Bo Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Bo Chen has a total of 61 co-authors. Here are the top co-authors by number of shared publications:
- Jiayu Dai
College of Science, National University of Defense Technology 1 , Changsha 410073
12 shared publications - Qiyu Zeng
National University of Defense Technology
5 shared publications - Xiaoxiang Yu
College of Science, National University of Defense Technology 1 , Changsha 410073
5 shared publications Dongdong Kang
College of Science, National University of Defense Technology , Changsha 410073
4 shared publications- Dongdong Kang
College of Science, National University of Defense Technology 1 , Changsha 410073
4 shared publications Fangyu Guo
College of Science, National University of Defense Technology , Changsha 410073
3 shared publications
Latest Publications
- 1 October 2026
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 165 (13)
Interlayer stress governed hexagonal graphite-to-hexagonal diamond transformation under uniaxial compression
The synthesis of high-purity bulk hexagonal diamond (HD) remains a formidable challenge, and a more detailed understanding of the transformation pathway from hexagonal graphite (HG) to HD remains elusive. Here, we elucidate the pathway selection mechanism and identify the favorable conditions for synthesizing high-purity bulk HD. Using deep potential molecular dynamics simulations, we systematically investigate the transformation under interlayer uniaxial compression across a broad temperature and pressure range. Our results demonstrate that interlayer stress governs the transformation pathway by regulating interlayer sliding of graphite layers. At low stress, the HG-to-HD transition proceeds through an intermediate orthorhombic graphite (OG) phase, whereas at high stress, it proceeds through a post-graphite phase; in both pathways, the final conversion to HD is completed via a buckling mechanism. Although temperature does not change the pathway type, it synergistically interacts with stress to regulate the formation of the AB′A-stacked OG phase, thereby determining the phase purity of HD. At 900 K and 55 GPa, an HD fraction of up to 86.7% is achieved. The interlayer stress-temperature map of the HD fraction exhibits an isolated island-like distribution, revealing an exceptionally narrow synthesis window. Moreover, the distribution of OG atom counts on the interlayer stress-temperature map provides a clear signature of the boundary separating the two transformation pathways. This work uncovers the pathway-selection mechanism for carbon phase transitions under extreme anisotropic stress conditions and establishes a theoretical framework for the controlled synthesis of high-purity bulk HD.
- 1 October 2026
- Published by AIP Publishing in The Journal of Chemical Physics
- Vol. 165 (13)
Interlayer stress governed hexagonal graphite-to-hexagonal diamond transformation under uniaxial compression
The synthesis of high-purity bulk hexagonal diamond (HD) remains a formidable challenge, and a more detailed understanding of the transformation pathway from hexagonal graphite (HG) to HD remains elusive. Here, we elucidate the pathway selection mechanism and identify the favorable conditions for synthesizing high-purity bulk HD. Using deep potential molecular dynamics simulations, we systematically investigate the transformation under interlayer uniaxial compression across a broad temperature and pressure range. Our results demonstrate that interlayer stress governs the transformation pathway by regulating interlayer sliding of graphite layers. At low stress, the HG-to-HD transition proceeds through an intermediate orthorhombic graphite (OG) phase, whereas at high stress, it proceeds through a post-graphite phase; in both pathways, the final conversion to HD is completed via a buckling mechanism. Although temperature does not change the pathway type, it synergistically interacts with stress to regulate the formation of the AB′A-stacked OG phase, thereby determining the phase purity of HD. At 900 K and 55 GPa, an HD fraction of up to 86.7% is achieved. The interlayer stress-temperature map of the HD fraction exhibits an isolated island-like distribution, revealing an exceptionally narrow synthesis window. Moreover, the distribution of OG atom counts on the interlayer stress-temperature map provides a clear signature of the boundary separating the two transformation pathways. This work uncovers the pathway-selection mechanism for carbon phase transitions under extreme anisotropic stress conditions and establishes a theoretical framework for the controlled synthesis of high-purity bulk HD.
