Using Gaussian

General Information

This page describes how to submit Gaussian Jobs on the computer cluster. Two caveats:

1. The methods given on the Gaussian, Inc. official website are not working on our system. The reason is that the job management system on our cluster is Torque which uses a different syntax from the original PBS system.
2. The method given below only supports parallel computing on a single node using the shared memory mechanism. For the parallel computing across different codes, Gaussian uses Linda, for which we did not buy the license, I think, instead of the common Message Parsing Interface (MPI) mechanism.

Job Submitting Script

Under your home directory, create a file with the following content. Let's call it q09. After the file is created, change its mode to be executable using the command, chmod +x g09.

#!/bin/bash

cat <<EOF >$1.job
#!/bin/csh
#PBS -N $1
#PBS -e $1.err
#PBS -o $1.log
#PBS -l nodes=1:ppn=12
#PBS -l mem=60GB
#PBS -l walltime=30:00
#PBS -W group_list=gaussian
source /usr/local/g09/bsd/gau-unlimit
cd $PWD
g09 <$1.gjf >& $1.out
EOF

qsub $1.job -q batch
rm $1.job

In the above script, the following will be carried out in sequential mode.

1. Create a qsub job script. Here, the default values use all 12 processors on a single node and 60 GB memory. The wall time limit is set to 30 minutes. You are advised to change ppn, walltime and mem values according to your needs but to leave the rest intact.

1. PBS -W group_list=gaussian sets the group under which the job will be run on the batch. If you do not set this, you may not have permission to use files in the gaussian directories

1. qsub submits the job to the system using the "batch" job sequence. You may change this value accordingly.
2. The qsub job script is removed. You may also remove *.err and *.log files safely as well since no valuable information is kept there.

Submit a Gaussian Job

Now you are ready to submit your jobs to the system as easy as it is used to be. You may use Link0 commands or Default. Route to specify the resource usage. For example, if your Gaussian input file is called test1.gjf, use the following command to submit the job.

q09 test1

Please note:

1. No extension should be given while the input file should have a "gjf" extension.
2. q09 should be in your binary search path. If it is not, use the full file name with the path information.
3. After the job is finished, you may just remove the *err and *log files.