Hi Andy,
Thanks for looking into this. I have noticed that many sqm SCF
failures occur with ligands that include phosphate groups, which is in
line with your observation of SCF convergence difficulties with highly
negatively charged structures.
The ANP structure in the context of the 2ews PDB entry doesn't really
allow for the "optimized" geometry delivered by mopac with it's
internal H-bonds to the phosphates, since in 2ews those phosphates are
occupied by H-bonding to the protein backbone and coordinating a
magnesium ion. Unfortunately sqm couldn't possibly run on ANP in the
larger context of the entire 2ews system, where the negative phosphate
charges would be stabilized by ANP's environment. Maybe if the
magnesium ion had been included in the sqm input (with the appropriate
formal charge change) it would have worked?
Anyway, here's hoping for a trickier sqm sometime in the future! :-)
--Eric
On Jan 19, 2011, at 11:34 PM, Andreas Goetz wrote:
> Hi Eric,
>
> I had a look at the geometries and sqm input files. The first thing
> I noticed is that the unprotonated system has a charge of -4. In
> general, SCF convergence for negatively charged systems is more
> difficult to achieve than for neutral or positively charged systems
> (although this problem is more pronounced for ab initio methods with
> flexible basis sets than for semiempirical methods).
>
> That being said, I managed to optimize the geometry with mopac2009.
> However, also mopac had serious SCF convergence problems in the
> first iterations of the geometry optimization. sqm does not have
> problems with the geometry that has been optimized by mopac
> (attached). If you look at the optmized geometry, the phosphate
> group makes hydrogen bonds to the hydroxyl groups which should
> stabilize the negative charge in a similar fashion as protonation.
>
> Conclusion: the molecule at your input geometry has a very tricky
> electronic structure and sqm needs implementation of additional
> tricks next to DIIS (mopac2009 unfortunately does not tell exactly
> what it is doing).
>
> All the best,
> Andy
> <am1-mopac2009-opt.xyz>
>
> On Jan 19, 2011, at 1:08 PM, Eric Pettersen wrote:
>
>> Indeed it is an SCF convergence problem. I guess my point is that
>> this file fails to converge while the other file doesn't, despite the
>> fact that the only difference between them is the proton added to O3G
>> (in the file that converges). The coordinates are otherwise
>> identical
>> (and the charge naturally differs by 1). The starting heavy atom
>> geometry comes directly from the 2.05A 2ews crystal structure.
>>
>> Actually I believe that Note 6 on p. 84 is out of date in that all
>> the
>> current default settings for sqm are now the same or even more lax
>> than the values given on p. 84. Perhaps someone should revise that
>> note?
>>
>> --Eric
>>
>> On Jan 19, 2011, at 12:21 PM, Qiantao Wang wrote:
>>
>>> This is the SCF convergence problem. You may change the convergence
>>> criteria
>>> see if that helps. Please refer to Note 6 on p. 84 of the
>>> AmberTools Users'
>>> Manual for more details.
>>> Make sure your starting geometry is reasonable.
>>>
>>> Best
>>> Qiantao
>>>
>>> On Tue, Jan 18, 2011 at 4:30 PM, Eric Pettersen <pett.cgl.ucsf.edu>
>>> wrote:
>>>
>>>> On Jan 18, 2011, at 12:55 PM, Qiantao Wang wrote:
>>>>
>>>>
>>>>> Did you get any error messages?
>>>>
>>>> The end of the sqm.out file that failed is:
>>>>
>>>>
>>>> --------------------------------------------------------------------------------
>>>> RESULTS
>>>>
>>>> --------------------------------------------------------------------------------
>>>>
>>>>
>>>> QMMM: ERROR!
>>>> QMMM: Unable to achieve self consistency to the tolerances
>>>> specified
>>>> QMMM: No convergence in SCF after 1000 steps.
>>>> QMMM: E = -0.1136E+07 DeltaE = -0.9349E+02 DeltaP = 0.3035E+00
>>>> QMMM: Smallest DeltaE = -0.9270E+00 DeltaP = 0.4112E+00 Step
>>>> = 35
>>>>
>>>> --Eric
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> AMBER-Developers mailing list
>>>> AMBER-Developers.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>>>>
>>> _______________________________________________
>>> AMBER-Developers mailing list
>>> AMBER-Developers.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>>
>>
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
> --
> Dr. Andreas W. Goetz
> San Diego Supercomputer Center
> Tel : +1-858-822-4771
> Email: agoetz.sdsc.edu
> Web : www.awgoetz.de
>
>
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Jan 20 2011 - 12:00:04 PST
