Raman intensities in version 6.1 and 6.3 differ

Just Got Here

7:08:10 AM PDT - Tue, Aug 26th 2014

Hi there,

running the same Raman job with version 6.1 and 6.3, I get very similar vibrational frequencies but largely differing Raman intensities:

Excerpt from 6.1 .normal file:

 23  1001.75        0.7072

 24  1084.59        1.5235

 25  1096.68      208.5233

Excerpt from 6.3 .normal file:

 23  1001.75        0.3229

 24  1084.59        0.8551

 25  1096.68       50.6998

Is anybody aware of major changes in the Raman part of the code (or the CPKS solver)?
Have any defaults related to the evaluation of Raman intensities changed in 6.3?

Any comment is appreciated!
- Thomas

Here is the input deck:
START 4mba
MEMORY total 1000 mb
TITLE "4MBA : 6-31G* raman"
GEOMETRY NOCENTER NOAUTOZ NOAUTOSYM

LOAD format xyz "4mba.opt.xyz"

END
BASIS SPHERICAL

* library 6-31G* file /misc/condmat/niehaus/nim29319/opt/nwchem-6.3/src/basis/libraries/

END
DFT

DIRECT

ITERATIONS 500

XC xpbe96 cpbe96

GRID nodisk

END
DRIVER

MAXITER 100

END
RAMAN

NORMAL

LORENTZIAN

LOW 1000

HIGH 1800

WIDTH 15.0

END
PROPERTY

RESPONSE 1 8.8559e-2

DAMPING 0.007

END
HESSIAN

THRESH 1d-6

END
TASK DFT raman

Gets Around

1:48:17 PM PDT - Tue, Aug 26th 2014

I can't pinpoint when the change occurred but there is definitely a change. I found large intensity differences looking at an old version of the h2o_raman_3 test case vs. the current version.

The file release.notes.6.1 appeared at svn revision 21888.

Here's the old reference log file, introduced at revision 21120 : http://pastebin.com/jC0EG7x6
And the new reference log file, introduced at revision 23607: http://pastebin.com/Sr1HvR7V

Data from .normal file, 21120:

 #  freq [ 1/cm]  S [Ang**4/amu]
 ------------------------------------------------------------
   1  1591.50        0.8354
   2  3680.21      115.5964
   3  3783.86       28.4166
 ------------------------------------------------------------
  1533.82    0.2982E-04
  1536.13    0.3221E-04
  1538.45    0.3490E-04
  1540.76    0.3796E-04
  1543.08    0.4143E-04
  ...

Data from .normal file, 23607:

 #  freq [ 1/cm]  S [Ang**4/amu]
 ------------------------------------------------------------
   1  1591.48        0.8353
   2  3680.25      115.5960
   3  3783.91   259483.9293
 ------------------------------------------------------------
  1533.79    0.1541E-02
  1536.11    0.1546E-02
  1538.42    0.1552E-02
  1540.74    0.1558E-02
  1543.05    0.1565E-02
  ...

The .normal files start out similarly in both versions but intensities soon diverge wildly. I don't know if the values generated by newer code are a vast improvement or the opposite.

The first notable difference in the log files is that under each DFT Final Molecular Orbital Analysis the MO Center values printed are significantly different. The second is that under NWChem Dynamic CPHF Module output the number of equations solved and maximum subspace are different:

21120:

Iterative solution of linear equations
  No. of variables     1740
  No. of equations        3
  Maximum subspace       30

23607:

Iterative solution of linear equations
  No. of variables      870
  No. of equations        3
  Maximum subspace       33

In between the commits of the old and new reference .normal files there are a few possibly relevant commits that touch the Raman property code:

------------------------------------------------------------------------
r23263 | niri | 2012-12-09 10:38:17 -0800 (Sun, 09 Dec 2012) | 1 line
Changed paths:
   M /trunk/src/property/CalcPerturbedTDPmat1.F
   M /trunk/src/property/aor_r1_beta.F
   M /trunk/src/property/aor_r1_beta_anl.F
   M /trunk/src/property/aor_r1_tensor.F
   M /trunk/src/property/aoresponse_driver_new.F
   M /trunk/src/property/fiao_b1_movecs.F
   M /trunk/src/property/fiao_f1_movecs.F
   M /trunk/src/property/giao_b1_movecs.F
   M /trunk/src/property/hnd_giaox_zora.F
   M /trunk/src/property/hnd_gshift_zora.F
   M /trunk/src/property/prop_input.F
   M /trunk/src/property/prp.F
   M /trunk/src/property/raman.F
   M /trunk/src/property/raman_input.F
   M /trunk/src/property/task_raman.F

response updates from Fredy

------------------------------------------------------------------------
r22112 | niri | 2012-03-08 18:16:15 -0800 (Thu, 08 Mar 2012) | 1 line
Changed paths:
   M /trunk/src/property/prop_input.F
   M /trunk/src/property/raman.F
   M /trunk/src/property/task_raman.F

raman fixes for static polarizability

Gets Around

4:02:39 PM PDT - Tue, Aug 26th 2014
Apart from the different intensity magnitudes, the plotted curves do not even have similar shapes. This is again for h2o_raman_3 .normal data, as created by NWChem revision 21120 vs. 23607. Revision 21120: http://www.sciencemadness.org/cc/21120.png Revision 23607: http://www.sciencemadness.org/cc/23607.png

Forum Regular

5:06:11 PM PDT - Tue, Aug 26th 2014
Hi, We checked in a few bug fixes earlier this year. We are in the process of finalizing the 6.5 release and will look into this. I will let you know. Thanks. Best, -Niri niri.govind@pnl.gov

Forum >> NWChem's corner >> Running NWChem