Force Field Reference
Interaction potentials, functional forms, parameters, and implementation details. NeuroCGMD uses a compact bead-based force model with bonded, nonbonded, and optional electrostatic terms; additional refinement terms can be injected by QCloud and the residual layer.
▶ BONDED INTERACTIONS — maintain chain connectivity and local geometry
Harmonic Bonds
V(r) = ½k(r − r0)²
The standard harmonic bond potential between connected CG beads. Provides the bonded backbone connectivity that maintains chain integrity during dynamics.
Reference bonded parameters used in the documented starter configuration:
Reference bonded parameters used in the documented starter configuration:
k = 1250 kJ/(mol·nm²)
r0 = 0.37 nm
The high spring constant ensures bonds oscillate around equilibrium rather than stretch significantly. For typical CG timesteps of 0.02 ps, this provides stable dynamics without constraints.Harmonic Angles
V(θ) = ½kθ(θ − θ0)²
Angular potential between three consecutive bonded beads. Controls chain stiffness and local geometry. Angle parameters can be system-specific or derived from mapped AA distributions.
nonbonded interactions
▶ NONBONDED INTERACTIONS — van der Waals and electrostatics
Lennard-Jones (Shifted)
V(r) = 4ε[(σ/r)12−(σ/r)6] − V(rcut)
The standard 12-6 Lennard-Jones potential with a shift at the cutoff distance. The r−12 term models Pauli repulsion at short range. The r−6 term models van der Waals dispersion attraction.
The potential is shifted so V(rcut) = 0, ensuring energy continuity at the cutoff and preventing discontinuous jumps that cause energy drift.
The potential is shifted so V(rcut) = 0, ensuring energy continuity at the cutoff and preventing discontinuous jumps that cause energy drift.
ε = 2.0 kJ/mol — well depth (interaction strength)
σ = 0.47 nm — effective bead diameter
rmin = 21/6σ ≈ 0.528 nm — equilibrium distance
The force is:F(r) = (24ε/r)[2(σ/r)12 − (σ/r)6]
Coulomb Electrostatics
V(r) = qiqj / (4πε0r)
Optional real-space Coulomb potential for charged beads. Applied when the force field includes electrostatic interactions (e.g., charged amino acid side-chains in CG representation). Can be combined with reaction-field or shifted cutoff methods.
efficient pair search
▶ NEIGHBOR LIST — O(N) spatial hashing
O(N) Cell-List Neighbor Search
Spatial hashing for efficient pair finding
Divides the simulation box into cubic cells sized to the interaction cutoff. Each particle is assigned to a cell based on position. Pair interactions are computed only between particles in the same or 26 adjacent cells.
Computational complexity: O(N) rather than O(N²) for brute-force pair enumeration.
The list is rebuilt every step in the current implementation. Future versions may add buffered neighbor-list updates to reduce rebuild frequency.
Computational complexity: O(N) rather than O(N²) for brute-force pair enumeration.
The list is rebuilt every step in the current implementation. Future versions may add buffered neighbor-list updates to reduce rebuild frequency.