1. Introduction

In recent years, there has been growing interest in methods for generating and detecting radiation at terahertz (THz) frequencies (considered to lie between 300 GHz to 10 THz). Established photoconductive THz systems are typically based on femtosecond pulsed lasers driving optical switches fabricated from low-temperature-grown (LT)-GaAs [1]. In an alternative approach, continuous-wave THz (cw-THz) radiation can be emitted and detected by photomixing two cw lasers in a photoconductor. The difference in frequency between the two lasers is tuned to the THz region, and - providing the photoconductor can respond sufficiently rapidly - the conductance of the device is modulated at the difference frequency [2 ,3]. Compared to pulsed THz switches, cw-THz generation places more stringent requirements on the material properties - suitable photomixer materials require a very high dark resistivity (to allow the applied bias to become large without generating an excessive current), high carrier mobilities, and sub-picosecond carrier lifetimes [3]. Engineering these properties is not straightforward, but can be achieved in GaAs through low-temperature molecular beam epitaxial (MBE) growth followed by post growth annealing [4].

For reasons of cost, robustness and flexibility, for many THz applications it would be beneficial to deliver the laser radiation using fiber-optic cables. However, LT-GaAs is not an efficient photoconductor for excitation at fiber-optic wavelengths (1.06 μm, 1.3 μm, and 1.55 μm), because mid-band gap impurity state carrier transport and two-photon processes are inefficient [5]. Recently, we demonstrated a method for successfully manufacturing LT-In_0.3Ga_0.7As [6], which is compatible with 1.06 μm laser excitation [7], despite the fact that earlier reports had indicated that LT-growth of InGaAs might not be feasible [8,9]. Whilst our original demonstration restricted the use of this material to a pulsed THz system, in this paper, we demonstrate that the excellent material properties are also sufficient to enable both the generation and homodyne detection of cw-THz radiation, using LT-In_0.3Ga_0.7As photomixer devices.

Photomixers may also be implemented as coherent THz detectors, operating in a homodyne detection mode. Similar criteria are imposed upon the lifetime and resistivity as for the emission. There have been no previous reports of these materials being successfully demonstrated in coherent homodyne detectors. This may in part be because cw-THz detection is considered to be a very critical test of the material properties - as demonstrated in this paper. The ultrashort carrier lifetime is required to maximize the depth of modulation of the conductance, while high resistivities are simultaneously necessary to minimize the Johnson noise in the device.

2. Experimental technique

The cw-THz system is shown schematically in Fig. 1. Two independent external-cavity, tunable diode lasers operating around 830 nm are made collinear using a 50:50 beamsplitter. The combined beams are focused onto the emitter and detector photomixers, with combined optical powers of 50 mW and 20 mW, respectively.

 

Fig. 1. Schematic diagram showing the cw-THz apparatus with the combining beamsplitter, optical delay line, and THz photomixers. The inset on the left shows the spiral antenna, and to the right, the interdigitated fingers located at the central feed of the spiral.

Download Full Size | PDF

The photoconductive In_0.3Ga_0.7As layers were grown on semi-insulating (SI)-GaAs by MBE at a nominal growth temperature of 230°C. The excellent material properties were achieved by applying a post-growth anneal, in which the surfaces were passivated by contact with semi-insulating GaAs wafers. The 10 minute anneal was performed at a temperature of 450°C, below the 500–600°C commonly employed. Full details of the MBE growth, and low-temperature annealing technique can be found in reference [6]. The resulting material exhibits a minimum in carrier lifetime as a function of anneal temperature and, simultaneously, an enhanced material resistivity, relative to untreated InGaAs, of around 10⁴ Ωcm [6]. Time-resolved pump-probe photoreflectance measurements at 1.06 μm excitation indicated a sub-500 fs carrier lifetime [6].

The identical emitter and detector consist of 4×4 μm interdigitated fingers [10], shown as an inset to Fig. 1, loading self-complementary spiral antennas, also shown inset. The bias applied across the emitter electrodes was modulated at 25 kHz, and the resulting cw-THz radiation coupled out of the substrate using a 25 mm hyper-hemispherical silicon lens, producing a collimated THz beam. A similar lens couples the THz beam to the detector, via two plane aluminum 2 inch mirrors. The dc homodyne output signal is detected by a lock-in amplifier referenced to the emitter bias modulation. A time delay stage is used to change the phase between the optical beat and the THz beam, at the detector. The dc detected signal is maximized when the THz beam and optical beat arrive at the detector in-phase; therefore the time delay maps out an interferogram from which both the THz electric field amplitude and phase can be measured [3]. The THz frequency was tuned by a mechanical adjustment of the grating which forms the output coupler of the Littrow external cavity.

3. Results

Figure 2 shows the power, at a frequency of 0.4 THz, for an applied emitter bias varying between 0 and 7.5 V, and using a 30 ms lock-in time constant. Shown inset is the interferogram measured in the time domain, and its Fourier transform, at 7 V, which corresponds to a photocurrent of 0.4 mA. Optical heterodyne theory predicts that the THz power should vary quadratically with bias, assuming Ohmic transport within the photoconductive emitter [2]. However, for an applied bias in excess of 2 V, we measure a super-quadratic dependence. This has been observed in LT-GaAs photomixers, and the non-Ohmic behavior attributed to space charge effects, which result from the application of very high electric fields in the presence of recombination-limited transport [2].

 

Fig. 2. THz power as a function of applied bias for an all LT-InGaAs cw-photomixer system. Shown in the insets are an interferogram (top left) and normalized Fourier transform (bottom right) for an applied bias of 7 V.

