I want to scan one bond of my system of interest, while simultaneously computing the forces at each step. The "force" keyword which is suggested by Gaussian is not compatible with the "scan" jobtype... . A suggested workaround I discovered would be to use the "opt=(modredundant)" jobtype. However, this invokes an energy minimization at each scan step, which is undesirable since I want to apply this to larger systems further down the line. I have a attached a minimal example input file.
Hello Armin Penz
Though I am not sure about this, you may try to calculate the numerical gradient (Force) for the bond scan.
For each scan step, you might calculate energies by varying bond distance in very small steps (ca. 0.0001 Bohr) for both forward and backward directions. The change in energy might be small. Then, you should be able to get the force by using the finite difference formula.
I computed the forces for the water molecule using B3LYP/6-31G*. I found that the analytical force for the O-H bond is 0.0055, while the numerical gradient (using a simple Symmetric derivative) for the same is 0.0053. The difference of 0.0002 units could be due to the method and step length I used to calculate the derivative.
The procedure may initially seem tedious but can be automated with simple scripts.
Although I do not have validation for the above procedure, you might be able to get some references.
I hope this helps! And feel free to reach out.
Hello Dilawar Singh Sisodiya
Thank you for bringing up this workaround! However, it seems I was not clear enough in formulating my question, since I want to calculate the force acting on each atom in the system and not just the total force.
I am sorry, Armin Penz.
Then, you may manually change the bond length and use the "Force" keyword. You will get gradients in the Cartesian Coordinate format in the log file giving information for forces on each atom.
I hope I gave reliable answer and if yes, then you can also try to automate the above procedure.
Yes Dilawar Singh Sisodiya , this is a valid workaround. Although it generates a lot of individual files and isn't as clean as I was hoping for a solution.
Anyway, thanks a lot for taking your time!
Hello Armin Penz
I think the number of your scanning steps is very low. On the other hand, increasing the number of steps increases the CPU time. I put a complex scan sample in the attachment. In this scan, both the bond and the angle have been examined. You can also get help from MolQube for CPU. Email them: [email protected]
Hello Maryam Farajpour Mojdehi ,
The input file I provided is intended to be a minimal example, that's why I set the step number to such a low number. Moreover, I am not interested in scanning the angle of my structures of interest. The essence of my question is the calculation of forces employing the "scan" jobtype.
- Badges
- Science topic