Article
Chemistry, Physical
Chenchen Song, Todd J. Martinez, Jeffrey B. Neaton
Summary: The new diagrammatic approach allows for automatic derivation of analytical nuclear gradients for THC methods, ensuring performance comparable to energy calculations with roughly three times the cost. Its application to THC-MP2 and multi-reference perturbation theory leads to more efficient computations and developments in analytical nuclear gradients.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Fang Liu, Michael Filatov, Todd J. Martinez
Summary: Conical intersections are crucial for understanding photochemistry, and the SI-SA-REKS method provides a computationally efficient alternative for studying excited state dynamics in large molecular systems. Our recent implementation demonstrates the promise of SI-SA-REKS with accurate conical intersection topography and energetics for several examples, showing sub-quadratic scaling as a function of molecular size.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Chemistry, Physical
Chenchen Song, Jeffrey B. Neaton, Todd J. Martinez
Summary: A new method for CASPT2 analytical gradients is proposed in this study, utilizing the supporting subspace method and MP2, Fock derivatives, which can calculate gradients more efficiently, reduce computational costs, and has wide applications in fields such as ab initio molecular dynamics simulations and geometry optimization.
JOURNAL OF CHEMICAL PHYSICS
(2021)
Article
Meteorology & Atmospheric Sciences
Sheide Chammas, Qing Wang, Tapio Schneider, Matthias Ihme, Yi-fan Chen, John Anderson
Summary: This study demonstrates the use of tensor processing units (TPUs) to simulate low clouds, providing valuable insights into their role in climate. The simulations conducted using TPUs show unprecedented speed and scalability, allowing for the generation of large datasets for training climate models.
JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS
(2023)
Article
Computer Science, Interdisciplinary Applications
Qing Wang, Matthias Ihme, Yi-Fan Chen, John Anderson
Summary: This study develops a computational fluid dynamics (CFD) simulation framework for fluid-flow prediction on the Tensor Processing Unit (TPU) platform, utilizing low-Mach approximation and finite-difference method. The research shows that the framework performs well both numerically and analytically.
COMPUTER PHYSICS COMMUNICATIONS
(2022)
Article
Multidisciplinary Sciences
Blake Armstrong, Alessandro Silvestri, Raffaella Demichelis, Paolo Raiteri, Julian D. Gale
Summary: Crystallization of alkaline earth metal carbonates from water is significant for biomineralization and environmental geochemistry. Computer simulations provide atomistic insights and quantitative determinations of thermodynamics for individual steps in complement to experimental studies. A revised force field for aqueous alkaline earth metal carbonates is introduced in this article, reproducing solubilities and hydration free energies efficiently on graphical processing units. The performance of the revised force field is compared to previous results for important properties relevant to crystallization.
PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY A-MATHEMATICAL PHYSICAL AND ENGINEERING SCIENCES
(2023)
Review
Chemistry, Physical
Umberto Raucci, Hayley Weir, Sukolsak Sakshuwong, Stefan Seritan, Colton B. Hicks, Fabio Vannucci, Francesco Rea, Todd J. Martinez
Summary: This article outlines methods to eliminate the barriers preventing the wider chemistry community from performing quantum chemistry calculations. These methods include GPU-accelerated quantum chemistry in the cloud, AI-driven natural molecule input methods, and extended reality visualization. The article also highlights the exciting applications of these methods in computing and visualizing spectra, 3D structures, molecular orbitals, and other chemical properties.
ANNUAL REVIEW OF PHYSICAL CHEMISTRY
(2023)
Article
Chemistry, Analytical
Piotr Perek, Aleksander Mielczarek, Dariusz Makowski
Summary: In recent years, cinematography and other digital content creators have been increasingly using Three-Dimensional (3D) imaging technology. However, producing a correct 3D image can still be a complex task. This paper presents a prototype of a diagnostic system that supports the film crew in calibrating stereoscopic cameras and analyzing 3D depth. The system uses GPU for efficient video analysis and can process full HD resolution video streams.
