about vibrational frequency

[candy]

Dear guys,
I have a question about the vibration frequency calculation. I use two methods to calculate vibrational frequency of a same system, but the results are different.
the system: O2---> 2O on Pd/Au(100) surface. At beginning, the saddle point of the reaction was searched by using dimer method, and then I calculated the vibrational frequencies of the saddle point by two methods ( relax two oxygen atoms, freeze the other atoms in both methods)
Method (Ⅰ)
INCAR file :
--------------------------------------------------
SYSTEM = O2 on Pd/Au (100)
ISTART = 0 ; ICHARG=2
ISMEAR = 2 ; SIGMA = 0.2
IALGO = 48 ;ENCUT=400
ALGO = V ; EDIFF = 1E-6
PREC=normal
dynamic:
NSW=100; POTIM = 0.024
IBRION = 5; NFREE = 2
LREAL = A
-------------------------------------------------
There are six frequencies in OUTCAR:
f( cm-1) 381.64, 362.41, 332.20, 318.03, 192.57, -278.35
there is only one imaginary frequency of the saddle point.
Method (Ⅱ):
1)Making DISPLACECAR : dymseldsp.pl (POSCAR1) (POSCAR2) (atom to include) displacement
atom to include=2, displacement =0.005
2)Doing dynamical matrix calculation based on information on website : http://theory.cm.utexas.edu/vtsttools/dynmat/
INCAR file:
-------------------------------------------------
system= O2-2O DM calculation
ISTART=0; ICHARG=2
ENCUT=400
ISMEAR=2; SIGMA=0.2
IBRION=3 ;POTIM=0
EDIFF=1E-7; NSW=7
LREAL=Auto; ICHAIN=1
PREC=normal
--------------------------------------------------
3)producing freq.dat : using dymmatrix.pl to produce freq.dat for saddle point
freq.dat:
270.335069 cm^{-1} ... 1
99.485934 cm^{-1} ... 1
147.534581 cm^{-1} ... 0
311.180434 cm^{-1} ... 0
366.284366 cm^{-1} ... 0
428.587689 cm^{-1} ... 0
The result of the two methods are different, there is only one imaginary frequency in methodⅠ , however, the number of imaginary frequency become 2 in method Ⅱ.
Why the results of two method are different? Which one is correct? Did I find the right saddle point? Please give me some suggestion, thank you in advance!

[graeme]

You could try lowering the displacement to 0.002 in the second method and see if the frequencies change, or if they are converged with respect to the finite difference step size.

Also, you should try to include some of the surface metal atoms. Again, check to see if the modes of the molecule are unchanged. I would expect that including the nearest surface atoms will be important for the normal modes.