Article
Materials Science, Composites
Daili Feng, Zihao Zhao, Xinxin Zhang, Yanhui Feng
Summary: Sugar alcohol phase change materials (SAPCMs) have attracted attention in thermal storage due to their large energy storage density, wide melting temperature range, and excellent thermal cycling stability. This study investigates the use of carbon-based nanoadditives to enhance the thermal properties of SAPCMs, specifically mannitol and erythritol. Experimental and molecular dynamics methods were used to analyze the thermal conductivity improvement and reduction in undercooling and latent heat difference. The study also highlights the impact of interfacial thermal resistance on the effectiveness of carbon nanoadditives.
COMPOSITES SCIENCE AND TECHNOLOGY
(2023)
Article
Polymer Science
Victor M. Nazarychev, Sergey V. Lyulin
Summary: In recent decades, there has been a growing interest in improving the thermal conductivity of polymers for the development of new thermal interface materials (TIM) for electronic and electrical devices. In this study, the effects of uniaxial deformation on the thermal conductivity of thermoplastic polyimides were examined using atomistic computer simulations. The results showed that the thermal conductivity coefficient is anisotropic in different directions and significantly increases in the direction parallel to the deformation.
Article
Chemistry, Physical
Zheyong Fan, Yanzhou Wang, Penghua Ying, Keke Song, Junjie Wang, Yong Wang, Zezhu Zeng, Ke Xu, Eric Lindgren, J. Magnus Rahm, Alexander J. Gabourie, Jiahui Liu, Haikuan Dong, Jianyang Wu, Yue Chen, Zheng Zhong, Jian Sun, Paul Erhart, Yanjing Su, Tapio Ala-Nissila
Summary: We present the latest advancements in machine-learned potentials based on the neuroevolution potential framework and their implementation in the open-source package GPUMD. The accuracy of the models is improved by enhancing the radial and angular descriptors, and their efficient implementation in graphics processing units is described. Comparisons with state-of-the-art MLPs demonstrate the superior accuracy and computational efficiency of the NEP approach. Additionally, an active-learning scheme based on the latent space of a pre-trained NEP model is proposed, and three Python packages are introduced for integrating GPUMD into Python workflows.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Zequn Wang, Meng An, Kunliang Zhang, Dongsheng Chen, Xuhui Sun, Xin Wang, Yuejin Yuan, Junwen Shi, Jianchun Wu
Summary: Developing semiconducting materials with suitable band gap for next-generation electronic devices has become a hot research topic. The newly synthesized monolayer carbon boron (C3B) material, with outstanding electronic properties and indirect bandgap, shows promise as an alternative to graphene. This study investigates the thermal conductivity of monolayer C3B and graphene, as well as interfacial phonon transport across heterointerfaces. The reduced thermal conductivity of C3B compared to graphene is attributed to the decreased phonon group velocity and phonon relaxation time caused by the introduction of periodic boron atoms in the C3B sheet. The study also discusses the influences of temperature and strain on the thermal conductivity of C3B, and analyzes the thermal transport across graphene|C3B heterointerfaces at different temperatures. These findings provide insights for understanding the thermal transport properties of two-dimensional materials and have potential implications for the design of thermal management and thermoelectric materials.
SURFACES AND INTERFACES
(2023)
Article
Physics, Fluids & Plasmas
Anthony Saliou, Philippe Jarry, Noel Jakse
Summary: By using large-scale computer simulations and supervised learning, the relationship between excess entropy and diffusion was analyzed for the Lennard-Jones potential. The results showed a strong correlation between excess entropy and the potential energy landscape (PEL). The exponential law in liquids was found to be linked with the landscape-influenced regime of the PEL, while the power-law behavior in fluid-like systems corresponded to the free diffusion regime.
Article
Materials Science, Multidisciplinary
Simone Perego, Daniele Dragoni, Silvia Gabardi, Davide Campi, Marco Bernasconi
Summary: Models of the amorphous phase of GeTe compound were generated using molecular dynamics simulations, by depositing individual atoms on a substrate. This material has applications in phase change memories. The simulations, using up to 8000 atoms, utilized a machine learning potential developed by the group. The structural properties of the films were found to be similar to those generated by quenching from the melt, with some differences in bonding and atomic geometries.