Download Full Size | PDF

To examine the frequency response of a single photomixer device, and to allow a direct power comparison with an equivalent LT-GaAs emitter, the photomixer detector was replaced by a silicon composite bolometer cooled to 4 K. The LT-InGaAs photomixer performance was characterized as a function of THz frequency, by tuning of the output frequency of one of the diode lasers. The final plane mirror was replaced by an f/1 off-axis parabolic mirror to focus the THz beam into the input cone of the bolometer. A bias of 7 V was again applied to the emitter electrodes, this time modulated at 190 Hz, limited by the bolometer response time. The 190 Hz output signal from the bolometer was amplified using a 40 dB pre-amplifier, and then detected using a lock-in amplifier with a time constant of 300 ms. The results are shown in Fig. 3. The response of the antenna is relatively flat up until approximately 400 GHz. However, above 700 GHz, the power rolls off at a rate of approximately 6 dB/octave with a signal detected up to a frequency of 1.14 THz. The roll-off at 6 dB/octave implies the photomixer response in this frequency range is limited only by the carrier lifetime of the photomixer [11]. At higher frequencies (>1 THz), a 12 dB/octave roll-off would be expected, limited also by the antenna capacitance [11]. For comparison, the THz power detected from an identical device fabricated from LT-GaAs, is also shown in Fig. 3. To facilitate comparison of the best available device performance, in both cases the devices were driven close to the point of breakdown. For the GaAs device, this corresponded to a photocurrent of 0.75 mA, an incident laser power of 32 mW, and an applied bias of 20 V. The 6 dB/octave and 12 dB/octave roll-off regions are evident for the LT-GaAs antenna. The relevant system and material parameters for both measurements are summarized in Table 1.

 

Fig. 3. Plots of the roll-off in photomixer emitted power (measured using a bolometer) as a function of frequency for LT-GaAs and LT-InGaAs.

Download Full Size | PDF

Table 1. Comparison of the carrier lifetime (τ), and illuminated device resistance (R_device), along with the current (I), bias (V), and dissipated electrical power (P), for both LT-GaAs and LT-InGaAs emitters shown in Fig. 3.

View Table

3. Conclusions

Currently, it is clear that LT-InGaAs devices do not perform as well as those manufactured from LT-GaAs, with an approximate 15 dB discrepancy across the entire frequency range. In part, this is probably associated with the lower resistivity of the LT-InGaAs (see Table I). For similarly designed LT-GaAs photomixers, our measurements indicate that bias voltages in excess of 30 V can routinely be applied without device failure, with super-quadratic dependence appearing for biases above around 7 V. This is consistent with other literature reports, in which Brown et al have convincingly demonstrated the importance of highly resistive materials in accessing high emitter powers [2]. From Table I, it may be seen that the reduced resistivity of InGaAs has a large effect upon the dc power it is able to dissipate before breakdown, limiting the applied bias. Since it has already been shown that the THz power varies super-quadratically with bias, the difference in breakdown voltage alone adequately explains the 15 dB difference. A similar comparison using homodyne detection indicates a total system deficit of approximately 20 dB in comparison to an LT-GaAs system, when considering the combined effect of both emission and detection. Therefore further material innovations, aimed at improving the resistivity of the InGaAs devices, would be beneficial, both in improving the emitter output powers, and in minimizing the noise in the detectors.

With a smaller band gap, it must be accepted that InGaAs will typically exhibit values of electrical resistivity and breakdown field that are lower than LT-GaAs. However, if a significant reduction in cost and size can be realized using this technology, then a short term performance deficit may prove to be commercially acceptable.

To overcome this inherent problem, there are several prospects for the improvement of our LT-InGaAs material. In particular, several reports have recently described the incorporation of buried metallic nanoparticles in InGaAs, compensated by Be, to increase the resistivity of the material [12,13]. We expect this technique will also be of potential use to our material. Other avenues of research have recently emerged, such as ion-implantation of InGaAs [14], or ErAs doping of GaAs or InGaAs [12,15]. Each of these techniques produces short carrier lifetimes, (to below 300 fs in the case of reference [14]), but these studies have not yet simultaneously focused on the resistive properties of the material.

We also anticipate substantial improvements from better matching of the band gap to the laser diode wavelengths. With 830 nm excitation, the absorption depth in LT-InGaAs is lower than that expected from excitation close to the band edge. An increase in local carrier densities might result in premature device breakdown, and the excess energy of the photoexcited carriers will additionally alter the dynamics of carrier transport. Furthermore, any carrier transfer to satellite valleys in the conduction band, resulting from the excess energy and applied bias, would degrade the carrier lifetime and photoconductor temporal response. Detrimental effects of above-band-edge excitation have been demonstrated in LT-GaAs photomixers illuminated at an excitation wavelength of 585 nm, which showed a significantly reduced performance [11]. Regardless of the precise mechanism, the use of LT-InGaAs photomixers with 1.06 μm diode lasers is expected to lead to an immediate and fundamental improvement. The THz power is proportional to the product of the two laser wavelengths [16]. Therefore, an additional improvement of 63% is expected for 1.06 μm excitation, relative to 830 nm, for similar optical powers.

In summary, for the first time, an all LT-InGaAs cw-THz photomixing system has been demonstrated using inexpensive laser diodes. Photomixers fabricated on epitaxial low-temperature grown In_0.3Ga_0.7As are used for the generation and homodyne detection of radiation from microwave frequencies to beyond 1.0 THz.