The analysis of experimental results with Python often requires writing many code scripts which all need access to the same set of functions. In a common field of research, this set will be nearly the same for many users. The qspec Python package was developed to provide functions for physical formulas, simulations and data analysis routines widely used in laser spectroscopy and related fields. Most functions are compatible with numpy arrays, enabling fast calculations with large samples of data. A multidimensional linear regression algorithm enables a King plot analyses over multiple atomic transitions. A modular framework for constructing lineshape models can be used to fit large sets of spectroscopy data. A simulation module within the package provides user-friendly methods to simulate the coherent time-evolution of atoms in electro-magnetic fields without the need to explicitly derive a Hamiltonian.
使用 Python 分析实验结果通常需要编写许多代码脚本，这些脚本都需要访问同一组函数。在一个共同的研究领域，这套函数对许多用户来说几乎都是一样的。qspecPython 软件包的开发目的是为激光光谱学及相关领域广泛使用的物理公式、模拟和数据分析例程提供函数。大多数函数都与 numpy 数组兼容，可以快速计算大量数据样本。通过多维线性回归算法，可以对多个原子跃迁进行 King plot 分析。构建线形模型的模块化框架可用于拟合大量光谱数据集。软件包中的一个模拟模块提供了用户友好的方法，用于模拟原子在电磁场中的相干时间演变，而无需明确推导哈密顿。
{"title":"The qspec Python package: A physics toolbox for laser spectroscopy","authors":"Patrick Müller, Wilfried Nörtershäuser","doi":"arxiv-2409.01417","DOIUrl":"https://doi.org/arxiv-2409.01417","url":null,"abstract":"The analysis of experimental results with Python often requires writing many\u0000code scripts which all need access to the same set of functions. In a common\u0000field of research, this set will be nearly the same for many users. The qspec\u0000Python package was developed to provide functions for physical formulas,\u0000simulations and data analysis routines widely used in laser spectroscopy and\u0000related fields. Most functions are compatible with numpy arrays, enabling fast\u0000calculations with large samples of data. A multidimensional linear regression\u0000algorithm enables a King plot analyses over multiple atomic transitions. A\u0000modular framework for constructing lineshape models can be used to fit large\u0000sets of spectroscopy data. A simulation module within the package provides\u0000user-friendly methods to simulate the coherent time-evolution of atoms in\u0000electro-magnetic fields without the need to explicitly derive a Hamiltonian.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204116","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Machine learning potentials (MLPs) have become an indispensable tool in large-scale atomistic simulations because of their ability to reproduce ab initio potential energy surfaces (PESs) very accurately at a fraction of computational cost. For computational efficiency, the training data for most MLPs today are computed using relatively cheap density functional theory (DFT) methods such as the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) functional. Meta-GGAs such as the recently developed strongly constrained and appropriately normed (SCAN) functional have been shown to yield significantly improved descriptions of atomic interactions for diversely bonded systems, but their higher computational cost remains an impediment to their use in MLP development. In this work, we outline a data-efficient multi-fidelity approach to constructing Materials 3-body Graph Network (M3GNet) interatomic potentials that integrate different levels of theory within a single model. Using silicon and water as examples, we show that a multi-fidelity M3GNet model trained on a combined dataset of low-fidelity GGA calculations with 10% of high-fidelity SCAN calculations can achieve accuracies comparable to a single-fidelity M3GNet model trained on a dataset comprising 8x the number of SCAN calculations. This work paves the way for the development of high-fidelity MLPs in a cost-effective manner by leveraging existing low-fidelity datasets.
