[gmx-users] Energy drift combining tabulated potentials and group cut-off.
Dear GROMACS Community, I am trying to do a simulation using a tabulated potential which forces me to use the group cut-off. I get a drift in the conserved quantity of -600KJ/mol/ns. My system consists of 1000 rigid TIP4P water molecules and a Na+ ion. I imagine that eventhough it will soon be deprecated the group cutoff scheme still works correctly. I have tryed to change the parameters without success and in any case the ones I use seem reasonable (in my experience using other programs). The rdfs of the system look normal so I ruled out topology problems. It seems that the problem is in the group cut-off since when I change the tabulated potential for a regular vdw the problem persists. I have followed the instructions of: http://www.gromacs.org/@api/deki/files/94/=gromacs_nb.pdf to implement the tabulated potentials. Is this the kind of energy drift acceptable? I have pasted a copy .mdp Thanks for your help, Sergio Perez-Conesa integrator = md dt = 0.001 nsteps = 10 init-step = 0 cutoff-scheme = group nst-list = 1 verlet-buffer-tolerance = 0.0005 ns-type = grid rlist = 1.3 pbc = xyz coulombtype = PME-switch rcoulomb = 1. rcoulomb-switch = 0.95 pme-order = 4 fourierspacing = 0.1 ewald-rtol = 1.e-5 vdwtype = user rvdw = 1.0 DispCorr = No tcoupl = v-rescale tc-grps = System ;nsttcouple = 1 tau-t = ref-t = 300.0 constraints = all-angles constraint-algorithm = LINCS lincs_iter = 1 lincs_order = 4 energygrps = NA OW energygrp_table = NA OW comm-mode = linear -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org.
Re: [gmx-users] Energy drift combining tabulated potentials and group cut-off.
Hi, There's not easy answers to such questions. A practical implementation of MD that can actually do most relevant kinds of science in relevant time frames can't have bitwise conservation. Drift accumulates from multiple aspects of the implementation, too, including the system size, and constraint algorithm, which can dominate over the effects of missing interactions over the lifetime of pair lists. See the discussion in the GROMACS reference manual, e.g. figure 3.5. One proper question to ask is at what point your protocol correctly models whatever relevant physical data exists that is relevant to what you want to observe. Drift that is too large is likely to affect observables, but will affect different ones differently. Mark On Fri, Sep 1, 2017 at 12:12 PM wrote: > Dear GROMACS Community, > > I am trying to do a simulation using a tabulated potential which forces > me to use the group cut-off. I get a drift in the conserved quantity of > -600KJ/mol/ns. My system consists of 1000 rigid TIP4P water molecules > and a Na+ ion. I imagine that eventhough it will soon be deprecated the > group cutoff scheme still works correctly. I have tryed to change the > parameters without success and in any case the ones I use seem > reasonable (in my experience using other programs). The rdfs of the > system look normal so I ruled out topology problems. > > It seems that the problem is in the group cut-off since > when I change the tabulated potential for a regular vdw the problem > persists. I have followed the instructions of: > > http://www.gromacs.org/@api/deki/files/94/=gromacs_nb.pdf > > to implement the tabulated potentials. > > Is this the kind of energy drift acceptable? I have pasted a copy .mdp > > Thanks for your help, > > Sergio Perez-Conesa > > integrator = md > dt = 0.001 > nsteps = 10 > init-step = 0 > cutoff-scheme = group > nst-list = 1 > verlet-buffer-tolerance = 0.0005 > ns-type = grid > rlist = 1.3 > pbc = xyz > coulombtype = PME-switch > rcoulomb = 1. > rcoulomb-switch = 0.95 > pme-order = 4 > fourierspacing = 0.1 > ewald-rtol = 1.e-5 > vdwtype = user > rvdw = 1.0 > DispCorr = No > tcoupl = v-rescale > tc-grps = System > ;nsttcouple = 1 > tau-t = ref-t = 300.0 > constraints = all-angles > constraint-algorithm = LINCS > lincs_iter = 1 > lincs_order = 4 > energygrps = NA OW > energygrp_table = NA OW > comm-mode = linear > -- > Gromacs Users mailing list > > * Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before > posting! > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > * For (un)subscribe requests visit > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or > send a mail to gmx-users-requ...@gromacs.org. > -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org.
Re: [gmx-users] Energy drift combining tabulated potentials and group cut-off.
