The best achievable accuracy depends on the spectrometer, of course. The light source used for calibration comes on second place. You can use for instance a low-pressure calibration lamp (e.g., Hg, Ar, Xe, Ne, etc.) which produce sharp and narrow spectral lines with wavelengths known to 5 digits. This is sufficient for spectrometers of resolution 0.5 - 1 Å (e.g., ~50 cm focal length) equipped with CCD cameras. For such spectrometers the dependence wavelength versus pixel number on the CCD is nonlinear and even if the non-linearity is accounted for using the correct formula, small errors in the parameters of this formula (focal length, included angle of the spectrometer, etc.) will produce small errors in the measured absolute wavelengths of the Raman peaks, which, however, will be larger than the accuracy of the calibration. Double monochromators with equipped with photomultiplier and entrance/exit slits are another story. In that case the spectrum is obtained by subsequent acquisition of the spectral intensity for each wavelength. The main factor affecting the accuracy is the repeatability of the mechanical movement of the spectrometer which, of course, depends on the quality and condition of the mechanics. Good spectrometers like SPEX 1404, for example, will have maximum resolution of 0.05 Å and repeatability better than that!

Finally, the best calibration just for Raman measurements is probably the line of the exciting laser itself. Even it is known with not so good accuracy (say, 532 nm instead of the real value of 531.9 nm, i.e., absolute error of 0.1 nm), the Raman line positions are always relative to the laser line and they will appear at wavelengths which are shifted approximately by the same amount (0.1 nm in our example) from their real wavelengths. (Provided that the "wrong" laser line of 532 nm instead of 531.9 nm is used for calibration.) However, the calculated Raman shifts will have much better accuracy even though they have been calculated using slightly wrong wavelengths (of the laser and the Raman lines). One can easily check that the same absolute error in the laser wavelength and the measured absolute wavelength will not introduce any notable error in the calculated Raman shift.

Thank you, Ivan.

I agree with Ivan Ivanov's answer, but another phenomenon should be noticed, the wavelength of the laser may be has the relationship with the laser power. I use Verdi-V5 from Coherent, the wavelength at very low power is different with the wavelength at high power.

Good point, Ke Lin! Therefore, if the calibration is done with the laser wavelength, it should be done when the power is fixed before the measurement of the Raman spectrum.

Isn't it better if we use some Raman shift frequency standards(for e.g. cyclohexane, naphthalene,benzene etc. ) as they do not depend on laser wavelength.

after all the mentioned important notes, you may want to use a piece of silicon wafer and do the Raman spectra. If your instrument works well, you have to be able to see a sharp peak at 520 or 521 cm-1 (wavenumber). Silicon is absolutely the best material people use for calibration purposes. It has no fluorescence effect and no photo bleaching effect and if your instrument is OK you will see a sharp peak.

Thank you, dear colleagues. It is helpful discussion.

Using the 520.7 cm-1 peak of silicon as a standard can be a good way to proceed, as highlighted by Hossein Taghinejad, because of a good knowledge about the peak position (see related papers). Furthermore, you can change the laser line wavelength and use the same standard, another convenient thing about using silicon. Don't forget to check the alignement each day because small variations can be introduced in a short time, even in rooms where temperature is perfectly controlled.

Just make sure to calibrate over the entire range of the spectrometer you wish to collect data. The laser line should be about zero, but may differ if you do not calibrate in that range. Raman shifts are of course relative, so the best choice depends on the spectral range you are interested in. The Dr. McCReery has a good list on his group page. (I would also give his spec book a read and the website can be found through google.)

I di agree with the whole spectral range calibration and the use of different standands. The alignment must be checked daily also (even sometimes twice a day as it may change if the room is not temperature-controlled).

I use Indene (Raman Shift), Carbon tetrachloride (Raman Shift), Standard Neon Lamp and Silicon wafer ; depending upon the spectral range and the laser to calibrate the Raman system.

According to Mc Creery's works, the following standards have been tested on different instruments:

1. naphthalene

2. 1,4 bis (2-methylstyryl) benzene (BMB)

3. sulfur

4. 50/50 (v/v) toluene/acetonitrile

5. 4-acetamidophenol

6. benzonitrile

7. cyclohexane

8. polystyrene

We often use Teflon as a standard.

Diamond (1332 cm-1) provides a useful calibration line, as do the Raman modes of atmospheric oxygen and nitrogen.

Currently developing a calibration library with Si, PS, LiNbO3.  As I shift gears into the over 1000 cm-1 I find that I want a few more materials.  This thread has pointed me to the right direction.