1. Introduction

Dual-color infrared (IR) detections are crucial for discrimination of objects and imaging under varying atmospheric conditions, and thus attracted significant research interest. Beside the fact that different wavelength regions are associated with different practical applications, detecting an object’s IR emission at multiple wavelengths is beneficial to remove background clutter and sunlight interference information and leave the useful information of the target object, which greatly improves the accuracy of detection [1–4]. In these regions, quantum well infrared photodetectors (QWIPs) have been proved to be one type of important third-generation infrared photodetectors, with a number of potential advantages including the use of standard manufacturing techniques based on mature III-V group materials growth and processing technologies, high thermal stability and intrinsic radiation hardness [5,6]. However, in spite of all achieved developments in multicolor detection for QWIPs [7–10], there are several important deficiencies such as bias-dependent wavelength detection, difficulties in simultaneous detection with multi-wavelength [11]. As an alternative solution, quantum cascade detectors (QCDs) seem to be very competitive. In 2010, A. Rostami et al. have proposed and numerically investigated a dual-color detector based on quantum cascade structures which can detect two different wavelengths through two independent output current paths simultaneously, and introduced two different structures for the proposed idea [11]. In 2012, S. Sakr et al. have designed and demonstrated a room temperature dual-band GaN-based QCD exhibiting a photoresponse at 1.7 and 1 µm wavelength simultaneously [12]. In 2016, Liang Li et al. proposed and numerically investigated a surface plasmonic coupled mid-long-infrared two-color quantum cascade detector, and the detection wavelengths are peaked at 4.4 µm and 9.0 µm [13]. Benefiting from the asymmetric intersubband energy level structure, QCDs work at zero bias, and show advantages of very low noise level and negligible thermal load, which makes QCDs promising in large focal plan array and small pixel applications [14–18]. Moreover, QCDs can operate at higher temperatures than QWIPs theoretically, which is very attractive [14–19]. QCDs have become a major competitor for multicolor detection. However, QCD has its own obstacles in optical coupling mechanism. According to the intersubband (ISB) selection rule, infrared absorption of QCDs is possible only when the electric field vector of the radiation has a component perpendicular to the quantum well layers. Most work on QCDs has hitherto been directed to using a 45° polished edge to couple radiation into the absorption zone. Yet for most technical applications, planar device geometry and large area illuminating are essential. Recent years, antenna-coupled microcavity structures have been actively investigated as tools to harness the interaction between the metal-semiconductor interface and light confined in extremely sub-wavelength regions of the space, which is a promising solution to enhancing the performance of detection with normal incidence [20–23].

In this paper, a mid-wave/long-wave dual-color infrared QCD was constructed and enhanced by an array of metallic nano-antennas. As the electromagnetic field was compressed into size and distribution optimized sub-wavelength microcavities where the wavelength-selective photodetection takes place, the antenna array improves the responsivity of the device. Meanwhile, owing to the much larger photonic area than the electrical area, the Johnson noise limited detectivity has been greatly increased. These results pave the way to the development of multicolor detections.

2. Methods

Our QCD wafer was grown on a semi-insulating (SI) InP substrate utilizing molecular beam epitaxy (MBE) method. 5 periods of active region with response wavelengths of 5.7 µm (217 meV) and 9.6 µm (128 meV) were sandwiched between a 150 nm bottom contact layer and a 50 nm top contact layer. The active region was designed according to an optimized coupled miniband diagonal-transition structure, each period of the active region consists of 8 In_0.52Al_0.48As/In_0.53Ga_0.47As coupled wells. The layer sequence of one period of structure in nanometers is as follows: 4.5/6.25/3.0/4.9/2.8/3.95/2.5/3.4/4.4/3.25/4.6/3.25/4.9/3.1/1.8/6.7, where the In_0.52Al_0.48As barrier layers are in bold font, the In_0.53Ga_0.47As well layers are in normal font, and the underlined layers are doped to n = 1×10¹⁸ cm⁻³. The band structure is self-consistently calculated via Schrödinger-Poisson equations and presented in Fig. 1. The dual-color detection is achieved by a vertical-transition in the active quantum well and a diagonal-transition across the adjacent quantum wells [24,25].

 

Fig. 1. Calculated conduction band diagram and relevant wave function of one period of QCD. The optical transitions are indicated by the black arrows. The electrons are excited from the lower to the upper state. From there, they can escape into the extraction region, eventually reaching the ground level of the next cascade and contributing to the photocurrent. The optical transition of the mid wave happens between two energy levels in one well (E₀  →  E_T), while the optical transition of the long wave occurs between two energy levels in two adjacent wells (E₀  →  E_mini). The calculated ratio of the theoretical relationship between the spectral responsivities due to the vertical transition at 217  meV and the diagonal transition at 128  meV is 0.67.