{"title":"Data-Efficient Construction of High-Fidelity Graph Deep Learning Interatomic Potentials","authors":"Tsz Wai Ko, Shyue Ping Ong","doi":"arxiv-2409.00957","DOIUrl":"https://doi.org/arxiv-2409.00957","url":null,"abstract":"Machine learning potentials (MLPs) have become an indispensable tool in\u0000large-scale atomistic simulations because of their ability to reproduce ab\u0000initio potential energy surfaces (PESs) very accurately at a fraction of\u0000computational cost. For computational efficiency, the training data for most\u0000MLPs today are computed using relatively cheap density functional theory (DFT)\u0000methods such as the Perdew-Burke-Ernzerhof (PBE) generalized gradient\u0000approximation (GGA) functional. Meta-GGAs such as the recently developed\u0000strongly constrained and appropriately normed (SCAN) functional have been shown\u0000to yield significantly improved descriptions of atomic interactions for\u0000diversely bonded systems, but their higher computational cost remains an\u0000impediment to their use in MLP development. In this work, we outline a\u0000data-efficient multi-fidelity approach to constructing Materials 3-body Graph\u0000Network (M3GNet) interatomic potentials that integrate different levels of\u0000theory within a single model. Using silicon and water as examples, we show that\u0000a multi-fidelity M3GNet model trained on a combined dataset of low-fidelity GGA\u0000calculations with 10% of high-fidelity SCAN calculations can achieve accuracies\u0000comparable to a single-fidelity M3GNet model trained on a dataset comprising 8x\u0000the number of SCAN calculations. This work paves the way for the development of\u0000high-fidelity MLPs in a cost-effective manner by leveraging existing\u0000low-fidelity datasets.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204117","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In therapeutic focused ultrasound (FUS), such as thermal ablation and hyperthermia, effective acousto-thermal manipulation requires precise targeting of complex geometries, sound wave propagation through irregular structures and selective focusing at specific depths. Acoustic holographic lenses (AHLs) provide a distinctive capability to shape acoustic fields into precise, complex and multifocal FUS-thermal patterns. Acknowledging the under-explored potential of AHLs in shaping ultrasound-induced heating, this study introduces a roadmap for acousto-thermal modeling in the design of AHLs. Three primary modeling approaches are studied and contrasted using four distinct shape groups for the imposed target field. They include pressure-based (BSC-TR and ITER-TR), temperature-based (IHTO-TR), and machine learning (ML)-based (GaN and Feat-GAN) methods. New metrics including image quality, thermal efficiency, control, and computational time are introduced. The importance of evaluating target pattern complexity, thermal and pressure requirements, and computational resources is highlighted for selecting the appropriate methods. For lightly heterogeneous media and targets with lower pattern complexity, BSC-TR combined with error diffusion algorithms provides an effective solution. As pattern complexity increases, ITER-TR becomes more suitable, enabling optimization through iterative forward and backward propagations controlled by different error metrics. IHTO-TR is recommended for highly heterogeneous media, particularly in applications requiring thermal control and precise heat deposition. GaN is ideal for rapid solutions that account for acousto-thermal effects, especially when model parameters and boundary conditions remain constant. In contrast, Feat-GaN is effective for moderately complex shape groups and applications where model parameters must be adjusted.
{"title":"A Roadmap to Holographic Focused Ultrasound Approaches to Generate Thermal Patterns","authors":"Ceren Cengiz, Zekeriya Ender Eger, Pinar Acar, Wynn Legon, Shima Shahab","doi":"arxiv-2409.01323","DOIUrl":"https://doi.org/arxiv-2409.01323","url":null,"abstract":"In therapeutic focused ultrasound (FUS), such as thermal ablation and\u0000hyperthermia, effective acousto-thermal manipulation requires precise targeting\u0000of complex geometries, sound wave propagation through irregular structures and\u0000selective focusing at specific depths. Acoustic holographic lenses (AHLs)\u0000provide a distinctive capability to shape acoustic fields into precise, complex\u0000and multifocal FUS-thermal patterns. Acknowledging the under-explored potential\u0000of AHLs in shaping ultrasound-induced heating, this study introduces a roadmap\u0000for acousto-thermal modeling in the design of AHLs. Three primary modeling\u0000approaches are studied and contrasted using four distinct shape groups for the\u0000imposed target field. They include pressure-based (BSC-TR and ITER-TR),\u0000temperature-based (IHTO-TR), and machine learning (ML)-based (GaN and Feat-GAN)\u0000methods. New metrics including image quality, thermal efficiency, control, and\u0000computational time are introduced. The importance of evaluating target pattern\u0000complexity, thermal and pressure requirements, and computational resources is\u0000highlighted for selecting the appropriate methods. For lightly heterogeneous\u0000media and targets with lower pattern complexity, BSC-TR combined with error\u0000diffusion algorithms provides an effective solution. As pattern complexity\u0000increases, ITER-TR becomes more suitable, enabling optimization through\u0000iterative forward and backward propagations controlled by different error\u0000metrics. IHTO-TR is recommended for highly heterogeneous media, particularly in\u0000applications requiring thermal control and precise heat deposition. GaN is\u0000ideal for rapid solutions that account for acousto-thermal effects, especially\u0000when model parameters and boundary conditions remain constant. In contrast,\u0000Feat-GaN is effective for moderately complex shape groups and applications\u0000where model parameters must be adjusted.