Silicon-compatible germanium (Ge) coherent light sources have received considerable attention recently resulting in both lasing and photoluminescence (PL) in strained Ge, for which sufficient tensile strain reduces the relatively small indirect to direct bandgap (BG) difference to zero, thus making Ge a direct gap material, albeit at a relatively low energy. Previously we explained the cause of the intense, low temperature PL observed for many MBE-grown Si1-xGex epitaxial layers as being due to nm-thick Ge nanocrystals (NCs) embedded in the SiGe epilayers, as in the plan-view TEM image and schematic insets of Fig. 1. We showed that the Ge NC peak PL energy depends on several factors, namely: (a) Ge bandgap variation with vertical strain; (b) effect of the Ge-fraction in the host SiGe epilayer on the vertical lattice constant; (c) effect of strain on the SiGe bandgap energy; (d) confinement shifts in the Ge NCs; and (e) confinement shifts in the SiGe epilayers. A model incorporating these effects was developed that effectively explained the PL energy dependence for the SiGe samples comprising over 60 separate epilayer configurations (Fig. 1). The experimental data is presented in Fig. 1, which also contains on the left the PL spectra for three Ge NC samples, including the one to be discussed later in this paper. From the analysis of the data we were able to use confinement for the vertical size of the Ge NCs, which increased linearly with vertical tensile strain.Many of the Ge NC samples exhibited a very bright luminescence, with as high as 3% internal quantum efficiency. For one of the samples the PL was imaged in video format for broad band radiation from 1200 to 1900 nm. One such image is displayed in Fig. 2 for a sample at low temperature comprised of a SiGe/Si multiple quantum well (MQW) structure, with 40 epilayers of 7.6 nm thick Si0.75Ge0.25 separated by 20 nm of Si on a Si(001) substrate. In this case the excitation (100 mW at 514.5 nm in a 1 mm diameter beam) at normal incidence was at the center of the 3 mm diameter sample mask aperture. The exciting radiation was excluded from the camera aperture with a long-wave pass filter, which absorbed light with wavelengths below 1200 nm. On the right of the sample mask in the image, the sample extends with its (110) cleaved edge sticking out at a small angle (~20o) with respect to the edge of the sample mask. The PL is apparently brightest at the point of excitation, but is also quite intense at the edge of the sample, more so at the right hand edge. The trace below the image is the PL intensity profile along the thin yellow line, which indicates an intensity at the edge of the sample comparable to that at the excitation location. In this experiment the incident excitation light was polarized roughly parallel to the wafer edge.In the sample imaged, the MQW system is thick enough and has a high enough effective refractive index to act as a thin film waveguide in the infrared, which could account for much of the edge brightening in Fig. 2. In addition to the waveguiding in the thin film structure, we expect some bulk waveguiding in the silicon substrate, as the back surface is likely smooth enough and the silicon material is transparent in this wavelength range. The luminescence along the sample edge was brightest perpendicular to the excitation polarization, like fluorescence anisotropy. A broadband PL image is shown in Fig. 3 for the excitation focused to a 0.1 mm spot near the edge of the sample, where there a second bright PL spot. Two PL intensity profiles are shown, one through the excitation and edge spots and one along the wafer edge. We see that the edge brightening has also occurred, although the overall PL intensity is not so greatly increased on focusing, possible due to NC saturation effects possible when more than one exciton is captured by a Ge NC. However, we see from Fig. 3 that the edge emission is in a spot not greatly larger than the extent of the emission at the excitation spot. This observation suggests that the emission is quite collimated in the MQW waveguide, much more so than would be expected from the simple direction effects based on excitation-emission polarization that are normally seen for bulk luminescence. This lack of divergence between the excitation and the edge might be indicative of an optical amplification process occurring in the MQW waveguide. Figure 1
Read full abstract