Visible (VIS) and near-infrared (NIR) dual-mode imaging technology can capture rich visual information, aiding humans in better understanding and interacting with the world. This technology, beyond human vision, has a wide array of potential applications in fields such as machine vision, biomedical imaging, environmental monitoring, and military and defense, attracting extensive attention from researchers [1–4]. For example, Tian’s group effectively improved clinical image-guided tumor surgery by assembling visible and infrared cameras into a multispectral camera system. However, this approach involves two separate optical systems, leading not only to a bulky system but also to issues with image fusion [5,6]. The side-by-side configuration of VIS and NIR photodetectors (PDs) is an effective strategy for enhancing system integration, but it can lead to large pixel sizes in image sensors and difficulties in resolving optical crosstalk. More critically, traditional silicon/germanium (Si/Ge)-based materials have complex preparation processes and non-adjustable bandgaps, making it difficult to implement spectral regulation engineering and limiting the development of vertical stacking VIS/NIR dual-mode PDs with compact pixels.

The booming novel materials such as organics, 2D materials, and perovskites (PVKs) have offered new solutions to overcome previous challenges [7–11]. Scientists have engineered VIS-NIR switchable dual-mode vertically stacked PDs, showcasing their potential in dual-color imaging applications. For instance, researchers have increased the thickness of the organic or PVK active layer to induce a spatially confined photocarrier distribution effect, designing single component PDs with switchable VIS-NIR dual-mode response [12–14]. But the thick active layer leads to the device requiring a high bias voltage to operate. To reduce the operating bias, researchers further proposed vertically stacking two materials with different bandgaps to maintain a thin active layer while still achieving the spatially confined photocarrier distribution effect in dual-mode PDs. As a result, PVK/organic [15], PVK/PVK [16,17], MoS₂/Ge [18], and PVK/Si [19] stacked dual-mode PDs have been developed, featuring low-voltage-driven switchable VIS and NIR photodetection. However, these devices have not yet achieved the necessary level of spectral controllability to replace Si photodiodes that are coupled with filters [20,21]. Moreover, these multi-junction PDs are structurally complex and require dual bias polarity operations, resulting in complex fabrication processes for planar array devices and challenging readout circuit designs. Therefore, the pursuit of simple yet efficient VIS-NIR dual-mode PDs for excellent dual-mode imaging remains a considerable challenge.

Herein, we propose a simple device architecture that enables one PD to detect VIS-NIR dual-mode light at zero bias. The device’s structure is PVK-Au/Si/Ag, where PVK serves as an optical filter layer and Si acts as the active layer. The asymmetric electrodes design (Au/Si/Ag) enables the device to have excellent detection performance at zero bias, and the double-fishbone electrode design allows the device to support both top and bottom illumination modes. In the top mode, light is directly absorbed by silicon, so the device exhibits excellent VIS-NIR broadband light response characteristics. In the bottom mode, the PVK films absorbs visible light, which does not contribute to the photocurrent, so only NIR light reaches the silicon layer and generates photocurrents, enabling a response in the NIR spectrum. Thus, we design a minimalist self-powered VIS-NIR dual-mode PD that can switch between modes simply by flipping it over.

Although our PD’s design is simple and requires no bias, creating a high-density array imaging sensor remains challenging. It is also a common issue for novel semiconductor materials in imaging sensor development. To address this, our group has developed a novel single-pixel imaging technique, offering a fresh perspective for the advancement of new-materials based PDs [22]. Single-pixel imaging, as a novel computational imaging technology, encodes the spatial information of objects into structured light patterns, thereby enabling imaging functionality with only a single detector combined with backend decoding algorithms [23]. This technique enables spatially resolved imaging using just a single PD (rather than array image sensor) coupled with single-pixel imaging algorithms, presenting new opportunities for advanced imaging technologies such as non-visible light imaging, wide-angle imaging, and multispectral imaging [24–26]. In this work, we develop a distinctive dual-mode single-pixel imaging system based on the proposed dual-mode PD. A broad-spectrum VIS-NIR image and a pure NIR image can be obtained with the PD working on the top mode and bottom mode, respectively. Furthermore, a pure VIS image and an NIR-enhanced broad-spectrum image can be calculated by the two images from the top mode and bottom mode. Therefore, unprecedented VIS-NIR dual-mode imaging results are obtained by the proposed PVK-Si-based mode-switching single-pixel imaging system, demonstrating the novel dual-mode PDs’ unique advantages.

