Quantum cascade detectors (QCDs), a new type of photovoltaic quantum well infrared photodetector (QWIP), have been presented and made great progress over the past decades^[1,2]. Typically, the wells and the barriers of QCDs are composed of two different semiconductor materials, such as GaAs/AlGaAs, InGaAs/InAlAs, and InGaAs/AlAsSb. In addition, based on bound-to-bound intersubband (ISB) transitions in a built-in asymmetric conduction band structure, QCDs exhibit the characteristics of tailorable operation wavelength, low noise photovoltaic mode, and room temperature operation, which are advantageous over well-established detectors^[3,4]. Meanwhile, as a unipolar photodetector, the subpicosecond electron ISB scattering time makes QCDs have a very high frequency response, thereby providing them with application potential in fields such as free-space optical communication and on-chip sensing^[5–7].

So far, QCDs are able to cover a wide spectral range from the near- to far-infrared spectral region^[8–14]. In the very long wave infrared (VLWIR) region (λ∼14–30μm), QCDs have attracted special attention for their potential applications in astronomical observation, atmospheric monitoring, and intrusion detection fields^[15]. Many quantum designs for VLWIR QCDs have been proposed^[13,16–18] to decrease the dark current noise and improve responsivity, such as the energy ministeps^[17] and twin-well absorption region^[16]. Certain improvement in the device performance was demonstrated, but these structures are mainly narrowband designs resulting in an optimized structure targeting only one specific wavelength. Therefore, a highly robust quantum band structure that is insensitive to wavelength tuning is urgently needed for VLWIR QCDs.

In this Letter, a modular miniband diagonal transition (MDT) band structure strategy is proposed for VLWIR QCDs. This structure enables the QCDs to tune the wavelength in a modular manner. The theoretical simulation shows that the response wavelength of QCDs with different well widths can cover the range of 14–29.5 µm while maintaining high extraction efficiency (>90%) and dipole matrix element (DME>1.38nm). Three devices with different well widths were fabricated with the center wavelengths of 14, 16, and 18 µm, respectively. At 10 K, the detectivities of all the three samples exceeded 1×1010 Jones, with corresponding responsivities of 7, 11, and 2.5 mA/W, respectively. The experimental results are in good agreement with the numerical simulations, which proves the capability of the modular design based on the MDT scheme for wide wavelength tuning in the VLWIR.

In order to obtain a VLWIR QCD band structure that is insensitive to wide tuning, two aspects must be considered. First, the band structure should be modular, so that wavelength tuning can be achieved by adjusting for specific modules rather than for the whole active region. Second, the proposed quantum band structure can effectively address the inherent issues of electron thermal backfill and nonradiative diagonal transition in VLWIR QCD. In response to these points, an MDT structure is proposed, as shown in Fig. 1. The diagonal transition enables optical transitions between two adjacent wells, which overcomes the electron extraction scheme based on resonant tunneling (which requires stringent energy level alignment), thus realizing the modular design of band structures. The band structure consists of absorption (blue) and extraction (red) modules, and by matching different functional modules, the device can achieve the control of parameters such as wavelength, electron transport, resistance, and dark current. In addition, the smaller wavefunction overlap of the diagonal transition also helps to improve thermal backfilling of the upper level electrons. In the extraction module, the miniband is utilized as the upper level to enhance electron extraction efficiency, while the increased spatial gap between the ground state and the step levels effectively suppresses the nonradiative transition process. Furthermore, the miniband consists of four levels with a bandwidth of 10.1 meV; this effectively broadens the response bandwidth accordingly.

