Hello Kafoumba,

-First of all you could try with modify the inital guess (imaginary frequency means that you start near a TS and not close to a minima). 

-You also could try MP2 method or other dft. B3LYP is not a very good functional for metals, you could change to B3PW91 or others.

-Maybe the problem comes from the basis sets (but I think that is rare). In this case, you could change to other basis sets with different pseudo-potentials. See the link in order to acquire them. If you have some problem in introduce basis set manually, sent to me a message.

Finally, I also attach a guide about optimization that could be useful to you.

I hope it helps you,

Joaquim Rius

https://bse.pnl.gov/bse/portal

Technical Report Gaussian Guide. Section 2. Optimization Methods. DFT, MP2 and CCSD.

thank you Joaquim, I can not help acknowledging your help. so I will try with functional B3PW91. however,  my problem is that, between five isomers of the same complex, four converge to the minimum state with the same method.  only the trans one does not. therefore, it becomes difficult to compare them. so to perform the comparison, it becomes compulsory for me to find out a technique to come through the issue.

You know, I have always worked with B3LYP since I study ruthenium complexes. it is the first time the optimization of a complex leads to TS.

my best regards

I think that this imaginary frequency do not correspond to a realistic TS, but an structure with one imaginary freq sure is not a minima on the PES. On the other hand, B3PW91 could imply an improve to the B3LYP results but it is not "popular". Even so, sometimes is found some systems that converge with one or other. 

Also, the interaction in the new position of the ligand can imply new effects. Could I see one correct isomers and with the imaginary frequency, you could send via private message or mail? I hope I could help you better if I have more information.

thank you again for your help Joaquim.

I've even tried the functional B3PW91,  I still have the same imaginary frequency.

I have enclosed the input file.

the regarding complex is the trans isomer.

if you open the file with GaussView, you find out in detail the method I used for calculation. whereas the output file, I ve some trouble uploading it. perhaps if I have your mail, I will send it to you in due time. meanwhile, the results that indicate the imaginary frequency are given below

B3PW91/LANL2DZ: NImag=1 and freq= -19.40 cm-1

B3LYP/LANL2DZ: NImag=1 and freq=-17.84 cm-1

see you.

Hi Kafoumba,

Being at Rutgers University (RU), I have an obligation to be interested in Ru complexes :-;

Your trans complex does not want to have C2 symmetry.  I tried PBE and Hartree-Fock (with LANL2DZ) and the calculation immediately breaks symmetry (to C1). It is possible to stay in C2 symmetry with B3LYP and B3PW91 but you end up in a TS with the low value imaginary frequencies you show above. The vibrational mode is a rocking/twist of the entire complex and has B symmetry (not A symmetry). Hence, following the mode will break point group symmetry from C2 to a lower symmetry point group. There is only one lower symmetry point group to go to, namely C1 (no symmetry). The trans complex does not want to be in C2 symmetry.

The energy lowering upon distortion is small, 0.42 kcal/mol ~ 150 cm-1 with B3LYP/LANL2DZ (E(C2) = -1759.04414944, E(C1) = -1759.04482272), which is fully consistent with the very small value of the imag freq.

i wouldn't worry about it. The distortion may be related to the wave functions being 'unstable' (in C2 symmetry) but that is another can of worms.

Best

kk-j

PS. I noticed that you are asking for very little memory with the %mem command. Your calculation(s) would go a lot faster, if you actually have access to more memory (and processors).

Dear kafounba

Have you tried to make your optimization without using

Any symmetry constrain?  If you used try to remove it. Once i did 

Some computatipns at Ru complexes by using DFT/b3lyp mrthod with success. Also be aware that all these complexes suffer by multireference states, single reference approximation might be inadequate.Check this by a small Cas(2,2), that is 2 el distributed in 2 orbitals. The occupation numbers provide an indication for the reliability of the single reference approach.

