imaginary frequencies before the translations and rotations are projected out


Just Got Here
Dear NWChem community,

I have done a frequency calculation on an optimized geometry for adenine using the cc-pVTZ basis set and the SSB-D functional. Initially, the first ten vibrational modes (out of 45 total; not copying them all here to save space since all of the imaginaries are in the first ten) contain several imaginary frequencies, leading me to believe that this structure isn't actually a minimum on the potential energy surface.
  
    1     -134.664 ||      -0.010               0.273            -0.014
    2     -127.584 ||       0.020               0.045            -0.030
    3     -113.121 ||      -0.020               0.086             0.038
    4      -97.375 ||      -0.046              -0.026             0.377
    5      -81.505 ||       0.025              -0.013            -0.193
    6      -78.833 ||      -0.031               0.035             0.172
    7      110.606 ||       0.144               0.041             0.175
    8      189.684 ||       0.104              -0.076             0.047
    9      258.368 ||      -0.017              -0.008             0.021
   10      281.460 ||       0.046              -0.100             0.331



However, after the translations and rotations are projected out, all of the imaginary frequencies are set to zero.


   
    1        0.000 ||    0.005906           0.136         5.758       1.852
    2        0.000 ||    0.000145           0.003         0.142       0.046
    3        0.000 ||    0.003436           0.079         3.349       1.077
    4        0.000 ||    0.001648           0.038         1.606       0.517
    5        0.000 ||    0.000294           0.007         0.287       0.092
    6        0.000 ||    0.001687           0.039         1.645       0.529
    7       95.360 ||    0.003451           0.080         3.364       1.082
    8      186.209 ||    0.000804           0.019         0.784       0.252
    9      257.951 ||    0.000167           0.004         0.163       0.053
   10      280.046 ||    0.005875           0.136         5.727       1.842



So my question is: do I treat this structure like it is not a minimum (and go back and look at my optimization parameters, use a finer grid, re-evaluate my functional and basis set choices, tighten convergence thresholds, etc) or do I say "well, after the translations and rotations were removed there were no imaginaries, so I'm good to go?" I'm a little uncomfortable with the second option unless anyone has some solid rationale for it.

Forum Regular
The short answer is that you probably want to re-optimize the geometry and re-calculate the normal modes with tighter thresholds (you shouldn't have to re-evaluate your functional/basis set choices unless you have other reasons for doing so). For a little more discussion on the matter, I recommend the following white paper, http://gaussian.com/vib/


Forum >> NWChem's corner >> General Topics