Return to Steve Plimpton's home page

Molecular Dynamics - Parallel Algorithms

I work in the area of classical molecular dynamics (MD). It's an atomistic simulation method where:

I've worked on several parallel algorithms that are useful for MD simulations. They are described in the next section.

I've also written several parallel MD codes that I distribute freely. They are available from this page.

Performance of these parallel codes for a variety of systems (atomic, polymer, biomolecular, metal, granular) on several parallel platforms are discussed on the Benchmark page of the LAMMPS website.

Simulation efforts that I've been involved in are discussed on this page. Work of others using these codes is listed on the Publications page of the LAMMPS website, along with pictures and movies.



This paper gives an overview of the LAMMPS MD code, including its algorithms, code design, and applications:

LAMMPS - A flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales, A. P. Thompson, H. M. Aktulga, R. Berger, D. S. Bolintineanu, W. M. Brown, P. S. Crozier, P. J. in 't Veld, A. Kohlmeyer, S. G. Moore, T. D. Nguyen, R. Shan, M. J. Stevens, J. Tranchida, C. Trott, S. J. Plimpton, Comp Phys Comm, 271, 108171 (2022). (abstract)


This paper describes 3 classes of parallel algorithms suitable for short-range MD force fields: so-called atom-, force-, and spatial-decomposition algorithms. All 3 are implemented and compared in the paper, but the only one I "invented" was force-decomposition which was work with Bruce Hendrickson at Sandia.

In a nutshell, atom-decomposition methods assign a subset of atoms permanently to each processor, force-decomposition methods assign a subset of pairwise force computations to each proc, and spatial-decomposition methods assign a sub-region of the simulation box to each proc.

Fast Parallel Algorithms for Short-Range Molecular Dynamics, S. J. Plimpton, J Comp Phys, 117, 1-19 (1995). (abstract)

The Lennard-Jones codes discussed in the paper that implement the various parallel algorithms are available for download here.


These two papers describe how MD kernels like force calcluation and neighbor list formation can be formulated to run efficiently Intel CPUs and Xeon Phis:

Increasing Molecular Dynamics Simulation Rates with an 8-Fold Increase in Electrical Power Efficiency, W. M. Brown, A. Semin, M. Hebenstreit, S. Khvostov, K. Raman, S. J. Plimpton, SC16 Conference, SLC, Utah, Nov 2016. (abstract)

Optimizing legacy molecular dynamics software with directive-based offload, W. M. Brown, J.-M. Y. Carrillo, N. Gavhane, F. M. Thakkar, S. J. Plimpton, Comp Phys Comm, 195, 95-101 (2015). (abstract)

This paper discusses how LAMMPS was designed to make it flexible and extensible in an open-source context:

Developing community codes for materials modeling, S. J. Plimpton and J. D. Gale, Current Opinion in Solid State and Materials Science 17, 271-276 (2013). (abstract)

This paper discusses computational attributes of many-body potentials and their performance as implemented in LAMMPS:

Computational Aspects of Many-body Potentials, S. J. Plimpton and A. P. Thompson, MRS Bulletin, 37, 513-521 (2012). (abstract)

These papers describe how MD kernels like force calcluation and neighbor list formation and PPPM can be formulated to run efficiently on hybrid processors, meaning ones that have both a multicore CPU and a GPU:

Implementing Molecular Dynamics on Hybrid High Performance Computers - Short Range Forces, W. M. Brown, P. Wang, S. J. Plimpton, A. N. Tharrington, Comp Phys Comm, 182, 898-911, (2011). (abstract)

Implementing Molecular Dynamics on Hybrid High Performance Computers - Particle-Particle Particle-Mesh, W. M. Brown, A. Kohlmeyer, S. J. Plimpton, A. N. Tharrington, Comp Phys Comm, 183, 449-459 (2012). (abstract)

This paper describes how our parallel MD code LAMMPS is structured at the outer level to enable it to be used as one tool in conjunction with others, e.g. as part of a coupled multiscale calculation:

Software components for parallel multiscale simulation: an example with LAMMPS, B. Frantzdale, S. J. Plimpton, M. S. Shephard, Engineering with Computers, 26, 205-211 (2010). (abstract)

This paper describes how the virial/pressure tensor is computed in LAMMPS for many-body potentials, both in serial and parallel:

General formulation of pressure and stress tensor for arbitrary many-body interaction potentials under periodic boundary conditions, A. P. Thompson, S. J. Plimpton, W. Mattson, J Chem Phys, 131, 154107 (2009). (abstract)

This paper describes the coupling of our LAMMPS MD code to the POEMS multi-body dynamics solver:

Substructured molecular dynamics using multibody dynamics algorithms, R. M. Mukherjee, P. S. Crozier, S. J. Plimpton, K. S. Anderson, Intl J of Non-Linear Mechanics, 43, 1045-1055 (2008). (abstract)

This paper discusses a perdynamic model we added to our LAMMPS MD code to enable mesoscale/continuum modeling of material properties:

Implementing peridynamics within a molecular dynamics code, M. L. Parks, R. B. Lehoucq, S. J. Plimpton, S. A. Silling, Comp Phys Comm, 179, 777-783 (2008). (abstract)

This paper discusses algorithms we added to our LAMMPS MD code to enable efficient modeling of mixtures with widely disparate particles sizes:

Accurate and Efficient Methods for Modeling Colloidal Mixtures in an Explicit Solvent using Molecular Dynamics, P. J. in 't Veld, S. J. Plimpton, G. S. Grest, Comp Phys Comm, 179, 320-329 (2008). (abstract)


This paper describes the pros and cons of the 3 parallel algorithms mentioned above in the context of molecular systems, where one must also compute intra-molecular forces - e.g. bond, angle, torsional terms within each molecule's topology.

Parallel Molecular Dynamics Algorithms for Simulation of Molecular Systems, S. J. Plimpton and B. A. Hendrickson, chapter in Parallel Computing in Computational Chemistry, edited by T. G. Mattson, published by the American Chemical Society, Symposium Series 592, 114-132 (1995). (abstract)


This paper describes how to use the force-decomposition algorithm with embedded atom method (EAM) potentials which are commonly used for metals and alloy systems. We implemented the idea in a code called ParaDyn, which is a parallelization of the serial DYNAMO EAM code of Stephen Foiles and Murray Daw, and which is available for download here.

Parallel Molecular Dynamics With the Embedded Atom Method, S. J. Plimpton and B. A. Hendrickson, in Materials Theory and Modelling, edited by J. Broughton, P. Bristowe, and J. Newsam, MRS Proceedings 291, Pittsburgh, PA, 1993, p 37. (abstract)


The extension of the force-decomposition idea to molecular MD is described in this paper. We used these ideas in a Sandia code called ParBond; it has since been superceded by our LAMMPS MD code.

A New Parallel Method for Molecular-Dynamics Simulation of Macromolecular Systems, S. J. Plimpton and B. A. Hendrickson, J Comp Chem, 17, 326-337 (1996). (abstract)


LAMMPS is our current production-scale MD code (suitable for molecular or atomic systems). Two of its parallel algorithms are discussed in this paper.

Particle-Mesh Ewald and rRESPA for Parallel Molecular Dynamics Simulations, S. J. Plimpton, R. Pollock, M. Stevens, in Proc of the Eighth SIAM Conference on Parallel Processing for Scientific Computing, Minneapolis, MN, March 1997. (abstract)


This paper discusses projections of what will be possible with classical MD using current techniques on high-end computers of the future.

Computational Limits of Classical Molecular-Dynamics Simulations, S. J. Plimpton, Computational Materials Science, 4, 361-364 (1995). (abstract)