Figure 1(a) illustrates the fabrication process of our minimalist dual-mode PD, which utilizes a thermal evaporation technique to sequentially deposit Au, Ag, and PVK on a Si substrate. Both Au and Ag are deposited using a mask evaporation method, allowing the device to support both top and bottom illumination modes. The use of a simple asymmetric electrode design can create a potential difference at the interface [27], allowing the device to operate at zero bias, thus eliminating the need for complex high-temperature doping and interfacial modification processes [28]. The MAPbI3 (CH3NH3PbI3) PVK films are deposited on top of the Au electrode using a co-evaporation method [29], avoiding deposition over the Ag electrode due to the high reactivity of iodine (I) in MAPbI3 with silver (Ag), which can lead to degradation of the device performance [30]. Figure 1(b) presents a schematic diagram of the device structure, which supports two operational modes: bottom illumination for infrared band response and top illumination primarily for visible band response, as shown in the inset image. In top mode, direct silicon absorption of light yields a superior VIS-NIR light response. Conversely, in bottom mode, PVK film absorption of visible light is non-contributory to photocurrent, allowing only NIR light to stimulate photocurrents in the silicon, thus focusing the response on the NIR spectrum. Figure 1(c) shows a schematic of the dual-mode single-pixel camera designed using this PD, demonstrating its capability to simultaneously support visible light imaging for surface profiling and near-infrared imaging for vascular mapping, offering a new imaging tool for humans to understand the complex world.

The detection performance of the device with top illumination is tested, as shown in Fig. 2. Figure 2(a) plots the current-voltage (I-V) curves of the device under the light and dark conditions. It indicates that the device has an extremely low dark current density of 0.1nA/cm2 and a high on/off ratio of ∼106 at zero bias, suggesting that the device will possess self-driven photodetection capabilities. Figure 2(b) shows the transient photocurrent response curve of the device (660 nm, 0.1 mW), with its regular square-wave demonstrating the device’s light-switching characteristics and fast response speed. We further test the relationship between photocurrent and light power as shown in Fig. 2(c). The linear dynamic range (LDR) of the device reaches 199 dB, which can be calculated by LDR=20log(Pmax/Pmin), where Pmax and Pmin are 36 pW and 0.3 W, respectively. Figure 2(d) shows the spectra-dependent responsivity and special detectivity (D*). It indicates the device at top illumination exhibits broadband photoresponse covering from 400 nm to 1100 nm. The highest responsivity and D* are up to 172mAW−1 and 5.56×1013 Jones, respectively, at ∼900nm with zero bias. Briefly, our designed dual-mode PD under top illumination mode possesses a broadband response primarily in the visible range and demonstrates good photodetection performance, laying the foundation for high-quality VIS single-pixel imaging.

The organometal halide perovskites, as the star materials in recent years, have been widely used in photovoltaics, LEDs, and PDs due to their bandgap tunability and optoelectronic properties. Here, we employ the MAPbI3 perovskite film as a visible-light cutoff filter, enabling the device to have a narrowband near-infrared spectral response in bottom illumination modality. Figure 3(a) displays the X-ray diffraction (XRD) pattern of the MAPbI3 film, and the well-defined crystalline orientation suggests a high-quality film. The inset in Fig. 3(a) shows the crystal structure of the MAPbI3 perovskite. Figure 3(b) shows the SEM image of the PVK film, revealing that the film is a dense polycrystalline film with grain sizes around 200–300 nm. Figure 3(c) shows the absorption and PL spectra of the PVK film, indicating that the absorption cutoff of the device is around 780 nm, which perfectly covers the entire visible light region. The PL spectrum is a symmetrical single peak located at the absorption cutoff position, suggesting that the film is a direct bandgap semiconductor. Figure 3(d) provides the transmission spectrum of the film, which is measured by depositing the PVK films on a transparent glass substrate. It further indicates that the PVK film possesses filtering characteristics within the visible range, with its transmittance being suppressed by nearly 1000 times. The transmittance of the PVK film in the NIR spectrum can exceed 70%, with a tendency to increase with wavelength. The above characterization data indicate that the PVK film is a visible light filter, which will endow our designed PD with a near-infrared narrowband response characteristic.