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"22 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Daoqi Liu, Tao Shan, Maokun Li, Fan Yang, Shenheng Xu
In this work, we propose a deep learning-based imaging method for addressing the multi-frequency electromagnetic (EM) inverse scattering problem (ISP). By combining deep learning technology with EM physical laws, we have successfully developed a multi-frequency neural Born iterative method (NeuralBIM), guided by the principles of the single-frequency NeuralBIM. This method integrates multitask learning techniques with NeuralBIM's efficient iterative inversion process to construct a robust multi-frequency Born iterative inversion model. During training, the model employs a multitask learning approach guided by homoscedastic uncertainty to adaptively allocate the weights of each frequency's data. Additionally, an unsupervised learning method, constrained by the physical laws of ISP, is used to train the multi-frequency NeuralBIM model, eliminating the need for contrast and total field data. The effectiveness of the multi-frequency NeuralBIM is validated through synthetic and experimental data, demonstrating improvements in accuracy and computational efficiency for solving ISP. Moreover, this method exhibits strong generalization capabilities and noise resistance. The multi-frequency NeuralBIM method explores a novel inversion method for multi-frequency EM data and provides an effective solution for the electromagnetic ISP of multi-frequency data.
{"title":"Multi-frequency Neural Born Iterative Method for Solving 2-D Inverse Scattering Problems","authors":"Daoqi Liu, Tao Shan, Maokun Li, Fan Yang, Shenheng Xu","doi":"arxiv-2409.01315","DOIUrl":"https://doi.org/arxiv-2409.01315","url":null,"abstract":"In this work, we propose a deep learning-based imaging method for addressing\u0000the multi-frequency electromagnetic (EM) inverse scattering problem (ISP). By\u0000combining deep learning technology with EM physical laws, we have successfully\u0000developed a multi-frequency neural Born iterative method (NeuralBIM), guided by\u0000the principles of the single-frequency NeuralBIM. This method integrates\u0000multitask learning techniques with NeuralBIM's efficient iterative inversion\u0000process to construct a robust multi-frequency Born iterative inversion model.\u0000During training, the model employs a multitask learning approach guided by\u0000homoscedastic uncertainty to adaptively allocate the weights of each\u0000frequency's data. Additionally, an unsupervised learning method, constrained by\u0000the physical laws of ISP, is used to train the multi-frequency NeuralBIM model,\u0000eliminating the need for contrast and total field data. The effectiveness of\u0000the multi-frequency NeuralBIM is validated through synthetic and experimental\u0000data, demonstrating improvements in accuracy and computational efficiency for\u0000solving ISP. Moreover, this method exhibits strong generalization capabilities\u0000and noise resistance. The multi-frequency NeuralBIM method explores a novel\u0000inversion method for multi-frequency EM data and provides an effective solution\u0000for the electromagnetic ISP of multi-frequency data.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"26 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Silvia Cipiccia, Wiebe Stolp, Luca Fardin, Ralf Ziesche, Ingo Manke, Matthieu Boone, Chris Armstrong, Joachim R. Binder, Nicole Bohn, Alessandro Olivo, Darren Batey
Ptychography is a scanning coherent diffraction imaging technique successfully applied in the electron, visible and x-ray regimes. One of the distinct features of ptychography with respect to other coherent diffraction techniques is its capability of dealing with partial spatial and temporal coherence via the reconstruction algorithm. Here we focus on the temporal and clarify the constraints which affect the energy resolution limits of the ptychographic algorithms. Based on this, we design and perform simulations for a broadband ptychography in the hard x-ray regime, which enables an energy resolution down to 1 eV. We benchmark the simulations against experimental ptychographic data of an NMC battery cathode material, attaining an energy resolution of 5 eV. We review the results, discuss the limitations, and provide guidelines for future broadband ptychography experiments, its prospective application for single acquisition x-ray absorption near edge structure imaging, magnetic dichroism imaging, and potential impact on achieving diffraction limited resolutions.
