Ge-on-Si photodiodes have been firstly reported more than twenty years ago opening the way for the integration of IR photodetectors on Si. A tremendous development has been done, moving from vertically illuminated, stand-alone devices, to waveguide integrated arrays of photodetectors and CMOS integrated imagers. Usually, the Ge epilayer act as the absorbing material for the NIR radiation, while Si acts only as a substrate. Here we report on a dual-band Ge-on-Si photodetector where light detection can take place both within the Ge epilayer and the underlying Si substrate: the device responsivity can thus be tuned from the VIS to the SWIR spectral range by means of an external bias. Therefore, the presented device can be implemented as a VIR-SWIR CMOS compatible imager.Two photodiodes in a back-to-back configuration are formed by p-doping the back side of a high resistivity Si wafer and by epitaxially growing a p-i-n heterojunction Ge-on-Si layer on the front side (Fig.1a)). An external bias can be applied to the whole layer stack by means of ohmic contacts formed on the top Ge layer and on the wafer backside (Fig.1b)). When a positive bias is applied between the top and bottom contact, the Ge diode is forward biased while the Si diode is reversed biased. In such a configuration the photocurrent generated within the Si substrate by back illuminating the device, will flow through the external circuit and be detected. By reversing the bias polarity, still maintaining a backside illumination, the photocurrent will be generated within the Ge epilayer: in this way it is possible to tune the device responsivity simply by controlling the external bias. A TCAD model has been set up to optimize doping levels, in particular that of the Si substrate, determining the most effective position the p-n junction within the Si photodiode and investigating the role played by substrate thickness[1].According with the simulation outcome, a p-type layer has been formed on the backside of an intrinsic Si wafer by spin-on-dopant obtaining a boron-doped layer with an average concentration of 2×1019cm-3 and a thickness of ≈300nm. The epitaxial growth has been performed by low-energy plasma-enhanced chemical vapor deposition (LEPECVD).Firstly, a 200nm thick phosphorous-doped Si layer has been deposited at 760°C followed by 3 µm thick nominally undoped Ge, grown at 500°C. Six thermal annealing cycles between 600-780°C have been performed in-situ to reduce the threading dislocation density. As a final step 150nm of boron-doped Ge have been deposited.Ge mesas with dimensions ranging between 100-500µm have been realized by standard UV-lithography and dry etching on the Ge-on-Si epitaxial layer. The mesas have been etched down to the intrinsic Si surface to electrically isolate the different devices on the same chip. The top ohmic contacts have been realized by e-beam deposition of a Ti/Al metal stack and lift-off. Different bottom ohmic contact have been tested both by depositing two rectangular stripes of Ti/Au or by depositing a transparent ITO layer on the backside of the chip.Optical characterization was performed by means of a lamp-monochromator set-up. The VIS/SWIR bias dependent dual-band operation of the device is demonstrated in Fig.1c), where the spectral responsivity is shown for two different applied voltages. The peak responsivities are 0.41A/W and 0.63A/W at 960nm and 1520nm, respectively. The specific detectivity D* of the device has been found to be 7·1011cmHz1/2/W and 2·1010cmHz1/2/W in the VIS and NIR, respectively[2].VIS-SWIR single-pixel imaging was performed by mounting the chip on a rastering system comprising a three-axis stage. In this way the pixel can scan the image plane behind a piano-convex lens. Two images, one in the VIS and one in the SWIR, can be obtained by repeating the scanning and changing the polarity of the applied bias at the device Fig.4d). A coloured VIS image has been obtained by repeating the measurements three times with three different optical bandpass filters (450nm, 550nm and 650nm) resulting in three images with R, G, and B components, which have been combined to obtain the final VIS image.In conclusion, we have designed and fabricated a voltage-tunable dual-band photodetector operating in the VIS and SWIR range. The device is based on a couple of p-i-n photodiodes connected back-to-back formed within a Ge-on-Si epitaxial structure. The device architecture enables to electronically select the shorter (400-1100nm) or longer (1000-1600nm) wavelength range with a relatively low applied voltage. VIS-SWIR single-pixel imaging has been performed.[1] A.De Iacovo, et al., Journal of Lightwave Technology, 37(14), 3517(2019).[2] E.Talamas Simola, et al., Opt. Express 27, 8529-8539(2019). Figure 1
Read full abstract