Article
Geochemistry & Geophysics
Anders Kjaer-Rasmussen, Matthew P. Griffiths, Denys Grombacher, Jakob Juul Larsen
Summary: This study investigates the issue of powerline noise in surface nuclear magnetic resonance (NMR) measurements and proposes two new methods to significantly accelerate its removal speed. One method is based on projection to determine the powerline model, and the other utilizes high-performance parallel computations offered by graphical processing units (GPUs).
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Information Systems
Qingjiang Xiao, Shiqiang Du, Jinmei Song, Yao Yu, Yixuan Huang
Summary: In this study, a new method WHLR-MSC is introduced for multi-view subspace clustering, which utilizes weighted tensor nuclear norm to accurately capture class discrimination information of sample distribution and hyper-Laplacian graph regularization to capture local geometric structure of data. Extensive experiments demonstrate the effectiveness of the proposed WHLR-MSC method on benchmark image datasets.
Article
Chemistry, Physical
Chenchen Song
Summary: This paper presents a state-averaged complete active space self-consistent field method in polarizable continuum model (PCM) for studying photoreactions in solvents. The optimization of solute wavefunctions and PCM surface charges of the solvent simultaneously minimizes the state-averaged free energy. The method supports both fixed and dynamic weights, with the weights automatically adjusted based on energy gaps. The paper also discusses the implementation of the method using seven basic operations, allowing for high-performance computation on graphical processing units. Results demonstrating the accuracy and performance of the implementation are presented, along with applications in conical intersection search and photoreaction energy pathways in solvents.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Minggang Guo, Zhifan Wang, Yanzhao Lu, Fan Wang
Summary: This study implemented the calculation of triples in the CCSD(T) approach with spin-orbit coupling on a GPU using single-precision data. The results showed that the use of single-precision data for certain calculations is acceptable in terms of accuracy, but mixed-precision calculations are required for the (T) correlation energy.
Article
Engineering, Electrical & Electronic
Rubens Luiz Rech Junior, Sujit Malde, Carlo Cazzaniga, Maria Kastriotou, Manon Letiche, Christopher Frost, Paolo Rech
Summary: In this study, we investigate the reliability of Google's coral TPUs in different neutron irradiation environments, and find that despite high error rates, most neutron-induced errors only have a slight impact on convolution outputs and do not alter the detection or classification of CNNs. The study provides valuable information for designing CNNs to prevent neutron-induced events from leading to misdetections or misclassifications.
IEEE TRANSACTIONS ON NUCLEAR SCIENCE
(2022)
Article
Computer Science, Hardware & Architecture
R. Lily Hu, Damien Pierce, Yusef Shafi, Anudhyan Boral, Vladimir Anisimov, Sella Nevo, Yi-fan Chen
Summary: Recent advancements in hardware accelerators such as Tensor Processing Units (TPUs) have not only sped up computations for machine learning, but also for scientific modeling and computer simulations, as demonstrated in this study. By utilizing TPUs for distributed scientific computing, we were able to solve partial differential equations (PDEs) for simulating fluid physics and modeling riverine floods. The results show that TPUs achieve a significant speedup of two orders of magnitude compared to CPUs. Access to running physics simulations on TPUs is publicly available through the Google Cloud Platform, and we have released a Python interactive notebook version of the simulation.
INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS
(2022)
Proceedings Paper
Automation & Control Systems
Rubens Luiz Rech Junior, Paolo Rech
Summary: This paper investigates the reliability of Google's Coral TPU on atmospheric neutrons and provides valuable information on how to evaluate and improve the reliability of CNNs executed on a TPU through experimental data.