层析成像技术是一种扫描相干衍射成像技术，已成功应用于电子、可见光和 X 射线领域。与其他相干衍射技术相比，层析成像技术的一个显著特点是它能通过重建算法处理部分空间和时间相干。在此，我们将重点放在时间上，并阐明影响层析成像算法能量分辨率限制的约束条件。在此基础上，我们设计并模拟了硬 X 射线条件下的宽带层析成像技术，它能使能量分辨率低至 1 eV。我们根据 NMC 电池阴极材料的实验层析成像数据对模拟进行了基准测试，达到了 5 eV 的能量分辨率。我们回顾了这些结果，讨论了其局限性，并为未来的宽带层析成像实验提供了指导，其在单次获取 X 射线吸收近边缘结构成像、磁分色成像方面的应用前景，以及对实现衍射有限分辨率的潜在影响。
{"title":"Electronvolt energy resolution with broadband ptychography","authors":"Silvia Cipiccia, Wiebe Stolp, Luca Fardin, Ralf Ziesche, Ingo Manke, Matthieu Boone, Chris Armstrong, Joachim R. Binder, Nicole Bohn, Alessandro Olivo, Darren Batey","doi":"arxiv-2409.00703","DOIUrl":"https://doi.org/arxiv-2409.00703","url":null,"abstract":"Ptychography is a scanning coherent diffraction imaging technique\u0000successfully applied in the electron, visible and x-ray regimes. One of the\u0000distinct features of ptychography with respect to other coherent diffraction\u0000techniques is its capability of dealing with partial spatial and temporal\u0000coherence via the reconstruction algorithm. Here we focus on the temporal and\u0000clarify the constraints which affect the energy resolution limits of the\u0000ptychographic algorithms. Based on this, we design and perform simulations for\u0000a broadband ptychography in the hard x-ray regime, which enables an energy\u0000resolution down to 1 eV. We benchmark the simulations against experimental\u0000ptychographic data of an NMC battery cathode material, attaining an energy\u0000resolution of 5 eV. We review the results, discuss the limitations, and provide\u0000guidelines for future broadband ptychography experiments, its prospective\u0000application for single acquisition x-ray absorption near edge structure\u0000imaging, magnetic dichroism imaging, and potential impact on achieving\u0000diffraction limited resolutions.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204144","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The calculation of material phonon thermal conductivity from density functional theory calculations requires computationally expensive evaluation of anharmonic interatomic force constants and has remained a computational bottleneck in the high-throughput discovery of materials. In this work, we present a machine learning-assisted approach for the extraction of anharmonic force constants through local learning of the potential energy surface. We demonstrate our approach on a diverse collection of 220 ternary materials for which the total computational time for anharmonic force constants evaluation is reduced by more than an order of magnitude from 480,000 cpu-hours to less than 12,000 cpu-hours while preserving the thermal conductivity prediction accuracy to within 10%. Our approach removes a major hurdle in computational thermal conductivity evaluation and will pave the way forward for the high-throughput discovery of materials.
从密度函数理论计算中计算材料声子热导率需要对谐波原子间力常量进行计算昂贵的评估，这一直是高通量材料发现过程中的计算瓶颈。在这项工作中，我们提出了一种机器学习辅助方法，通过对势能面的局部学习来提取谐波力常数。我们在 220 种不同的三元材料上演示了我们的方法，评估非谐波力常数的总计算时间从 480,000 cpu 小时减少到不到 12,000 cpu 小时，减少了一个数量级以上，同时保持了 10%以内的热导率预测精度。我们的方法消除了计算热导评估中的一大障碍，将为高通量材料发现铺平道路。
{"title":"Accelerating Phonon Thermal Conductivity Prediction by an Order of Magnitude Through Machine Learning-Assisted Extraction of Anharmonic Force Constants","authors":"Yagyank Srivastava, Ankit Jain","doi":"arxiv-2409.00360","DOIUrl":"https://doi.org/arxiv-2409.00360","url":null,"abstract":"The calculation of material phonon thermal conductivity from density\u0000functional theory calculations requires computationally expensive evaluation of\u0000anharmonic interatomic force constants and has remained a computational\u0000bottleneck in the high-throughput discovery of materials. In this work, we\u0000present a machine learning-assisted approach for the extraction of anharmonic\u0000force constants through local learning of the potential energy surface. We\u0000demonstrate our approach on a diverse collection of 220 ternary materials for\u0000which the total computational time for anharmonic force constants evaluation is\u0000reduced by more than an order of magnitude from 480,000 cpu-hours to less than\u000012,000 cpu-hours while preserving the thermal conductivity prediction accuracy\u0000to within 10%. Our approach removes a major hurdle in computational thermal\u0000conductivity evaluation and will pave the way forward for the high-throughput\u0000discovery of materials.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"9 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xiuying Zhang, Linqiang Xu, Jing Lu, Zhaofu Zhang, Lei Shen