The entire tuning process is realized by changing the width of the W1 well in the absorption module. As the width of W1 well decreases, the E1 level gradually rises, and the corresponding transition energy between E1 and Emini gradually decreases, resulting in a longer response wavelength. The tuning range of the E1 level is about 39.6 meV, which is approximately one longitudinal optical (LO)-phonon energy. The inset in Fig. 1 shows the detection wavelengths at different W1 well widths. When the well width is varied in the range of 5–7 nm, the response wavelength covers the range of 14–29.5 µm. In addition, to investigate the effect of MDT structure on the performance of devices for different response wavelengths, the extraction efficiency and DME were calculated, as shown in Fig. 2. As the well width increases, the extraction efficiency exhibits a trend of initially increasing and then decreasing, but the minimum value still exceeds 90%. This weak wavelength dependence of the extraction efficiency proves the high robustness of the MDT structure. The DME is another important parameter reflecting the QCD performance, and its squared value is proportional to the ISB absorption strength. As seen in Fig. 2, the DME is positively correlated with the response wavelength, with values of 1.38 and 2.3 nm at 14 and 29.5 µm, respectively. It is worth noting that the DME at the 14 µm wavelength is already close to 1.42 nm for the E1−E2 vertical transition and much higher than 0.98 nm^[11] for the bound-to-bound diagonal transitions.

Under illumination, the electrons in the ground state E1 absorb photons and a diagonal transition to the miniband is excited. After the transportation in the miniband, two energy steps separated by the E3 and E4 levels are passed through by LO phonon scattering to reach the ground state E5 of the next period, and finally are collected by the external circuit. For different detection wavelengths, since they have the same extraction module, the corresponding extraction efficiency and DME are essentially the same. Therefore, device performance is mainly affected by two transition processes, namely, E1−Emini and E3/4−E5. When the detection wavelength increases, the energy gaps for both E1−Emini and E3/4−E5 decrease. At this point, electrons on the E1 level are more likely to be affected by thermal noise during the diagonal transition process, leading to certain electron losses. Similarly, the electron transport efficiency from the E3/4 level to the E5 level also decreases because the energy gap between them no longer satisfies an LO-phonon energy. Moreover, the probability of thermal escape of electrons at the E5 level to the E3/E4 level also increases.

To verify the capability of widely tailorable wavelength coverage of the MDT structure, three samples with the same material structure except for different absorption transition well (W1) widths in the active region, corresponding to 7.0, 6.5, and 5.9 nm, were grown and fabricated. The active region is based on lattice-matched In0.53Ga0.47As/In0.52Al0.48As materials. The layer information (composition and thickness) of one cascade period can be found in the caption of Fig. 1. The QCD wafers were grown on a semi-insulating InP substrate by a metal–organic chemical vapor deposition (MOCVD) system equipped with a close-coupled showerhead growth chamber. During the growth process, the growth chamber was maintained at a low-pressure environment of 100 mbar. For the designed structure, the growth temperature and rate of the active region were controlled at 600°C and ∼0.2nm/s, respectively. All samples were composed of the active region with 30 repetitions inserted between the upper (200 nm) and lower (500 nm) contact layers. The doping concentrations of Si in the absorption well (W1) and the contact layers are 1×1017cm−3 and 2×1018cm−3, respectively.

The material quality and internal structure of the samples were investigated by high-resolution X-ray diffraction (XRD). Figure 3 presents the XRD rocking curves of the three samples. Clearly, the satellite peaks of all samples are very narrow (15–20 arcsec), indicating that they possess good periodicity and low interface roughness. According to the position of the satellite peak, the periodic thicknesses of the three samples were obtained as 58.852, 59.408, and 60.203 nm, respectively, within 1.86% thicker than the nominal design values. Nevertheless, the variation in the periodic thickness among samples due to the different widths of the W1 well corresponds well with the measured values.

Finally, the wafer was processed into 200μm×200μm mesas using standard lithography and wet-etching processes. A 450-nm-thick SiO2 sidewall passivation layer was then deposited using plasma-enhanced chemical vapor deposition. Ohmic contacts were fabricated by the deposition of Ti/Au on the top and bottom contact layers. To perform photocurrent measurements, the facets of the sample were polished into 45° to meet the QW selection rules requirement for TM-polarized excitation.

