Scanning cathodoluminescence microscopy: A unique approach to atomic-scale characterization of heterointerfaces and imaging of semiconductor inhomogeneities

Abstract

Luminescence experiments provide a powerful and nondestructive approach to the ex situ investigation of semiconductor heterointerfaces which might be buried up to several μm below the surface in a given complex sample structure. Combined with the ability of taking images simply by scanning the exciting focused electron beam across the area under investigation, lateral fluctuations of electronic properties like the variation of the fundamental band gap Eg(x,y) can be directly visualized by scanning cathodoluminescence (CL). The novel experimental approach, cathodoluminescence wavelength imaging (CLWI), which involves recording of a complete CL spectrum at every scanning position (x,y), yields direct 3D images of the atomic-scale morphology of quantum wells (QWs) as sensed by the QW exciton: similar to the tip of a scanning tunneling microscope, the exciton samples the local fluctuations of QW thickness Lz and transforms this structural information Lz(x,y) into a spectral one, the lateral variation of band gap Eg(x,y) and thus the CL emission wavelength λ(x,y). Topological maps of QW interfaces can thus be recorded at various positions and at various magnifications. The interface roughness can be investigated statistically at lateral resolution starting with the diameter of the QW exciton up to the mm regime. The same experimental principle for recording λ(x,y) and Eg(x,y) maps is successfully applied for the analysis of patterned structures. In the nonlattice-matched system GaAs on Si, the lateral strain variation causes Eg(x,y) fluctuations and can thus be directly imaged by CLWI. Metalorganic chemical vapor deposition grown GaAs layers on micropatterned Si(001) substrates show strongly inhomogeneous doping with Si impurities. By means of CLWI the strong increase of this Si incorporation in the vicinity of free {111} surfaces is measured and Si concentration maps are recorded across the complete sample pattern.