Short MD with honest diagnostics
The inputs on this page are educational templates that have not been executed for this course. Use an authorized installation, verify version-specific options and establish your own convergence.
What a short trajectory can teach
Isolated water teaches Born–Oppenheimer MD: solve electrons, compute forces, advance nuclei, repeat. It does not establish liquid thermodynamics, diffusion, converged IR spectra or reaction rates. Classical nuclei omit quantum effects relevant to hydrogen. Start from optimized coordinates and tested electronic settings in an adequate nonperiodic box. Set PROJECT water_nve, RUN_TYPE MD and replace the entire old MOTION block. Add GLOBAL/SEED 24680 to record velocity initialization; a seed is not a bitwise cross-architecture guarantee.
&MOTION
&MD
ENSEMBLE NVE
STEPS 200
TIMESTEP [fs] 0.5
TEMPERATURE [K] 300
ANGVEL_ZERO .TRUE.
&END MD
&PRINT
&TRAJECTORY
FORMAT XYZ
&EACH
MD 1
&END EACH
&END TRAJECTORY
&VELOCITIES
&EACH
MD 1
&END EACH
&END VELOCITIES
&RESTART
&EACH
MD 20
&END EACH
&END RESTART
&END PRINT
&END MOTION
The nominal duration is 100 fs=0.1 ps. TEMPERATURE initializes otherwise-unsupplied velocities, not an NVE heat bath. ANGVEL_ZERO removes initial rotation in this nonperiodic teaching example; record degrees-of-freedom treatment. It is ignored when initial velocities are explicitly supplied. Three atoms give strongly fluctuating instantaneous kinetic temperature, not a bulk measurement. MD controls.
Energy before animation
Read .ener headers for steps, fs time, kinetic energy, temperature, potential and conserved quantity. In uncomplicated NVE assess kinetic plus potential, but do not assume a thermostat's conserved quantity excludes bath contributions. MD diagnostics.
Test timesteps at equal physical duration, not equal steps. Reuse saved positions and velocities for exact initial-state comparison: same nominal temperature or random seed alone is not the analysis contract. A 0.5 fs step is cautious for unconstrained O–H motion but still requires testing force/integration accuracy.
Add a thermostat for the intended ensemble
In a separate water_nvt project, change ENSEMBLE NVE to NVT and insert this inside MD, leaving other settings:
CSVR time constant controls coupling; 100 fs is a teaching choice. Shorter coupling can perturb dynamical correlations. Temperature should fluctuate around a target distribution, not exactly equal 300 K each step. Do not hide bad NVE drift with a strong bath that removes accumulating energy while retaining inaccurate forces. CSVR reference.
Exercise
Compare two NVE timesteps at equal duration and an NVT run with a documented common starting state. Identify relevant energy diagnostics and expected fluctuations, inspect bond/angle traces, and state unsupported conclusions. This tiny short run does not justify bulk statistical averages; correlated samples require equilibration and uncertainty analysis.
Previous：Geometry optimization and boundary conditions · Next：Analysis, restarts and HPC