Transition State Optimization with Dimer Method

[Natalie Austin]

Hello,

I recently attempted to calculate a transition state for CO2 dissociation using CI-NEB. Unfortunately, the NEB did not converge after 66 steps ( It ran once up to 32 steps then I had to resubmit it to run another 34 steps) and so I stopped the calculation because I noticed that the MAX FORCE for the NEB calculation was just fluctuating(see below). Would anyone be able to indicate why this is the case? I used the same input for a located a transition from physisorbed CO2 to chemisorbed CO2 on the same metal system and I did not have this problem.

I did come across a case in the steps where all criteria was met except for the MAX force as shown below:

 RMS DISPLACEMENT =   0.00013     [ 0.00100]                               (YES)
 MAX DISPLACEMENT =   0.00225     [ 0.00300]                               (YES)
 RMS FORCE        =   0.00025     [ 0.00100]                               (YES)
 MAX FORCE        =   0.00407     [ 0.00300]                               ( NO)

So I took that structure and ran vibrational analysis on it and found that it had 3 imaginary frequencies as shown below:

 VIB|                        NORMAL MODES - CARTESIAN DISPLACEMENTS
 VIB|
 VIB|                         1                    2                    3
 VIB|Frequency (cm^-1)  -374.496175          -281.692265           -60.923863
 VIB|Intensities           0.000000             0.000000             0.000000
 VIB|Red.Masses (a.u.)    14.024323            16.040216            53.709296
 VIB|Frc consts (a.u.)    -0.000706            -0.000259            -0.000002


As a previous user suggested I then tried to use the Transition State Dimer method to see if I could guide this structure to a transition state by providing the vectors in DIMER _VECTOR for the mode corresponding to the state that I was looking for. I was able to locate a potential transition state. I've providing the transition state coordinates below. Does it make sense to use CI-NEB and reoptimize with the Transition state search in cases where it is difficult to converge the NEB?  Is this a common occurrence? I'm just trying to make sure that I'm locating transition states in an appropriate way.

Thanks,

Natalie
 

[luca]

Dear Natalie
the dimer method is a tool to search
TS structures and it can be used to refine (or validate) TS provided by
other methods.
I have just looked your input, the dimer method section is ok; however,
the calculation of the frequency is very sensitive, so, I would suggest
you to increase the accuracy, decreasing RMS displacement and so on.
This is my usual set-up:
RMS_DR 1.0E-04 # 1.0E-04
MAX_DR 3.0E-04 # 3.0E-05
RMS_FORCE 1.0E-04 # 1.0E-04
MAX_FORCE 3.0E-04 # 3.0E-05
Actually, to obtain a single ifreq more tight parameters are required.
Pay attention, the dimer method shows a slow convergence, so it requires
many steps.
In the attached file you can find parameters that I used in
DOI:10.1021/jo3023439

Cheers,
Luca

On Thu, 2016-05-05 at 17:27 -0700, Natalie Austin wrote:
> Hello,
>
> I recently attempted to calculate a transition state for CO2
> dissociation using CI-NEB. Unfortunately, the NEB did not converge
> after 66 steps ( It ran once up to 32 steps then I had to resubmit it
> to run another 34 steps) and so I stopped the calculation because I
> noticed that the MAX FORCE for the NEB calculation was just
> fluctuating(see below). Would anyone be able to indicate why this is
> the case? I used the same input for a located a transition from
> physisorbed CO2 to chemisorbed CO2 on the same metal system and I did
> not have this problem.
>
> I did come across a case in the steps where all criteria was met
> except for the MAX force as shown below:
>
> RMS DISPLACEMENT = 0.00013 [ 0.00100]
> (YES)
> MAX DISPLACEMENT = 0.00225 [ 0.00300]
> (YES)
> RMS FORCE = 0.00025 [ 0.00100]
> (YES)
> MAX FORCE = 0.00407 [ 0.00300]
> ( NO)
>
> So I took that structure and ran vibrational analysis on it and found
> that it had 3 imaginary frequencies as shown below:
>
>

> As a previous user suggested I then tried to use the Transition State
> Dimer method to see if I could guide this structure to a transition
> state by providing the vectors in DIMER _VECTOR for the mode
> corresponding to the state that I was looking for. I was able to
> locate a potential transition state. I've providing the transition
> state coordinates below. Does it make sense to use CI-NEB and
> reoptimize with the Transition state search in cases where it is
> difficult to converge the NEB? Is this a common occurrence? I'm just
> trying to make sure that I'm locating transition states in an
> appropriate way.
>
> Thanks,
>
> Natalie
>
>

> --
> You received this message because you are subscribed to the Google
> Groups "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to cp2k+uns...@googlegroups.com.
> To post to this group, send email to cp...@googlegroups.com.
> Visit this group at https://groups.google.com/group/cp2k.
> For more options, visit https://groups.google.com/d/optout.

[Natalie Austin]

Hello Lucas,

Thank you for your input. I have already ran a vibration frequency calculation using the coordinates obtained from the Dimer Method TS calculation. I was able to locate a single imaginary frequency using the attached input. My MAX FORCE was 0.0004 in that calculation.
So although I was able to locate a single imaginary mode, if I were to tighten the criteria for the TS calculation, I could see completely different results in the vibrational analysis? 


Thanks,

Natalie

[luca]

Hi Natalie,
tightening convergence criteria you should observe a commensurate
conformational change, thus, I do not aspect huge changes otherwise, the
system is not really converged.
Anyway, it is system dependent. To fully characterize the TS (or more
TSs for multiple path reaction) you should perform other TS searches,
for example, starting from different conformation of the system. You can
refine the characterized TS re-doing the TS search starting from the
last TS geometry obtained, but using a new randomized dimer directions
set.

good luck,

Luca

[Megha Anand]

Dear Luca,

I am trying to produce the results from your paper : http://pubs.acs.org/doi/full/10.1021/jo3023439

Will you please provide the input for TS search for any of the transition states from your paper.

Thank you so much for your help and time!

Best regards,

Megha