summaryrefslogtreecommitdiff
path: root/sim/src/BUGS
blob: 98fa8699ed74f8014f4cdebb7bf5829bcf61a98b (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
 Copyright 2006 Nanorex, Inc.  See LICENSE file for details. 

Known issues with the simulator/minimizer.

When minimizing a complex structure, the RMS force value can fluctuate
from one iteration to the next.  This is a natural result of the
minimization process, since it is the potential energy which is being
minimized, not the force value.  When an actual minimum is approached,
the force should be low, since the gradient (force) must be zero at
the exact minimum location.

For a complex structure, the shape of the potential surface is also
complex.  Long, narrow "valleys" in the potential surface may arise.
If the minimizer drops into one of these valleys, it may bounce from
side to side as it progresses down the valley.  If the sides of the
valley have differing slopes, the RMS force value will fluctuate with
each "bounce."

The RMS force value may drop below the exit threshold of 1 pN on one
side of such a valley, causing the minimizer to terminate.  It may be
obvious to the user that this is not a minimum energy configuration,
and the minimizer may be restarted.  At that point, the minimizer may
bounce to the other side of the valley with a resulting increase in
the RMS force.

The minimizer also exits if it detects that it is making little
forward progress, and it may do so even if the RMS force value is
quite high.  This can happen if the calculated gradient points in a
direction which does not lead downhill along the actual potential
energy surface.  This could be due to a bug in the gradient code, or
due to the fact that the calculated gradient can only be an
approximation of the actual gradient.  Different stretch and bend
parameters may be coupled in a complex structure, and that coupling is
not directly represented in the gradient calculation.

The end result is that minimizing a structure could result in a new
structure with a higher RMS force value than the input structure.  It
should be the case that the new structure will always have a lower
potential energy than the input structure, even if the RMS force value
may sometimes be higher.