I am calculating the transition state for tautomerization processes with opt = (TS,CalcAll,Noeigentest) with the method of calculation M062X/6-31 + g (d, p), but I cannot find it and QST2 does not work. Could someone tell me what is the best level of calculation to locate transition states because DFT fails in some cases?
Its almost always your input structure. Remember, a TS calculation and a GS calculation are basically the same. Both seek a minimum on a potential energy surface, its just that in the TS case this minimum corresponds to a maximum on another surface. Thus, your input structure energy needs to be higher in energy than the TS energy or it will collapse.
(1) First scan the bond (hopefully one bond) at the HF level using a large step size.
(2) Refine the step size near the maximum you found in the bond scan.
(3) Repeat the scan at your intended level of theory with further refinement. The fact that you have restricted a geometry optimization mode in your scan will lead to the higher energy structure you need.
(4) Calculate the frequency at the maximum energy structure you found and check the imaginary mode in molden or elsewhere to confirm the nuclear displacements.
(5) Remove any other negative modes by adding or subtracting their displacements to the nuclear coordinates and redo the bond scan if this step was needed.
(6) Now run your TS optimizaion by reading in your calculated correct frequencies.
(7) readfreq should be enough, calcall may be required later. You may want to guess=mix,always if you have electronic instabilities.
(8) If this fails your level is not appropriate and increase the basis and repeat.
(9) if that fails, sacrifice a lamb.
DFT with 6-311**G(d,p) or 6-31G* may yield better results to find transition state.
Its almost always your input structure. Remember, a TS calculation and a GS calculation are basically the same. Both seek a minimum on a potential energy surface, its just that in the TS case this minimum corresponds to a maximum on another surface. Thus, your input structure energy needs to be higher in energy than the TS energy or it will collapse.
(1) First scan the bond (hopefully one bond) at the HF level using a large step size.
(2) Refine the step size near the maximum you found in the bond scan.
(3) Repeat the scan at your intended level of theory with further refinement. The fact that you have restricted a geometry optimization mode in your scan will lead to the higher energy structure you need.
(4) Calculate the frequency at the maximum energy structure you found and check the imaginary mode in molden or elsewhere to confirm the nuclear displacements.
(5) Remove any other negative modes by adding or subtracting their displacements to the nuclear coordinates and redo the bond scan if this step was needed.
(6) Now run your TS optimizaion by reading in your calculated correct frequencies.
(7) readfreq should be enough, calcall may be required later. You may want to guess=mix,always if you have electronic instabilities.
(8) If this fails your level is not appropriate and increase the basis and repeat.
(9) if that fails, sacrifice a lamb.
If you are using Gaussian0X you can try the QST3 method that uses as input also a guess of the TS (be carefulle that it should be the last input structure not that one in the middle !!!) It is better than the opt=TS procedure that frequenty fails.
The suggestions of Peter are fine (hopefully without sacrifying any lamb ! ) In this respect QST3 helps a lot. On the other hand the quest of a TS has still some flawor of an art. I should suggest to use a not too large basis set because larger basis set tend to reduce the curvature of the PES and the TS became floppy making the procedure harder. I would suggest to stick to a 6-31g(d) adding a p function only on the proton involved in the process. You can refine it in the following with larger basis set and using the previous Hessian which should already have just one negative eigenvalue.
What is your molecule ? If you send me the structure I could give you more suggestions cause I did many work on this kind of processes.
Good luck !
Hello:
Yes. I am surprized of the suggestions for diffuse functions. These will make convergence difficult. Stick with the lower basis set for now like Francvesco suggests. Also, we are assuming that you are looking at a proton transfer and so the proton needs some basis functions to describe it properly. I guess also the TS will be ionic and diffuse function can help here, so after the low level, refinment at the suggested basis sets seems like a good idea.
Best,
Peter
I would like to thank everyone, especially Peter and Francesco for their comments and suggestions!. And yes I am looking at a proton transfer and we decided to use this functional and basis, as in previous studies at this university, where this combination gives a good description.
After a change in calculation method was able to converge to the TS, I did:
1) I did a Opt = (TS, CalcAll, Noeigentest, MaxCycle = 500, MaxStep = 500) PM6 geom = connectivity, obtaining only one imaginary frequency.
2) Then I refine the calculation with Opt = (TS, CalcFC, Noeigentest, MaxCycle = 500, MaxStep = 500) M06-2X/6-31 + (d, p) geom = connectivity at the end of the TS optimization converged well .
3) Finally, the final structure obtained was made a calculation of frequencies (Freq M06-2X/6-31 + (d, p)), obtaining the desired imaginary frequency.
Thank you for your help!
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry