The fundamental bandgap of Si makes it transparent and insensitive to light with wavelengths longer than 1110 nm (or equivalently, less than 1.12 eV). This inconveniently excludes most of the infrared spectrum, including the most technologically useful portion in telecommunications (e.g. the 1.31 µm and 1.55 µm bands). Alternatives such as HgCdTe, III-V, and Ge-based infrared photodiodes can be expensive and fall short especially in terms of their integration with the existing Si-CMOS manufacturing process. There is thus a strong incentive to improve the infrared detection capability of Si, with potential for further applications in photovoltaics, optical computing, and infrared imaging. In the last decade, enhanced near-to-mid infrared photoresponse in Si has been demonstrated in Si hyperdoped with transition metals such as Ti and Au, fabricated from high dose ion implantation followed by nanosecond pulsed laser melting, which can, in principle, incorporate up to a few atomic percent of transition metal impurities in Si. However, these transition metal-hyperdoped Si photodiodes were demonstrated to have extremely low external quantum efficiencies (< 0.01%) that do not appear to be consistent with measured high levels of absorption in the near-infrared. Further, there remains ambiguity in the literature about the presence and role of extended crystalline defects in affecting the observed optical, electrical, and photodiode characteristics.
This talk sets out to explore some of these questions, and will be divided into two parts. In part one, defect characterisation experiments performed on Au-hyperdoped Si and a similarly prepared control Si diode (without Au hyperdoping) are presented. Processing-induced vacancy-type deep levels were measured in both samples from deep level transient spectroscopy, and its implication to photodiode performance is discussed. Analysis of optical and photocurrent characteristics of these samples revealed several shortcomings that limit photo-carrier collection efficiency in our Au-hyperdoped Si photodiodes. Alternative device architectures to overcome some of these limitations and preliminary optimization efforts are presented. On the other hand, the second part of this talk focuses on exploring Si hyperdoping possibilities beyond Au, notably for Ag and Ti. Extensive physical characterization revealed that it is not possible to achieve hyperdoping of Ag or Ti in Si due to near-complete surface segregation and highly defective epitaxy during pulsed laser melting, leaving Au as the only viable impurity to date for achieving the required level of hyperdoping in Si for efficient sub-bandgap light absorption. Factors that limit successful transition metal hyperdoping in Si are discussed, as well as some potentially fruitful avenues for future work.