The band structure of the M-structure superlattice is shown in Fig. 1(a). Its periodic structure was composed of 18 monolayers (MLs) InAs/3 MLs GaSb/5 MLs AlSb/3 MLs GaSb. The thin AlSb layer in the M-structure affects both electrons in the conduction band and holes in the valence band. It forms a more effective barrier that blocks the tunneling of the electrons in the two adjacent InAs wells. For holes, the AlSb layer causes the GaSb layer to become a double quantum well, which elevates the first and the second valence bands. Therefore, with the high electron effective mass in the M-structure, the transport of electrons in the depletion region due to diffusion and tunneling would have more resistance, and the dark current is expected to decrease.^[3] Our device was grown using a molecular beam epitaxy (MBE) system (Veeco Epi Gen II) based on an N-type GaSb (100) substrate. The substrate was double-side polished to avoid strain relaxation and interface related defects.^[13] The structure and band diagram of the device are shown in Fig. 1(b). First, an N-type GaSb buffer with thickness of 900 nm was grown between the substrate and the red channel contact region, which ensures an atomically smooth surface for the device growth. Then, the 2 μm thick lightly Be-doped (p ≈ 4–6 × 10¹⁶ cm⁻³) active region superlattice for the red channel was grown with 14 MLs of InAs and 7 MLs of GaSb per period, while the blue channel grown subsequently had an 11 MLs InAs/7 MLs GaSb design with the same total thickness. Both Si-doped N regions (n ≈ 1 × 10¹⁸ cm⁻³) were composed of a 0.5 μm thick M-superlattice. The M-structure barrier regions with thickness of 0.5 μm were undoped. A 0.2 μm Be-doped (p ≈ 8 × 10¹⁶ cm⁻³) Al_0.2Ga_0.8Sb barrier was sandwiched between the two channels. Finally, the entire structure was capped off with a 0.02 μm thick N-type (n ≈ 2 × 10¹⁸ cm⁻³) InAs layer.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (a) The band diagram of the M-structure, where E1b and H1b represent the band edges of the first electron energy levels and the first hole energy levels in the minibands, respectively. (b) The structure diagram and (c) band alignment of the LW/LW dual-color detector, where the electron QFL and hole QFL represent quasi Fermi levels.

The band alignment and distribution of electrical field intensity are shown in Fig. 1(b). The valence band offset between the red and blue channels was almost eliminated. This means that the device could be operated under low bias voltage. The electrical field drop is small across both absorber regions, since there is a significant amount of field drop across the wider-bandgap barrier layers. This reduction in electric field will lead to small depletion regions and hence a reduction in the generation–recombination current (J_gr) and tunneling current.

The mismatch of each superlattice was measured using symmetric (004) x-ray diffraction. The high-resolution x-ray diffraction (HRXRD) pattern is shown in Fig. 2(a). Table 1 lists the parameters of each superlattice in the device. All of the contained superlattices’ periodic thicknesses match well with our design. The full width at half maximum (FWHM) shows the high crystalline quality of the epi-layers. The morphology of the sample was characterized by differential interference microscopy (Nikon ECLIPSE) (Fig. 2(b)) and atomic force microscopy (AFM) (Fig. 2(c)). In an area of 870 μm × 650 μm, the epi-layer exhibits flatness and non-defective morphology. The AFM scanning result shows the atomic step edge. And the root-mean-square (RMS) of roughness is 1.9 Å over a 10 μm × 10 μm area.

Fig. 2.

	Figure Option ViewDownloadNew Window
	Fig. 2. (a) HRXRD pattern, (b) differential interference micrograph, and (c) AFM image of the device.

Table 1.

Table 1.

Table 1. The mismatch, period, and FWHM of 0th peaks of different superlattices. . Superlattice Blue channel Red channel M-structure Mismatch/(″) 14 191 191 Theoretical period/Å 58 68 92 Actual period/Å 58 68 91 FWHM/arcsec 36 39.6 39.6

Table 1. The mismatch, period, and FWHM of 0th peaks of different superlattices. .

The mesas with different sizes (200 μm, 300 μm, 500 μm) were defined through lithography (SUSS MA6) and fabricated by dry etching through the InAs cap layer (top contact) and each region into the red channel N-type contact (bottom contact). The SiO₂ hard etching mask (PECVD STS) and inductively coupled plasma (ICP OXFORD plasmalab System 100) system were chosen to etch the mesa. The sidewall was passivated using anodic sulfide, then covered by a 200 nm SiO₂ layer. The SiO₂ layer was deposited using ion beam sputtering (OXFORD optofab3000). After the above steps, a second photolithography and ICP were performed to open the window on the passivation layer. The final metal electrode was deposited using electron beam evaporation (SKY EB700-I).

