1. Introduction

Terahertz waves have significant potential for varied applications. They are efficient for the safe detection of prohibited substances like drugs [1], explosives [2], and lethal gases [3]. They are ideal for security and environmental applications. THz waves have also been shown to be very sensitive to polymorphic structures [4] making them particularly interesting to the pharmacological industry. High contrast imaging systems, based either on continuous-wave or pulsed THz sources, offer major advantages in security [5], microelectronic [6] and medical domains [7–9]. The THz band is also becoming increasingly popular in wireless ultrafast communication systems [10] and proving promising for tracking/authentication/identification systems [11]. With continued efforts towards the development of compact, cost-effective, robust and reliable tabletop THz sources and detectors, exciting developments in the THz technology can be expected.

In many terahertz time-domain spectroscopy (THz-TDS) systems, pulsed THz waves are radiated from photoconductive antennas [12] that are generally made of GaAs-based semiconductor materials and triggered by femtosecond pulses from a Ti:sapphire laser at about 800 nm. Such a THz setup is sensitive to vibrations and small changes in ambient temperature and relative humidity. Using instead high power femtosecond fiber laser sources emitting at 1030, 1300 and 1550 nm (based on Yb-, Pr- and Er-doped fiber gain media, respectively) it is now possible to build environmentally robust and compact THz spectroscopy and imaging systems. However, when shifting to long wavelengths, different materials have to be considered in order to replace GaAs-based photoconductors now turned transparent. Interesting results have been reported by many research teams at 1550 nm on using efficient terahertz devices made with InGaAs ternary compounds, which were produced via various fabrication approaches [13–18]. These approaches are briefly described in the next section. In light of these works, we propose to investigate the use of quaternary III–V photoconductive materials (i.e., InGaAsP) modified by a post-growth fabrication process based on high fluence Fe ion implantation at low temperature (83 K) and optimized by thermal annealing treatments. We are presenting characterization results of the material optimization and conclude the work by showing results on optical generation of broadband (2 THz) electrical pulses at 1550 nm from Fe-implanted InGaAsP photoconductive antennas.

2. Rationale and fabrication of Fe-implanted InGaAsP

Band-to-band photocarrier generation with femtosecond Er-doped fiber lasers is efficient when the photoconductor band gap energy E_g is smaller than 0.80 eV (i.e., the photon energy at 1550 nm). The ternary alloy In_0.53Ga_0.47As meets this criteria (E_g = 0.74 eV) and has been the subject of many developments in recent years. Its fabrication by low-temperature (LT) growth was studied first and led to short carrier lifetimes [19]. However, it was found to electrically underperform LT grown GaAs. The dominant traps, assigned to point defects, are too shallow for adequate electrical isolation. As a result, extra processing has to be performed in order to increase the material resistivity after annealing. Carrier compensation with berylium doping does not allow for a sufficient increase of the resistivity. Some authors suggest to reduce slightly the gallium content (thus increasing the material band gap) for superior results [13]. Other epitaxy-based approaches were tested recently and photoconductors exhibiting suitable properties for making THz devices have been obtained by i) incorporating ErAs metallic precipitates in Be-doped In_0.53Ga_0.47As [14], ii) incorporating iron impurities that act as mid-gap states in In_0.53Ga_0.47As [15], or iii) using an especially designed InGaAs/InAlAs multilayer structure for which absorption and photocarrier trapping occur in different layers of the structure [16]. Be-doped plasma-assisted epitaxy has also been used to fabricate ultrafast III–V compounds [20] but such materials have not yet been exploited for optical THz emission. Despite the high level of material control that can be achieved by all these epitaxial growth techniques, the development of new structures and growth recipes may be particularly time consuming (i.e., requires numerous wafer pieces and growth hours) depending on the number of parameters to be optimized. Standard epitaxial In_0.53Ga_0.47As layers can also be damaged by ion beam processing, which is a popular post-growth fabrication technique that reduces the carrier lifetime [21]. For instance, ultrafast ternary layers were produced after damaging them with heavy ions (i.e., by irradiation [18] or implantation [17, 22]). It was found that energetic atomic recoils from the collision cascade triggers the formation of a high number of small defect clusters which further evolve into defect complexes after subsequent thermal annealing treatments. These complexes, unlike point defects, are believed to increase the ternary’s electrical resistivity through forming deep carrier traps and efficient carrier recombination centers. When iron is the implanted species, it appears to combine with defect complexes and contribute to raise the resistivity. This is probably due to its capacity for producing extra mid-gap states [22]. However, resistivity levels discussed in the literature in relationship to ultrafast irradiated or Fe-implanted In_0.53Ga_0.47As are typically lower than those pertaining to materials fabricated by epitaxial growth only [13–16]. High resistivity is desirable since Joule heating becomes a problem when operating photoconductive emitters under high DC electric fields. Overall, the resistivity of most ultrafast materials based on In_0.53Ga_0.47As struggles to exceed 500 Ωcm, sometimes with low Hall mobility of less than 100 cm²V⁻¹s⁻¹, a situation which prompts to look for alternatives.

