Growth Studies of Doped SiGeSn/Strained Ge(Sn) Heterostructures

Large-scale integration of Si optoelectronics is still hampered due to the lack of efficient group IV light emitting materials. One promising way to overcome this obstacle is to introduce tensile strain in Ge in order to reduce the energy difference between the Γ- and the L-valley in the conduction band. A fully Si compatible technology for straining Ge becomes feasible by the use of large lattice constant strain relaxed (Si)GeSn buffer layers. Additionally, it has been theoretically shown that SiGeSn ternary alloys are ideal candidates as cladding layers in SiGeSn/sGe(Sn) heterostructures for photonic devices. In this contribution we present studies on (i) epitaxial strain engineering of SiGeSn alloys (to strain Ge) and (ii) the in-situ doping of such ternary compounds.

The Sn-based heterostructures shown here have been grown by Reduced-Pressure Chemical Vapor Deposition (RP-CVD) using Si₂H₆, Ge₂H₆ and SnCl₄. Low growth temperatures (350 – 400°C) have been employed to avoid Sn precipitation and obtain high substitutional Sn concentrations in 200-300 nm thick, partially strain relaxed (Si)GeSn buffer layers (SRB) for strained Ge (sGe). SiGeSn:P and SiGeSn:B layers grown in the same temperature range using B₂H₆ (100 ppm in H₂) and undiluted PH₃ precursors have been investigated by means of electrochemical C-V (ECV) profiling, sheet resistance measurements and ToF-SIMS.

Fig. 1a shows the reciprocal space map (RSM) of 70 nm strained Ge grown on a 250 nm Ge_0.89Sn_0.11 SRB. This method allows an accurate determination of the in-plane and out-of-plane lattice constant of the GeSn SRBs and the overgrown sGe. The GeSn peak is obtained above the cubic lattice line indicating residual compressive strain. The degree of relaxation is about 73 %. Since the sGe peak lies on the pseudomorphic line of the SRB this top layer is tensile strained (1.4 %) coherently to the GeSn. The strain levels for several strained Ge layers grown on GeSn SRBs are summarized in Fig 1b. The strain level in overgrown sGe(Sn) layers can be adjusted by the choice of the Sn content and degree of relaxation of the buffer. In-situ doping is the method of choice to dope SiGeSn ternary layers due to the reduced thermal stability of Sn-based materials. The resistivity of SiGeSn:B and SiGeSn:P layers is shown in Fig. 1c as a function of the dopant partial pressure. The resistivity decreases for SiGeSn:B and increases for SiGeSn:P layers as the dopant partial pressure increases. In the latter case, a reduced layer quality has been found for PH₃ partial pressures above 10 Pa indicated by an increase of the RBS/Channeling minimum yield, c_min (inset). The growth rates for doped ternaries do not show any dopant partial pressure dependence at least at a constant growth temperature of 425°C (inset). Interestingly, the growth rate for SiGeSn:P is higher than for SiGeSn:B layers which is comparable to that of undoped SiGeSn (44.5 nm/min). More B atoms are incorporated on substitutional lattice sites for high B₂H₆ partial pressures, leading to increased dopant activation up to 2x10¹⁹ cm^-3 without dopant segregation. Within the investigated process window, the ECV profiling of a p-n junction with the highest achieved doping concentrations of 2x10¹⁹ cm^-3 (p-type) and 8x10¹⁹ cm^-3 (n-type) is shown in Fig. 1d. The very low Χ_min of about 5 %, comparable to that of the Ge-VS underneath, indicates high single crystalline quality with high dopant substitutionality without Sn or P segregation.

Fig. 1 (a) XRD-RSM of a highly (1.4 %) tensile strained Ge layer grown on a partially relaxed GeSn buffer layer. (b) 3d plot of several tensile strained Ge layers on differently alloyed GeSn layers. (c) Resistivity as a function of the dopant partial pressure for SiGeSn:B and SiGeSn:P layers. The minimum channeling yields and growth rates are shown in the insets. (d) ECV profile of a SiGeSn p-n junction. The RBS/Channeling spectra are shown in the inset.