We report for the first time on the thorough experimental and theoretical assessment of the noise limit of mechanical stress and temperature measurements with micro-Raman spectroscopy. A comprehensive study has been performed in which, for different incident laser light intensities and acquisition times, 1000 Raman spectra of mono-crystalline silicon were acquired per setting. Curve fitting was employed to obtain the peak positions of all the spectra, from which the standard deviations of the measured peak positions were obtained versus the total accumulated amount of laser light incident on the sample during one measurement.
It has been found that the noise in the obtained peak position decreases as 1/sqrt(n) over more than three decades of the incident amount of laser light. At very low light conditions, the noise decreases as 1/n. By comparing the experimental results obtained to recent theoretical work, we show that the acquisition is limited by photon shot noise over most of the range and is limited by electronic detector noise at very low light conditions only. Pixelation errors do not play a role.
It is concluded that the low electronic noise of typical Raman spectroscope detectors is overkill for the investigation of mechanical stress and temperature in silicon and other materials with comparable peaks, as it has absolutely no influence on the noise level of such an experiment. Maximum Raman signal intensity on the detector and high quantum efficiency detection are more important. Copyright © 2013 John Wiley & Sons, Ltd.