Question

When calculating with gaussian 09, if the frequency becomes negative what should be done?

I am interested in calculating & calculate frequency with gaussian 09

Topics

0 / 0  ·  31 Answers  ·  7818 Views

• Arnab Chakrabarty · Process Systems Enterprise Limited
Those will be imaginary frequency or transition state. See
http://www.gaussian.com/g_whitepap/vib.htm
• Immanuel Gadaczek · University of Bonn
If you only have one negative frequency, than you have reached a transition state, but if there are more then one you have a shitty structure and you should re-optimize it.
• Melis Gedik · University of Alberta
I would recommend :
1.viewing the log file in gaussview or molden
2. Locate the imaginary frequency
3. Move the atoms in the direction of this vibration
4. Re-optimize starting from this structure

I have encountered this problem many times and this method has always worked for me :)
Another consideration is what Immanuel pointed out, so it depends if this is a ground state optimization or transition state ofcourse.
• Mehboob Alam · University of Calcutta
Hi Razieh,

I agree with Melis. But you must remember that if you are doing any geometry optimization and you get a very high negative frequency (even if you get all YES for the optimization criteria) then you need to reoptimize the geometry following the steps as Melis suggested. But if you get a very small frequency (around 10-50 or even 100 cm-1) then you can ignore that frequency and the final structure may be taken as a good geometry for the next calculations (However it is always advisable to fully optimize the structure).
I must point out one thing that while doing geometry optimization for a group of molecules or metal clusters a small negative frequency is an unavoidable problem (appears almost every-time I perform such job).

Hope it'll help.

Best of luck
Mehboob
• Luke Burke · Rutgers, The State University of New Jersey
Well... for a Physics answer, they are not negative frequencies but they are imaginary since frequencies are computed from the square roots of the force constants. The normal modes can become negative if that mode leads to a lowering of the energy. For a Chemistry answer, do the usual things in Organic Chemistry : minimize the number of eclipsing interactions, account for planarity in maximizing resonance and l.p. interactions, use the trans conformation for methyls when they enter into pi-CH2 hyper-conjugation...
• Chi-Wei Chen · National Taiwan University
Try another geometry and opt again !
• Heath Watts · Pennsylvania State University
Hi Razleh,
The convergence criteria might not be tight enough. One imaginary frequency indicates that you are at a saddle point (transition state), which is a potential energy maximum rather than a PES minimum.

Use:

OPT=(Tight) Int=(Grid=Ultrafine) CPHF=(Grid=Fine) FREQ Geom=Check Guess=Read

rather than simply:

OPT FREQ

Int and CPHF tell G09 to use a more accurate numerical integration grid. Geom=Check and Guess=Read will start the calculation from the last structure in your checkpoint file. Remember to remove your Carteisan coordinates (or Z-matrix) from your input file when using Geom=Check Guess=Read.

If this does not work. Look at the model in Gaussview to determine what the imaginary mode looks like. Slightly change the bond distance or angle in the direction of the imaginary mode, save the new geometry, and re-minimize the structure.

This paper can be useful:
www.gaussian.com/g_whitepap/vib.htm

Good luck!
Heath
• Mehboob Alam · University of Calcutta
Hi Heath,

You are right but I have a query cum comment (Please correct me if I'm wrong).

I think Grid=ultrafine is applicable to DFT methods only.

The keyword Opt=Tight will take much longer time than usual and for larger molecules it will take even larger time to complete. Sometimes opt=tight may cause convergence failure also.

No doubt, the keywords you suggested can result a better geometry than the normal OPT Freq keywords. But computational time must also be considered if the system is large enough. And in order to avoid the imaginary frequency your second method (Also that of Melis) i.e. changing the geometry in the direction of those imaginary frequencies is a good option.

Best of luck,
Mehboob
• Heath Watts · Pennsylvania State University
HI Mehboob,
Yes, the grid only applies to DFT methods. The original question did not specify the method, so I presumed that Razieh was using DFT; this assumption could be incorrect. I agree with your comment about computational time. Thanks.
Best regards,
Heath
• Mehboob Alam · University of Calcutta
Thanks Heath...:-)
• Mohammad Alam · Aligarh Muslim University
i agree with Melis
• Aqeel Ali · University of Basrah
Dear Razieh, I think you mean the vibrational frequencies. The negative frequencies resulted at any frequency computations. Negative values apear for first numbers only. Another two or three positive values are not real frequencies. The other values are theoretical values.
• Zoubeyr Boudene · IFP Energies nouvelles
You can also modify the convergence criterium for geometrical optimization: IOp(1/7= .....), the default value is 300
• Debajyoti Bhattacharjee · Tezpur University
I am also agree with Melis & Mehboob..Thanks Heath for the paper.
• Mehdi Bayat · Buali Sina University
there are several way.
for example you can use from another geometry as a starting input file.
in second way you can freezed the first negative frequency with Chemcraft software and used from this new geometry as a input file.
• Mohanned Al-anber · University of Basrah
All the above persons gave you very good suggestion, you must check all that
• Sunwoo Kang · Samsung Display, South Korea
Here is simple way to eliminate negative frequency.

