375GM-bandwidth photoconductive detector

Much progress has been made over the past several years in the development of high-speed photodiode detectors. A detection bandwidth of 105 GHz together with a responsivity of 0.1 A# have been reported for a metalsemiconductor-metal (MSM) photodiode.’ The most common approach to increasing the bandwidth in MSM photodiodes (at least up to 100 GHz) is to shorten the carrier transit time by reducing the electrode spacing. However, achievement of bandwidths > 100 GHz requires more than simply further reducing the electrode spacing. Monte Carlo simulation of the intrinsic response for a photodiode with 0.1~pm electrode spacing, for example, shows a response tail persisting for picoseconds and having an integrated energy comparable to the main signa1.53 This tail is caused by the long transit time of the photogenerated holes, which is almost 10 times that of electrons. The response times of photoconductive detectors, on the other hand, can be quite fast because they are determined solely by the carrier lifetime of the material that is used. Recently, low-temperature-grown GaAs (LT GaAs) has been applied to ultrafast and high-power’ optical switching. The subpicosecond carrier lifetime,6 high mobility ( > 200 cm’/V s), and high breakdown field strength ( > 100 kV/cm) of LT GaAs make this material ideal for electrical pulse generation and gating. Such applications have so far required use of moderate-to-high peak optical powers. We now report a LT GaAs-based photoconductive detector that takes advantage of the high breakdown field capability of LT GaAs to greatly improve sensitivity. in a photodiode, a reduction in electrode spacing improves speed with little change in sensitivity. In a photoconductive detector, by contrast, such a reduction increases the sensitivity with little change in speed. Decreasing the carrier transit time across the semiconductor gap to a value comparable to the carrier lifetime increases the photocurrent gain to a value of unity.7 For a carrier lifetime of 1 ps, this condition is met when the electrode spacing, i.e., the actual gap between electrodes, is 0.1 pm. A further decrease in the electrode spacing could increase the photoconductive gain to an even higher value, provided the metal-semiconductor contacts are ohmic. With unity gain, the responsivity of a LT GaAs photoconductive detector becomes comparable to that of a photodiode. To demonstrate this principle, we fabricated a LT GaAs photoconductive detector with interdigitated electrodes having finger widths and spacings of 0.2 pm. A 1.5~pm-thick LT GaAs layer was grown on a (100) semiinsulating GaAs substrate that has a 0.4~pm-thick conventionally grown undoped buffer layer. The LT GaAs layer growth was performed using molecular beam epitaxy at a substrate temperature of 190 “C, followed by annealing at 600 “C for 10 min in an arsenic overpressure. The interdigitated electrodes were fabricated of 300&2000-A Ti/Au using a JEOL JBX SDIIF direct-write electron-beam lithography system. The active area of the detector is 6.5 ~7.6 pm’. Coplanar transmission line electrodes of 500A/2500-A Ti/Au, with 20+m widths and spacings and 5-mm lengths, were also fabricated on the LT GaAs using conventional optical lithography. As shown in Fig. 1, the LT GaAs photoconductive detector was placed between coplanar transmission lines (Z, = 90 (1) to assure good coupling of the generated electrical pulse to the propagating mode and also to eliminate parasitic losses. The detector was not antireflection (AR) coated in this initial work. A reference transmission line of identical dimensions, but without the interdigitated-electrode detector, was also fabricated on the wafer to determine the system response. The technique of sliding-contact switching, which provides the excitation between the lines without the interdigitated-electrode detector, can have a response CO.5 ps.8 The photoconductive detector was characterized using the technique of external electro-optic (EO) sampling,’ as depicted in Fig. 1. A balanced, colliding pulse, modelocked dye laser operating at 610 nm with a repetition rate of 100 MHz was used, which produces 150-fs pump and probe pulses. Although not shown in the figure, the EO sampling crystal spanned both the detector/transmissionline assembly and the reference transmission line, so that the translation of the pump and the probe beams required to make either measurement was only -200 r(Lm. A bias of 10 V dc was applied to the detector before breakdown occurred, corresponding to a breakdown field strength of 500 kV/cm. This value is more than twice the highest that has been reported for LT GaAs under dc-bias conditions, which was obtained using 20-,um-spaced electrodes.’ Our result represents a better measurement of the actual breakdown field strength for LT GaAs, since our 0.2~pm electrode spacing confines the electric field to the 1.5~pm-thick LT GaAs epilayer. The dark current for 1 V applied to the detector is 100 pA. At 8 V (400 kV/cm),