- Open Access
- 23 September 2026
- Published by Oxford University Press (OUP) in National Science Review
- Vol. 13 (20)
Non-equilibrium nucleation and ultrafast growth of ice VII under ramp compression
Under ramp compression, water can rapidly freeze into ice VII within nanoseconds, in stark contrast to its freezing behavior under ambient conditions, which typically takes several seconds. Understanding this rapid freezing process is essential for planetary science and high-pressure physics. However, the underlying mechanisms, particularly at the molecular scale, remain poorly understood. Here, we employ deep neural network-based molecular dynamics simulations to investigate the ramp compression of water. Two notable phenomena are observed. (1) Extreme ramp compression delays nucleation to pressures exceeding those predicted by classical nucleation theory, indicating non-equilibrium nucleation. (2) The formation of a thermally elevated nucleus and an interfacial layer with high rotational mobility accelerates hydrogen bond rearrangement, thereby facilitating ice growth. Inspired by these findings, a scaling law is proposed to predict the transition pressure to ice VII under ramp compression, revealing that the metastable water can be overdriven far beyond its known limits. By bridging atomistic dynamics with macroscopic phase transitions, our work provides transformative insights into phase transitions under dynamic loading.
- 23 September 2026
- Published by Oxford University Press (OUP) in National Science Review
- Vol. 13 (20)
Open AccessNon-equilibrium nucleation and ultrafast growth of ice VII under ramp compression
Under ramp compression, water can rapidly freeze into ice VII within nanoseconds, in stark contrast to its freezing behavior under ambient conditions, which typically takes several seconds. Understanding this rapid freezing process is essential for planetary science and high-pressure physics. However, the underlying mechanisms, particularly at the molecular scale, remain poorly understood. Here, we employ deep neural network-based molecular dynamics simulations to investigate the ramp compression of water. Two notable phenomena are observed. (1) Extreme ramp compression delays nucleation to pressures exceeding those predicted by classical nucleation theory, indicating non-equilibrium nucleation. (2) The formation of a thermally elevated nucleus and an interfacial layer with high rotational mobility accelerates hydrogen bond rearrangement, thereby facilitating ice growth. Inspired by these findings, a scaling law is proposed to predict the transition pressure to ice VII under ramp compression, revealing that the metastable water can be overdriven far beyond its known limits. By bridging atomistic dynamics with macroscopic phase transitions, our work provides transformative insights into phase transitions under dynamic loading.
- 14 September 2026
- Published by AIP Publishing in Applied Physics Letters
- Vol. 129 (11)
Experimental observation of a fourfold difference in thermal conductivity yet comparable maximum phonon mean free path in α -alumina and α -gallium oxide
Lighter-element materials usually possess higher lattice thermal conductivity (LTC) due to both higher group velocities and longer mean free paths (MFPs) than their heavier-element counterparts. In this work, using broadband frequency-domain thermoreflectance, we measure the LTC and MFP spectra of corundum-structured α-Al2O3 and α-Ga2O3 crystals and report comparable maximum MFPs yet an approximately fourfold difference in LTC. For α-Al2O3, diffusive transport persists down to a thermal penetration depth of 135 nm, yielding an LTC of 34.9 ± 1.1 W m−1 K−1 and a maximum MFP smaller than 135 nm. For α-Ga2O3, ballistic transport appears below 120 nm, exhibiting an LTC of 9.2 ± 0.53 W m−1 K−1 and a maximum MFP of about 120 nm. This counterintuitive finding challenges the conventional expectation, providing experimental insights into the fundamental understanding and design of thermal transport in functional materials.
- 14 September 2026
- Published by AIP Publishing in Applied Physics Letters
- Vol. 129 (11)
Experimental observation of a fourfold difference in thermal conductivity yet comparable maximum phonon mean free path in α -alumina and α -gallium oxide
Lighter-element materials usually possess higher lattice thermal conductivity (LTC) due to both higher group velocities and longer mean free paths (MFPs) than their heavier-element counterparts. In this work, using broadband frequency-domain thermoreflectance, we measure the LTC and MFP spectra of corundum-structured α-Al2O3 and α-Ga2O3 crystals and report comparable maximum MFPs yet an approximately fourfold difference in LTC. For α-Al2O3, diffusive transport persists down to a thermal penetration depth of 135 nm, yielding an LTC of 34.9 ± 1.1 W m−1 K−1 and a maximum MFP smaller than 135 nm. For α-Ga2O3, ballistic transport appears below 120 nm, exhibiting an LTC of 9.2 ± 0.53 W m−1 K−1 and a maximum MFP of about 120 nm. This counterintuitive finding challenges the conventional expectation, providing experimental insights into the fundamental understanding and design of thermal transport in functional materials.
