Ge / SiGe Multi Quantum Well Fabrication by Using Reduced Pressure Chemical Vapor Deposition

In this paper we have deposited structures comprising a stack of 10 periods made of 15 nmthick Ge multi quantum well (MQW) enclosed in 15 nm-thick Si0.2Ge0.8 barrier have been deposited on SiGe virtual substrates (VS) featuring different Ge contents in the 85% 100% Ge range to investigate the influence of heteroepitaxial strain on the Si0.2Ge0.8 and Ge growth. With increasing Ge concentration of the VS, growth rate of the Si0.2Ge0.8 in the MQW increases. Si incorporation into the Si0.2Ge0.8 layer becomes also slightly higher. However, almost no influence of the growth rate is observed for Ge growth in the MQW. We argue that the increased tensile strain promotes the Si reaction at the surface. In the case of the Si0.2Ge0.8 growth on Ge, we observe a smeared interface due to the Ge segregation during the growth. Furthermore, we observe that this interface width increases with increasing Ge concentration of VS. We attribute this observation to the increased segregation of Ge driven by the increased strain energy accumulated in the in the Si0.2Ge0.8 layers. We also observed that the MQW layer “filters-out” threading dislocations formed in the VS.

Introducing high quality SiGe / Ge superlattice (SL) structure has a potential to realize high efficient Ge multi quantum well (MQW) cascade laser [23][24][25][26][27]. In order to fabricate the SiGe / Ge MQW cascade laser, high quality and precisely controlled SiGe / Ge SL layer fabrication process is required. To realize the high crystal quality and precisely controlled SiGe / Ge MQW structures, influence of strain on the Ge and SiGe growth has to be investigated. In this study, we fabricated Si0.2Ge0.8 / Ge MQW structures on Ge and SiGe virtual substrate (VS) with different Ge content and discuss the influence of strain on Si0.2Ge0.8 and Ge growth. In this paper, influence of strain on the Si0.2Ge0.8 and Ge growth rate as well as steepness of the interface are also discussed in addition to data already published in [28].

Experimental
Epitaxial Si0.2Ge0.8 / Ge MQW fabrication is carried out by using a reduced pressure (RP) chemical vapor deposition (CVD) system. Si(100) substrates are used. H2 is used as a carrier gas and SiH4 and GeH4 are used as growth precursor gases. For a Ge VS preparation, ~2 µm thick Ge is deposited by following procedure. After standard radio corporation of America (RCA) clean, the Si(100) wafer is loaded into the RPCVD reactor and baked at 1000 o C in RP H2 to remove native SiO2. After that the wafer is cooled down to 350 o C. During the cooling carrier gas is changed from H2 to N2 at 600 o C to form a hydrogen-free Si surface. After temperature stabilization, a Ge seed layer is deposited using N2-GeH4 gas mixture. Afterwards the wafer is heated up to 550 o C in H2 environment and the deposition is continued by growing the main Ge layer part using H2-GeH4. To improve crystal quality of the 2 µm thick Ge, cyclic annealing at 800 o C is performed by interrupting the Ge deposition process at 800 o C for several times [29][30][31]. To fabricate SiGe VS with 95%, 90% and 85% Ge content, ~500 to ~600 nm thick SiGe layer deposition using H2-SiH4-GeH4 gas system followed by postannealing at this reverse graded buffer approach, high crystal quality of SiGe VS deposition is possible comparable to that of direct SiGe deposition on Si substrate [32]. After the VS fabrication, the wafers are unloaded from the RPCVD reactor and cleaned by HF dip and loaded into the RPCVD reactor again. Then the wafer is baked at 800 o C to remove native oxide. Afterwards 10 cycles of Ge and Si0.2Ge0.8 layers are deposited at 500 o C using H2-GeH4 and H2-SiH4-GeH4 system, respectively. Target thickness of the Ge and the SiGe layers are 15 nm. The 10 cycles of the Si0.2Ge0.8 / Ge MQW deposition is performed on the VS with 85%, 90%, 95% and 100% Ge content with identical process condition to check influence of the strain on the Si0.2Ge0.8 and Ge growth.
X-ray diffraction (XRD) is used for periodicity and degree of relaxation measurements.
Cross section transmission electron microscope (TEM) is used for crystallinity and profile analysis. Energy dispersive X-ray spectrometry (EDX) and X-ray photoelectron spectroscopy (XPS) with monochromated Al Kα (1486.7 eV) are used for Si composition analysis of the deposited Ge / SiGe SL. Threading dislocation density (TDD) is measured by combination of Secco defect etching and angle view scanning electron microscope (SEM) measurement.

