Greetings,

Which algorithm you are using to search for the TS? TS (Berny), QST2 or QST3? Because considering your explanation it seems that your structure guess is the problem, let me try to explain:

If you as looking for the TS using the Berny algorithm, you'll have to feed the software with the TS guess structure only, and that guess should be well studied/proposed so it is really close to the saddle point you want. Otherwise you won't find the right TS and just get rubbish results.

If you are using QST2, you must feed the software with the guess structure for the reactants and the guess structure for the products so the algorithm will try to guess the TS and optimize it. This algorithm seems an insteresting option because you don't need to guess the TS but it is troublesome and most often than not the results are not what you want (if you get any).

If you are using QST3, you must feed the software with guess structures for reactants, TS and products, it is a much better approach than the QST2 but sometimes it is hard to guess three reasonable structures that lead you to a good result. My advice is always paying special attention to the TS and making sure everything makes perfect chemical sense.

If you give us more details, we can surely help you further.

Lucas Gian Fachini

Thanks for the reply.

I am using TS(Berny).

I am running two structure-optimized molecules through a scan calculation to find a plausible structure that yields only one imaginary vibration, and confirming that the transition state of a similar reaction in a previous study has a similar structure, with the following keywords.

#p opt=(TS,modred) freq b3lyp/gen pseud=read

The functions used are 6-31+G(d,p) and lanl2dz, and the reaction part is specified as follows

1 59 D

61 63 D

63 64 D

1 64 D

The reaction part is plausible for a transition state of a six-membered ring, but even though we start there, the two molecules output independent states.

I see, I'd work on the TS you are feeding to the algorithm, is almost always the problem regarding the Berny approach.

The scan is a good option to find a starting point, but are you doing the scan relaxed or constrained?

If you are doing the coinstrained scan you'll have to tweak the structure so it is closer to the saddle point (even do some, 3 or 4, common opt cycles to get to a less energetic structure near the TS). That is necessary because if the molecule structure you are proposing as a whole is too energetic (thus, far from the TS) it makes it impossible to get to the right result. This happens mostly because the algorithm is not going to be able to accurately judge a saddle point that is orders of magnitude less energetic than the starting point of the TS optimization. Always remember that TS solving is actually a N dimensional problem that can get really problematic when N is large. So, sometimes a dozen of bad located hydrogens can make (all together) an energetic displacement bigger than the reaction barrier you are looking for, making the TS OPT almost impossible.

If you are doing the relaxed scan, you could actually tweak it a little, so you get to a more clear TS, sometimes it is possible to tweak an angle and get to a better result, but this depends on the chemical sense you have regarding your system.

Finding TS is more of an art than it looks, sometimes you just see yourself tweaking the positions of one or two atoms so the system makes a more clear chemical sense and leads to a better solution.

Lucas Gian Fachini

I see that the slightest difference has a much bigger impact than I had thought.

It is true that the scan may have been very rough. I was doing it by moving the distance between two atoms in the reaction part by 0.1. (Is this what you mean by "constrained"?).

I will try to do a more detailed scan around the top of the energy peak.

Thank you very much.

The term for a scan where every angle/bond lenght but the reaction coordinate is constrained is "rigid scan". I used the term "constrained" instead of "rigid" because sometimes people just constrain some angles/bond lenghts during the reaction coordinate scan to get faster computations (so it is not a rigid scan, but it is not a relaxed one as well). The idea is that coinstraints are useful but can also hinder the resolution of problematic TS.

I wish you luck on this endeavor, keep looking around the TS you want with angles/bond lenghts tweaking and never forget to try to employ as much of your chemical sense as you can, it helps a lot.

Some possible explanations:

1. Initial guess: The initial guess for the TS calculation is crucial. Even though you used the optimized structures from the scan calculation, it's possible that the initial guess for the TS calculation is not accurate enough. You might consider providing a better initial guess, such as manually modifying the coordinates to bring the molecules closer together or using interpolation methods to generate an intermediate structure between the reactant and product.

2. Convergence criteria: The default convergence criteria for TS calculations in Gaussian might not be stringent enough for your system. You could consider tightening the convergence criteria by specifying tighter optimization and frequency convergence thresholds using the `Opt` and `Freq` keywords, respectively.

3. Constraints: It's important to ensure that the necessary constraints are properly imposed during the TS calculation. Constraints, such as freezing specific bonds or angles, can help guide the optimization towards the desired transition state. You might need to review and adjust the constraints you're using for the TS calculation.

4. Noeigentest: The `noeigentest` keyword can be used to disable the eigentest in Gaussian, which can improve stability in some cases. However, it's generally recommended to use the `noeigentest` keyword primarily for smaller systems. For larger systems with 84 atoms, the eigentest is usually beneficial for stability and convergence. Therefore, it's advisable to include the `noeigentest` keyword for the TS calculation.

5. Method and basis set: The choice of the computational method and basis set can affect the accuracy and convergence of TS calculations. You might consider using a higher level of theory or a different basis set to improve the results. Additionally, for catalytic reactions, it's often beneficial to use functionals or methods that are known to describe bond-breaking and forming processes accurately.

Hope it helps:credit AI