1. Introduction

The mid-infrared spectral region is the portion of the electromagnetic spectrum, that is highly suitable for spectroscopy of gases and liquids by addressing their fundamental bending and stretching vibrations [1,2]. It also hosts multiple atmospheric transmission windows, extending to the long-wavelength infrared (LWIR). Especially, the low susceptibility towards atmospheric turbulences and small attenuation in the LWIR [3], together with the fast modulation capabilities of intersubband devices [4,5], make this spectral range highly promising for unlocking novel telecom applications [5,6]. One crucial component in telecommunication receiver systems are suitable high-performance detectors. While HgCdTe-based interband photodetectors (MCTs) are very sensitive, moderate bandwidth (low-GHz range) detectors, they have low saturation thresholds, limiting their application in heterodyne detection. MCTs are in particular not suitable for monolithic integration due to their incompatible material system. In contrast, photoconductive quantum well infrared photodetectors (QWIPs) [7,8], show high responsivities and improved high frequency behaviour (»10 GHz) [9], due to short intrinsic carrier lifetimes. One major drawback of QWIPs is their pervasive dark current, originating from biased detector operation [10]. Quantum cascade detectors (QCDs), demonstrated in 2002, support high-speed and room temperature operation [11,12]. Based on QC laser (QCL) structures relying on intersubband transitions, they allow unbiased, i.e. low dark-current, detection, opening pathways for low-noise applications [13]. Even more important for compact heterodyne detection is their linearity and high saturation intensity [2]. The realization of monolithic devices, integrating active and passive components like lasers, detectors and waveguides on one chip, makes them suitable for sensing and telecommunication applications [14–16]. Extending the telecom-bandwidth to mid-infrared frequencies mitigates bandwidth and availability limitations in current fifth generation (e.g. Long-Term Evolution (LTE)) communication systems, arising from demanding high-speed breakthrough technologies like unmanned mobility, industry 4.0, e-health [17,18] and satellite communication [19]. Moreover, mid-infrared free space optical links have small device footprints, low power consumption and high modulation speeds [20,21] with high connection stability and reliability even under harsh weather conditions including fog and rain. Literature shows that systems operating in the 8-12 $\mu$m atmospheric window are suitable candidates for such systems [22], but limited due to device performance at longer wavelengths and need further improvements. QCL devices covering this wavelength range were already successfully used as atmospherically robust transmitter units [23–26]. The first 9 $\mu$m QCDs operating at room temperature took advantage of the GaAs/AlGaAs material system [5]. We present room-temperature InGaAs/InAlAs ridge QCDs at this wavelength and show an experimental comparison between similar single and multi-period QCDs in our study. The reduction of the number of periods, in order to increase the device responsivity and its impact on the noise behaviour, is an important parameter to optimize QCDs for implementation into monolithic heterodyne detectors.

