1. Introduction

The generation of coherent terahertz radiation from photoconductive antenna (PA) has attracted considerable interest since it is a way to reach the intermediate terahertz frequency range. The best terahertz performance is achieved by photoconductive antenna excited by ∼0.8μm optical pulses and made from low-temperature-grown (LTG) GaAs material, because this material associates both subpicosecond carrier lifetime and high resistivity [1]. The use of a lower-bandgap semiconductor, such as In_0.53Ga_0.47As, allows cheap, compact and turnkey terahertz spectroscopy setups based on erbium fiber (Er:fiber) lasers, which can produce sub-picosecond pulses at a central wavelength λ=1.55 μm. Moreover, as the bandgap of the In_0.53Ga_0.47As is smaller than the GaAs one, In_0.53Ga_0.47As semiconductor shows a lower Γ-valley-effective mass which result in a higher electron mobility and greater terahertz power is expected [2]. Photoconductive antennas based on Fe-implanted or heavy ion-irradiated In_0.53Ga_0.47As excited by 1.55μm laser pulses have been investigated and their efficiencies either as terahertz emitter [3,4] or as terahertz detector [5] have been demonstrated. However, whereas ionic irradiation or implantation is an efficient method to reduce in a controlled way the carrier lifetime in semiconductor layers, only few works [6] focus on the study of how the carrier lifetime impacts on the emitted terahertz signal characteristics in In_0.53Ga_0.47As PA. Such investigation turns out to be crucial for applications as carrier lifetime strongly influences the spectral distribution and the power of radiation emitted by PA [7–10]. In this letter, we present a detailed study of the effect of the carrier lifetime on the terahertz signal characteristics emitted by PA made on Br⁺-irradiated In_0.53Ga_0.47As material; we investigated four In_0.53Ga_0.47As emitters with different carrier lifetimes, namely > 1ns, 4.2 ps, 0.7 ps and 0.3 ps. The variations of the temporal waveforms and the average radiated powers for various carrier lifetimes are experimentally analyzed and compared to predictions of analytical models of charge transport. The spectral bandwidth is found to increase with the decrease of the carrier lifetime on the emitter. Improvement in the average power of the emitted terahertz radiation is also observed for emitters with short carrier lifetimes as a consequence of the increase of the maximum possible bias voltage with the increase of the ion irradiation dose. The power radiated by ion-irradiated In_0.53Ga_0.47As PA excited by 1550 nm wavelength optical pulses is compared to that emitted by similar LTG-GaAs PA excited by 780 nm wavelength optical pulses.

2. Devices

Undoped 1-μm-thick n-type In_0.53Ga_0.47As layers were epitaxially grown by gas-source MBE on semi-insulating InP:Fe substrates. A mesa etching process was used to define In_0.53Ga_0.47As absorbing area of 82×14 μm² and 7×14 μm² on the InP substrate for the emitter and the detector, respectively. The layers were then irradiated by 11 MeV heavy ions (Br⁺) at irradiation doses from 4×10¹⁰cm^-2 to 1×10¹²cm^-2. Furthermore, a thin cap layer of optical transparent silicon nitride is grown to protect the device from oxidation and to provide an antireflection coating. The electrode patterns were fabricated by metal evaporation and a conventional lift-off photolithographic technique. Then, the antenna structure is located at the center of a 20-mm-long coplanar transmission line consisting of simple coplanar striplines patterned onto the InP substrate. The latter are made of two 5-μm-wide, 0.5-μm-thick Ti/Au strips separated by 80 μm and 30 μm for the emitters and the detector, respectively. For the detector, a gap of 5 μm between the contacts is added.

3. Time domain measurements

In the time domain experiment, 250 fs optical pulses with a repetition rate of 14.3 MHz, delivered by a passively mode-locked fiber laser (Calmar Optcom) operating at 1550 nm were used to excite the emitter and the detector. The optical pump beam, with an average power of 3 mW, was focused on the photoconductive antennas on a spot size of about 10 μm, near the anode of the antennas. High-resistivity Si hyperhemispherical substrate lenses, with a 10mm diameter, were attached back to the emitter and the receiver antennas. The intensity of the terahertz radiation was modulated using a mechanical chopper. The detector was placed 5 cm away from the emitter. The optical probing beam, with an average power of 3 mW, was focused on the PC antennas on a spot size of about 5 μm. The current induced by the probe beam and the terahertz radiation in the detector is amplified and processed with a lock-in digital amplifier. Note that the InP substrate do not absorb the 0.8 eV photons (since the laser fluences are relatively low[11]) and its contribution on the measured waveforms is negligible.

