Chemistry Reference
In-Depth Information
3.
The algorithm is easy to run in parallel with no costly communication
between processors.
Ordinary PC clusters can be used.
4.
All the trajectories are reactive as in any reaction path method.
This is in
contrast to initial value normal and MTS MD methods, in which many
trajectories do not end at the desired state. This also enhances the efficient
use of computational resources.
5.
The SDEL formulation is general.
It is not limited to processes with large
energy barriers, single barriers, or those with exponential kinetics. This
makes SDEL more versatile than other reaction paths methods.
6.
The algorithm produces an interpolation between the minimum energy
path (MEP) and a true classical trajectory.
Hence, even trajectories with
low resolution can be useful in qualitative reaction path studies.
But the algorithm also has several disadvantages including:
1.
The trajectories are approximate.
High-frequency motions, which can be
important in certain dynamical events, are not resolved.
2.
The computations are expensive.
Trajectories for systems with
1000
atoms require a parallel resource of near 20 CPU-s at current processor
speeds. However, cluster of computers of this size are becoming common in
computational chemistry labs.
3.
The length formulation makes it difficult to estimate the time scale of the
process.
SDEL can provide information about the relative sequence of
events but not absolute times. This is a limitation shared by all reaction
path methods.
4.
Thermodynamic properties are approximate and quantitative kinetic
properties are inaccessible.
The removal of high-frequency modes due to
the large step size affects computed thermodynamic properties and
transition probabilities. Enthalpic properties of slow variables, however,
are affected only slightly.
98
5.
The final solution depends on the initial guess for the trajectory.
No global
optimization protocol will generate the true minimum for the target
function
in an acceptable time for a large system. In the applications of
SDEL, the initial guess is an approximate MEP obtained with a self-penalty
walk algorithm and most of the solutions obtained correspond to
trajectories in local minima somewhere near the initial guess. The
implementation of less biased procedures, which sample trajectories
connecting structures in configuration space,
is a subject of ongoing
research.
118
6.
The current SDEL algorithm uses an implicit solvent model.
This is not a
fatal flaw of SDEL; computations of trajectories with an atomistic
description of the environment are possible. The assumption of time
separation between the relaxation to equilibrium of water molecules and
the solvated molecule (e.g., a protein) makes the calculations viable. The
configuration for the water molecules can then be determined using a