2. Device design and fabrication

In this work, we demonstrate two different QCDs designed for the target wavelength of 9.124 $\mu$m: 1) a 15-period device with high-specific-detectivity, and 2) a high-responsivity single-period device. The facet-illuminated ridge devices are optimized for low-noise, high-responsivity operation. This geometry is required for the development of a monolithic photonic integrated circuit (PIC) device [27,28]. For monolithical integration of active and passive components, as for example needed for a heterodyne detection system, ridge geometry can be directly implemented. Laser, waveguide, and detector can be fabricated aligned to each other on a single chip, preventing difficult alignment procedures between components. A trade off between the optical active area and the effective electric area has to be found. Small electrical areas lead to high device resistivities and thus improve the signal to noise ratio (SNR) while large optical areas increase the device response. Mesa structures, illuminated from the top-side, show the same size of electrical and optical area while the optical area can be increased through different structures, like photonic crystal cavities or plasmonic lenses [29,30]. Ridge devices, on the other hand, can be fabricated narrow and short, reducing the electrical active area. By incoupling light from the side facet, optimal absorption lengths are ensured even though the optical active area is small compared to mesa-geometries [31,32]. Further, light polarized in growth direction can be directly detected without additional gratings [33], compared to, for example, surface detecting devices [34]. One of the most commonly used material systems in mid-IR and terahertz photonics is InGaAs/InAlAs, lattice matched to InP substrate. We used the same material system in this work grown by molecular beam epitaxy (MBE). The alternating stacking of InGaAs and InAlAs creates a series of quantum wells, forming energy bands through the coupling of the individual energy states. One period is typically referred to as the part of the active region that contains the optical active quantum well together with the extractor/injector sections transporting the electrons from one active region period to the next. The extraction mechanism is a reason why QCDs are highly suitable candidates for data transmission: the lifetime of the typically involved LO phonons is < 1 ps, making QCDs fast devices with very high saturation thresholds [35]. QCDs are typically realized with 5 to 50 periods in their active region, with exceptions like single period devices [36] as in this work. The simulated band structures were obtained via a self-consistent calculation scheme of the eight-band $k \cdot p$ model and the Poisson equation [37,38]. As the former yields the subbands spatial distribution based on the band structure, the latter finds the corresponding Fermi levels with consideration of the dopant densities. The Hartree potentials are calculated as the electrostatic potential of the electron charge densities. Figure 1 shows the simulated band structure of the single period device at room temperature and at zero-bias operation. The optical transition occurring in the active well is indicated by the red arrow, while the current flow through the device is indicated by the black arrows. The active zone is embedded in a waveguide and composed as follows (InAlAs layers in bold, Si-doped layers underlined, in nanometers): 2.5, 7.2, 2.5, 6.6, 2.5, 6.6, 3.5 (2 x $10^{17}$ cm$^{-3}$), $9.4$ (8 x $10^{17}$ cm$^{-3}$), 4.5, 3.3, 4.1, 3.5, 4.1, 4.1, 3.4, 4.4, 3.4, 5, 3, 6.5, 3 (3 x $10^{17}$ cm$^{-3}$). The layer sequence of the active region stack, including the injector/extractor structure of the 15 period device in nanometers is 4.5, 3.2„ 4.3, 3.3, 3.9, 3.5 (1 x $10^{17}$ cm$^{-3}$), 3.9, 3.8, 4.1, 4.4, 3.7, 3.7, 4.7, 5.1, 3.7, 5.3, 2.8 (1 x $10^{17}$ cm$^{-3}$), 9.4 (2 x $10^{17}$ cm$^{-3}$), with InAlAs barrier layers in bold and Si-doped layers underlined. In Fig. 2, a cutout showing one period (between the dashed lines) of the simulated bandstructure of the multiperiod design, is depicted. This stack is repeated 15 times and embedded in a top, and bottom charge balancing region, easing the formation of ohmic contacts. In both designs, the series resistances of the bulk InGaAs/InP interfaces were reduced by the implementation of a contact chirped superlattice structure. Figure 1 and 2 allow a quantitative explanation of how the number of periods used in a design, will finally influence the detector performance and its figures of merit. The responsivity $R$ of a QCD, the figure of merit describing the produced photocurrent per incoupled optical power, is proportional to 1/N, with N being the number of periods [10]. On the other hand, a higher number of periods will increase the device resistance, improving its noise behaviour. This trade-off is the typical challenge in QCD design. The devices used in this work were fabricated as 200-$\mu$m-long and 14-$\mu$m-wide ridges, with a height of approximately 7 $\mu$m. After patterning the ridges with UV lithography and reactive ion etching (RIE), the surface was passivated with SiN using plasma enhanced chemical vapor deposition (PECVD). The devices were electrically contacted through a sputtered layer of titanium as an adhesion promoter, and gold. After cleaving to produce clean facets, the device was soldered to a copper base plate and bonded to a printed circuit board. Figure 3 shows a SEM picture of the cleaved facet of a 15-period detector ridge with indicated layers. The active region can be seen as the rectangle in a slightly brighter grey tone compared to the surrounding cladding layers. The SiN passivation layer can be identified as a dark, nearly black line between the substrate and the gold contact layer. The quality of the cleaved facet is a crucial parameter, determining the effectivity of incoupling the light.

