AMBER Archive (2002)

Subject: Re: Calculating atomic charges for AMBER with Gaussian98W

From: M. L. Dodson (bdodson_at_scms.utmb.EDU)
Date: Sat Nov 23 2002 - 09:36:27 CST

On 23 Nov, Thomas Steinbrecher wrote:
> Dear Amber users,
>
> I have a question concerning the calculation of atomic
> charges for small organic molecules with RESP charge
> fitting.
>
> I use Antechamber with the -fo gcrt option to create a
> Gaussian Job file. I plan to run this with Gaussian98W,
> to use the output with Antechamber again to create a .prep
> file, as suggested in the manual on p. 70-71 and p. 259.
>
> My problem is that the Gaussian runs take very long
> (Running on a 433MHz Intel computer with 128 MB RAM and
> plenty of disk space. My molecule contains ca. 20 heavy and
> 30 hydrogen atoms). If I check the run after 24 hours, I
> find the following in my .out file:
>
> "...
>  Item Value Threshold Converged?
>  Maximum Force 0.078301 0.000450 NO
>  RMS Force 0.006788 0.000300 NO
>  Maximum Displacement 0.852774 0.001800 NO
>  RMS Displacement 0.155835 0.001200 NO
> ..."
>
> I assume this means, that the geometry optimization is far
> from being complete because the values are so much larger
> than the Thresholds.
>
> My question is, is this behaviour to be expected?
>

Yes. 6-31G* calculations are serious computations. And your
molecule is not small for these kinds of calculations, nor is
your computer particularly fast by contemporary standards. And I
expect the Windows Gaussian versions to be slower than the Unix
versions, but I expect you can't do anything about that.

> How long do Gaussian Runs from the antechamber created job
> files approximately take for "typical" organic ligands (no
> metal atoms, ca 20 C, O and N atoms)?
>

Depends on how close the initial geometry is to the optimum
geometry for that level of theory.

> Is there any way to speed up the calculations? I thought
> about removing the SCF=tight command from my G98 input
> file, but I'm not sure if the atomic charges calculated
> that way are still valid for use in AMBER.
>

What may help is to approach the optimum geometry stepwise:
First do a geometry optimization at a lower theory level, say
3-21G (or possibly even lower, followed by 3-21G). Then use
that geometry to start your 6-31G* calculation. You don't say
where you got the initial geometry you input to antechamber, so
predictions are impossible.

My experience for large "new residue" amber simulations is that
the initial Gaussian run is a large fraction of the total time
for the whole computation. Use the best possible initial
geometry. And crystal structures may not be very close to the
QM optimum if there are a lot of degrees of freedom with low
barriers in the molecule (e.g., as for a dinucleotide).

Good luck.
Bud Dodson

> Sorry for the length of my post.
>
> Kind Regards,
>
> Thomas 

-- 
M. L. Dodson                                bdodson_at_scms.utmb.edu
409-772-2178                                FAX: 409-772-1790