To take some confusion out of this:
a) Gaussian has an NDDO implementation including d orbitals. This supports only PM6, paremeters for MNDO/d or AM1/d have not been added by Gaussian. Gaussian's NDDO implementation does indeed not support external electric fields (and thus electronic embedding) but for Wei's problem this does not matter.
b) All NDDO methods in sqm (Amber) use the same integral routines, irrespective of the Hamiltonian in use. When PM6, AM1/d and MNDO/d were added to sqm, I have carefully tested that integrals and matrix elements for MNDO/d and PM6 are correctly calculated and sqm reproduces energies and geometries as obtained with reference codes.
c) Given that we can trust the MNDO/d and PM6 implementation in Amber reasonably well (see b)), and Gaussian is able to converge the SCF for PM6 (and possibly other codes for MNDO/d, which I have not tested), the problems that Wei encounters are most likely due to the SCF convergence algorithms (not) implemented in sqm. Wei's simulation simply reached a point of the PM6 (and MNDO/d) potential energy surface that cannot be converged with what is presently implemented. There are many tricks to improve SCF convergence and if time permits, I will look into this at some point - others are welcome to help.
All the best,
Andy
On Sep 28, 2012, at 10:54 AM, Brian Radak wrote:
> Hi all,
>
> Our group (although not me personally) was responsible for the MNDO/d and
> AM1/d implementations in SQM, but not PM6, as that was added by the Walker
> group. A quick glance at the code (qm2_parameters.F90) shows that MNDO/d
> defaults to d-orbitals on Sulfur while AM1/d does not (only Phosphorus and
> Magnesium are parameterized for now and mixing those is somewhat dubious).
>
> If you are using the Gaussian interface then, as far as I know, there is no
> d-orbital implementation whatsoever. There is also no electronic embedding
> which would greatly complicate things in solution. NDDO type calculations
> in Gaussian and SQM are therefore not likely to be comparable at all.
>
> Regards,
> Brian
>
> On Fri, Sep 28, 2012 at 1:00 PM, Wei Lin <w5lin.ucsd.edu> wrote:
>
>> Thank you for your reply. We also did few more tests for the same system
>> with MNDO/d and AM1/d (both use d orbitals for Sulfur if we are not
>> mistaken). It appears that MNDO/d crashes also, but AM1/d and PM6
>> through Gaussian interface work well. We guess it may be related to the
>> way d-orbitals are handled in the PM6 sqm implementation.
>>
>> Thanks,
>> Wei
>>
>>
>> On 09/28/2012 08:50 AM, Matthew Zwier wrote:
>>> We've observed this also, with systems with large spatial extents and
>>> containing D orbitals. Localized systems work fine, systems without D
>>> orbitals of any extent work fine. This is occurring in solution for
>>> us.
>>>
>>> MZ
>>>
>>> On Fri, Sep 28, 2012 at 10:09 AM, Gerald Monard
>>> <gerald.monard.univ-lorraine.fr> wrote:
>>>> Hi,
>>>>
>>>> I run some tests on your files. It seems indeed that only PM6 crashes at
>>>> some point. Other semiempirical hamiltonians do not crash (I tested AM1,
>>>> PM3, PM3-PDDG, and RM1). May be there is a problem in the PM6 potential
>>>> energy surface describing your system.
>>>>
>>>> Sincerely,
>>>>
>>>> G. Monard.
>>>>
>>>> On 09/28/2012 10:58 AM, Gerald Monard wrote:
>>>>> Hi,
>>>>>
>>>>> IMHO, I'm not sure that it is really a SCF convergence problem in your
>> case.
>>>>> Your system is a charge separated system ( (CH3NH3^+)2 + NH4^+
>> (SO4H^-)2
>>>>> ) that you try to simulate in gas phase at T=800K. There is a really
>>>>> great chance that during an QM-MD, your system will blow up at some
>>>>> point. In a certain chemical sense, I am not sure that this is
>> something
>>>>> wrong.
