Zihan Yan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zihan Yan has a total of 124 co-authors. Here are the top co-authors by number of shared publications:
- Yizhou Zhu
Department of Materials Science and Engineering, Westlake University, Hangzhou, China
9 shared publications - Zheyong Fan
Bohai University
4 shared publications Benrui Tang
College of Physical Science and Technology Bohai University Jinzhou China
3 shared publications- Boyi Situ
College of Physics Science and Technology & Microelectronics Industry Research Institute, Yangzhou University, Jiangsu 225009, China
3 shared publications Hao Yang
School of Physical Science and Technology ShanghaiTech University Shanghai China
3 shared publications- Yusong Tu
Yangzhou University
3 shared publications
Latest Publications
- Open Access
- 7 September 2026
- Published by Springer Nature in npj Energy Materials
- Vol. 1 (1) , 14
A perspective on training machine learning force fields for solid-state electrolyte materials
Machine learning force fields enable high-accuracy modeling of solid-state electrolytes (SSEs). This perspective evaluates dataset size, reference quality, and model architectures. We show that crystalline frameworks of bulk SSEs favor efficient learning, prioritizing data quality over quantity. Crucially, force RMSE does not reliably predict transport performance. By analyzing locality and benchmarking frameworks, we provide practical guidelines for modeling bulk SSEs to accelerate the development of next-generation solid-state batteries.
- 7 September 2026
- Published by Springer Nature in npj Energy Materials
- Vol. 1 (1) , 14
Open AccessA perspective on training machine learning force fields for solid-state electrolyte materials
Machine learning force fields enable high-accuracy modeling of solid-state electrolytes (SSEs). This perspective evaluates dataset size, reference quality, and model architectures. We show that crystalline frameworks of bulk SSEs favor efficient learning, prioritizing data quality over quantity. Crucially, force RMSE does not reliably predict transport performance. By analyzing locality and benchmarking frameworks, we provide practical guidelines for modeling bulk SSEs to accelerate the development of next-generation solid-state batteries.
- Open Access
- 8 August 2026
- Published by Wiley in Advanced Energy Materials
- p. e71368
Computational Discovery of Novel Sulfide Lithium Superionic Conductors via a Structural‐Feature‐Constrained Generative Model
Recent advances in generative models have opened new avenues for discovering novel materials with desired properties by steering generation toward property constraints. However, for properties that are scarce, inconsistent across different sources, or difficult to obtain, such as Li ionic conductivity or activation energy, directly incorporating property constraints into generative models remains challenging. To address this challenge, a generative strategy is proposed based on structural feature constraints rather than property constraints, utilizing structural features strongly correlated with target properties to steer the generation process. Focusing on lithium sulfides, a simple descriptor is constructed based on the average anion‐anion coordination number to identify the body‐centered cubic anion framework, a structural feature known to facilitate fast Li ionic diffusion. By incorporating this structural feature constraint into a generative model, a set of ternary lithium sulfides with body‐centered cubic anion framework is generated. Among them, 21 materials exhibit room‐temperature ionic conductivities exceeding 0.1 mS/cm according to machine learning molecular dynamics simulations, confirming their superionic behavior. This structural‐feature‐constrained generative model not only accelerates the discovery of lithium superionic conductors but also provides a general framework and universal strategy for the inverse design of materials with desired properties.
- 8 August 2026
- Published by Wiley in Advanced Energy Materials
- p. e71368
Open AccessComputational Discovery of Novel Sulfide Lithium Superionic Conductors via a Structural‐Feature‐Constrained Generative Model
Recent advances in generative models have opened new avenues for discovering novel materials with desired properties by steering generation toward property constraints. However, for properties that are scarce, inconsistent across different sources, or difficult to obtain, such as Li ionic conductivity or activation energy, directly incorporating property constraints into generative models remains challenging. To address this challenge, a generative strategy is proposed based on structural feature constraints rather than property constraints, utilizing structural features strongly correlated with target properties to steer the generation process. Focusing on lithium sulfides, a simple descriptor is constructed based on the average anion‐anion coordination number to identify the body‐centered cubic anion framework, a structural feature known to facilitate fast Li ionic diffusion. By incorporating this structural feature constraint into a generative model, a set of ternary lithium sulfides with body‐centered cubic anion framework is generated. Among them, 21 materials exhibit room‐temperature ionic conductivities exceeding 0.1 mS/cm according to machine learning molecular dynamics simulations, confirming their superionic behavior. This structural‐feature‐constrained generative model not only accelerates the discovery of lithium superionic conductors but also provides a general framework and universal strategy for the inverse design of materials with desired properties.
