Hi Karl,
I have rescanned the ct-c-c-ct MP2 profile of 2,3-butanedione with three
different basis sets and obtained the following results:
https://docs.google.com/open?id=0BwkQMO2EgiyoMWE0Y2M2N2MtMjliYi00ZDYwLTlhY2MtNTQ2NWE3NmQyNTgw
Again, the profile obtained with MP2/6-31G* (the same method reported by
Wang et al, 2004 in the development of GAFF), shows the same local minima
at +- 30 degrees I was seeing before. This is quite different from the
profile shown in Figure 3 of Wang et al.
The MP2/6-31++G(d,p) profile I calculate seems exactly like the one you
produced, therefore I would speculate that you would find the same profile
at MP2/6-31G* as I did. The optimized geometries with a dihedral angle of
-60 degrees are shown in the linked-to figure; they are very similar except
for slight changes in hydrogen positions.
Rescanning the PK values of the V2 potential (using your partial atomic
charges) and comparing the GAFF/MM profile to a
MP4/aug-cc-pVDZ//MP2/aug-cc-pVDZ QM profile, I find that PK=1.4 now creates
a best fit between the MM and QM profiles with an RMSD of 0.137 over 13
points at 30 degree intervals. This is close to the PK value of 1.2
determined with Parmscan.
https://docs.google.com/open?id=0BwkQMO2EgiyoNmRhYTE2ZjMtZTQyMC00OWU5LWIzM2ItMjAzYTRhN2M0NzJl
Based on these results, I am troubled by the discrepancies between my QM
results and those reported in Wang et al. In the parametrization I am
trying to do, I wanted to stay as close as possible to the methods used in
the GAFF paper, however, it now seems that it would be wiser to switch to a
more complete basis set.
-ryan
On Wed, Feb 8, 2012 at 8:11 AM, Karl N. Kirschner <
kkirsch.scai.fraunhofer.de> wrote:
> Hi Ryan,
>
> I can basically reproduce your MM results and the Wang et al QM results.
> However, I took a slightly different approach for the quantum calculations
> and the partial atomic charges. Note that due to symmetry, one only needs
> to do a heavy atom rotation from 0-180 degrees (I also did this in 30
> degree increments). I wasn't able to do the single-point calculation at the
> MP4 level, instead I opted for a larger basis set at the MP2 level.
>
> For the QM constraint optimization about the C-C-C-C torsion angle I used
> MP2//6-31++G(d,p). I performed single-point calculations on these
> geometries at MP2/aug-cc-pVDZ (i.e. MP2/aug-cc-pVDZ//MP2//6-31++G(d,p)).
> For partial atomic charges I used R.E.D., with HF/6-31G(d) electrostatic
> potential using the Connolly surface algorithm, and a 2 stage RESP fit
> using weighting factors qwt=.0005/.001. (3 molecular orientations were used
> in the R.E.D. calculations.) All of this was done using GAMESS.
>
> My partial atomic charges are slightly different than yours, but
> ultimately I don't think it matters too much for the internal heavy atom
> rotation of this molecule. Regardless, here are the charges:
>
> 1 C1 -0.7300 0.2420 0.0000 c 1 MOL 0.5239
> 2 C2 0.7300 -0.2420 0.0000 c 1 MOL 0.5239
> 3 C3 1.7980 0.8190 0.0000 c3 1 MOL -0.2739
> 4 H4 1.6930 1.4560 0.8710 hc 1 MOL 0.0861
> 5 H5 1.6930 1.4560 -0.8710 hc 1 MOL 0.0861
> 6 H6 2.7700 0.3460 0.0000 hc 1 MOL 0.0861
> 7 C7 -1.7980 -0.8190 0.0000 c3 1 MOL -0.2739
> 8 H8 -1.6930 -1.4560 0.8710 hc 1 MOL 0.0861
> 9 H9 -1.6930 -1.4560 -0.8710 hc 1 MOL 0.0861
> 10 H10 -2.7700 -0.3460 0.0000 hc 1 MOL 0.0861
> 11 O11 -0.9520 1.4110 -0.0000 o 1 MOL -0.5082
> 12 O12 0.9520 -1.4110 -0.0000 o 1 MOL -0.5082
>
> Attached you can see the resulting potential energy curves. I can
> essentially reproduce the gaff curve and your curve. What I noticed is that
> the resulting molecular mechanics curve is extremely sensitive to the V2
> parameter! By changing the parameter from 1.2 to 2.2 the resulting curve at
> 0 degrees goes from a transition state to a high energy minima.
>
> Concerning the QM calculations, both MP2 calcuations have no local minima
> at 30 degrees, which align well with Wang et al MP4 curve. It is clear that
> the shape of this potential is dependent upon the basis set, and most
> likely the theory level too. I would double check your 30 degree
> conformation (perhaps there is an issue with the improper torsion geometry
> about the ketone functionality).
>
> Concerning the RMSD. I obtain a lower RMSD (averaged over all 7
> conformations, not just the minima) for gaff than for your parameter, which
> is opposite of what you find. For comparison to my MP2//6-31++G(d,p)
> geometries, I obtain an RMSD value of 0.056 for gaff, and 0.058 when using
> your V2.
>
> I don't think this answers your question(s) directly, but I hope that
> helps some :) .
>
> Cheers,
> Karl
>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Feb 10 2012 - 10:30:02 PST