Hello, Thank you very much for your thoughtfull response. The structural properties seem to be represented correctly so I guess it should work. I know that drifts are something we must live with but since it appeared when changing the system to a tabulated potential+a group cutoff I was not sure if I was making some mistake. Sergio El 01/09/2017 12:48, Mark Abraham escribió: > Hi, > > There's not easy answers to such questions. A practical implementation of > MD that can actually do most relevant kinds of science in relevant time > frames can't have bitwise conservation. Drift accumulates from multiple > aspects of the implementation, too, including the system size, and > constraint algorithm, which can dominate over the effects of missing > interactions over the lifetime of pair lists. See the discussion in the > GROMACS reference manual, e.g. figure 3.5. > > One proper question to ask is at what point your protocol correctly models > whatever relevant physical data exists that is relevant to what you want to > observe. Drift that is too large is likely to affect observables, but will > affect different ones differently. > > Mark > > On Fri, Sep 1, 2017 at 12:12 PM wrote: > >> Dear GROMACS Community, >> >> I am trying to do a simulation using a tabulated potential which forces >> me to use the group cut-off. I get a drift in the conserved quantity of >> -600KJ/mol/ns. My system consists of 1000 rigid TIP4P water molecules >> and a Na+ ion. I imagine that eventhough it will soon be deprecated the >> group cutoff scheme still works correctly. I have tryed to change the >> parameters without success and in any case the ones I use seem >> reasonable (in my experience using other programs). The rdfs of the >> system look normal so I ruled out topology problems. >> >> It seems that the problem is in the group cut-off since >> when I change the tabulated potential for a regular vdw the problem >> persists. I have followed the instructions of: >> >> http://www.gromacs.org/@api/deki/files/94/=gromacs_nb.pdf >> >> to implement the tabulated potentials. >> >> Is this the kind of energy drift acceptable? I have pasted a copy .mdp >> >> Thanks for your help, >> >> Sergio Perez-Conesa >> >> integrator = md >> dt = 0.001 >> nsteps = 10 >> init-step = 0 >> cutoff-scheme = group >> nst-list = 1 >> verlet-buffer-tolerance = 0.0005 >> ns-type = grid >> rlist = 1.3 >> pbc = xyz >> coulombtype = PME-switch >> rcoulomb = 1. >> rcoulomb-switch = 0.95 >> pme-order = 4 >> fourierspacing = 0.1 >> ewald-rtol = 1.e-5 >> vdwtype = user >> rvdw = 1.0 >> DispCorr = No >> tcoupl = v-rescale >> tc-grps = System >> ;nsttcouple = 1 >> tau-t = ref-t = 300.0 >> constraints = all-angles >> constraint-algorithm = LINCS >> lincs_iter = 1 >> lincs_order = 4 >> energygrps = NA OW >> energygrp_table = NA OW >> comm-mode = linear >> -- >> Gromacs Users mailing list >> >> * Please search the archive at >> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before >> posting! >> >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists >> >> * For (un)subscribe requests visit >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or >> send a mail to gmx-users-requ...@gromacs.org. -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org.
Re: [gmx-users] Energy drift combining tabulated potentials and group cut-off.
Hi, Your earlier email implied that the drift was present when you used the group cutoff-scheme, whether or not you used tabulated interactions. But just now you implied something different. Unbuffered cutoffs with the group scheme are not comparable with e.g. the default buffering used with the Verket scheme. Mark On Mon, Sep 4, 2017 at 10:56 AM wrote: > Hello, > > Thank you very much for your thoughtfull response. The structural > properties seem to be represented correctly so I guess it should work. I > know that drifts are something we must live with but since it appeared > when changing the system to a tabulated potential+a group cutoff I was > not sure if I was making some mistake. > > Sergio > > El 01/09/2017 12:48, Mark Abraham escribió: > > > Hi, > > > > There's not easy answers to such questions. A practical implementation of > > MD that can actually do most relevant kinds of science in relevant time > > frames can't have bitwise conservation. Drift accumulates from multiple > > aspects of the implementation, too, including the system size, and > > constraint algorithm, which can dominate over the effects of missing > > interactions over the lifetime of pair lists. See the discussion in the > > GROMACS reference manual, e.g. figure 3.5. > > > > One proper question to ask is at what point your protocol correctly > models > > whatever relevant physical data exists that is relevant to what you want > to > > observe. Drift that is too large is likely to affect observables, but > will > > affect different ones differently. > > > > Mark > > > > On Fri, Sep 1, 2017 at 12:12 PM wrote: > > > >> Dear GROMACS Community, > >> > >> I am trying to do a simulation using a tabulated potential which forces > >> me to use the group cut-off. I get a drift in the conserved quantity of > >> -600KJ/mol/ns. My system consists of 1000 rigid TIP4P water molecules > >> and a Na+ ion. I imagine that eventhough it will soon be deprecated the > >> group cutoff scheme still works correctly. I have tryed to change the > >> parameters without success and in any case the ones I use seem > >> reasonable (in my experience using other programs). The rdfs of the > >> system look normal so I ruled out topology problems. > >> > >> It seems that the problem is in the group cut-off since > >> when I change the tabulated potential for a regular vdw the problem > >> persists. I have followed the instructions of: > >> > >> http://www.gromacs.org/@api/deki/files/94/=gromacs_nb.pdf > >> > >> to implement the tabulated potentials. > >> > >> Is this the kind of energy drift acceptable? I have pasted a copy .mdp > >> > >> Thanks for your help, > >> > >> Sergio Perez-Conesa > >> > >> integrator = md > >> dt = 0.001 > >> nsteps = 10 > >> init-step = 0 > >> cutoff-scheme = group > >> nst-list = 1 > >> verlet-buffer-tolerance = 0.0005 > >> ns-type = grid > >> rlist = 1.3 > >> pbc = xyz > >> coulombtype = PME-switch > >> rcoulomb = 1. > >> rcoulomb-switch = 0.95 > >> pme-order = 4 > >> fourierspacing = 0.1 > >> ewald-rtol = 1.e-5 > >> vdwtype = user > >> rvdw = 1.0 > >> DispCorr = No > >> tcoupl = v-rescale > >> tc-grps = System > >> ;nsttcouple = 1 > >> tau-t = ref-t = 300.0 > >> constraints = all-angles > >> constraint-algorithm = LINCS > >> lincs_iter = 1 > >> lincs_order = 4 > >> energygrps = NA OW > >> energygrp_table = NA OW > >> comm-mode = linear > >> -- > >> Gromacs Users mailing list > >> > >> * Please search the archive at > >> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before > >> posting! > >> > >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > >> > >> * For (un)subscribe requests visit > >> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or > >> send a mail to gmx-users-requ...@gromacs.org. > -- > Gromacs Users mailing list > > * Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before > posting! > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > * For (un)subscribe requests visit > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or > send a mail to gmx-users-requ...@gromacs.org. -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org.