The working mechanism is similar to that of the dual-band detector which was mentioned in Ref. [14]. The optical characterization of the front-side illuminated detector was performed at 77 K using a Fourier transform infrared (FTIR) spectrometer (Bruker Vertex70) and a calibrated black-body (SYSTEMS) source at a temperature of 500 °C. The FTIR spectrometer was calibrated by using a standard DTGS detector without anti-reflection (AR) coating. Figure 3 shows the optical performance of the detector with diameter 300 μm. It is defined as forward bias when the applied bias on the top contact is positive with respect to the bottom contact. Otherwise, it is defined as reverse bias. The blue channel exhibits a peak responsivity (R_i) of 2.4 A/W at 6.8 μm under 20 mV, while the red channel shows a peak R_i of 4.2 A/W at 9.1 μm under a bias of −60 mV. The 50% QE cutoff wavelengths are measured at 7.6 μm and 11.5 μm. The peak QEs of the blue and red channels are 44% at 6.3 μm and 57% at 9.1 μm, respectively. From Fig. 3(c), the bias range above −20 mV corresponds to the transport of minority electrons in the blue channel, while the bias range below −20 mV corresponds to the transport of minority electrons in the red channel. The applied bias is small enough to be suitable for focal plane arrays (FPAs) operation. In order to quantify the optical cross-talk, we use a selectivity parameter (S) defined as follows:^[15]

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. (a) Spectral responsivity and (b) quantum efficiency under −60 mV, −20 mV, and 20 mV at 77 K. (c) The variation of quantum efficiency with applied bias for wavelengths of 9.1 μm and 6.3 μm at 77 K.

Table 2.

Table 2.

Table 2. Comparison with published results. . Structure Red channel peak QE/% Applied bias/mV Blue channel peak QE/% Applied bias/mV Ref. [11] N-M-π-P-P-P-π-M-N 45@6.3 μm –50 43@10.2 μm 150 Ref. [12] pBp 38@6.5 μm 90 34@9.81 μm –100 This work N-M-π-B-π-M-N 44@6.3 μm 20 57@9.1 μm –60

Table 2. Comparison with published results. .

The dark current density (J) was measured using Agilent B1500 in the bias range from 0.3 V to −0.3 V at 77 K. As shown in Fig. 4(a), the blue and red channels of the device with diameter 300 μm show dark currents of 8.5 × 10⁻⁵ A/cm² at 20 mV and 4.6 × 10⁻⁴ A/cm² at −60 mV, respectively. The differential resistance and area products RA are 201 Ω ·cm² under 20 mV and 67 Ω ·cm² under −60 mV at 77 K. The linear fitting of inverse RA vs. P/A is shown as a function of P/A for the devices with diameter 500 μm, 300 μm, and 200 μm in Fig. 4(b), and the bulk-limited resistances of the blue and red channels are 1000 Ω · cm² and 179 Ω ·cm², respectively. The surface resistivity is equal to 2.9 × 10⁴ Ω · cm and 9.1 × 10³ Ω · cm for the blue and red channels, respectively, at 77 K. For both channels, the relatively high surface resistivity shows suppressed surface related currents, but the bulk-limited resistance is not large enough to block the bulk leakage.^[12] As a result, the main dark current comes from bulk leakage. Compared with the results published in Refs. [11] and [12], the dark current density of our blue channel is relatively high. The bulk leakage was fitted by referencing the published result.^[16] As shown in Fig. 4(c), the main leakage in both channels under operating bias is J_gr. The M-structure barrier in the blue channel fails to take most part of the applied bias. Therefore, a large part of the p-side depletion region lies within the absorber layer, giving rise to large J_gr. A narrow width of the depletion region in the absorber is beneficial for decreasing J_gr. It may be useful to decrease the depletion region width in the blue channel by using the periodic structure of 18 MLs InAs/5 MLs GaSb/5 MLs AlSb/5 MLs GaSb to replace the currently used M-structure. And the high density of recombination centers in the absorber also results in large J_gr. Another main dark current source is the trap-assisted tunneling current (J_TAT) that depends on the trap density and the effective mass of the electrons. And in theory, the recombination centers and trap states can be eliminated by optimizing the superlattices’ growth conditions.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (a) Dark current density J, differential resistance and area products RA of the dual-color detector with mesa diameter 300 μm. (b) The inverse RA vs. P/A is shown as a function of P/A for the detectors with different sizes. (c) Measured and modeled dark current of the LW/LW dual-color detector under bias from −0.3 V to 0.3 V.

Figure 5(a) collects dark current densities of the blue and red channels for comparison with “Rule 07”. The cutoff wavelength is taken as the point of a 50% response. The experimental data of the blue channel show considerably greater leakage current than Rule 07 by some orders of magnitude, while the current density of the red channel is close to Rule 07. The trend is that it is generally more difficult to control dark current at the shorter wavelength.^[17] The specific detectivity (D^*) spectrum is shown in Fig. 5(b). At 77 K, the maximum D* for the blue channel is 5.0 × 10¹¹ cm·Hz^1/2/W at 6.8 μm under 0 mV, while the red channel has 3.1 × 10¹¹ cm·Hz^1/2/W at 9.1 μm under −60 mV. The specific detectivity can be enhanced by increasing the absorber thickness and inhibiting dark current density.

Fig. 5.

	Figure Option ViewDownloadNew Window
	Fig. 5. (a) The 77 K dark current densities plotted against cutoff wavelength for the blue channel under 20 mV and the red channel under −60 mV. The solid line indicates the dark current density calculated using the empirical Rule 07 model. (b) The specific detectivity spectrum of both channels at different applied bias voltages.