Click to view
The multigrator reduces fluctuations in the conserved quantity in NPT simulations when using limited-precision arithmetic. (a) The conserved quantity (enthalpy H) is shown as a function of time for simulations of a reference lipid/water system, using a conventional integrator with double-precision (red) and single-precision (black) arithmetic, and the multigrator with single-precision arithmetic (blue). (b) The same data as in panel (a), but zoomed in on the last 200 ps time interval. (c) The energy fluctuations in the single-precision conventional integrator mirror fluctuations in system volume. Volume is shown as a function of time for the same time interval as in panel (b).
Click to view
The multigrator better achieves the target pressure in simulations with multiple time step methods compared to a conventional integrator, even with high-precision arithmetic. (a) We performed 500 ns NPT simulations with a target pressure of 1 atm (solid horizontal line) of the same lipid/water reference system with a conventional integrator (red) and the multigrator (blue), both using double-precision arithmetic and both using an r-RESPA multiple time step scheme where long-range electrostatic interactions are evaluated at every third time step. Pressure is estimated by averaging over time windows of the form t ∈ [T − τ, T], where T is the end time of the simulation (500 ns) and τ is the window size. Estimates are plotted as a function of window size. (b) Same as (a), but with a target pressure of 10 atm for both simulations. (c) When the conventional integrator is used without a multiple time step scheme (i.e., all forces computed at every time step), it achieves the target pressure (1 atm, as in (a)).
Article metrics loading...
In molecular dynamics simulations, control over temperature and pressure is typically achieved by augmenting the original system with additional dynamical variables to create a thermostat and a barostat, respectively. These variables generally evolve on timescales much longer than those of particle motion, but typical integrator implementations update the additional variables along with the particle positions and momenta at each time step. We present a framework that replaces the traditional integration procedure with separate barostat, thermostat, and Newtonian particle motion updates, allowing thermostat and barostat updates to be applied infrequently. Such infrequent updates provide a particularly substantial performance advantage for simulations parallelized across many computer processors, because thermostat and barostat updates typically require communication among all processors. Infrequent updates can also improve accuracy by alleviating certain sources of error associated with limited-precision arithmetic. In addition, separating the barostat, thermostat, and particle motion update steps reduces certain truncation errors, bringing the time-average pressure closer to its target value. Finally, this framework, which we have implemented on both general-purpose and special-purpose hardware, reduces software complexity and improves software modularity.
Full text loading...