Towards electronic structure-based ab-initio molecular dynamics simulations with hundreds of millions of atoms

Schade, Robert; Kenter, Tobias; Elgabarty, Hossam; Lass, Michael; Schütt, Ole; Lazzaro, A.; Pabst, Hans; Mohr, Stephan; Hutter, Jürg; Kühne, Thomas D.; Plessl, Christian

doi:10.1016/j.parco.2022.102920

Cited by 31 publications

(38 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It should be noted that Desmond only supports single-GPU execution, but this demonstrates how the scalability of Allegro enables it to reach practical performance comparable to that of current, much less accurate methods through the use of current supercomputing hardware. Finally, we can favorably compare the speeds we achieve simulating the 44M atom HIV capsid-3.9-8.7 timesteps/s on 512-1280 nodes-to a previous effort to simulate a 62M atom HIV capsid at quantum accuracy in [32], which achieved 0.0005 timesteps/s on 384 nodes.…”

Section: B Scalabilitymentioning

confidence: 99%

Learning local equivariant representations for large-scale atomistic dynamics

et al. 2023

View full text Add to dashboard Cite

A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.

show abstract

Section: B Scalabilitymentioning

confidence: 99%

Learning local equivariant representations for large-scale atomistic dynamics

et al. 2023

View full text Add to dashboard Cite

show abstract

“…122 In DFT-based MD, only very recently scalability over thousands of GPUs has been achieved exploiting innovative linear scaling approaches and sparse algebra methods within an extended tight-binding scheme. 64 These observations indicate the necessity to develop innovative algorithms and statistical mechanics based methods beyond standard MD approaches as a route towards exascale DFT QM/MM MD, an idea already explored in the context of semiempirical QM/MM simulations. 113 As a very flexible multiscale framework, MiMiC is an excellent candidate to bring DFT QM/MM MD simulations to the exascale by coupling codes running on GPUs and exploiting massively parallel free energy methods.…”

Section: Discussionmentioning

confidence: 99%

“…Despite the many efficient DFT QM/MM software available, [56][57][58][59][60][61][62][63][64][65][66][67] to the best of our knowledge, scarce information can be found in the literature regarding their strong scaling in pure QM/MM MD applications. In this respect, the Multiscale Modeling in Computational Chemistry (MiMiC) QM/MM framework 68,69 that couples CPMD 70 (QM) and GROMACS 71 (MM), represents a notable exception.…”

Section: Introductionmentioning

confidence: 99%

Drug Design in the Exascale Era: A Perspective from Massively Parallel QM/MM Simulations

Raghavan

Paulikat

Ahmad

et al. 2023

Preprint

View full text Add to dashboard Cite

The initial phases of drug discovery - in silico drug design - could benefit from first principle Quantum Mechanics / Molecular Mechanics (QM/MM) molecular dynamics (MD) simulations in explicit solvent, yet many applications are currently limited by the short time scales that this approach can cover. Developing scalable first principle QM/MM MD interfaces fully exploiting current exascale machines - so far an unmet and crucial goal - will help overcome this problem, opening the way to the study of the thermodynamics and kinetics of ligand binding to protein with first principle accuracy. Here, taking two relevant case studies involving the interactions of ligands with rather large enzymes, we showcase the use of our recently developed massively scalable MiMiC QM/MM framework (currently using DFT to describe the QM region) to investigate reactions and ligand binding in enzymes of pharmacological relevance. We also demonstrate for the first time strong scaling of MiMiC QM/MM MD simulations with parallel efficiency above 70\% with over 40,000 cores. Thus, among many others, the MiMiC interface represents a promising candidate towards exascale applications by combining machine learning with statistical mechanics based algorithms tailored for exascale supercomputers.

show abstract

“…Theoretical chemistry has always been one of the main drivers of the development of high-performance computing and computers (HPC) , and is now aiming at saturating the resources of exascale computers. , Since the calculation of electronic structure and dynamics is ultimately NP-hard, this means that classical computing may soon face an exponential wall, not the least when it comes to energy consumption, information is physical . Here, quantum computers may in principle provide exponential advantage for quantum chemistry through quantum superposition and entanglement.…”

Section: Introductionmentioning

confidence: 99%

Complete Active Space Methods for NISQ Devices: The Importance of Canonical Orbital Optimization for Accuracy and Noise Resilience

Triviño

Delcey

Wendin

2023

J. Chem. Theory Comput.

View full text Add to dashboard Cite

To avoid the scaling of the number of qubits with the size of the basis set, one can divide the molecular space into active and inactive regions, which is also known as complete active space methods. However, selecting the active space alone is not enough to accurately describe quantum mechanical effects such as correlation. This study emphasizes the importance of optimizing the active space orbitals to describe correlation and improve the basis-dependent Hartree−Fock energies. We will explore classical and quantum computation methods for orbital optimization and compare the chemically inspired ansatz, UCCSD, with the classical full CI approach for describing the active space in both weakly and strongly correlated molecules. Finally, we will investigate the practical implementation of a quantum CASSCF, where hardware-efficient circuits must be used and noise can interfere with accuracy and convergence. Additionally, we will examine the impact of using canonical and noncanonical active orbitals on the convergence of the quantum CASSCF routine in the presence of noise.

show abstract

Towards electronic structure-based ab-initio molecular dynamics simulations with hundreds of millions of atoms

Cited by 31 publications

References 46 publications

Learning local equivariant representations for large-scale atomistic dynamics

Learning local equivariant representations for large-scale atomistic dynamics

Drug Design in the Exascale Era: A Perspective from Massively Parallel QM/MM Simulations

Complete Active Space Methods for NISQ Devices: The Importance of Canonical Orbital Optimization for Accuracy and Noise Resilience

Contact Info

Product

Resources

About