Due to their high initial energy, the irradiating ions are implanted in the InP substrate beyond 3 μm, and uniform damage profiles through the In_0.53Ga_0.47As layer are created, as suggested by calculations using the “Stopping Range of Ions in the Matter” software [12]. The damages in the In_0.53Ga_0.47As layers are only host atom displacements distributed essentially in defect condensates. These defect clusters have deep energy levels acting as efficient capture and recombination centers for free carriers. The lifetimes of electrons in the conduction band have been determined by pump-probe differential transmission experiments. Four In_0.53Ga_0.47As emitters with carrier lifetimes of >1 ns, 4.2 ps, 0.7 ps and 0.3 ps were made by varying the Br⁺ irradiation dose. The carrier lifetime in the detector was 0.3 ps. The signal waveforms emitted from the four In_0.53Ga_0.47As PA are shown in Fig. 1. These results are obtained using an applied bias voltage of 7 V for the long carrier lifetime sample and bias voltages higher than 15V for samples with picosecond carrier lifetimes respectively.

In the approximation of the dipole antennas, the emitted transient field waveforms in the far field approximation is related to the temporal derivative of the transient current: E_THz ∝ ∂J(t)/∂t. For each emitter, a main positive peak is observed in Fig. 1, resulting from the ultrafast rise of the surge current by the photocarrier injection and the subsequent carrier acceleration under the bias field of the PC antennas. For the samples with picosecond carrier lifetimes, i.e. the ion-irradiated samples, the main positive peak is followed by a negative peak, attributed to the decay of the current governed by the carrier trapping. As the carrier lifetime decrease, the negative peak is more intense and happens earlier. The transition from unipolar radiation to asymmetrically bipolar and almost nearly symmetrical bipolar transition is a direct consequence of the decrease of the carrier lifetime. The origin of the negative peak observed to occur before the main positive peak is attributed either to the pulse reshaping effect due to the frequency dependent THz beam focus on the detector antenna, which has an effect equivalent to that of a high-pass filter to the THz radiation [13] or to the carrier lifetime in the detector material [14]. Small oscillations are observed after the negative peak and may be the results of resonance effects in the emitting antenna and of plasma type oscillations.

 

Fig. 1. Terahertz radiation waveforms from Br⁺-irradiated In_0.53Ga_0.47As emitters. The solid lines represent the measured waveforms and the dashed lines the calculated waveforms. The carrier lifetime reported on each graph is the carrier lifetime extracted from optical pump-probe differential transmission measurements.

Download Full Size | PDF

To correlate these experimental evidences of the effect of the carrier lifetime on the emitted waveforms to theoretical predictions, the detected THz waveforms were fitted by the analytical expression of the detected photocurrent given by Duvillaret et al [15]:

with 1/$\tilde{\tau }$_em = 1/τ_em + 1/δτ_em and τ_em, τ_rec, τ_las, δτ_em are the carrier lifetime in the emitter, the carrier lifetime in the detector, the laser pulse duration and the carrier collision time respectively. This model is based on the assumption that the free-carrier relaxation in both emitter and detector is governed by a single exponential decay law. Since Fe:InP material shows dispersion in the terahertz domain, a broadening of the pulses due to their travel in the Fe:InP substrates of the two antennas (the terahertz beam propagates through a total thickness of 0.7mm of Fe:InP) must be accounted. The detected photocurrent is then calculated using the convolution integral of $j\left(t\right)=\int dt\prime j_{\mathrm{rec}}\left(t\prime \right)\exp \left(\frac{-{\left(t-t\prime \right)}^2}{2\Delta t}\right)$ with Δt = Δnl/c ≈ 200fs since we measured, by time domain terahertz spectroscopy, the maximum variation of refraction index Δn in Fe:InP in the considered frequency range to be approximatively 0.09. The saturation effects are not taken into account as the optical excitation fluences involved in these measurements are relatively low. Note that this model does not integrate high pass and low pass filter to account for the effect of the detector size.

The dotted lines are the waveforms calculated with τ_{re c} = 330 fs, τ_las = 250 fs and δτ_em = 180 fs[16]. We did not compare the absolute amplitudes since they are sensitively dependant on the critical positioning of the high-resistivity Si hyperhemispherical lens on the detector. The amplitude of the waveforms has thus been determined to fit the amplitude of the positive peak. The carrier lifetime of the emitters was then the only remaining adjustable parameter in the fit. The best fit is obtained for carrier lifetime values of 0.3ps, 0.9 ps, 3.9 ps which are very close to the values measured by optical pump-probe differential transmittance experiments. The amplitude of the negative peak and the following rising edge are related to the carrier lifetime in the emitter. Whereas the experimental and theoretical amplitudes of this negative peak are well matched, the extra oscillations, which are superimposed to the experimental waveforms during this rising stage, lead to a discrepancy between theoretical and experimental waveforms. Nevertheless, the overall shapes of the measured waveforms and their dependence on the carrier lifetime in the emitter are well described by the analytical model.

