major difference between pbe and b3-lyp

[foea]

Hello ^_^

I am currently working with a carbonnanotube structur (see attachment) and came across the following problem:
I wanted to calculate the energy with the PBE-functional but even after some days not even the first iteration step had finished. But when I used the B3-LYP-functional under the same computational conditions in about an hour the first iteration step was completed and in some hours the whole calculation had finished.
Since for comparison reasons I would prefer to use the PBE-functional I would be gratefull if you could tell me what I could change so that the calculation will run in a reasonable time with PBE.
(Note: For PBE I also used ri and marij with the default settings. Both calculations were carried out with the def2-TZVP basis in version 6.2)

Thanks, foea

[uwe]

Hello,

was that a parallel run?

Regards,

Uwe

[foea]

Yes it was a parallel run.

[uwe]

Hello,

RI-DFT needs more memory, with the default settings about 1 GB for your input with def2-TZVP basis set. If the calculation is stuck before the first SCF iteration is done (which should take on a single core about 5 minutes), this looks like a problem with the linear algebra when treating the RI matrices.

Did you set the stack size limit as described here: http://www.turbo-forum.com/index.php?topic=23.0 ?

Some people also had problems with the parallel linear algebra routines in Turbomole 6.2 which lead to a couple of changes in the source code and the possibility to switch of the usage of ScaLAPACK (http://www.netlib.org/scalapack/ in its Intel-optimized version http://software.intel.com/en-us/articles/intel-mkl/) in Turbomole 6.3.

If increasing the stack size limit (by changing /etc/security/limits.conf) does not help, upgrading to the latest Turbomole version should do the job.... just contact the support.

Regards,

Uwe

[foea]

Hello,

Increasing the stack size limit did not help.

Greetings,
foea