Article
Mathematics, Applied
Wing Tat Leung, Yating Wang
Summary: For time-dependent problems with high-contrast multiscale coefficients, a multirate partially explicit splitting scheme is introduced to achieve efficient simulation with the desired accuracy. The stability of the multirate methods is analyzed for the partially explicit scheme. Local error estimators are derived and an adaptive local temporal refinement framework is proposed to improve computational efficiency.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2022)
Article
Mathematics, Applied
Rujeko Chinomona, Daniel R. Reynolds
Summary: This work introduces a new class of high-order accurate methods for multirate time integration of systems of ordinary differential equations, supporting mixed implicit-explicit treatment. By utilizing an infinitesimal formulation for the fast time scale, the methods provide flexibility for the slow time scale. Order conditions on the IMEX-MRI-GARK coefficients are derived to ensure both third and fourth order accuracy for the overall multirate method.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2021)
Article
Mathematics, Applied
Yating Wang, Wing Tat Leung
Summary: In this paper, a space and time adaptive framework is proposed for solving flow problems in multiscale media. The proposed method uses a stable temporal splitting scheme and constructs multiscale subspaces to handle fast-flow and slow-flow parts separately.
COMPUTERS & MATHEMATICS WITH APPLICATIONS
(2023)
Article
Mathematics
Eric T. Chung, Yalchin Efendiev, Wing Tat Leung, Wenyuan Li
Summary: This work extends previous research on developing partially explicit methods for linear multiscale problems to nonlinear problems, proposing a splitting approach and deriving a stability condition that requires contrast-independent spaces for slow components of the solution. Numerical results show that the proposed methods provide results similar to implicit methods with a time step independent of the contrast.
Article
Nuclear Science & Technology
H. J. Uitslag-Doolaard, K. Zwijsen, F. Roelofs, M. M. Stempniewicz
Summary: Increasing computational power allows the nuclear community to combine existing knowledge and develop tools for simulating interacting phenomena. This paper introduces the development of a code-coupling tool called myMUSCLE and verifies its feasibility through simulations of multiscale thermal-hydraulic applications.
NUCLEAR SCIENCE AND ENGINEERING
(2023)
Article
Multidisciplinary Sciences
Jun Gong, Juling Zhang, Wenqiang Guo, Zhilong Ma, Xiaoyi Lv
Summary: A new short text classification method is proposed based on explicit and implicit multiscale weighting semantic information interaction. The method utilizes word vector model, convolutional neural networks, and long short-term memory to obtain the explicit and implicit representations of short text. A multiscale convolutional neural network is then used to obtain the explicit and implicit multiscale weighting semantics. Experimental results demonstrate that this method outperforms existing short text classification algorithms and models on multiple datasets.
Article
Mathematics, Applied
Dmitry Ammosov, Aleksandr Grigorev, Sergei Stepanov, Aleksei Tyrylgin
Summary: In this paper, a new approach based on hybrid explicit-implicit learning (HEI) is proposed to solve the poroelasticity problem in a fractured medium. The spatial approximation is done using the finite element method with linear basis functions, while the time approximation is achieved through an explicit-implicit scheme. The method incorporates fixed strain and fixed stress splitting schemes to simplify the calculations.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2023)
Article
Mathematics, Applied
D. A. Ammosov, S. P. Stepanov, A. A. Tyrylgin, N. V. Malysheva, L. S. Zamorshchikova
Summary: In this paper, a new mathematical model of language interactions considering bilingualism is proposed. The model incorporates diffusive and convective language spreads with language exchange terms, resulting in a coupled system of partial differential equations. A finite element approximation and a partial learning approach using a Deep Neural Network are developed for the mathematical model. The numerical results demonstrate that the proposed approach achieves good accuracy while reducing computational costs.
JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS
(2023)
Article
Computer Science, Information Systems
Jianfei Wang, Wen Cao
Summary: In the era of big data, a novel multiscale spatiotemporal correlation method is proposed to capture and quantify the uncertainty of spatiotemporal information. The method categorizes spatiotemporal information into explicit and implicit types based on uncertainty levels, and employs spatiotemporal cubes and a benchmark scale to interpret and determine the certainty of each spatiotemporal item. The experimental results demonstrate the effectiveness of the proposed method in capturing spatiotemporal implicit information and enabling better utilization of multiscale spatiotemporal data, providing valuable insights for identifying novel objects and associations.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2023)
Article
Mathematics
Aleksei Tyrylgin, Sergei Stepanov, Dmitry Ammosov, Aleksandr Grigorev, Maria Vasilyeva
Summary: In this paper, a new method based on hybrid explicit-implicit learning is proposed to solve the poroelasticity problem in dual continuum heterogeneous media. The method utilizes deep neural network to learn the implicit part of the flow and combines it with discrete empirical interpolation method and proper orthogonal decomposition for linear interpolation. The results demonstrate fast and accurate predictions.
