frequency calculations

[apple]

I am doing vibrational frequency calculations at the gamma point for a metal surface with Au dimer adsorbed on it. I get two imaginary frequencies, which would imply a second order saddle point. However, one of the frequencies is very small (8.5 cm-1), so I don't know whether I should consider it important to call the configuration a second order saddle point. I used stringent parameters to optimize the structure (EDIFFG = -1E-03, EDIFF = 1E-05) before running freq. calculation with EDIFFG = -1E-04, EDIFF = 1E-06). I tried to reoptimize the structure by moving the atoms in the direction of the imaginary frequency vibration, but I would always end up with this dimer configuration and two imaginary frequencies. Given the accuracy of VASP, are the two imaginary frequencies large enough for the structure to be called a second order saddle point?

Below see the output:
1 f = 7.389286 THz 46.428252 2PiTHz 246.480034 cm-1 30.55964 meV
2 f = 7.331773 THz 46.066889 2PiTHz 244.561616 cm-1 30.321787 meV
3 f = 6.244021 THz 39.232338 2PiTHz 208.278101 cm-1 25.823203 meV
4 f = 6.2146 THz 39.047486 2PiTHz 207.29675 cm-1 25.701531 meV
5 f = 5.93462 THz 37.288317 2PiTHz 197.957607 cm-1 24.543624 meV
6 f = 5.745973 THz 36.10301 2PiTHz 191.665006 cm-1 23.763441 meV
7 f = 5.57623 THz 35.036487 2PiTHz 186.003008 cm-1 23.061442 meV
8 f = 5.439878 THz 34.179763 2PiTHz 181.454799 cm-1 22.497536 meV
9 f = 5.258353 THz 33.039207 2PiTHz 175.399774 cm-1 21.746808 meV
10 f = 5.001709 THz 31.426661 2PiTHz 166.839033 cm-1 20.685411 meV
11 f = 4.5607 THz 28.655722 2PiTHz 152.128565 cm-1 18.861545 meV
12 f = 4.554786 THz 28.618567 2PiTHz 151.931314 cm-1 18.837089 meV
13 f = 3.263116 THz 20.50276 2PiTHz 108.845816 cm-1 13.495166 meV
14 f = 2.483955 THz 15.607151 2PiTHz 82.855824 cm-1 10.272817 meV
15 f = 2.461872 THz 15.4684 2PiTHz 82.119218 cm-1 10.181489 meV
16 f = 1.108203 THz 6.963043 2PiTHz 36.965661 cm-1 4.58316 meV
17 f/i= 0.254954 THz 1.601921 2PiTHz 8.504341 cm-1 1.054404 meV
18 f/i= 0.626128 THz 3.934078 2PiTHz 20.885379 cm-1 2.589458 meV

Can someone comment, please.

[alex]

Hi apple,

I'm using far tighter criteria for structure optimisation. E.g. EDIFFG = -1E-04, EDIFF = 1E-06 down to EDIFFG = -1E-06, EDIFF = 1E-08.
However, if you are doing the second derivatives numerically, you'll end up with some imaginary modes around zero wavenumbers. The two you have, are a little too larger if you want to do it precisely. If it's just for a rough overview, I think it's ok.
If you are doing statistical thermodynamics afterwards, you have to become better.

Cheers,

alex

[apple]

Thanks Alex for your comment. I will try to reoptimize the structure with more stringent criteria to see what happens.