 

Fig. 1. Band structure of the single period design. The current path is visualized by the black arrows while the optical transition is pointed out by the red arrow.

Download Full Size | PDF

 

Fig. 2. Band structure of the 15-period design. The vertical dotted lines indicate the layers of one single period, which is repeated 15 times. The black arrows show the current path while the optical transitions are highlighted by the red arrows.

Download Full Size | PDF

 

Fig. 3. Scanning electron microscope picture of the facet of a 15-period device. The individual layers are indicated.

Download Full Size | PDF

3. Characterization

Both samples were characterized optically and electrically. I-V curves were measured with a Keithley 2601B source-meter for different device temperatures under dark conditions in order to extract the differential resistance at zero bias $R_0$. For the plotted I-V curves we refer the reader to Figures S1 in the supplementary material. The extracted zero bias resistance $R_0=dV/dI$ was plotted on a logarithmic scale as function of the inverse measurement temperature. The slope of the resulting curve in the arrhenius plot, shown in Fig. 4 describes the activation energy of the quantum well structure, and thus its transition energy. We extracted activation energies of $E_{A 15-per}$ = 111.3 meV for the 15-period device and $E_{A single}$ = 126.1 meV for the single period device, respectively. The low room-temperature resistances of 18 $\Omega$ and 264 $\Omega$ of the single period device and the 15-period device, respectively, can be explained by the low transition energy at a wavelength of 9.124 $\mu$m. Using a single period design lowers the resistance by simply reducing the number of subsequent barriers. The spectral characterization of the devices was conducted with an external-cavity QCL in pulsed operation, able to cover a wavelength range from $8-12$ $\mu$m. Absolute responsivity measurements were done with an Alpes Lasers QCL in continuous wave operation at a wavelength of $9.28$ $\mu$m. After the laser beam shape measurement and power calibration with the help of an automatized X-Y stage and an optical powermeter, the spectral responsivity curves were recorded at room temperature using a transimpedance- and a lock-in amplifier and scaled to the absolute values measured at $9.28$ $\mu$m. To examine the free-space detector coupling and calculate the incoupled optical power with high precision, FEM simulations were performed. Figure 5 illustrates the spatial distribution of the optical mode intensity for the two QCD designs extracted with the simulation software COMSOL Multiphysics. The detector ridge structures were modeled according to the waveguide design. The effective mode area ($A_{eff}$) was calculated from the simulations as shown in Eq. (1),

 

Fig. 4. Arrhenius plot: natural logarithm of the differential zero bias resistance plotted as a function of the reciprocal temperature. The slope of the curves is proportional to the activation energy $E_A$.

Download Full Size | PDF

 

Fig. 5. Simulation of the spatial distribution of the optical mode intensitiy inside the waveguide. The active regions are highlighted by the red arrows.

Download Full Size | PDF

 

Fig. 6. Spectral responsivity curves acquired at room temperature. The 15-period device is shown in blue, the single period device is shown in red. The black dashed line marks the target wavelength of 9.124 $\mu$m.

Download Full Size | PDF

4. Conclusion

Concluding our work, we designed, fabricated and characterized two different types of quantum cascade detectors, targeting a wavelength of 9.124 $\mu$m. As expected, the single period device shows a higher peak responsivity of $R_P$ = 0.411 A/W compared to the 15-period device with $R_P$ = 0.111 A/W. Due to the single-period devices very low resistance, it shows a slightly higher noise equivalent power of $NEP_1$ = 83.6 pW/$\sqrt {Hz}$ compared to the 15-period device with $NEP_{15}$ = 80.9 pW/$\sqrt {Hz}$, at room temperature. Device characterization, particularly the normalization of incoupled power and the alignment of the ridge devices, is quite challenging for this geometry. Due to the low resistances of the devices, the electrical measurements required careful data analysis and interpretation. We see that the reduction of the number of periods effectively increases the device responsivity, with the drawback of a decrease in device resistance. Future challenges include the simulation and design of a single-period structure with a higher zero-bias resistance at room temperature in order to improve its noise behaviour and thus its usability in photonic integrated circuits.