KXKe Xu

Ke Xu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Ke Xu has a total of 106 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A polyimide-based thermal bridge

    by Haichang Guo, Zhiyao Jiang, Guanghua Zeng, Weiheng Li, Dongliang Ding, Ke Xu, Xiaoliang Zeng, Yimin Yao, Dongyan Xu, Bai Song, et al.

    Thermal management across micro- to macro-scale critically depends on accurate characterization of intermediate-scale materials, such as individual fibers and thin films. Yet, measuring the thermal conductivity of a single fiber remains rather challenging. Here, we present a polyimide (PI)-based thermal bridge for accurate thermal transport measurement of single fibers and related microscale samples. By addressing key challenges in hybrid material integration and flexible substrate interconnection, our approach retains the mechanical flexibility of PI while achieving compatibility with conventional microfabrication processes. The PI suspension beams exhibit thermal conductance of ∼12 μW K−1, within one order of magnitude of similar devices made of SiNx, while their vertical stiffness is more than two orders of magnitude higher, providing exceptional stability for sample manipulation. Calibration using a standard 20 μm diameter platinum wire yields a measured thermal conductivity of 70.6 ± 3.6 W m−1 K−1, within 5% of the literature value (71.7 W m−1 K−1). With its robust mechanical design and reliable thermal performance, this PI-based thermal bridge offers a versatile platform for accurate characterization of individual fibers and films and holds great promise for advancing thermal management research.

  • Open Access

    CGNEP‐MB‐pol: A Single‐Site Coarse‐Grained Machine Learning Potential for Water

    by Ke Xu, Fuyin Yin, Yue Zhang, Ting Liang, Zherui Chen, Chao Zhang, Jibao Lu, Jianbin Xu

    Coarse‐grained (CG) molecular dynamics (MD) can greatly extend the accessible time and length scales for water, provided that the reduced model captures key structural, dynamical, and thermodynamic properties. Here, we introduce a CG machine learning potential (MLP) for water, named CGNEP‐MB‐pol, which integrates a one‐molecule to one‐bead mapping with the neuroevolution potential (NEP) framework and MB‐pol reference data, thereby aiming to alleviate, within the liquid‐water and ice‐Ih states examined here, the state dependence and limited transferability that often constrain conventional CG models. Without altering the NEP descriptor or network architecture, the model is trained on CG coordinates and labels derived from atomistic configurations through a two‐step dataset refinement strategy, combining multi‐state training based on forces, energies, and virials. CGNEP‐MB‐pol remains compatible with the GPUMD + NEP workflow while delivering a substantial speedup over its all‐atom counterpart. It reproduces mapped reference forces, radial distribution functions of liquid water and ice, and the overall temperature dependence of self‐diffusion and viscosity. Two‐phase simulations capture ice growth, melting, and solid–liquid coexistence, yielding a melting temperature of approximately 274 K. This study establishes a practical route for constructing CG‐MLPs from high‐fidelity atomistic reference models.

  • Top Cited Papers
    Open Access

    qNEP: A Highly Efficient Neuroevolution Potential with Dynamic Charges for Large-Scale Atomistic Simulations

    by Zheyong Fan, Benrui Tang, Esmée Berger, Ethan Berger, Erik Fransson, Ke Xu, Zihan Yan, Zhoulin Liu, Zichen Song, Haikuan Dong, et al.

    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.


  • Accelerated phonon transport calculations for nanostructures: Combining neuroevolution potentials and compressed sensing

    by Fei Yin, Shixian Liu, Yiming Dong, A. A. Barinov, Ke Xu, V. I. Khvesyuk

    Thermal management in nanostructured devices necessitates the accurate and efficient prediction of phonon transport properties. However, solving the Boltzmann transport equation via first-principles calculations is often computationally prohibitive due to the extensive supercells required to model realistic nanostructures. In this work, we present an accelerated, automated workflow that synergizes neuroevolution potentials with compressed sensing techniques to efficiently extract high-order anharmonic force constants. We validate this approach using silicon thin films as a prototype, explicitly accounting for the complexities of surface reconstruction. Our results demonstrate that this framework achieves accuracy comparable to density functional theory while reducing the computational cost by several orders of magnitude. The workflow successfully reproduces phonon dispersion relations and captures the temperature- and size-dependent trends of lattice thermal conductivity, incorporating the critical contribution of inter-mode coherence. This methodology offers a scalable and robust solution for the high-throughput thermal characterization of low-dimensional materials.


  • Open Access

    PYSED: A tool for extracting kinetic-energy-weighted phonon dispersion and lifetime from molecular dynamics simulations

    by Ting Liang, Wenwu Jiang, Ke Xu, Hekai Bu, Zheyong Fan, Wengen Ouyang, Jianbin Xu

    Machine learning potential-driven molecular dynamics (MD) simulations have significantly enhanced the predictive accuracy of thermal transport properties across diverse materials. However, extracting phonon-mode-resolved insights from these simulations remains a critical challenge. Here, we introduce pysed, a Python-based package built on the spectral energy density (SED) method, designed to efficiently compute kinetic-energy-weighted phonon dispersion and extract phonon lifetime from large-scale MD simulation trajectories. By integrating high-accuracy machine-learned neuroevolution potential (NEP) models, we validate and showcase the effectiveness of the implemented SED method across systems of varying dimensionalities. Specifically, the NEP-driven MD-SED accurately reveals how phonon modes are affected by strain in carbon nanotubes, as well as by interlayer coupling strengths and the twist angles in two-dimensional molybdenum disulfide. For three-dimensional systems, the SED method effectively establishes the thermal transport regime diagram for metal-organic frameworks, distinguishing between particlelike and wavelike propagation regions. Moreover, using bulk silicon as an example, we show that phonon SED can efficiently capture quantum dynamics based on path-integral trajectories. The pysed package bridges MD simulations with detailed phonon-mode insights, delivering a robust tool for investigating thermal transport properties with detailed mechanisms across various materials.