PHYSICA STATUS SOLIDI-RAPID RESEARCH LETTERS
(2023)
Article
Thermodynamics
L. Klochko, J. Noel, N. R. Sgreva, S. Leclerc, C. Metivier, D. Lacroix, M. Isaiev
Summary: Investigating the properties of phase change materials (PCMs) is crucial for improving heat storage systems and thermal regulation devices. This research combines molecular simulations and experimental measurements to gain a better understanding of the thermophysical and rheological characteristics of PCMs, with a focus on n-hexadecane. The results are compared to experimental data to validate the simulations.
INTERNATIONAL COMMUNICATIONS IN HEAT AND MASS TRANSFER
(2022)
Article
Polymer Science
Yaguang Sun, Kaiwei Wan, Wenhui Shen, Jianxin He, Tong Zhou, Hui Wang, Hua Yang, Xinghua Shi
Summary: Recycling and reprocessing of conventional thermosetting polymers have gained attention due to environmental concerns. This study focuses on covalent adaptable networks (CANs) which incorporate functional groups capable of reversible exchange reactions into polymer networks, altering the topology arrangement and achieving stress relaxation. The researchers developed a machine-learning force field to describe the exchange reactions of polyimine CANs and provided insights into reaction mechanisms and energy profiles through enhanced sampling methods.
Article
Chemistry, Physical
Ali K. Shargh, Niaz Abdolrahim
Summary: A deep learning framework is used to design nanoporous silicon nitride (NPN) membranes with improved or prescribed strength values. The predictions of the framework are validated by physics-based simulations. The optimized microstructural heterogeneity suggested by the framework leads to strength improvement of NPN membranes compared to conventional membranes with homogenous microstructures.
NPJ COMPUTATIONAL MATERIALS
(2023)
Article
Chemistry, Physical
Mohammed Guerboub, Steve Dave Wansi Wendji, Carlo Massobrio, Assil Bouzid, Mauro Boero, Guido Ori, Evelyne Martin
Summary: It is found that the thermal conductivity of amorphous systems is not sensitive to the details of the atomic structure, despite the difference in local structure between different models. The behavior is rationalized in terms of extended vibrational modes.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Materials Science, Multidisciplinary
Zheyong Fan, Zezhu Zeng, Cunzhi Zhang, Yanzhou Wang, Keke Song, Haikuan Dong, Yue Chen, Tapio Ala Nissila
Summary: In this research, a neuroevolution-potential framework for generating neural network-based machine-learning potentials trained with an evolutionary strategy is developed. The atomic environment descriptor is constructed using Chebyshev and Legendre polynomials. The NEP method, implemented in GPUMD package, achieves high computational speed and provides per-atom heat current information.
Article
Mechanics
S. M. Kazem Manzoorolajdad, Hossein Hamzehpour, Jalal Sarabadani
Summary: The electrokinetic transport in a neutral system was studied by investigating an aqueous NaCl solution confined in a nanochannel with two similar parallel phosphorene walls. Different black, blue, red, and green phosphorene allotropes were studied under the presence of an external electric field in the x and y directions. The results showed the increase in Stern layer thickness with negative electric surface charge density (ESCD) and roughness ratio, as well as the occurrence of Debye-Huckel regime, intermediate regime, and flow reversal with increasing absolute value of negative ESCD.
Article
Thermodynamics
Victor M. Nazarychev, Artyom D. Glova, Igor V. Volgin, Sergey Larin, Alexey Lyulin, Sergey Lyulin, Andrey A. Gurtovenko
Summary: This study investigates the effects of different atomistic force field models on the calculation of thermal conductivity of n-eicosane samples using molecular dynamics simulations. The results show significant differences in the performance of different models in crystalline and liquid states, highlighting the challenging task of selecting an appropriate model for thermal conductivity calculations.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2021)
Article
Chemistry, Physical
Kim Lopez-Guell, Nicolas Forrer, Xavier Cartoixa, Ilaria Zardo, Riccardo Rurali
Summary: Crystal phase engineering allows for the manipulation of phonon transport in periodic nanostructures, such as twinning superlattices. This study focuses on GaAs and InAs twinning superlattices and identifies two distinct transport regimes, one where each interface behaves as an independent scatterer and another where a segment with closely spaced interfaces acts as a metamaterial with its own thermal properties.
