Jul 2026· Journal of Physical Chemistry Letters· Vol 17, pp. 8585-8594· 0 citations· 36 references
Medicine
TL;DR
DynStabNet is designed as a fast surrogate model to accelerate the prescreening of candidate structures prior to subsequent MLIP- or DFT-based phonon validation, and achieves an accuracy of 97% while reducing the evaluation time per structure from several hours to approximately 1 ms.
Abstract
Semiconductor materials are widely used in electronic, optoelectronic, and energy applications. While DFT-based phonon calculations provide highly accurate assessments for dynamical stability of structures, their prohibitive computational cost poses a significant bottleneck for large-scale materials screening. Herein, we develop DynStabNet, an E(3)-equivariant graph neural network (E3GNN) framework that learns dynamical stability from phonon-informed data, enabling rapid prediction without the need for explicit phonon calculations at the inference stage. Rather than serving as a replacement for first-principles phonon calculations, DynStabNet is designed as a fast surrogate model to accelerate the prescreening of candidate structures prior to subsequent MLIP- or DFT-based phonon validation. Specifically, a crystal structure generation model is utilized to produce diverse candidates, and a pretrained machine learning potential is employed to rapidly compute their phonon spectra to construct the training data set. The E(3)-equivariant architecture enables DynStabNet to capture the complex relationships between crystal structures and their dynamical stability. Consequently, unstable configurations can be rapidly eliminated early in the materials design pipeline. The DynStabNet model achieves an accuracy of 97% while reducing the evaluation time per structure from several hours to approximately 1 ms. By dramatically accelerating the overall screening process, this approach provides an efficient framework for large-scale materials screening.
An adaptation of the Δ-ML strategy for quantum property prediction of transition metal complexes is presented, which consistently achieves higher accuracy in the prediction of high-fidelity targets, while demonstrating improved data efficiency and out-of-domain transferability.
Hannes Kneiding, David Balcells· Chemistry· 0 citations
Crystals are cornerstone materials for semiconductors and renewable energy, yet their discovery is hindered by the prohibitive cost of Density Functional Theory (DFT). While geometric graph neural networks have advanced Crystal Structure Prediction (CSP) and energy estimation, existing methods treat these tasks as disparate problems with task-specific architectures, failing to exploit cross-task synergies. In this paper, we propose UniPath, a novel framework based on Learnable-Time Flow Matching that unifies crystal structure and energy prediction within a single probability path. Our key insight is modeling unstable crystals as an intermediate state along the probability path from a prior to the stable distribution. We introduce two pivotal mechanisms: (1) a learnable time parameter that dynamically identifies the intermediate state, and (2) an auxiliary velocity field that parameterizes tailored flows to synchronize the intermediate and unstable distributions. We theoretically prove that under optimal velocity fields, these paths converge to a consistent trajectory, enabling a unified backbone to maximize shared representation learning capabilities. Extensive evaluations on CSP and energy prediction benchmarks demonstrate that UniPath significantly outperforms task-specific baselines. Notably, UniPath achieves over 50% higher Match Rate on MPTS-52 and over 30% lower energy MAE on MPT-MPTS-52. Furthermore, UniPath generalizes effectively to small molecules, surpassing conventional models in molecular conformation generation. The source code is publicly available at https://github.com/GLAD-RUC/UniPath.