Article
Mathematics, Applied
Haijin Wang, Anping Xu, Qi Tao
Summary: This paper presents the optimal error estimates of the semi-discrete ultra-weak discontinuous Galerkin method for solving one-dimensional linear convection-diffusion equations, and analyzes the stability and error estimates of the corresponding fully discrete schemes by coupling with a specific Runge-Kutta type implicit-explicit time discretization. Numerical experiments are conducted to verify the theoretical results.
JOURNAL OF COMPUTATIONAL MATHEMATICS
(2022)
Article
Computer Science, Software Engineering
Samar A. Aseeri, Anando Gopal Chatterjee, Mahendra K. Verma, David E. Keyes
Summary: The fast Fourier transform (FFT) has wide applications in various frequency related studies. To improve the efficiency of FFT in large-scale computations, researchers propose using parallel computing and aligning the all-to-all communication of FFT with the physical connections of the Dragonfly topology to achieve better scaling and reduce communication time.
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
(2023)
Article
Computer Science, Hardware & Architecture
Marc T. Henry de Frahan, Jon S. Rood, Marc S. Day, Hariswaran Sitaraman, Shashank Yellapantula, Bruce A. Perry, Ray W. Grout, Ann Almgren, Weiqun Zhang, John B. Bell, Jacqueline H. Chen
Summary: This article introduces the application of PeleC in reacting flow simulations for combustion, which is developed using the AMReX library and targeted at supercomputers. The performance and scalability of PeleC are verified through development and simulation experiments on multiple GPU architectures.
INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS
(2023)
Article
Chemistry, Multidisciplinary
Yan Zhao, Kaiyue Jiang, Can Li, Yufeng Liu, Gucheng Zhu, Michele Pizzochero, Efthimios Kaxiras, Dandan Guan, Yaoyi Li, Hao Zheng, Canhua Liu, Jinfeng Jia, Mingpu Qin, Xiaodong Zhuang, Shiyong Wang
Summary: Individual quantum nanomagnets based on metal-free multi-porphyrin systems have been synthesized. The magnetic coupling between porphyrins was tuned by converting specific porphyrin units to their radical or biradical state. The resulting chains exhibit different magnetic properties, with gap excitation in S = 1/2 antiferromagnets and distinct end states in S = 1 antiferromagnets.
Article
Chemistry, Physical
Mingu Kang, Shiang Fang, Jonggyu Yoo, Brenden R. Ortiz, Yuzki M. Oey, Jonghyeok Choi, Sae Hee Ryu, Jimin Kim, Chris Jozwiak, Aaron Bostwick, Eli Rotenberg, Efthimios Kaxiras, Joseph G. Checkelsky, Stephen D. Wilson, Jae-Hoon Park, Riccardo Comin
Summary: The authors use high-resolution angle-resolved photoemission spectroscopy to determine the microscopic structure of three-dimensional charge order in AV(3)Sb(5) (A = K, Rb, Cs) and its interplay with superconductivity. The observed difference in charge order structure between CsV3Sb5 and the other compounds potentially explains the double-dome superconductivity in CsV3(Sb,Sn)(5) and the suppression of T-c in KV3Sb5 and RbV3Sb5. These findings provide fresh insights into the phase diagram of AV(3)Sb(5).
Article
Computer Science, Theory & Methods
Marcin Rogowski, Samar Aseeri, David Keyes, Lisandro Dalcin
Summary: We introduce mpi4py.futures, which is a lightweight, asynchronous task execution framework for Python using MPI for interprocess communication. It follows the interface of Python's concurrent.futures package and allows scaling over multiple nodes. We discuss its design, implementation, feature set, and compare its performance to other solutions, showing its superiority in most cases.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Physics, Multidisciplinary
Nikita V. Tepliakov, Johannes Lischner, Efthimios Kaxiras, Arash A. Mostofi, Michele Pizzochero
Summary: In this study, a new perspective on the electronic structure of armchair graphene nanoribbons is presented using simple model Hamiltonians and ab initio calculations. The research demonstrates that the energy-gap opening in these nanoribbons is caused by the breaking of a hidden symmetry through long-ranged hopping of pi electrons and structural distortions at the edges. This hidden symmetry can be restored or manipulated through in-plane lattice strain, enabling continuous energy-gap tuning, the emergence of Dirac points at the Fermi level, and topological quantum phase transitions. This work establishes an original interpretation of the semiconducting properties of armchair graphene nanoribbons and provides guidelines for their rational electronic structure design.
