When Small Things Become a Big Deal

Computer-simulated atomic motion answers real-world questions like “How do things break?”

Simulation of stretching of a silver nanowire accurately shows the entire process from “necking” (thinner regions in the wire) to the formation of a new phase (red portion in the last image).
Image courtesy of Los Alamos National Laboratory
Simulation of stretching of a silver nanowire accurately shows the entire process from “necking” (thinner regions in the wire) to the formation of a new phase (red portion in the last image). In the center images, atomic features called “stacking faults” – the red regions in the necked portion – form. As the stretching continues, the structure evolves back to a “perfect” structure (all-blue image), but with a smaller diameter. In the final image, even further strain has caused formation of a new phase (red). Impressively, each microsecond of atom movement during the stretching of the nanometer-thick wire required only one wall-clock minute on LANL’s Roadrunner high-performance computer, possible only by using accelerated dynamics methods developed at LANL.

The Science

What happens when you stress a material–like stretching a silver nanowire? Advanced computational methods allow simulation of the atomic movements for relatively long simulation times in a short amount of computing time to better answer this and other questions. Recent developments in the simulation approach, the Parallel Replica Dynamics (ParRep) method, allow direct comparison between simulations and experiments for the size of materials and timescales required to understand real-world phenomena.

The Impact

Pushing atomistic simulations to timescales that are relevant for comparison to experiments is crucial for understanding many processes in materials science–processes that could influence new energy technologies. The ParRep method stands to become a workhorse of materials science as it enables simulations that can be directly compared to laboratory experiments and real-world phenomena, including pyrolysis (thermal decomposition), diffusion of hydrogen, interactions at liquid/solid interfaces, fracture, and related deformations (such as stretching of nanowires).


Molecular dynamics (MD) is a computer simulation technique that can calculate the movement and interactions of collections of atoms and molecules. MD is considered a workhorse of computational materials science. In principle, it can be used to predict many material properties without introducing approximation or assumptions beyond the “interaction potential,” a mathematical description of forces between atoms as a function of their geometry. Given this potential, MD simply applies Newton’s famous second law to determine how those atoms move through space. Because MD can be used to evaluate the dynamics for a large number of atoms in three dimensions, it is an extremely powerful and flexible tool to study materials. However, these qualities are computationally expensive and consume large amounts of supercomputer time, limiting the achievable and practical system sizes and simulation times. While the size limitation can be efficiently addressed with massively parallel implementations of MD, allowing for the simulation of trillions of atoms, the same approach usually cannot extend the timescales much beyond microseconds. In this research led by Los Alamos National Laboratory, a technique called the Parallel Replica Dynamics method provides an alternative, parallel-in-time integration strategy. For specific atomic configurations, the ParRep method can simulate the evolution of an atomic structure efficiently over very long timescales–orders of magnitude beyond the microsecond limitation of traditional approaches. The ParRep method is particularly suited to the coming age of exascale computing. A recent review paper demonstrates the usefulness of ParRep by presenting different examples of materials simulations where access to long timescales was essential to study the physical regime of interest. Sixteen years after its first introduction, with a new understanding of its generality and the ever-increasing availability of parallel processing, the ParRep method is coming of age.


Arthur F. Voter, Danny Perez, and Blas P. Uberuaga
Los Alamos National Laboratory
Los Alamos, NM 87545
afv@lanl.gov, danny_perez@lanl.gov, and blas@lanl.gov


This work was supported by the DOE Office of Science, Office of Basic Energy Sciences, and the Los Alamos National Laboratory (LANL) Laboratory Directed Research and Development program.


D. Perez, B. P. Uberuaga, and A. F. Voter, “The parallel replica dynamics method – Coming of age.” Computational Materials Science 100, 90 (2015). [DOI: 10.1016/j.commatsci.2014.12.011]

Highlight Categories

Program: BES , MSE

Performer: DOE Laboratory