InGaAsP quaternary alloys offer some extra design flexibility over In_0.53Ga_0.47As. Lattice-matching with InP is met by adjusting the molar fraction y in (In_0.53Ga_0.47As)_y(InP)_1−y. Its bandgap can be tuned from 0.74 eV up to 1.35 eV. It becomes possible to tailor the absorption band edge of the quaternary photoconductor close to any of the aformentioned rare-earth fiber laser wavelengths. This could prove useful for avoiding intervalley carrier scattering [23], particularly with the use of Yb-doped fiber sources. As a rule of thumb, semiconductors with higher bandgaps show larger intrinsic resistivity. For use with Er-doped fiber lasers, band-to-band absorption at 1550 nm is allowed for molar fractions of about y = 0.9 or greater. Relative to In_0.53Ga_0.47As, a small increase in the band gap is possible (E_g = 0.79 eV) which corresponds to a factor of about 2 or 3 in higher intrinsic resistivity. Amongst the fabrication techniques listed here, we select the Fe implantation post-growth process, given that the epitaxy of quaternary alloys is based on already rigorous calibrations.

In this context, undoped InGaAsP layers (1.5 μm thick) were grown on semi-insulating InP wafers (75 mm diameter) by metal-organic chemical vapor deposition. The quaternary layer was sandwiched between thin (100 nm) undoped InP buffer and capping layers. Two wafers were selected to assess the end-to-end reproducibility of the material processing. Their room-temperature photoluminescence (PL) peak wavelengths were 1565 nm (wafer 1) and 1575 nm (wafer 2). Offsets of −240 arc sec (wafer 1) and −380 arc sec (wafer 2) in the high resolution X-ray diffraction rocking curve were observed for the (004) reflection, revealing a small lattice mismatch. Thus the starting material was In_1−xGa_xAs_yP_1−y, with x = 0.39 and y = 0.865 for wafer 1 and with x = 0.38 and y = 0.86 for wafer 2. The level of compressive strain (0.1% and 0.15%) was acceptable and layer coherence was maintained.

Using a Tandetron 1.7 MV accelerator (High Voltage Engineering Europa), one quarter of each wafer was implanted with an incident beam of Fe ions, impacting the surface at a 7 degree angle. The ion fluence level was set around 10¹⁵ cm⁻² which produced better overall results in previous Fe-implanted In_0.53Ga_0.47As studies [22]. Five energies (0.25, 0.5, 1, 1.8, 2.5) MeV and respective fluences (0.11,0.22,0.33,0.44,1) × 10¹⁵ cm⁻² were determined by srim software simulations [24] and optimized for flat damage and uniform iron profiles through the whole quaternary layer. Summing for all energies, a total fluence of 2.1 × 10¹⁵ cm⁻² was implanted, resulting in an average Fe concentration of 1.1 × 10¹⁹ cm⁻³ in the InGaAsP layer. At these high Fe fluences, the quaternary layer is heavily damaged, if not amorphized [17]. srim software simulations predicted 6.5 atomic displacements per atom (using 10 eV as the displacement energy for In, Ga, As, and P atoms). According to Too et al. [25], Marcinkevičius et al. [26] and Subramaniam et al. [27], the implantation temperature can also affect significantly the electrical properties of Fe-implanted In_0.53Ga_0.47As at high fluences. The InGaAsP samples were implanted at 83 K for that very reason. The cold implantation suppresses detrimental dynamic defect annealing that may otherwise occur during room temperature implantations [28].

The damaged quaternary layers were then processed by rapid thermal annealing (RTA) in order to modify and optimize their electrical and optical properties. For each wafer, a series of cleaved samples were maintained for 30 s at distinct temperatures between 400 °C to 800 °C in a nitrogen filled thermal processing chamber (Jipelec, Jetfirst). A 100 mm silicon wafer (which served as a susceptor) and a silicon proximity cap protected both top and bottom surfaces from phosphorous desorption. Prior to characterization, the InP capping layer was removed by selective wet chemical etching.

3. Effects of Fe ion implantation and RTA on critical properties for THz emission

3.1. Electrical Hall measurements

The carrier transport in Fe-implanted InGaAsP layers was investigated by Hall measurements (at 300 K, in dark ambient conditions). Resistivity, carrier density and Hall mobility were estimated assuming a uniform layer, a single carrier type and a unity scattering factor. Measurements were made in the van der Pauw geometry on 6 mm × 6 mm samples with indium contacts alloyed at 300 °C for 60 s. Hall coefficients were recorded at 0.36 T. Measurements on Fe-implanted InGaAsP samples reported in Fig. 1 pertain to two wafers. Properties of the as-grown InGaAsP epilayers are also tabulated in the inset of this figure. Hall coefficients values are negative in all cases, the electrons being the majority carrier in the as-grown material.

 

Fig. 1 Effects of a 30 s RTA (400 °C to 800 °C) on (a) resistivity, (b) free carrier density and (c) Hall mobility of implanted InGaAsP with an Fe fluence of 2.1 × 10¹⁵ cm⁻² at 83 K. Data for as-implanted samples are indicated at 300 °C, which was the In alloying temperature. Two wafers were used and their as-grown parameters are tabulated here. The solid lines are only guides for the eye.

Download Full Size | PDF

For both wafers, the Hall transport results show similar influence of the Fe implantation and RTA processes. At the lowest annealing temperature (300 °C), necessary for In alloying, we find that ion implantation at high fluences causes a drastic reduction of the Hall mobility, compared to the as-grown material. By increasing the RTA temperature up to 500 °C, we observe an important decline of the free carrier density (Fig. 1b) along with a progressive recovery of the Hall mobility (Fig. 1c). Then, maximum resistivity values of 2500 Ωcm and 1400 Ωcm and minimal carrier density values of about 10¹³ cm⁻³ were achieved at 600 °C and 500 °C for wafer 1 and wafer 2, respectively (Figs. 1a and 1b). The resistivity increases by more than a 10⁴ factor with respect to the as-grown material. The Hall mobility peaks at slightly different RTA temperatures. Still, the highest resistivity sample of both series show a reasonable mobility, of about 400 cm²V⁻¹s⁻¹. By further increasing the RTA temperature beyond 600 °C, our results show a rapid reduction of the material resistivity associated with a strong increase of the carrier density and a gradual decrease of the carrier mobility.

Low mobility hopping conduction is expected to be the actual carrier transport mechanism in an amorphized quaternary layer [22]. Accordingly, our implanted samples annealed at the lowest temperature do show a 100-fold reduction in mobility with respect to as-grown layers. As the annealing temperature increases, mobile defects can diffuse, annihilate or agglomerate to form complexes along with the implanted Fe that act as deep level carrier traps. Because the thermal annealing reduces the number of defects and restores to some extent the crystalline quality of the layer, carrier mobility tends to increase, especially between 400 °C and 550 °C (i.e., channels of conductive paths gradually open as the temperature increases). At RTA temperatures above 600 °C, the upturn in free carrier density follows a carrier release mechanism. Either shallow donor defects are forming, such as As or P antisites [22], or donor impurities intrinsic to the material are re-activated. Other possible mechanisms might be related to the dissociation of Fe defect complexes, or related to the in-diffusion of interface contaminants such as C or Si, or else the out-diffusion of Fe at high annealing temperatures, as it typically does [25]. Detailed analysis of impurity depth profiles with secondary ion mass spectrometry and advanced capacitance-voltage measurements would be required for a proper assignment. These shallow levels are probably ionized which explains the gradual decrease of carrier mobility with temperature. The activation of Fe, which is expected at high temperatures, did not cause the resistivity to re-increase, as observed by Carmody et al. with Fe-implanted In_0.53Ga_0.47As [22].