Figure 2 displays the normalized Fourier transform amplitude spectra of the temporal waveforms. Just as the temporal behaviour depends on the carrier lifetime in the emitter, the spectral peaks depend on these carrier lifetimes, being shifted to higher frequency when decreasing the carrier lifetime. The maximum of the spectrum is shifted from a frequency inferior to 0.05 THz for the un-irradiated In_0.53Ga_0.47As PA to a frequency of 0.38 THz for the most irradiated In_0.53Ga_0.47As PA (1×10¹²cm^-2).

 

Fig. 2. Normalized spectra of the terahertz waveforms for emitters with different carrier lifetime. Inset: Frequency of peak emitted terahertz power as a function of carrier lifetime. The experimental data are represented by triangles and the values extracted from the model by solid line.

Download Full Size | PDF

The inset of Fig. 2 shows how the frequency at which the terahertz emission peaks when varying the carrier lifetime, and as one can see, theory and experiment agree fairly well. Therefore we have shown that higher bandwidth terahertz emission is obtained with the decrease of the carrier lifetime, consistently with theoretical expectations based on Duvillard et al’s model.

5. Time integrated measurements

In order to confirm that the carrier lifetime of the emitter can be adjusted to reach a specific terahertz spectral range, absolute terahertz powers emitted by the PA devices have been investigated. The terahertz radiation was detected by a 3 mm-diameter silicon diamond composite bolometer, located just in front of the Si lens placed on the back of the PA emitters. The optical sensitivity of the bolometer is 4.6 10⁵ V/W (included preamplifier gain). The PA are excited by 200 fs optical pulses at a central wavelength of 1550 nm at 80 MHz repetition rate produced by an optical parametric oscillator pumped by a mode-locked Ti:sapphire laser. The mode-locked Ti:sapphire laser delivered 200 fs optical pulses at a central wavelength of 780 nm. For these measurements, the In_0.53Ga_0.47As PA devices were operating at their maximum possible bias voltage, which remains below the electric breakdown and also below the threshold for thermal runaway. The maximum bias voltages were 16V, 28 V, 40 V and 60 V for the samples with carrier lifetimes of >1 ns, 4.2 ps, 0.7 ps and 0.3 ps respectively and were found to be essentially proportional to the dark resistivity of the ion-irradiated In_0.53Ga_0.47As layers. Previous works have shown that the dark resistivity of the In_0.53Ga_0.47As layer increases with the irradiation dose as the result of the degradation of Hall electron mobility due to scatterings on neutral defects [17]. Figure 3 represents the dependence of the radiated power on the incident laser power. The maximum achieved power of 0.8 μW is delivered by photoconductive antennas having 0.7 ps carrier lifetime. The emitted terahertz power increases with the increase of the ion irradiation dose until the lifetime is reduced down to 0.7 ps. However, at low incident optical power, experimental and theoretical works have shown that the terahertz power decreases within less than a factor 3 with the decrease of the carrier lifetime from nanosecond down to sub-picosecond values [2, 9]. Moreover, the degradation of the photo-carrier mobility induced by ion irradiation is expected to also reduce the terahertz power [2]. As the terahertz power is proportional to the square of the applied electric field, the increase of the terahertz power observed in Fig. 3 is then attributed to the strong increase of the maximum possible bias voltage. For PA with 0.3 ps carrier lifetime, the decrease of the radiated power is attributed to carrier recombination that occurred before the end of the optical pulse as the carrier lifetime is comparable to the pulse duration. The performances of these ion-irradiated In_0.53Ga_0.47As PAs have been directly compared to those of a typical LTG GaAs antenna by measuring the delivered terahertz power in the same experimental set up and with the same calibrated bolometer. The LTG GaAs device was a similar 80-μm-wide coplanar stripline antenna deposited onto LTG GaAs having a carrier lifetime of ∼ 1 ps. The positioning of the incident optical beam and of the Si lens was carefully optimized for the two measurements, and the actual maximum values are estimated to be reached within a few tenths of percent. The power delivered by the LTG GaAs PA biased at 60 V and excited by optical pulse trains of 36 mW average optical power at 780 nm central wavelength was 0.47 μW, in agreement with the measured powers reported by other group [9]. The terahertz powers emitted by ion-irradiated In_0.53Ga_0.47As PA excited by 1550 nm optical pulses, which reached a maximum value of 0.8 μW, are then comparable with or greater than that emitted by LTG-GaAs PA excited by 780 nm optical pulses.

 

Fig. 3. Bolometer output measured as a function of the average laser power driving the emitter for the photoconductive antennas with different carrier lifetime. The solid curve is the theoretical curve fitted to the data.

