Summary: Loacting a transition state

Hi all,

I would like to thank those who responded my questions on locating a transition state.  All of them are quite instructive for me. Special thanks go to  David Gallagher and James J.P. Stewart who actually located the transition state for my system (See below). I could also locate a transition state.

Responses exhibited wide spectra of perception of the transition state search:

|---- Easy                                              David Gallagher, James J.P. Stewart, David Close
|---- Accurate methods                    Valentine Ananikov, Alexander Martins
|---- Difficult                                          Irene Newhouse, Bin Shan
|---- Black art                                      Markus Dittrich
|----More than one TS                     Stephen Bowlus
|---- Existence of TS                        David Shobe

 David Gallagher did find the transition state for the hydrolysis reaction of acetylcholine with water using PM5 method in CAChe version of MOPAC 2002. Using PM3 in MOPAC, James Stewart proved that the structure derived from the above PM5 method has only one negative frequency. I used the PM5 TS structure with PM3 in Gaussin98 [#P RHF/PM3 opt=(TS,CalcFC)], however, the job stopped yielding two negative frequencies (Optimization stopped.   -- Wrong number of Negative eigenvalues: Desired=  1 Actual=  2 ). The suggestion by Valentine Ananikov, Alexander Martins and Bin Shan  [opt=(modredundant,TS,noeigen)] solved this problem.

In the above, we assume that there is a transition state for a given reaction. David Shobe raised a question whether or not the transition structure exists, pointing out a no barrier reaction for TiCl4 + H2O --> TiCl4(H2O). The existence of a TS may be considered from the free energies for the reactants and the products. If the products are higher in free energy than the reactants, there must be a barrier. However, for the opposite case, there is no grantee that the barrier exists. Are anyone aware of any criteria for the existence of the transition state?



My posting was


I am trying to locate a transition state for a system about 30 atoms using an empirical
method. So far I haven't had any success. I would appreciate if you could
give me any suggestions.

I am interested in a hydrolysis reaction pathway, using PM3 in Gaussian 98:

R-C-O-C-R' + H*-O'H ---> R-C-O-H*  + H-O'-C-R'

I have done the following:
1) opt=Ts ; This always failed due to incorrect number of negative
2) Scan  ; I have not tried much.
3) opt=ModRedundant: Scanning succeeded, but the population analysis after
the scanning seemed failed. The output did not indicate why it failed, but
the heading of population analysis appeared, followed by signal 11 error.
4) opt=qst2; There is (are) one or two statement(s) "stationary point was
found" (the distances were the same for all, but the angles and torsions were slightly different in the two stationary points) in the output, but the optimization (?) kept going and failed; Optimization aborted
--- No acceptable step or Inconsistency: ModMin=   2.

I optimized two reactant molecules individually and moved them close to make
a reaction. The products were also optimized. The order of the atoms are
exactly the same.


Genzo Tanaka

Here are the responses for my questions

This should be easy, but what exactly is R-C-O-C-R meant to be?

An ether (R-CH2-0-CH2-R) or an anhydride (R-CO-O-CO-R) or what?

If you send me the exact structures, I'll try and run it for you.

David Gallagher

Although, the old MOPAC PM3 hamiltonian seems to have difficulty in locating a transition state for your reaction, the latest PM5 method in MOPAC 2002 nailed it first go with a saddle calculation.

The attached Zip file includes pictures and the MOPAC output of the gas phase transition state for your hydrolysis reaction. In the picture, the bond thickness indicates the relative bond order and the numbers are the atom distances in Angstroms. The force calculation shows a single negative vibration and the IRCs verify that this is the only transition state along this particular reaction path for the concerted reaction. The new PM5 method has been re-parameterized for higher accuracy and MOPAC 2002 includes new algorithms. It seems to be much faster and more reliable for finding transition states than the older MOPAC methods such as PM3 and AM1. Also, I was able to model the transition state with the COSMO water solvent field. As expected, the energy was much lower.

Although, the old PM3 seems to work fine for a simple ester hydrolysis, the charged quaternary ammonium group seems to confuse it. The best transition state structure I got out of PM3 had a second negative vibration due to a methyl rotation. However, the bond orders, atom distances and negative vibration all indicated a proton transfer only (not the C-O bond break), so I don't think that this was the correct transition state for the reaction.

If you need more information about MOPAC 2002 you can view the manual at The pictures and calculations were done with the CAChe version of MOPAC 2002.

I hope this helps
David Gallagher, Fujitsu  
Attached are two MOPAC data sets.  In the first, bug.mop, I used David's
geometry from his PM5 result.  I changed PM5 to PM3, and ran the TS.  This
runs without problem.

From the ARC file, I made bug1.mop.  This runs the FORCE calculation and
confirms that there is exactly 1 imaginary frequency.

So, in summary, both PM3 and PM5 (and, I assume, all the other methods)
give more-or-less the same transition state.

As far as the system being difficult, my apologies, but what was difficult?

(I use MOPAC as stand alone, because I'm more comfortable with that.)

Best wishes,


I'm not really an expert on TS optimizations, although I was willing to help
someone who was even further back on the learning curve than I was.  And
I've never had much luck with opt=qst2.  You've checked that the atoms
correspond properly between reactant and product, which is good.  I think
sometimes opt=qst2 has trouble if the path from reactants to products is
indirect: opt=qst2 will try to find a path that takes the reagent *through*
the other molecule, instead of going *around* it.

I don't even know what signal 11 error is.  It sounds like a Unix message.
Two possibilities that come to mind are that there was a time limit for the
job which ran out, or that the system administrator "killed" it for some

Given what you've told me, I would try the opt=modredundant again, after I
found out what Signal 11 error means and had done whatever was necessary to
avoid it.  Oh, and you may want to send your question to the CCL list, since
someone else may have a better idea!

--Dave S.  
I think of these as being multi-step reactions.  Maybe you're trying to find
a TS that doesn't exist, because you skipped a step?  

> I am interested in a hydrolysis reaction pathway, using PM3 in
> Gaussian 98:
> R-C-O-C-R' + H*-O'H ---> R-C-O-H* + H-O'-C-R'

It turns out that in one of the examples that poisoned my mind against
opt=qst2, the transition state was not found for an equally good reason: the
reaction was an association reaction TiCl4 + H2O --> TiCl4(H2O), and as it
turns out the energy is monotonic from reactants to product: there is no
activation barrier.

Anyway, I wish you luck in solving your transition state puzzles.

--David Shobe  
Transition states are always a difficult problem.  You will have to be
pretty close to the actual state to get it to optimize.  Yes, that's as
circular as it sounds!

I would strongly urge you to search the literature for similar transition
states & use those structures to guide you in your initial guess.

Irene Newhouse  

The results of the first step are very common.  But this is not a problem.
You must look at the output with software that will show the vibration
frequencies.  Only one of them will involve motion along the coordinates
of the reaction path you are trying to follow.  Look at the other negative
frequencies.  You will see that they have nothing to do with your reaction
coordinates.  It may be some low frequency involving the wagging of an
exo-cyclic group.  Find the atoms involved in the unwanted motion.  Go to
the Standard Orientation near the end of the output and disturb these
coordinates a little (maybe change them 0.1 A).  Run the OPT (TS) again
with these new coordinates.  You will quickly come down off the local
minimia you are stuck on and be at the true TS.
 Regards, Dave Close.  

First, I would further research whether the (apparently) four center reaction that is being done here is really feasible (or is that the question you are trying to answer?)  I would guess there are two TS that are encountered in this reaction.  For example, addition of water to form RCO-  ... +H2O'CR' (tight) ion pair, followed by hydrogen transfer to form the two neutral species.
Regardless of the mechanism, I would use something like SADDLE in Mopac or CHAIN in Ampac to get an estimate of the possible TS, then refine the TS using eigenvector following or some gentle gradient reduction method.  The general idea is to come up with an approximate starting geometry that is somewhere on the reaction coordinate (not just two molecules in space; this is where some inference about the probable mechanism is needed); a final product geometry that is on the other side of the hump, and then work both ends against the middle to find an approximate TS.  Then refine.  Then do IRC calculations to assure that you are actually doing the chemistry you think.  There is a section in the Mopac manual about all this, and it is included in an early JCAMD (around 1996, I think).
Sorry if this is redundant to what you know ...
Stephen Bowlus    <stephen.bowlus;at;>
I can't tell you much concrete. Finding transition states aka
looking for saddle point is a black art. I am currently
doing something similar for a 20 atom system using ab initio methods and
don't have much experience with semi-empirical ones.

In general you have to start as close as possible to the true
transitions state when starting a saddle point search. Otherwise
you won't be able to get it. You could make an educated guess or
first constrain some degrees of freedom you think are important
while looking for saddle points. Comparing the energies can give
you a hint in which direction you have to go.
Just moving two optimized molecules close to each other won't
do in 99.9% of the cases. Your transition state will be
'something inbetween'. Maybe a partial transfer of your groups,
partial bond formation... You will have to play around.
If you have a good guess about a possible reaction coordinate
contrain you system such that you mimick the reaction, monitoring
the energies. That should allow you to position your system as
close to the transition state as possible.

In any case. Once you think you found the transition state you
have to make sure that you have found a saddle point
by recalculating your hessian matrix which has to have one and
only one imaginary frequency. Otherwise you're back to where
you were before.

Good luck,

Markus Dittrich   <markus;at;>

opt(TS, noeigentest) should help.

I suggest you HF/3-21G instead of PM3 for preliminary calculations
and B3LYP/6-31G* for more accurate data.

regards, Valentin Ananikov.  

       First of all, it's a good idea use DFT methods instead of
semiempirical ones. The results ara more reliable.

       You can use the keyword OPT=(TS, NoEigenTest) with a good guess
for the geometry. Futhermore, the keyword OPT=QST3 it's a good option
too. Don't forget use a good basis set, like 6-31G.

       Try these sugestions.

       Good luck,

       Alexander Martins.  
   Genzo,try something like this

# opt(modredundant,TS,noeigen,CalcFC) rb3lyp/6-31g(d) nosymm

CalcFC calculates the initial force constants be computed at the first point
and helps to determine which direction the atoms are to to move.The keyword
noeigen and nosymm are also important.
  And it is also extremely important to specify a good initial
guess.Finding transition state is by no means an easy task...according to my
  Good luck.

Bin Shan  

Attachment: bug1.mop
Description: Binary data

Attachment: bug.mop
Description: Binary data