- Open Access
- 3 June 2026
- Published by Wiley in Materials Genome Engineering Advances
- Vol. 4 (2) , e70074
GPUMDkit: A User‐Friendly Toolkit for GPUMD and NEP
Machine‐learned interatomic potentials have revolutionized molecular dynamics simulations by providing quantum–mechanical accuracy at empirical–potential speeds. The graphics processing unit molecular dynamics (GPUMD) package, featuring the highly efficient neuroevolution potential (NEP) framework, has emerged as a powerful tool in this domain. However, the complexity of force field development, active learning, and trajectory post‐processing often requires extensive manual scripting, imposing a steep learning curve on new users. To address this, we present GPUMDkit, a comprehensive and user‐friendly toolkit that streamlines the entire simulation workflow for GPUMD and NEP. GPUMDkit integrates a suite of essential functionalities, including format conversion, structure sampling, property calculation, and data visualization, accessible through both interactive and command‐line interfaces. Its modular, extensible architecture ensures accessibility for users of all experience levels while allowing seamless integration of new features. By automating complex tasks and enhancing productivity, GPUMDkit substantially lowers the barrier to using GPUMD and NEP programs. This article describes the program architecture and demonstrates its capabilities through practical applications. Trial Registration: GitHub Repository: https://github.com/zhyan0603/GPUMDkit
- 3 June 2026
- Published by Wiley in Materials Genome Engineering Advances
- Vol. 4 (2) , e70074
Open AccessGPUMDkit: A User‐Friendly Toolkit for GPUMD and NEP
Machine‐learned interatomic potentials have revolutionized molecular dynamics simulations by providing quantum–mechanical accuracy at empirical–potential speeds. The graphics processing unit molecular dynamics (GPUMD) package, featuring the highly efficient neuroevolution potential (NEP) framework, has emerged as a powerful tool in this domain. However, the complexity of force field development, active learning, and trajectory post‐processing often requires extensive manual scripting, imposing a steep learning curve on new users. To address this, we present GPUMDkit, a comprehensive and user‐friendly toolkit that streamlines the entire simulation workflow for GPUMD and NEP. GPUMDkit integrates a suite of essential functionalities, including format conversion, structure sampling, property calculation, and data visualization, accessible through both interactive and command‐line interfaces. Its modular, extensible architecture ensures accessibility for users of all experience levels while allowing seamless integration of new features. By automating complex tasks and enhancing productivity, GPUMDkit substantially lowers the barrier to using GPUMD and NEP programs. This article describes the program architecture and demonstrates its capabilities through practical applications. Trial Registration: GitHub Repository: https://github.com/zhyan0603/GPUMDkit
- Top Cited PapersOpen Access
- 20 April 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (9) , 4787-4801
qNEP: A Highly Efficient Neuroevolution Potential with Dynamic Charges for Large-Scale Atomistic Simulations
Although electrostatics can be incorporated into machine-learned interatomic potentials, existing approaches are computationally very demanding, limiting large-scale, long-time simulations of electrostatics-driven phenomena such as dielectric response, infrared activity, and field–matter coupling. Here, we extend the neuroevolution potential (NEP), a highly efficient machine-learned interatomic potential, to a charge-aware framework (qNEP) by introducing explicit, environment-dependent partial charges. Each ionic partial charge is represented by a neural network as a function of the local descriptor vector, analogous to the NEP site-energy model. This formulation enables the direct prediction of the Born effective charge tensor for each ion and, consequently, the polarization. As a result, dielectric properties, infrared spectra, and coupling to external electric fields can be evaluated within a unified framework. We derive consistent expressions for the forces and virials that explicitly account for the position dependence of the partial charges. The qNEP method has been implemented in the free-and-open-source GPUMD package with support for both Ewald summation and particle–particle particle–mesh treatments of electrostatics. We demonstrate the accuracy and efficiency of the qNEP approach through representative applications to water, Li7La3Zr2O12, BaTiO3, and a magnesium–water interface. These results show that qNEP enables accurate atomistic simulations with explicit long-range electrostatics, scalable to million-atom systems on nanosecond time scales using consumer-grade GPUs.
