[gmx-users] Re: QMMM with ORCA

[Gerrit Groenhof]

Dear Xiaohu,

Thanks for bringing this up. The comment has been there for ever. Since I could not think of an application where one would not be using pbc at the time. However, your clusters prove me wrong.

In any case, as a work-around, you may want to use a bigger box, with long enough cut-offs, no PME.

Hope this gets you started,

Gerrit

On 27 Jan 2011, at 21:02, Xiaohu Li wrote:

> Dear GMX code writters,
>          Could anyone tell me why this comments in the code mdlib/qmmm.c appear?(version 4.5.3)
> at line ~ 714, in the beginning of subroutine update_QMMMrec
> ======================================================
> /* updates the coordinates of both QM atoms and MM atoms and stores
>    * them in the QMMMrec.  
>    *
>    * NOTE: is NOT yet working if there are no PBC. Also in ns.c, simple
>    * ns needs to be fixed!  
> */
> ======================================================
>         First of all, I have a non-PBC job with electronic embedding and it fails right in this routine. So I guess this is the reason, for non-PBC, it crashes. But for oniom, it went through. Is there any simple fix to have the non-PBC qmmm work? I'm also interested in doing non-PBC calculation(large cluster).
> 
> 
> Xiaohu
> 
> On Thu, Jan 27, 2011 at 9:11 AM, Christoph Riplinger <cri at thch.uni-bonn.de> wrote:
> Hi Xiaohu,
> 
> We are using gromacs/ORCA quite a lot. It works like any of the other interfaced programs and whether you should use it depends on your needs to the QM part of the QM/MM calculation.
> 
> I am using gromacs 4.0.7, but I downloaded the 4.5.3 version to test it.
> 
> I could not reproduce your problem with ONIOM. For the electrostatic embedding bug, I agree. This didn't work in my test calculation, but my gdb told me that the abort occured before the qm_orca.c was even called. I suspect the bug is in the general qmmm.c code. I do not have access to the other QM-programs and cannot test this, so if you have, could you please run the same job with a different QM-program. 
> 
> Hope that helps
> 
> Christoph
> 
> 
> 
> 
> On 01/26/2011 06:48 AM, Xiaohu Li wrote:
>> 
>> Hi, All,
>>      I'm trying to see if anybody has experience of using the interface of gromacs and ORCA(since it's free). I know that the following link gave information on how
>> http://wwwuser.gwdg.de/~ggroenh/qmmm.html#code
>>      But.....But, the gromacs in the above link is quite old(3.2). I download the latest 4.5.3 and followed the instructions in the above link and I was trying to optimize an simple cluster(no pbc) where part of it are treated using QM. here is the example mdp file
>> =========================================================================================================================
>> title               =  cpeptide
>> integrator          =  steep   ;integrator includes energy minimization algorithms
>> dt                  =  0.002    ; ps !
>> nsteps              =   10000
>> nstlist             =  1
>> ns_type             =  simple
>> rlist               =  3.0
>> rcoulomb            =  3.0
>> coulombtype         = cut-off
>> vdwtype             = cut-off
>> rvdw                = 3.0
>> pbc                 =  no
>> periodic_molecules  =  no
>> constraints         = none
>> energygrps          = qm_part mm_part
>> ; QM/MM calculation stuff
>> QMMM = yes
>> QMMM-grps = qm_part
>> QMmethod = rhf
>> QMbasis = 3-21G
>> QMMMscheme = oniom
>> QMcharge = 0
>> QMmult = 1
>> ;
>> ;       Energy minimizing stuff
>> ;
>> emtol               =  60   ; minimization thresold (kj/mol.nm-1)    1 hartree/bohr= 49614.75241 kj/mol.nm-1  1 kj/mol.nm-1=2.01553e-5 hartree/bohr
>> emstep              =  0.01  ; minimization step in nm
>> =========================================================================================================================
>> I set up the BASENAME and ORCA_PATH as told in the instruction.
>> first of all, the normal electronic embedding just simply gave segmentation fault error right after the it prints information on number of steps of optimization.
>> 
>> So I switch to ONIOM, this time, at least, orca is called and energy and gradient are both generated. However, when it comes to read the energy and gradient, it always crashes when tried to read gradient, this is at line 346 source code src/mdlib/qm_orca.c
>> ============================================
>> sscanf(buf,"%lf\n", &QMgrad[k][XX]);
>> ============================================
>> a segmentation fault error is printed. If I replace the &QMgrad[k][XX] by an temporary variable temp 
>>  sscanf(buf,"%lf\n", &temp);
>> temp gets the correct value and if I use,
>> QMgrad[k][XX]=temp 
>> and tries to print QMgrad[k][XX], a bus error will be printed.
>> I did some research online, seems that usually this implies an memory bug in the code which is the most difficult bug one can ever encounter.
>> So has anyone successfully used gromacs and orca to do QMMM?
>> Generally, would anyone recommend using gromacs to do QMMM?
>> 
>> Cheers,
>> Xiaohu
> 
> 
> 
> 
> -- 
> =================================================
> Xiaohu Li
> Post-doctoral Research Associate
> Department of Chemistry
> Northwestern University
> 2145 Sheridan Road
> Evanston IL 60208
> U.S.A
> =================================================
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110128/cdc7bc63/attachment.html>