Figure 1. Detailed Structural and Electrical Characteristics of the BP/Ge Heterojunction. (a) Schematic diagram of the BP/Ge heterojunction, depicting the configuration of p-type BP and p+-type Ge. (b) Optical microscope image of the assembled BP/Ge heterostructure, showcasing the meticulous alignment of the layers. The inset features an AFM image, highlighting the surface uniformity. (c) High-resolution cross-sectional TEM image of the BP/Ge interface, evidencing the seamless and impurity-free bonding between BP and Ge. Scale bar: 5 nm. (d) Energy band profiles of BP and Ge prior to contact, showing potential alignments and initial band conditions. (e) Experimentally measured I-V output characteristics of the P-type BP/Ge van der Waals heterojunction, displaying reverse rectification behavior that underscores the heterojunction’s distinctive electronic properties. (f) TCAD simulated band diagrams of the BP/Ge heterojunction with p-doped Ge, detailing the equilibrium state (red line) and the state post-carrier-induced band inversion (blue line), offering insights into the light-induced dynamic electronic interactions within the heterojunction.
Figure 1a illustrates the schematic of our BP/Ge van der Waals heterojunction, which incorporates p-type BP and p+-type Ge (doping with Ga ions). The detailed methods for device fabrication are outlined in the Methods section. In this configuration, BP serves as the anode, while Ge functions as the cathode. Figure 1b shows the device as viewed under an optical microscope; the inset features an atomic force microscopy (AFM) image, highlighting the device surface's flatness and uniformity. Supplementary Figure S1 provides measurements of the thicknesses of the BP and Al2O3 flakes, which are 40 nm and 60 nm, respectively.
To confirm the high quality of the interface post-transfer, we employed cross-sectional Transmission Electron Microscopy (TEM). Figure 1c shows TEM images at the BP/Ge interface, revealing clean atomic interfaces free from lattice mismatches, impurities, or gaps, with a notable single-layer thickness of BP at precisely 5.4 Å. The absence of interface oxide layers, which could impede carrier transport, is verified, highlighting the effectiveness of our nitrogen-filled glove box fabrication environment. Energy Dispersive X-ray Spectroscopy (EDS) mapping in Supplementary Figure S2 further corroborates these findings by distinctly labeling elements like Ge, P, and O. Additionally, Raman spectroscopy performed at three different positions within the heterostructure (Supplementary Figure S3) confirms the structural integrity and composition of the materials.
The design of the energy band structure is crucial for the functionality of semiconductor photodetectors and is illustrated in Fig. 1d. The accuracy of our band design was confirmed through Kelvin Probe Force Microscopy (KPFM), with measured work functions of 4.5 eV for BP and 4.56 eV for Ge (Supplementary Figure S4). Predictions based on the built-in electric field's direction suggested that the heterojunction would demonstrate forward cutoff and reverse conduction output characteristics. These predictions were corroborated by the I-V characteristics of our fabricated device, as depicted in Fig. 1e.
Figure 1f displays the simulated energy bands of the device using TCAD, indicating that holes in Ge diffuse towards BP, forming a P+P heterojunction. A notable weak potential barrier at the valence band (0.06 eV), depicted in blue, plays a crucial role in the transport properties of photogenerated carriers, especially under self-driven conditions. This barrier induces charge accumulation upon light exposure, leading to the inversion of the built-in electric field at BP, as illustrated by the blue line. Further theoretical validations through TCAD band simulations and device output I-V curves of the NP Ge/BP heterojunction are provided in Supplementary Note 1 and Supplementary Figure S5.
The heterojunction's response to photon irradiation exhibits distinctive behaviors depending on the energy range. High-energy photons (visible to near-infrared) trigger significant carrier recombination at the interface, adversely affecting detectability. In contrast, under low-energy photon (mid-wavelength infrared) exposure, the potential barrier hinders carrier migration along the built-in electric field, resulting in the accumulation of photogenerated holes and subsequent p-doping of BP. This accumulation leads to energy band inversion in BP, markedly enhancing infrared detectability. Given the heterojunction comprises two p-type semiconductors, hole transport predominates. After energy band inversion, photogenerated holes are impeded from crossing the interface, effectively reducing carrier recombination driven by interfacial transport and thereby enhancing detection capabilities at longer wavelengths. This light-induced band inversion not only augments selective infrared photoresponse but also curtails dark current, significantly improving the device's sensitivity and advancing its potential to approach the BLIP limit.
