Questions on grid size in dft

[evgeniy]

Hi,

I wonder what is the largest grid size which is possible in dft (dscf) calculations?
I am asking because there seems to be a relatively large (in my opinion) error in energy
(0.000155 hartree) even between "m4" and "7" grids. Also I wonder how
influential the "diffuse" parameter is, I mean what error in energy should one expect between
diffuse=1 and diffuse=6, especially in the case when quite diffuse basis sets are used.

Best regards,
Evgeniy

[Arnim]

Hi Evgeniy,

gridsize 7 is indeed the largest grid. You might wanna use the key words 'reference' for a combination of settings and thresholds. Just replace the gridsize line with reference, see http://www.cosmologic.de/data/DOK_HTML/node336.html#33232

The diffuse parameter might lead to smoother potential energy surfaces for non-bonded systems with diffuse basis sets. I don't know if it has any other influence.

Best,
Arnim

[evgeniy]

Hi Arnim,

Thanks very much for your reply!

By the way, there the option (keyword)

$natural orbitals

which is a bit unclear for me what it does. Could you please explain it.
Many thanks!

Best regards,
Evgeniy

[Arnim]

Hi Evgeniy,

when setting up an UHF calculation, define will asks you:
DO YOU REALLY WANT TO WRITE OUT NATURAL ORBITALS ?  DEFAULT=n
If case of yes, this will be written in control:
$natural orbitals    file=natural
$natural orbital occupation    file=natural

With pointval you can plot them, e.g.
$pointval nmo 55-60

Cheers,

Arnim

[evgeniy]

Hi Arnim,

Thanks for your detailed reply!
I guess these UHF nat orbitals are produced by diagonalizing
the UHF density, am I right.

Yes, I am interested in visualizing them. I wonder if it is possible
to generate a molden file for these orbitals?

Regards,
Evgeniy

[Arnim]

Hi Evgeniy,

the contruction is somehow connected to a Loewdin orthonormalization. The IUPAC definition is here: http://goldbook.iupac.org/NT07079.html. Better look it up in the literature.

You can cheat the tm2molden by putting the nmos in a mos file.
In control comment out (#) the uhf entries and put in the normal closed shell lines:
$scfmo   file=mos
$closed shells
 a       1-nn                                    ( 2 )
Just run
sdg "natural orbitals" > mos
and replace
$natural orbitals    scfconv=0   format(4d20.14)
with
$scfmo    scfconv=7   format(4d20.14)
Then run tm2molden. That should produce a molden.input file.

Best,

Arnim