A semiempirical kinetic model is presented which maps out the thermal budget for processing of strained layer devices through epitaxial growth and postgrowth anneals. Misfit strain relaxation in Si1−xGex/Si heterostructures by the injection and propagation of a/2 〈110〉 60°-type misfit dislocations has been studied for a range of geometries and dimensions. Strained layer superlattices, Si1−xGex alloy layers, 0<x<0.3, and alloy layers with unstrained Si capping layers of thickness 0 to 400 nm were grown by molecular-beam epitaxy on (100) Si substrates and subjected to post-growth thermal cycles. Velocity and nucleation rate data from Nomarski interference microscopy of defect-etched surfaces were correlated with electron beam induced current microscopy transmission electron microscopy and x-ray diffraction results to define Arrhenius relationships for misfit dislocation injection rates and propagation velocities. A unified kinetic model for misfit strain relaxation that incorporates both nucleation and propagation is then developed, which is applicable for all heterostructures and thermal cycles in the low dislocation density regime <103 mm−1. Nonuniform strain distribution in graded device heterostructures is considered by defining the effective stress acting on misfit dislocations for an arbitrary geometry. The effective stress was varied from 0 to 750 MPa in Si1−xGex/Si heterostructures by varying both layer dimensions and Ge concentration. Misfit dislocation nucleation rates varied from 10−3 to 103 mm−2 s−1 and misfit extension velocities of 25 nm s−1 to 2 mm s−1 were obtained over the temperature range 450–1000 °C for anneals of duration 5–2000 s. Activation energies, stress exponents, and prefactors in the Arrhenius relations were found to be independent of Ge concentration, effective stress, and heterostructure geometry allowing a comprehensive model to be developed. The onset of strain relaxation during epitaxial growth cycles (the ‘‘apparent critical thickness’’ or metastability limit) characteristic of molecular-beam epitaxy and chemical vapor deposition was measured and correlated with the simulation of misfit dislocation injection and propagation in typical growth sequences. The kinetic model is also used to define the maximum time-temperature envelope, or thermal budget (t, T), for the misfit dislocation-free processing of Si1−xGex/Si heterostructures subjected to post-growth thermal treatme

No abstract

We have directly imaged the evolution of surface cusps during strained-layer epitaxy. The cusps arise naturally as a result of gradients in the surface chemical potential. High stress concentrations at the cusp tip have important implications for strain relaxation in the film via dislocation nucleation.

The reduction of the dislocation density in relaxed SiGe/Si heterostructures using a low-temperature Si͑LT-Si͒ buffer has been investigated. We have shown that a 0.1 m LT-Si buffer reduces the threading dislocation density in mismatched Si 0.85 Ge 0.15 /Si epitaxial layers as low as ϳ10 4 cm Ϫ2. Samples were grown by both gas-source molecular beam epitaxy and ultrahigh vacuum chemical vapor deposition.

Jesson et al. Reply: The new structure (RS3) proposed by us in our Letter [1] is not identical to a structure RS2 (sometimes referred to as RH2) considered previously [2][3][4][5]. The phases have identical space groups (R3m) but different site occupancies as clearly indicated in our Letter [1]. LeGoues et al. [6] suggest that they did not discuss the difference between the phases because it is "small." However, we note that a 10% difference in site occupancy is readily measurable experimentally [1], so that the two phases are physically distinct. It is important to point out that LeGoues and co-workers have made no previous mention of any possible structure other than RS2 despite four papers [2-5] detailing their observations of the ordered phase. We note, in particular, that the theoretical work of Kelires and Tersoff [7] was applied by LeGoues et al. [2] to support their prior experimental determination of RSI [3] and not RS3. Furthermore, no experiments were undertaken to investigate the structural differences between RS2 and RS3, even though this is straightforward by x-ray diffraction as shown in our Letter [1], A major point of our Letter [l] is that ordering in Si x Geix alloys is linked to evolving surface morphology. In particular, we demonstrated that the average lateral size of the ordered domains is around 40 nm, of exactly the same dimensions as the islands seen to occur during growth. It is the locally vicinal surfaces presented by the islands which enable bilayer step flow to be established at unexpectedly low temperatures [8]. These regions, once established, are rather stable so that the ordered domains propagate up into the epilayer during growth. Thus, the surface morphology dictates the size of the domains.This explains the large domains observed in the Comment if it is appreciated that the surface morphology is unlikely to be flat as assumed by LeGoues et al. [6]. Rather, the observation of 0.5 fim domains implies the existence of surface mounds or "islands" of lateral dimensions 0.5 fim and height 90 A (assuming that bilayer step flow is stable for angles of one degree as shown in our Letter [1]). Unfortunately, the cross sectional transmission electron microscopy data presented by LeGoues et al. [6] lacks sufficient sensitivity to detect such features. The most likely explanation for the formation of a moundlike morphology lies in the inherent instability of the ideally flat (001) surface. Mounds of similar dimensions have recently been shown to occur for GaAs epitaxy [9], and a possible explanation was given in terms of a Schwoebel barrier. We believe, therefore, that ordering is again driven by surface morphology as in the case of coherent islanding as reported in our Letter [1], the only difference being the scale and physical origin of the island templates involved.Our ordering model, based on our earlier work on steps [10,11], is, therefore, fully consistent with all experimental data, including observations of interfacial ordering in ultrathin (Si w Ge rt ) p superlattices [12]. F...