- 20 April 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (9) , 4787-4801
Top Cited PapersOpen AccessqNEP: A Highly Efficient Neuroevolution Potential with Dynamic Charges for Large-Scale Atomistic Simulations
Although electrostatics can be incorporated into machine-learned interatomic potentials, existing approaches are computationally very demanding, limiting large-scale, long-time simulations of electrostatics-driven phenomena such as dielectric response, infrared activity, and field–matter coupling. Here, we extend the neuroevolution potential (NEP), a highly efficient machine-learned interatomic potential, to a charge-aware framework (qNEP) by introducing explicit, environment-dependent partial charges. Each ionic partial charge is represented by a neural network as a function of the local descriptor vector, analogous to the NEP site-energy model. This formulation enables the direct prediction of the Born effective charge tensor for each ion and, consequently, the polarization. As a result, dielectric properties, infrared spectra, and coupling to external electric fields can be evaluated within a unified framework. We derive consistent expressions for the forces and virials that explicitly account for the position dependence of the partial charges. The qNEP method has been implemented in the free-and-open-source GPUMD package with support for both Ewald summation and particle–particle particle–mesh treatments of electrostatics. We demonstrate the accuracy and efficiency of the qNEP approach through representative applications to water, Li7La3Zr2O12, BaTiO3, and a magnesium–water interface. These results show that qNEP enables accurate atomistic simulations with explicit long-range electrostatics, scalable to million-atom systems on nanosecond time scales using consumer-grade GPUs.
- Top Cited Papers
- 29 January 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66 (3) , 1406-1413
Improving Robustness and Training Efficiency of Machine-Learned Potentials by Incorporating Short-Range Empirical Potentials
Machine learning force fields (MLFFs) are powerful tools for materials modeling, but their performance is often limited by training data set quality, particularly the lack of rare event configurations. This limitation undermines their accuracy and robustness in long- and large-scale molecular dynamics simulations. In this work, we demonstrate a simple yet general approach to improving robustness and training efficiency by utilizing a hybrid MLFF framework that integrates an empirical short-range repulsive potential. Using solid electrolyte Li7La3Zr2O12 (LLZO) as a model system, we show that purely data-driven MLFFs fail to prevent unphysical atomistic clustering in extended simulations due to inadequate short-range repulsion. In contrast, the hybrid force field eliminates these artifacts, enabling stable long-time simulations, which are critical for studying various properties of LLZO. Our approach also reduces the need for extensive active learning and performs well with just 25 training configurations. By combining physics-driven constraints with data-driven flexibility, this approach is compatible with most existing MLFF architectures and establishes a universal paradigm for developing robust, training-efficient force fields for complex material systems.
