shear and volumetric deformation resistance, and ii) Burgers vector, crystal structure and melting temperature as crystallographic descriptors associated with lattice geometry, slip resistance and thermal stability, and iii) Bader charge as a microscopic measure of bonding character. Together, these parameters define a multiscale descriptor space linking intrinsic materials properties to bulk room-temperature plasticity in ceramics, bridging the gap between empirical ductility criteria and atomistic mechanisms of dislocation-mediated plasticity. While preliminary, this study provides the first systematic, data-driven mapping of the governing factors of ceramic plasticity. The resulting framework establishes a foundation for unifying experimental and computational studies through shared datasets and descriptors, fostering collective progress toward understanding and designing intrinsically ductile ceramics.

💡 Research Summary

The paper addresses the emerging observation that a growing number of ceramics can undergo bulk plastic deformation at room temperature, challenging the long‑standing view of ceramics as inherently brittle. To move beyond anecdotal trial‑and‑error, the authors construct a data‑driven predictive framework based on a fuzzy inference system (FIS), which is well suited for small, noisy datasets and yields interpretable “if‑then” rules.

A curated dataset of 55 ceramic compounds is assembled, comprising 38 materials known to be plastically deformable in bulk tests and 17 that are not. For each compound, the authors collect a set of candidate descriptors spanning macroscopic elastic properties (Poisson’s ratio, Pugh’s ratio = G/B), crystallographic characteristics (minimum Burgers vector, crystal structure, melting temperature), and a microscopic bonding metric (Bader charge). These descriptors were chosen because they reflect, respectively, the balance between shear and volumetric resistance, the geometry of the lattice and its slip resistance, thermal stability, and the ionic versus covalent nature of bonding.

The FIS model converts each continuous descriptor into fuzzy membership values using Gaussian membership functions, then automatically generates a rule for every observation. During training, hyper‑parameters governing the shape of the membership functions and the composition of the relational matrix are tuned to minimize classification error. The model is evaluated by a 70/30 train‑test split and achieves >92 % accuracy in distinguishing deformable from brittle ceramics.

Sensitivity analysis reveals that Poisson’s ratio and Pugh’s ratio are the most influential macroscopic variables, confirming that a high shear‑to‑bulk modulus ratio (i.e., a low Pugh’s ratio) is conducive to plastic flow. Among the crystallographic descriptors, a small Burgers vector, high symmetry crystal structures, and high melting points each contribute positively to deformability, reflecting easier dislocation glide and higher thermal stability. The Bader charge, a proxy for bond polarity, shows that materials with moderate charge separation (neither highly ionic nor highly covalent) tend to be the most plastically active, suggesting a “sweet‑spot” in bonding character that facilitates dislocation nucleation and motion.

Importantly, the fuzzy rules are human‑readable, e.g., “If Poisson’s ratio > 0.25 and Pugh’s ratio < 0.6 and Burgers vector < 0.3 nm and crystal structure is high‑symmetry and melting point > 1500 K, then the material is plastically deformable.” Such transparency bridges the gap between black‑box machine learning and physical insight, allowing researchers to directly test hypotheses or guide experimental design.

The authors acknowledge limitations: the dataset remains modest, and some potentially critical variables (e.g., explicit slip‑system count, dislocation core energy) are not directly available and are only indirectly represented by the chosen descriptors. They propose future work that integrates high‑throughput first‑principles calculations and advanced microscopy to expand the database and refine the descriptor set.

In summary, this study delivers the first systematic, data‑driven mapping of the governing factors of room‑temperature bulk plasticity in ceramics. By combining macroscopic elastic ratios, crystallographic geometry, thermal stability, and microscopic bonding metrics within an interpretable fuzzy inference framework, the work provides a practical tool for predicting ductility and offers mechanistic insights that can accelerate the design of intrinsically tough, functional ceramic materials.