CCL:G: Gaussian09 Error

From: Jamin Krinsky <jamink(-)berkeley.edu>
Subject: CCL:G: Gaussian09 Error
Date: Fri, 30 Jul 2010 10:24:09 -0700
 Sent to CCL by: Jamin Krinsky [jamink[]berkeley.edu]
 Gadzikano,
 Before you said the frequency calculation worked with the small basis
 set. Now it is dying in the first part of your job (before the Link1)?
 Also, neither of us were suggesting that you do frequencies using a
 small basis set and then optimize again with a larger one, which is
 what your input is doing.
 To be clear, you are specifying that the job be run on 8 processors
 each on 8 nodes, for a total of 64 processors. Your are only
 specifying 128 MW of memory per 8 processors. Is that all the RAM
 available on each node? If there is more you should use it. Unless
 there is something strange about about your system (tiny disks, etc.)
 you should be able to run a frequency calculation on a 150-atom
 peptide using 6-31G(d) without a problem.
 Your issue MAY be a network problem with LINDA related to your two-part job.
 I would suggest doing the following two separate calculations, meaning
 run the first one, and when it finishes submit the second one using
 the optimized geometry.
 Job #1
 %NProcShared=8
 %nprocl=8
 %chk=start1.chk
 %mem=128mw
 # b3lyp/6-31G(d) opt
 title
 Molec. Spec.
 Now copy start1.chk to start2.chk
 Job #2
 %NProcShared=8
 %nprocl=8
 %chk=start2.chk
 %mem=(Specify the maximum amount of memory your can, per node of 8 processors)
 # b3lyp/6-31G(d) freq geom=allcheck guess=read
 In addition to diagnosing possible LINDA issues, this could get around
 some problems with memory allocation that Gaussian has with multi-step
 jobs.
 Jamin
 On Fri, Jul 30, 2010 at 2:30 AM, Gadzikano Munyuki gadzie===gmail.com
 <owner-chemistry*o*ccl.net> wrote:
 > Hi Jamin and Robert
 >
 > I did what you are suggesting before and i was getting the same error...
 > that was the input file i used.
 >
 > %NProcShared=8
 >  %nprocl=8
 >  %chk=start1.chk
 >  %mem=128mw
 >  # b3lyp/3-21G freq geom=allcheck guess=read
 >
 > --Link1--
 > %NProcShared=8
 >  %nprocl=8
 >  %chk=start1.chk
 >  %mem=128mw
 > #n b3lyp/6-31G opt=readfc geom=allcheck scf=(qc,maxcycle=1000)
 >
 >
 > On Thu, Jul 29, 2010 at 10:57 PM, Jamin Krinsky jamink-,-berkeley.edu
 > <owner-chemistry]_[ccl.net> wrote:
 >>
 >> Gadzikano, Robert,
 >>
 >> It is hard for me to tell whether it is a memory or disk space error,
 but
 >> the job is definitely running out of resources during the calculation
 of
 >> force constants. Robert is right that you normally don't need to
 calculate
 >> forces at the beginning of a geometry optimization (the opt=calcfc
 keyword),
 >> however you will need to do a frequency calculation at the end (which
 will
 >> fail similarly with the resources you are allocating). The suggestion
 of
 >> doing the frequencies using a smaller basis is not sound: a vibrational
 >> analysis on a structure optimized at a different level of theory is
 >> meaningless, because the stationary point will be located at different
 >> places in coordinate space for different methods (in this case basis
 set
 >> size).
 >>
 >> Regards,
 >> Jamin
 >>
 >>
 >> On Thu, Jul 29, 2010 at 12:09 PM, Robert McGibbon
 rmcgibbo[A]princeton.edu
 >> <owner-chemistry,ccl.net> wrote:
 >>>
 >>> Gadzikano,
 >>> I'm not entirely sure what your problem is, as I've never seen that
 >>> error. However, I would not recommend using the 'opt=calcfc'
 keyword.
 >>> Computing the force constants at the beginning is very costly and
 could be
 >>> where your error is coming. It might be preferable to use the force
 >>> constants from a freq job with the 3-21G basis set as the initial
 guess
 >>> Hessian instead - this can be done with the opt=readfc keyword.
 >>> Just my $0.02,
 >>> -robert
 >>>
 >>> On Thu, Jul 29, 2010 at 4:40 AM, Gadzikano Munyuki
 gadzie(!)gmail.com
 >>> <owner-chemistry.:.ccl.net> wrote:
 >>>>
 >>>> Hi all,
 >>>>
 >>>> I am trying to do an optimization job of a calcium-peptide
 complex using
 >>>> gaussian09. the peptide is quite big, it  has 10
 residues  . I managed to
 >>>> optimize the structure without any problems with a smaller
 basis set 3-21G.
