A minimum has vanishing gradients and only positive eigen modes. A (first-order) transition state (saddle point) is characterized by having one negative mode. With a transition state search the optimizer will go uphill in the direction of the lowest (nonzero) eigenmode and downhill in all other degrees of freedom. In our example it would follow mode 8. Let us give it a try from the minimum.
Choose in the 'Main Options' panel the task 'TransitionState'
We have just calculated a Hessian (with the frequency run) so we'd better use it.
Go to the 'Optimization' panel, click on the plus button next to 'Initial Hessian From:' Select with the file dialog 'H3_freq.runkf'.
Save the project as 'H3_ts' and run it.
The most likely outcome, however, is that the optimizer stops immediately, because the gradients are zero. Therefore, we need to help the optimizer a bit.
Move the rightmost atom a tiny bit to the right (increase the x value by 0.01).
(Here I increased the values next to H(3)).
Run it again, and now it will run for more cycles. After it has finished, open adfmovie
The last frame looks like
The third H atom ends up exactly in the middle of the (repeated) H1 and H2 atoms. Let us finally check that we are indeed in the transition state.
In the 'Main Options' panel select the task 'Frequencies' Save the project as 'H3_ts_freq' and run it. Afterwards, open adfspectra and click on the 'NormalModes' menu
You should see
We have found a geometry with vanishing gradients with one weak negative vibrational mode. We have succeeded in finding a transition state.