Figure 3(e) demonstrates the spectral response/rejection characteristics of the device under the action of the PVK filter layer, that is, in the bottom illumination modality. The spectral rejection ratio is in a mirrored relationship with the transmission spectrum of the PVK film, where the maximum rejection ratio can reach 105. The spectral response curve exhibits a near-infrared narrowband response characteristic, providing a significant guarantee for perceiving only infrared imaging information. The highest responsivity and full width at half maximum (FWHM) of the PD in bottom mode are 127mAW−1 and about 200 nm, respectively. Figure 3(f) presents the light switch response characteristics of the device at a modulation frequency of 5 kHz, with a rise time of 25 μs and a fall time of 34 μs. The response time is defined as the time interval between 10% and 90% of the signal strength. The ultra-fast response speed of our device is attributed to the Au/Si Schottky junction. Since the Schottky junction is formed at the interface of silicon, the charge transport process is very fast under the action of the built-in electric field. Such rapid response time enables the PD to meet the requirements for the digital micromirror device (DMD) operating at high frequencies, thereby increasing imaging speed. Hence, we have successfully designed a novel PD with VIS-NIR dual-mode modulation, proving its high photodetection performance and suitability for dual-mode single-pixel imaging.

In order to better reflect the advantages of our device, we compared the performance of other VIS-NIR dual-mode PDs. As shown in Table 1, our device performance is not only capable of mode conversion at zero bias, but also has outstanding detection performance.Table 1.

Performance Comparison of the VIS-NIR Dual-Mode PDs

To bypass the manufacturing complexity of high-density pixel arrays, we develop a novel dual-mode single-pixel imaging system based on our designed PVK-Au/Si/Ag PD. Figure 4(a) shows the schematic diagram of the dual-mode single-pixel imaging system, which consists of a DMD, a broadband light source, lens groups, the imaging object, the PD, and a computer. In this configuration, the DMD-based projection system is responsible for illuminating the imaging object with pre-processed Fourier patterns. The PD then receives the diffusely reflected light intensity signals under different projection patterns. By flipping the PD, two mode signal datasets can be acquired. The acquired data are ultimately fed into a computer, where imaging algorithms process the data into an image. Here, we employ a four-step phase-shifting Fourier single-pixel imaging algorithm, as shown in Fig. 4(b). Through algorithmic analysis, we can obtain dual-mode imaging results from two different bands.

To demonstrate the advantages of our proposed dual-mode single-pixel imaging system in acquiring superhuman visual information, we design an imaging object as shown in Fig. 5(a). The imaging object consists of nine circles and three letters, where the top three circles can absorb the full spectrum of VIS to NIR light, while the six circles in the middle and bottom only absorb VIS light. To better demonstrate the system’s signal separation capability, we use NIR ink, which is invisible to the naked eye but absorbs NIR light, to write the letters XDU within the middle row of circles. The faint visibility of XDU in the photograph [Fig. 5(a)] taken by a mobile phone is due to water trace. The PD records the diffuse reflection light signal for each pattern under both top and bottom modes, yielding two I-m curves as depicted in Figs. 5(b) and 5(c), where I is the obtained current illuminated by a pattern and m is the number of patterns. Using the DMD for high-speed projection (10 kHz), we projected a total of 32,768 pre-encoded Fourier patterns within ∼3s. Subsequently, inverse Fourier transformations are applied to the two I-m curves to reconstruct the images for the I-m curves, as shown in Figs. 5(d) and 5(e). It is evident that the quality of the reconstructed images significantly improved with an increased m. In the top modality, both VIS and NIR signals are detectable because the PD’s response spectrum covers 400–1100 nm. Since the VIS and NIR information is mixed, this mode cannot differentiate between VIS and NIR information of the object. In the bottom modality, the images only reveal areas with NIR absorption, especially the XDU letters, which are distinctly isolated. Additionally, the unique ringing artifacts caused by the low sampling rate of single-pixel imaging are notably mitigated in the bottom mode (such as an image with m=2048). This suggests that enhancing the purity of detection signals can effectively improve imaging quality and suppress ringing effects, a finding that will inspire fast single-pixel imaging. Specifically, in this experiment, we only need to project 2048 patterns to recognize the letters “XDU” in the NIR image, reducing the imaging time from ∼3s to ∼0.2s, significantly enhancing image recognition efficiency. In addition, to obtain pure VIS images, we subtract the bottom mode images from the top mode images, which removed the XDU letters from the images, as shown in Fig. 5(f). Even more, we can also obtain images with enhanced NIR information by fusing the top and bottom dual-mode images [Fig. 5(g)], suggesting that this technique will have positive implications in applications such as blood vessel mapping as shown in Fig. 1(c).

