Device design and working principle. Figure 2a depicts the architecture of the multi-dimensional optical information MWIR photodetector. The top P-type bP and the middle N-type MoS2 form the PN junction. Simultaneously, the MoS2 is combined with the P-type b-AsP at the bottom to form an NP junction. This b-AsP/MoS2/bP arrangement creates a vertical structure of PNP heterojunction, where bP and b-AsP serve as the absorbing layers for MWIR1 and MWIR2, respectively, and MoS2 acts as the barrier layer to block hole transport. The properties of b-AsP, MoS2, and bP field effect transistor (FET) have been characterized (Supplementary Fig. 1). The back-to-back photodiode configuration is commonly employed for dual-band detection38–41. Unlike the three-terminal configuration with a common electrode, the two-terminal configuration only requires one readout circuit electrode, simplifying the device structure and increasing the fill factor, thereby advancing high-resolution detectors.
Figure 2b shows the mechanism of photodetector bias-switchable polarization-resolved detection. During the stacking of heterojunctions, the crystal orientation of bP and b-AsP does not overlap, allowing bP and b-AsP to be sensitive to different linearly polarized light. In the configuration shown in Fig. 2b, b-AsP is sensitive to linearly polarized light in the x direction, while bP is sensitive to linearly polarized light in the y direction. To effectively block majority carriers while maintaining good transport capacity for photoexcited minority carriers, the thickness of MoS2 is less than 20 nm, as confirmed by atomic force microscopy (AFM). The thickness of the absorption layer is flexible. The PNP vdWs heterojunction is fabricated through dry transfer stacking after mechanical stripping to obtain the required materials. After depositing electrodes, a thin layer of h-BN is transferred to the top as a passivation layer. The image of cross-section high-resolution transmission electron microscopy (TEM) and Energy dispersive electron spectroscopy (EDS) at the b-AsP/MoS2/bP heterojunction interface is taken (Supplementary Fig. 2). Further details of the fabrication process can be found in the Methods section. For this experiment, two types of b-AsP with different percentages of As atoms, namely b-As0.6P0.4 and b-As0.83P0.17, are used. The crystal structures of bP and b-AsP with different arsenic atomic ratios were verified by Raman spectroscopy (Supplementary Fig. 3). Figure 2c presents the polarization transmission spectra of bP and b-AsP, measured by Fourier transform infrared spectroscopy (FTIR). From Fig. 2c, it can be observed that both bP and b-AsP exhibit anisotropic optical absorption characteristics, with the absorption cutoff edge of b-AsP shifting towards longer wavelengths as the percentage of arsenic atoms increases.
To investigate bias selection characteristics, photocurrent mapping images were obtained under different bias voltages. Figure 3 presents the operational mechanism of the heterojunction, which utilizes positive/negative bias voltages to select different response channels. In Fig. 3a, an optical image of the device is shown, with the contours of b-AsP, MoS2, and bP highlighted by red, gold, and blue dashed lines, respectively. The schematic diagram of the device structure is displayed in the upper right portion of Fig. 3a, with b-AsP, MoS2, and bP stacked from bottom to up. The device can be divided into three regions based on the type of junction area: Region Ⅰ represents the b-AsP/MoS2 PN junction area, region Ⅱ represents the MoS2/bP NP junction area, and region Ⅲ represents the PNP heterojunction composed of b-AsP/MoS2/bP, as depicted in Fig. 3b, 3c, and 3d, respectively. To gain insights into the underlying mechanisms of heterojunction photoresponse at different bias voltages, photocurrent mapping images were acquired at bias voltages of 0.5 V and − 0.5 V, as illustrated in Fig. 3e and 3i, respectively. These measurements were conducted under the illumination of an 830 nm laser, with electrode A1 serving as the source and electrode B2 as the drain. It is evident that under a positive bias voltage of 0.5 V, the photoresponse originates from bP (Fig. 3e). The band structure schematics of regions Ⅰ, Ⅱ, and Ⅲ under a bias of 0.5 V are presented in Fig. 3f, 3g, and 3h, respectively. The internal electric fields of the b-AsP/MoS2 junction 1 and MoS2/bP junction 2 are denoted as Ein1 and Ein2, respectively. The applied electric field generated by the bias voltage is denoted as Eex. When operating under a positive bias, junction 1 is forward biased and junction 2 is reverse biased. When the positive bias voltage increases to |Eex| > |Ein1|, the photogenerated holes produced by b-AsP layer cannot be transported to the drain due to the presence of the barrier layer, resulting in recombination with electrons and no photocurrent generation, as shown in Fig. 3f. The bP layer absorbs photons and produces photogenerated electron holes, which are separated by the electric