DL-FIND on CHC Server
This is a description of DL-FIND on CHC server, with which you can perform MEP (Minimum Energy Potential) Search between 2 geometric points. It can also find a path similar to IRC. Therefore, one can locate TS too. In this description, therefore, I use the isomerization of hydrogen peroxide as an example.
1. Add the following line in your .cshrc
setenv LD_LIBRARY_PATH /home/share/lib
Also add /home/share/local/sbin to your path. Finally, run ascr script to prepare the temporary scratch file directories on all the compute nodes.
2. Due to the many files generated by DL-FIND, you should create a directory for each project.
Check /home/cchoi/dl-find/test/H2O2 directory.
3. Optimize the two structures (reactant and product)
Check /home/cchoi/dl-find/test/H2O2/reactant.inp and product.inp
4. Put the the two optimized geometries in *.xyz file.
Be careful! The xyz format is different from that of GAMESS. The * in this case is peroxide. But it should be different for different project.
Check /home/cchoi/dl-find/test/H2O2/peroxide.xyz
5. Create a *.inp with the RUNTYP=GRADIENT option
Note that runtyp MUST be gradient! Otherwise, it would not run. The * in this case is peroxide. But it should be different for different project.This is because the DL-FIND utilizes the energy and gradient of gamess calculations. The geometry of $DATA in this file is NOT used. Therefore, any geometry is good. However, you must provide the same atoms with the same atom sequence. Therefore, the best practice is to simply copy the reactant.inp to peroxide.inp
Check /home/cchoi/dl-find/test/H2O2/peroxide.inp
Note for Ice Surface Calculation: Ice surface is frozen during calculation and it is in $EFP group. DL-FIND only automatically modify $DATA for the image points. Therefore, simply put $EFP group in the *.inp file, which can allow the ice surface calculations of QM/EFP runs too.
5. Setting DL-FIND parameters in input.peroxide
If you don't provide an input.peroxide, queue WILL generate it. So don't worry if you don't know what to do. Just go to the next step.
6. Submit a job by qfind
The necessary files for DL-FIND are peroxide.inp and peroxide.xyz. For test run, just copy these two files to your directory. Then, you can submit a job by simply typing
qfind peroxide
The qfind will ask the number of images, which is the number of mid points between the two structures you provided.
7. Checking the status of calculations
DL-FIND generates peroxide.out. So you can
tail -f peroxide.out
8. Resubmit
DL-FIND can be resubmitted if it is terminated. Now you have input.peroxide. Simply change restart option from no to yes in the file. Then, type
qfind peroxide
9. Checking the values
All the necessary infos are in peroxide.out. But it contains all the infos for every iteration steps. The rundir/nebpath.xyz contains the infos of last iteration. So simply type
grep E rundir/nebpath.xyz | awk '{$1=$2=""; print $0}'
will show you only the energies of each images. Then, you can plot like below.
10. Killing running job
Please use qcln node_name. For instance, qcln i9k1 If you submitted more than one node. You need to qcln for every node.
1. Add the following line in your .cshrc
setenv LD_LIBRARY_PATH /home/share/lib
Also add /home/share/local/sbin to your path. Finally, run ascr script to prepare the temporary scratch file directories on all the compute nodes.
2. Due to the many files generated by DL-FIND, you should create a directory for each project.
Check /home/cchoi/dl-find/test/H2O2 directory.
3. Optimize the two structures (reactant and product)
Check /home/cchoi/dl-find/test/H2O2/reactant.inp and product.inp
4. Put the the two optimized geometries in *.xyz file.
Be careful! The xyz format is different from that of GAMESS. The * in this case is peroxide. But it should be different for different project.
Check /home/cchoi/dl-find/test/H2O2/peroxide.xyz
5. Create a *.inp with the RUNTYP=GRADIENT option
Note that runtyp MUST be gradient! Otherwise, it would not run. The * in this case is peroxide. But it should be different for different project.This is because the DL-FIND utilizes the energy and gradient of gamess calculations. The geometry of $DATA in this file is NOT used. Therefore, any geometry is good. However, you must provide the same atoms with the same atom sequence. Therefore, the best practice is to simply copy the reactant.inp to peroxide.inp
Check /home/cchoi/dl-find/test/H2O2/peroxide.inp
Note for Ice Surface Calculation: Ice surface is frozen during calculation and it is in $EFP group. DL-FIND only automatically modify $DATA for the image points. Therefore, simply put $EFP group in the *.inp file, which can allow the ice surface calculations of QM/EFP runs too.
5. Setting DL-FIND parameters in input.peroxide
If you don't provide an input.peroxide, queue WILL generate it. So don't worry if you don't know what to do. Just go to the next step.
6. Submit a job by qfind
The necessary files for DL-FIND are peroxide.inp and peroxide.xyz. For test run, just copy these two files to your directory. Then, you can submit a job by simply typing
qfind peroxide
The qfind will ask the number of images, which is the number of mid points between the two structures you provided.
7. Checking the status of calculations
DL-FIND generates peroxide.out. So you can
tail -f peroxide.out
8. Resubmit
DL-FIND can be resubmitted if it is terminated. Now you have input.peroxide. Simply change restart option from no to yes in the file. Then, type
qfind peroxide
9. Checking the values
All the necessary infos are in peroxide.out. But it contains all the infos for every iteration steps. The rundir/nebpath.xyz contains the infos of last iteration. So simply type
grep E rundir/nebpath.xyz | awk '{$1=$2=""; print $0}'
will show you only the energies of each images. Then, you can plot like below.
Please use qcln node_name. For instance, qcln i9k1 If you submitted more than one node. You need to qcln for every node.
Comments
Post a Comment