JOURNAL OF PHYSICAL CHEMISTRY C
(2022)
Article
Chemistry, Physical
Kim Lopez-Guell, Nicolas Forrer, Xavier Cartoixa, Ilaria Zardo, Riccardo Rurali
Summary: Crystal phase engineering can alter phonon transport, and twinning superlattices exhibit two transportation mechanisms depending on interface number and spacing.
JOURNAL OF PHYSICAL CHEMISTRY C
(2022)
Article
Materials Science, Multidisciplinary
Dario Baratella, Daniele Dragoni, Marco Bernasconi
Summary: Thermoelectric effects are crucial in phase-change memories for programming operations. The electrothermal modeling of the device requires knowledge of transport coefficients, such as electrical and thermal conductivities, and the Seebeck coefficient at various temperatures. In this study, using density functional molecular dynamics simulations, the transport coefficients of liquid Ge2Sb2Te5 were calculated at temperatures slightly above its melting point.
PHYSICA STATUS SOLIDI-RAPID RESEARCH LETTERS
(2022)
Article
Chemistry, Multidisciplinary
Stefano Cecchi, Inaki Lopez Garcia, Antonio M. Mio, Eugenio Zallo, Omar Abou El Kheir, Raffaella Calarco, Marco Bernasconi, Giuseppe Nicotra, Stefania M. S. Privitera
Summary: This study investigates the composition of Ge-rich Ge2Sb2Te5 alloy and finds that it is less prone to decompose and segregate germanium.
Article
Chemistry, Physical
Marius Herbold, Joerg Behler
Summary: In recent years, various machine learning potentials (MLPs) have been developed to accurately represent high-dimensional potential-energy surfaces (PESs), and most of these MLPs rely on atomic energy contributions and forces derived from local chemical environments. This study proposes a method to determine structurally converged molecular fragments based on Hessian analysis, which can provide reliable atomic forces. The method serves as a locality test and allows estimation of the importance of long-range interactions.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Letter
Physics, Multidisciplinary
Giuseppe Barbalinardo, Zekun Chen, Haikuan Dong, Zheyong Fan, Davide Donadio
PHYSICAL REVIEW LETTERS
(2022)
Article
Materials Science, Multidisciplinary
Simone Perego, Daniele Dragoni, Silvia Gabardi, Davide Campi, Marco Bernasconi
Summary: Models of the amorphous phase of GeTe compound were generated using molecular dynamics simulations, by depositing individual atoms on a substrate. This material has applications in phase change memories. The simulations, using up to 8000 atoms, utilized a machine learning potential developed by the group. The structural properties of the films were found to be similar to those generated by quenching from the melt, with some differences in bonding and atomic geometries.
PHYSICA STATUS SOLIDI-RAPID RESEARCH LETTERS
(2023)
Article
Physics, Multidisciplinary
Janos Daru, Harald Forbert, Joerg Behler, Dominik Marx
Summary: The study introduces a framework to extend the accuracy of coupled cluster theory from small systems to large condensed phase systems using high-dimensional neural network potentials. This automated approach enables high-quality coupled cluster molecular dynamics and demonstrates its application in liquid water.
PHYSICAL REVIEW LETTERS
(2022)
Article
Chemistry, Physical
Trent Barnard, Steven Tseng, James P. Darby, Albert P. Bartok, Anders Broo, Gabriele C. Sosso
Summary: SOAP_GAS is a computational tool that uses genetic algorithms to optimize parameters for SOAP descriptors, leading to enhanced predictive capabilities.
