Transition State Optimizations

The easiest way to find a transition state is to first do a constrained geometry optimization where the constraints are the bonds that are being made and broken. Typically you want these to be ~0.3 Angstroms longer than the optimum, e.g. O-H optimum is ~0.95 A, so O-H in transition state might be ~1.25 A; another example: C-O optimum is ~1.45 A, so C-O in transition state might be ~1.75 A; for breaking a pi-bond, C=C is 1.33 A optimally and C-C is 1.5 A optimally so the intermediate case may be ~1.42 A.

If your initial structure (usually built in Spartan) is far away from these constraints, you don't want to apply these transition state constraints immediately, but rather in steps. For example, if you wanted C-O to be ~1.75 A, but your starting structure had them at 3.0 A, you would constrain these in several steps: first 2.7 A, then 2.4 A, then 2.0 A, then 1.75 A.

Once you have the constrained optimization done, replace the keyword igeopt=1 with igeopt=2. Then replace each constraint with #active and run the job.

When the job is done, check the distances in the transition state, if they are still somewhat near the ones you had initially constrained (i.e., intermediate bond forming/breaking distances), you've probably found the transition state. In your .out file look at the last optimization step (just before the box that reads Geometry Optimization Complete) and you should find a section that reads "TS search will follow eigenvector..." In the list of eigenvectors check that at least some (perhaps not all) of the bonds being made and formed are in this eigenvector.

Another way to check would be to run a frequency calculation. (You can do a quick and dirty one first before spending the computational time to run an analytical frequency calculation.) If there is one negative frequency in the range of 600-1600 cm(-1), it's probably right!

If the transition state falls to reactants or products, try a slightly different geometry biased the opposite way. For example, if the transition state fell to the products, bias your initial geometry a little closer to the reactants.

If the first eigenvector that is followed does not include any bond-stretches, use keyword itrvec=-2. If it correctly picks the first eigenvector but then deviates, try keyword ifollow=1

Alternatively, you can use keyword itrvec=-6 if you have #active eigenvector coordinates you want it to follow.