Dear Joaquim, Karsten and Aggelos,

I'm thanking you for  helping me. sorry for being late, I've 1 processor computer to do my calculations. therefore, it must take many days before I get a result. anyway,

Joaquim:

as you advised me, I tried to move one of the naphtol group towards the other Cl atom. subsequently, the structure changed from C2 to Ci symmetrical. thus, when I performed the calculation with the same method (B3LYP/Lanl2dz) then I got a minimum local with NImag=0. but the energy here is higher than that of C2 symmetry which has NImag=1. I ve respectively G(Ci) =-1756.00 and G(C2) = -1758.64. what do you think about both results? must I keep the Ci or try again to improve until I get C2 symmetry with minimum local?

Karsten:

actually, the trans-RuCl2(Nazpy)2 shall display a C2 symmetry.  I think that this molecule must necessarily be symmetrical though both Cl atoms are in opposite position. therefore, both ligands Nazpy shall be equivalent. so I'm convinced that having C1 as symmetry for that molecule is not suitable.

Aggelos:

why do you remove the symmetry? to me the objective of theory tends to predict and to explain the experience. If I remove the symmetry constrain, all the five molecules will be C1 symmetrical. therefore, performing a calculation like NMR will be far from reality. don't you think?

Anyhow, I expect keeping in contact with you to discuss theoretically about ruthenium complexes.

my best regards

If G(Ci)-G(C2) ~ -1756.00 - (-1758.64)  ~2.6 Hartrees ~ 1500 kcal/mol with the same method/basis set, then the Ci structure is useless.

I agree fully that the optimized trans structure 'should' display C2 symmetry according to conventional chemistry wisdom and experience. But it doesn't, at least not with a single determinant method like standard DFT. There is a very, very small imaginary frequency at the stationary point located with C2 symmetry imposed. The energy difference, E(C2) - E(C1) ~ 0.5 kcal/mol (see above), is very small, so bond lengths and angles are very similar in the two structures.

The single determinant wave function used in DFT seemingly does not want to display C2 symmetry for trans-RuCl2(Nazpy)2. It is unstable to symmetry relaxation to C1. Gaussian has the capability to analyze the wave function further to see whether it is a symmetry problem (which it certainly looks like) or a restricted vs unrestricted problem (the keyword  is 'stable'---with some options).

This type of problem ('it should have XY symmetry but the calculation seemingly says XZ symmetry, where XZ is a point group of lower order than XY) is not unique to Ru complexes. It happens fairly regularly with transition metal complexes, because the single determinant wave function approximation breaks down more readily for these types of systems. Single determinants work great for most closed-shell organic molecules.

kk-j

Dear Kafounba,

Keep in mind that a geometry optimization is a symmetry conserving process.

What ever symmetry is denoted at the beginning (C2v, D2H)..) this will be retained throughout the optimization procedure. In order to prove that you reached a minimum, either global/local, you have to compute the vibrational spectrum. If at least negative freq. exists, this means that the initial symmetry  does not correspond to a minimum structure rather than a transition state.

Aggelos

Be careful with symmetry and G09. G03, G98, .........G80 would not break point group symmetry spontaneously during geometry optimization. If the input geometry was D2h, say, and the wave function after a few optimization steps only had D2 symmetry (i.e. the molecule would go non-planar), then the job would stop with a statement like: 'Omega........change of point group detected. Optimization stopped'.

G09 will change the point group symmetry during optimization without telling you. That's what happened in my HF and PBE calculations on 'Kafounba's Complex'.

You have to enforce the symmetry in G09, if you definitely want a specific point group.

There is little doubt that a proper wave function for trans-RuCl2(Nazpy)2 will generate a C2 symmetry molecular structure. It looks like single determinant approaches are not sufficiently proper.

kk-j

A Transition State (TS) is a first-order saddle point on the Potential Energy Surface (PES). At a TS, the matrix of second derivatives of the energy with respect to nuclear coordinates (the Hessian') has one negative eigenvalue (negative force constant), hence the presence of a so-called 'imaginary' vibrational frequency..It is not necessary to have the vibrational frequency larger than 100i cm-1 for the stationary point to be characterized as a TS (of some sort).