Results and Discussion
In Fig. 1a and 1b, Si0.2Ge0.8 and Ge thicknesses in the Si0.2Ge0.8 / Ge MQW deposited on VSs with 85%, 90%, 95% and 100% Ge content are shown. In the case of the Si0.2Ge0.8 layer ( Fig. 1a), the thickness on a Si0.15Ge0.85 VS is ~17 nm. The Si0.2Ge0.8 layer thickness increases with increasing Ge concentration of VS from 85% to 100%. This result indicate that growth rate of the Si0.2Ge0.8 layer is increased by increasing tensile strain. On the other hand, in the case of Ge layer in the MQW (Fig. 1b), there is almost no influence of Ge concentration of the VS on the Ge layer thickness. Based on these results, it seems that enhancement of surface reaction of SiH4 occurs on higher tensile strained SiGe surface.
In Fig. 2, Si concentration in the SiGe layer in the MQW measured by TEM-EDX and intensity ratio of Si 2p and Ge 3d measured by XPS is shown. Because the escape depth of photoelectron by XPS measurement is below 10 nm and top SiGe layer thickness is 17 nm to 18 nm (as shown in Fig. 1a), only photoelectrons from top SiGe layer are visible and the photoelectron intensity from the Ge layer underneath is negligible. By the TEM-EDX analysis, Si concentration in the SiGe layer in MQW on Si0.15Ge0.85 VS is ~18%, and tends to increase slightly by increasing the Ge concentration in the VS. The increase is very small, but the increase of Si concentration is also supported by XPS analysis. Therefore, it seems that the increase of the Si incorporation with increasing Ge concentration in the VS (i.e. increasing tensile strain in the growing Si0.2Ge0.8 layer) is reliable. Based on the results of Fig. 1a, b and but not for the Ge layer (Fig. 1b) could be that there is an increased reaction of SiH4 on higher tensile strained Si0.2Ge0.8 due to higher surface energy. At 500 o C, SiH4 reaction is in the surface reaction limited regime, but GeH4 reaction is in the mass transport limited regime. Therefore, the reaction of the SiH4 is increased by higher strain, but the reaction of the GeH4 is less sensitive for surface energy of the Ge surface because the reaction is mainly limited by the mass transport.
In Fig. 3, cross section STEM ( Fig. 3a and 3b) and EDX ( Fig. 3c and 3d) of 10 cycles of Si0.2Ge0.8 / Ge MQW are shown. Fig. 3a and 3c depict samples with MQW grown on Ge VS and Fig 3b and 3d with MQW grown on Si0.15Ge0.85 VS. For both Fig. 3a and 3b, no defects are observed in STEM images, indicating good crystal quality of the MQW. In both EDX images ( Fig. 3c and 3d), steep profiles are observed at the interface of Ge on Si0.2Ge0.8 layer, however the interface of Si0.2Ge0.8 on Ge is smeared. The smeared interface could be caused by Ge segregation into Si0.2Ge0.8 [33], to the lower surface energy density of Ge as compared to Si.
Next, influence of strain on the interface steepness is discussed. Because the resolution of the TEM images is influenced by the TEM lamella thickness, relation between the interface thickness measured by TEM and the TEM sample lamella thickness are summarized in Fig. 4.
The interface thickness is determined as the thickness of transition area of Ge concentration between Ge and Si0.2Ge0.8. As shown in Fig 4, with increasing the TEM sample lamella thickness, slight increase of the interface thickness is observed for both interfaces, Ge on  (Fig. 5b), position of the Ge VS peak indicates 0.18% of tensile strain again. Position of Si0.15Ge0.85 VS peak shows 0.29% of tensile strain, due to partial relaxation from the Ge VS. Subpeaks of SL from the Si0.2Ge0.8 / Ge MQW are also observed. Estimated periodicity of the SL is 30.9 nm. The Qx position of the SL peaks is same as Si0.15Ge0.85 VS peak, indicating the Si0.2Ge0.8/Ge MQW layer is pseudomorphically grown on the Si0.15Ge0.85 VS.
In Fig.6, improvement of TDD by the Si0.2Ge0.8 / Ge MQW deposition on Ge or SiGe VS with various Ge content is summarized. In the case of the Si0.2Ge0.8 / Ge MQW deposition on Si0.15Ge0.85 VS, the TDD is the same level as that of the Si0.15Ge0.85 VS. With increasing Ge concentration of the SiGe VS, improvement of the TDD is observed by the 10 cycles of Si0.2Ge0.8 / Ge MQW deposition. Approximately, a 50% reduction of the TDD is observed for the Ge VS sample. It is not possible to achieve a TDD reduction of the same extend by adding the same thickness (300 nm) of simple Ge layer on the ~2 µm thick Ge VS without annealing steps [29,31,34]. We attribute the the decrease of TDD by the Si0.2Ge0.8 / Ge MQW deposition on the Ge VS might be the same mechanism as reverse graded buffer. The TD network seems to go downwards by tensile strained SiGe growth on Ge [32]. However, further investigations are required to clarify the detailed mechanism behind this observed effect.

Summary and Conclusion
Ten cycles of 15 nm thick Si0.2Ge0.8/15 nm thick Ge MQW are fabricated on SiGe virtual substrate with 85% -100% Ge content to discuss the influence of strain on SiGe and Ge growth.
In the case of Si0.2Ge0.8 growth on Ge MQW layer, smeared interface is observed due to segregation of Ge atom. On the Ge VS, slightly higher Ge segregation into the Si0.2Ge0.8 is observed compared to that on SiGe VS with 85% Ge content due to higher tensile strain in the Si0.2Ge0.8 layer. Small increase of growth rate and Si concentration of the Si0.2Ge0.8 layer is observed by increasing Ge content of the SiGe VS from 85% to 100%. With increasing Ge content in the VS, higher reduction of the TDD is observed by depositing the 10 cycles of Si0.2G0.8/Ge MQW.