Strictly Research: AlAs/GaAs(100) lattice matching with Phosphorus

Faebian Bastiman

The AlAs/GaAs(100) system is widely employed in III-V device architecture. The system benefits from a large band gap difference combined with a very small lattice mismatch. The small mismatch and the induced compressive stain become significant when growing thick layers. The critical dislocation thickness, tc, has been shown empirically to be approximately 1.1 µm.

A number of elements can introduce tensile strain into this system and lattice match the overlayer to the substrate. Nitrogen, N, possess a band gap bowing co-efficient of up to -125meV/%, which means the strain balancing comes with a reduction in the bandgap. Boron, B, introduces tensile strain with a much smaller but still negative -1meV/%. Phorphorus, P, introduces tensile strain whilst increasing the band gap by +20meV/%. The AlAs1-xPx/GaAs(100) therefore is an excellent candidate for lattice matching the alloy and increasing the carrier confinement compared to pure AlAs/GaAs(100). Furthermore, AlP and AlAs are readily miscible at the optimum GaAs-system growth temperature which greatly improves the alloy quality and simplifies growth.

The AlAs96P4 alloy can be seen to be nearly lattice matched at room temperature, though its no longer the case at the growth temperature of ~600°C.  The respective thermal expansion coefficients of GaAs and AlAs means that GaAs expands at a faster rate (Å/°C). A phenomenon that means the two binary systems reach a natural lattice matched condition around ~880 °C. A maximum thickness before dislocation of >4 µm is observed for in the AlAs98P2/GaAs(100). AlAs98P2 and GaAs are therefore lattice matched at 600 °C.

The lattice matched alloy can be grown in two ways. Either a lattice match ternary: AlAsP or a pair of strain compensating binaries: AlAs/AlP. In the ternary the group V flux ratios set the composition. For AlAs/AlP the composition is set with the individual layer thicknesses. The ternary composition can be very simply taken from XRD peak splitting. AlP has its own tc of ~3.2ML and hence a more complicated super-lattice (SL) structure is necessary.  The more convoluted AlAs/AlP SL requires additional analysis but can provide information on layer thicknesses and hence provides an additional growth rate check.  Contrary to popular belief the ternary forms in a straightforward manner and only a small P BEP is needed, utilising a Vecco two zone (red-white) P cracker with the bulk at 50 °C (c.f. typical 90 °C for AlP) provides ample flux and is readily repeatable to with 0.1%. The lower P flux greatly reduces P contamination and unwanted P alloying with GaAs layers. AlAs/AlP theoretically provides higher [P] composition precision due to the sub-ML accuracy of MBE, however shutter transients are difficult to predict especially for high growth rate and low cycle times. Furthermore the necessity to purge background As and P during the growth of AlP and AlAs respectively increases the overall growth time. For most applications the simpler to implement but lower precision of the AlAs98P2 ternary is perfectly suitable, the higher complexity, higher precision binary offers a greater degree of design freedom with respect to layer structure and refractive index interface design.