2011/8/1 chemistry <[email protected]>: > I've installed VOTCA 1.2 and now it works with the tables for bonds > and angle in case of monomer. But i have another problem, i have three > different beads (A, B and C) and that's why i have 6 non-bonded > interactions. I've checked all target RDFs and i did not see any > artifacts on it. But on the "C-C.pot.new" which used in the first > iteration the potential in the region from 0.33 to 0 instead of > increasing is decreasing and has the value "-4937870000" in zero > point. And another problem that after first iteration on two of my > RDFs (A-A.dist.new and C-C.dist.new) appear strange peaks in region > where probability should be zero. When i'd checked the trajectory > confout.gro i really saw between some of the molecules very short > distances for A-A and C-C. How is it possible if the potential for all > interactions, except C-C.pot.new, very strong on this distances and in > this case the beads should not come so close to each other? And can > you please explain to me to fully understand how exactly gromacs works > with the tables for the non-bonded interactions? It is hard to say what goes wrong, have you installed the VOTCA version from Mercurial (using build.sh) ? We recently fixed some small issue in there related to calculating the initial potential: <http://code.google.com/p/votca/source/list?repo=csg&r=stable>
However, it seems to me that the automatic extrapolation failed somewhere. Go to step_000 and have a look at -NAME.dist.tgt that are the target distributions VOTCA uses internally, it is a spline interpolated version of the given targets. -NAME.pot.new that is the k_BT*log(rdf) and then extrapolated exponentially on the left side and extrapolated constant on the right side. If the dist.tgt or pot.new in step_000 look incorrect, you should check your given target distributions again. Sometimes it is easier to create an initial guess for the potential by hand. It can be put in the maindir under the name 'NAME.pot.in'. If the dist.tgt and pot.new in step_000 look correct, go to step_001 and have a look at: -NAME.dist.new - the distribution calculated from the simulation, maybe it is too rough, then you have to increase the length of the simulation in the mdp file. -NAME.dpot.new - the potential update calculated from NAME.dist.new -NAME.pot.new - sum of NAME.pot.cur and NAME.dpot.new Sometimes it also helps to set 'min' in seetings.xml to a value unequal 0 (in your case 0.33 or so) to prevent VOTCA from calculating an update in the region r < min. Cheers, Christoph > > Thanks a lot in advance, > Sergio > > On Jul 16, 4:42 pm, Christoph Junghans <[email protected]> wrote: >> 2011/7/16 chemistry <[email protected]>:> Hi, >> >> > In my case i have the similar shape of the RDFs coming from the >> > iterations to the reference RDFs. But the RDFs coming from the >> > iterations do not converge to the reference RDFs, that's why i want to >> > try to change the numerical factor (0<ai<1) during the iterations, >> > maybe it will help to reach needed value of the potential. And back to >> > my previous question: "how one can change, for example, after third >> > iteration the numerical parameter and start fourth iteration with a >> > new numerical factor." >> >> We do not support something like this. >> It is not hard to implement it, but things like this are better done >> by hand anyway. >> >> 1.) Do 3 iterations >> 2.) Change the scale factor >> 3.) Do more iterations >> >> Cheers, >> >> Christoph >> >> >> >> > Thanks