AMBER Archive (2005)

Subject: Re: AMBER: R.E.D

From: FyD (fyd_at_u-picardie.fr)
Date: Sun Mar 06 2005 - 19:42:20 CST

Dear Wai Keat Yam,

> Hello there. I would like to express my gratitude to
> the development of R.E.D, which has indeed help me a
> lot in calculating charges, to be used in AMBER.

R.E.D. is only the application of the following work:
Cieplak et al. J. Comput. Chem. 1995 16 1357-1377

R.E.D.-II present numerous improvments compared to R.E.D.-I (the version you
have I think) thanks in particular to the AMBER mailing discussions. We can send
you the new version which present a really better manual and also new tutorials.
We worked a lot on the new manual.

> R.E.D works very well with g03, as my input file
> (minimized structure)was done by g03 and other
> calculation (MEP) also using g03.

R.E.D. works well with GAMESS-US, g94 (thanks to Menno D. & Vlad C.), g98 & g03.
We will implement the use of GAMESS-UK in R.E.D.-III (thanks to the help of
Thomas P.)

> However,
> from the output file (mol_o.mol2), I found that 2 of
> my H atom (out of 118 atoms) is charging negatively
> (-0.0293, -0.0893). I would like to ask
> whether this is possible? Could I still use these
> values in AMBER later??

Yes, here is what we found for the EtOH molecule:
C5 0.3922
H6 -0.0545 (methylene group).

> Also, my input file was minimized with B3LYP/6-31G
> (d,p) and R.E.D used HF/6-31G* to generate MEP
> calculation. Is this reasonable?

As Piotr answered, only HF/6-31G* is implemented in R.E.D.-I since it is the
theory level which has been used by the Kollman's group in the Cornell et al.
FF development. My understanding is that the HF/6-31G* theory level is
_required_ to compute the MEP (see the manual why). However, which theory level
has to be used for the geom. opt. is less clear for me. Personnaly, I use
B3LYP/6-31G* quite often but _only_ in geom. opt. We added information about
the use of HF/6-31G* in R.E.D. in the R.E.D.-II manual.

However, the use of B3LYP instead of HF in geom. opt. might have a tricky
disadvantge: the implementation of the B3LYP method in GAMESS & Gausssian is
different. This means that the RESP (and ESP charges) will be different (not
reproducible) for the same structure/minimum if you compare the results
obtained with 2 QM soft. We ran a lot of tests and caraterized which conditions
give different RESP charges using the 2 QM soft: This is summarized in the
R.E.D. manual version II. Here is what you can find in the R.E.D.II manual about
the use of B3LYP in geom. opt.:
"Another example is the utilization of the B3LYP method in geometry
optimization. Different mathematical equations for this method (based on the
VWN5 or VWN3 correlation functional) are implemented in each program. If this
method is used to optimize geometries (i. e. the R.E.D. code has to be
modified), it leads to slightly different geometries and RESP or ESP charges."

Thank you for this very kind email, regards,

  F.-Y. Dupradeau
DMAG EA 3901 & Faculte de Pharmacie, Amiens, France
The Scripps Research Institute, La Jolla, CA, USA
       --
http://www.u-picardie.fr/labo/lbpd/FyD.htm
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu