Michelle Coote

Computer Aided Chemical Design

Software Usage

Gaussian Error Message Translation

Leaf on Thu Mar 17 08:21:10 2011

The following are copied from http://biowulf.nih.gov/apps/gaussian/
Gaussian errors are not always straightforward to interpret. Something as simple as a "file not found" can seem baffling and cryptic. Here is a collection of errors and their translations:

Gaussian Error Translation to English
Error termination in NtrErr:
ntran open failure returned to fopen.
Segmentation fault
Can't open a file.
Internal consistency error detected in FileIO for unit 1 I= 4 J=0 I Fail= 1. Gaussian is limited to 16 GB of scratch space on the 32-bit nodes.
Out-of-memory error in routine UFChkP (IEnd= 12292175 MxCore= 6291456)
Use %Mem=12MW to provide the minimum amount of memory required to complete this step.
Error termination via Lnk1e at Thu Feb 2 13:05:32 2006.
Default memory (6 MW, set in $GAUSS_MEMDEF) is too small for unfchk.
galloc: could not allocate memory.: Resource temporarily unavailable Not enough memory.
Out-of-memory error in routine... Not enough memory.
End of file in GetChg.
Error termination via Lnk1e ...
Not enough memory.
IMax=3 JMax=2 DiffMx= 0.00D+00
Unable to allocate space to process matrices in G2DrvN:
NAtomX= 58 NBasis= 762 NBas6D= 762 MDV1= 6291106 MinMem= 105955841.
Gaussian has 6 MW free memory (MDV1) but requires at least 106 MW (MinMem).
Estimate disk for full transformation -677255533 words. Semi-Direct transformation. Bad length for file. MaxDisk has been set too low.
Error termination in NtrErr:
NtrErr Called from FileIO.
The calculation has exceeded the maximum limit of maxcyc.
Erroneous read. Read 0 instead of 6258688.
fd = 4
Disk quota or disk size exceeded. Could also be disk failure or NFS timeout.
Erroneous write. Write 8192 instead of 12288.
fd = 4
orig len = 12288 left = 12288
Disk quota or disk size exceeded. Could also be disk failure or NFS timeout.
PGFIO/stdio: Permission denied
PGFIO-F-/OPEN/unit=11/error code returned by host stdio - 13.
File name = /gaussian/Gau-#####.inp
In source file ml0.f, at line number 177
The user does not have write permission for $GAUSS_SCRDIR.

Choosing A Number Of Processors

Leaf on Thu Mar 17 08:11:32 2011

The following are copied from http://biowulf.nih.gov/apps/gaussian/
However, not all calculation types parallelize well or at all. In fact, most run best as single-threaded processes. Only HF, DFT, CCSD(T), CIS, and MP2/MP4 jobs will benefit from running with a %NProcShared=4 on dual core nodes. All others should be run with %NProcShared=2 (or 1).
The following table shows the best use of Gaussian with respect to the number of processors:

Method Energy Gradient / Opt Freq / Hessian
HF 4 4 4
HDFT 4 4 4
Pure DFT 4 4 4
MP2 4 3 1-2
MP3 1 1  
MP4 2-4    
MP5 1    
CCD 1 1  
CCSD 1 1  
CCSD(T) 2-4    
CIS 4 3  
CISD 1 1  
AM1 1 1  

Gaussian Utility Program

Leaf on Thu Mar 17 09:24:37 2011

The following are copied from http://biowulf.nih.gov/apps/gaussian/
There are a number of utility programs available. These programs can be run both on the login node and on the cluster.
NOTE: The utilities are by default the Gaussian09 version. To run the Gaussian03 version of the utility, add the prefix 'g03' to the utility name. For example, to run ghelp for g03, type g03ghelp.
For more information, type ghelp at the prompt.

Program Function
c8609 Converts checkpoint files from previous program versions to Gaussian09 format.
chkchk Displays the route and title sections from a checkpoint file.
cubegen Standalone cube generation utility.
cubman Manipulates Gaussian-produced cubes of electron density and electrostatic potential (allowing them to be added, subtracted, and so on).
formchk Converts a binary checkpoint file into an ASCII form suitable for use with visualization programs and for moving checkpoint files between different types of computer systems.
freqchk Prints frequency and thermochemistry data from a checkpoint file. Alternate isotopes, temperature, pressure and scale factor can be specified for the thermochemistry analysis.
freqmem Determines memory requirements for frequency calculations.
gauopt Performs optimizations of variables other than molecular coordinates.
ghelp On-line help for Gaussian.
mm Standalone molecular mechanics program.
newzmat Conversion between a variety of molecular geometry specification formats.
rwfdump Dumps the file index and data from a read-write or checkpoint file.
testrt Route section syntax checker and non-standard route generation.
unfchk Convert a formatted checkpoint file back to its binary form (e.g., after moving it from a different type of computer system).

Gaussian Glitch

Anya on Thu Mar 17 13:24:37 2011

Just want to update you on our progress with Rika in solving the Gaussian error (g_write...). I was getting this error for QCISD(T) optimisation+frequencies jobs and for CCSD(T) single point jobs. I was running them with Gausiian 09 rev. A.02. However, we tried Gaussian 09 Rev. B.01 and they finished successfully. The allocations (lf, lw, time) were regular, nothing extraordinary. I also had to change the gaussian version in usa script to make jobs run with B.01. Another important thing was to run those calcs as separate jobs (instead of multi-step ones with --Link1--).


Ivan Rostov on Thu Mar 17 13:58:36 2011

do you aware that the last two versions of GAMESS (dated by March and October 2010) have a bug in the the two electron integral transformation part? The bug affect CCSD and MP2 results for systems including elements not in first two rows of the Periodic Table, if program run in parallel. See, here for more information . The GAMESS-US team promises come up with a fix anyday soon. The March 2010 version is installed on our system ( called by the rungms.11 script)

Gaussian Parallel JOBS

Rika on Thu Mar 17 14:01:41 2011

I've run a few testjobs and understand a lot more about the MP2 algorithms now. Basically, there are three algorithms specified in input FullDirect,Direct and SemiDirect. If you give it a lot of memory Gaussian chooses the FullDirect algorithm, of which it says in the user manual:
"This is seldom a good choice, except for machines with very large main memory and limited disk."
For your 4PE job described below you gave it 16Gb memory which for the FullDirect algorithm was only enough to store 1 OVN intermediate requiring 60+15 passes compared to 12 + 4 passes for the 1PE job i.e. about 5 times as much computation. That same 4PE job but with 42 Gb memory allowing it to do 4 occupied batches at a time takes 99:43:14. Being FullDirect the jobs only took 1.3 Gb disk. This can be reduced further by forcing the in-between algorithm (as described below for 8 processors). The best performance comes from the intermediate (Direct) algorithm where Gaussian tries to make best use of both memory and disk driven by fitting batches of OVN in memory and pushing intermediates out to disk. I played around with some memory disk combinations and got the following:

Memory Disk No. Passes Walltime
Your 1PE job 15Gb 1.3Gb 12 + 4 167 hrs
8PE (FullyDirect) 46Gb 1.3Gb 32 + 10 62 hrs
8PE (SemiDirect) 9Gb 9Gb 80 + 80 135 hrs
8PE (Direct) 2Gb 275Gb 7 + 7 44 hrs
8PE (Direct) 2Gb 388Gb 5 + 5 40 hrs
8PE (Direct) 2Gb 479Gb 4 + 4 40 hrs
8PE (Direct) 40Gb 388Gb 5 + 5 73 hrs

Note that the trade-off between disk use and no. passes for the 40 hour jobs. Note also that giving excessive memory doesn't necessarily help. I'm thinking that may be a load balancing thing. Bottom line is when comparing scaling you have to compare like with like (same amount of compute as judged by no. of passes), which is further complicated in the MP2 algorithm which hides how much of the intermediates are recomputed, stored to memory or stored to disk.

ADF convergence for dissociative species

Leaf on Fri Mar 18 14:44:40 2011

Should always check whether adf is converged when running solvent calculations.
If ADF has convergence problem, usually in fragment.run, put the following keywords before SAVE line.

! The default value of 0.2 may be too high for some systems
! so we change it to 0.1
Mixing 0.1
! The default value DIIS ok=0.5 might make DIIS kick in too soon
! thus making it unstable. The cycle starting from which DIIS is
! enforced may also be shifted forward.
DIIS ok=0.01 cyc=20
! Set the max number of SCF cycles to 100 but you can use an
! even larger value.
Iterations 100

G3(MP2)-RAD energy for H radical

Anya on Thu Mar 24 14:20:23 2011

The electron correlation contribution to the energy of hydrogen radical is zero, and therefore the energies we get from HF and ROMP2 and URCCSD(T) calculations for G3(MP2)-RAD procedure should be equal (though, knowing that there is no need to do them all, HF is enough - but we often use standard scripts that do this for us). However, if you do calculate these energies for H- radical, you will get -0.4982329 hartrees for ROHF, ROMP2 and URCCSD(T) /6-31G(d), but ROMP2/GTMP2Large yields in - 0.4998179 hartrees. Thus, for G3(MP2)-RAD electronic energy instead of -0.5001209 hartrees (when all energies are the same) you will get -0.5017059 hartrees. This means your overall electronic energy would be underestimated for 4.2 kJ/mol. This problem occurs when using Gaussian both 2003 RevE.01 and 2009 RevB.01.

G3(MP2)-RAD energy for H radical

Dave on Thu Mar 24 15:23:15 2011

The G3(MP2)-RAD electronic energy for a one-electron system (neglecting the HLC) is HF/6-31G(d) + HF/gtmp2large - HF/6-31G(d) = HF/gtmp2large For the hydrogen atom, this is -0.4998179 hartrees. Using g09, UCCSD(T), ROCCSD(T), UMP2 and ROMP2 give same results for a given basis set for hydrogen, as they should. Your G3(MP2)-RAD and the HF/gtmp2large energies differ by 1.888 millihartrees, which is exactly the Higher Level Correction (HLC) for a system with one alpha electron and no beta electrons. Therefore, there is no problem. See J. Chem. Phys. (2003), 118(11), 4849 for details.

TChem usage

ilhan on Fri Apr 1 04:26:31 2011

Is there an easy way to calculate partition functions using TChem without gaussian output files but the necessary data in it, such as optimized geometries, frequencies, energies, etc.?

tchem usage

on Fri Apr 1 12:39:46 2011

Read through the tchem perl script to see what it does. thermochem-auto.f is probably what you want to hack. Dave

Create Conformers from XYZ

Leaf on Wed Apr 13 11:12:50 2011

This is a easier way to create your conformers!
The code "rotbond.pl" can be downloaded from the Programs page.

To use, simply have the bond desired for rotation saved in a file "$mol.bonds"
And given a xyz file that is called "$mol.xyz"
rotbond.pl $mol

If you have a transition structure with elongated bond, you also need to put atom number 1 and 2 that connects the elongated bond, ie
rotbond.pl $mol $tsatom1 $tsatom2

This code can also be used for generating rotational fragments for hindered rotor.

Never run a compound job with an RO method

Leaf (via Carl) on Fri May 20 09:35:25 2011

Both versions of G09 (A.02 and B.01) behave the same, it is just that the job you run with B.01 was a compound job ("Opt Freq"), while the job you run with A.02 was just an "Opt" job. If you want to run a frequency calculation on an optimized geometry with RO methods, you will need to do the two calculations (geometry optimization and frequency calculation) separately. For example:

# ROHF/3-21G* opt=(grad,maxcyc=300)


0 2
O 1 B1
C 2 B2 1 A1
D23 154.87231
D24 -85.12767
D25 34.87232
B4 2.08098
B11 1.96642

# ROHF/3-21G* geom=check guess=read freq=noraman


0 2

Remember that since analytic second derivatives for RO methods are not available, the frequency calculation is done by numerical differentiation of the analytic RO gradients. There is no need to specify "Freq=Numer" in this second job step as the program will do that by default when using the "Freq" keyword with an RO method.

The current behavior of G09 is that an RO method will turn into an U (or R) method in any job step of a compound job that will involve calculation of energy derivatives and only do RO for the job steps of a compound job that are only single point energy calculations. This is used in compound model chemistry jobs such as W1RO, ROCBS-QB3, etc. in which the geometry optimization and frequency calculation job steps are done with U (or R) methods, while the high level single point energy calculations are done with RO methods. Although this behavior was applied on purpose, it had the side effect of affecting RO "Opt Freq" compound jobs as you have found. The correct behavior for the RO "Opt Freq" case would have been to do RO geometry optimization followed by RO numerical frequency calculation, just like what I show in the example above. We will correct this in a future release. In the meantime, you can proceed by separating the "Opt" and "Freq" job steps as I showed in the example above.

Freq=HinderedRotor Keyword

Leaf on Fri May 20 09:39:45 2011

Gaussian 03 (e01) sometimes give the following error message when the keyword is used:
"Problem with the number of degrees of freedom"
Whileas Gaussian 09 (a02-intel) works fine.

A recent HR paper is worth a read.

When ROHF/6-31G* energy differs (more than 10^-4 Hartree) between Gaussian and Molpro

Leaf on Fri May 20 09:49:51 2011

1. Check symmetry of the molecule.

2. Keyword to try in Gaussian:
a) guess = indo
b) guess = local, works well for TS when two states are close.
c) guess = alter, and switch "alpha" electron from occ to virtual. works for TS when homo lumo close. You can also check molpro MO energies, and alter accordingly!
d) run a smaller basis set, and read in the guess.
e) run a UHF and read in the guess.
f) run a DFT and read in the guess.

3. Keyword to try in Molpro:
a) {rhf;start,h0;}
b) {rhf;start,atden;}
c) {rhf;shift,0.0,0.0;} or alter these numbers to something else..
d) {rhf;ipol,dm;}
e) run a smaller basis set, and read in the guess.
f) run a UHF and read in the guess.
g) run a DFT and read in the guess.
h) run a basis set with diffuse function, and read in the guess. works for raft type of radicals.

4. RUN Q-Chem!

If you have success with some other keyword, please feel free to add into this list..

COSMO-RS for heavy atoms

Leaf (via Junming) on Fri May 20 11:33:38 2011

ADF has a new (available in adf 2010.01) basis set and relativistic effect correction for heavy atoms. (Z>=37)
Please modify your input file if you have heavy atoms in the system.

TZP small core basis set (for all Z ≤ 36)
TZ2P small core basis set basis set for Iodine (for all Z ≥ 37)
the Becke Perdew exchange-correlation functional
the relativistic scalar ZORA Hamiltonian
integration accuracy 6

Basis type TZP
core Small
GGA Becke Perdew
relativistic scalar zora


Junming on Wed Aug 31 10:25:35 2011

I recently encountered the following error when running ADF COSMO-RS calculations on XE and VU: Platform-MPI licensed for SCM. bindcpu() failed forrtl: severe (174): SIGSEGV, segmentation fault occurred MPI Application rank 0 exited before MPI_Finalize() with status 174 mv: cannot stat `TAPE21': No such file or directory mv: cannot stat `TAPE13': No such file or directory This error is easily resolved if you submit the input as a multi-processor job (i.e. nproc > 1).

High Level Correction for triplet state

Anya on Wed Oct 19 15:50:04 2011

When using the g3eng script to calculate the HLC be aware that it does not discriminate between singlet and triplet states, which means in the latter case it mistreats the number of alpha and beta electrons. For example, for oxygen (triplet in a ground state) the script results in HLC=-0.0564780, which is a value for a singlet, while a triplet has HLC=-0.0550030.

Bug in Gaussian 03 CCSD(T) calculation

Anya on Fri Jan 6 13:03:54 2012

Gaussian 03 Rev. D03 was used to calculate the CCSD(T)/6-31G(d) single-point energy of an azaphenalene-based oxoammonium cation (charge=1, multiplicity=1). Obtained results were: E(SCF)=-745.38484718 and E(CCSD(T))=-748.0267517 Hartree. This was used to calculate the oxidation potential of the corresponding nitroxide radical / oxoamminium cation couple, and the obtained Eox was different from an experimental value by over 0.500 V (MAD for other system was within 0.050 V). Recently CCSD(T) energy of this cation was recalculated using Gaussian 09 Rev. B01 and Molpro 2009.1. Both packages gave identical results: E(SCF) - same as above, but E(CCSD(T))=-748.0074939 Hartrees, which is 50 kJ/mol higher than above. However, Eox calculated using this newer CCSD(T) number was within only 0.040 V from experiment. Conclusion: bug in Gaussian 03 resulted in a wrong CCSD(T) energy for the right SCF energy.

Few notes on molpro

leaf on Fri Jan 20 11:16:49 2012

1. To load molpro 2010 version, please do not load any other packages. (so get rid of the "module load" lines in your ~/.profile)
2. TZVP in earlier molpro version means "Ahlrichs-tzvp" but in 2010 means Turbolmole's "def2-tzvp", and all the others in the same family as well. Completely different things!
3. If you want to use G3MP2Large basis set in molpro, Kaili has written a program to convert basis set from gaussian input format into molpro input format (translate.py), please download it here:
There's also a script to convert basis set from gaussian output format into molpro input format (g2mbasis.pl), please download it here:

Note, G3MP2Large basis set is not defined for heavy atoms, I use aug-cc-pvtz-PP with ECP.
4. include in molpro can only reads certain number of lines, so, you can translate all basis into one file and ecp into another file, then include both of them in the input file.
5. molpro 2010.1_nf2 is a placeholder, do not use it.

Gaussian Rev B bug

Carl (via Leaf) on Tue Mar 13 17:09:11 2012

Often users wish to include a frequency calculation after having performed a geometry optimization. A typical command line in the data file might look like: uhf/6-31G* opt=(ts,ef) freq To our surprise, we discovered that G09 would often not perform the frequency calculation at the same level of theory that the optimization was done at. Needless to say, this is really annoying. The output file suggested that it was performed at the correct level, but on close inspection of the data, it became clear that there were unexplained differences such as energies not matching, or odd values in the frequency calculation for radicals. After reporting this, Gaussian confirmed that this problem existed and they also confirmed that: 1. Restricted open-shell (RO) optimizations would ALWAYS result in unrestricted (U) frequencies. 2. They could not guarantee that there were no other problems of this nature. They also confirmed that the problem only arises when you combine an optimization keyword with a frequency keyword in the same data file. If a frequency calculation is performed on the optimized geometry in a separate calculation, everything appears to be fine. We have also confirmed this. Advice: is you have been using RevB of G09, I strongly recommend that you check to see that you actually have frequencies that are done at the correct level of theory. Look in the Archive entries from each stage of the calculation and make sure that the data match (eg. energies, ). Apparently this bug has been fixed in RevC.01.

Spin-forbidden reactions - finding the minimum energy crossing point (MECP) on conical intersect

Richmond on Thu May 30 11:01:20 2013

Jeremy Harvey's code for finding MECP in spin forbidden reaction which is coupled to G09 (http://www.chm.bris.ac.uk/pt/harvey/jeremy.htm) is found in /projects/x69/bin/ . Spin forbidden reactions might apply to radical systems, eg scission of peroxo O-O bonds (singlet to triplet or vice versa). It would be wise to run PES scans for the 2 non-adiabatic channels of your elementary step in interest before hitting on the code. A good guess would consume less time to find the MECP! If you need any technical assistance you can get hold of me.

on Wed Mar 16 09:38:27 2016



test on Wed Mar 16 09:38:38 2016