In recent years, it has been demonstrated that irradiation by intense ultrashort laser pulses leads to the enhancement of material optical and electronic properties [1–3]. This technique allows the realization of new electronic surface states as well as a broad variety of micro- and nano-structured texturization [4,5] depending on the femtosecond pulse-train characteristics or the presence of specific gas compounds. For example, hyperdoping of silicon has been achieved with nonequilibrium concentration of dopants in combination with controlled surface texturing allowing close to 100% absorption over a wider spectral range than the intrinsic material [6,7]. Similar results on gallium arsenide (GaAs) have recently been reported, where femtosecond laser ablation has shown strong absorption extended below the bandgap under low-light, broadband, continuous-wave illumination [8]. For the latter, the enhanced properties arise from the creation of micrometer-scale periodic ripples, with the creation of surface plasmon polaritons (SPPs) as a possible explanation for this surface patterning [9,10]. This periodic pattern acts as an antireflection layer, thus leading to enhanced photoabsorption [8,11]. The ability to manipulate doping, photoabsorption, or carrier lifetimes in materials via this ultrafast ablation technique provides a new approach to engineering optoelectronic devices with enhanced performance, e.g., photovoltaics or terahertz (THz) devices.

Currently, the development of cost-effective radiation sources and detectors in the THz range is an active field of research. For example, photoconductive devices have allowed THz time-domain spectroscopy (TDS) to become a very powerful tool to investigate materials or chemical compounds as well as for pharmaceutical and security applications [12–15]. With the emergence of a compact commercial THz-TDS system, one focus is the realization of photoconductive emitters and detectors based on low-cost and high-efficiency substrates. For this purpose, methods improving the photoabsorption, emitted THz spectral bandwidth, or reducing the carrier lifetime of GaAs have been investigated, such as implantation of ions [16], chemical passivation [17], or epitaxial growth techniques [18]. However, most of these methods are not time or cost efficient. Alternately, the femtosecond-laser-ablation technique provides a fast and cheaper alternative for the realization of materials with engineered optoelectronic properties for THz applications.

In this Letter, we demonstrate and study the functioning of a THz photoconductive antenna device processed on femtosecond-laser-ablated GaAs and compare it to an antenna fabricated on the same semi-insulating (SI)-GaAs, without the laser ablation. We observe counter-intuitive phenomena; namely, for high optical excitation, the ablated device generates more efficient THz, despite exhibiting a lower photocurrent and lower carrier mobility. On the other hand, for low optical excitation, the ablated device shows poorer THz emission. To understand this behavior, we study the I–V characteristics in conjunction with the emission properties, as well as the lifetime and photoconductivity of the carriers in the two devices using an optical-pump THz probe (OPTP) [19]. We also measure the increase in photoabsorption and doped carrier concentrations in the two devices resulting from the ablation process. For the high-photoexcitation regime, we find that increased photoabsorption in the ablated device results in greater emission, but the very short carrier lifetimes result in lowered photocurrent measurements. In the low-photoexcitation regime, we find that the presence of doped carriers arising from the ablation process diminishes the THz generation due to free carrier absorption.

Figure 1(a) is a photo of a fabricated photoconductive bowtie antenna composed of two gold electrodes with a 50-μm gap. The surface of a second identical antenna is irradiated with a femtosecond pulse laser in order to ablate the surface as shown in Fig. 1(b). The laser system used for ablation delivers 120-fs pulses centered at 800 nm at a 1-kHz repetition rate. The surface ablation is realized with an average power of 5 mW, and the beam scans the surface at 800 μm/s [8]. The ablated pattern presents 20-μm-large grooves defined by the scanning step of the translation stage as seen in Fig. 1(c). The mean depth of the grooves is 2.35 μm. The ablated pattern alternates grooves with 3-μm-diameter grains and periodic ripples owing to the SPP formation [shown in Fig. 1(d)]. The ripples are perpendicular to the grooves and laser polarization. Their length is 5.2 μm, and their width 0.5 μm. The period of the SPP pattern is 740 nm, close to the ablation laser wavelength [20].

 

Fig. 1. Optical microscope images of (a) nonablated GaAs and (b) femtosecond-laser-ablated photoconductive antennas. (c) Scanning electron microscope picture of the area within the red rectangle of (b) showing 20-μm-wide grooves. (d) Zoom within the green rectangle of (c) showing the formation of ripples.