Re: [gmx-users] Energy drift combining tabulated potentials and group cut-off.
Hello, My apollogies, the drift is present whether I use tabulated potentials or not. The email's title I chose is misleading. If I am not mistaken I am using a buffer when I put rcut= 1.0 and rlist=1.3 , but this was not entirely clear to me using with the manual. Sergio El 05/09/2017 12:29, Mark Abraham escribió: > Hi, > > Your earlier email implied that the drift was present when you used the > group cutoff-scheme, whether or not you used tabulated interactions. But > just now you implied something different. Unbuffered cutoffs with the group > scheme are not comparable with e.g. the default buffering used with the > Verket scheme. > > Mark > > On Mon, Sep 4, 2017 at 10:56 AM wrote: > > Hello, > > Thank you very much for your thoughtfull response. The structural > properties seem to be represented correctly so I guess it should work. I > know that drifts are something we must live with but since it appeared > when changing the system to a tabulated potential+a group cutoff I was > not sure if I was making some mistake. > > Sergio > > El 01/09/2017 12:48, Mark Abraham escribió: > > Hi, > > There's not easy answers to such questions. A practical implementation of > MD that can actually do most relevant kinds of science in relevant time > frames can't have bitwise conservation. Drift accumulates from multiple > aspects of the implementation, too, including the system size, and > constraint algorithm, which can dominate over the effects of missing > interactions over the lifetime of pair lists. See the discussion in the > GROMACS reference manual, e.g. figure 3.5. > > One proper question to ask is at what point your protocol correctly models > whatever relevant physical data exists that is relevant to what you want to > observe. Drift that is too large is likely to affect observables, but will > affect different ones differently. > > Mark > > On Fri, Sep 1, 2017 at 12:12 PM wrote: > > Dear GROMACS Community, > > I am trying to do a simulation using a tabulated potential which forces > me to use the group cut-off. I get a drift in the conserved quantity of > -600KJ/mol/ns. My system consists of 1000 rigid TIP4P water molecules > and a Na+ ion. I imagine that eventhough it will soon be deprecated the > group cutoff scheme still works correctly. I have tryed to change the > parameters without success and in any case the ones I use seem > reasonable (in my experience using other programs). The rdfs of the > system look normal so I ruled out topology problems. > > It seems that the problem is in the group cut-off since > when I change the tabulated potential for a regular vdw the problem > persists. I have followed the instructions of: > > http://www.gromacs.org/@api/deki/files/94/=gromacs_nb.pdf > > to implement the tabulated potentials. > > Is this the kind of energy drift acceptable? I have pasted a copy .mdp > > Thanks for your help, > > Sergio Perez-Conesa > > integrator = md > dt = 0.001 > nsteps = 10 > init-step = 0 > cutoff-scheme = group > nst-list = 1 > verlet-buffer-tolerance = 0.0005 > ns-type = grid > rlist = 1.3 > pbc = xyz > coulombtype = PME-switch > rcoulomb = 1. > rcoulomb-switch = 0.95 > pme-order = 4 > fourierspacing = 0.1 > ewald-rtol = 1.e-5 > vdwtype = user > rvdw = 1.0 > DispCorr = No > tcoupl = v-rescale > tc-grps = System > ;nsttcouple = 1 > tau-t = ref-t = 300.0 > constraints = all-angles > constraint-algorithm = LINCS > lincs_iter = 1 > lincs_order = 4 > energygrps = NA OW > energygrp_table = NA OW > comm-mode = linear > -- > Gromacs Users mailing list > > * Please search the archive at > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before > posting! > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists > > * For (un)subscribe requests visit > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or > send a mail to gmx-users-requ...@gromacs.org. -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org. -- Gromacs Users mailing list * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting! * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists * For (un)subscribe requests visit https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-requ...@gromacs.org.