a lot in advance, >> > Sergio. >> >> > On Jul 15, 2:49 pm, Dominik Fritz <[email protected]> wrote: >> >> Hi Sergio, >> >> >> have you checked, that you treat intramolecular exclusions in a >> >> consistent way? >> >> When you obtain the reference RDFs, you typically exclude >> >> (intramolecular) pairs of beads, which are interacting via bonded >> >> interactions. >> >> The same exclusions have to be used for the RDFs coming from the >> >> iteration steps; otherwise your potentials will most probably diverge. >> >> >> Cheers, >> >> >> Dominik >> >> >> On 07/15/2011 01:52 PM, chemistry wrote: >> >> >> > Hey! >> >> >> > In general i'm trying to reproduce the article ("Coarse-grained >> >> > computer simulations of polymer/fullerene bulk heterojunctions for >> >> > organic photovoltaic applications", J.Chem.Theory.Comput., Vol.6, No. >> >> > 2, 2010) to be familiar with the ibi procedure. With the help of VOTCA >> >> > I've got all bonded interaction potentials, but i still have a problem >> >> > with getting the non-bonded interaction potentials. I have the same >> >> > RDFs like in this article. During the first three iterations CG >> >> > potentials going down from the initial guess to the desired value, but >> >> > after third-fifth iteration it going up and diverge. The scale >> >> > parameter seems to be just change how fast or how slow it will >> >> > diverge. Is it something else how one can influence on the >> >> > convergency? >> >> >> > Thanks a lot in advance, >> >> > Sergio >> >> >> > On Jul 12, 5:28 pm, Christoph Junghans <[email protected]> wrote: >> >> >> >> You forget adding "scale" to the "post_update" list. >> >> >> >> Cheers, >> >> >> >> Christoph >> >> >> >> 2011/7/12 chemistry <[email protected]>: >> >> >> >>> Hey! >> >> >> >>> Can you please explain how exactly I can set and variate numerical >> >> >>> factor ai (0<ai<1) for U(n+1) = U(n) + ai*kb*T*ln(P(n)/P(ref)). >> >> >>> Because after 10 iteration instead of convergence I have a big >> >> >>> divergence. I've specified already simple pressure correction and it >> >> >>> works: >> >> >>> <cg> >> >> >>> ... >> >> >>> <inverse> >> >> >>> <p_target>1</p_target> >> >> >>> <post_update>pressure</post_update> >> >> >>> <post_update_options> >> >> >>> <pressure> >> >> >>> <do>0 0 1</do> >> >> >>> <type>simple</type> >> >> >>> <simple> >> >> >>> <scale>0.0003</scale> >> >> >>> </simple> >> >> >>> </pressure> >> >> >>> </post_update_options> >> >> >> >>> Which update script I should use for this purpose and in which section >> >> >>> of settings.xml file I need to specify this option? >> >> >> >>> Thanks a lot in advance, >> >> >>> Sergio. >> >> >> >>> On Jul 10, 8:12 pm, "Victor Rühle" <[email protected]> wrote: >> >> >> >>>> Use the post update script scale and its corresponding option >> >> >> >>>>http://www.votca.org/Documentation/reference-cginteraction >> >> >>>> -- >> >> >>>> Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail >> >> >>>> gesendet. >> >> >> >>>> chemistry <[email protected]> schrieb: >> >> >> >>>> Hello! in U(n+1) = U(n) + ai*kb*T*ln(P(n)/P(ref)) where one can set >> >> >>>> numerical factor ai (0<ai<1). At the moment my calculation is >> >> >>>> diverges, so in general how one can reach better convergency? Kind >> >> >>>> regards, Sergio On Jul 7, 5:03 pm, Christoph Junghans >> >> >>>> <[email protected]> wrote: > 2011/7/7 chemistry >> >> >>>> <[email protected]>:> Is it possible do not use tables_*.xvg and >> >> >>>> instead of it set the > > parameters for bond, angle and dihedral