Author Topic: egrad error: nfun > 10  (Read 7401 times)

sv

  • Newbie
  • *
  • Posts: 6
  • Karma: +0/-0
egrad error: nfun > 10
« on: May 30, 2007, 11:14:01 PM »
Hi all,
I am trying to calculate
gradient for nth excited state of a molecule and I get this error when I run egrad:

 2e-integral 1st. derivatives will be neglected if expon. factor < 0.100000E-06
 outer boundary for 2e-integral derivative neglection is 0.191406E-08

 nfun > 10 not supported by <edabc>
 
 MODTRACE: no modules on stack

 nfun > 10
 egrad ended abnormally
 egrad ended abnormally


I used b3-lyp method with TZVP basis set.
Please advise.
Thanks in advance

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 486
  • Karma: +0/-0
Re: egrad error: nfun > 10
« Reply #1 on: May 31, 2007, 10:57:21 AM »
Hi,

you should have been a bit more precise: you are not using B3-LYP, but B3-LYP_Gaussian - right?

B3-LYP_Gaussian has mainly been implemented to make it possible to compare Turbomole total energies with those of old Gaussian calculations, to avoid re-calculation of old Gaussian jobs with Turbomole.

Since there are no TDDFT or TDHF analytical gradients available in Gaussian, there is nothing to compare with. Therefore B3-LYP_Gaussian has not been tested with egrad, and calculations with that functional will not be done.

In Turbomole 5.9, the error message should look like that:

Functional nfun > 10 not supported by <edabc>
Please use b3-lyp instead of b3-lyp_Gaussian !!


Regards,

Uwe

sv

  • Newbie
  • *
  • Posts: 6
  • Karma: +0/-0
Re: egrad error: nfun > 10
« Reply #2 on: June 05, 2007, 07:22:31 PM »
thanks uwe,
I figured that out. I was doing b3-ylp_gaussian first. Thanks.

wiking

  • Newbie
  • *
  • Posts: 4
  • Karma: +0/-0
Re: egrad error: nfun > 10
« Reply #3 on: December 20, 2007, 08:02:42 PM »
Hi!

I'm worried, because I've just got this message (Please use b3-lyp instead of b3-lyp_Gaussian) from egrad. More jobs are waiting for excited state geometry optimization...
1. Is that mean that my optimization will failure?
2. Do I have to start once again from basic state optimization with b3-lyp functional (not _Gaussian)?
3. Is that error related only to b3-lyp_Gaussian?

Thanks in advance!
Marcin
« Last Edit: December 20, 2007, 09:11:43 PM by wiking »

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 486
  • Karma: +0/-0
Re: egrad error: nfun > 10
« Reply #4 on: December 21, 2007, 11:43:47 AM »
Hi,

there is just one single reason why you should use b3-lyp_Gaussian at all in Turbomole: If you want to compare total ground state B3-LYP energies with old Gaussian jobs. For all other purposes: use b3-lyp and forget the b3-lyp_Gaussian functional.

The b3-lyp_Gaussian functional was only implemented for that purpose, and the functional derivatives have only been tested up to the level that is needed for that comparison. Programs that use higher derivatives will stop to prevent the usage of an untested feature. The error message is therefore nothing to worry about.

It seems that you are doing excited state geometry optimizations, and hence there are no total ground state energies to compare.

If you do Turbomole calculations, please just choose the b3-lyp functional (without _Gaussian) in the future.

Regards,

Uwe