MOLECULAR SYSTEMS DESIGN & ENGINEERING
(2023)
Article
Chemistry, Physical
Tsz Wai Ko, Jonas A. Finkler, Stefan Goedecker, Joerg Behler
Summary: Significant progress has been made in the development of machine learning potentials (MLPs) for atomistic simulations. Incorporating the electrostatic potential arising from the charge distribution in the atomic environments as a descriptor significantly improves the quality and transferability of MLPs. An electrostatically embedded fourth-generation high-dimensional neural network potential (ee4G-HDNNP) augmented by pairwise interactions demonstrates impressive capabilities for NaCl as a benchmark system.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Hiroki Chikuma, Gentoku Takasao, Toru Wada, Patchanee Chammingkwan, Joerg Behler, Toshiaki Taniike
Summary: We have successfully accelerated the determination of catalyst nanostructures by using a high-dimensional neural network potential (HDNNP). The training set's structural diversity is crucial for building HDNNPs for MgCl2/TiCl4 clusters with computationally tractable sizes. The resulting HDNNPs significantly accelerated the structure determination and yielded consistent results with density functional theory (DFT). Additionally, we developed a multistep adaptive procedure to construct HDNNP for MgCl2/TiCl4 clusters consistent in size and composition with prior DFT results.
JOURNAL OF PHYSICAL CHEMISTRY C
(2023)
Review
Physics, Multidisciplinary
Giuliano Benenti, Davide Donadio, Stefano Lepri, Roberto Livi
Summary: Energy transfer in small nano-sized systems can be very different from that in their macroscopic counterparts due to reduced dimensionality, interaction with surfaces, disorder, and large fluctuations. We provide an overview of recent advances in understanding non-diffusive heat transfer in these systems through nonequilibrium statistical mechanics and atomistic simulations. The underlying basic properties leading to violations of standard diffusive heat transport and the effects of long-range interaction and integrability on non-diffusive transport are discussed. We also explore the applications of these features in thermal management, rectification, and improving energy conversion efficiency.
RIVISTA DEL NUOVO CIMENTO
(2023)
Article
Chemistry, Physical
Alea Miako Tokita, Joerg Behler
Summary: The introduction of modern Machine Learning Potentials (MLPs) has revolutionized atomistic simulations, allowing large-scale simulations of extended systems. MLPs can achieve the accuracy of electronic structure calculations with proper training and validation. This Tutorial outlines the key steps for training reliable MLPs and provides a general example.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Multidisciplinary
Regine Herbst-Irmer, Xiaobai Wang, Laura Haberstock, Ingo Koehne, Rainer Oswald, Joerg Behler, Dietmar Stalke
Summary: Phosphorus exists in several different allotropes, with white phosphorus being the most important. In addition to the known three polymorphs, a new polymorph, delta-P-4, has been discovered with a structure similar to alpha-Mn but distinct from alpha-P-4. DFT calculations show that delta-P-4 is metastable at a similar energy level to gamma-P-4.
Article
Chemistry, Physical
Marius Herbold, Joerg Behler
Summary: Machine learning potentials (MLP) enable accurate atomistic simulations at a fraction of the cost of electronic structure calculations. Most MLPs construct the potential energy as a sum of atomic energies given by local chemical environments. Training MLPs often requires computationally demanding large systems to obtain reference forces, but this work demonstrates that small density-functional theory calculations of molecular fragments can be used to learn transferable forces for extended systems, illustrated using high-dimensional neural network potentials for metal-organic frameworks.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Dilshana Shanavas Rasheeda, Alberto Martin Santa Daria, Benjamin Schroeder, Edit Matyus, Joerg Behler
Summary: In recent years, the potentials of machine learning for atomistic simulations have gained considerable attention in chemistry and materials science. Although many new approaches have been developed, it is still challenging to assess the reliability of modern machine learning potentials in reproducing the subtle details of multi-dimensional potential-energy surfaces for large condensed systems. On the other hand, there is a lack of investigations on moderately sized systems that enable the application of tools for thorough and systematic quality-control.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2022)
