Dear Chunsik

the intensity of the Raman signal depends on the specific bond you are addressing, that is on its Raman cross section. On the other hand, the photon noise (fluctuation) follows a Poisson statistics: for N photons the noise is proportional to N1/2. If your detector "sees" only the Raman photons, the stronger the signal (or the acquisition time) the better the signal-to-noise ratio: In fact in this case the S/N ratio improves as N / N1/2=N1/2.

If you detector "sees" also another source of photons (not only fluorescence but also stray light coming from elastically scattered light) then the above consideration fails, as the noise is usually determined by this second source, which is usually much stronger than the Raman signal. In this case the Raman signal can be smaller than the fluctuations associated to the spurious photons source, and it is very difficult to retrieve it. You see the only way to improve the signal in this case is to avoid spurious photons reaching the detector, avoiding fluorescence or stray light.

I hope these cnsiderations can be useful.

Mauro

Raman is nonlinear with excitation, unlike the the fluorescence - use this

If you are experiencing fluorescence problems, then changing the excitation wavelength can sometimes help, at the expense of Raman photon production efficiency.  Excitation towards the blue end of the spectrum is most efficent at producing Raman photons, but more likely to induce fluorescence.  Moving to the red end, the opposite is true. Typically at Kaiser we use green (532 nm) for gases, where fluorescence is not an issue, red (785 nm) for most applications, since it hits a sweet spot giving some fluorescence rejection but still having good Raman photon production efficiency, and near-infrared (1000 nm) when fluorescence is the prevelent process using 785 nm.  In most cases, 785 nm is preferred because you can use good sensitive CCD cameras.

You may be seeing improvements with lower laser power because you are not saturating your detector.  Fluorescence is more efficient than Raman and so you must adjust exposure for the fluorescence you are seeing.  You can sometime subtract off a fluorescence background, but you have to sum many spectra together because the shot noise will come from the TOTAL signal.  You can subtract signals, but you can't subtract noise, so your only choice to eliminate the noise is averaging/summing spectra.

The last way to eliminate fluorescence is somewhat sample dependent, but with certain samples you can see fluorescence decrease over time -- this is called photobleaching. 

Why not just use fluorescence instead of Raman?  1) fluorescence does not have well-defined peaks, 2) fluorescence can be quenched by things like oxygen, and 3) the fluorescence signal may not relate to the component of interest.  There is a reason why fluorescence is seldom used as an analytical tool.

Nice answer.