Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2601.08077 · The Journal of Physical Chemistry A

Accelerating Density Fitting with Adaptive Precision and 8-Bit Integer on AI Accelerators

Hua Huang, Wenkai Shao, Jeff R. Hammond

Published Apr 17, 2026• 2 citations
arXivHTMLPDF

Abstract

The emergence of artificial intelligence (AI) accelerators like NVIDIA Tensor Cores offers new opportunities to speed up tensor-heavy scientific computations. However, applying them to quantum chemistry is challenging due to strict accuracy demands and irregular data patterns. We propose an adaptive precision algorithm to accelerate the density fitting (DF) method with Gaussian basis sets on AI accelerators using 8-bit integer (INT8) arithmetics. Implemented in the GPU-accelerated PySCF package, the algorithm is tested on more than 20 molecular systems with different NVIDIA GPUs. Compared to the standard FP64 code, our algorithm is up to 204% faster on an RTX 4090 gaming GPU and up to 364% faster on an RTX 6000 Ada workstation GPU without compromising the converged energy. This work demonstrates a practical approach to using AI hardware for reliable quantum chemistry simulations.

DOI
10.1021/acs.jpca.6c00225

Related papers

Accelerating Quantum Chemistry with Emulated Double-Precision Matrix Multiplication on GPUs

Giuseppe M. J. Barca, Cole Brower

The evaluation of exact exchange and the treatment of electron correlation remain major computational challenges in electronic-structure theory. Over the past decade, GPU computing has substantially accelerated these calculations, extending the scale and complexity of electronic-structure problems that can be addressed computationally. More recently, GPU architectures developed to meet the demands of artificial intelligence have introduced new opportunities for acceleration, with Tensor Cores providing reduced-precision arithmetic throughput far exceeding that of native double-precision (FP64) operations. Exploiting this performance advantage while preserving the numerical accuracy required in quantum chemistry remains a significant challenge. Here, we investigate the Ozaki-I and Ozaki-II floating-point emulation schemes in the RI-HF, RI-MP2, and RI-CCSD implementations of EXESS on NVIDIA GB200 GPUs. We demonstrate that Ozaki-II at 52-bit precision achieves numerical accuracy comparable to native FP64, delivering complete calculation speedups of up to 1.26× (cc-pVDZ) and 1.50× (cc-pVTZ) for RI-HF, 2.55× and 3.01× for RI-MP2, and 3.11× and 2.15× for RI-CCSD. The energy to solution improves by 1.22× and 2.21× for the representative RI-HF and RI-MP2 cc- pVTZ calculations and by 2.37× for the RI- CCSD cc-pVDZ calculation. Reducing the precision to 47 bits provides further acceleration while maintaining total energy deviations below 0.1 kcal mol -1 across all systems examined.

•Nov 11, 2026•0 citations

Enhancing PySCF-based quantum chemistry simulations with modern hardware, algorithms, and Python tools

Zhichen Pu, Qiming Sun

The PySCF package has emerged as a powerful and flexible open-source platform for quantum chemistry simulations. However, the efficiency of electronic structure calculations can vary significantly depending on the choice of computational techniques and hardware utilization. In this paper, we explore strategies to enhance research productivity and computational performance in PySCF-based simulations. First, we discuss graphics processing unit acceleration for selected PySCF modules. Second, we demonstrate algorithmic optimizations for particular computational tasks, such as initial guess manipulation, second-order self-consistent field methods, multigrid integration, and density fitting approximation, to improve convergence rates and computational efficiency. Finally, we explore the use of modern Python tools, including just-in-time compilation and automatic differentiation, to accelerate code development and execution. These approaches present a practical guide for enhancing the use of PySCF’s capabilities in quantum chemistry research.

arXiv:2506.06661•APL Computational Physics•Sep 1, 2025•16 citations

Introducing GPU Acceleration into the Python-Based Simulations of Chemistry Framework

Rui Li, Qiming Sun, Xing Zhang +1 more

High Resolution Image Download MS PowerPoint Slide We introduce the first version of GPU4PySCF, a module that provides GPU acceleration of methods in PySCF . As a core functionality, this provides a GPU implementation of two-electron repulsion integrals (ERIs) for contracted basis sets comprising up to g functions using the Rys quadrature. As an illustration of how this can accelerate a quantum chemistry workflow, we describe how to use the ERIs efficiently in the integral-direct Hartree–Fock build and nuclear gradient construction. Benchmark calculations show a significant speedup of 2 orders of magnitude with respect to the multithreaded CPU Hartree–Fock code of PySCF and the performance comparable to other open-source GPU-accelerated quantum chemical packages, including GAMESS and QUICK, on a single NVIDIA A100 GPU.