Download Full Size | PDF

We compare the performances of the femtosecond laser-ablated SI-GaAs with the nonablated SI-GaAs when used as THz photoconductive emitters for THz-TDS. The setup is based on a Ti:Sa oscillator system delivering 45-fs pulses centered at 800 nm at a 4-MHz repetition rate. The emission from the antennas is collected through a set of off-axis parabolic mirrors. Electro-optic detection based on a 1-mm-thick ZnTe crystal is used. Both antennas were biased with a 10-Vpp square modulation at 10 kHz. Figure 2(a) presents a comparison between the THz-detected transients generated by the ablated and nonablated GaAs-based photoconductive antennas for three different characteristic optical fluences ($10\mathrm{mJ}/{\mathrm{cm}}^2$, $200\mathrm{\mu J}/{\mathrm{cm}}^2$, and $2.2\mathrm{\mu J}/{\mathrm{cm}}^2$) exciting the 50-μm antenna gaps. For high fluence ($10\mathrm{mJ}/{\mathrm{cm}}^2$), the ablated antenna shows the best performance and 65% improved efficiency. For intermediate fluence ($200\mathrm{\mu J}/{\mathrm{cm}}^2$), ablated and nonablated devices exhibit similar performances. For low fluence ($2.2\mathrm{\mu J}/{\mathrm{cm}}^2$), we observe better emission from the antenna based on regular SI-GaAs.

 

Fig. 2. (a) Terahertz (THz)-generated transients obtained for three different optical excitations ($2.2\mathrm{\mu J}/{\mathrm{cm}}^2$, $200\mathrm{\mu J}/{\mathrm{cm}}^2$, and $10\mathrm{mJ}/{\mathrm{cm}}^2$) from the ablated device (blue) and the nonablated device (red). (b) THz peak-to-peak electric field amplitude extracted from (a). (c) Photocurrent versus optical fluence. (d) Comparison of the THz emission (green) and photocurrent (purple) relative efficiency calculated as the ratio between the ablated and nonablated devices from (b) and (c).

Download Full Size | PDF

Figure 2(b) summarizes the measured peak-to-peak THz field as a function of the pump fluence illuminating the two devices. We clearly identify two regimes—a low photoexcitation regime, where the ablated antenna results in weaker emission compared to the nonablated one, and a high-photoexcitation regime, where the ablated antenna shows enhanced emission over the nonablated one. It also shows that the ablated device exhibits a lower saturation effect that is known to occur for optical fluences above $200\mathrm{\mu J}/{\mathrm{cm}}^2$ [21]. Figure 2(c) presents a comparison of the photocurrent in the same conditions of optical excitation and applied bias, i.e., with the same 45-fs, 800-nm, 4-MHz repetition rate laser system and 10-V bias. In contrast with the THz emission measurements, the regime of high-optical excitation shows a significantly weaker photocurrent for the ablated antenna as compared to the nonablated one. Indeed, one generally expects that the material showing the higher photocurrent would give the best THz emitter. This is summarized in Fig. 2(d), which depicts the relative efficiency obtained from Figs. 2(b) and 2(c), i.e., the ratio of the ablated over the nonablated, showing up to 65% enhancement of the THz emission, despite exhibiting only a third of the photocurrent compared to the nonablated material. To investigate this discrepancy, we perform OPTP measurements on both the samples to measure the lifetime and photoconductivity of the excited carriers.

On Figs. 3(a) and 3(b) are plotted the OPTP differential transmission signals for the (a) ablated GaAs and (b) nonablated GaAs for optical fluence varied from $1.6\mathrm{\mu J}/{\mathrm{cm}}^2$ to $2.2\mathrm{mJ}/{\mathrm{cm}}^2$. In order to cover three orders of magnitude for the optical pump exciting the samples, two different setups based on two different laser systems were used to perform the OPTP measurements. From 1.6 to $12\mathrm{\mu J}/{\mathrm{cm}}^2$, we use a 4-MHz, 650-nJ, 45-fs, 800-nm source laser, thus allowing for a high signal-to-noise ratio even at low optical fluence. In this case, a THz probe is generated via an interdigitated photoconductive antenna as described in Ref. [22] biased with a 10-kHz modulation and 20-Vpp applied voltage. For high fluence, a 1-kHz Ti:Sa (100 fs, 800 nm) amplifier system is employed allowing us to reach up to $2.2\mathrm{mJ}/{\mathrm{cm}}^2$. The THz probe is generated via optical rectification in a 1-mm-thick ZnTe crystal. Figure 3(b) shows the expected response from regular SI-GaAs, i.e., a long carrier lifetime extending above 600 ps. For fluence higher than $200\mathrm{\mu J}/{\mathrm{cm}}^2$, we observe a saturated response resulting in a complete screening of the THz probe by the photoexcited carriers. Figure 3(a) shows the differential THz transmission response from the ablated material measured for the same pump fluences as the nonablated antenna. For low pump fluences, the negative differential transmission of the THz probe is lower in the ablated material, which rapidly becomes comparable to that in the nonablated material at pump fluences of $\sim 200\mathrm{\mu J}/{\mathrm{cm}}^2$ and higher, showing similar saturation effects as the nonablated material. Figure 4(a) conveys this point by plotting the amplitude of the OPTP differential signal versus pump fluence at zero delay for fluences up to $200\mathrm{\mu J}/{\mathrm{cm}}^2$. These results are consistent with the generation of THz transients from the two antennas as discussed in Fig. 2(a)—at low fluences, the OPTP signals and the THz emission are weaker in the ablated antennas, while at higher fluences, the OPTP signals and THz emission are comparable or stronger in the ablated antennas. It also shows that the enhancement of THz emission does not originate from the ablation of the electrodes, as OPTP measurements were performed on non-patterned materials. These measurements suggest that at high pump fluences, the ablated material exhibits higher photoconductivity, which we attribute to the higher photoabsorption in the ablated materials acting as an antireflection layer as previously reported in Ref. [8]. This is also confirmed by FTIR measurements in these specific samples. Figures 4(c) and 4(d) present the transmittance in the mid-IR range (0.1–0.8 eV) and NIR (from 1.1 to 1.4 eV), respectively. Whereas the nonablated GaAs shows a below bandgap flat transmission of about 50%, the ablated material presents significantly lower transmission (for all energies ranging from 0.2 to 1.4 eV), as well as a prominent transmission dip in the mid-IR, indicating the previously reported presence of below bandgap states and higher photoabsorption originating from the ablation process [8,23,24].