- 23 June 2026
- Published by American Physical Society (APS) in Physical Review B
- Vol. 114 (3)
Path-dependent nucleation and growth mechanisms in graphite-to-diamond transformation
- 23 June 2026
- Published by American Physical Society (APS) in Physical Review B
- Vol. 114 (3)
Path-dependent nucleation and growth mechanisms in graphite-to-diamond transformation
- 28 May 2026
- Published by American Physical Society (APS) in Physical Review B
- Vol. 114 (3)
Melting behavior of MgSiO 3 , CaSiO 3 , and SiO 2 at Earth’s mantle conditions based on machine learning potentials
- 28 May 2026
- Published by American Physical Society (APS) in Physical Review B
- Vol. 114 (3)
Melting behavior of MgSiO 3 , CaSiO 3 , and SiO 2 at Earth’s mantle conditions based on machine learning potentials
- Open Access
- 27 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 592-598
PFD Scheme: A Hierarchical Training Framework for Scaling Reliable Machine-Learned Interatomic Potentials to Complex Systems
The incorporation of foreign elements into alumina is recognized as an effective approach to suppressing the γ→α phase transformation and consequently enhancing its thermal stability. Nevertheless, the underlying stabilization mechanisms have yet to be systematically investigated. Furthermore, owing to the substantial costs and time-intensive nature of conventional experimental synthesis and computational simulations, research efforts have predominantly focused on mono-doping systems. The exploitation of intrinsic elemental properties and their synergistic interactions constitutes a critical strategy for achieving superior comprehensive performance in the design and development of novel materials. However, the paucity of high-fidelity interatomic potentials has hindered atomic-scale mechanistic studies of complex structures in multi-principal-element systems, thereby precluding high-throughput screening of effective co-doping strategies. Herein, we propose the pre-training–fine-tuning–distillation (PFD) scheme, a hierarchical training framework for scaling reliable machine-learned interatomic potentials to multi-element doping alumina systems. This work demonstrates that fine-tuning pre-trained models improves prediction accuracy by an order of magnitude, while synthetic distillation enables at least 20 times acceleration in molecular dynamics simulations for downstream tasks with negligible accuracy loss, yielding more efficient and lightweight models. The successful application of the proposed model to multi-element doping alumina systems establishes the foundation for high-throughput screening of effective co-doping strategies in alumina, while also furnishing a transferable research paradigm for high-fidelity simulations of multi-principal-element systems.
- 27 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 592-598
Open AccessPFD Scheme: A Hierarchical Training Framework for Scaling Reliable Machine-Learned Interatomic Potentials to Complex Systems
The incorporation of foreign elements into alumina is recognized as an effective approach to suppressing the γ→α phase transformation and consequently enhancing its thermal stability. Nevertheless, the underlying stabilization mechanisms have yet to be systematically investigated. Furthermore, owing to the substantial costs and time-intensive nature of conventional experimental synthesis and computational simulations, research efforts have predominantly focused on mono-doping systems. The exploitation of intrinsic elemental properties and their synergistic interactions constitutes a critical strategy for achieving superior comprehensive performance in the design and development of novel materials. However, the paucity of high-fidelity interatomic potentials has hindered atomic-scale mechanistic studies of complex structures in multi-principal-element systems, thereby precluding high-throughput screening of effective co-doping strategies. Herein, we propose the pre-training–fine-tuning–distillation (PFD) scheme, a hierarchical training framework for scaling reliable machine-learned interatomic potentials to multi-element doping alumina systems. This work demonstrates that fine-tuning pre-trained models improves prediction accuracy by an order of magnitude, while synthetic distillation enables at least 20 times acceleration in molecular dynamics simulations for downstream tasks with negligible accuracy loss, yielding more efficient and lightweight models. The successful application of the proposed model to multi-element doping alumina systems establishes the foundation for high-throughput screening of effective co-doping strategies in alumina, while also furnishing a transferable research paradigm for high-fidelity simulations of multi-principal-element systems.