Download Full Size | PDF

For achieving an optimum plasmonic and QCD interaction, the spectral matching of plasmonic resonance to QCD absorption peak and the spatial matching of plasmonic field region to the QCD absorption region must be satisfied. The antenna-coupled microcavity has the property of extraordinary optical transmission because of the generation of surface plasmons (SPs), and the peak transmission wavelengths are given by λ = 2sn_eff, where n_eff = 3.3 is the effective refractive index, and the patch antenna size (s) corresponding to the absorbing wavelength can be deduced.

Based on the analysis above, the device was designed and fabricated, which is shown in Fig. 2. An array of metal-semiconductor-metal microcavities, which provides a strong sub-wavelength electric field confinement and acts as antennas, was fabricated. The device consists of two 20×20 periodic square microcavity arrays nested within each other, whose patch edge sizes are s₁= 1.5 µm and s₂= 0.9 µm, and electrically connected by 200 nm wide wires, respectively. According to λ = 2sn_eff, the patch with s = 1.5 µm is therefore in resonance with λ = 9.6 µm, and s = 0.9 µm is in resonance with λ = 5.7 µm. The period of both arrays is p = 5 µm, in the same order of magnitude as the wavelength. The active region has been patterned into an array of square patches by means of electron beam exposure and inductively coupled plasma (ICP) etching, and inserted between the Ti/Au ground plane on the bottom and the metallic array on the top. The connecting wires are in the diagonal direction of the square platform. To achieve effective optical coupling, a thin active region is needed, and that’s why different with most QCDs, our wafer has only 5 period of active region. As the last step of the device fabrication process, the top and bottom metal electrode [Ti (40 nm)/Au (300 nm)] structures were formed using electron beam deposition and lift-off processes, while the top metal electrode was electrically insulated with the bottom metal electrode by a 450-nm-thick SiO₂ layer. After gold wire bonding, the detector chips were then mounted on an oxygen-free copper heat sink and fixed on the cold finger of a well-designed liquid-nitrogen cryostat to perform device characterization.

 

Fig. 2. (a) Scanning electron microscope image of the detector, with a partial enlargement of the electrically connected microcavities. The main parameters of the array are indicated in the diagram. Region A: The two 20×20 periodic square microcavity arrays with different sizes nested within each other. Region B: the top metal electrode. Region C: SiO₂ for electrical insulate. Region D: Ti/Au ground plane. (b) Schematic diagram of a single microcavity structure.

Download Full Size | PDF

The electric field component distribution of SPs for the structure was obtained through a commercial COMSOL Multiphysics finite-element simulations, and presented in Fig. 3(a). The electric field distribution follows a standing-wave pattern, with the node at the center of the square and the antinode at the edges, showing strong field localization in the semiconductor region. This indicates that these microcavities have strong coupling effects on the light with specific wavelengths. The TM mode is perturbed at the corners by the connecting wires, which results in a co-sinusoidal dependence on the light polarization of a normally incident wave in the antenna-coupled device, and this is the reason why the absorption efficient of incident light cannot reach 100%.

 

Fig. 3. Simulation results of the device. (a) The distribution of the vertical component of the electric field |E_z| at the center plane of the microcavity with wavelength of incident light (a.1) λ₁ = 9.6 µm and (a.2) λ₂ = 5.7 µm. (b) Calculated reflectivity spectrum of the device. The absorption peaks at 9.7 µm and 5.7 µm are caused by the microcavities with patch edge sizes s₁= 1.5 µm and s₂= 0.9 µm, respectively, which is in good agreement with our design.

Download Full Size | PDF

In our structure, the microcavity increases the responsivity of the device by enhancing the local field in the thin active region. Meanwhile, the photon collection area of the detector, A_coll, was extended by the antenna effect, being much larger than the electrical area σ = s² of the device. As the photocurrent is proportional to A_coll, whereas the dark current is proportional to σ, there is a substantial reduction of the dark current with respect to the case σ = A_coll for the same number of collected photons. As a result, there is a net increase of the operating temperature of the detector.

To quantify the above detector (called “Patch”) performance, a reference detector called “Mesa”, was processed into a 200-µm-square mesa with light coupling through a 45° polished substrate edge. This comparison revealed the intrinsic photoresponse of the detector. In Fig. 4, we compared the responsivity spectra at 77 K, zero bias for the two configurations, obtained with a Nicolet 6700 Fourier Transform Infrared Spectrometer (FTIR) with a KBr beam splitter and the internal glow-bar source. To obtain the absolute value of the responsivity, the FTIR spectra was calibrated with a blackbody source at 1000°C. The Patch device showed a peak responsivity of 4.1 mA/W at the long wave (10.4 µm), and 0.6  mA/W at the mid wave (5.8 µm), while the Mesa device showed a peak responsivity of 0.45  mA/W at the long wave (9.9 µm), and 0.22  mA/W at the mid wave (5.7 µm), respectively. In other words, the Patch device showed a 9.1 times enhancement of the responsivity at the long wave and a 2.7 times enhancement at the mid wave. The drift of the peak position at the longwave is caused by the dispersion effect. The comparison of different peak responsivity illustrates that our structure is more advantageous for long-wave devices. There are two reasons. One is that there is a higher tolerance for wavelength drift at the long wave, as shown in Fig. 3(b), and the other is that the long wave corresponds to a larger square platform size, and the perturb caused by the same size connection wire is relatively smaller.