arXiv:2407.09700•The Journal of Physical Chemistry A•Jan 23, 2025•78 citations

Python-Based Quantum Chemistry Software in the Era of Quantum Computing: Educational Tools, Platforms, and Research Integration for 2025

Cyprian Moyo

This presentation was given at the 45th SACI National Convention of the South African Chemical Institute, themed “Chemistry: A Window for Change.” It surveys Python-based quantum chemistry software in 2025 and asks how GPU acceleration, cloud access to quantum hardware, and interactive teaching platforms can be brought into chemistry education and research, especially in African institutions. Contents: Foundations. The quantum mechanical basis of computational chemistry and its main approximations: Born–Oppenheimer, Hartree–Fock, DFT, perturbation theory and coupled cluster. How method and basis set together define a “level of theory.” Where computational and data-driven chemistry sit among the paradigms of science, from empirical through computational to data-driven and AI. The software landscape. Quantum chemistry codes available through South Africa’s Centre for High Performance Computing (CHPC), and the Python ecosystem around MolSSI QCArchive (QCEngine, QCFractal, QCPortal). These give students one interface that works across many electronic-structure programs. Why quantum computing. The caffeine example shows how exact molecular simulation quickly outgrows classical resources, which motivates qubit-based approaches. The talk compares major quantum hardware and cloud providers on educational access and African partnerships, and gives an overview of Africa’s software-first quantum strategy. GPU-accelerated quantum chemistry. GPU4PySCF (v1.4.0, April 2025): reported speedups of up to about 1000× for density-fitting algorithms over single-core PySCF, and 20–50× for general DFT over a 32-core CPU node. CPU–GPU switching through to_gpu() / to_cpu(), and example applications. Educational platforms and pedagogy. Cloud quantum education programmes (IBM Quantum, Microsoft Azure Quantum, Google Quantum AI, QubitxQubit), interactive Jupyter-based teaching (eChem), and the main barriers to teaching quantum chemistry with matching technological fixes. A staged learning pathway runs from traditional theory through Python, interactive notebooks, GPU computing and real quantum hardware (e.g., VQE) to research integration. Applications and outlook. Industry case studies in drug discovery, energy and materials. Strategic recommendations for bringing GPU acceleration, quantum hardware access and proven educational platforms into quantum chemistry training. Author: Cyprian B. Moyo, Nelson Mandela University and National Institute for Theoretical and Computational Sciences (NITheCS), South Africa. Acknowledgements: CHPC/NICIS, NITheCS, the Wits Quantum Initiative (WitsQ), and IBM Quantum.

•Nov 12, 2026•0 citations

Calculation of Quantum Chemical Two-Electron Integrals by Applying Compiler Technology on GPU

Gábor János Tornai, István Ladjánszki, Ádám Rák +2 more

In this article, we present an effective approach to calculate quantum chemical two-electron integrals over basis sets consisting of Gaussian-type basis functions on graphical processing unit (GPU). Our framework generates several different variants called routes to the same integral problem with different integral algorithms (McMurchie-Davidson, Head-Gordon-Pople, and Rys) and precision. Each route is benchmarked on more GPU architectures, and with this data, a model is fitted to select the best available route for an integral task given a GPU architecture. Moreover, this approach supports the computation of high angular momentum orbitals up to g effectively on GPU, tested up to cc-pVQZ-sized basis sets. Rigorous analysis is shown regarding the effectiveness of our method. Molecule simulations with several basis sets are measured using NVIDIA GTX 1080 Ti, NVIDIA P100, and NVIDIA V100 cards.