to >> >> >>>> the topol.top with > > equilibrium bond length, angle and force >> >> >>>> constants. In this case ibi- > > procedure for my system works. > > >> >> >>>> Yes, if the potential can be reasonably approximated by a harmonic > >> >> >>>> form. Just compare the table to k/2*(r-r_0)^2 ! > > > And how one >> >> >>>> can choose the exact simulation time for one iteration for > > >> >> >>>> different systems? > > There is no rule. The iteration has to be >> >> >>>> long enough, so that the > distributions of the interactions, you >> >> >>>> want to iteratively refine, are > sampled reasonable. Too long >> >> >>>> iterations just waste your computer >> >> >>>> time > ;-) > I usually do 100 short iterations follow by 50 longer >> >> >>>> ones, but that > is my personal recipe. > > Cheers, > > Christoph > >> >> >>>> > > > > Thanks a lot in advance, > > Sergio > > > On Jul 5, 5:04 pm, >> >> >>>> Victor Ruehle <[email protected]> wrote: > >> Hey, > > >> 1) I >> >> >>>> recommend to use version 1.2. The functionality + arguments of > >> >> >> >>>> convert_potentials have slightly changed. > > >> 2) in the files you >> >> >>>> sent me, the bad sampled regions were not cut and > >> therefore the >> >> >>>> extrapolate failed (zoom in at table_a0.xvg where the > >> >> >> >>>> extrapolated region at the left starts). This will lead to a very > >> >> >>>> >> strong peak in the force and is probably what messes up your run. >> >> >>>> > >> Please make sure to really cut the very rough points at the >> >> >>>> left and > >> right of bonded potentials. This was done in the sed >> >> >>>> commands i send > >> you. Either adjust the number of points there >> >> >>>> or better skip this > >> command and do it manually. > > >> Victor > >> >> >>>> > >> 2011/7/5 chemistry <[email protected]>: > > >> > Hey,> > >> >> >> >>>> > the format of the table_*.xvg which was created after >> >> >>>> post-processing > >> > looks like this: > > >> > table_a1.xvg: > > >> >> >>>> >> > 0.0000000000e+00 8.6955642509e+07 0.0000000000e+00 > >> > >> >> >>>> 5.0000000000e-02 8.6862417509e+07 1.8640000000e+06 > >> > >> >> >>>> 1.0000000000e-01 8.6769242509e+07 1.8630000000e+06 > >> > >> >> >>>> 1.5000000000e-01 8.6676117509e+07 1.8620000000e+06 > >> > >> >> >>>> 2.0000000000e-01 8.6583042509e+07 1.8610000000e+06 > > >> > and >> >> >>>> yesterday I've created the same format of the table like in your > >> >> >>>> >> > case by myself. I used for this C-programing and described the >> >> >>>> > >> > potential with the help of harmonic potential. And now it >> >> >>>> works well. > > >> > new one table_a1.xvg: > > >> > 0.000000 >> >> >>>> 1201.461834 0 > >> > 0.050000 1200.343519 0 > >> > 0.100000 >> >> >>>> 1199.225724 0 > >> > 0.150000 1198.108451 0 > >> > 0.200000 >> >> >>>> 1196.991698 0 > > >> > In both cases the potentials look similar >> >> >>>> around the minimum. Is it > >> > possible that the GROMACS in >> >> >>>> someway cannot read correctly the > >> > >> >> >> >>>> table_*.xvg? > > >> > Kind regards, > >> > Sergio > > >> > On Jul 4, >> >> >>>> 7:50 pm, Victor Ruehle <[email protected]> wrote: > >> >> Hey, > > >> >> >> >>>> >> can you please attach your table_?.xvg > > >> >> Cheers, > >> >> >> >> >>>> Victor > > >> >> 2011/7/4 chemistry <[email protected]>: > > >> >> >> >>>> >> > Hey, Victor! > > >> >> > With table_b1.xvg and table_a1.xvg >> >> >>>> from your tutorial simulation for > >> >> > propane works well. > >> >> >> >>>> >> > I'd tried to process the potentials for propane myself and >> >> >>>> after when > >> >> > I started the simulations I've got the same >> >> >>>> mistake like for my > >> >> > system: > > >> >> > "Fatal error: > >> >> >> >>>> >> > A tabulated bond interaction table number 1 is out of the table >> >> >>>> range: > >> >> > r 3.157974, between table indices 3157 and 3158, >> >> >>>> table length 501" > > >> >> > Seems that I have some problem with >> >> >>>> creating the tables_*.xvg > >> >> > Can you please check the way of >> >> >>>> post-processing of the potentials. > >> >> > At the beginning I had >> >> >>>> the same shape of the potential like in your > >> >> > >> >> >>>> case (I've got it with csg_boltzmann from all-atom MD simulation of >> >> >>>> > >> >> > propane at T=200 K): > > >> >> > bond: > > >> >> > sed -e >> >> >>>> '1,5d' -e 's/$/ i/' AB.pot | tac | sed -e '1,4d' | tac > AB.cut > >> >> >> >>>> >> > csg_call table smooth AB.cut AB.smooth > >> >> > csg_resample >> >> >>>> --in AB.smooth --out AB.refined --grid 0::0.001:0.5 > >> >> > >> >> >>>> csg_call table extrapolate --function quadratic AB.refined >> >> >>>> AB.pot.cur > >> >> > csg_call --options table.xml convert_potential >> >> >>>> xvg --type bonded > >> >> > AB.pot.cur table_b1.xvg > > >> >> > >> >> >>>> where table.xml: > > >> >> > <cg> > >> >> > <inverse> > >> >> > >> >> >>>> <gromacs> > >> >> > <pot_max>1e8</pot_max> > >> >> > >> >> >>>> <table_end>0.5</table_end> > >> >> > >> >> >>>> <table_bins>0.002</table_bins> > >> >> > </gromacs> > >> >> > >> >> >>>> </inverse> > >> >> > </cg> > > >> >> > angle: > > >> >> > sed -e >> >> >>>> 's/$/ i/' ABA_angle.pot | tac | sed -e '1,2d' | tac > ABA.cut > >> >> >> >>>> >> > awk '{print $1/3.141592654*180.0,$2,$3,$4}' ABA.cut > >> >> >>>> ABA.new.cut > >> >> > csg_call table >> >> >>>> smooth ABA.new.cut ABA.smooth > >> >> > csg_resample --in ABA.smooth >> >> >>>> --out ABA.refined --grid 0::0.05:180 > >> >> > csg_call table >> >> >>>> extrapolate --function quadratic ABA.refined > >> >> > ABA.pot.cur > >> >> >>>> >> >> > csg_call --options table.xml convert_potential xvg --type >> >> >>>> bonded > >> >> > ABA.pot.cur table_a1.xvg > > >> >> > where >> >> >>>> table.xml: > > >> >> > <cg> > >> >> > <inverse> > >> >> > >> >> >>>> <gromacs> > >> >> > <pot_max>1e8</pot_max> > >> >> > >> >> >>>> <table_end>180</table_end> > >> >> > >> >> >>>> <table_bins>0.05</table_bins> > >> >> > </gromacs> > >> >> > >> >> >>>> </inverse> > >> >> > </cg> > > >> >> > In both cases the potentials >> >> >>>> look similar around the minimum, but > >> >> > afterwords they are >> >> >>>> totally different. I wanted to get the same > >> >> > table_*.xvg >> >> >>>> like in your example. > > >> >> > Yours respectfully, > >> >> > >> >> >>>> Sergio > > >> >> > On Jul 1, 11:34 am, Victor Ruehle >> >> >>>> <[email protected]> wrote: > >> >> >> Be aware, if you have a bug in >> >> >>>> your coarse-grained topology, it might > >> >> >> >> >> >>>> immediately crash the simulations similar to the problem you > >> >> >> >> >>>> >> described. The csg_map will give a reasonable initial structure, >> >> >>>> this > >> >> >> should do well if your coarse-grained >> >> ... >> >> read more » > > -- > You received this message because you are subscribed to the Google Groups > "votca" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/votca?hl=en. > > -- Dr. Christoph Junghans Votca Core Developer Web: http://www.votca.org -- You received this message because you are subscribed to the Google Groups "votca" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/votca?hl=en.
