Pengfei Suo
Pengfei Suo has a total of 15 co-authors. Here are the top co-authors by number of shared publications:
Hongzhen Tian
School of Materials Science and Engineering, Beihang University, Beijing 100191, People’s Republic of China
3 shared publications- Lin-Wang Wang
State Key Laboratory of Superlattices and Microstructures, Institute of Semiconductors, Chinese Academy of Science, Beijing 100083, People’s Republic of China
3 shared publications Xingxing Wu
Technology Department, Beijing Lonxun Quantum Co., Ltd., Beijing 100192, People’s Republic of China
3 shared publications- Hanping Li
State Key Laboratory of Advanced Separation Membrane Materials, School of Chemistry & School of Material Science and Engineering & School of Chemical Engineering and Technology, Tiangong University, Tianjin, 300387, PR China
1 shared publication - Hongxing Xu
Wuhan University
1 shared publication - Jing Shi
School of Physics and Technology, Wuhan University, Wuhan 430072, China
1 shared publication
Latest Publications
- Open Access
- 6 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Towards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- 6 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessTowards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- Open Access
- 22 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Towards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- 22 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessTowards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- Open Access
- 6 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Towards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- 6 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessTowards Scalable and Efficient Machine-Learning Force Fields: The MatPL package and Its Advancements on Neuroevolution Potentials
The development of accurate and efficient machine-learning interatomic potentials is a transformative technology in computational materials science and molecular dynamics simulations. Here, we introduce the Materials Potential Library (MatPL) , a comprehensive software framework integrating multiple machine-learning potential architectures, with a particular emphasis on advancing the neuroevolution potential (NEP) framework. MatPL incorporates gradient-based optimizers to improve training stability and accuracy. A key innovation is its highly optimized, GPU-native interface for LAMMPS, seamlessly integrated with LAMMPS’s robust Message Passing Interface (MPI) infrastructure. This design transcends the limitations of single-node GPU execution, enabling efficient cross-node scalability without compromising the computational efficiency intrinsic to NEP. MatPL is a high-performance and versatile platform for next-generation atomistic modeling, balancing between high accuracy, training efficiency, and large-scale simulation capability.
- Open Access
- 31 January 2026
- Published by Elsevier in Materials Today Bio
- Vol. 37, 102862
Self-passivated bilayer black phosphorus QDs based multifunctional nanoparticles for tumor immune reprogramming
- 31 January 2026
- Published by Elsevier in Materials Today Bio
- Vol. 37, 102862
Open AccessSelf-passivated bilayer black phosphorus QDs based multifunctional nanoparticles for tumor immune reprogramming
- Open Access
- 20 May 2022
- Published by MDPI AG in Nanomaterials
- Vol. 12 (10) , 1746
Cost-Effective Calculation of Collective Electronic Excitations in Graphite Intercalated Compounds
Graphite/graphene intercalation compounds with good and improving electrical transport properties, optical properties, magnetic properties and even superconductivity are widely used in battery, capacitors and so on. Computational simulation helps with predicting important properties and exploring unknown functions, while it is restricted by limited computing resources and insufficient precision. Here, we present a cost-effective study on graphite/graphene intercalation compounds properties with sufficient precision. The calculation of electronic collective excitations in AA-stacking graphite based on the tight-binding model within the random phase approximation framework agrees quite well with previous experimental and calculation work, such as effects of doping level, interlayer distance, and interlayer hopping on 2D plasmon and 3D intraband plasmon modes. This cost-effective simulation method can be extended to other intercalation compounds with unlimited intercalation species.- 20 May 2022
- Published by MDPI AG in Nanomaterials
- Vol. 12 (10) , 1746
Open AccessCost-Effective Calculation of Collective Electronic Excitations in Graphite Intercalated Compounds
Graphite/graphene intercalation compounds with good and improving electrical transport properties, optical properties, magnetic properties and even superconductivity are widely used in battery, capacitors and so on. Computational simulation helps with predicting important properties and exploring unknown functions, while it is restricted by limited computing resources and insufficient precision. Here, we present a cost-effective study on graphite/graphene intercalation compounds properties with sufficient precision. The calculation of electronic collective excitations in AA-stacking graphite based on the tight-binding model within the random phase approximation framework agrees quite well with previous experimental and calculation work, such as effects of doping level, interlayer distance, and interlayer hopping on 2D plasmon and 3D intraband plasmon modes. This cost-effective simulation method can be extended to other intercalation compounds with unlimited intercalation species.