Well, it's not easy.

The way gaussian and most other programs do normal mode analysis do get ZPE is physically meaningful only for stationary points.

If you're interested in the reactions energy barrier, you can calculate ZPE for reactants, transition state and products and add it to the energy - that makes perfect sense and should improve your results. However, for all other points on PES that is not that straightfoward. Clary group did some work on that, but limited only to very simple reactions (e.g. http://pubs.acs.org/doi/full/10.1021/jp9616019) and using their own code, which allows to calculate (actually project) frequencies at each point using curvlinear coordinate system.

Dear Fred,

You might try to use "Freq=Projected" option in Gaussian to find the vibrational frequencies perpendicular to your path. I am not sure about the accuracy of this method though.

Bartosz,

I am interested in points along the PES including stationary points. Most paper don't actually mention how they have included the ZPEs (it might seem a common sense!). Thanks. I will have a look at this paper hopefully I will get an idea.

Fred

Daniel,

This might be it!

Thanks.

Daniel,

Here is my gaussian input. I want to compute ZPEs on all the points along the IRC. If I add Freq=projected to the IRC line, gaussian complains that the two can't be combined. Do you have an idea where I should include it? Do I need to use a different approach?

%chk=cis_trans_TS2.chk

%mem=4096MB

#P UB3LYP/6-311++G(d,p) Opt=(QST3,MaxCycle=1000,ModRedundant) Freq SCF(Tight,MaxCycle=1000) Test

cis-HNNO

0 2

O

N 1 1.217

N 2 1.241 1 137.6

H 3 1.043 2 107.6 1 0.0

1 2 3 4

trans-HNNO

0 2

O

N 1 1.208

N 2 1.254 1 131.4

H 3 1.031 2 105.5 1 180.0

1 2 3 4

TS

0 2

O

N 1 1.209

N 2 1.231 1 177.0

H 3 1.056 2 112.8 1 0.0

1 2 3 4

--Link1--

%chk=cis_trans_TS2.chk

%mem=4096MB

#P UB3LYP/6-311++G(d,p) Geom=AllCheck IRC=(DVV,RCFC,MaxPoints=50,Report=Read) Guess=Read Test

1 2 3 4

Fred,

In order to compute PES along your reaction path that includes ZPEs you need to do the following steps:

1) Locate the transition state (TS). I wouldn't use QST3 here. Your reaction is rather simple so I would try "Opt(TS,NoEigentest)" method instead where you need to provide a guess of the TS structure.

2) If step 1) converges you should confirm if the structure that you found is indeed the TS. So you have to run a frequency job. You should then check if there is only one imaginary frequency. If that is the case then I would advice you to further confirm it by visualizing this vibrational mode in some drawing software. You should check if the displacement of the atoms along this mode is in the right direction (e.g. is it possible to form product(s) by following this mode?). If that is the case then it is your TS. You can also check the 6 low lying frequencies that Gaussian reports slightly above the first Frequency results line. They correspond to the rotations and translations (so 6 modes). They should be close to zero.

3) The next thing is to repeat steps 1-2) tightening the convergence criteria and passing the found TS structure from one job to another (you can do it via chk file by using the command "Geom=AllCheck". I would run "Tight" and "VTight" jobs with "Integral=Ultrafine" and "Integral=Superfine" respectively.

4) Once you are happy with your TS you should backup its chk file. Please bear in mind that the accuracy of your TS will have a huge impact on the IRC results.

5) Once you complete steps 1-4) you can use TS chk file to start IRC job. You may play here with the number of steps and its size and convergence criteria. Please, remember that if you want to start IRC again (from scratch) after some unsuccessfull run you need to use old TS chk file and not the one produced by the previous IRC job.

6) Once you have all the points that you need along the reaction path you need to save each of them into a separate gaussian input file. Then you should specify exactly the same level of theory that you used for previous jobs and then you should add Freq=Projected keyword.

For training purposes you may try to reproduce some of the results from this paper: DOI: 10.1021/jp953644f

where a lot of details along the reaction path were computed.

Daniel

Thanks Daniel,

Now the roadmap seems clear!