field and eventually collected by the electrode to form a photocurrent, as shown in Fig. 3g. The band structure of PNP heterojunction under the positive bias opening operating voltage is shown in Fig. 3h. In another word, when the positive bias voltage increases to |Eex| > |Ein1|, the holes generated in the b-AsP layer, whether photogenerated or thermal excitation, will be blocked by the barrier layer35, and the bias applied at this time can be defined as the operating voltage Vop of junction 2 (bP channel, λ1 cut-off wavelength). Therefore, the PNP heterojunction with the barrier band structure designed in our device can not only reduce the electrical crosstalk of the dual-band photodetector but also suppress the dark current generated by the non-absorbing layer when operating at the opening voltage34. When the reverse bias reaches the open operating voltage, the situation reveres, as shown in Fig. 3i (b-AsP channel, λ2 cut-off wavelength). The change of PNP heterojunction band structure is opposite to that under negative bias, as shown in Fig. 3j, 3k, and 3l, respectively. Thus, the dual-band response is obtained under two different bias polarities.
To study the transition states before reaching the operating voltage when switching between positive and negative bias voltages, we also analyzed the photocurrent mapping of the device at 0.1 V, 0 V, and − 0.1 V bias (Supplementary Fig. 4). A detailed discussion of the physical process of the transition states and the analysis of reaching the open operating voltage Vop is provided in Supporting Information Section 1. Supplementary Fig. 5 and Supplementary Fig. 6 illustrate the photocurrent mapping images of device 2 and device 3, respectively, under different bias voltages, providing further evidence for the stable reliability of our photodetectors.
Photodetector characterization. Figure 4a depicts the schematic diagram of opening MWIR1 and MWIR2 by switching the bias voltage. To investigate the polarization-resolved spectral response of the dual-band photodetector, we utilized a FTIR spectrometer. The measurement details can be found in the Methods section. In Fig. 4b, a comparison of the polarization spectral responses between MWIR1 and MWIR2 of the photodetector (b-As0.83P0.17/MoS2/bP) is presented, alongside the atmospheric transmission spectrum. The dual-band responses, with λ1 cut−off at ~ 4.2 µm and λ2 cut−off at ~ 4.9 µm, are clearly observed in Fig. 4b. Notably, 4.2 µm corresponds to the absorption peak position of carbon dioxide, and our experimental setup operates in the atmospheric environment. Consequently, the spectral response of b-AsP exhibits a spectral gap at 4.2 µm. As shown in Fig. 4b, the response spectrum of b-AsP covers blue and red bands, and the response trend of the blue band aligns generally with that of bP. In practical applications, we can designate the response signal of bP as MWIR1. By multiplying a proportional coefficient (which is affected by absorptivity due to material thickness and stack position) to the response signal of b-AsP and subtracting the response signal of bP under identical conditions, we can define MWIR2. As a result, our photodetector is capable of discerning between MWIR1 (2.5 to 4.2 µm) and MWIR2 (4.3 to 4.9 µm) response spectra, enabling polarization detection within their respective spectral response ranges. This implies that our unit photodetector can directly distinguish spectral and polarization information. Therefore, this heterogeneously stacked vdWs heterojunction, exhibiting distinct absorption spectra and anisotropic spectral response, offers a promising avenue for the advancement of multi-dimensional optical information photodetectors in the future.
Take a step further, the detection performance of the photodetector for blackbody radiation at room temperature is evaluated. The measurement setup for blackbody radiation is illustrated in Supplementary Fig. 7a, with the photodetector placed in a vacuum Dewar. By utilizing 450°C blackbody as the light source, the output characteristic curve of the dual-band MWIR1/MWIR2 photodetector is shown in Supplementary Fig. 7b, demonstrating the sensitivity of our photodetector to blackbody radiation at room temperature. The effective incident blackbody radiation power can be calculated using the Eq. 36: \({P}_{bb}=\eta \frac{\sigma ({T}_{b}^{4}-{T}_{0}^{4}){A}_{b}{A}_{D}}{2\sqrt{2}\pi {L}^{2}}\), where η = 0.9 represents the transmittance of the Dewar window, σ is the Stefan-Boltzmann constant, Tb is the blackbody temperature, T0 is the operating temperature of the photodetector, Ab is the area of the blackbody radiation aperture, AD is the area of the active region of the sample, and L is the distance from blackbody to detector. The blackbody responsivity \({R}_{bb}\) given by Rbb = Iph/Pbb. Based on the data extracted from Supplementary Fig. 7b, we can calculate the blackbody responsivity of the MWIR1 photodiode (Vd = 0.5 V) and MWIR2 photodiode (Vd = -0.5 V) to be 10.1 A/W and 4.0 A/W, respectively.