Figure 2. Room-Temperature Photoresponse Characteristics of the BP/Ge Heterojunction. (a) Photogenerated carrier transport properties and energy band structure under λ < 1875 nm stimuli, where both BP and Ge absorb incident photons, leading to interface recombination. (b) Measured dark current and photocurrent under 520 nm laser irradiation. (c) Photocurrent response as a function of polarization angle under 520 nm laser illumination, with experimental data and fitted curves highlighting the anisotropic optical properties of the heterojunction. (d) Photogenerated carrier transport properties and energy band structure for wavelengths 1875 nm < λ < 3000 nm, where only BP absorbs photons and carrier recombination within these wavelength ranges is observed. The built-in electric field is directed towards Ge. (e) Dark current and photocurrent measurement under 2000 nm laser irradiation, showing an increase in open-circuit voltage with increasing laser power. (f) Photocurrent mapping under mid-wavelength infrared irradiation, revealing light absorption within the heterojunction region. (g) The photogenerated carrier transport properties and energy band structure under 3000 nm < λ < 4000 nm stimuli. The interface recombination is inhibited and the built-in electric field reverses, pointing towards BP. (h) Measurement of dark current and photocurrent under 4000 nm laser irradiation. (i) The α-dependence, derived from the relationship between optical power and photocurrent I∝Pα, across the operation wavelength. Higher values in the bandwidth where the energy band is reversed suggest reduced carrier recombination compared to traditional cases in (a) and (d).
To thoroughly evaluate the optoelectrical performance of the BP/Ge heterojunction, we began with I-V measurements to assess contact formation. Ideal ohmic contact formation, as shown in Supplementary Figure S6, is essential for observing subtle changes in the energy band structure. Building on this foundational understanding, we explored the optical response of the BP/Ge heterojunction under laser irradiation at 520 nm, 2000 nm, and 4000 nm. Figure 2a details the transport behavior of photogenerated carriers under 520 nm laser irradiation. With photon energy exceeding the bandgaps of both Ge and BP, both materials absorb the incident photons, leading to photogenerated carriers. High-energy photons enable hot carriers to overcome the potential barrier at the heterojunction interface and transport along the built-in electric field direction. Notably, the bilateral generation of photogenerated carriers results in substantial carrier recombination at the interface.
As depicted in Fig. 2b, the dark current and the photocurrent following 520 nm laser irradiation are presented, maintaining Ge as the anode and BP as the cathode. The direction of the photogenerated electric potential shift with increasing optical power confirms that the built-in electric field directs from BP to Ge, aligning with typical energy band design. The polarization-resolved performance at zero bias voltage is shown in Fig. 2c, with photocurrent mappings at various polarization angles (see Supplementary Figure S7) confirming that the polarized response originates from the heterojunction. The photocurrent decreases significantly as the angle shifts from 0° to 135°, reaching a minimum at 90°, and then gradually increases, consistent with BP's anisotropic properties.
For wavelengths in the range 1876 nm to 3000 nm, the energy exceeds the bandgap of BP but is less than that of Ge. Figure 2d displays the transport behavior of photogenerated carriers after irradiating the device within this wavelength range. Here, only BP unilaterally produces photogenerated carriers, reducing recombination probability compared to the bilateral absorption seen at shorter wavelengths. We further measured the dark current and photocurrent under 2000 nm laser irradiation as depicted in Fig. 2e. From the photogenerated electric potential, it is clear that the built-in electric field still points from BP to Ge. Verification that this built-in electric field resides in the heterojunction region was achieved through mid-wavelength optical current mapping, as shown in Fig. 2f.
These results for wavelengths less than 3000 nm, where photon energies exceed the combined bandgap of Ge and BP (Eg(BP) + 0.06 eV), confirm the alignment between the measured photogenerated electric potential, photocurrent, and the built-in electric field, thereby validating our device design and theoretical predictions. To elucidate the unique characteristic of light-induced energy band inversion, we extensively investigated the photoresponses at wavelengths greater than 3000 nm. Figure 2g illustrates the inversed energy band induced by the lower photon energy between 3000 nm and 4000 nm. In this scenario, the built-in electric field reverses and, remarkably, the recombination of dark carriers is inhibited, providing a feasible pathway towards ultrahigh sensitivity by suppressing interface recombination.
Figure 2h displays the dark and photocurrents under 4000 nm laser irradiation. As expected, the photogenerated electric potential shifts to negative, distinctly different from the positive responses at shorter wavelengths. This anomalous behavior is primarily due to the inability of low-energy photons (less than 0.4 eV) to surpass the potential barrier at the BP/Ge valence band interface, resulting in charge accumulation and subsequent inversion of the energy bands. Unlike previous configurations, the inversed energy band configuration prevents photogenerated hole transport from crossing the interface, as depicted in Fig. 2g, significantly reducing the likelihood of photogenerated carrier recombination at the interface.