In this work, we design a VIS-NIR dual-mode PD with a simple structure of PVK-Au/Si/Ag. This device can support two modes at zero bias: a broad-spectrum VIS-NIR response under top illumination and a narrowband NIR response under bottom illumination. The device exhibits photodetection performance with a highest responsivity and specific detectivity reaching 172mAW−1 and 5.56×1013 Jones, respectively, and a response speed of 25/34 μs (rise time/fall time). To demonstrate the superiority of the PD, we develop a dual-mode single-pixel imaging system, showcasing decent imaging results. The imaging results indicate that our dual-mode system can not only effectively differentiate VIS and NIR information but also significantly reduce ringing artifacts at low sampling rates, enhancing the speed of near-infrared signal perception. We believe that our newly developed dual-mode single-pixel imaging method will offer fresh perspectives for the field of intelligent perception.

Fabrication of the device employs a full thermal evaporation process, whereby Au and Ag fishbone electrodes with a 120 nm thickness are first deposited on the cleaned top and bottom surfaces of the n-Si layer, respectively. Evaporation is performed when the pressure of the coating machine is reduced to 10−4Pa, and the evaporation rate is 0.5Å/s. Subsequently, a dual-source co-evaporation technique is used to deposit MAPbI3 PVK film with a 400 nm thickness on the side of the Au electrode. MAI (CH3NH3I, Greatcell) and PbI2 (TCI) were co-deposited in a thermal evaporation chamber. Source heating began after the chamber pressure was reduced to below 5×10−4Pa. The deposition rate of PbI2 was stabilized at 0.50Å/s, controlled via a gold-plated quartz crystal microbalance, with the temperature maintained between 280°C and 310°C. Meanwhile, the temperature of the MAI source was adjusted to ensure a deposition rate of 0.70Å/s, monitored by a nearby quartz microbalance, with temperatures ranging from 180°C to 190°C. The final perovskite films had a thickness of approximately 400 nm. The active area of the device is 0.3cm2 defined by the fishbone electrode region.

XRD patterns of MAPbI3 PVK films are measured by Rigaku, Miniflex600. Absorption spectra are measured by a UV-VIS spectrophotometer (UV-2600, Shimadzu, Japan). SEM images are obtained by using a scanning electron microscope (ZEISS ULTRA 55). Detection performance characterizations are measured by a Keithley source meter (2601B) with a series of lasers with different output wavelengths. Illumination intensity is modified using a series of attenuators (0.01%, 0.1%, 1%, 10%, 50%) and is standardized with an optical power meter equipped with a conventional Si photodiode (S10-14010). The spectral response curve is measured by a Keithley 2601A source meter and a monochrome light (Enlitech, Si detector S10-14010). The standardized EQE was tested by using the QE-R3011 system (Enlitech’s QE-R system), the light spot size is 0.01cm2, and the light intensity of different wavelengths is distributed around 1mW/cm2.

VIS/NIR continuous spectrum light is from a white/NIR LED with electric power of ∼3/15W. The camera lens is from a commercial projector. The DMD (V-7001VIS, ViALUX) has 1024×768 micromirrors with up to 22727 Hz projection rate. The PD is self-powered and I-V amplificated by a photodiode amplifier (PDA200C, Thorlabs). DAS is a data acquisition card (NI USB-6361, National Instruments).

Acknowledgment. Authors thank the National Natural Science Foundation of China, the National Young Talent Program, the Shaanxi Young Top-notch Talent Program, the Key Research and Development Projects of Shaanxi Province, and Xidian University Specially Funded Project for Interdisciplinary Exploration for help identify collaborators for this work.