Alloy effects on the Raman spectra of Si1 xGexand calibration protocols for alloy compositions based on polarization measurements

The Raman spectra of bulk and epitaxial Si 1− x Ge x alloys (0 < x < 1) were studied with the aim of establishing a metrology for the determination of alloy compositions from the first-order Si–Si, Ge–Ge and Si–Ge optical phonon Raman modes. However, the lineshapes and intensities of these modes are distorted by various subsidiary features including (i) a broad quasi-amorphous feature on the low-energy side of the Ge–Ge mode and (ii) peaks in the range of 340–500 cm −1 , very close to the main modes and believed to arise from various types of Si–Ge clusters. Standard curve-fitting procedures cannot cope with such background features and fail to give reliable values for the intensities of the main lines. We demonstrate that the background features can be minimized by protocols based on polarization measurements. One of these methods that reveals less distorted Ge–Ge and Si–Si modes involves subtraction of the first-order-forbidden spectrum from the allowed spectrum in a parallel polarization configuration. The other involves measurements in a cross-polarization geometry in which the background is suppressed. These protocols allow better estimates of the integrated phonon intensities for a wide range of alloy compositions by usual curve-fitting, and improvements in the ratios of the I Ge–Ge / I Si–Si integrated intensities by factors as large as 8 were obtained for Si-rich samples. The simpler procedure for calibration based on phonon frequency positions, sufficient for bulk alloys, may be incorrect for heterostructures due to the additional strain dependence of the frequencies superimposed on the alloy compositional dependence. Therefore, a protocol based on the integrated phonon intensities is applicable and performs better in both strained-layer devices and bulk alloys.