Quantum-mechanics-based transport simulation is of importance for the design of ultra-short channel field-effect transistors (FETs) with its capability of understanding the physical mechanism, while facing the primary challenge of the high computational intensity. Traditional machine learning is expected to accelerate the optimization of FET design, yet its application in this field is limited by the lack of both high-fidelity datasets and the integration of physical knowledge. Here, we introduced a physics-integrated neural network framework to predict the transport curves of sub-5-nm gate-all-around (GAA) FETs using an in-house developed high-fidelity database. The transport curves in the database are collected from literature and our first-principles calculations. Beyond silicon, we included indium arsenide, indium phosphide, and selenium nanowires with different structural phases as the FET channel materials. Then, we built a physical-knowledge-integrated hyper vector neural network (PHVNN), in which five new physical features were added into the inputs for prediction transport characteristics, achieving a sufficiently low mean absolute error of 0.39. In particular, ~98% of the current prediction residuals are within one order of magnitude. Using PHVNN, we efficiently screened out the symmetric p-type GAA FETs that possess the same figures of merit with the n-type ones, which are crucial for the fabrication of homogeneous CMOS circuits. Finally, our automatic differentiation analysis provides interpretable insights into the PHVNN, which highlights the important contributions of our new input parameters and improves the reliability of PHVNN. Our approach provides an effective method for rapidly screening appropriate GAA FETs with the prospect of accelerating the design process of next-generation electronic devices.
基于量子力学的输运模拟对于超短沟道场效应晶体管（FET）的设计非常重要，它能够理解物理机制，但同时也面临着计算强度高的主要挑战。传统的机器学习有望加速场效应晶体管的优化设计，但由于缺乏高保真数据集和物理知识的整合，机器学习在这一领域的应用受到了限制。在这里，我们引入了一个物理集成神经网络框架，利用内部开发的高保真数据库预测 5 纳米以下全栅极 (GAA) FET 的传输曲线。数据库中的传输曲线收集自文献和我们的第一原理计算。除了硅之外，我们还将不同结构相的砷化铟、磷化铟和硒纳米线作为场效应晶体管的沟道材料。然后，我们建立了一个物理知识集成超矢量神经网络（PHVNN），在预测传输特性的输入中加入了五个新的物理特征，取得了 0.39 的足够低的平均绝对误差。特别是，目前约 98% 的预测残差都在一个数量级之内。利用 PHVNN，我们有效地筛选出了对称 p 型 GAA 场效应晶体管，这些晶体管具有与当时型晶体管相同的性能指标，这对于制造同质 CMOS 电路至关重要。最后，我们的自动微分分析为 PHVNN 提供了可解释的见解，突出了新输入参数的重要贡献，提高了 PHVNN 的可靠性。我们的方法为快速筛选合适的 GAA FET 提供了一种有效的方法，有望加快下一代电子器件的设计进程。
{"title":"Physics-integrated Neural Network for Quantum Transport Prediction of Field-effect Transistor","authors":"Xiuying Zhang, Linqiang Xu, Jing Lu, Zhaofu Zhang, Lei Shen","doi":"arxiv-2408.17023","DOIUrl":"https://doi.org/arxiv-2408.17023","url":null,"abstract":"Quantum-mechanics-based transport simulation is of importance for the design\u0000of ultra-short channel field-effect transistors (FETs) with its capability of\u0000understanding the physical mechanism, while facing the primary challenge of the\u0000high computational intensity. Traditional machine learning is expected to\u0000accelerate the optimization of FET design, yet its application in this field is\u0000limited by the lack of both high-fidelity datasets and the integration of\u0000physical knowledge. Here, we introduced a physics-integrated neural network\u0000framework to predict the transport curves of sub-5-nm gate-all-around (GAA)\u0000FETs using an in-house developed high-fidelity database. The transport curves\u0000in the database are collected from literature and our first-principles\u0000calculations. Beyond silicon, we included indium arsenide, indium phosphide,\u0000and selenium nanowires with different structural phases as the FET channel\u0000materials. Then, we built a physical-knowledge-integrated hyper vector neural\u0000network (PHVNN), in which five new physical features were added into the inputs\u0000for prediction transport characteristics, achieving a sufficiently low mean\u0000absolute error of 0.39. In particular, ~98% of the current prediction residuals\u0000are within one order of magnitude. Using PHVNN, we efficiently screened out the\u0000symmetric p-type GAA FETs that possess the same figures of merit with the\u0000n-type ones, which are crucial for the fabrication of homogeneous CMOS\u0000circuits. Finally, our automatic differentiation analysis provides\u0000interpretable insights into the PHVNN, which highlights the important\u0000contributions of our new input parameters and improves the reliability of\u0000PHVNN. Our approach provides an effective method for rapidly screening\u0000appropriate GAA FETs with the prospect of accelerating the design process of\u0000next-generation electronic devices.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204146","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xuyang Guo, Kirk W. Madison, James L. Booth, Roman V. Krems
