[gmx-users] QM/MM simulations

[Justin Lemkul]

On 4/24/14, 9:34 AM, Iamkaant wrote:
> Dear Justin!
> I'm very sorry to waste your time, but I've got an issue with ORCA support
> in GROMACS and hope you can help me.
> I'm installing gromacs-4.6.5 on my Dell PowerEdge R720 machine running Suse
> 12.3 x64 ( 3.7.10-1.28-desktop core). Installed version of ORCA is
> orca_3_0_0_linux_x86-64.
> 0)Previos steps according to the  installation guide
> <http://www.gromacs.org/Documentation/Installation_Instructions>  .
> 1) cmake
> /*cmake .. -DGMX_MPI=ON -DGMX_QMMM_PROGRAM=ORCA
> -DCMAKE_INSTALL_PREFIX=/opt/GROMACS-ORCA/ -DGMX_DOUBLE=on
> -DFFTW_LIBRARY=/usr/lib64/libfftw3.so.3*/
>
> I get following output. CMake-output.txt
> <http://gromacs.5086.x6.nabble.com/file/n5016041/CMake-output.txt>
> Contents of CMakeCache file is here.  CMakeCache.txt
> <http://gromacs.5086.x6.nabble.com/file/n5016041/CMakeCache.txt>
> The GMX_QMMM_PROGRAM variable is properly set in CMakeCache.txt:
>
> /*# grep GMX_QMMM_PROGRAM CMakeCache.txt *
> GMX_QMMM_PROGRAM:STRING=ORCA
> //ADVANCED property for variable: GMX_QMMM_PROGRAM
>   GMX_QMMM_PROGRAM-ADVANCED:INTERNAL=0/
>
> and in src/config.h
>
> /*# grep "#define GMX_QMMM_ORCA" src/config.h *
> #define GMX_QMMM_ORCA/
>
> 2) make & make install, everything without errors.
> 3) I create .mdp-file with following QMMM parameters
>
> /QMMM                     = yes
> QMMM-grps                = QMatoms
> QMmethod                 = RHF
> QMbasis                  = 3-21G
> QMMMscheme               = normal
> QMcharge                 = 0
> QMmult                   = 1/
>
> and  .ORCAINFO file (as stated in ORCA manual)
>
> /# Optimization step in the Lennard-Jones and point charges field of the MM
> atoms
> ! QMMMOpt
> # file containing the Lennard Jones coefficients for the Lennard-Jones
> interaction
> #%LJCoefficients "temp.LJ"
> # file containing the point charges for the electrostatic interaction
> #%pointcharges "temp.pc"
> %geom
> # calculate the exact Hessian before the first optimization step
> Calc_Hess true
> # in case of a TS optimization the updated Hessian of the previous
> # TS optimization run is read instead of calculating a new one
> read_temp_Hess true
> end/
>
> Variables are set as follows: $BASENAME=topol (name of .tpr and .ORCAINFO
> files) and  $ORCA_PATH=/opt/orca_3_0_0_linux_x86-64
> I run
> //*opt/GROMACS-ORCA/bin/mdrun_mpi_d -v*/
> and get
>
> /...
> Back Off! I just backed up md.log to ./#md.log.8#
> Reading file topol.tpr, VERSION 4.6.5 (double precision)
> Using 1 MPI process
> QM/MM calculation requested.
> Layer 0
> nr of QM atoms 24
> QMlevel: RHF/3-21G
> -------------------------------------------------------
> Program mdrun_mpi_d, VERSION 4.6.5
> Source code file: /home/ak87/Downloads/gromacs-4.6.5/src/mdlib/qmmm.c, line:
> 778
> Fatal error:
> Ab-initio calculation only supported with Gamess, Gaussian or ORCA.
> For more information and tips for troubleshooting, please check the GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> -------------------------------------------------------
> Halting program mdrun_mpi_d
> --------------------------------------------------------------------------
> MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
> with errorcode -1.
> NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
> You may or may not see output from other processes, depending on
> exactly when Open MPI kills them.
> --------------------------------------------------------------------------/
>
> It is quite strange since such problem was noted in earlier versions of
> GROMACS and have been solved by your patch. Therefore, it should not arise
> in version 4.6.5, particularly because all variables connected to ORCA are
> properly defined in Cmake config files.

My previous patch only addressed a missing #cmakedefine statement, but did not 
solve the problem entirely.  Please note the commentary on 
http://redmine.gromacs.org/issues/1286.

It's sort of a wonder that the code even compiled.  I got 4.6.2 to compile with 
my patch to config.h.cmakein, but when trying again with the latest master 
branch, compilation fails spectacularly with several errors in call_orca() and 
init_orca().  So clearly, the code has problems.  I patched the existing master 
branch to fix the compilation issues (please see 
https://gerrit.gromacs.org/#/c/3395/), but that is certainly no guarantee that 
the underlying code does anything sensible.  The fact that 4.6.5 compiles but 
still triggers the fatal error is worrisome, but given that the development 
branch acts as it should (i.e. chokes on incorrect code), there's not much point 
troubleshooting 4.6.5.

If you want to continue debugging, please check out the latest git master and 
apply the patch linked above.  It would be interesting to see if that has any 
effect, but again, given the fairly obvious problems in the function calls, 
you'll likely find more things to deal with.  The QM/MM interface is long 
overdue for an overhaul, and I believe there is work being done on that, but the 
code in the master branch should be considered a bit tenuous, at best.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================