The current calculation is causing a discontinuity in the dihedral at the ends, and I think this is occasionally leading to issues in the simulation, at least I think this is the reason. Please see attached.
Do we need to add back in the end calculation as you were mentioning? Maybe that was needed. Thanks. Josh On Saturday, March 4, 2017 at 11:45:23 PM UTC-5, Christoph Junghans wrote: > 2017-03-04 11:26 GMT-07:00 Joshua Moore <[email protected] <javascript:>>: > > 1) Angles > > > > I realized I was not using the development version. Sorry about this. > This > > issue seems to have been corrected in the development version. I'm > still > > not exactly what the change was to fix this, but it is correct in the > > development version. > > > > 2) Dihedrals > > > > I found two issues > > > > a) In potential_to_lammps.sh, there was a scaling to degrees for the > "scale" > > step. This seems to be the cause of it dropping points in the smoothing > > step. If you remove this being applied to the $bondtype="dihedral", > this > > issue goes away. > The issue is that the table beginning and ending should be converted > to degrees as well in the case of dihedral potentials. For the "angle" > type this issue doesn't appear as the table always goes from 0 to 180 > (both in degrees already.) > > > > > However, there was a second issue related to the units LAMMPS expects > for > > the dihedral. I originally thought the default was radians, but it > appears > > it is degrees. We need to add this to table_to_tab.pl so that LAMMPS > knows > > the table is in radians. Then remove the previous check for dihedrals > when > > it is scaling. > > > > } elsif ($type eq "dihedral" ) { > > printf(OUTFILE "VOTCA"\n"); > > printf(OUTFILE "N %i RADIANS\n\n",$#r+1); > > . > > . > > . > > > > LAMMPS seems to be finicky for the derivative for the dihedrals, and I > think > > the amount of noise in the distribution is causing an issue for short > runs > > and LAMMPS is complaining that the dihedral table has inconsistent > forces. > > I'm assuming VOTCA is generating these from a spline? The issue seems > to be > > alleviated with longer runs. However, LAMMPS has an option to leave off > the > > forces for the dihedral table and let LAMMPS calculate them. I'll look > into > > this some more to see if this helps. > > > > I'll work on a pull request if you agree. > Thanks Josh! Your pull request at > <https://github.com/votca/csg/pull/202> seems to fix the two issues > above. > > Christoph > > > > > Thanks. > > > > Josh > > > > > > > > > > > > On Friday, March 3, 2017 at 9:34:52 PM UTC-5, Joshua Moore wrote: > >> > >> Hello, > >> > >> I am seeing some issues with angle and dihedral translation issues. > >> > >> 1) Angles > >> > >> In the attached ANGLE_ISSUE.PNG, I illustrate the issue. > >> > >> On step000, the initial angle potential is inverted correctly from the > >> angle distribution. The unit conversion appears to be correct. On > step001, > >> it also is translated corrected during the scale, smoothing, > extrapolation, > >> interpolation, and shifting steps. However, it appears when the LAMMPS > >> table file is written for the angle, something goes wrong. Here I > compare > >> to the initial inversion by hand which is shifted in step000, which > accounts > >> for the difference. The smoothing, extrapolation, interpolation steps > fall > >> on top of one another, and the shifting step is shifted of course. > >> > >> Again here the issue appears to be the lammps table writing step, > although > >> I haven't been able to find the issue in the code. Any ideas? > >> > >> 2) Dihedrals > >> > >> See attached DIHEDRAL_ISSUE.PNG > >> > >> Here, there may very well be a LAMMPS table writing issue as with the > >> angles, but on step001 the first issue that occurs is on the smoothing > >> steps, where it drops most of the data points and fits only 11 points. > This > >> is the cause for the difference. This is then populated through to the > next > >> steps including writing the LAMMPS table file. Although the smooth, > >> extrapolation, interpolation, and shifting steps are in radians, the > scaling > >> step writes the file in degrees, although it appears this is not > >> subsequently used. This scaling step is correct in the translation of > the > >> potential and the number of points. The next step I think after this > is the > >> smoothing step which again drops most of the points and appears to be > the > >> major issue here. > >> > >> I think the dihedral issue might be a general issue in VOTCA and not > >> necessarily a problem with interfacing with LAMMPS? > >> > >> > >> Thanks in advance. > >> > >> Josh > >> > >> > > -- > > You received this message because you are subscribed to the Google > Groups > > "votca" group. > > To unsubscribe from this group and stop receiving emails from it, send > an > > email to [email protected] <javascript:>. > > To post to this group, send email to [email protected] > <javascript:>. > > Visit this group at https://groups.google.com/group/votca. > > For more options, visit https://groups.google.com/d/optout. > > > > -- > Christoph Junghans > Web: http://www.compphys.de > -- You received this message because you are subscribed to the Google Groups "votca" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/votca. For more options, visit https://groups.google.com/d/optout.