 

Fig. 3. OPTP negative differential transmission of (a) ablated GaAs and (b) nonablated GaAs for pump fluences varied from $1.6\mathrm{\mu J}/{\mathrm{cm}}^2$ to $2.2\mathrm{mJ}/{\mathrm{cm}}^2$.

Download Full Size | PDF

 

Fig. 4. (a) OPTP negative differential transmission at zero delay versus pump fluence extracted from Fig. 3 for the ablated (blue) and nonablated GaAs (red). (b) Fast decay time component obtained from bi-exponential decay function fit of the OPTP signals for the ablated GaAs. (c) Comparison of the transmittance between the ablated and nonablated GaAs in the mid-IR range and (d) in the near-IR range.

Download Full Size | PDF

We also observe from Fig. 3(a) that the dynamics of the photocarriers is given by a bi-exponential decay rate, and the ablated material exhibits much shorter carrier lifetime than the nonablated one. Figure 4(b) presents the fast component of the decay time obtained by fitting the OPTP differential signals of Fig. 3(a) with a bi-exponential decay function. It demonstrates that the decay is shorter as the pump fluence increases, with the decay time decreasing from 12 to 3 ps as the pump fluence varies from 1.6 to $32\mathrm{\mu J}/{\mathrm{cm}}^2$. For higher fluence, as saturation occurs owing to complete screening of the THz probe, fitting is not reliable. At high pump fluences, the short carrier lifetimes in the ablated materials explain the perceived contradiction between the THz-generation properties and the photocurrent measurements for the two antennas (although shorter carrier lifetime does not affect the generated THz amplitude). Indeed, as it is a DC measurement, one can access only the average photocurrent. Therefore, with a long lifetime, photocarriers in nonablated GaAs will contribute more, whereas with a sub-10-ps lifetime, photocarriers in ablated GaAs will show very little contribution. We note that our observation of lower photocurrent in the ablated device with 50-fs, 800-nm pulses does not contradict previous observations [8], which were done under different experimental conditions—CW, broadband excitation.

The low-fluence behavior of the THz emission properties of the two antennas can be understood by noting that the ablation process typically results in an $n$-doped material. Hall effect measurements on the ablated material indicate $n$-doping with a sheet concentration of $1.2\times {10}^{12}{\mathrm{cm}}^{-2}$ and a low carrier mobility of $5{\mathrm{cm}}^2/\mathrm{Vs}$. Thus at low fluences, despite the increased photoconductivity of the ablated material due to greater photoabsorption, the emitted THz transients are absorbed via the native $n$-doped carriers resulting from the ablation process. As one gets to higher pump fluences, with photocarrier densities significantly larger than the doped carrier concentration, one begins to observe the effects of improved photoabsorption via enhanced THz generation in the ablated antennas.

In conclusion, we demonstrate the first THz photoconductive antenna based on femtosecond laser-ablated GaAs with up to 65% improved THz-generation efficiency compared to SI-GaAs antennas. Our results are explained by measuring and understanding the impact of femtosecond-laser-ablation on the optoelectronic properties of SI-GaAs. Further optimization and directions of study involve the orientation of the ablated grooves with respect to the electrodes, the polarization of the optical excitation, or other parameters involved in the ablation process. The surface structure of the ablated device determined by ablation parameters such as wavelength, power, pulse duration, and scanning speed can also be optimized [9,10]. The ripples seen in our device correspond to the wavelength of 800-nm ablating pulses, and are subwavelength compared to THz wavelengths. Thus, we estimate that variations in wavelength of the ablating laser over the visible and NIR will not significantly alter the performance of the device from this perspective. Overall, we show that the femtosecond-laser-ablation process provides a powerful tool to engineer material properties such as doping, photoabsorption, and carrier lifetime, thus allowing for a new route toward cost-effective, efficient THz devices and optoelectronic materials.