3.2. Optical absorption measurements

The optical transmission spectra of the samples were measured with an arc lamp source (Oriel, model 66881) and a spectrum analyzer (Ando, model AQ6317). The samples were illuminated at normal incidence with a collimated beam (0.7 mm spot size). The power spectral density was kept below −65 dBm/nm. A 1 nm resolution bandwidth gave an adequate dynamic range. At this resolution, the Fabry-Perot interferences, caused by multiple reflections from the facets of the 600 μm thick InP substrate, are partially smoothed out. Signals transmitted through and without the sample were recorded subsequently, at each wavelength position.

Room-temperature transmission loss spectra of the as-grown, implanted and annealed samples are shown in Fig. 2. The inset shows the fraction of the light absorbed by the samples around 1550 nm obtained after subtracting the amount of light reflected by the sample’s facets. For the as-grown quaternary layer, we observe a step-like transmission enhancement at 1575 nm which corresponds to the absorption band edge of this material. For wavelengths longer than the band edge wavelength, the small 2.4 dB transmission loss is caused by 27% facet reflections. After the implantation of Fe at a fluence of 2.1 × 10¹⁵ cm⁻², the band edge signature is no longer visible and the transmission loss becomes larger by more than 3 dB at 1550 nm. The increased absorption factor at short wavelengths might be caused by strong scattering of light on large defects present in the amorphized layer. The absence of a sharp transmission feature at 1575 nm, on the other hand, can be explained by extra optical transitions involving trap states that are distributed within the band gap of the material. Annealing the sample at 600 °C reduces the transmission loss to the level of the as-grown material and extra absorption is still observed for wavelengths longer than 1550 nm due to the remaining energy-distributed trap states. The total absorption of the laser light at 1550 nm by this 1.5 μm thick InGaAsP photoconductive layer is only 0.5. This corresponds to an optical attenuation coefficient of 4700 cm⁻¹. By contrast, In_0.53Ga_0.47As has an absorption coefficient of about 7500 cm⁻¹ at 1550 nm [29]. A ternary layer of the same thickness would absorb about 0.75 of an incoming light. These relatively small factors need to be taken into account when coupling these photoconductors to Er-doped fiber laser sources.

 

Fig. 2 White light transmission loss spectra of as-grown, implanted and annealed In-GaAsP/InP samples from wafer 1. The spectrum analyzer resolution bandwidth = 1 nm, the scanning step = 1 nm, and a 10-point Savitsky-Golay smoothing was applied. The y-axis reference level is located at 2.4 dB, it is the average transmission loss from a cavity with facet reflection of 27 %. The figure inset shows the optical absorption factor within the 1.5 μm-thick quaternary layer around 1550 nm.

Download Full Size | PDF

3.3. Photocarrier trapping time measurements

3.3.1. Experimental setup

A schematic of the transient differential reflectivity setup used to estimate the carrier trapping time of Fe-implanted InGaAsP layers is shown in Fig. 3. An Er-doped fiber laser source developed by the Institut National d’Optique was used for these measurements. This source delivers 250 fs pulses at a repetition rate of 20 MHz, and with an average power of 1 W. The linearly polarized output beam, centered at 1550 nm, is split into cross-polarized pump and probe beams. Their power ratio is set by the rotation of a half-wave plate placed before the polarizing beam splitter. In the context of our study, the pump and probe powers are set to 10 mW and 300 μW, respectively. The pump and probe beams are focused on the sample and the overlapping spots have full widths at half maximum of 20 μm and 10 μm, respectively.

 

Fig. 3 Schematic of the transient differential reflectivity setup configured for testing semiconductor chips (SC) at 1550 nm.