Specific detectivity D* is an important parameter for characterizing the performance of IR photodetectors and can be determined using the following Eq. 20:
$${D}^{*} = \frac{\sqrt{AΔ f}}{NEP}=\frac{R\sqrt{AΔ f}}{{I}_{n}}$$
1
where A is the effective area of the photodetector, Δf is the bandwidth, R is the responsivity, In is the noise current, and NEP is the noise equivalent power. The main sources of noise in vdWs material photodetectors are typically shot noise and Johnson noise36, 42:
$${I}_{n} =2q{I}_{d}Δ f+\frac{4{k}_{B}TΔ f}{{R}_{shunt}}$$
2
where q is the elementary charge, Id is the dark current, kB is the Boltzmann constant, T is the temperature, and Rshunt is the shunt resistance. By combining equations 1 and 2, the specific detectivity D* of the MWIR1 photodiode and the MWIR2 photodiode can be calculated as 2.1×1010 cmHz1/2W−1 and 1.9×1010 cmHz1/2W−1, which is comparable with commercial photodetector.
Moreover, we conducted an investigation into the speed of the dual-band photodetector to a modulated laser at room temperature in the atmosphere. Supplementary Fig. 8 showcases the response of the MWIR1 photodiode and MWIR2 photodiode to the modulated laser with a wavelength of 1550 nm. Notably, both photodiodes exhibit a remarkably fast response to the laser, with rise and fall times of approximately 0.5 µs. Furthermore, Fig. 4c displays the frequency response of the MWIR1 photodiode and MWIR2 photodiode to a 1550 nm modulated laser, with both photodiodes reaching their − 3 dB points at ~ 700 kHz. Our device response speed is faster that of previously reported bP and b-AsP photodetectors27–30, 43–45.
Temperature measurement with colorimetric method. Dual-band IR photodetectors are more accurate and reliable than monochrome IR photodetectors in remote temperature measurement11, 46, 47. According to Wien's displacement law, the product of the absolute blackbody temperature and the wavelength corresponding to the maximum radiation intensity is a constant (λmT = b). This implies that as the temperature of an absolute blackbody increases, the maximum radiation power shifts towards the shorter wavelengths (as shown in Supplementary Fig. 8a). In Planck's law under Wien's approximation, the emission power of the target can be expressed as47:
$$M=\frac{2\pi h\epsilon b{\lambda }^{-5}Δ \lambda }{{e}^{\frac{hc}{{k}_{B}\lambda T}}}$$
3
Here, h is Planck's constant, ε is the emissivity, b is the Wien constant, λ is the wavelength, Δλ is the wavelength width, c is the speed of light, and T is the temperature of the object. When the target being measured behaves as a blackbody, the temperature obtained from the target is both accurate and reliable. However, if the target behaves differently from a blackbody, knowledge of the target object’s emissivity becomes essential for accurate data acquisition from a monochrome photodetector. In contrast, dual-band photodetectors eliminate the necessity of knowing the object’s emissivity, resulting in more precise and reliable temperature measurement compared to monochrome photodetectors.
It determines the temperature of an object based on the ratio of the radiation energy in two adjacent bands of the object’s IR radiation46. The ratio can be expressed as:
$$Q=\frac{{M}_{1}}{{M}_{2}}=\frac{{\epsilon }_{1}Δ {\lambda }_{1}}{{\epsilon }_{2}Δ {\lambda }_{2}}{\left(\frac{{\lambda }_{2}}{{\lambda }_{1}}\right)}^{5}exp\left[\frac{hc}{{k}_{B}T}\left(\frac{1}{{\lambda }_{2}}-\frac{1}{{\lambda }_{1}}\right)\right]$$
4
$$lnQ={C}_{1}+{C}_{2}{T}^{-1}$$
5
In this context, C1 and C2 represent the system constants. It is worth noting that colorimetric temperature measurement remains unaffected by emissivity, rendering it suitable for absolute and remote temperature measurement applications.
