Gaussian jobs generate large amounts of intermediate data that are too enormous for checkpoint files to store. Instead, this data is stored in RWF files, which are designed to handle large data volumes. If a Gaussian job is interrupted due to power failure or other reasons, the RWF files contain important data that can be recovered using the Restart keyword, which resumes previously failed jobs. Normally, if a job terminates successfully, RWF files are deleted automatically because of the % sign in front of the keyword, but using %NoSave after %RWF prevents this default deletion. Two important files are:
1. %Chk: Often useful after the job finishes, so place it after %NoSave.
2. %RWF: Should be stored in a location with enough space, such as a local scratch file system. An interrupted geometry optimization can be restored using the Restart keyword in the route section:
Sample program example
%rwf=/location to store read/write file
%NoSave
%Mem=4GB
%Chk=filename.chk
#p B3LYP/6-311++G(d, p) Opt=Restart Freq Integral=UltraFine SCF=XQC
job name
1 2 #charge and multiplicity
Other program options
The Opt=Restart option always resumes from the last completed point in the previous optimization. If the Opt keyword previously included additional choices, you can typically just include Restart with those options.
%Chk=filename.chk
%# Opt=(QST2, Restart)…….
If CalcFC or RCFC was included in the original options, you can exclude it if force constant computation is unnecessary, such as in optimizations beyond the initial step. However, by adding CalcFC to the option list along with Restart, the force constants will be computed at the start of the restarted optimization process. If you want to start from a specific point, then use
# Geom=(Allcheck, step=n)
You can initiate a new geometry optimization from a specific point in a previous optimization using the Step option with the Geom keyword. Regardless of whether the previous optimization completed was interrupted, or failed. You can retrieve the desired structure from the existing filename.chk file without modifying it using the following notation:
%OldChk=old filename.chk
%Chk=new filename.chk
# B3LYP/6-311++G(d, p) Opt Geom=(Allcheck, step=20) Guess=Read
This will start the optimization of the structure at point 20 of the previous optimization steps:
IRC can also be started by just using Restart
%Chk=filename.chk
# IRC=Restart …
This feature can restart an IRC calculation that was not successfully completed.
Be aware that restarting requires the filename.chk file from the previous job, and some job types also need RWF files. If any required file type is deleted, the job cannot be restarted.
The Restart option for the Freq keyword allows for restarting a numerical frequency calculation. It resumes the numerical frequency job from the last successfully completed geometry. If a numerical frequency job fails, it can be restarted from its checkpoint file by replicating the original job’s route section and adding the Restart option to the Freq=Numer keyword/option. No additional input is necessary.
%Chk=filename.chk
# Freq=(Numer,Restart)
Restart from RWF (option for large data files)
Below is an example demonstrating how to restart interrupted Gaussian 16 calculations using %RWF and filename.Chk files. I have managed to successfully resume both the frequency calculation and the EOM-CCSD excited state calculation from their last completed steps. It’s important to highlight that the EOM-CCSD calculation is highly computationally intensive (scaling approximately as N6 where N is the basis functions), making the restart capability invaluable for saving significant time. Additionally, please note that the Checkpoint file is essential in conjunction with the %RWF file.
%rwf=/home/temp/Gau-1727.rwf, 600GB #Location of the rwf file, disk space
%NoSave
%Mem=8GB
%Chk=filename.chk
# p restart
job name
For comprehensive details on Gaussian 16, including its capabilities, supported keywords, installation procedures, usage guidelines, and limitations, please refer to the complete documentation on I/O files and related information available at link to documentation. This resource provides thorough insights into utilizing Gaussian 16 effectively for various computational tasks.
Leave a Reply