Details
There are numerous challenges with running molecular dynamics that are not present with ordinary DFT calculations.
Here are a few challenges, and the solutions chosen by presto:
-
Edge effects can result in unrealistic solvation: only a vanishingly small percentage of solvent molecules are near an interface in a real (macroscopic) reaction,
but a large fraction of the system is near the boundary for small spheres of solvent.
Accordingly, presto confines the system of interest to the center of the solvent sphere with a shallow harmonic potential,
ensuring that several solvent shells insulate the system from any boundaries.
-
An actual microdroplet of organic solvent would evaporate rapidly in the gas phase.
To address this, presto applies a restoring force to any molecules which stray farther than a user-defined radius from the origin.
This enforces a realistic solvent density and prevents evaporative cooling.
-
Real reactions are in thermodynamic equilibrium with an external heat bath, and thus maintain a steady temperature by changing their total energy.
However, conventional âthermostatsâ for molecular dynamics simulations result in unphysical dynamics,
either by scaling particle velocities (Berendsen) or by inducing collisions with an imaginary heat bath (Langevin).
presto employs the Langevin thermostat, but only for the outer solvent shells not directly in contact with the system of interest.
Since the different layers of the droplet are in thermal equilibrium,
this has the effect of controlling the entire systemâs temperature while preserving realistic (non-Langevin) dynamics for the solute and its immediate environment.
-
For all but the fastest reactions, the timescales accessible by molecular dynamics are insufficient to observe processes of interest.
(At room temperature, it will take roughly 17,000 fs to observe a process with âG = 4.0 kcal/molâwhich is still a relatively low barrier for an organic reaction!)
Accordingly, presto permits the addition of pairwise constraining potentials to study non-ground states (akin to a "scan" in electronic structure programs).
The transition state can be located using biased sampling methods, and individual reaction trajectories can then be started from the region of the transition state.
See the tutorial for more details.