Heterogeneous Integration of Thin-film Germanium on Silicon and Steel

Abstract

Integration of germanium (Ge) on low cost substrates such as silicon and steel open several applications such as low cost III-V photovoltaics, beyond silicon CMOS and optoelectronics applications. However, growth of germanium on silicon and steel is non-trivial owing to lattice mismatch of 4% and 50 % respectively. So far successful growth of Ge on Si has been shown by using complex methods such as molecular beam epitaxy (MBE) and Ultra High Vacuum Chemical vapour deposition (UHV-CVD). Despite their advantages the methods are not suitable for scale-up owing to their high-cost and low throughput. Alternatives such as liquid phase epitaxy from metal catalyst are industry scalable but lead to defective Ge films due to metal incorporation. In this work, we have developed a novel process for growth of epitaxial germanium films on silicon using liquid phase crystallization. The films are grown by melting and re-crystallization from pure “liquid phase” of germanium instead of a metal solution, thus eliminating problem of metal contamination. The liquid phase crystallization method is relatively simple and low-cost, and the Ge films so obtained are comparable in quality to MBE and CVD methods. The LPC method is wafer-scale and suitable for batch processing, making it commercially viable. The LPC Ge films are characterized for material quality using SEM, XRD, Raman measurements and for device quality using lifetime and Hall measurements. The films are found to be epitaxial with excellent electronic properties. The LPC Ge was used to demonstrate two novel application: a) low cost III-V photovoltaics, and b) on-chip Ge devices on Si. One way to significantly reduce cost of GaAs photovoltaics is to grow high-quality GaAs on low-cost Si wafers. However, direct integration of GaAs on Si leads to highly defective films due to their lattice mismatch. Unlike Si, Ge is lattice matched to GaAs. So better quality GaAs-on-silicon films can be grown by integrating an epitaxial Ge buffer, grown using LPC, between GaAs and Si. The GaAs films were grown on the Ge on Si substrates using MOCVD, and characterized using SEM, XRD and Raman measurements. To examine the device quality, room and low temperature photoluminescence (PL) and time resolved PL measurements were done. Results show that the GaAs films grow epitaxially on Ge-on-Si with room for further optimization. Ge photo-detectors find use in next-generation silicon photonics and as IR sensors. LPC Ge is a candidate for on-chip Ge-based IR devices. As proof of concept, MSM were fabricated on LPC-grown Ge-on-Si films using aluminum metal as contacts. The detectors were found to be responsive in IR range from 1100 nm to 1700 nm with moderate responsivity. Even more challenging than Si, is integration of Ge on steel. Steel is robust and low-cost substrate and stable at high temperatures. These qualities make Ge-on-steel interesting for low-cost, large-area, and flexible solar cells. However direct growth of any semiconductor on steel leads to defective films, primarily due to diffusion of iron from steel into semiconductor; deteriorating the electronic properties of the semiconductor. Introducing a “diffusion barrier” between steel and the semiconductor layer may solve this problem. Titanium nitride (TiN) was found to be a good barrier-layer because it is stable at high temperatures, does not contaminate semiconducting over layers and has very low iron diffusivity. TiN on steel as iron diffusion barrier was characterized using Secondary Ion Mass Spectrometry (SIMS). It was found that the diffusion estimation is non-trivial due to contribution from both bulk and grain boundaries which was for the first time explained and observed carefully in this work. The study reveals the importance of effective diffusion estimation for designing barrier layer for devices on steel. Thereafter, germanium was, for the first time, successfully crystallized on TiN on steel using laser annealing of amorphous Ge films as a first step towards realization of cGe/TiN/Steel stack for III-V integration on steel. Laser based annealing was used owing to low thermal budget requirement for device on steel to reduce iron diffusion problems. This work provides a low cost and scalable solution to the problem of integrating Ge on Si and steel with initial proof of concept applications.