- 29 January 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66 (3) , 1406-1413
Top Cited PapersImproving Robustness and Training Efficiency of Machine-Learned Potentials by Incorporating Short-Range Empirical Potentials
Machine learning force fields (MLFFs) are powerful tools for materials modeling, but their performance is often limited by training data set quality, particularly the lack of rare event configurations. This limitation undermines their accuracy and robustness in long- and large-scale molecular dynamics simulations. In this work, we demonstrate a simple yet general approach to improving robustness and training efficiency by utilizing a hybrid MLFF framework that integrates an empirical short-range repulsive potential. Using solid electrolyte Li7La3Zr2O12 (LLZO) as a model system, we show that purely data-driven MLFFs fail to prevent unphysical atomistic clustering in extended simulations due to inadequate short-range repulsion. In contrast, the hybrid force field eliminates these artifacts, enabling stable long-time simulations, which are critical for studying various properties of LLZO. Our approach also reduces the need for extensive active learning and performs well with just 25 training configurations. By combining physics-driven constraints with data-driven flexibility, this approach is compatible with most existing MLFF architectures and establishes a universal paradigm for developing robust, training-efficient force fields for complex material systems.
- Open Access
- 16 January 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Suppressing solvent adducts via coordination competition enables scalable perovskite photovoltaics
Scalable coating methods are indispensable for the commercialization of perovskite photovoltaics. However, fundamental divergences in crystallization dynamics hinder the direct adaptation of spin-coating optimization strategies. Furthermore, the limited understanding of crystallization control under scalable conditions constrains the fabrication of high-quality perovskite films. Herein, we identify the solvent-precursor interaction time (τint) as the critical, yet previously overlooked, kinetic parameter governing film quality in scalable processes. We demonstrate the prolonged τint inherent to blade coating stabilizes solvent-adduct phases, increases solvent retention, and ultimately degrades film crystallinity. To resolve this, we introduce a dynamic coordination competition strategy that modulates the precursor coordination equilibrium, thereby effectively shortening τint and yielding high-crystallinity films with enhanced phase purity. Consequently, the blade-coated devices deliver power conversion efficiencies (PCEs) of 26.5% (0.0665 cm2) and 22.9% (728.0 cm2). Our findings provide a kinetic knob for crystallization control and establish a robust protocol for large-area manufacturing of high-quality perovskite films.
- 16 January 2026
- Published by Springer Nature in Nature Communications
- Vol. 17 (1)
Open AccessSuppressing solvent adducts via coordination competition enables scalable perovskite photovoltaics
Scalable coating methods are indispensable for the commercialization of perovskite photovoltaics. However, fundamental divergences in crystallization dynamics hinder the direct adaptation of spin-coating optimization strategies. Furthermore, the limited understanding of crystallization control under scalable conditions constrains the fabrication of high-quality perovskite films. Herein, we identify the solvent-precursor interaction time (τint) as the critical, yet previously overlooked, kinetic parameter governing film quality in scalable processes. We demonstrate the prolonged τint inherent to blade coating stabilizes solvent-adduct phases, increases solvent retention, and ultimately degrades film crystallinity. To resolve this, we introduce a dynamic coordination competition strategy that modulates the precursor coordination equilibrium, thereby effectively shortening τint and yielding high-crystallinity films with enhanced phase purity. Consequently, the blade-coated devices deliver power conversion efficiencies (PCEs) of 26.5% (0.0665 cm2) and 22.9% (728.0 cm2). Our findings provide a kinetic knob for crystallization control and establish a robust protocol for large-area manufacturing of high-quality perovskite films.
- Top Cited PapersOpen Access
- 3 August 2025
- Published by Wiley in Materials Genome Engineering Advances
- Vol. 3 (3)
GPUMD 4.0: A high‐performance molecular dynamics package for versatile materials simulations with machine‐learned potentials
This paper provides a comprehensive overview of the latest stable release of the graphics processing units molecular dynamics (GPUMD) package, GPUMD 4.0. We begin with a brief review of its development history, starting from the initial version. We then discuss the theoretical foundations for the development of the GPUMD package, including the formulations of the interatomic force, virial and heat current for many‐body potentials, the development of the highly efficient and flexible neuroevolution potential (NEP) method, the supported integrators and related operations, the various physical properties that can be calculated on the fly, and the GPUMD ecosystem. After presenting these functionalities, we review a range of applications enabled by GPUMD, particularly in combination with the NEP approach. Finally, we outline possible future development directions for GPUMD.
