Hello Maxime,

Maybe the initial guess is not too much near of the point. A thing that you can try is change a little bit the geometry (bond distances, angles...), also, in order to do it fast, you could try with lower basis set and without "calcall" (which is more useful when only the two last parameters says NO and you think that the results is good). Other think to see is, if there is any imaginary freq (if it is yes, the initial guess is wrong).

In addition, depending on the system, it could be a "DFT problematic system". If you are not much sure about it, you can perform a MP2 calculation in order to ensure that is a geometric problem and not a DFT one. Remember in add extra RAM memory to increase efficiency for MP2 calculations.

I hope it helps you,

Joaquim

It depends on your structure’s size, reality, atoms and ..., . I suggest start with semi-empirical methods and then with HF. Subsequently, try with M06/(a small basis set) as  Finally you can optimize it with your demanded method. If this does not work you can test quadratically convergent SCF procedure (scf=qc) if your calculation is not restricted open shell.

Good luck!

Dear Maxime.

Try:

Opt=(Z-matrix,MaxCycle=300,CalcFC) GFINPUT IOP(6/7=3)

Good luck,

Rafik

As suggested above, you're most likely stuck in a geometry that is not the minimum you're looking for and gaussian optimized can't move in any reasonable direction. Modify the geometry slightly and try reoptimizing - if it doesn't help modufy the geometry more and try reoptimizing.

SCF(qc) or SCF(xqc) also helps in some cases. Other ideas: you can try also opt=z-matrix or opt=cartesian - these have usually completely different convergence process that the redundant coordinates (it doesn't matter how our input geometry is specified - this is a completely different problem). 

Finally, make sure you;re using Int=Ultrafine (for any DFT), adding opt=tight is also usually a good idea. 

If none of these (try different combinations of the keywords above) help, change the DFT functional to a similar one.

You can limit the displacement with the opt=maxstep. If your structure is oscillating too much, sometimes trying a lower maxstep can help it converge into a narrow minimum.

Hello Maxime,

Does it shows signs of convergence or does it oscillate? I would suggest first trying to converge using another method, either Hartree-Fock with a small basis set or a semi-empirical method like HF-3c or PM6. Then go back to DFT. Why choosing the M06 exchange-correlation functional with Pople's basis set? What type of system is this (fully organic or are metals involved?)

Best

Gilles

Use this link, it has covered your problem with detailed explanation:

http://gaussian.com/faq3/

Hope this helps... :-)

Regards

Ajay Khanna