Probing the behavior of vibrational modes in semiconductors is central to understand their basic structural, optical, and phononic properties. This is particularly true for the emerging GeSn and SiGeSn. This family of group-IV semiconductors is of growing interest for optoelectronic applications, notably due to the possibility of tuning the bandgap to cover a broad range from the short-wave infrared to the mid-infrared. An indirect to direct transition in GeSn alloys is expected for Sn incorporation higher than 9%, which can lead to the development of high efficiency light emission and detection devices. The growth of such alloys has however proven challenging due to the low solubility of Sn in Ge (≈1%). In addition, any residual strain means more tin incorporation is required to obtain a direct-bandgap [1]. The alloying process enables tuning the electronic and optoelectronic properties, but it also comes with enhanced lattice disorder thus affecting the lattice vibrational modes. Raman spectroscopy is commonly used to assess the role of composition y and strain ε in shaping the lattice vibrations. Curiously enough, previous works on GeSn mainly focused on analyzing the behavior of Ge-Ge mode. This is due to the broad use of 488 nm [2] or 532 nm [3], [4] excitation lines, which yield a very weak signal of the Ge-Sn mode. Alternatively, the use of a 633 nm excitation enables a clear detection of all Raman modes in GeSn layers independently of their composition. This is plausibly attributed to the fact that this wavelength might be close to resonance with the alloy’s E1 gap [5]. There are very few reports on the identification of the Ge-Sn mode [6], but quantitative analyses of the effects of composition and strain on this mode remain conspicuously missing in literature. With this perspective, this work presents a detailed study of Raman vibrational modes in Sn-rich (7-18 at.%) GeSn semiconductors. Samples were grown in a low-pressure chemical vapor deposition (LP-CVD) reactor using monogermane and tin-tetrachloride precursors. The samples consist of one or two GeSn layer(s) on a Ge virtual substrate (VS) on a Si wafer. Using X-ray diffraction (XRD), reciprocal space mappings were performed on all samples to retrieve the composition and the strain. Raman measurements have been carried out on an InVia Microscope from Renishaw with a 633 nm laser. While Voigt or Lorentzian functions are commonly used for fitting peaks, they cannot reproduce the asymmetric broadening typical to GeSn Raman peaks (Fig. 1). This asymmetry is due to alloying as substitutional Sn atoms break the translational symmetry and lead to a relaxation of the momentum selection rule [7]. To perform the fits, we are therefore using exponentially modified gaussian (EMG) functions, which can better reproduce the line shape of the Raman modes. A typical Raman spectrum appears in Fig. 1. Two-dimensional linear regressions are then performed to decouple the influence of strain and composition on the Raman shift of both modes. The resulting functions are illustrated in Fig. 2 and the coefficients appear in table 1. The planar fits accurately represent the distributions, as confirmed by the relatively small error on both a and b parameters (<10%) and the coefficients of determination near 0.99. Furthermore, the calculated Raman shift of the Ge-Ge mode in the limit of a pure and completely relaxed Ge layer is equal to the value obtained for bulk germanium, and the a and b parameters are comparable to those found in earlier studies for the Ge-Ge mode. Based on these detailed Raman studies, an exhaustive discussion of the influence of lattice strain and Sn content on GeSn vibrational modes will be presented. Acknowledgements The authors thank J. Bouchard for the technical support, and NSERC Canada (Discovery, SPG, and CRD Grants), Canada Research Chair, Canada Foundation for Innovation, Mitacs, FRQNT, Institut de l’énergie Trottier and MRIF Québec for support.
Read full abstract