- 3 August 2025
- Published by Wiley in Materials Genome Engineering Advances
- Vol. 3 (3)
Top Cited PapersOpen AccessGPUMD 4.0: A high‐performance molecular dynamics package for versatile materials simulations with machine‐learned potentials
This paper provides a comprehensive overview of the latest stable release of the graphics processing units molecular dynamics (GPUMD) package, GPUMD 4.0. We begin with a brief review of its development history, starting from the initial version. We then discuss the theoretical foundations for the development of the GPUMD package, including the formulations of the interatomic force, virial and heat current for many‐body potentials, the development of the highly efficient and flexible neuroevolution potential (NEP) method, the supported integrators and related operations, the various physical properties that can be calculated on the fly, and the GPUMD ecosystem. After presenting these functionalities, we review a range of applications enabled by GPUMD, particularly in combination with the NEP approach. Finally, we outline possible future development directions for GPUMD.
- Open Access
Improving robustness and training efficiency of machine-learned potentials by incorporating short-range empirical potentials
Machine learning force fields (MLFFs) are powerful tools for materials modeling, but their performance is often limited by training dataset quality, particularly the lack of rare event configurations. This limitation undermines their accuracy and robustness in long-time and large-scale molecular dynamics simulations. In this work, we present a hybrid MLFF framework that integrates an empirical short-range repulsive potential and demonstrates improved robustness and training efficiency. Using solid electrolyte Li$_7$La$_3$Zr$_2$O$_{12}$ (LLZO) as a model system, we show that purely data-driven MLFFs fail to prevent unphysical atomistic clustering in extended simulations due to inadequate short-range repulsion. In contrast, the hybrid force field eliminates these artifacts, enabling stable long-time simulations, which are critical for studying various properties of LLZO. The hybrid framework also reduces the need for extensive active learning and performs well with just 25 training configurations. By combining physics-driven constraints with data-driven flexibility, this approach is compatible with most existing MLFF architectures and establishes a universal paradigm for developing robust, training-efficient force fields for complex material systems.
Open AccessImproving robustness and training efficiency of machine-learned potentials by incorporating short-range empirical potentials
Machine learning force fields (MLFFs) are powerful tools for materials modeling, but their performance is often limited by training dataset quality, particularly the lack of rare event configurations. This limitation undermines their accuracy and robustness in long-time and large-scale molecular dynamics simulations. In this work, we present a hybrid MLFF framework that integrates an empirical short-range repulsive potential and demonstrates improved robustness and training efficiency. Using solid electrolyte Li$_7$La$_3$Zr$_2$O$_{12}$ (LLZO) as a model system, we show that purely data-driven MLFFs fail to prevent unphysical atomistic clustering in extended simulations due to inadequate short-range repulsion. In contrast, the hybrid force field eliminates these artifacts, enabling stable long-time simulations, which are critical for studying various properties of LLZO. The hybrid framework also reduces the need for extensive active learning and performs well with just 25 training configurations. By combining physics-driven constraints with data-driven flexibility, this approach is compatible with most existing MLFF architectures and establishes a universal paradigm for developing robust, training-efficient force fields for complex material systems.
- 19 November 2024
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 36 (23) , 11551-11557
Impact of Lithium Nonstoichiometry on Ionic Diffusion in Tetragonal Garnet-Type Li7La3Zr2O12
- 19 November 2024
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 36 (23) , 11551-11557
Impact of Lithium Nonstoichiometry on Ionic Diffusion in Tetragonal Garnet-Type Li7La3Zr2O12
- 2 August 2024
- Published by American Chemical Society (ACS) in ACS Applied Energy Materials
- Vol. 7 (16) , 7019-7024
Elucidating Anisotropic Ionic Diffusion Mechanism in Li3YCl6 with Molecular Dynamics Simulations
- 2 August 2024
- Published by American Chemical Society (ACS) in ACS Applied Energy Materials
- Vol. 7 (16) , 7019-7024
Elucidating Anisotropic Ionic Diffusion Mechanism in Li3YCl6 with Molecular Dynamics Simulations