PROCEEDINGS OF THE 2022 DESIGN, AUTOMATION & TEST IN EUROPE CONFERENCE & EXHIBITION (DATE 2022)
(2022)
Article
Chemistry, Physical
Edward G. Hohenstein, B. Scott Fales, Robert M. Parrish, Todd J. Martinez
Summary: The study presents a quartic-scaling implementation of CCSD based on low-rank tensor hypercontraction (THC) factorizations, which significantly reduces computational complexity and enables accurate correlation energy calculations.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Christoph Bannwarth, Todd J. Martinez
Summary: The article presents an approach to unify ab initio and semiempirical electronic structure code paths. By separating the wavefunction ansatz and the matrix representations of operators, the Hamiltonian can refer to either an ab initio or semiempirical treatment. The authors built a semiempirical integral library and interfaced it to the GPU-accelerated electronic structure code TeraChem. This approach enables the combination of semiempirical Hamiltonians with the full functionality of the ab initio electronic structure code.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Vinicius Wilian D. Cruzeiro, Yuanheng Wang, Elisa Pieri, Edward G. Hohenstein, Todd J. Martinez
Summary: This article introduces how to set up the GPU-accelerated electronic structure program TeraChem as an electronic structure server, which can be easily accessed by third-party client programs. The client interface, called TeraChem protocol buffers (TCPB), has been designed for ease of use and compatibility with multiple programming languages. By incorporating the TCPB client into Amber for QM/MM simulations, significant time savings and speedup have been achieved compared to prior implementations. The practical application of TCPB is demonstrated through the computation of the free energy profile of a model chromophore.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Chenchen Song
Summary: This paper presents a state-averaged complete active space self-consistent field (SA-CASSCF) method in the AMOEBA polarizable water model for rigorous simulation of non-adiabatic molecular dynamics with nonequilibrium solvation effects. The method solves the molecular orbital and configuration interaction coefficients, and the induced dipoles on solvent atoms, by minimizing the state averaged energy. The method is implemented by partitioning the calculations of different terms between the quantum chemistry and molecular mechanics codes.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Review
Chemistry, Physical
Umberto Raucci, Hayley Weir, Sukolsak Sakshuwong, Stefan Seritan, Colton B. Hicks, Fabio Vannucci, Francesco Rea, Todd J. Martinez
Summary: This article outlines methods to eliminate the barriers preventing the wider chemistry community from performing quantum chemistry calculations. These methods include GPU-accelerated quantum chemistry in the cloud, AI-driven natural molecule input methods, and extended reality visualization. The article also highlights the exciting applications of these methods in computing and visualizing spectra, 3D structures, molecular orbitals, and other chemical properties.
ANNUAL REVIEW OF PHYSICAL CHEMISTRY
(2023)
Editorial Material
Chemistry, Physical
Michele Ceriotti, Lasse Jensen, David E. Manolopoulos, Todd Martinez, David R. Reichman, Francesco Sciortino, C. David Sherrill, Qiang Shi, Carlos Vega, Lai-Sheng Wang, Emily A. Weiss, Xiaoyang Zhu, Jenny Stein, Tianquan Lian
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
David M. G. Williams, Eirik F. Kjonstad, Todd J. Martinez
Summary: It is found that the geometric phase effect is correctly reproduced around defective excited-state conical intersections in coupled cluster theory, indicating that these intersections are local artifacts.
JOURNAL OF CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Soren Holm, Pablo A. Unzueta, Keiran Thompson, Todd J. Martinez
Summary: In this study, a graph neural network model is developed and trained to correct the basis set incompleteness error between a small and large basis set at the RHF and B3LYP levels of theory. The results show that fitting an ML model to correct the BSIE is better at generalizing to systems not seen during training compared to fitting to the total potential. Acceptable performance is achieved when the training data sufficiently resemble the systems one wants to make predictions on.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Multidisciplinary
Wei-Hai Fang, Todd J. Martinez, Lihong Liu
Summary: Light-driven molecular motors, such as the chiral N-alkyl imines reported by Greb and Lehn, show promising potential in material and biological systems. However, the mechanism of unidirectional rotation in these motors is not fully understood. Through computational study, we find that the location and energetics of conical intersections alone cannot explain the motor's mechanism, and dynamic effects and out-of-plane distortions play a crucial role. Our results provide insights on how to improve the efficiency of photoisomerization in this class of molecular motors.