First thing is change the structure based on the vibrational mode of negative frequency.
Second thing is opt=calcall. This is very useful if the system is not quite large
You can choice one of them.
But remember this, large system = use first solution small system = use calcall
• Mer Mercurate · University of Illinois, Urbana-Champaign
My case is like that:
Mode # Freq Infrared
1 -12 0.03
2 -7.0 0.5
3 -5.5 0.3
4 11.o 0.07
Then all possitive after that.
Are -12, -7, -5.5 the negative frequency?
Thanks
SCF restricted and unrestricted energies, gradients and frequencies, and RO energies and gradients. Look at log file in gaussview then locate the Im frequency and move atoms in the direction of this vibration and finally start again from this structure
• Xingyong Wang · Nanjing University
I agree with Melis. And there are little programs that are used to fulfill the purpose of "Move the atoms in the direction of this vibration". Or you can write one by yourself. Good luck!
• Francesca Rondinelli · University of Naples Federico II
I agree with Melis too! With some organic systems (antioxidants), I got good results by using Tight keyword. Xingyong, what are the programs that move the atoms in the direction of the vibration?
• Mohammad Alam · Aligarh Muslim University
you may use GaussView software for this purpose.
• Francesca Rondinelli · University of Naples Federico II
...our beloved Gaussview! Thanks!
• Jan Hrušák · Academy of Sciences of the Czech Republic
All above is relevant and it will certainly help. However, there are also cases where the negative vibrational frequency is caused by severe troubles in energy evaluations (even for analytical gradients) and it is often related to problems of the wavefunction (multireference character, instability etc.). We have documented some cases in the past. If all the hints you have got do not help, just send me a mail.
• Juan Alvarez-Idaboy · Universidad Nacional Autónoma de México
In addition to previously suggested approaches, Specially if you are using one of the Minnesota functionals as suggested by Prof. Truhlar himself you may need to increase the grid, The keyword is int=grid=ultrafine. Sorry if some one mentioned it before, I read the answers very fast.
• Viorel Chihaia · Institute of Physical Chemistry
Usually, the negative frequencies for the complex systems have low values and correspond to the delocalized vibrational modes. In such cases you may scale (contract or relax) the system to activate these vibrational modes.
• C.Yohannan Panicker · TKM College of Arts and Science
Add the coordinates of imaginary frequency value to the coordinates of standard output . Then optimize and calculate frequency
yohannanpanicker
• Saifful Kamaluddin Muzakir · Universiti Malaysia Pahang
Hi Melis Gedik,

I am referring to your explanation as follow:
"3. Move the atoms in the direction of this vibration"

How can we check the direction of the vibration?
• Bartosz Trzaskowski · University of Warsaw
Saifful: take a look at your Gaussian output. Locate the imaginary frequency: below it you will have all the atoms in your molecule and each oft hem will have a set of coordinates (x, y, z) - these show you how much this particular atom is moving during that vibration. Look for large numbers which correspond to large movements/vibrations of certain atoms and then move thse atoms slightly in that direction.

Or better: visualize the frequency corresponding to the imaginary frequency in GaussView (commercial) or Molden (free) and then manually move the atoms in the direction of the vibration. You can adjust the coordinates of certain atoms in Molden/GaussView or just edit them in the Gaussian input file.
• Stella Kritikou · Texas A&M University
It really depends on what you are looking for.
If you are searching for a transition state, you need to check if the frequency represents what actually happens in your reaction. For example you might have a low imaginary freq that is not helpful or you can have one that has a higher freq and for example represents a hydrogen moving, a bond that has been made or broken etc, a rotation etc.
In the case that you are looking for a transition state but the node you get is not the one you were looking for, I would first suggest to run the same calculation but to specify different or more reductant coordinates.
Now if you have more than one imaginary frequencies that means that you have a higher order saddle point. In that case you can check your starting geometry to make sure it was an adequate enough guess and you could try displacing atoms (there is a manual displacement option in Gauss view) anytime you visualize a vibration.
In case you have a few low lying imaginary nodes you can try changing the optimization to tight and vtight (you need to add the int=ultrafine keyword as well), or maybe changing the step size (it sometimes helps, depending on your system).
If you only have one imaginary frequency and you are looking for a minimum I would suggest to start from the two displaced geometries that you get from the vibration and rerun and see what you get.
Hope this helps :)
• Marutheeswaran Senivasan · Pondicherry University
I have one doubt
which level of theory and basis set is suitable for finding TS of the reaction path?