Software-hardware co-design accelerates materials simulations

Xiaozhe Wang; Wei Zhang; En Ma

doi:10.1088/1674-4926/26010039

J. Semicond. > Just Accepted

NEWS AND VIEWS

Software-hardware co-design accelerates materials simulations

Xiaozhe Wang, Wei Zhang and En Ma

+ Author Affiliations

Center for Alloy Innovation and Design (CAID), State Key Laboratory for Mechanical Behavior of Materials, Xi’an Jiaotong University, Xi’an 710049, China

Author Bio:

Xiaozhe Wang is currently an assistant professor at Xi'an Jiaotong University. He received his bachelor’s degree and doctoral degree both at Xi’an Jiaotong University. Dr. Wang focuses on materials synthesis and device fabrications of phase-change materials for non-volatile memory and neuro-inspired computing applications.

Wei Zhang is currently a professor at Xi'an Jiaotong University. Prof. Zhang is the director of Center for Alloy Innovation and Design (CAID), and serves as a committee member for the European Phase-Change and Ovonic Symposium (E\PCOS). His current research interests include phase-change materials for non-volatile memory and neuro-inspired computing, first-principles materials design, as well as machine-learned interatomic potentials. Prof. Zhang has published 104 archival papers, including 3 in Science and 8 in Nature Materials/Electronics/Reviews Materials. His publications have received more than 6,800 citations (Google Scholar) with an h index of 41.

En Ma En (Evan) Ma is currently a professor at Xi'an Jiaotong University. Prof. Ma is an elected Fellow of the Materials Research Society (MRS), ASM International, the American Physical Society (APS), and TMS. In 2025, Prof. Ma was elected a Fellow of the European Academy of Sciences. Prof. Ma's research interest focuses on metastable materials, including amorphous alloys, chalcogenide phase-change memory alloys and multi-principal element high-entropy alloys. He has published approximately 450 archival papers, including 33 in Science, Nature, and Nature Materials/Physics/Electronics, and another 24 papers in Nature Communications. His publications have received more than 70,000 citations (Google Scholar) with an h index of 134.

Corresponding author: Wei Zhang, wzhang0@mail.xjtu.edu.cn; En Ma, maen@xjtu.edu.cn

DOI: 10.1088/1674-4926/26010039CSTR: 32376.14.1674-4926.26010039

References

[1]	Horton M K, Huck P, Yang R X, et al. Accelerated data-driven materials science with the Materials Project. Nat Mater, 2025, 24(10): 1522 doi: 10.1038/s41563-025-02272-0
[2]	Huang B, von Rudorff G F, von Lilienfeld O A. The central role of density functional theory in the AI age. Science, 2023, 381(6654): 170 doi: 10.1126/science.abn3445
[3]	Friederich P, Häse F, Proppe J, et al. Machine-learned potentials for next-generation matter simulations. Nat Mater, 2021, 20(6): 750 doi: 10.1038/s41563-020-0777-6
[4]	Zhou Y X, Zhang W, Ma E, et al. Device-scale atomistic modelling of phase-change memory materials. Nat Electron, 2023, 6(10): 746 doi: 10.1038/s41928-023-01030-x
[5]	Zhou Y X, Thomas du Toit D F, Elliott S R, et al. Full-cycle device-scale simulations of memory materials with a tailored atomic-cluster-expansion potential. Nat Commun, 2025, 16(1): 8688 doi: 10.1038/s41467-025-63732-4
[6]	Wang G J, Wang C R, Zhang X G, et al. Machine learning interatomic potential: Bridge the gap between small-scale models and realistic device-scale simulations. iScience, 2024, 27(5): 109673 doi: 10.1016/j.isci.2024.109673
[7]	Li H, Wang Z, Zou N L, et al. Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation. Nat Comput Sci, 2022, 2(6): 367 doi: 10.1038/s43588-022-00265-6
[8]	Lanza M, Pazos S, Aguirre F, et al. The growing memristor industry. Nature, 2025, 640(8059): 613 doi: 10.1038/s41586-025-08733-5
[9]	Wang X Z, Wang R B, Sun S Y, et al. Amorphous phase-change memory alloy with no resistance drift. Nat Mater, 2026, 25(3): 456 doi: 10.1038/s41563-025-02361-0
[10]	Xu M, Wang S C, He Y G, et al. Efficient modeling of ionic and electronic interactions by a resistive memory-based reservoir graph neural network. Nat Comput Sci, 2025, 5(12): 1178 doi: 10.1038/s43588-025-00844-3

Fig. 1. (Color online) Software-hardware co-design architecture and performance metrics for accelerating materials simulations. (a) Schematic of the RGNN architecture for the prediction of materials properties. (b) Resistive-memory-based hybrid analogue-digital computing system for accelerating RGNN inference. (c) Estimated computational complexity per MD step for AIMD and for the proposed software-hardware co-design. (d) Decomposition of the training cost of the RGNN in comparison with a fully trained GNN. (e) Breakdown of inference energy across different hardware platforms. Adapted with permission from Ref. [10], Springer Nature Limited.

DownLoad: Full-Size Img PowerPoint

[1]	Horton M K, Huck P, Yang R X, et al. Accelerated data-driven materials science with the Materials Project. Nat Mater, 2025, 24(10): 1522 doi: 10.1038/s41563-025-02272-0
[2]	Huang B, von Rudorff G F, von Lilienfeld O A. The central role of density functional theory in the AI age. Science, 2023, 381(6654): 170 doi: 10.1126/science.abn3445
[3]	Friederich P, Häse F, Proppe J, et al. Machine-learned potentials for next-generation matter simulations. Nat Mater, 2021, 20(6): 750 doi: 10.1038/s41563-020-0777-6
[4]	Zhou Y X, Zhang W, Ma E, et al. Device-scale atomistic modelling of phase-change memory materials. Nat Electron, 2023, 6(10): 746 doi: 10.1038/s41928-023-01030-x
[5]	Zhou Y X, Thomas du Toit D F, Elliott S R, et al. Full-cycle device-scale simulations of memory materials with a tailored atomic-cluster-expansion potential. Nat Commun, 2025, 16(1): 8688 doi: 10.1038/s41467-025-63732-4
[6]	Wang G J, Wang C R, Zhang X G, et al. Machine learning interatomic potential: Bridge the gap between small-scale models and realistic device-scale simulations. iScience, 2024, 27(5): 109673 doi: 10.1016/j.isci.2024.109673
[7]	Li H, Wang Z, Zou N L, et al. Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation. Nat Comput Sci, 2022, 2(6): 367 doi: 10.1038/s43588-022-00265-6
[8]	Lanza M, Pazos S, Aguirre F, et al. The growing memristor industry. Nature, 2025, 640(8059): 613 doi: 10.1038/s41586-025-08733-5
[9]	Wang X Z, Wang R B, Sun S Y, et al. Amorphous phase-change memory alloy with no resistance drift. Nat Mater, 2026, 25(3): 456 doi: 10.1038/s41563-025-02361-0
[10]	Xu M, Wang S C, He Y G, et al. Efficient modeling of ionic and electronic interactions by a resistive memory-based reservoir graph neural network. Nat Comput Sci, 2025, 5(12): 1178 doi: 10.1038/s43588-025-00844-3