JOURNAL OF THE AMERICAN CHEMICAL SOCIETY
(2023)
Article
Multidisciplinary Sciences
Y. Liu, D. M. Sanchez, M. R. Ware, E. G. Champenois, J. Yang, J. P. F. Nunes, A. Attar, M. Centurion, J. P. Cryan, R. Forbes, K. Hegazy, M. C. Hoffmann, F. Ji, M. F. Lin, D. Luo, S. K. Saha, X. Shen, X. J. Wang, T. J. Martinez, T. J. A. Wolf
Summary: Combining ultrafast electron diffraction and ab initio dynamics simulations, the authors visualize the structure of a pericyclic minimum, also known as a pericyclic minimum, in real time in a photochemical reaction. Electrocylic reactions involve the simultaneous formation and cleavage of sigma and pi bonds through a cyclic structure. This research provides experimental evidence for the importance of this structure in electrocyclic reactions.
NATURE COMMUNICATIONS
(2023)
Article
Chemistry, Physical
Keiran C. Thompson, Todd J. Martinez
Summary: Scalable numerical solutions to the time dependent Schrodinger equation remain an outstanding goal in theoretical chemistry. Here we present a method which adaptively adjusts a dictionary of basis functions to the dynamics of the system using recent breakthroughs in signal processing. Our results show that for two low-dimensional model problems, the size of the basis set does not grow quickly with time and is weakly dependent on dimensionality. The method primarily utilizes energies and gradients of the potential, suggesting its potential application in on-the-fly ab initio quantum wavepacket dynamics.
Article
Chemistry, Physical
Wenwen Xu, David M. M. Sanchez, Umberto Raucci, Hantao Zhou, Xinning Dong, Mingqiu Hu, Christopher J. J. Bardeen, Todd J. J. Martinez, Ryan C. C. Hayward
Summary: By incorporating diarylethene microcrystals into a polyethylene terephthalate matrix, highly ordered and compliant photomechanical composites with superior performance compared to single crystals are achieved. These composites exhibit rapid response times, sustain a high level of performance over numerous cycles, and generate high work densities.
Article
Chemistry, Multidisciplinary
Rui Xu, Jan Meisner, Alexander M. Chang, Keiran C. Thompson, Todd J. Martinez
Summary: Our recent success in utilizing GPUs to speed up quantum chemistry computations has led to the creation of the ab initio nanoreactor, a computational framework for automatic reaction discovery and kinetic model construction. In this work, we apply the ab initio nanoreactor to study methane pyrolysis, uncovering the elementary reactions through GPU-accelerated simulations and refining the reaction paths using transition state theory. With 53 species and 134 reactions, the kinetic model derived from the discovered reactions is validated against experimental data and literature models. We also demonstrate the effectiveness of local brute force and Monte Carlo sensitivity analysis for identifying important reactions and improving the accuracy of the kinetic model.
Article
Chemistry, Physical
Alexander M. Chang, Jan Meisner, Rui Xu, Todd J. Martinez
Summary: This study examines the effectiveness of using metadynamics, attractive potentials, and local thermostats for accelerating reaction discovery. By constructing different reaction networks, it is found that a combination of accelerating forces is best suited for reaction discovery.
JOURNAL OF PHYSICAL CHEMISTRY A
(2023)
Article
Chemistry, Physical
Lucas M. Cornetta, Todd J. Martinez, Marcio T. do N. Varella
Summary: Electron induced dissociation reactions play a significant role in various fields. However, simulating dissociation electron attachment (DEA) dynamics is challenging. In this study, a modified AIMS method was proposed to simulate the DEA dynamics of 5-bromouracil, successfully considering all vibrational modes.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2022)