Abstract Tu Hoang

Since 1965 the number of transistors on a single integrated circuit (IC) for minimum component costs has been almost doubling each 18 months, a phenomenon which is known as Moore’s law. The most advanced ICs contain more than a billion transistors and the line-width used in the next generation CMOS-processes is 32 nm. It is expected that in the coming decade the exponential increase in complexity will start saturating because we are approaching fundamental limits, and new approaches are being explored to push the development further after the saturation of conventional microelectronic technology. Instead of electrical interconnects, optical integration is suggested to be used in microelectronics. Furthermore, silicon nowadays remains the first material for IC technology. Meanwhile, a compact high-speed efficient silicon light source being suitable for on-chip integration process is still missing. The aim of this research is to investigate this missing component.

This dissertation presents an exploration on infrared-light emitting probability in silicon light emitting devices (Si-LEDs) with new approaches to improve the routinely low emission efficiency of silicon. It starts with a theoretical calculation of the possible internal quantum efficiency in bulk-Si LEDs and SOI-LEDs, after that a model for calculating the relationship between internal and external efficiencies is introduced. Following that the experimentally achieved results from our realized LED structures implementing the theoretical predictions are exhibited subsequently.

The probability ratio of radiative and non-radiative recombination in silicon determines the light emission efficiency of a Si-LED. The light emitting from silicon is inefficient because the unwanted non-radiative recombination processes are faster than the desired radiative transitions of electrons and holes. The highest possible internal quantum efficiency of photoluminescence and electroluminescence for a bulk Si-LED and an SOI-LED is theoretically calculated in a wide range of excess carrier concentration from 10¹⁵-10²⁰ /cm³. This calculation shows that a high purity level of silicon wafer is required for a high efficient light source. Furthermore, the non-radiative recombinations, such as Auger and Shockley-Read-Hall, occur mostly at the junctions and at the Si/SiO₂ interfaces. Thus, if one can inhibit the carriers from reaching these areas the radiative recombination will be enhanced consequently, i.e., the light emission efficiency is improved. Those information can be found in chapter 2 of this dissertation.

In chapter 3, the optical analysis equipment used in this work is described. Calibration of the optical system is an important issue to confirm the reliability of the obtained results. All information about the equipment used and procedures carried out in the calibration process of the optical system can be found in this chapter.

The importance of the purity of the silicon material is examined in the experimental series presented in chapter 4. The lattice defects were introduced in silicon wafer at different levels by implantation of either boron or silicon ions. When all device variations are realized on the same wafer the undesired variation from fabrication processes are excluded. The efficiency comparison of the emitted light between devices is made to investigate how dislocation-loops affect the light emission probability. A conclusion drawn from these experiments is that the Si-LEDs with less defects (high purity) will emit more light at the band to band wavelength of 1.15 µm. The created defects in silicon crystalline however do act as emitting centers at communications wavelength of 1.3-1.5 µm. These luminescence peaks are known as D-line emission (chapter 5). A wide range analysis of the dislocation-loop related emission is carried out at liquid nitrogen temperature up to room temperature. It is shown that the band-to-band recombination and the D-line are competing each other, i.e., the D-line emission is stronger in the LEDs with more dislocation loops. However, the D-line emission caused by dislocation loops is negligible at room temperature, whereas, at a higher temperature the band-to-band emission is stronger.

Chapter 6 presents a new approach in which the limitations of bulk-Si LEDs caused by the lack of a spatial confinement; such as: low switching speed, difficulty in formation of compact optical integrated structures, cross-talk problems; can be improved by using SOI technology. The key element for that improvement is that the emitting region in our SOI LEDs is a well confined area on a high quality SOI layer. An efficiency improvement of almost two orders compared to the reported thus far on SOI is the result of the successful utilization of carrier confinement effects in the realized SOI-LED structures. The way of using the standard local oxidation (LOCOS) technique to create two ultrathin access layers close to the p⁺/p and n⁺/p junction, respectively, in a p⁺/p/n⁺ structure is given. By doing that, three confinement mechanisms implemented at the same time in that emitting device are electrical field effects, geometrical effects, and band-gap widening effects. Simulation data supporting for those experimental results are also demonstrated in this chapter.

In conclusion, the success of this research project gives a contribution to the recent understanding of light emission from silicon and opens a highly potential application for high-speed photonic integrated circuits.