Navigation and service


A Dynamic Cutoff Method for the Large-scale Atomic/Molecular Massively Parallel Simulator

Scientific area

Classical molecular dynamics simulations.

Short description

LAMMPS is a general engine for molecular dynamics simulations capable of studying a broad cross-section of materials, including polymers, crystals, metals, and biomolecules. The major design emphases of LAMMPS are its parallelizability and extensibility, as users can design and implement their own libraries and packages for nearly every task, including integrating Newton's laws of motion, calculating forces and energies, and real-time data processing and output. To overcome the well documented scalability limits arising in simulations of interfacial systems, we developed a dynamic cutoff method (DCM) for interfacial systems of arbitrarily large size, and incorporated it into LAMMPS. The idea consists in adopting a cutoff-based method in which the cutoff is chosen on a particle-by-particle basis, according to the distance from the interface. Computationally, the challenge is shifted from the long-range solvers to the detection of the interfaces and to the computation of the particle-interface distances. For these tasks, we deployed linear-time algorithms that do not rely on global communication patterns; as a result, LAMMPS(DCM) is suited for large systems of particles and massively parallel computers.

Error comparisonAbsolute error in the z-direction (i.e., perpendicular to the interface) for the planar system with 19,200 particles. Each cross corresponds to a single particle. DCM uses a minimum cutoff of 3.0 sigma.


  • 458,752 cores on BlueGene/Q

Weak scaling of LAMMPS on JUQUEENWeak scaling test of LAMMPS comparing the dynamic cutoff method (DCM) and the particle-particle particle-mesh solver (PPPM).

Programming language and model

  • C++, Python
  • MPI, OpenMP

Tested on platforms

  • BlueGene/Q
  • x86

Application developers

LAMMPS itself is developed at SANDIA with many authors.

The Dynamic Cutoff Method is developed by:

Paul Springer
Ahmed Ismail
Paolo Bientinesi

Aachen Institute for Advanced Study in
Computational Engineering Science
RWTH Aachen
Schinkelstr. 2
52062 Aachen


Paul Springer
Aachen Institute for Advanced Study in
Computational Engineering Science

(Text and images provided by the High-Performance and Automatic Computing group)