Download Full Size | PDF

The pump beam illuminates the sample surface at normal incidence and its intensity is modulated at 1 kHz with a mechanical chopper. The probe beam is reflected by the sample surface and then collected and redirected on two separate photodiodes. The output of photodiode 1 is sent to a lock-in amplifier for a measurement of the pump-induced reflectivity (ΔR) signal. The output of photodiode 2 is low-pass filtered (the cut-off frequency being much lower than the chopper frequency) giving rise to the average reflectivity signal. A pinhole is inserted into the detection path and spatially filters the reflected probe beam such that the detected signal mainly originates from a 1 μm spot size on the sample. The pinhole also reduces significantly the level of pump scattering from the semiconductor surface. A high-pass optical filter and a polarizer are placed in front of photodiode 1 in order to cut photoluminescence at 1565 nm and the remaining pump intensity, respectively. Under these experimental conditions, the noise level is low enough to detect relative changes of reflectivity as small as 0.001%. Finally, a variable delay line is used to record ΔR/R signals at different time delays between the pump and probe pulses hitting the sample surface.

Differential reflectivity measurements were made with wafer 2 sample series, which was annealed at temperatures between 400 °C and 750 °C. We chose this sample series because of better overall surface quality (as compared to wafer 1 series). We found out that small particles tend to appear on the silicon susceptor after many RTA cycles. This influences the quality of the InGaAsP top surface, which is facing the susceptor during the RTA process. The wafer 2 sample series was processed just after replacing the susceptor with a new one.

3.3.2. Results and analysis

Figure 4 shows the differential reflectivity signals (circles in Fig. 4a) which are plotted on a semi-log graph as a function of the pump-probe time delay. The experimental curves exhibit exponential decay behaviours. The reflectivity signal from the sample annealed at 400 °C decays very rapidly with a single exponential. As the annealing temperature increases, the decay time becomes longer and a second exponential decay has to be taken into account, in order to describe the experimental data at longer pump-probe time delays.

 

Fig. 4 (a) Normalized differential reflectivity measurements (circles) and fitted curves (solid lines) for InGaAsP (wafer 2) implanted with Fe at a fluence of 2.1 × 10¹⁵ cm⁻² at 83 K, after 30 s RTA at various temperature. (b) Amplitude ratio A₁/A₂ and (c) decay times τ₁ and τ₂ are plotted against the RTA temperature, after their extraction from curve fitting.

Download Full Size | PDF

Those differential reflectivity signals could be modeled by two convolution products involving the laser probe pulse, the pump pulse, and the impulse response of the sample:

Relative pump-induced variations of the Fresnel reflection coefficient ΔR/R are generally small enough (no more than a few percent) to be considered linearly dependent on the excess electron-hole pair density [30]. The analysis, however, is often complicated by contributions of several mechanisms that affect the total absorption change Δα(t) and its related real refractive index change Δn(t) [31, 32]. Since the maximum of the ΔR/R signal varies almost linearly with the excitation pump power (from 1 mW to 10 mW), the hypothesis we made was that the transient differential reflectivity is more or less proportional to N_ph(t) (which is either the number of photo-created electron-hole pairs due to band-to-band absorption or the number of free electrons in the conduction band due to absorption from deep levels).

Figures 4c and 4b show the photocarrier population decay times and amplitude ratio extracted from the fitting procedure applied to each sample. The shortness in decay times observed with all samples is somehow related to capture and recombination mechanisms associated with traps located directly at the surface of the semiconductor and within the photoexcited region. After a 400 °C anneal, intrinsic and implantation-induced surface defect densities remain so high that photocarriers are captured very efficiently by surface traps (τ₁ < 0.2 ps). For samples annealed at temperatures of 500 °C and above, a slower second exponential decay time τ₂ is observed. The relative amplitude of this second recombination channel increases with the RTA temperature. This behavior might be explained by the back-diffusion of deep photocarriers towards the surface since this transport channel opens up once carriers initially photoexcited near the surface partially fill surface traps. Fig. 4c shows that both decay times (τ₁ and τ₂) become longer as the annealing temperature increases, although τ₁ seems to saturate at about 0.45 ps. These trends are related to the thermal annealing process which reduces the density of ion implantation induced defects at the surface and within the 1.5 μm thick epilayer. The upper value of 0.45 ps for τ₁ is most likely related to remaining intrinsic surface defects that cannot restored by thermal annealing. Explanations presented here might not be unique. For instance, the contribution of two independent populations of carriers also give rise to a double exponential decay. In order to validate these working hypotheses, further experiments are planned. The influence of the laser pump power on the transient differential reflectivity curves is going to be studied over a wide range of pump powers.

