Introduction

`PATHSAMPLE` is the `OPTIM` driver program to implement a
discrete path sampling (DPS) construction of stationary point
databases and perform kinetic analysis.[1,2,3,4,5]
This implementation provides a number of different methods for growing stationary
point databases in order to produce samples that are kinetically relevant.
The main difference from the old version `PATHSAMPLE.2.0` is that the geometrical test for
distinguishing stationary points has been changed to use the minimum distance.
This usage is consistent with `OPTIM` and the tolerance is set using the
keyword *GEOMDIFFTOL*. Subsequent enhancements of the code are not implemented
in `PATHSAMPLE.2.0`.

`PATHSAMPLE` relies upon `OPTIM` [6] to perform geometry optimisation,
especially double-ended searches for pathways between specified minima.
In the simplest case, all the `OPTIM` calls will be Dijkstra connect runs
[7] (this is the case if the *CONNECTIONS* parameter is less than or equal to one,
see §4).
Please refer to the `OPTIM` documentation for full details of this program [6].

The DPS technique is designed to calculate rate constants
between two different regions (A and B) on the potential energy surface,
where a region is defined by a set of local minima.[1,2,3,4,5,8,9,10]
In `PATHSAMPLE.2.0` and above, this objective is achieved by building up
a stationary point database using successive double-ended connection runs between
local minima, where connections are chosen according to the minimum distance
between the minima and their predicted committor probabilities.
Note that that this approach is different from the original philosophy,
which focused upon the rates associated with individual discrete paths and created
new paths by perturbing old ones.

David Wales 2017-09-24