Troubleshooting Gaussian calculations

Troubleshooting Gaussian calculations
by Kathrin Helen Hopmann — last modified Jul 09, 2010 02:26 PM
A list of error messages and possible solutions
Gaussian calculations can fail with various error messages. Some error messages from .out and .log files – and possible solutions – have been compiled here to facilitate problem solving.
These are divided into:
Syntax and similar errors
Memory and similar errors
Convergence problems
Errors in solvent calculations
Errors in log files
Syntax and similar errors:
End of file in ZSymb.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l101.exe
Solution: The blank line after the coordinate section in the .inp file is missing.
Unrecognized layer “X”.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l101.exe
Solution: Error due to syntax error(s) in coordinate section (check carefully). If error is “^M”, it is caused by DOS end-of-line characters (e.g. if coordinates were written under Windows). Remove ^M from line ends using e.g. emacs. To process .inp files from command line, use sed -i  ‘s/^M//’ File.inp  (Important: command does not work if ^M is written as characters – generate ^M on command line using ctrl-V ctrl-M).
Solution: Check .inp carefully for syntax errors in keywords
RdChkP:  Unable to locate IRWF=0 Number=  522.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l401.exe
FileIO operation on non-existent file.
[…] Error termination in NtrErr:
NtrErr Called from FileIO.
Solution: Operation on .chk file was specified (e.g. geom=check, opt=restart), but .chk was not found. Check that:
%chk= was specifed in .inp
.chk has the same name as .inp
.chk is in the same directory as .inp
run script transports .chk to temporary folder upon job start. Run scripts downloaded here should do this.
The combination of multiplicity N and   M electrons is impossible.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l301.exe
Solution: Either the charge or the multiplicity of the molecule was not specified correctly in .inp.
 Specify only one instance of %LindaWorkers and do not use %NProcLinda.
 Contradictory % commands.
Solution: You probably have lines in your input file specifying %NProcLinda and %NProcShared while trying to run g09. Remove these lines and resubmit your job. See more info about this: Gaussian 09
Memory and similar errors:
Out-of-memory error in routine RdGeom-1 (IEnd=       1200001 MxCore=          2500)
Use %mem=N MW to provide the minimum amount of memory required to complete this step
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l101.exe
Not enough memory to run CalDSu, short by     1000000 words.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l401.exe
[…] allocation failure:  
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l1502.exe
Solution: Specify more memory in .inp (%mem=Nmb). Possibly, also increase pvmem value in run script. Especially solvent calculations can exhibit allocation failures and explicit amounts of memory should be specified.
galloc:  could not allocate memory.
Solution: The %mem value in .inp is higher than pvmem value in run script. Increase pvmem or decrease %mem.
Probably out of disk space. Write error in NtrExt1
 Solution: /scratch space is most likely full. Delete old files in temporary folder.  
PrsmSu:  requested number of processors reduced to:   2 ShMem   1 Linda
By default, Gaussian reduces the number of processors on a parallel calculation whenever there is not enough memory allocated to perform a given amount of computational work using all specified processors. For example, if one specifies:
this means that there will be about 1 GB of memory available per processor in this parallel run. If a given part of the calculation requires 2 GB of memory per processor to run in parallel, then the program will automatically reduce the number of processors to 2, in order for the calculation to continue. If one wants to use all 4 processors in such case, one would have to allocate more memory, e.g. %mem=8GB.
Convergence problems:
Density matrix is not changing but DIIS error= 1.32D-06 CofLast= 1.18D-02.
The SCF is confused. Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/linda-exe/l502.exel
Solution: Problem with DIIS. Turn it off completely, e.g. using SCF=qc, or partly by using SCF=(maxconventionalcycles=N,xqc), where N is the number of steps DIIS should be used (see SCF keyword).
Convergence criterion not met. SCF Done:  E(RHF) =  NNNNNNN     A.U. after  129 cycles
[…] Convergence failure — run terminated. Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/linda-exe/l502.exe
Solution: One SCF cycle has a default of maximum 128 steps, and this was exceeded without convergence achieved. Possible solution: In the route section of input file, specify SCF=(MaxCycle=N), where N is the number of steps per SCF cycles. Alternatively, turn of DIIS (e.g. by SCF=qc) (see SCF keyword).
Problem with the distance matrix.
Error termination via Lnk1e in /pkg/gaussian/g03/l202.exe
 Solution: Try to restart optimization from a different input geometry.  
New curvilinear step not converged. Error imposing constraints
Error termination via Lnk1e in /pkg/gaussian/g03/l103.exe
Solution: Problem with constrained coordinates (e.g. in OPT=modredun calculation). Try to restart optimization from a slightly different input geometry.
Optimization stopped.   — Number of steps exceeded,  NStep= N
[..] Error termination request processed by link 9999.
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l9999.exe
Solution: Maximum number of optimization steps is twice the number of variables to be optimized. Try increasing the value by specifying OPT=(MaxCycle=N) in .inp file, where N is the number of optimization steps (see OPT keyword). Alternatively, try to start optimization from different geometry.
Errors in solvent calculations:
AdVTs1: ISph= 2543 is engulfed by JSph= 2544 but Ae( 2543) is not yet zero!
Error termination via Lnk1e in /global/apps/gaussian/g03.e01/g03/l301.exe
Solution: Problem is related to building of the cavity in solvent calculations. One possible solution is to change the cavity model (default in g03 is UAO, can be changed by adding RADII keyword in section below coordinates in the .inp file, e.g. RADII=UFF, see SCRF keyword).
Hydrogen   X has 2 bounds. Keep it explicit at all point on the
potential energy surface to get meaningful results.
Solution: In UAO cavity model, spheres are placed on groups of atoms, with hydrogens assigned to the heavy atom, they are bound to. If assignment fails (e.g. because heavy atom-H bond is elongated), cavity building fails. Possible solutions: a) use cavity model that also assigns spheres to hydrogens (e.g. RADII=UFF) or b) Assign a sphere explicity on problematic H atom (use SPHEREONH=N, see SCRF keyword)
=>> PBS: job killed: walltime N exceeded limit M
signal number 15 received.
Solution: Job did not finish within specified wall time. Retrieve .out and .chk files from temporary folder /global/work/$USER/$JOB (or $PBS_JOBID) and restart calculation if possible (using e.g. opt=restart or scf=restart).  
cp: cannot stat $JOB.inp: No such file or directory
Solution: The .inp file is not in the directory from where the job was submitted (or its name was misspelled during submission. If error reads: cp: cannot stat $JOB.inp.inp, the .inp file was submitted with extension).
ntsnet: unable to schedule the minimum N workers
Solution: The value of %NprocLinda=N in the .inp file is higher than the number of nodes asked for during submission. Make sure these values match.
Connection refused […] died without ever signing in
Sign in timed out after 0 worker connections. Did not reach minimum (N), shutting down
Solution: Error appears if you run parallel calculations but did not add this file to your $HOME directory: .tsnet.config containing only the line: Tsnet.Node.lindarsharg: ssh (see also guidelines for submission).


슈퍼컴퓨팅 전문 기업 클루닉스/ 상무(기술이사)/ 정보시스템감리사/ 시스존 블로그 운영자

You may also like...

페이스북/트위트/구글 계정으로 댓글 가능합니다.