Русская версия English version

On accelerating computation through quick evaluation of particle proximity in the method of molecular dynamics

A.A. Kharitonov, I.F. Yasinskiy

Vestnik IGEU, 2017 issue 4, pp. 50—55

Download PDF

Abstract in English: 

Background: The method of molecular dynamics (MMD) is widely applied to studying continuous media processes. It is used in the design of nuclear reactors to analyse the effects of radiation on matter, in the calculation of hydraulic, filtration and temperature conditions to justify the reliability and safety of hydraulic structures, and in a number of other energy problems solved by the mechanics of continuous media. An important MMD problem that has to be addressed is the choice of a method for estimating the proximity of the particles making up the simulated system. The selection criteria for this method depend on the researcher's experience, while proximity estimation is a computationally complex procedure, and it significantly inhibits the ability to study large objects. The aim of the study is to accelerate calculations in the molecular dynamics method by developing an improved technique for estimating the particle proximity.

Materials and methods: We used methods of molecular dynamics, Eulerian and Lagrangian particle proximity evalution methods as well as methods of applied mathematics such as numerical integration of differential equations by the Runge-Kutta method, estimation of the error in the results and computer modeling of physical processes.

Results: A combined method for estimating the proximity of particles is proposed based on the sorting of particles along one of the spatial coordinates and sequential clipping of the removed particles in accordance with the metric function. The proposed method for estimating the proximity of particles is studied using the example of the physical problem of gas-dust cloud dynamics. Numerical experiments with sequential and parallel software models have shown an speed increase in the computations made in accordance with the proposed approach in comparison with the traditional Eulerian and Lagrangian proximity schemes by 2,2 times in a system of 103 particles and up to 200 times in a system of 106 particles.

Conclusion: A promising approach is proposed for estimating the proximity of particles in molecular dynamics. Reducing the time spent on modeling molecular processes speeds up the study of large systems consisting of more than 106 particles. The method is an alternative to the known Eulerian and Lagrangian methods and is compatible with all kinds of potential interactions. It can be used in many applied fields, in particular in the molecular dynamics mathematical modeling. The reliability of the results is confirmed by the similarity of the test models and the models studied in power engineering.

References in English: 

1. Krivtsov, A.M., Krivtsova, N.V. Metod chastits i ego ispol'zovanie v mekhanike deformiruemogo tela [Particle method and its use in deformable body mechanics]. Dal'nevostochnyy matematicheskiy zhurnal DVO RAN, 2002, vol. 3, no. 2, рр. 254–276.

2. Dimova, S., Puzynin, I., Puzynina, T., Tukhliev, Z., Hristov, I., Hristova, R., Chernogorova, T., Sharipov, Z. Nepreryvno-atomisticheskoe modelirovanie protsessov vzaimodeystviya tyazhelykh ionov s metallami na vysokoproizvoditel'nykh vychislitel'nykh sistemakh [Continuum-atomistic modeling of interaction processes of heavy ions with metals by using high performance computer systems]. Trudy VII Mezhdunarodnoy konferentsii «Raspredelennye vychisleniya i grid-tekhnologii v nauke i obrazovanii» (GRID 2016) [Proceedings of the 7th International Conference «Distributed Computing and Grid-technologies in Science and Education» (GRID 2016)]. Dubna, Russia, 2017. CEUR Workshop Proceedings (CEUR-WS.org).

3. Kapustin, P.E. Modelirovanie GPU-tsirkoniya metodom molekulyarnoy dinamiki [Modeling of zinconium HCP by the molecular dynamics method]. Izvestiya Samarskogo nauchnogo tsentra Rossiyskoy Akademii Nauk, 2013, vol. 15, no. 4–5, pp. 1131–1136.     

4. Kholmurodov, Kh.T., Altayskiy, V.M., Puzynin, I.V., Dardin, T., Filatov, F.P. Metody molekulyarnoy dinamiki dlya modelirovaniya fizicheskikh i biologicheskikh protsessov [Methods of molecular dynamics for modeling physical and biological processes]. Fizika elementarnykh chastits i atomnogo yadra, 2003, vol. 34, issue 2.

5. Rahman, A. Correlations in the motion of atoms in liquid argon. Phys. Rev., 1964, 136A, pp. 405–411.

6. Haile, J.M. Molecular dynamics simulation. Wiley, 1992.

7. Shaytan, K.V. Molekulyarnaya dinamika peptidov [Molecular dynamics of peptides]. Dinamicheskie modeli protsessov v kletkakh i subkletochnykh nanostrukturakh [Dynamic models of processes in cells and subcellular nanostructures]. Moscow; Izhevsk, 2009, p. 101.

8. Kroketon, K.A. Fizika zhidkogo sostoyaniya [Physics of liquid state]. Moscow, 1978.

9. Galimov, E.M., Krivtsov, A.M., Zabrodin, A.V., Legkostupov, M.S., Eneev, T.M., Sidorov, Y.I. Dinamicheskaya model' obrazovaniya sistemy Zemlya-Luna [A dynamic model of the Earth-Moon system formation]. Geokhimiya, 2005, no. 11, pp. 1137–1149.

Ключевые слова на русском языке: 
методы молекулярной динамики, оценка близости, параллельные вычисления, моделирование молекулярных процессов, алгоритм сортировки
Ключевые слова на английском языке: 
molecular dynamics, proximity estimation, parallel computations, molecular process simulation, sorting algorithm
The DOI index: 
10.17588/2072-2672.2017.4.050-055
Downloads count: 
6