numerical frequencies and frozen atoms

Clicked A Few Times

4:21:18 AM PDT - Wed, Jul 18th 2012
Hello All: I am trying to run some calculations involving the vibration of an aquo ion embedded within a frozen shell meant to approximate the crystalline hydrate environment. Both regions are treated with DFT and the entire system (aquo ion and shell) have C2V symmetry. When invoking geometry, I set the "actlist" to include only the atoms in the central aquo ion. This works fine for the optimization, but when I do the frequency task, it seems NWChem is also carrying out displacements for the frozen region (i.e. only atoms 1-25 are "active", but the code is starting now to do displacements on atom 29). I just want the aquo ion to vibrate in the fixed shell. Moreover, its hard for me to understand the purpose of displacing the fixed atoms as they aren't involved in the optimization and so the forces on them are not zero. Another confusing thing is that when I look at the list of atoms that are getting displaced (grep for "atom:" ) they are not necessarily the symmetry-unique atoms. Although not every atom is being displaced, so symmetry is clearly being used in some way. I was under the impression that once I set the actlist, this was set for every task unless I explicitly unset it. Do I need to tell the numerical frequencies module about actlist explicitly? It seems to me that the vibration of a molecule (or molecular ion) embedded in a fixed cage is a well-defined problem, but maybe my thinking is wrong? The cluster is designed so that the system is neutral overall, if that matters. Thanks for any insight! Jim Rustad

Forum Vet

2:20:37 PM PDT - Wed, Jul 18th 2012
Jim, It is not a 100% clear to me if the calculation of a numerical hessian for active atoms only is working. Reading the active atom section, it states that active atoms is supported in driver, but not in stepper, and stepper is used for the hessian calculation. I went through some code, and it suggests that people have worked on it. You could try setting the active list as follows: set gen_hess:actlist <your list of atoms> and see what this does. We will put this on our list to check. Bert Quote:Jrrustad Jul 18th 11:21 am Hello All: I am trying to run some calculations involving the vibration of an aquo ion embedded within a frozen shell meant to approximate the crystalline hydrate environment. Both regions are treated with DFT and the entire system (aquo ion and shell) have C2V symmetry. When invoking geometry, I set the "actlist" to include only the atoms in the central aquo ion. This works fine for the optimization, but when I do the frequency task, it seems NWChem is also carrying out displacements for the frozen region (i.e. only atoms 1-25 are "active", but the code is starting now to do displacements on atom 29). I just want the aquo ion to vibrate in the fixed shell. Moreover, its hard for me to understand the purpose of displacing the fixed atoms as they aren't involved in the optimization and so the forces on them are not zero. Another confusing thing is that when I look at the list of atoms that are getting displaced (grep for "atom:" ) they are not necessarily the symmetry-unique atoms. Although not every atom is being displaced, so symmetry is clearly being used in some way. I was under the impression that once I set the actlist, this was set for every task unless I explicitly unset it. Do I need to tell the numerical frequencies module about actlist explicitly? It seems to me that the vibration of a molecule (or molecular ion) embedded in a fixed cage is a well-defined problem, but maybe my thinking is wrong? The cluster is designed so that the system is neutral overall, if that matters. Thanks for any insight! Jim Rustad

Clicked A Few Times

3:17:31 PM PDT - Wed, Jul 18th 2012
(sorry if this does not maintain the thread) Hi Bert- Thanks very much for checking this out. I wonder, since the displacements for all the active atoms have finished (all the active atoms are together in the beginning of the geometry) could I kill the job and then read in the hessian by hand? I ask because I am pretty sure you can restart from an aborted numerical hessian and the code is smart enough to pick up where it left off, suggesting that at any given intermediate time the hessian file is in a self-contained, rereadable state. I suppose the problem would be dealing with the symmetry...it might require some actual cleverness on my part to reconstruct a partial block hessian from the active atoms from the "post-mortem" hessian file. Unless the whole 3N x 3N block is filled in... in other words, all the symmetry is already applied when the columns of the hessian are written? So I could just cut out the first 3N rows and 3N columns? But I will try your suggestion in any case and let you know what happens. Thanks, Jim R.

Clicked A Few Times

6:29:44 PM PDT - Wed, Jul 18th 2012
Well, its especially humiliating when you find the answer to your question on the mailing archive, and find you were the one who posed the question! I have no memory of this at all from five years ago. Anyway, this does appear to be working, although symmetry is not used (the code is simply doing all the active atoms) Sorry for the trouble! Jim R. >>> Jim, >>> >>> Hessian does not use actlist input, it actually resets it >>> during calculations for internal code purposes. >>> >>> Instead try using >>> >>> set gen_hess:actlist .... >>> >>> Marat >>> >>> -----Original Message----- >>> From: owner-nwchem-users@emsl.pnl.gov >>> [1] On Behalf Of James Rustad >>> Sent: Thursday, October 04, 2007 10:04 AM >>> To: nwchem-users@emsl.pnl.gov >>> Subject: [NWCHEM] actlist and hessian >>> >>> Dear Users, >>> >>> I notice the "actlist", which specifies which atoms are mobile in a >>> geometry optimization, does not get considered when calculating a >>> hessian (calculated by finite difference). The hessian goes through >>> and >>> steps all the atoms instead of just the active ones. >>> >>> I have an older version (4. something) >>> >>> Does anybody know if the actlist concept has been implemented in >>> hessian >>> calculations in version 5? >>> >>> Thanks >>> Jim Rustad

Forum >> NWChem's corner >> General Topics