For most simple reactions, i.e., reactions where a stable stationary point goes to another stable stationary point, the transition state can be located using LOCATE-TS. When LOCATE-TS fails, there are definite procedures that can be followed.
User errors: Most failures of LOCATE-TS are due to user errors. This sounds harsh, but during the testing phase of LOCATE-TS several failures were encountered. In every case a fault in the chemistry of the reaction was found. Two typical faults were: (A) There was an intermediate stable minimum between the reactant and the product, and (B) There was a typographical error - a simple user mistake. Given that enzymes lower activation barriers, if the calculated barrier is large, carefully examine the reaction to see if there could be an alternative step that would lower the barrier.
In the first stage of LOCATE-TS structures A and B are moved towards each other. At the start of this process, a weak force is used to roughly orientate A and B, then the force is increased so that the two structures move up the barrier, and the distance between them gets less and less. Finally, when the distance is small enough (the largest motion of any atom is less than about 0.7 Å) the geometries of the two structures are averaged, and the resulting structure is used as the starting point for the gradient minimization. At this point, three files are printed out: <file>_30p0_average.mop, <file>_30p0_first.mop, and <file>_30p0_second.mop. (This assumes that the final pull was 30 kcal/mol/Å.)
Add more steps to the orientation process: Look at <file>_30p0_first.mop and <file>_30p0_second.mop. These should be obviously related to structures A and B . If they are not, and you're sure that the the starting points are correct, try using more steps in LOCATE-TS.
Check the starting geometry for the gradient minimization: Have a look at the file <file>_30p0_average.mop, this is the "best guess" for the starting point for refining the transition state. If the starting structure for gradient minimization does not look correct, the failure likely occurred in the gradient minimization, so use a GUI to examine the geometry. If the geometry does not look like the transition state for the reaction, carefully examine the reactants and products to verify that the reaction is the one that was intended.
Alternative gradient minimization methods: If the starting structure for gradient minimization looks correct, but the transition state did not refine correctly, try refining the transition state by alternatively minimizing the heat of formation while holding the atoms in the transition state fixed and minimizing the gradient of the atoms in the transition state while holding all other atoms fixed. This can be done in a single run, but requires more human effort to set up the data set correctly. For more help, please write to openmopac@gmail.com.