On Thu, May 16, 2013 at 11:49 AM, David Winogradoff <dwino218.gmail.com>wrote:
> Hey guys,
>
> Thanks Jason for the thorough reference. I am currently comparing 2 prmtop
> files, and their LENNARD_JONES_ACOEFF (& ...BCOEFF) sections are almost
> identical with a few discrepancies. In your reference, you comment that
> there are [NTYPES X (NTYPES + 1)] / 2 entries. Could you explain why there
> are this many entries and how to know which pair of atoms corresponds to
> which entry?
>
NTYPES x (NTYPES + 1) / 2 comes from combinatorics. The A and B
coefficients are pairwise terms, so each term corresponds to an interaction
between a pair of atoms (think of it as a matrix of dimension NTYPES by
NTYPES where each matrix element A_i,j is the A-coefficient between atom
types i and j). This matrix is symmetric (since the X->Y interaction is
the same as the Y->X interaction). Therefore, you only need half of the
NTYPES^2 terms. You also need to include the diagonal of that coefficient
matrix, which is where the extra +1 comes from (so you really need slightly
more than half of the matrix elements: half of the non-diagonals and all of
the diagonals).
There's also a note in those sections that tells you to refer to the
ATOM_TYPE_INDEX and NONBONDED_PARM_INDEX sections to see how the ACOEF and
BCOEF sections are indexed.
The topology file is a complicated beast. There are a few sections with
real 'parameters' (bond force constants, equilibrium values, angle
parameters, dihedral parameters, LJ parameters, charges, and masses). Most
other sections are filled with pointers into various other sections of the
topology file. It is a strong reflection of how the energy and force
routines were originally implemented in Amber, but it does not make for a
human-friendly file format.
While I will certainly not urge against learning the intricacies of the
prmtop file, I will warn against trying to grok topology files of large-ish
systems directly. I would consider myself to be an expert on prmtop files,
and even for trivial questions, I typically use a program to extract the
information I want. I find ParmEd quite helpful here (in large part
because I wrote it) -- there are a variety of commands available to extract
almost any kind of information you want (including the ability to write out
an frcmod file with all of the parameters contained inside).
HTH,
Jason
P.S. One other comment: 'diff' (and meld, vimdiff, etc.) are very limited
tools for comparing prmtop equivalence. Obviously if 2 files are
identical, so too is the energy function defined by those prmtops.
However, 2 prmtops can give identical energies, yet be completely
different according to diff tools.
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 16 2013 - 09:30:02 PDT
