Hierarchical parallelization of divide‐and‐conquer density functional tight‐binding molecular dynamics and metadynamics simulations

Nishimura, Yoshifumi; Nakai, Hiromi

doi:10.1002/jcc.26217

Cited by 10 publications

(8 citation statements)

References 123 publications

(216 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The mechanism was confirmed complementarily by the minimum energy pathway obtained by nudged elastic band (NEB) calculations with a more realistic model, in which the BR system was embedded in lipid membranes consisting of 27 1-palmitoyl-2-oleoyl- sn -glycero-3-phosphocholin (POPC) molecules and solvated by 920 water molecules, totaling 10119 atoms. Such large-scale QM-MD/MTD simulations and NEB calculations can be realized by utilizing three pillars: the linear-scaling divide-and-conquer (DC) technique, − cost-effective density-functional tight-binding (DFTB) method, − and massively parallel implementation with the hybrid of MPI and OpenMP. − Owing to these pillars, the computational overhead of the whole QM system in the DC-DFTB calculation is in principle comparable with that of the small QM system in the usual QM/MM simulations, when conducted in a massively parallel machine with a number of CPU cores. Thus, the DC-DFTB method allows us to effectively treat all the interactions quantum-mechanically in large molecular systems in a consistent manner.…”

Section: Introductionmentioning

confidence: 99%

Quantum-Mechanical Molecular Dynamics Simulations on Secondary Proton Transfer in Bacteriorhodopsin Using Realistic Models

et al. 2021

Self Cite

View full text Add to dashboard Cite

Bacteriorhodopsin (BR) transports a proton from intracellular to extracellular (EC) sites through five proton transfers. The second proton transfer is the release of an excess proton stored in BR into the EC medium, and an atomistic understanding of this whole process has remained unexplored due to its ubiquitous environment. Here, fully quantum mechanical (QM) molecular dynamics (MD) and metadynamics (MTD) simulations for this process were performed at the divide-and-conquer density-functional tight-binding level using realistic models (∼50000 and ∼20000 atoms) based on the time-resolved photointermediate structures from an X-ray free electron laser. Regarding the proton storage process, the QM-MD/MTD simulations confirmed the Glu-shared mechanism, in which an excess proton is stored between Glu194 and Glu204, and clarified that the activation occurs by localizing the proton at Glu204 in the photocycle. Furthermore, the QM-MD/MTD simulations elucidated a release pathway from Glu204 through Ser193 to the EC water molecules and clarified that the proton release starts at ∼250 μs. In the ubiquitous proton diffusion in the EC medium, the transient proton receptors predicted experimentally were assigned to carboxylates in Glu9 and Glu74. Large-scale QM-MD/MTD simulations beyond the conventional sizes, which provided the above findings and confirmations, were possible by adopting our DCDFTBMD program.

show abstract

Section: Introductionmentioning

confidence: 99%

Quantum-Mechanical Molecular Dynamics Simulations on Secondary Proton Transfer in Bacteriorhodopsin Using Realistic Models

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…The present PA-based DFTB Ehrenfest MD algorithm was implemented into the developmental version of the D cdftbmd code. , The distributed memory parallelization was carried out using MPI. All the computations reported below were conducted on a multinode computation system, where two Intel Xeon Gold 6148 central processing units (CPUs), comprising a total of 40 cores, were installed in each node with 192 GB/node of memory.…”

Section: Resultsmentioning

confidence: 99%

Scalable Ehrenfest Molecular Dynamics Exploiting the Locality of Density-Functional Tight-Binding Hamiltonian

Uratani

Nakai

2021

J. Chem. Theory Comput.

Self Cite

View full text Add to dashboard Cite

To explore the science behind excited-state dynamics in high-complexity chemical systems, a scalable nonadiabatic molecular dynamics (MD) technique is indispensable. In this study, by treating the electronic degrees of freedom at the density-functional tight-binding level, we developed and implemented a reduced scaling and multinode-parallelizable Ehrenfest MD method. To achieve this goal, we introduced a concept called patchwork approximation (PA), where the effective Hamiltonian for real-time propagation of the electronic density matrix is partitioned into a set of local parts. Numerical results for giant icosahedral fullerenes, which comprise up to 6000 atoms, suggest that the scaling of the present PA-based method is less than quadratic, which yields a significant advantage over the conventional cubic scaling method in terms of computational time. The acceleration by the parallelization on multiple nodes was also assessed. Furthermore, the electronic and structural dynamics resulting from the perturbation by the external electric field were accurately reproduced with the PA, even when the electronic excitation was spatially delocalized.

show abstract

“…These features are required not only for current NISQ devices but also for future fault-tolerant quantum devices with tens of thousands of qubits and sufficient error correction. Moreover, the DC method, with high parallel efficiency on a graphical processing unit , or supercomputers, , would effectively utilize a cluster of quantum devices with a relatively small number of qubits and quantum devices with massive qubits.…”

Section: Discussionmentioning

confidence: 99%

Quantum Algorithm of the Divide-and-Conquer Unitary Coupled Cluster Method with a Variational Quantum Eigensolver

Yoshikawa

Takanashi

Nakai

2022

J. Chem. Theory Comput.

Self Cite

View full text Add to dashboard Cite

The variational quantum eigensolver (VQE) with shallow or constant-depth quantum circuits is one of the most pursued approaches in the noisy intermediate-scale quantum (NISQ) devices with incoherent errors. In this study, the divide-and-conquer (DC) linear scaling technique, which divides the entire system into several fragments, is applied to the VQE algorithm based on the unitary coupled cluster (UCC) method, denoted as DC-qUCC/VQE, to reduce the number of required qubits. The unitarity of the UCC ansatz that enables the evaluation of the total energy as well as various molecular properties as expectation values can be easily implemented on quantum devices because the quantum gates are unitary operators themselves. Based on this feature, the present DC-qUCC/VQE algorithm is designed to conserve the total number of electrons in the entire system using the density matrix evaluated on a quantum computer. Numerical assessments clarified that the energy errors of the DC-qUCC/VQE calculations decrease by using the constraint of the total number of electrons. Furthermore, the DC-qUCC/VQE algorithm could reduce the number of quantum gates and shows the possibility of decreasing incoherent errors.

show abstract

Hierarchical parallelization of divide‐and‐conquer density functional tight‐binding molecular dynamics and metadynamics simulations

Cited by 10 publications

References 123 publications

Quantum-Mechanical Molecular Dynamics Simulations on Secondary Proton Transfer in Bacteriorhodopsin Using Realistic Models

Quantum-Mechanical Molecular Dynamics Simulations on Secondary Proton Transfer in Bacteriorhodopsin Using Realistic Models

Scalable Ehrenfest Molecular Dynamics Exploiting the Locality of Density-Functional Tight-Binding Hamiltonian

Quantum Algorithm of the Divide-and-Conquer Unitary Coupled Cluster Method with a Variational Quantum Eigensolver

Contact Info

Product

Resources

About