Dear Alvin,

Statistically, a numerical frequency calculation has a higher chance of
running into
SCF convergence problems, because many geometries each need to be
treated individually.

We plan to introduce in the next ADF release, ADF2006, analytic second
derivatives for GGA functionals
as well. This implementation should not only be faster than the current
default of numerical
frequencies, but should normally also not run into the practical
problem that you are encountering now,
because only one SCF (for the equilibrium geometry) is needed.

The current version already allows such analytic frequency calculations
at the LDA level (SD program).

Best regards,
Stan

On Feb 19, 2006, at 7:30 AM, Alvin Chang wrote:

> Thank you, I'm going to try again.
> And how about the parameter of Mixing ?
> set it the default value or much smaller value, or disable it ?
>
> Best Regards,
>
> On 2/19/06, Marcel Swart <m.swart_at_few.vu.nl> wrote:
> This happens many times, and especially when you crank up the
> convergence
> of the SCF; two things may help in those cases, either increasing the
> accuracy of the integration, but more importantly, set the DIIS to
> kick in later:
>
> INTEGRATION 6.0 6.0
> SCF
>   converge 1.0e-6 1.0e-6
>   diis ok=0.01
> END
>
> or sometimes even:
>
> INTEGRATION 6.0 6.0
> SCF
>   converge 1.0e-6 1.0e-6
>   diis ok=0.001 cyc=15
> END
>
> On Saturday, February 18, 2006, at 02:21 PM, Alvin Chang wrote:
>
> > Hi, all.
> >
> > Today I met a problem about failure of SCF convergence in freq
> > calculation after the geom opt calculation.
> >
> > The end of logfile says
> > " <Feb18-2006> <23:36:46> More iterations needed. NOT CONVERGED
> > <Feb18-2006> <23:36:54> 300 ErrMat 0.00000365 MaxEl -0.00000128
> > <Feb18-2006> <23:36:55> >>>> TOTEN
> > <Feb18-2006> <23:37:14> >>>> POPAN
> > <Feb18-2006> <23:37:14> >>>> DEBYE
> > <Feb18-2006> <23:37:14> >>>> AMETS
> > <Feb18-2006> <23:37:14> SCF NOT COMPLETELY CONVERGED
> > <Feb18-2006> <23:37:14> STOP GEOMETRY ITERATIONS
> > <Feb18-2006> <23:37:14> NORMAL TERMINATION
> > <Feb18-2006> <23:37:14> END
> > "
> >
> > I customized the same parameter in opt and the following freq
> > calculation in SCF block, shown as
> > "scf
> > iterations 300
> > converge 1e-7 1e-6
> > Mixing 0.2
> > end
> > "
> > Other option was
> > converge grad=1e-4 in opt;
> > Integration 5.5 in opt , 6.5 in freq;
> >
> > In opt, SCF convergence was OK, while in freq, it failed in "
> > Frequency Displacement 29".
> >
> > Cartesian coordinates was employed in both calculations of the "
> > lanthanide + alkane " system.
> >
> > So, what's the matter of my calculations ?
> >
> > Thanks in advance.
> >
> > Best regrads
> >
> > --
> > so let us be up and doing
> > with a heart for any fate
> > still pursuing and still achieving
> > learn to labor and to wait
> ––––––––––––––––––––––––––––––––––––––––––––
> dr. Marcel Swart
>
> Theoretische Chemie (kamer R152)
> Vrije Universiteit Amsterdam
> Faculteit der Exacte Wetenschappen
>
> De Boelelaan 1083
> 1081 HV Amsterdam
> The Netherlands
>
> T  +31-(0)20-5987619
> F  +31-(0)20-5987629
> E  m.swart_at_few.vu.nl
> W  http://www.few.vu.nl/~swart
> ––––––––––––––––––––––––––––––––––––––––––––
> Starting May 1, 2006:
>
> ICREA researcher at
> Institut de Química Computacional
> Universitat de Girona
>
> Campus Montilivi
> 17071 Girona
> Catalunya (Spain)
> ––––––––––––––––––––––––––––––––––––––––––––
>
>
>
>
> --
> so let us be up and doing
> with a heart for any fate
> still pursuing and still achieving
> learn to labor and to wait
Dr. S.J.A. van Gisbergen      
Scientific Computing & Modelling NV
Theoretical Chemistry, Vrije Universiteit
De Boelelaan 1083
1081 HV Amsterdam
The Netherlands                                
vangisbergen_at_scm.com  
http://www.scm.com
T: +31-20-5987626    
F: +31-20-5987629      
Received on 2006-02-20 10:03:14