Skip to content

fort.13 File

The fort.13 file reports the total force-field error during a ReaxFF force‑field optimization. The error is computed from the cost function defined in trainset.in and reflects how well the current force‑field parameters reproduce the reference data.

This file is only generated during force field optimization runs.

Purpose

  • Track how the total error changes as individual force‑field parameters are varied.
  • Provide input for the parabolic extrapolation procedure used to determine optimized parameter values.
  • Act as a high‑level convergence and diagnostics file for parameter fitting.

When Is fort.13 Written?

fort.13 is written when ReaxFF is run with:

  • iopt = 1 (force field optimization mode)
  • A valid params file
  • A valid trainset.in defining the cost function

Each parameter listed in params produces four entries in fort.13.

Example: fort.13 Output

91.5209  ! Total FF error using 45.00 for parameter 1 1 1
91.5893  ! Total FF error using 55.00 for parameter 1 1 1
91.6449  ! Total FF error using 50.00 for parameter 1 1 1
91.5209  ! Total FF error using optimized value for parameter 1 1 1 (see fort.79)

91.6038  ! Total FF error using 9.7423 for parameter 1 2 1
91.4330  ! Total FF error using 9.9391 for parameter 1 2 1
91.5175  ! Total FF error using 9.8407 for parameter 1 2 1
91.3896  ! Total FF error using optimized value for parameter 1 2 1 (see fort.79)

............ ! Next parameter listed in params

Interpretation

For each parameter listed in the params file, ReaxFF performs:

  1. Lower-bound test
    Parameter value decreased by the search interval

  2. Upper-bound test
    Parameter value increased by the search interval

  3. Baseline test
    Original force‑field value

  4. Optimized test
    Value obtained from parabolic extrapolation (see fort.79)

Each of these runs produces one line in fort.13.

Relation to Other Files

File Role
params Defines which parameters are optimized and their search bounds
trainset.in Defines the cost function
fort.13 Reports total error for each trial
fort.79 Contains details of the parabolic extrapolation
fort.99 Breaks down individual error contributions

Practical Notes

  • Lower values in fort.13 indicate a better fit to the training set.
  • The optimized value is always the 4th entry per parameter.
  • Sudden increases in error often indicate:
  • Overly large search intervals
  • Conflicting training-set targets
  • Over‑constrained linked parameters (koppel2)

ReaxKit Usage

In ReaxKit, fort.13 is typically used to:

  • Track optimization progress
  • Detect unstable parameter updates
  • Summarize force‑field quality during fitting workflows