Choosing the integrator

Perhaps the next most important choice is what kind of integration algorithm to use. Above we did a constant-temperature (NVT) algorithm (the actual algorithm is of the Nosé-Hoover type). For constant energy (NVE) runs we create the integrator as follows, here passing only the time-step:

itg = IntegratorNVE(timeStep=0.0025)

(Technical note: this creates an object of the same type as IntegratorNVT, but here it defaults to NVE mode—in fact in either case one can switch thermostatting on or off using the SetThermostatOn() method).

Additional integrators available in RUMD are listed below

Name of Integrator Description Arguments              
IntegratorNVE The NVE Leap-frog algorithm. timeStep              
IntegratorNVT The Nos$\acute{e}$-Hoover NVT algorithm. timeStep, targetTemperature, thermostatRelaxationTime              
IntegratorNPTAtomic NPT algorithm. (JCP 101, N 5, 1994)
timeStep, targetTemperature, thermostatRelaxationTime, targetPressure, barostatRelaxationTime
IntegratorNVU Algorithm conserving the total potential energy. dispLength, potentialEnergy              
IntegratorMMC Metropolis NVT Monte Carlo. dispLength, targetTemperature              
IntegratorIHS Energy minimization via the Leap-frog algorithm. timeStep              
IntegratorSLLOD Shear an atomic system in the xy-plane using the SLLOD equations. timeStep, strainRate              
IntegratorMolecularSLLOD Shear a molecular system in the xy-plane using the SLLOD equations. timeStep, strainRate              
IntegratorNPTLangevin Algorithm which generates the NPT ensemble using Langevin dynamics timeStep, targetTemperature, friction, targetPressure, barostatFriction, barostatMass              

The above integrators are chosen in the usual way with named arguments as given in the table. In the case of IntegratorNPTAtomic the user must choose suitable relaxation times for the thermostat and the barostat. An example of reasonable values for the thermostat and barostat relaxation times for the LJ system at $T=2.0$ and $p=5.0$ are thermostatRelaxationTime=0.4 and barostatRelaxationTime=10.0. For IntegratorNPTLangevin the friction may be chosen over a wide range, depending on how strongly individual particles should be coupled to the heat bath. For the box dynamics, note that the ratio of barostatMass to barostatFriction is a relaxation time which should be reasonable for the system (at least long compared to the time step).