14
$\begingroup$

I am trying to learn best practices for force matching ab initio molecular dynamics (AIMD) data with nonlinear least-squares minimization to generate interatomic potentials for use in classical MD.

I understand the concepts of selecting appropriate reference structures (for the expected classical MD behavior) and fitting to physically-justified functional forms. However, the art of selecting the right weighting factors, the right amount of reference structures, and judging if a set of potential parameters are fit properly appears to be highly open-ended problem with no single solution.

I was wondering if anyone knows of good resources (or has any advice) on gaining mastery and judgment over this process.

$\endgroup$
4
  • 1
    $\begingroup$ What is AIMD? Can you link to some examples of problems like this? $\endgroup$ May 21 '20 at 2:35
  • 4
    $\begingroup$ It stands for ab-initio molecular dynamics $\endgroup$
    – anneb101
    May 21 '20 at 2:44
  • $\begingroup$ Would this work $\endgroup$
    – Cody Aldaz
    Jul 11 '20 at 2:29
  • 1
    $\begingroup$ Why would you want to fit using a fixed functional form? $\endgroup$
    – Saleh
    Jul 11 '20 at 20:52