- Open Access
- 3 December 2025
- Published by AIP Publishing in Matter and Radiation at Extremes
- Vol. 11 (1)
Three-step formation of diamonds in shock-compressed hydrocarbons: Dissociation, species separation, and nucleation
The accumulation and circulation of carbon and hydrogen contribute to the chemical evolution of ice giant planets. Species separation and diamond precipitation have been reported in carbon–hydrogen systems and have been verified by static and shock compression experiments. Nevertheless, the dynamic formation processes underlying these phenomena remain insufficiently understood. In combination with a deep learning model, we demonstrate that diamonds form through a three-step process involving dissociation, species separation, and nucleation processes. Under shock conditions of 125 GPa and 4590 K, hydrocarbons decompose to give hydrogen and low-molecular-weight alkanes (CH4 and C2H6), which escape from the carbon chains, resulting in C/H species separation. The remaining carbon atoms without C–H bonds accumulate and nucleate to form diamond crystals. The process of diamond growth is associated with a critical nucleus size at which the dynamic energy barrier plays a key role. These dynamic processes of diamond formation provide insight into the establishment of a model for the evolution of ice giant planets.
- 3 December 2025
- Published by AIP Publishing in Matter and Radiation at Extremes
- Vol. 11 (1)
Open AccessThree-step formation of diamonds in shock-compressed hydrocarbons: Dissociation, species separation, and nucleation
The accumulation and circulation of carbon and hydrogen contribute to the chemical evolution of ice giant planets. Species separation and diamond precipitation have been reported in carbon–hydrogen systems and have been verified by static and shock compression experiments. Nevertheless, the dynamic formation processes underlying these phenomena remain insufficiently understood. In combination with a deep learning model, we demonstrate that diamonds form through a three-step process involving dissociation, species separation, and nucleation processes. Under shock conditions of 125 GPa and 4590 K, hydrocarbons decompose to give hydrogen and low-molecular-weight alkanes (CH4 and C2H6), which escape from the carbon chains, resulting in C/H species separation. The remaining carbon atoms without C–H bonds accumulate and nucleate to form diamond crystals. The process of diamond growth is associated with a critical nucleus size at which the dynamic energy barrier plays a key role. These dynamic processes of diamond formation provide insight into the establishment of a model for the evolution of ice giant planets.
- Open Access
- 25 November 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 10464
Picosecond-scale heterogeneous melting of metals at extreme non-equilibrium states
Extreme electron-ion non-equilibrium states, generated by ultrafast laser excitation, lead to melting processes that are fundamentally different from those under conventional thermal equilibrium and remain not fully understood. Through neural network-enhanced multiscale simulations of tungsten and gold nanofilms, we identify electronic pressure relaxation as critical to heterogeneous phase transformations. This nonthermal expansion generates a density decrease that enable surface-initiated melting far below equilibrium melting temperatures, creating electronic pressure-driven solid-liquid interface propagation at a high speed of 2500 ms −1 —tenfold faster than that of thermal heterogeneous melting mechanisms. Simulated time-resolved X-ray diffraction signatures distinguish this nonthermal expansion from thermal expansion dynamics driven by thermoelastic stress. These results establish hot-electron-mediated lattice destabilization as a universal pathway for laser-induced structural transformations, providing new insights for interpreting time-resolved experiments and controlling laser-matter interactions.
- 25 November 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 10464
Open AccessPicosecond-scale heterogeneous melting of metals at extreme non-equilibrium states
Extreme electron-ion non-equilibrium states, generated by ultrafast laser excitation, lead to melting processes that are fundamentally different from those under conventional thermal equilibrium and remain not fully understood. Through neural network-enhanced multiscale simulations of tungsten and gold nanofilms, we identify electronic pressure relaxation as critical to heterogeneous phase transformations. This nonthermal expansion generates a density decrease that enable surface-initiated melting far below equilibrium melting temperatures, creating electronic pressure-driven solid-liquid interface propagation at a high speed of 2500 ms −1 —tenfold faster than that of thermal heterogeneous melting mechanisms. Simulated time-resolved X-ray diffraction signatures distinguish this nonthermal expansion from thermal expansion dynamics driven by thermoelastic stress. These results establish hot-electron-mediated lattice destabilization as a universal pathway for laser-induced structural transformations, providing new insights for interpreting time-resolved experiments and controlling laser-matter interactions.
