In recent years, van der Waals (vdW) heterostructures, formed by vertically stacking two-dimensional materials, have garnered significant attention from both academia and industry due to their unique interfacial mechanical and physical properties. Research in this area has emerged as a new interdisciplinary frontier. Notable examples include the superconductivity observed in magic-angle twisted bilayer graphene and the superlubricity (ultra-low friction and wear) in graphene/h-BN heterostructures. Experiments have shown that these properties are closely intertwined with the moiré superlattices formed at the heterointerfaces. To understand the underlying mechanisms, it is crucial to accurately describe the long-range vdW interlayer interactions.
The study of vdW heterostructures presents significant computational challenges. While first-principles calculations offer high accuracy in describing these systems, they are computationally intensive and impractical for simulating large-scale moiré superlattice systems. On the other hand, molecular dynamics (MD) simulations provide a more computationally feasible approach but face limitations in accuracy. The accuracy of molecular dynamics simulations heavily depends on the adopted force fields. The commonly used combination of bond-order potentials for intralayer interactions and Lennard-Jones potentials for interlayer interactions in vdW layered materials fails to accurately predict and explain experimental observations. This limitation hinders a deeper understanding of the mechanisms behind the interfacial mechanical and physical properties of vdW heterostructures. Consequently, it becomes challenging to offer reliable experimental guidance for enhancing or tuning their performance, thereby restricting the broader application of these novel materials.
To address this challenge, our research group has been dedicated to developing a unified method for rapidly constructing accurate anisotropic interlayer potentials (ILPs) for various van der Waals heterostructures. These ILPs are then combined with either bond-order potentials or machine learning potentials (MLPs) that describe intralayer interactions, enabling realistic MD simulations. MLPs offer superior accuracy compared to empirical bond-order potentials in describing the mechanical and physical properties of vdW layered materials within individual layers. However, both approaches struggle to adequately capture long-range interlayer interactions. Conversely, ILPs excel at accurately describing interlayer vdW interactions but are not designed to model intralayer interactions. By combining ILPs with MLPs, we leverage the strengths of each approach, creating a more comprehensive and accurate framework for simulating vdW heterostructures.
Then we integrate the developed computational framework into the well-known open-source MD software LAMMPS and GPUMD, providing efficient and reliable computational methods for large-scale simulations of vdW heterostructures. Utilizing these advanced computational tools, one can systematically investigate factors influencing the interfacial mechanical and physical properties of these heterostructures, uncover underlying mechanisms, and develop corresponding theoretical models. Our goal is to offer reliable guidance for the experimental design of vdW heterostructures with superior performance, thereby promoting their widespread application in smart devices and advanced manufacturing.
A key advantage of our computational framework lies in its modular and versatile nature. The ILPs and MLPs for various vdW layered materials can be developed independently and then seamlessly combined. This approach allows for the simulation of a wide range of vdW heterostructure combinations and multilayer systems without the need to retrain the force fields for each new heterostructure. This flexibility significantly enhances the framework's transferability and efficiency in studying diverse heterostructures. To date, our research group has successfully developed multiple ILPs and MLPs for various van der Waals interfaces. In line with our commitment to advancing scientific research and collaboration, we have made all of these potentials open-source, enabling researchers worldwide to access, utilize, and build upon our work in their own studies of vdW heterostructures. By providing these tools to the scientific community, we aim to accelerate progress in understanding and harnessing the unique properties of vdW heterostructures, potentially leading to breakthroughs in materials science and nanotechnology.
The study of vdW heterostructures presents significant computational challenges. While first-principles calculations offer high accuracy in describing these systems, they are computationally intensive and impractical for simulating large-scale moiré superlattice systems. On the other hand, molecular dynamics (MD) simulations provide a more computationally feasible approach but face limitations in accuracy. The accuracy of molecular dynamics simulations heavily depends on the adopted force fields. The commonly used combination of bond-order potentials for intralayer interactions and Lennard-Jones potentials for interlayer interactions in vdW layered materials fails to accurately predict and explain experimental observations. This limitation hinders a deeper understanding of the mechanisms behind the interfacial mechanical and physical properties of vdW heterostructures. Consequently, it becomes challenging to offer reliable experimental guidance for enhancing or tuning their performance, thereby restricting the broader application of these novel materials.
To address this challenge, our research group has been dedicated to developing a unified method for rapidly constructing accurate anisotropic interlayer potentials (ILPs) for various van der Waals heterostructures. These ILPs are then combined with either bond-order potentials or machine learning potentials (MLPs) that describe intralayer interactions, enabling realistic MD simulations. MLPs offer superior accuracy compared to empirical bond-order potentials in describing the mechanical and physical properties of vdW layered materials within individual layers. However, both approaches struggle to adequately capture long-range interlayer interactions. Conversely, ILPs excel at accurately describing interlayer vdW interactions but are not designed to model intralayer interactions. By combining ILPs with MLPs, we leverage the strengths of each approach, creating a more comprehensive and accurate framework for simulating vdW heterostructures.