Journal of Chemical Theory and Computation•Sep 10, 2019•57 citations

Speeding Up Hartree–Fock in JuliaChem with Density Fitting

John J. Hayes, Mark S. Gordon

In this work, the density fitting (DF) approximation is added to the restricted Hartree-Fock (RHF) implementation in the JuliaChem computational chemistry code. Utilizing a DF algorithm that uses symmetry and integral screening, a significant reduction in time to compute the Fock matrix is achieved. The symmetry and screening DF-RHF techniques were adapted to be performed on graphics processing units (GPUs), which are well suited to perform the matrix multiplications that comprise the bulk of the Fock build time in DF-RHF. The JuliaChem DF-RHF GPU algorithm employs a novel approach that automatically switches between two DF-RHF algorithms depending on the number of basis functions in the calculation. The JuliaChem GPU DF-RHF implementation demonstrates up to 2× speedup for Fock build times compared to the existing best-in-class GPU DF-RHF implementation by operating directly on screened intermediate matrices. Due to the high portability of the Julia language code, the JuliaChem CPU and GPU DF-RHF implementations could be benchmarked on a variety of CPU and GPU architectures from multiple hardware vendors.

Journal of Chemical Theory and Computation•Apr 6, 2026•0 citations
The emergence of artificial intelligence (AI) accelerators like NVIDIA Tensor Cores offers new opportunities to speed up tensor-heavy scientific computations. However, applying them to quantum chemistry is challenging due to strict accuracy demands and irregular data patterns. We propose an adaptive precision algorithm to accelerate the density fitting (DF) method with Gaussian basis sets on AI accelerators using 8-bit integer (INT8) arithmetics. Implemented in the GPU-accelerated PySCF package, the algorithm is tested on more than 20 molecular systems with different NVIDIA GPUs. Compared to the standard FP64 code, our algorithm is up to 204% faster on an RTX 4090 gaming GPU and up to 364% faster on an RTX 6000 Ada workstation GPU without compromising the converged energy. This work demonstrates a practical approach to using AI hardware for reliable quantum chemistry simulations.
The evaluation of exact exchange and the treatment of electron correlation remain major computational challenges in electronic-structure theory. Over the past decade, GPU computing has substantially accelerated these calculations, extending the scale and complexity of electronic-structure problems that can be addressed computationally. More recently, GPU architectures developed to meet the demands of artificial intelligence have introduced new opportunities for acceleration, with Tensor Cores providing reduced-precision arithmetic throughput far exceeding that of native double-precision (FP64) operations. Exploiting this performance advantage while preserving the numerical accuracy required in quantum chemistry remains a significant challenge. Here, we investigate the Ozaki-I and Ozaki-II floating-point emulation schemes in the RI-HF, RI-MP2, and RI-CCSD implementations of EXESS on NVIDIA GB200 GPUs. We demonstrate that Ozaki-II at 52-bit precision achieves numerical accuracy comparable to native FP64, delivering complete calculation speedups of up to 1.26× (cc-pVDZ) and 1.50× (cc-pVTZ) for RI-HF, 2.55× and 3.01× for RI-MP2, and 3.11× and 2.15× for RI-CCSD. The energy to solution improves by 1.22× and 2.21× for the representative RI-HF and RI-MP2 cc- pVTZ calculations and by 2.37× for the RI- CCSD cc-pVDZ calculation. Reducing the precision to 47 bits provides further acceleration while maintaining total energy deviations below 0.1 kcal mol -1 across all systems examined.
The PySCF package has emerged as a powerful and flexible open-source platform for quantum chemistry simulations. However, the efficiency of electronic structure calculations can vary significantly depending on the choice of computational techniques and hardware utilization. In this paper, we explore strategies to enhance research productivity and computational performance in PySCF-based simulations. First, we discuss graphics processing unit acceleration for selected PySCF modules. Second, we demonstrate algorithmic optimizations for particular computational tasks, such as initial guess manipulation, second-order self-consistent field methods, multigrid integration, and density fitting approximation, to improve convergence rates and computational efficiency. Finally, we explore the use of modern Python tools, including just-in-time compilation and automatic differentiation, to accelerate code development and execution. These approaches present a practical guide for enhancing the use of PySCF’s capabilities in quantum chemistry research.
