Learn how to ask questions correctly

Vitaly

korepanov@nctu.edu.tw

Dear Thomas,

thank you! Now I understand what was the problem. The explanation on the linear dependence issue in the manual is clear.

Also thank you for your tips.

Yours, Vitaly

On Sat May 3 '14 12:21pm, Thomas Pijper wrote

---------------------------------------------

>Dear Vitaly,

>It seems your system suffers from partial linear dependence, as indicated by the message:

>

* * * WARNING * * * ------------------------------------------------------------------------------ THE OVERLAP MATRIX HAS 71 EIGENVALUES BELOW 3.3E-04. THE SMALLEST OF THESE IS 5.68385E-08. THIS INDICATES A PARTIAL LINEAR DEPENDENCE IN YOUR ATOMIC BASIS.

>This is a common problem with basis sets that contain diffuse functions, such as the three basis sets that you've attempted to use. Please have a look at the manual for tips on how to remove the dependency. Feel free to post if you need assistance with this.

>A few other tips:

>- Sadlej should be used with D5=.T. See page 68 of the manual for a list of sets that should be used with this option. Enabling this option may also reduce the dependence in your basis set.

>- I do not fully agree with the use of $CONTRL ICUT=20 ITOL=30 INTTYP=HONDO $END and $SCF NCONV=10 FDIFF=.FALSE. $END for Raman calculations. I think that ICUT=20 is too tight; ICUT=11 is typically accurate enough (one can choose 12 or 13 if a higher accuracy is required). Furthermore, I think that NCONV=10 is way too tight. Combined with a partial linear dependence, such a criterion will most likely cause your SCF to never converge. NCONV=6 or 7 should be tight enough. In the case of partial linear dependence, NCONV=7 has my preference.

>

>

>Kind regards,

>Thom

>

>

>

>

>On Sat May 3 '14 4:32am, Vitaly wrote

>-------------------------------------

>>Dear Thomas,

>>here is the typical case. First input, then output.

>>Yours, Vitaly

>>

>>

>>On Fri May 2 '14 3:27pm, Thomas Pijper wrote

>>--------------------------------------------

>>>Dear Vitaly,

>>>There can be various reasons for an SCF convergence problem. Can you please post the input and output file? This would allow me and other board members to take a closer look at your problem.

>>>

>>>

>>>Kind regards,

>>>Thom

>>>

>>>

>>>On Fri May 2 '14 10:20am, Vitaly wrote

>>>--------------------------------------

>>>>Dear colleagues,

>>>>I am trying to calculate the Raman spectra of organic cation, and experience some problems.

>>>>The geometry optimization was OK, the Hessian also. But when I try to calculate Raman spectrum, I get the error "FIELD-FREE SCF CALCULATION DID NOT CONVERGE". I know what the manual says about it:

>>>>"Finite field calculations require large basis sets, and extraordinary accuracy in the wavefunction. To converge the SCF to many digits is sometimes problematic, but we suggest you use the input to increase integral accuracy and wavefunction convergence, for example

>>>> $CONTRL ICUT=20 ITOL=30 INTTYP=HONDO $END

>>>> $SCF NCONV=10 FDIFF=.FALSE. $END"

>>>>But it doen't help. I tried

>>>>1) TZV basis with 2p, d, f and two diffsp=1 diffs=1

>>>>2) aug-ccpvtz (external)

>>>>3) Sadlej pvtz (external)

>>>>with hybrid functionals (MPW1PW91 and B3LYP).

>>>>Nothing works.

>>>>Did anyone face such problems? Are there any hints to get over this?

>>>>V.

Sat May 3 '14 3:13pm

This message read