Download Full Size | PDF

For all samples, the radiated power increases quadratically at low incident optical powers and saturates at higher incident optical powers. This saturation is attributed to the screening of the applied bias field by the radiation field and the space-charge field, which contribute to the collapse of the total electric field acting on the carriers at high carrier density. For small optical size excitation (small-aperture emitters), models based on Monte Carlo method have shown that the space charge induced by separating carriers is the dominant contribution to the screening of the applied electric field [18,19]. Therefore, to interpret the observed saturation of the average radiated powers with incident optical powers, we considered the following simple model. As the pulse duration is smaller than the carrier lifetime (except for the most irradiated sample), the concentration of photoexcited electron-hole pairs is given by n ₀ = α(1-e^-αl)F_las/hν where F_las is the fluence of the incident optical pulse beam. Whatever the incident optical power, the maximum of the transient current is achieved at the end of the pulse duration τ_laser and is thus given by j _max = -en ₀ v(τ_las), neglecting for the sake of simplicity the minor current contribution from the holes. The photoexcited carrier density is high and then scattering works very efficiently to restore thermal equilibrium among carriers in characteristic time less than the optical pulse duration. The average velocity is then expressed by v(τ _max) = μE_loc(τ_las). The local field can be estimated from the Poisson’s equation when considering that during the time τ_las, the carriers have propagated a distance l=vτ_las. The resulting local electric field is given by E_loc(τ_las) = E_bias - qn ₀ v(τ_las)τ_las/ε. As hole velocity is much lower than electron velocity, the hole contribution to the space charge field can be neglected. The average velocity is then written by v(τ_las) = μE_bias/(1 + qμn ₀ τ_las/ε) and the resulting peak current is given by j _max = -en ₀ μE_bias/(1 + qμn ₀ τ_las/ε). As the shape of the waveform is not affected by the saturation effects, the peak amplitude of the terahertz waveform can then be expressed as $E_{max}\propto \frac{F_{\mathrm{laser}}}{F_0+F_{\mathrm{laser}}}$ with F₀ defined by εh ν/(qα(1-e^-αl)μτ_las and the average terahertz power as $P_{\mathrm{THz}}\propto {\left(\frac{F_{\mathrm{laser}}}{F_0+F_{\mathrm{laser}}}\right)}^2$. Note that this theoretical law is very close to the one obtains in large-aperture photoconductive antenna for which screening by the radiated field only is considered. The excellent fit of the experimental data by this theoretical law contributes to validate our simple analysis. Assuming that the only irradiation fluence-dependent parameter of F ₀ is the photoexcited carrier mobility, from the fit, a photoexcited carrier mobility ratio of 3 between the un-irradiated sample and the sample with 0.3 ps carrier lifetime was estimated. This ratio is clearly lower than the Hall carrier mobility ratio of 23 measured between these two samples. The difference is explained because Hall mobility is the low electrical field mobility obtained with a carriers in low density (10¹⁵cm^-3), whereas the mobility involved in these measurements results from higher electrical field and from a high concentration of photocreated carriers (few 10¹⁷cm^-3). The irradiation process affects much more the electron Hall mobility due to the defect scattering than the effective mobility of photoexcited carriers in the early stage after their creation. Moreover, as the ion irradiation process is found to increase the maximum possible bias voltage of In_0.53Ga_0.47As material, the ion irradiation of In_0.53Ga_0.47As layers results in an increase of both the bandwidth and the maximum power delivered by PA emitters.

 

Fig. 4. Emitted terahertz power as a function of the carrier lifetime at 0.1 THz (circle), 0.38 THz(triangle), 0.84 THz (diamond) and 1.2 THz (square) computed from time domain measurements with Bolometer detector normalization.

Download Full Size | PDF

This tendency is illustrated by Fig. 4 showing the radiated power as a function of the carrier lifetime in In_0.53Ga_0.47As PA, computed for different frequencies from time domain measurements with Bolometer detector normalization. The increase of the radiated terahertz power with the decrease of the carrier lifetime at high frequency has important practical implications for terahertz emitter design: the carrier lifetime of the emitter can be adjusted to reach a specific terahertz spectral range.

Conclusion

The influence of the carrier lifetime in Br⁺-irradiated In_0.53Ga_0.47As photoconductive antennas excited by 1550 nm wavelength femtosecond optical pulses on the characteristics of the emitted terahertz signal has been investigated. We demonstrate that the spectral bandwidth increase when decreasing the carrier lifetime and that the average power is improved in short carrier lifetime devices as the ion irradiation process is found to increase the maximum possible bias voltage of In_0.53Ga_0.47As material. The maximum power radiated by ion-irradiated In_0.53Ga_0.47As photoconductive antennas excited by 1550 nm wavelength optical pulses is measured to be 0.8 μW. This value is comparable with or greater than that emitted by similar low temperature grown GaAs photoconductive antennas excited by 780 nm wavelength optical pulses. This study has important practical implications for the design of ion–irradiated In_0.53Ga_0.47As photoconductive antennas used as terahertz emitter.