 >>>> However when i try to adopt a bigger basis set i am getting the
 error below
 >>>>
 >>>> Any help will be very appreciated....
 >>>>
 >>>>
 >>>>  -----------------------------
 >>>> --------------
 >>>>  # b3lyp/6-31G  Opt=calcfc geom=allcheck guess=read
 >>>> scf=(qc,maxcycle=1000)
 >>>>
 >>>>
 >>>>  -------------------------------------------
 >>>>  1/10=4,29=7,30=1,38=1/1,3;
 >>>>  2/12=2,40=1/2;
 >>>>  3/5=5,11=2,14=-4,16=1,25=1,30=1,71=2,74=-5,116=-2/1,2,3;
 >>>>  4/5=1/1;
 >>>>  5/5=2,38=6,98=1/2;
 >>>>  8/6=4,10=90,11=11/1;
 >>>>  11/6=1,8=1,9=11,15=111,16=1/1,2,10;
 >>>>  10/6=1/2;
 >>>>  6/7=2,8=2,9=2,10=2,18=1,28=1/1;
 >>>>  7/8=1,10=1,25=1/1,2,3,16;
 >>>>  1/10=4,30=1/3;
 >>>>
 >>>>
 >>>> ............................................................
 >>>>
 >>>>  Calling FoFCou, ICntrl=  3107 FMM=T
 I1Cent=   0 AccDes= 0.00D+00.
 >>>>  FoFDir/FoFCou used for L=0 through L=1.
 >>>>  End of G2Drv Frequency-dependent properties
 file   721 does not exist.
 >>>>  End of G2Drv Frequency-dependent properties
 file   722 does not exist.
 >>>>          
 IDoAtm=11111111111111111111111111111111111111111111111111
 >>>>          
 IDoAtm=11111111111111111111111111111111111111111111111111
 >>>>          
 IDoAtm=11111111111111111111111111111111111111111111111111
 >>>>          
 IDoAtm=11111111111111111111111111111111111111111111
 >>>>          
 Differentiating once with respect to electric field.
 >>>>
                
 with respect to dipole field.
 >>>>          
 Differentiating once with respect to nuclear coordinates.
 >>>>           There
 are   585 degrees of freedom in the 1st order CPHF.
 >>>> IDoFFX=5.
 >>>>     585 vectors produced by pass  0 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 1.01D+05 2.93D+02.
 >>>>  AX will form    18 AO Fock derivatives at
 one time.
 >>>>     585 vectors produced by pass  1 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 1.16D+04 1.47D+01.
 >>>>     585 vectors produced by pass  2 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 4.64D+01 5.44D-01.
 >>>>     582 vectors produced by pass  3 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 1.01D-01 1.97D-02.
 >>>>     582 vectors produced by pass  4 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 1.18D-04 4.58D-04.
 >>>>     327 vectors produced by pass  5 Test12=
 2.74D-13 1.00D-09 XBig12=
 >>>> 9.50D-08 1.64D-05.
 >>>>      16 vectors produced by pass  6
 Test12= 2.74D-13 1.00D-09 XBig12=
 >>>> 6.51D-11 3.18D-07.
 >>>>       3 vectors produced by pass 
 7 Test12= 2.74D-13 1.00D-09 XBig12=
 >>>> 4.67D-14 7.29D-09.
 >>>>  Inverted reduced A of dimension  3265 with in-core
 refinement.
 >>>>  NIJ > Max2 in MMCore.
 >>>>  Error termination via Lnk1e in
 >>>> /opt/gridware/gaussian/g09/linda-exe/l1002.exel at Thu Jul 22
 19:33:56 2010.
 >>>>  Job cpu time:  1 days 23 hours  4 minutes 27.7
 seconds.
 >>>>  File lengths (MBytes):  RWF=  31540
 Int=      0 D2E=      0
 Chk=     96
 >>>> Scr=      1
 >>>> failed to open execfile
 >>
 >>
 >>
 >> --
 >> Jamin L Krinsky, Ph.D.
 >> Molecular Graphics and Computation Facility
 >> 175 Tan Hall, University of California, Berkeley, CA 94720
 >> jamink,berkeley.edu, 510-643-0616
 >> http://glab.cchem.berkeley.edu
 >>
 >
 >
 --
 Jamin L Krinsky, Ph.D.
 Molecular Graphics and Computation Facility
 175 Tan Hall, University of California, Berkeley, CA 94720
 jamink*o*berkeley.edu, 510-643-0616
 http://glab.cchem.berkeley.edu