Thermal rate coefficients for some atomic collisions have been observed to be remarkably independent of the details of interatomic interactions at short range. This makes these rate coefficients universal functions of the long-range interaction parameters and masses, which was previously exploited to develop a self-defining atomic sensor for ambient pressure. Here, we employ rigorous quantum scattering calculations to examine the response of thermally averaged rate coefficients for atom-atom collisions to changes in the interaction potentials. We perform a comprehensive analysis of the universality, and the boundaries thereof, by treating the quantum scattering observables as probabilistic predictions determined by a distribution of interaction potentials. We show that there is a characteristic change of the resulting distributions of rate coefficients, separating light, few-electron atoms and heavy, polarizable atoms. We produce diagrams that illustrate the boundaries of the thermal collision universality at different temperatures and provide guidance for future experiments seeking to exploit the universality.
{"title":"Boundaries of universality of thermal collisions for atom-atom scattering","authors":"Xuyang Guo, Kirk W. Madison, James L. Booth, Roman V. Krems","doi":"arxiv-2409.00273","DOIUrl":"https://doi.org/arxiv-2409.00273","url":null,"abstract":"Thermal rate coefficients for some atomic collisions have been observed to be\u0000remarkably independent of the details of interatomic interactions at short\u0000range. This makes these rate coefficients universal functions of the long-range\u0000interaction parameters and masses, which was previously exploited to develop a\u0000self-defining atomic sensor for ambient pressure. Here, we employ rigorous\u0000quantum scattering calculations to examine the response of thermally averaged\u0000rate coefficients for atom-atom collisions to changes in the interaction\u0000potentials. We perform a comprehensive analysis of the universality, and the\u0000boundaries thereof, by treating the quantum scattering observables as\u0000probabilistic predictions determined by a distribution of interaction\u0000potentials. We show that there is a characteristic change of the resulting\u0000distributions of rate coefficients, separating light, few-electron atoms and\u0000heavy, polarizable atoms. We produce diagrams that illustrate the boundaries of\u0000the thermal collision universality at different temperatures and provide\u0000guidance for future experiments seeking to exploit the universality.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"12 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204137","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Nonlinear circuits serve as crucial carriers and physical models for investigating nonlinear dynamics and chaotic behavior, particularly in the simulation of biological neurons. In this study, Chua's circuit and Lorentz circuit are systematically explored for the first time through machine learning correlation algorithms. Specifically, the upgraded and optimized SINDy-PI model, which is based on neural network and symbolic regression algorithm, is utilized to learn the numerical results of attractors generated by these two nonlinear circuits. Through error analysis, we examine the effects of the precision of input data and the amount of data on the algorithmic model. The findings reveal that when the input data quantity and data precision fall within a certain range, the algorithm model can effectively recognize and restore the differential equation expressions corresponding to the two circuits. Additionally, we test the anti-interference ability of different circuits and the robustness of the algorithm by introducing noise into the test data. The results indicate that under the same noise disturbance, the Lorentz circuit has better noise resistance than Chua's circuit, providing a starting point for further studying the intrinsic properties and characteristics of different nonlinear circuits. The above results will not only offer a reference for the further study of nonlinear circuits and related systems using deep learning algorithms but also lay a preliminary theoretical foundation for the study of related physical problems and applications.