In summary, optimal electrical properties for making a photoconductive THz antenna device are produced for RTA temperatures ranging from 500 °C to 600 °C (see Fig. 1). In this range, the fast decay time in transient reflectivity measurements is subpicosecond. The presence of implantation-induced defects throughout the InGaAsP layer limits the second decay time to a few ps. With such small decay times (⩽ 3 ps), a THz-TDS system made with Fe-implanted InGaAsP THz photoconductive antennas is expected to show more than 1 THz of usable bandwidth.

4. Antenna fabrication and THz emission with Fe-implanted InGaAsP

THz emitters were developed using dipolar antennas fabricated on InGaAsP samples taken from wafer 1 which were cold-implanted with Fe ions. The samples were annealed for 30 s at 600 °C. Ohmic contacts were formed by e-beam deposition of a 285 nm thick layer of a standard mixture of Ni/AuGe/Au, followed by a 30 s anneal at 410 °C. Coplanar electrodes (10 μm wide, 8 mm long and spaced by a 120 μm gap) were patterned by standard lift-off photolithography. A current-voltage curve was recorded for this device and a dark current value of 8.4 μA was found at 1 V bias. With this particular electrode geometry, it corresponds to a layer sheet resistivity of 7.9 MΩ/sq and a material resistivity of 1200 Ωcm. The THz emitter was mounted on a holder and its electrodes were wire-bonded to a coaxial connector pair. A hemispherical high resistivity silicon lens was placed in contact with the back side of the device for an efficient collection of the THz radiation.

The broadband emission of the photoconductive antennas was recorded using a THz time-domain spectroscopy setup coupled to the Er-doped fiber laser source described previously. The excitation laser power was about 80 mW and the pump beam intensity was modulated at 1 kHz using a mechanical chopper. The antenna was operated at 50 V, well below thermal runaway. The 1550 nm excitation beam was focused into a spot size diameter of about 20 μm (using the 1/e intensity fall criteria). We did observe an enhancement of the emitted THz radiation by exciting near the anode of the antenna (the radiation was measured using a bolometer). This behavior might be related to the ionization of traps that enhances locally the electric field (as reported by Ralph and Grischkowsky [33]) but contrary to trap-enhanced-field THz sources made on GaAs, we observed a similar enhancement by exciting near the antenna’s cathode. Experiments are ongoing in order to better understand this behavior. The radiation emitted by the antenna was collimated and refocused on the THz detector using a combination of the silicon lens followed by four off-axis parabolic mirrors. This mirror configuration was chosen because our setup is also designed to perform absorption measurement and visible-pump –THz-probe experiments. The electric field of the THz pulses was measured with electro-optic sampling detection and a 10 mW probing beam in a 0.5 mm thick < 110 > ZnTe crystal. The THz signal was retrieved using a Pockels cell setup with two balanced photodiodes and a lock-in amplifier. THz traces were recorded point by point using a delay line. The spectra of these traces were then produced using numerical Fourier transforms.

Figure 5 shows the temporal trace of the THz pulses (emitted by the InGaAsP-based antenna) and its corresponding spectrum. The Fourier-transformed THz spectrum peaks at 500 GHz. It exhibits a usable spectral range of 2 THz, and a maximum amplitude signal-to-noise ratio of 25 dB. The spectral bandwidth is likely limited by the detector since a large mismatch exists between the group velocity of the 1550 nm probing pulse and the phase velocity of the THz pulse in ZnTe [15, 34]. Further measurements using a thin GaP electro-optic crystal are underway to confirm this. It is also known that the antenna’s geometry can have significant effect on the bandwidth. Results shown in the context of this study were obtained on the very first antennas made with already existing lithographic masks. Nevertheless, their overall characteristics are comparable to the best ones made on In_0.47Ga_0.53As:Fe photoconductive materials [15]. The complete spectral investigation of the InGaAsP emitter against bias voltage and pump power will be the subject of a separate publication.

