2017-02-13 18:24 GMT-07:00 Joshua Moore <jdmo...@ncsu.edu>: > Hi Christoph, > > Thanks again for your quick message. > > It seems to run ok after the update. > > However, I think the forces are not correct in the LAMMPS table files. I > might be mistaken. > > I'm attaching the input files needed to generate the run with > (toVotca.tar.gz) > > csg-inverse --options settings.xml > > Below is a plot of what the Bond and Angle force from the generated table > files (table_b1.pot and table_a1.pot) after step001. I'm taking a simple > derivative as -deltaU/delta_r or -deltaU/delta_theta, and they don't seem > to match up with what VOTCA is outputting for the forces in the file. I > believe for the angle, that LAMMPS is expecting -dU/dtheta. > We had a discussion about LAMMPS' table format a while ago here: <https://groups.google.com/d/msg/votca/xGJnFSrFXYQ/Thg8kIyoYBoJ> and the code was written by Frank here: <https://groups.google.com/d/msg/votca/Jxsdz6v9QKw/OUVfYan8CwAJ>
I think, for bonds the 4th column is -r*deltaU/delta_r, while for angle and dihedral it is just the negative derivative, but I guess the code was never really tested. Have a look at: < https://github.com/votca/csg/commit/6db1e9f22f4db0f0967e2ed525af31878cf73eaf > Let me know if it fixes it. Otherwise, send me a patch for table_to_tab.pl. Christoph > Consequently, the temperature blows up nearly immediately, and then you > see warnings in LAMMPS like "WARNING: 87 of 91 force values in table are > inconsistent with -dE/dr. Should only be flagged at inflection points > (../bond_table.cpp:380)". For some reason the angle potential is not > triggering this warning, but I believe it should because it also does not > appear to be consistent. (Note: checking the lammps source, it appears > this check is not in the angle table code, so that's probably the reason.) > > The energies and forces from the pair coefficients appear to be correct. > > Any ideas about what I might be doing wrong, or a potential fix? I'm also > not getting perfect agreement to the potential of mean force (e.g., > -kTln(Pr))for the step001 angle and bond potentials either, but this might > partially be due to the shifting you are doing. > > Thanks again for your help. > > Josh > > Bonds: > > > Angles: > > > > On Monday, February 13, 2017 at 6:11:08 PM UTC-5, Christoph Junghans wrote: > >> 2017-02-13 14:06 GMT-07:00 Joshua Moore <jdm...@ncsu.edu>: >> > Thanks. I see now in step001. It does convert the table_a1.pot file >> as you >> > said. >> > >> > The issue seems to be that the LAMMPS potential file generated converts >> the >> > min/max from radians to degrees, but the <step>0.01</step> isn't >> converted, >> > so it tries to write out > 180 degrees in the LAMMPS file (18001 >> points) or >> > up to 18001*0.01*180/3.1415 = 10314.12 degrees. Is there a special way >> to >> > specify this step size after the unit conversion? >> Sorry, I just realized there is a bug the potential conversion for >> lammps. The conversion factor is applied twice. >> 3.1415*(180/3.1415)**2=10313.5. >> >> I fixed that now: >> <https://github.com/votca/csg/commit/cb92bcd6b84a60b05698b5e >> 409fc9d125959f03e> >> >> That said, the support for lammps as a simulation backend is still >> pretty rudimentary in VOTCA. >> >> Christoph >> >> > >> > Thanks for all your help with this. >> > >> > >> > Here is tail of the table_a1.pot LAMMPS file. >> > >> > 17989 1.03064e+04 7.9841910e-04 6.9526739e-04 >> > 17990 1.03069e+04 3.9962120e-04 6.9677605e-04 >> > 17991 1.03075e+04 0.0000000e+00 6.9814120e-04 >> > 17992 1.03081e+04 -4.0036230e-04 6.9936284e-04 >> > 17993 1.03087e+04 -8.0138340e-04 7.0044097e-04 >> > 17994 1.03092e+04 -1.2029811e-03 7.0137559e-04 >> > 17995 1.03098e+04 -1.6050732e-03 7.0216670e-04 >> > 17996 1.03104e+04 -2.0075775e-03 7.0281430e-04 >> > 17997 1.03109e+04 -2.4104117e-03 7.0331839e-04 >> > 17998 1.03115e+04 -2.8134936e-03 7.0367897e-04 >> > 17999 1.03121e+04 -3.2167410e-03 7.0389604e-04 >> > 18000 1.03127e+04 -3.6200716e-03 7.0396961e-04 >> > 18001 1.03132e+04 -4.0234033e-03 7.0389966e-04 >> > >> > >> > On Monday, February 13, 2017 at 3:59:10 PM UTC-5, Christoph Junghans >> wrote: >> >> >> >> 2017-02-13 13:43 GMT-07:00 Joshua Moore <jdm...@ncsu.edu>: >> >> > Hi Christoph, >> >> > >> >> > Thank you for the quick reply. Maybe I am not setting something >> right >> >> > in >> >> > the settings.xml file, because the angle.pot file in step000 is >> still >> >> > from 0 >> >> > to 3.1415. Also my target distribution function is from 0 to 3.1415 >> in >> >> > radians, calculated with csg_stat. Do I need to convert that first >> >> > before >> >> > feeding it through csg_inverse to LAMMPS? >> >> step_000 doesn't run any simulation and hence there is no >> >> table_a1.pot, look in step_001. >> >> >> >> Christoph >> >> > >> >> > <bonded> >> >> > <!-- name of the interaction --> >> >> > <name>angle</name> >> >> > <min>0.0</min> >> >> > <max>3.1415</max> >> >> > <step>0.01</step> >> >> > <inverse> >> >> > <kBT>0.59616108</kBT> >> >> > <!-- target distribution --> >> >> > <target>angle.dist.tgt</target> >> >> > <lammps> >> >> > <kBT>0.59616108</kBT> >> >> > <table>table_a1.pot</table> >> >> > </lammps> >> >> > </inverse> >> >> > </bonded> >> >> > >> >> > >> >> > >> >> > On Monday, February 13, 2017 at 3:31:56 PM UTC-5, Christoph Junghans >> >> > wrote: >> >> >> >> >> >> 2017-02-13 12:56 GMT-07:00 Joshua Moore <jdm...@ncsu.edu>: >> >> >> > Hello, >> >> >> > >> >> >> > How do people handle the fact that the LAMMPS angle potential is >> in >> >> >> > degrees >> >> >> > and not radians? >> >> >> > >> >> >> > If I convert the angle distribution to degrees, will csg_inverse >> >> >> > calculate >> >> >> > the correct comparison? >> >> >> Before running the actual simulation in each iteration step, >> >> >> csg_inverse will convert its internal format (.pot file) to a >> LAMMPS >> >> >> table (name given in the xml file), it will do the conversion from >> >> >> radians to degrees by default. >> >> >> There is also a scaling factor for the non-bonded interactions in >> case >> >> >> you need to convert from nm to Angstrom. >> >> >> >> >> >> Christoph >> >> >> > >> >> >> > Thanks. >> >> >> > >> >> >> > 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 votca+un...@googlegroups.com. >> >> >> > To post to this group, send email to vo...@googlegroups.com. >> >> >> > 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 votca+un...@googlegroups.com. >> >> > To post to this group, send email to vo...@googlegroups.com. >> >> > 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 votca+un...@googlegroups.com. >> > To post to this group, send email to vo...@googlegroups.com. >> > 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 votca+unsubscr...@googlegroups.com. > To post to this group, send email to votca@googlegroups.com. > 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 votca+unsubscr...@googlegroups.com. To post to this group, send email to votca@googlegroups.com. Visit this group at https://groups.google.com/group/votca. For more options, visit https://groups.google.com/d/optout.