  • Critical Size Transitions in Silicon Nanowires: Amorphization, Phonon Hydrodynamics, and Thermal Conductivity

    by Ke Xu, Yuan Li, Dongliang Ding, Ting Liang, Jianyang Wu, Jianbin Xu

    Understanding the intrinsic thermal transport properties of ultrathin semiconductor nanowires with varying diameters is crucial for the efficient thermal management of next-generation nanoelectronic devices. Here, we developed high-fidelity machine-learning potential (MLP) within the fourth-generation neuroevolution potential framework to elucidate the interplay between structural evolution, amorphous transition behavior, and thermal transport in silicon nanowires (SiNWs), resolving long-standing discrepancies between simulations and experiments. The structure of SiNWs below 1.1 nm in diameter undergoes a complete amorphous transformation, which originates from an amorphous surface structure of 5-6 atomic layers. We identify a nonmonotonic dependence of thermal conductivity on nanowire diameter due to competition between N (Normal) and U (Umklapp) phonon scattering processes. At frequencies <1 THz, N-process scattering rates exceed U-process rates by 3 orders of magnitude in ultrafine SiNWs, enabling fluid-like phonon transport. This study underscores the transformative potential of high-fidelity MLP in unraveling complex nanoscale material behaviors.


  • Top Cited Papers
    Open Access

    Advances in modeling complex materials: The rise of neuroevolution potentials

    by Penghua Ying, Cheng Qian, Rui Zhao, Yanzhou Wang, Ke Xu, Feng Ding, Shunda Chen, Zheyong Fan

    Interatomic potentials are essential for driving molecular dynamics (MD) simulations, directly impacting the reliability of predictions regarding the physical and chemical properties of materials. In recent years, machine-learned potentials (MLPs), trained against first-principles calculations, have become a new paradigm in materials modeling as they provide a desirable balance between accuracy and computational cost. The neuroevolution potential (NEP) approach, implemented in the open-source GPUMD software, has emerged as a promising machine-learned potential, exhibiting impressive accuracy and exceptional computational efficiency. This review provides a comprehensive discussion on the methodological and practical aspects of the NEP approach, along with a detailed comparison with other representative state-of-the-art MLP approaches in terms of training accuracy, property prediction, and computational efficiency. We also demonstrate the application of the NEP approach to perform accurate and efficient MD simulations, addressing complex challenges that traditional force fields typically cannot tackle. Key examples include structural properties of liquid and amorphous materials, chemical order in complex alloy systems, phase transitions, surface reconstruction, material growth, primary radiation damage, fracture in two-dimensional materials, nanoscale tribology, and mechanical behavior of compositionally complex alloys under various mechanical loadings. This review concludes with a summary and perspectives on future extensions to further advance this rapidly evolving field.


  • Top Cited Papers

    Highly efficient path-integral molecular dynamics simulations with GPUMD using neuroevolution potentials: Case studies on thermal properties of materials

    by Penghua Ying, Wenjiang Zhou, Lucas Svensson, Esmée Berger, Erik Fransson, Fredrik Eriksson, Ke Xu, Ting Liang, Jianbin Xu, Bai Song, et al.

    Path-integral molecular dynamics (PIMD) simulations are crucial for accurately capturing nuclear quantum effects in materials. However, their computational intensity often makes it challenging to address potential finite-size effects. Here, we present a specialized graphics processing units (GPUs) implementation of PIMD methods, including ring-polymer molecular dynamics (RPMD) and thermostatted ring-polymer molecular dynamics (TRPMD), into the open-source Graphics Processing Units Molecular Dynamics (GPUMD) package, combined with highly accurate and efficient machine-learned neuroevolution potential (NEP) models. This approach achieves almost the accuracy of first-principles calculations with the computational efficiency of empirical potentials, enabling large-scale atomistic simulations that incorporate nuclear quantum effects, effectively overcoming finite-size limitations at a relatively affordable computational cost. We validate and demonstrate the efficacy of the combined NEP-PIMD approach by examining various thermal properties of diverse materials, including lithium hydride (LiH), three porous metal–organic frameworks (MOFs), liquid water, and elemental aluminum. For LiH, our NEP-PIMD simulations successfully capture the isotope effect, reproducing the experimentally observed dependence of the lattice parameter on the reduced mass. For MOFs, our results reveal that achieving good agreement with experimental data requires consideration of both nuclear quantum effects and dispersive interactions. For water, our PIMD simulations capture the significant impact of nuclear quantum effects on its microscopic structure. For aluminum, the TRPMD method effectively captures thermal expansion and phonon properties, aligning well with quantum mechanical predictions. This efficient GPU-accelerated NEP-PIMD implementation in the GPUMD package provides an alternative, accessible, accurate, and scalable tool for exploring complex material properties influenced by nuclear quantum effects, with potential applications across a broad range of materials.