Optical measurements of the QCD were performed using a Bruker Vertex 70v Fourier-transform infrared spectrometer and a thermal source. Light is irradiated onto the QCD from a 45° InP substrate. Figure 4(a) shows the normalized photocurrent spectra of three samples at 10 K. It is evident that effective light detection can still be achieved by simply changing the width of well W1. The peak wavelengths of the three samples, corresponding to 14, 16, and 18 µm, are in good agreement with the theoretical calculations in the inset of Fig. 1. The three distinct dips are attributed to the absorption by CO2 (667cm−1) and InP two-phonon (632 and 660cm−1). These dips can be mitigated separately by using a vacuum spectrometer and a substrate thinning scheme. In addition, the influence of the miniband as an upper level on spectral width has also been observed. Here, the spectral width is defined as the ratio of the wavelength difference (Δλ) of 10% above the baseline to the peak wavelength (λ). At 10 K, the spectral widths of all three samples exceeded 26%, with a maximum width of 44% for 14 µm. This result is 6.3 times wider than the bound-to-bound design (7%)^[11], and also exceeds the broadband QCD (27.3%) based on the multicore active region (26 differently designed active regions) structure^[19]. It is worth noting that the total spectral width of the three samples exceeded 10 µm, which represents a new design concept for broadband QCDs.

Using the photoresponse spectra and the blackbody photocurrent, the responsivity of the samples can be obtained, as shown in Fig. 4(b). The responsivity of all devices shows a trend of initially increasing and then decreasing with increasing temperature, which can be explained by the inefficient ionization of the donor dopants at low temperatures^[11]. Among the three samples, the 14 and 16 µm QCDs have similar performance, with responsivities of 7 and 11 mA/W at 10 K, respectively, and corresponding operating temperatures reaching 110 K and 130 K. Compared with the other two samples, the device performance of 18 µm QCD shows a slight degradation. The responsivity is 2.5 mA/W at 10 K, and the maximum operating temperature is 60 K. It is worth noting that this result is better than the reported QCDs at an identical wavelength^[17].

The product of the area of the device (A) and its R0 at 0 V bias is indicative of the Johnson noise in the detector. Based on the results of dark current-voltage (not shown here), the differential resistance-area products (R0A) are plotted as functions of temperature for the three devices, as depicted in Fig. 5(a). At low temperatures (<40K), R0A demonstrates nonlinearity. This is because the dark current noise of the device gradually changes from thermal noise to photon noise as the dominant factor. Because photon noise does not depend on device temperature, R0A tends to be a constant within this temperature range. Through linear fitting, the thermal activation energies of the three samples can be derived, which are 47.5 meV (14 µm), 45.4 meV (16 µm), and 60.6 meV (18 µm), respectively. Reflected in the conduction band diagram, these values primarily correspond to the energy between the E1 and E3/E4 levels. The four resonantly coupled wells forming the miniband increase the spatial separation of the two levels, which will effectively suppress the diagonal electronic leakage and thus improve the noise characteristics of the device.

From the peak responsivity, R0A, the Johnson noise-limited detectivity (DJ*=RPR0A/4KBT) for QCDs is estimated^[20]. Figure 5(b) shows the DJ* of the three devices along with the calculated background-limited detectivities (DBLIP*). At 10 K, the DJ* of 14, 16, and 18 µm QCDs are 1.53×1011, 4.2×1010, and 3.5×1010 Jones, respectively. These results are better than the previous VLWIR QCDs, with a detectivity of 2.2×109 Jones at 10 K^[18]. The DBLIP* corresponds to the value at which the detectivity no longer changes when the temperature drops to a critical temperature, and this critical temperature is the background-limited temperature (TBLIP). The DBLIP* for 14, 16, and 18 µm QCDs are 4.63×1010, 4.7×1010, and 4.84×1010 Jones, respectively, showing an upward trend with increasing wavelength. Their corresponding TBLIP are 23 K, 10 K, and 6 K, respectively. Within this temperature range, the device maintains a good temperature stability.

In this work, we propose a modular band structure for VLWIR QCD to achieve wide wavelength tuning. The MDT design realizes this modular structure and provides high extraction efficiency and absorption strength. In theory, a wide spectral coverage (∼14–30μm) can be achieved by independently adjusting the absorption module. Finally, three samples with different well widths were designed and fabricated to verify the feasibility of the proposed scheme. At 10 K, the response wavelengths of the three devices are 14, 16, and 18 µm, respectively. Among them, the 16 µm QCD shows the best performance, with a peak responsivity of 11 mA/W, a DJ* of 1.53×1011 Jones, and a maximum operating temperature of 130 K.