[molpro-user] MP2-F12

[Hans-Joachim Werner]

Dear Glen,
the disk space in canonical MP2-F12 larger than for MP2, since all the
integrals involving F12 have to be stored. It grows approximately as
No^2 Nv^2, where No and Nv are the numbers of occupied and virtual  
orbitals,
respectively.

The DIRECT option is actually not implemented for F12 methods, so I  
wonder
how it could work for you. I'll make a patch to exit with an error  
message
if direct is attempted. You should use df-mp2-f12 instead.

If the problem "Numerical result out of range" occurs with mp2-f12 or  
df-mp2-f12
and direct is not set, please first check your output if there is any  
error
message. If not, or if it does not help, please send the output  
showing the
problem.

Best regards
Joachim


-- 
Prof. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart, Germany
Tel.: (0049) 711 / 685 64400
Fax.: (0049) 711 / 685 64442
e-mail: werner at theochem.uni-stuttgart.de



Am 11.11.2010 um 20:14 schrieb Glen Jenness:

> Dear Users,
> I have a couple questions regarding MP2-F12.
>
> 1)  Is it normal behavior to use a large amount of diskspace when  
> using
> it?  I find even when I'm using it in direct that I can end up using a
> large amount of disk.
>
> 2)  Every so often I get an error message like:
>
> "Last System Error Message from Task 2:: Inappropriate ioctl for  
> device
>  2: ARMCI aborting 0 (0).
> system error message: Invalid argument
>  8: interrupt(1)
> Last System Error Message from Task 5:: Numerical result out of range
> Last System Error Message from Task 6:: Numerical result out of range
> Last System Error Message from Task 7:: Numerical result out of range
> "
>
> I'm guessing somewhere I'm either dividing by zero or generating  
> some very
> large number-but it's not immediately clear where this error would be
> coming from.
>
> Thank you for your time!
> Glen Jenness
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user