Then we integrate the developed computational framework into the well-known open-source MD software LAMMPS and GPUMD, providing efficient and reliable computational methods for large-scale simulations of vdW heterostructures. Utilizing these advanced computational tools, one can systematically investigate factors influencing the interfacial mechanical and physical properties of these heterostructures, uncover underlying mechanisms, and develop corresponding theoretical models. Our goal is to offer reliable guidance for the experimental design of vdW heterostructures with superior performance, thereby promoting their widespread application in smart devices and advanced manufacturing.
A key advantage of our computational framework lies in its modular and versatile nature. The ILPs and MLPs for various vdW layered materials can be developed independently and then seamlessly combined. This approach allows for the simulation of a wide range of vdW heterostructure combinations and multilayer systems without the need to retrain the force fields for each new heterostructure. This flexibility significantly enhances the framework's transferability and efficiency in studying diverse heterostructures. To date, our research group has successfully developed multiple ILPs and MLPs for various van der Waals interfaces. In line with our commitment to advancing scientific research and collaboration, we have made all of these potentials open-source, enabling researchers worldwide to access, utilize, and build upon our work in their own studies of vdW heterostructures. By providing these tools to the scientific community, we aim to accelerate progress in understanding and harnessing the unique properties of vdW heterostructures, potentially leading to breakthroughs in materials science and nanotechnology.
1, Force fields Implemented in LAMMPS:
(1) ILP for graphene with edges, boron nitride, and their heterostructures (Ref: Nano Lett. 18, 6009-6016 (2018))
https://docs.lammps.org/pair_ilp_graphene_hbn.html
(2)KC for graphene systems with edges(Ref: Nano Lett. 18, 6009-6016 (2018))
https://docs.lammps.org/pair_kolmogorov_crespi_full.html
(3) ILP for transition metal dichalcogenides and their heterostructures(Ref: J. Chem. Theory Comput. 17, 7237 (2021); J. Phys. Chem. A, 127, 46, 9820-9830 (2023))
https://docs.lammps.org/pair_ilp_tmd.html
(4) SAIP for heterostructures of hexagonal 2D materials and metals(Ref: J. Chem. Theory Comput. 17, 7215 (2021); J. Phys. Chem. C 2024 128 (16), 6836-6851)
https://docs.lammps.org/pair_saip_metal.html
(5) AIP for water and 2D materials(Ref: J. Phys. Chem. C. 127(18), 8704-8713 (2023); Langmuir 39(50), 18198-18207 (2023))
https://docs.lammps.org/pair_aip_water_2dm.html
(6) ILP for 2D diamond, graphyne, and their heterostructures（Ref: J. Phys. Chem. C 2023, 127, 18641−18651)
Updating……
(1) ILP for graphene with edges, boron nitride, and their heterostructures (Ref: Nano Lett. 18, 6009-6016 (2018))
https://docs.lammps.org/pair_ilp_graphene_hbn.html
(2)KC for graphene systems with edges(Ref: Nano Lett. 18, 6009-6016 (2018))
https://docs.lammps.org/pair_kolmogorov_crespi_full.html
(3) ILP for transition metal dichalcogenides and their heterostructures(Ref: J. Chem. Theory Comput. 17, 7237 (2021); J. Phys. Chem. A, 127, 46, 9820-9830 (2023))
https://docs.lammps.org/pair_ilp_tmd.html
(4) SAIP for heterostructures of hexagonal 2D materials and metals(Ref: J. Chem. Theory Comput. 17, 7215 (2021); J. Phys. Chem. C 2024 128 (16), 6836-6851)
https://docs.lammps.org/pair_saip_metal.html
(5) AIP for water and 2D materials(Ref: J. Phys. Chem. C. 127(18), 8704-8713 (2023); Langmuir 39(50), 18198-18207 (2023))
https://docs.lammps.org/pair_aip_water_2dm.html
(6) ILP for 2D diamond, graphyne, and their heterostructures（Ref: J. Phys. Chem. C 2023, 127, 18641−18651)
Updating……
2, Force fields implemented in GPUMD
(1) (1) Currently being updated (ILP for graphene, boron nitride, and their heterostructures)
(2) (2) Currently being updated (ILP for transition metal dichalcogenides and their heterostructures)
……
(1) (1) Currently being updated (ILP for graphene, boron nitride, and their heterostructures)
(2) (2) Currently being updated (ILP for transition metal dichalcogenides and their heterostructures)
……
3, Open Source Codes of force fields in GitHub
Our research group's GitHub account（ouyang-laboratory）, shares a variety of resources developed by our team, including open-source code, datasets, tools, and examples.
(1) https://github.com/ouyang-laboratory/gpumd.git (NEP+ILP computational framework for vdW interfaces)
(2) updating……
Our research group's GitHub account（ouyang-laboratory）, shares a variety of resources developed by our team, including open-source code, datasets, tools, and examples.
(1) https://github.com/ouyang-laboratory/gpumd.git (NEP+ILP computational framework for vdW interfaces)
(2) updating……