Semiconductor photodetectors for photon-starved applications in the short-wavelength infrared spectral region
Abstract
The design, fabrication and characterisation of planar geometry Ge-on-Si single-photon
avalanche diode (SPAD) detectors is described in this Thesis. These devices utilise a Si
avalanche multiplication layer, and an adjacent Ge layer to absorb short-wave infrared
incident photons. The innovative planar geometry design ensures the confinement of the
high electric field to the centre of the detector away from the exposed sidewalls resulting
in significantly reduced dark count rate (DCR).
Planar Ge-on-Si SPADs were fabricated and characterised in terms of single-photon
detection efficiency (SPDE), DCR, and timing jitter. These devices exhibited SPDE of
almost one order of magnitude greater than previously reported, with the highest SPDE
measured being 38%. The dark count rates per unit area were approximately 4 orders of
magnitude less than equivalent mesa devices. A record-low noise equivalent power of
4 × 10-17 WHz-1/2 was obtained, more than two orders of magnitude lower than the
previous best reported value. The lowest timing jitter of 26 µm diameter devices was
150 ps. These devices exhibited lower afterpulsing when compared to a commercial
InGaAs/InP SPAD detector, illustrating the potential for high count rate operation. An
investigation of an SPDE spectral dependence at different operating temperatures
revealed that efficient single-photon detection of 1550 nm wavelength light will require
an operating temperature of 245 K.
Laboratory-based light detection and ranging (LIDAR) experiments using the time-offlight approach were performed using an individual Ge-on-Si SPAD detector. This
approach allowed depth and intensity profiles of scanned targets to be reconstructed.
Based on these results, a parametric LIDAR model was used to estimate LIDAR
performance at long distances. For example, eye-safe sub-mW average laser power levels
would be sufficient for imaging at kilometre distances. It was demonstrated that by
employing appropriate image processing algorithms the total acquisition time can be
reduced down to a few seconds for a 10000 pixels image at kilometre range, illustrating
the potential for rapid three-dimensional imaging for automotive applications.