Songyou Li, Mingze Li, Qianpu Liu et al.· Proceedings of the 32nd ACM...· 1 citation
Machine learning potentials (MLPs) have emerged as game-changing tools for large-scale atomic simulations, overcoming the poor-scaling limitation intrinsic to traditional quantum mechanics (QM) methods. However, accurately incorporating electronic information remains a significant challenge for MLPs, particularly in efficiently computing the dynamic properties of matter under electric fields─a task central to topics such as infrared spectroscopy, interfaces under electric fields, and ferroelectric polarization. Herein, we report a physics-informed pairwise charge-transfer (PQT) theory to derive dynamic equations for macroscopic polarization that inherently conserve fundamental physical laws. Using the PQT theory, a Generalized Global Neural Network (GGNN) enhanced with the PQT mechanism is developed for the rapid prediction of dynamic properties under electric fields, applicable to both molecules and materials across the periodic table. Specifically, a generalized global data set comprising 3.18 million structures with atomic charges for 81 elements is utilized to pretrain a GGNN patched with PQT modules. Leveraging this pretrained GGNN-PQT potential, we can conveniently sample the potential energy surface under electric fields and fine-tune the potential using a small QM data set containing exact response properties at low cost. Our GGNN-PQT has linear scaling and introduces a low computational overhead compared to the standard GGNN, yet achieves both high speeds and low scaling. We demonstrate the performance of GGNN-PQT in computing dynamic response properties across a wide range of systems, including isolated molecules, adsorbed molecules, molecular crystals, liquid water, and ferroelectric materials.
Xin-Tian Xie, Zhen-Xiong Wang, Zhen-Xing Yang et al.· Journal of Chemical Theory a...· 0 citations
Crystal structure prediction (CSP) is a cornerstone technology for the efficient discovery and rational design of functional materials. Here, we propose a deep-learning-enabled dual-mode CSP framework that simultaneously supports two complementary tasks: predicting stable crystal structures for given elemental compositions and identifying chemically viable elemental substitutions for a predefined crystal topology. The framework employs an improved normalized structural fingerprint descriptor that integrates local coordination topology without global spatial-scale information. A cascaded site-probability model, composed of an autoencoder and sigmoid-based classifiers, is developed to predict the occupancy probabilities of 84 chemical elements at crystallographically distinct sites and to efficiently rank candidate structures accordingly. Remarkably, even when trained solely on topological information, the model autonomously captures elemental chemical similarity and intrinsic periodic trends, yielding chemically meaningful multielement probability distributions. Application of the proposed framework to high-throughput screening of superhard materials in the B–N system successfully identifies a thermodynamically, mechanically, and dynamically stable hexagonal BN phase, together with a metastable monoclinic B2N3 structure. Furthermore, elemental substitution screening on the zinc-blende prototype reveals two metastable compounds, In4Sb4 and Ga4Sb4. The proposed framework provides a low-cost, high-efficiency, data-driven strategy for the rapid discovery of inorganic materials.
Mandala is a modular software framework for learning block-sparse electronic-structure matrices with E(3)-equivariant graph neural networks that connects electronic-structure learning and observable-guided modeling while retaining a representation tied to quantum-mechanical operators rather than only scalar or vector targets as in MLIPs.
B. Brzoza, Wiktoria Szopa, Z. Elabid et al.· 0 citations
Predicting the structure-dependent dielectric responses of high-k oxides remains a fundamental bottleneck in the development of next-generation nanoelectronics, primarily due to the complex nature of ionic polarization. In this work, we propose a physics-informed hybrid framework designed for the performance prediction of these materials. We explicitly decouple the total dielectric constant into its electronic (εel) and ionic (εion) contributions. To capture the multi-body interactions and bond-angle distortions that govern εion, we employ the Atomistic Line Graph Neural Network (ALIGNN). Crucially, the extracted structure-aware representations are coupled with a pre-training strategy and refined via an XGBoost ensemble regressor within a stacking architecture. This approach achieves high predictive accuracy, yielding an R2 of 0.943 for εel and 0.791 for the inherently challenging εion. Furthermore, it reduces the log-domain mean absolute error (MAE) of the total dielectric constant to 0.073, corresponding to a physical-domain MAE of ≈2.9, demonstrating improvements over both pure tree-based baselines and vanilla graph networks. Deploying this framework for high-throughput screening, we evaluated candidate oxides against stringent criteria, including a wide bandgap threshold Eg > 4.0 eV to suppress leakage currents. The pipeline successfully identified three promising high-k candidates, (Sr3Hf2O7, SrHfO3, Li2HfO3) offering a physically interpretable and scalable route for data-driven performance predictions of advanced electronic materials.
