opt=cartesian will help you to remove this problem. Submit a job with this keyword. All the best.

Thank you for your help :)

Dear Subrashis,

Since Gaussian chooses the internal coordinates on its own (though You can change them using modredundant / addredundant) it sometimes simply makes a bad choice, which leads to convergence problems. This is particularly true for systems with a lot of flat (close to 180 degrees) angles. In such cases you can try by modifying your geometries

Hope it will help.

Thank you

and you might have tried opt=(cartesian) as Mr. Susanta has already suggested, it works sometimes but not always.

Good luck dear!!!!!!

opt=cartesian with nosyym will help to solve it

Sometimes, opt=cartesian keyword is not enough to solve internal coordinate error

Prof.Susanta,

Can you kindly display the coordinates what you have submitted apart from those keywords you displayed here? because on seeing that one can solve your problem easily with same keywords.

Dear all, 

I am really sory this answer is for the questioner and not for Prof. Susanta, it is wrongly written sory sir.

Thank you Venkatesan. I have solved the problem.

Dear bannerjee, 

How you solved the problem sir. Could you please explain here. So that if it is any small techniques involved we will also learn.

Just follow what Dr. Susanta has said. He has clearly explained what you need to do. 

Dear Subhrashis,

Dr. Susanta Das has already given the best possible solution of your problem. Still it is well worth to understand what is the source of this type of error. Lets first begin with the common question 'How to define the structure of a molecule?'  Commonly there are two way to define the structure 

1. Use internal coordinate (Z-matrix) where the the structure is defined by means of bond length ,angles and dihedral and needs 3n-6(for linear molecule 3n-5,n=number of atoms) number of variables to optimize. One common example of H2O is given below,

H

O           0.95

H            0.95           104.5                  (total 3*3-6=3 number of variables used, two bond length and one bond angle)

2. Use Cartesian coordinate of each atom(needed 3n number of variables for each atom x,y and z coordinate). Example of H2O molecule is given below.

H          0.000000 0.000000 0.000000

O          0.000000 0.000000 0.950000

H          0.919740 0.000000 1.187861

Now, optimizing in Cartesian coordinate is a bit expensive since it uses 3n number of variables to optimize rather than 3n-6(or 3n-5) number of variables used in internal coordinate. Therefore by default gaussian uses internal redundant coordinate (As Dr. Haamid Rasool Bhat has already mentioned) for optimization. (Don't get confused to see the gaussian input files (.com of .gjf) made by Gview always uses cartesian coordinate. G09 always convert those coordinates before calculation according to as specified in input section or as per default. See output file for clarification).

Now using internal coordinate is fine if it is able to define the molecular structure well but there are certain situation when internal coordinate don't work so well, one such example is when some part of the molecule is linear. Z- matrix does not support angle value =0 or 180(see P 105 of attached link for details). Hence, if some part of the molecule is linear or becomes linear during optimization Gaussian will show you the famous error 'error in internal coordinate' . That's a modest way to say that your internal coordinate is not accurate so please correct it or shift to Cartesian one.

Although for small linear molecules or rings it is possible to define a well constructed Z matrix by using dummy atoms( see screenshots for CO2 as generated by molden), still as Dr. Sunwoo Kang has mentioned in rare cases that is not enough. Also one must remember the fact that optimization in Cartesian coordinate is more expensive (as it uses 3n number of variables rather than 3n-6 or 3n-5). In bigger systems  those extra variables can literally kill you. 

http://as.wiley.com/WileyCDA/WileyTitle/productCd-0471333689.html

Thank you a lot. This is really a good information. It will help me and others to understand better.

(as it uses 3n number of variables rather than 3n-6 or 3n-5). In bigger systems  those extra variables can literally kill you. 

Is term of cost, is the difference between 3n and 3n-6 really big as n increases?

Total agreement with the good colleagues using keyword opt=cartesian and nosymm or symmetry=none. An outstanding summary about "How to solve the error message in Gaussian" as follow: https://docs.computecanada.ca/wiki/Gaussian_error_messages

https://www.researchgate.net/post/I_want_to_calculate_binding_energy_of_M-L_coordinate_bond_in_Schiff_base_metal_complexes_using_Gaussion_03 .

Dang T. Nguyen Thank you for your advice, I changed to Cartesian and solved the same problem with an optimization of a transition state. However, I have the following concern:

I'm using the chekpoint files from TS optimizations, to perform IRC calculations with the option Geom=Checkpoint. From my 23 transition states, only one optimization was giving me the error in internal coordinates, and I solved it using cartesian . However, when performing the mentioned IRC calculations, the one corresponding to that transition state optimized in cartesian is not working. Should I also use the option cartesian in the irc calculation for this particular ts?