 

Fig. 4. Responsivity spectra at 77 K at zero bias for Patch under normal incidence and Mesa under 45-degree angle incidence, obtained with a Nicolet 6700 Fourier Transform Infrared Spectrometer (FTIR). To obtain the absolute value of the responsivity, the FTIR spectra was calibrated with a blackbody source at 1000°C. The observed peak wavelength is in agreement with theoretical predictions. The inset shows the dark current- voltage (IV) characteristic at 77 K which is measured with a Keithley 2601A Source Meter.

Download Full Size | PDF

The inset shows the dark current-voltage (IV) characteristic at 77  K, measured with a Keithley 2601A Source Meter. Compared with the Mesa device, the dark current of the Patch device is reduced by two orders of magnitude, thanks to a high resistance caused by the reduced electrical area, which is important to take advantage of a low Johnson noise to increase the operating temperature and get high detectivity.

The peak responsivity $\textrm{R}_{\textrm{peak}}$ is plotted in Fig. 5 for the Mesa reference and for the Patch device. Furthermore, the detector performance is best evaluated through the Johnson noise limited detectivity ${\textrm{D}_{\textrm{J}}^{\ast}} = \textrm{R}_{\textrm{peak}} \sqrt{\frac{{\textrm{R}^0}\textrm{A}}{4k_{B}T}} $ and plotted in Fig. 6, where $\textrm{R}^0$ is the resistance of the detector at zero bias, A is the area of the detector, ${k_B}$ is the Boltzmann constant, T is the operating temperature of the detector. At 77  K, the Johnson noise limited detectivity of Patch equals 1.8×10⁹cm  Hz^1/2/W (Jones) at long wave (10.4 µm) and 2.6  10⁸cm·Hz^1/2/W (Jones) at mid wave (5.8 µm), while the Johnson noise limited detectivity of Mesa equals 5.2  10⁷cm·Hz^1/2/W (Jones) at 9.9 µm and 2.6  10⁷cm·Hz^1/2/W (Jones) at 5.7 µm, respectively. Compared with the reference Mesa device, the Johnson noise limited detectivity of Patch showed an obvious enhancement of the detectivity by nearly two orders of magnitude at the long wave and an order of magnitude at the mid wave compared with the Mesa device.

 

Fig. 5. Peak responsivity of QCD devices at various temperatures for Patch under normal incidence and Mesa under 45-degree angle incidence.

Download Full Size | PDF

 

Fig. 6. Dependence of Johnson noise limited detectivity of the Patch device and Mesa reference on temperature.

Download Full Size | PDF

The Mesa device could be characterized only up to 140  K because the photocurrent becomes undetectable at higher temperatures, while the Patch detector could be characterized up to 180  K at the long wave and 140  K at the mid wave. The reason why the mid wave disappears first with the increase of temperature is that our device is more advantageous for longer wave detection, as mentioned above. It is worth noting that for our device, even if the operating temperature is increased to 180 K, the responsivity and Johnson noise limited detectivity is still comparable to the Mesa device in the order of magnitude at 77  K.

3. Conclusion

In this paper, a dual-color QCD enhanced by dual-size antenna-coupled microcavity has been proposed and fabricated. An optimized coupled miniband diagonal-transition active region structure has been introduced to achieve the dual-color detection. Thanks to a much larger effective absorption area than the electrical area, which reduces the dark current, the Johnson noise limited detectivity has been increased by two orders of magnitude at the long wave and an order of magnitude at mid wave. Meanwhile, the operating temperature of the device has also been improved. In our design, the coupling wavelength of the antenna-coupled microcavity can be conveniently tuned to match the targeted intersubband transition energy by optimizing the size of patches, which brings important degrees of freedom in future design of infrared detectors. This structure improves the utilization efficiency of the incident light, in particular for long wave devices, as longer wavelength corresponds to larger size, and the perturb on the diagonal of the connecting wire is relatively slightly. This structure can be extended to a multicolor detector which can be considered as a spatial wavelength-to-current optoelectronic multiplexer. In addition, in the future, by adjusting the connection way of the electrodes, we can separate the responses of the two wavelengths. And then the detection wavelength can be switched and selected according to the actual application.