 

Fig. 5 (a) Time-domain signal emitted from a photoconductive antenna made of Fe-implanted InGaAsP annealed at 600 °C, excited by 250 fs pulses from an Er-doped fiber laser (P_pump = 80 mW) at V_bias = 50 V (4.2 kV/cm) and detected using a 0.5 mm thick ZnTe electro-optic crystal. (b) Amplitude spectrum obtained by fast Fourier transform of the temporal signal.

Download Full Size | PDF

5. Discussion

For use with Er-doped femtosecond lasers, thermally annealed Fe-implanted InGaAsP compares favorably to many of the materials proposed up to now (see Table 1). In fact, 20- to 30-fold improvements in resistivity over ultrafast Fe-implanted In_0.53Ga_0.47As reported by Carmody et al. [22] is quite remarkable. Both were implanted with MeV Fe ions at similar fluence and annealed at similar temperatures.

Table 1. Summary of Results Reported for Ultrafast Photoconductive Layers Grown on Semi-Insulating InP Which Were Used for THz Emission

View Table

The quaternary band gap energy E_g is only larger by 0.04 eV compared to the ternary value, which rules out any large band gap-related intrinsic resistivity enhancement with respect to Fe-implanted In_0.47Ga_0.53As. It is not clear whether the phosphorous present in the quaternary plays a critical role in forming Fe defect complexes. The multiple-energy cold implantation process is likely responsible for the improvement and clearly shows free carrier compensation as low as the 10¹³ cm⁻³ level. In a previous annealing study on n-doped In_0.47Ga_0.53As implanted with Fe at 77 K by Subramaniam et al. [27], the resistivity reached 950 Ωcm with very similar fabrication parameters.

The ion beam damage from high fluence MeV Fe implantation takes about a day to complete on a 75 mm wafer quarter, due to nA ion current levels of the Fe⁺ source. The 1/e absorption length (at 1550 nm) of the Fe-implanted quaternary is 2.1 μm. Therefore, the existing layer thickness could be increased for photoconductive efficiency improvements, but at the cost of lengthened implantation and lower resistance between the electrodes. Strong band tails, resulting from cold implantation with high Fe⁺ fluence, are helping to absorb the spectral content of femtosecond pulses beyond the primary InGaAsP band edge wavelength. It may indicate that similar physics found in highly disordered and amorphous semiconductors is indeed happening (i.e., multiple conduction involving extended states, tail states and hopping transport) [35]. Additional work has to be undertaken in order to better understand the origin of the material tail states and defect states along with their influences on carrier transport, carrier lifetime, and on THz emission.

6. Conclusion

We investigated the use of quaternary InGaAsP alloys in fabricating photoconductive THz emitters. Epitaxial InGaAsP layers (with PL wavelengths at 1565 and 1575 nm) underwent high fluence Fe ion implantation at 83 K. They were then processed by rapid thermal annealing and characterized. Interesting optoelectronic properties occurred at annealing temperatures between 500 °C and 600 °C. We report a high resistivity of about (1200 – 2500) Ωcm, a Hall mobility of 400 cm²V⁻¹s⁻¹, and a photocarrier trapping time of (0.3 – 3) ps. At 1550 nm, the optical absorption coefficient is 4700 cm⁻¹. Fe-implanted InGaAsP dipole antennas were fabricated for a THz-TDS validation. They were excited at 1550 nm with 250 fs pulses. The emitted spectral range exceeded 2 THz and had 25 dB of peak-to-noise ratio. Post-growth processing with cold, high fluence, Fe implantation was key to produce InGaAsP-based THz devices with good emitter characteristics.

Quaternary alloys allow to adjust the offset of their optical absorption edge to the laser emission wavelength. Implantation damage makes femtosecond pulse absorption possible beyond the primary PL wavelength. Besides 1.55 μm, other wavelengths can be exploited for THz technologies. Work is in progress to characterize the MeV Fe implantation process on a quaternary with E_g = 0.95 eV, which could be used as a photoconductive material for Pr-doped fiber lasers sources operating at 1.3 μm or Yb-doped fiber lasers sources operating around 1.0 μm.