As shown in Fig. 4d, the photoresponse of MWIR1 and MWIR2 to the blackbody source is depicted, with the Dewar operating at liquid nitrogen temperatures to ensure a favorable signal-to-noise ratio. Notably, Fig. 4e demonstrates that the signal ratio of MWIR2/MWIR1 in our dual-band photodetector remains relatively constant despite variations in detection distance under a constant blackbody temperature. This finding suggests that colorimetric temperature measurement remains unaffected by changes in distance. Furthermore, the relationship between the MWIR2/MWIR1 signal ratio and changes in blackbody temperature at a constant detection distance was examined (Fig. 4f). It was observed that as the blackbody temperature increased from 450°C to 1000°C, the MWIR2/MWIR1 signal ratio correspondingly increased from 0.026 to 0.145. By utilizing the blackbody response as a reference standard, remote temperature measurements of a soldering iron and a candle flame were conducted. The temperature of the electric soldering iron was measured at 495°C, which closely aligned with the actual set temperature of 500°C, within the system's error tolerance. Similarly, the temperature of the candle flame was measured at approximately 720°C. These results serve as evidence that our dual-band photodetector can effectively be employed for remote temperature measurement of various target objects. Notably, even when the blackbody response frequency of the photodetector reaches 500 Hz, the MWIR2/MWIR1 ratio remains constant, as depicted in Supplementary Fig. 8. Furthermore, considering that the blackbody radiation power is associated with the detection distance, the relationship between the responsivity and blackbody radiation power can be derived from Fig. 4e, as illustrated in Supplementary Fig. 10. This highlights the capability of our photodetector to detect light intensity information in addition to its dual-band functionality.
Bias-switchable polarization-resolved detection. To demonstrate the bias-switchable polarization-resolved detection, we conducted an investigation into the response of a heterojunction composed of b-As0.6P0.4/MoS2/bP to linear polarized light after switching different bias voltages. The optical photo of the device is presented in Fig. 5a. As shown in Fig. 5b and 5c, the relationship between the periodic increase and decrease of the photocurrent is depicted as the polarization angle of the incident light (λ = 4.6 µm) changes from 0° to 360° at a bias voltage of 0.4 V and − 0.4 V, respectively. The measured polarization extinction ratios were found to be 24.7 and 11.8 for the bias voltages of 0.4 V and − 0.4 V, respectively. Furthermore, Fig. 5d illustrates the polar plot of forward and reverse photocurrents, determined by polarization angles of the incident light. It is evident that as the polarization angle varies from 0° to 180°, the photocurrent under a bias of 0.4 V is smallest at 50° and largest at 140°. Conversely, under a bias of -0.4 V, the photocurrent is maximum at 30° and minimum at 120°. This result indicates that the AC direction of bP is 140°, while the AC direction of b-AsP is 30°, as shown by the blue and red arrows in Fig. 5a. The photocurrent and polarization angle at both 0.4 V and − 0.4 V adhere to the functional relation45: Iph = acos(2θ + φ) + b, with a phase difference of 110°. The phase difference is determined by the angle between the crystal direction of bP and b-AsP during the stacking of the heterojunction.
The origin of heterojunction polarization detection is further confirmed through polarized photocurrent mapping. Figures 5e and 5i present schematic diagrams of the circuit when the bias voltage is 0.4 V and − 0.4 V, respectively. Additionally, Figs. 5f, 5g, and 5h display polarization-resolved photocurrent mapping images of the MWIR1 photodiodes at polarization angles of 50°, 95°, and 140°, respectively, with the incident light wavelength set at 4.6 µm. It can be seen that the photocurrent is generated in the bP region of the device, and the photocurrent reach its maximum at 140°. Similarly, Figs. 5j, 5k, and 5l depict the polarization-resolved photocurrent mapping images of the MWIR2 photodiode at polarization angles of 30°, 75°, and 120°, respectively. The device generates photocurrent in the b-AsP region, and the photocurrent is maximum at 30°. It is evident that the trend of the current changing with the polarization angle aligns with that shown in Fig. 5d. This observation further confirms the polarization recognition capability of the device with a bias switch.