NeuroCGMD
Installation
Menu
Home  /  Reference  /  Integrators

Integrator Reference

Time integration algorithms for propagating the equations of motion. NeuroCGMD implements velocity-Verlet and BAOAB Langevin integrators with stochastic thermostats for canonical (NVT) sampling.

VELOCITY-VERLET LANGEVIN — primary integrator for production dynamics
Velocity-Verlet Langevin
Primary integrator — two force evaluations per step
The workhorse integrator for production dynamics. Implements the velocity-Verlet algorithm with an optional Langevin thermostat for NVT sampling.

Algorithm (per timestep):
1. vn+½ = vn + (Δt/2m) Fn
2. Apply thermostat: vdamped = e−γΔt v + √((1−e−2γΔt) kT/m) · ξ
3. xn+1 = xn + Δt · vn+½
4. Evaluate Fn+1 at new positions
5. vn+1 = vn+½ + (Δt/2m) Fn+1

The thermostat (step 2) is only applied when stochastic=True. The noise term ξ is drawn from a standard normal distribution, ensuring correct Boltzmann sampling at the target temperature.

Kinetic energy is computed at each step: Ek = ½ ∑ mi |vi
vn+½ = vn + Δt2m Ftotal  →  xn+1 = xn + Δt·vn+½  →  vn+1 = vn+½ + Δt2m Fn+1
Ftotal includes classical, QCloud, and ML contributions
alternative integrator
BAOAB LANGEVIN SPLITTING — symmetric splitting, one force eval per step
parameters
PARAMETERS — click to expand details
time_step
Default: 0.02 ps
Integration timestep. For CG dynamics, timesteps of 0.01–0.04 ps are typical. Stability depends on the fastest vibrational mode (bond frequency).
friction_coefficient
Default: from state
Langevin coupling constant γ (ps−1). Controls the strength of thermostat coupling. Higher values: stronger coupling, faster equilibration. Lower values: more natural dynamics.
stochastic
Default: False
Enable the stochastic Langevin noise term. Required for NVT sampling. When False, the integrator runs as deterministic velocity-Verlet (NVE-like).
thermal_energy_scale
Default: 1.0
Scaling factor for kT in the noise term. Normally 1.0 for standard simulations. Can be adjusted for replica exchange or simulated tempering protocols.