No abstract

The atomic structure of In 0.81 Ga 0.19 As/InP alloy layers was examined using in situ scanning tunneling microscopy. The ͑2ϫ3͒ reconstruction observed during growth by reflection high-energy electron diffraction represents a combination of surface structures, including a ␤2͑2ϫ4͒ commonly observed on GaAs͑001͒ and InAs͑001͒ surfaces, and a disordered ͑4ϫ3͒ that is unique to alloy systems. The proposed ͑4ϫ3͒ structure is comprised of both anion and cation dimers. Empty and filled states images show that the features reverse contrast with sample bias, in agreement with the model.

The atomic structure of compound semiconductor alloy surfaces is important for heteroepitaxial growth, as it has an impact on the subsequent microstructure of the film. For example, it has been suggested that random fluctuations in composition may initiate lateral composition that propagates through the remainder of the film. Also, interfacial abruptness is greatly influenced by the step edge roughness. We examined the morphology and surface reconstruction of In,Ga,,As alloy layers during growth and after annealing. Films of different compositions were grown by molecular beam eptiaxy on GaAs and InP (001) substrates to thicknesses less than the critical thickness for 3D islanding or misfit dislocation formation, and examined using in-situ Scanning Tunneling Microscopy (STM). Figure 1 shows the filled states S T M image of a nominally lattice matched 1~,,Ga,,~,As/bP surface that was deposited at T=480"C and BEP=lZxlO" torr. Even though the surface reconstruction is highly disordered, nearly 25% of the surface is coverage with a (4x3) reconstruction and 15% is covered with a ~(3x4). Models developed based on the STM data suggest that these reconstructions are terminated by both cation and anion dimers. The surface reconstruction during the growth of alloys with compositions away from the lattice parameter of InP(OO1) was (2x3) according to reflection high energy electron diffraction. However, STM shows that the surface is covered with a number of different reconstruction domains. Figure 2 shows a filled state STM image of In,,G~,,As/GaAs grown at T=486"C and BEP=16x106 torr. Although most of the surface is covered by disordered c(3x4), 34% of the surface consists of short segments of a2(2x4) reconstructed regions within the terrace. The secondary reconstruction for b,81Ga,l&/InP, which has the same amount of lattice mismatch, is 82(2x4). Figure 3 shows a filled state STM image of a 25 ML thick I~,,,G~,,,As/InP surface. It is interesting to note that the 82(2x4) rests upon the underlying (4x3) reconstruction, and that the coverage of the 82(2x4) decreases as the surface is annealed. For example, the 82(2x4) coverage of an as-grown Iq,,,Gq,,,As/InP layer is nearly 50%, and drops to 34% after a 25 minute anneal at the growth conditions. This behavior is similar to the dependence of the surface In concentration as a function of annealing time, and suggests that the 82(2x4) regions on this surface are related to In surface segregation. 0-7803-782GU03/$17.0~W3 IEEE 43