Simulations: The dark side

With the advance of computer hardware and software, predictive simulations are widely available and used in varying domains of research. But what happens when the limitations of certain simulations are reached? Daan Frenkel, a computational physicist and professor at the University of Cambridge, UK, highlights these issues in a paper published in The European Physical Journal Plus.

Without the appropriate amount of care in application, there are certain simulations that can lead one astray. The author points to the behavior of a macroscopic system, which can be imitated but not really reproduced with boundary conditions, unless some details are lost in the repetition of the system in all directions. When using small unit cells to represent parts of a crystal and the simulation box is not compatible with the crystal’s symmetry, then crystal will be strained.

Some simulations methods appear reasonable but are not. Case in point, using a Monte Carlo simulation to predict thermal properties of a system. Monte Carlo allows for computing mechanical properties while more dedicated simulation techniques are needed to compute free energies.

Frenkel also points out common myths and misconceptions about simulations. He argues that because a proper comparison mechanism between older and newer algorithms does not exist, newer doesn’t necessarily mean better. He points out that older methods that have been discredited in the past often re-immerge with new identities.

D. Frenkel, “Simulations: The dark side” European Physical Journal Plus.

- Amber Harmon