>>>>>
>>>>> I made some quick tests with your system (minimization before running
>>>>> the MD, 300K instead of 800K, vrand=10 instead of 100, etc.), but at
>>>>> some point there is always a proton transfer between an amine group and
>>>>> a sulfate, then the system explodes because of the energy released.
>>>>> If your system was soaked into a aqueous solvent for example, I am
>>>>> pretty sure that this would not happen.
>>>>>
>>>>> Sincerely,
>>>>>
>>>>> G. Monard.
>>>>>
>>>>>
>>>>> On 09/28/2012 01:45 AM, Wei Lin wrote:
>>>>>> Hi Everyone,
>>>>>>
>>>>>> I tried to run MD simulation with PM6 Hamiltonian for a small charged
>>>>>> cluster contained d-orbital element. After ~ 38 steps of simulation,
>>>>>> the system blows up with error message as follows:
>>>>>>
>>>>>> QMMM: WARNING!
>>>>>> QMMM: Unable to achieve self consistency to the tolerances specified
>>>>>> QMMM: No convergence in SCF after 1000 steps.
>>>>>> QMMM: Job will continue with unconverged SCF. Warning energies
>>>>>> QMMM: and forces for this step will not be accurate.
>>>>>> QMMM: E = -0.6307E+06 DeltaE = 0.1152E-01 DeltaP = 0.3365E-02
>>>>>> QMMM: Smallest DeltaE = 0.3550E-03 DeltaP = 0.2442E-02 Step =
>> 787
>>>>>>
>>>>>> Then, I set up another simulation for the same system with Gaussian
>>>>>> interface. The simulation runs well.
>>>>>>
>>>>>> Andy Goetz suggested that can be related to SCF convergence algorithms
>>>>>> in sqm. I am wondering if anyone knows how to overcome this problem.
>>>>>>
>>>>>> The mdin (both in sqm and Gaussian interface), inpcrd, prmtop and run
>>>>>> files are attached.
>>>>>>
>>>>>> Thanks,
>>>>>> Wei
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> AMBER mailing list
>>>>>> AMBER.ambermd.org
>>>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>>>
>>>> --
>>>>
>> ____________________________________________________________________________
>>>>
>>>> Prof. Gerald MONARD
>>>> Theoretical Chemistry and Biochemistry Group
>>>> SRSMC, Université de Lorraine, CNRS
>>>> Boulevard des Aiguillettes B.P. 70239
>>>> F-54506 Vandoeuvre-les-Nancy, FRANCE
>>>>
>>>> e-mail : Gerald.Monard.univ-lorraine.fr
>>>> tel. : +33 (0)383.684.381
>>>> fax : +33 (0)383.684.371
>>>> web : http://www.monard.info
>>>>
>>>>
>> ____________________________________________________________________________
>>>>
>>>>
>>>> _______________________________________________
>>>> AMBER mailing list
>>>> AMBER.ambermd.org
>>>> http://lists.ambermd.org/mailman/listinfo/amber
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
>
>
> --
> ================================ Current Address =======================
> Brian Radak : BioMaPS
> Institute for Quantitative Biology
> PhD candidate - York Research Group : Rutgers, The State
> University of New Jersey
> University of Minnesota - Twin Cities : Center for Integrative
> Proteomics Room 308
> Graduate Program in Chemical Physics : 174 Frelinghuysen Road,
> Department of Chemistry : Piscataway, NJ
> 08854-8066
> radak004.umn.edu :
> radakb.biomaps.rutgers.edu
> ====================================================================
> Sorry for the multiple e-mail addresses, just use the institute appropriate
> address.
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
Dr. Andreas W. Goetz
Assistant Project Scientist
San Diego Supercomputer Center
Tel : +1-858-822-4771
Email: agoetz.sdsc.edu
Web : www.awgoetz.de
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 28 2012 - 15:30:03 PDT