- 24 September 2025
- Published by AIP Publishing in Matter and Radiation at Extremes
- Vol. 10 (6)
Investigating phase dynamics of materials under laser-induced extreme conditions
Shock compression driven by nanosecond-laser techniques generates extreme pressure and temperature conditions in materials, enabling the study of high-pressure phase transitions and the behavior of materials in extreme environments. These dynamic high-pressure states are relevant to a wide range of phenomena, including planetary formation, asteroid impacts, spacecraft shielding, and inertial confinement fusion. The integration of advanced X-ray diffraction experimental techniques, from laser-induced X-ray sources and X-ray free-electron lasers, and theoretical simulations has provided unprecedented insights into material behavior under extreme conditions. This perspective reviews recent advances in dynamic high-pressure research and the insights that they can provide, concentrating on dynamical phase transitions, metastable and transient states, the influence of crystal orientation, microstructural changes, and the kinetic mechanism of phase transitions across a variety of interdisciplinary fields.
- 24 September 2025
- Published by AIP Publishing in Matter and Radiation at Extremes
- Vol. 10 (6)
Investigating phase dynamics of materials under laser-induced extreme conditions
Shock compression driven by nanosecond-laser techniques generates extreme pressure and temperature conditions in materials, enabling the study of high-pressure phase transitions and the behavior of materials in extreme environments. These dynamic high-pressure states are relevant to a wide range of phenomena, including planetary formation, asteroid impacts, spacecraft shielding, and inertial confinement fusion. The integration of advanced X-ray diffraction experimental techniques, from laser-induced X-ray sources and X-ray free-electron lasers, and theoretical simulations has provided unprecedented insights into material behavior under extreme conditions. This perspective reviews recent advances in dynamic high-pressure research and the insights that they can provide, concentrating on dynamical phase transitions, metastable and transient states, the influence of crystal orientation, microstructural changes, and the kinetic mechanism of phase transitions across a variety of interdisciplinary fields.
- Open Access
- 7 November 2024
- Published by Springer Nature in npj Computational Materials
- Vol. 10 (1) , 1-10
High-speed and low-power molecular dynamics processing unit (MDPU) with ab initio accuracy
Molecular dynamics (MD) is an indispensable atomistic-scale computational tool widely-used in various disciplines. In the past decades, nearly all ab initio MD and machine-learning MD have been based on the general-purpose central/graphics processing units (CPU/GPU), which are well-known to suffer from their intrinsic “memory wall” and “power wall” bottlenecks. Consequently, nowadays MD calculations with ab initio accuracy are extremely time-consuming and power-consuming, imposing serious restrictions on the MD simulation size and duration. To solve this problem, here we propose a special-purpose MD processing unit (MDPU), which could reduce MD time and power consumption by about 103 times (109 times) compared to state-of-the-art machine-learning MD (ab initio MD) based on CPU/GPU, while keeping ab initio accuracy. With significantly-enhanced performance, the proposed MDPU may pave a way for the accurate atomistic-scale analysis of large-size and/or long-duration problems which were impossible/impractical to compute before.
- 7 November 2024
- Published by Springer Nature in npj Computational Materials
- Vol. 10 (1) , 1-10
Open AccessHigh-speed and low-power molecular dynamics processing unit (MDPU) with ab initio accuracy
Molecular dynamics (MD) is an indispensable atomistic-scale computational tool widely-used in various disciplines. In the past decades, nearly all ab initio MD and machine-learning MD have been based on the general-purpose central/graphics processing units (CPU/GPU), which are well-known to suffer from their intrinsic “memory wall” and “power wall” bottlenecks. Consequently, nowadays MD calculations with ab initio accuracy are extremely time-consuming and power-consuming, imposing serious restrictions on the MD simulation size and duration. To solve this problem, here we propose a special-purpose MD processing unit (MDPU), which could reduce MD time and power consumption by about 103 times (109 times) compared to state-of-the-art machine-learning MD (ab initio MD) based on CPU/GPU, while keeping ab initio accuracy. With significantly-enhanced performance, the proposed MDPU may pave a way for the accurate atomistic-scale analysis of large-size and/or long-duration problems which were impossible/impractical to compute before.