High Resolution Image Download MS PowerPoint Slide We introduce the first version of GPU4PySCF, a module that provides GPU acceleration of methods in PySCF . As a core functionality, this provides a GPU implementation of two-electron repulsion integrals (ERIs) for contracted basis sets comprising up to g functions using the Rys quadrature. As an illustration of how this can accelerate a quantum chemistry workflow, we describe how to use the ERIs efficiently in the integral-direct Hartree–Fock build and nuclear gradient construction. Benchmark calculations show a significant speedup of 2 orders of magnitude with respect to the multithreaded CPU Hartree–Fock code of PySCF and the performance comparable to other open-source GPU-accelerated quantum chemical packages, including GAMESS and QUICK, on a single NVIDIA A100 GPU.
This presentation was given at the 45th SACI National Convention of the South African Chemical Institute, themed “Chemistry: A Window for Change.” It surveys Python-based quantum chemistry software in 2025 and asks how GPU acceleration, cloud access to quantum hardware, and interactive teaching platforms can be brought into chemistry education and research, especially in African institutions. Contents: Foundations. The quantum mechanical basis of computational chemistry and its main approximations: Born–Oppenheimer, Hartree–Fock, DFT, perturbation theory and coupled cluster. How method and basis set together define a “level of theory.” Where computational and data-driven chemistry sit among the paradigms of science, from empirical through computational to data-driven and AI. The software landscape. Quantum chemistry codes available through South Africa’s Centre for High Performance Computing (CHPC), and the Python ecosystem around MolSSI QCArchive (QCEngine, QCFractal, QCPortal). These give students one interface that works across many electronic-structure programs. Why quantum computing. The caffeine example shows how exact molecular simulation quickly outgrows classical resources, which motivates qubit-based approaches. The talk compares major quantum hardware and cloud providers on educational access and African partnerships, and gives an overview of Africa’s software-first quantum strategy. GPU-accelerated quantum chemistry. GPU4PySCF (v1.4.0, April 2025): reported speedups of up to about 1000× for density-fitting algorithms over single-core PySCF, and 20–50× for general DFT over a 32-core CPU node. CPU–GPU switching through to_gpu() / to_cpu(), and example applications. Educational platforms and pedagogy. Cloud quantum education programmes (IBM Quantum, Microsoft Azure Quantum, Google Quantum AI, QubitxQubit), interactive Jupyter-based teaching (eChem), and the main barriers to teaching quantum chemistry with matching technological fixes. A staged learning pathway runs from traditional theory through Python, interactive notebooks, GPU computing and real quantum hardware (e.g., VQE) to research integration. Applications and outlook. Industry case studies in drug discovery, energy and materials. Strategic recommendations for bringing GPU acceleration, quantum hardware access and proven educational platforms into quantum chemistry training. Author: Cyprian B. Moyo, Nelson Mandela University and National Institute for Theoretical and Computational Sciences (NITheCS), South Africa. Acknowledgements: CHPC/NICIS, NITheCS, the Wits Quantum Initiative (WitsQ), and IBM Quantum.
In this article, we present an effective approach to calculate quantum chemical two-electron integrals over basis sets consisting of Gaussian-type basis functions on graphical processing unit (GPU). Our framework generates several different variants called routes to the same integral problem with different integral algorithms (McMurchie-Davidson, Head-Gordon-Pople, and Rys) and precision. Each route is benchmarked on more GPU architectures, and with this data, a model is fitted to select the best available route for an integral task given a GPU architecture. Moreover, this approach supports the computation of high angular momentum orbitals up to g effectively on GPU, tested up to cc-pVQZ-sized basis sets. Rigorous analysis is shown regarding the effectiveness of our method. Molecule simulations with several basis sets are measured using NVIDIA GTX 1080 Ti, NVIDIA P100, and NVIDIA V100 cards.
In this work, the density fitting (DF) approximation is added to the restricted Hartree-Fock (RHF) implementation in the JuliaChem computational chemistry code. Utilizing a DF algorithm that uses symmetry and integral screening, a significant reduction in time to compute the Fock matrix is achieved. The symmetry and screening DF-RHF techniques were adapted to be performed on graphics processing units (GPUs), which are well suited to perform the matrix multiplications that comprise the bulk of the Fock build time in DF-RHF. The JuliaChem DF-RHF GPU algorithm employs a novel approach that automatically switches between two DF-RHF algorithms depending on the number of basis functions in the calculation. The JuliaChem GPU DF-RHF implementation demonstrates up to 2× speedup for Fock build times compared to the existing best-in-class GPU DF-RHF implementation by operating directly on screened intermediate matrices. Due to the high portability of the Julia language code, the JuliaChem CPU and GPU DF-RHF implementations could be benchmarked on a variety of CPU and GPU architectures from multiple hardware vendors.