rodft, frequency calculation


Clicked A Few Times
Hello All
 Can anyone please tell how to do frequency calculation in rodft level? Is analytical hessian is avaliable with rofdt?

Thank You
Sujitha

Gets Around
I don't know if any of the DFT methods have analytical frequencies. Here's an example of a rodft calculation.

Title "no2-geovib"

Start  no2-geovib

echo

charge 0

geometry units angstroms print xyz
  O        0.75760       -0.04378        0.06825
  N        2.09353       -0.04378        0.06825
  O        3.42945       -0.04378        0.06825
  symmetry c1
end

basis "ao basis" cartesian print
  * library "6-31G*"
end

dft
  mult 2
  rodft
  XC m06
  mulliken
end

driver
  default
end

task dft optimize
task dft freq

Forum Regular
Hi,

NWChem supports analytical DFT frequencies for closed shell and unrestricted open shell calculations. There are some additional limitations due to the requirement for 2nd order partial derivatives of the functional you are using. At the moment we don't have these for all functionals but I hope to have that problem fixed soon.

For RODFT the story is a bit different as there is an additional constraint (additional relative to unrestricted open shell) that the alpha and beta orbitals should be the same. This leads to an additional Lagrangian in the gradient expression with has been coded in NWChem. But we haven't worked out what this additional term means for the frequency calculations. So, you are right that if you do an RODFT frequency calculation then the 2nd derivative will be constructed by numerically differentiating the gradient.

Huub

Clicked A Few Times
Thank you for the responses.

The RODFT requires with "CGMIN" . But the frequency calculatios were not possible with this option. My attempts to perform frequency calculation by removing this keyword was also unsuccessful. The options I tried for getting frequencies were as follows. As I mentioned both of these options were unsuccessful.

Thank You.
Sujitha

OPTION 1

dft
 xc b3lyp
rodft
cgmin
mult 1
decomp
tolerances tight
grid xfine nodisk
end

task dft optimize
task dft freq numerical

OPTION 2


dft
 xc b3lyp
rodft
cgmin
mult 1
decomp
tolerances tight
grid xfine nodisk
end

task dft optimize

dft
nocgmin
end

task dft freq numerical

Gets Around
bug with DFT grid nodisk
Interesting, there seems to be a bug in using geometry optimization with DFT grid 'nodisk' option under these circumstances. This input runs quickly and terminates normally:

Title "no2-geovib"

Start  no2-geovib

echo

charge 0

geometry units angstroms print xyz
  O        0.75760       -0.04378        0.06825
  N        2.09353       -0.04378        0.06825
  O        3.42945       -0.04378        0.06825
  symmetry c1
end

basis "ao basis" cartesian print
  * library "6-31G*"
end

dft
  mult 2
  xc b3lyp
  rodft
  cgmin
  decomp
  tolerances tight
  grid xfine
end

task dft optimize
task dft freq numerical


This input produces progressively higher energy geometries until it runs out of optimization steps:

Title "no2-geovib"

Start  no2-geovib

echo

charge 0

geometry units angstroms print xyz
  O        0.75760       -0.04378        0.06825
  N        2.09353       -0.04378        0.06825
  O        3.42945       -0.04378        0.06825
  symmetry c1
end

basis "ao basis" cartesian print
  * library "6-31G*"
end

dft
  mult 2
  xc b3lyp
  rodft
  cgmin
  decomp
  tolerances tight
  grid xfine nodisk
end

task dft optimize
task dft freq numerical


@ Step       Energy      Delta E   Gmax     Grms     Xrms     Xmax   Walltime
@ ---- ---------------- -------- -------- -------- -------- -------- --------
@    0    -204.97333967  0.0D+00  0.22447  0.10581  0.00000  0.00000      4.4
@    1    -204.95545153  1.8D-02  0.18350  0.08650  0.02951  0.06260      8.8
@    2    -204.94675962  8.7D-03  0.16714  0.07879  0.01340  0.02843     12.8
@    3    -204.94250996  4.2D-03  0.15979  0.07532  0.00640  0.01357     16.7
@    4    -204.93999303  2.5D-03  0.15562  0.07336  0.00375  0.00795     19.4
@    5    -204.93749968  2.5D-03  0.15160  0.07147  0.00369  0.00782     22.2
@    6    -204.93503032  2.5D-03  0.14774  0.06965  0.00363  0.00769     25.1
@    7    -204.93258045  2.4D-03  0.14402  0.06789  0.00358  0.00758     28.7
@    8    -204.93005490  2.5D-03  0.14027  0.06613  0.00381  0.00778     33.6
@    9    -204.92697342  3.1D-03  0.13582  0.06403  0.00492  0.00949     61.0
@   10    -204.92422775  2.7D-03  0.13195  0.06222  0.00435  0.00844     66.9
@   11    -204.92145431  2.8D-03  0.12812  0.06043  0.00439  0.00854     82.9
@   12    -204.91882836  2.6D-03  0.12458  0.05877  0.00417  0.00808     89.2
@   13    -204.91620944  2.6D-03  0.12112  0.05716  0.00417  0.00807     96.7
@   14    -204.91367270  2.5D-03  0.11783  0.05564  0.00406  0.00783    104.7
@   15    -204.91120819  2.5D-03  0.11469  0.05419  0.00396  0.00762    112.6
@   16    -204.90881147  2.4D-03  0.11169  0.05280  0.00388  0.00743    121.1
@   17    -204.90640494  2.4D-03  0.10873  0.05144  0.00390  0.00747    128.4
@   18    -204.90411816  2.3D-03  0.10596  0.05017  0.00374  0.00713    135.9
@   19    -204.90189926  2.2D-03  0.10331  0.04896  0.00367  0.00694    142.7
@   20    -204.89974575  2.2D-03  0.10077  0.04780  0.00360  0.00677    149.3
@   20    -204.89974575  2.2D-03  0.10077  0.04780  0.00360  0.00677    149.3


EDIT: In case it matters, these results were obtained with the latest NWChem snapshot: http://nwchemgit.github.io/download.php?f=Nwchem-dev.revision25716-src.2014-06-09.tar.gz

I compiled using OpenBLAS 0.28 as BLAS and was running under 64 bit Ubuntu 14.04.

Forum Vet
Matt
Thanks for the bug report.
The gradient generated with 'grid nodisk' and rodft/cgmin are incorrect.
A fix has been checked into the svn repository


Forum >> NWChem's corner >> General Topics