Photorécepteurs 1,3-1,55 µm obtenus par intégration monolithique de photodiodes p-i-n et de transistors HEMT
This thesis deals with the monolithic integration of long wavelength (1.3-1.55 µm) p-i-n photodiodes with HEMT field effect transistors for the realization of wide bandwidth OEIC receivers for lightwave communication systems. Part of this work was carried out in collaboration with the Laboratory for Electromagnetic Fields and Microwave Electronics (IFH) of the ETH-Zürich. The p-i-n photodiode structure consisting of a n+-InP lower contact layer, a 1 µm undoped InGaAs absorption layer and a 100 nm p+-InGaAs upper contact layer was grown lattice-matched on an InP substrate by Chemical Beam Epitaxy (CBE). The p+-InGaAs layer allowed the fabrication of low resistivity ohmic contact. Being thin enough (100 nm) this layer did not induce any degradation of the quantum efficiency nor of the speed of response of the photodiodes. I-V measurements up to -25 V were performed at different temperatures to identify the processes causing the dark current. It was shown that the dark current at low voltage is dominated by a diffusion current and not by generation-recombination effects. At higher bias, the exponential behavior of the dark current with the applied reverse bias was well fitted by taking into account the effect of the electrical field on the generation process. To integrate horizontally the very different p-i-n and HEMT layer structures, a two-step growth approach was used. First, an AlInAs/InGaAs HEMT structure was grown lattice-matched on a semi-insulating InP substrate by Molecular Beam Epitaxy (MBE); except a thick AlInAs buffer layer (1,2 µm), this structure is conventional. Mesas down to the InP substrate were then formed using wet etching. Finally, in the second growth step, the InGaAs/InP p-i-n structure was grown by CBE. Comparisons between photodiodes grown in one-step and regrown photodiodes showed only minor difference; in particular the dark current of the regrown photodiodes remained very low : 130 pA at -5 V for 30-µm diameter diodes. The influence of the p-i-n regrowth on the two-dimensional electron gas (2-DEG) transport properties, that is to Say ns2D,μn2D and the channel sheet resistance Rcc, is more critical, as it was determined by both Hall and TLM measurements. In a first phase, the CBE regrowth was done using a SiO2 mask layer deposited by CVD on the top of the HEMT structure in order to achieve a selective area regrowth of the p-i-n structure. However this growth step, done at around 520°C for more than one hour, led to a severe degradation of the 2-DEG transport properties. After the first growth Rcc was 240 Ω/square and it became as high as 600 Ω/square after the p-i-n selective regrowth. Test samples consisting of AlInAs were prepared; an increase in the trap concentration was measured when these samples were exposed to a similar treatment. The results showed that the degradation was clearly due to the presence of the SiO2 mask layer and the CBE growth conditions themselves were not responsible. In order to overcome this problem, the SiO2 deposition step was omitted resulting in a non selective regrowth. In this case, the p-i-n material is also deposited on top of the HEMT structure. This scheme of integration is therefore only feasible if the undesired p-i-n layers can be selectively etched away without degrading the underlying HEMT structure. In a first set of experiments, this selective etching was performed using the n+-InP p-i-n lower contact layer as an etch stop. Satisfying results were obtained, since Rcc remained under 300 Ω/square. In a second approach, an InP cap layer was added to the HEMT structure before its patterning. This led to a very slight reduction in the 2-DEG properties with an increase of only 3 % of Rcc after the regrowth and the subsequent selective etching of the undesired p-i-n material. Using the latter scheme of integration, complete photoreceivers were successfully fabricated using standard contact photolithography, except for the definition of the 0.25-µm T-shape HEMT gates. The current gain cutoff frequency fT exhibited by single HEMTs was higher than 100 GHz and a flat photoresponse with a bandwidth of 18 GHz was obtained for the OEIC receivers. The average equivalent input noise current density was 11 pA/√Hz and the responsivity measured at 1.55 µm of the photodiodes was 0.5 A/W. These values led to a sensitivity of -18 dBm calculated for a data bit rate of 20 Gbit/s at a BER of 10-9.