{"title":"Exploring Nonlinear System with Machine Learning: Chua and Lorentz Circuits Analyzed","authors":"Zhe Wang, Haixia Fan, Jiyuan Zhang, Xiao-Yun Wang","doi":"arxiv-2408.16972","DOIUrl":"https://doi.org/arxiv-2408.16972","url":null,"abstract":"Nonlinear circuits serve as crucial carriers and physical models for\u0000investigating nonlinear dynamics and chaotic behavior, particularly in the\u0000simulation of biological neurons. In this study, Chua's circuit and Lorentz\u0000circuit are systematically explored for the first time through machine learning\u0000correlation algorithms. Specifically, the upgraded and optimized SINDy-PI\u0000model, which is based on neural network and symbolic regression algorithm, is\u0000utilized to learn the numerical results of attractors generated by these two\u0000nonlinear circuits. Through error analysis, we examine the effects of the\u0000precision of input data and the amount of data on the algorithmic model. The\u0000findings reveal that when the input data quantity and data precision fall\u0000within a certain range, the algorithm model can effectively recognize and\u0000restore the differential equation expressions corresponding to the two\u0000circuits. Additionally, we test the anti-interference ability of different\u0000circuits and the robustness of the algorithm by introducing noise into the test\u0000data. The results indicate that under the same noise disturbance, the Lorentz\u0000circuit has better noise resistance than Chua's circuit, providing a starting\u0000point for further studying the intrinsic properties and characteristics of\u0000different nonlinear circuits. The above results will not only offer a reference\u0000for the further study of nonlinear circuits and related systems using deep\u0000learning algorithms but also lay a preliminary theoretical foundation for the\u0000study of related physical problems and applications.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"27 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204140","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The use of structurally complex lattice defects, such as functional groups, embedded nanoparticles, and nanopillars, to generate phonon scattering is a popular approach in phonon engineering for thermoelectric applications. However, the theoretical treatment of this scattering phenomenon remains a formidable challenge, especially with regards to the determination of the scattering cross sections and rates associated with such lattice defects. Using the extended Atomistic Green's Function (AGF) method, we describe how the numerically exact mode-resolved scattering cross section sigma can be computed for a phonon scattered by a single lattice defect. We illustrate the generality and utility of the AGF-based treatment with two examples. In the first example, we treat the isotopic scattering of phonons in a harmonic chain of atoms . In the second example, we treat the more complex problem of phonon scattering in a carbon nanotube (CNT) containing an encapsulated C60 molecule which acts as a scatterer of the CNT phonons. The application of this method can enable a more precise characterization of lattice-defect scattering and result in the more controlled use of nanostructuring and lattice defects in phonon engineering for thermoelectric applications.
{"title":"Exact scattering cross section for lattice-defect scattering of phonons","authors":"Zhun-Yong Ong","doi":"arxiv-2408.17004","DOIUrl":"https://doi.org/arxiv-2408.17004","url":null,"abstract":"The use of structurally complex lattice defects, such as functional groups,\u0000embedded nanoparticles, and nanopillars, to generate phonon scattering is a\u0000popular approach in phonon engineering for thermoelectric applications.\u0000However, the theoretical treatment of this scattering phenomenon remains a\u0000formidable challenge, especially with regards to the determination of the\u0000scattering cross sections and rates associated with such lattice defects. Using\u0000the extended Atomistic Green's Function (AGF) method, we describe how the\u0000numerically exact mode-resolved scattering cross section sigma can be computed\u0000for a phonon scattered by a single lattice defect. We illustrate the generality\u0000and utility of the AGF-based treatment with two examples. In the first example,\u0000we treat the isotopic scattering of phonons in a harmonic chain of atoms . In\u0000the second example, we treat the more complex problem of phonon scattering in a\u0000carbon nanotube (CNT) containing an encapsulated C60 molecule which acts as a\u0000scatterer of the CNT phonons. The application of this method can enable a more\u0000precise characterization of lattice-defect scattering and result in the more\u0000controlled use of nanostructuring and lattice defects in phonon engineering for\u0000thermoelectric applications.","PeriodicalId":501369,"journal":{"name":"arXiv - PHYS - Computational Physics","volume":"172 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-08-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142204145","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：