PHYSICAL REVIEW LETTERS
(2023)
Article
Meteorology & Atmospheric Sciences
Shixuan Zhang, Christopher J. Vogl, Vincent E. Larson, Quan M. Bui, Hui Wan, Philip J. Rasch, Carol S. Woodward
Summary: Convergence testing is used to verify the discretization of the CLUBB model and identifies two aspects that contribute to noise in the solutions. Numerical limiters and nonlinear artificial diffusion are found to introduce undesirable artifacts, but smoothing the limiters and using linear artificial diffusion can restore the expected convergence. These improvements enhance the trustworthiness of CLUBB's solutions and convergence testing is proven to be a valuable tool for detecting pathologies in the model equation set.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Meteorology & Atmospheric Sciences
Chen Yang, Carl Ponder, Bei Wang, Hoang Tran, Jun Zhang, Jackson Swilley, Laura Condon, Reed Maxwell
Summary: Unprecedented climate change and anthropogenic activities have led to increasing ecohydrological problems, prompting the development of large-scale hydrologic modeling for solutions. However, scientific progress in tracking water parcels at large-scale with high spatiotemporal resolutions is lacking due to the absence of powerful modeling tools. In this study, a parallel framework for the particle tracking model EcoSLIM is demonstrated, showing significant speedup and excellent parallel performance. This parallel framework can be applied to other particle tracking models and is a promising tool for accelerating our understanding of the terrestrial water cycle and upscaling subsurface hydrology to Earth System Models.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Geosciences, Multidisciplinary
Danielle Tijerina-Kreuzer, Jackson S. Swilley, Hoang V. Tran, Jun Zhang, Benjamin West, Chen Yang, Laura E. Condon, Reed M. Maxwell
Summary: This study evaluates multiple data-driven approaches for estimating hydraulic conductivity and subsurface properties at the continental-scale. It provides a recommended Selected National Configuration 1 km resolution subsurface dataset for use in distributed large-and continental-scale hydrologic modeling.
Article
Chemistry, Multidisciplinary
Nikita V. Tepliakov, Ruize Ma, Johannes Lischner, Efthimios Kaxiras, Arash A. Mostofi, Michele Pizzochero
Summary: In this study, it is predicted that the recently fabricated heterojunctions of zigzag nanoribbons embedded in two-dimensional hexagonal boron nitride exhibit half-semimetallic behavior, with opposite energy shifts of the states residing at the two edges while maintaining their intrinsic antiferromagnetic exchange coupling. These heterojunctions undergo an antiferromagnetic-to-ferrimagnetic phase transition upon doping, where the sign of the excess charge controls the spatial localization of the net magnetic moments. This research holds promise for the development of carbon-based spintronics.
Article
Meteorology & Atmospheric Sciences
Chen Yang, Reed Maxwell, Jeffrey Mcdonnell, Xiaofan Yang, Danielle Tijerina-Kreuzer
Summary: This study utilizes GPU-accelerated particle tracking with integrated hydrologic modeling to quantify the variations in evapotranspiration (ET) age at a regional scale. The results reveal that topography-driven flow paths play a crucial role in shaping the spatial and temporal patterns of ET age.
JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES
(2023)
Article
Environmental Sciences
Mario A. Soriano, Reed Maxwell
Summary: This study demonstrates the potential of using machine learning metamodels to efficiently predict transit time distributions in large regions. The results of the metamodels show that using upstream watershed aggregation achieves the best performance in target predictions, and the importance of predictors and their relationships with input-output are consistent across different spatial aggregation types.
ENVIRONMENTAL RESEARCH COMMUNICATIONS
(2023)
Article
Physics, Fluids & Plasmas
Ishan Srivastava, Daniel R. Ladiges, Andy J. Nonaka, Alejandro L. Garcia, John B. Bell
Summary: We propose a numerical method to solve the nonisothermal, compressible Navier-Stokes equations for multispecies fluid mixtures with thermal fluctuations. The method utilizes staggered grid momenta, along with a finite volume discretization of thermodynamic variables, to solve the resulting stochastic partial differential equations. The numerical scheme simplifies the discretization of diffusive and stochastic momentum fluxes and provides clear boundary conditions involving pressure. Compared to a collocated scheme, the staggered grid scheme gives more accurate results for the equilibrium static structure factor of hydrodynamic fluctuations in gas mixtures and the long-ranged correlations of hydrodynamic fluctuations under nonequilibrium conditions. The method is validated through simulations of giant nonequilibrium fluctuations and fluctuation-driven Rayleigh-Taylor instability in gas mixtures, showing excellent agreement with theory and measurements from the direct simulation Monte Carlo method.
Proceedings Paper
Computer Science, Hardware & Architecture
Qinglei Cao, Sameh Abdulah, Hatem Ltaief, Marc G. Genton, David Keyes, George Bosilca
Summary: This paper discusses the importance and challenges of large-scale geospatial modeling in climate and weather prediction, focusing on the potential and advantages of mixed-precision arithmetic in accelerating these modeling processes. The effectiveness and accuracy of the mixed-precision method are validated through experiments on the Summit supercomputer.
2023 IEEE INTERNATIONAL CONFERENCE ON CLUSTER COMPUTING, CLUSTER
(2023)
Article
Physics, Multidisciplinary
Ziyan Zhu, Marios Mattheakis, Weiwei Pan, Efthimios Kaxiras
Summary: In this study, we introduce a deep neural network model called HubbardNet for variational determination of the ground-state and excited-state wave functions of the one-dimensional and two-dimensional Bose-Hubbard model. The model demonstrates excellent generalization ability and outperforms traditional methods in terms of computational efficiency and accuracy.
PHYSICAL REVIEW RESEARCH
(2023)