We also measured the short-circuit current (Isc) as a function of incident power (P) of the device at different wavelengths, with details provided in Supplementary Figure S8. In Fig. 2i, we plot the retrieved power index relative to the irradiated wavelength range, revealing three distinct regions: BP and Ge absorption, BP absorption with normal energy band, and BP absorption with inversed energy band. Consistent with the analysis above, the inversion region exhibits the largest index, confirming that photogenerated carrier recombination is lowest in this region. This provides us with a feasible route to achieve ultrahigh sensitivity that approaches the background limit between 3000 to 4000 nm at room temperature.
To further assess the infrared light-induced energy band inversion characteristic, Fig. 3a displays I-T curves under various wavelength illuminations ranging from 2700 to 4200 nm. The device demonstrates a robust negative photodetection capability for all stimuli wavelengths exceeding 3000 nm (E = 0.413 eV), defined as the inversion wavelength λR. This negative photoresponse is a result of the sophisticated energy band design of the device, particularly the barrier in the valence band. When photon energy exceeds 0.413 eV, photogenerated carriers can surmount this barrier and are efficiently collected by the electrodes. Conversely, when photon energy is below 0.413 eV, photogenerated holes accumulate at the apex of the valence band, unable to cross the barrier, thus forming an inversion layer at the interface and reversing the built-in electric field within BP. This energy threshold is almost identical to the sum of Eg(BP) and ΔEV (0.34 eV + 0.061 eV = 0.401 eV). Notably, no photocurrent generation is observed at 4200 nm, aligning with the cutoff bandgap of BP as confirmed by photoluminescence testing (Supplementary Figure S9).
For heterojunctions with lower doping concentrations of Ge, the energy band inversion point λR is notably lower than in higher doping concentrations. As depicted in Fig. 3b, simulated wider space charge regions in these lower doping concentration heterojunctions favor charge accumulation, reducing tunneling likelihood and resulting in a blue-shifted λR . To verify this, a heterojunction with a lower doping concentration of Ge (6.8×1014 cm− 3) was fabricated. The results, shown in Fig. 3c, confirm a blue shift of λR to 1950 nm, consistent with our theoretical predictions.
The presence of phenomena such as photothermoelectric effects and interfacial defect states can lead to the reverse transport of photogenerated currents. Given that the response times associated with these effects typically exceed milliseconds, to eliminate this possibility, we measured the device's response time under MWIR stimuli as illustrated in Fig. 3d, finding rise and fall times of 396 µs and 446 µs, respectively. Thus, the influence of both effects can be confidently excluded. A more detailed discussion is provided in Supplementary Note 2.
Furthermore, we simulated the evolution of the valence band energy (Ev) of the heterojunction at various temperatures, as shown in Fig. 3e. The potential barrier difference decreases with increasing temperature, which, coupled with rising intrinsic excitation, leads to a nonlinear increase in carrier transport. Figure 3f presents the I-V curve across varying temperatures, with the inset illustrating the temperature-dependent variation of Ge's bandgap. Figure 3g outlines the photocurrent's dependency on temperature and bias voltage, with the dashed line indicating the current contours. This nonlinear growth in current with temperature highlights the thermal stability of the device. Additionally, we measured the energy band reversal point at 220K, finding that λR occurs at 2900 nm, corroborating our predictions (see Supplementary Figure S12).
To illustrate the pivotal role of light-induced energy band inversion in enhancing detection capabilities up to the background limit, we conducted photoresponse measurements under blackbody illumination at various temperatures, as shown in Fig. 4a. The device exhibited significant responsiveness, with photocurrent increasing consistently as the blackbody temperature rose. Additionally, as depicted in Fig. 4b, we assessed the low-frequency noise of the device without any bias. Under a background illumination of 300K, the system noise overlapped with the device noise, registering less than 8 fA⋅Hz− 1/2, indicative of the device's ultra-low noise level.
With the measured photocurrent and noise current, we calculated the detectivity and responsivity of the device to the blackbody at different temperatures at 0 V, shown in Fig. 4c. The blackbody responsivity Rbb is defined as \({\text{R}}_{bb}={I}_{ph}/{P}_{bb}\), where \({P}_{bb}\) is the effective incident blackbody radiation power, calculated by \({P}_{bb}=\frac{\sigma \left({T}_{b}^{4}-{T}_{0}^{4}\right){A}_{b}{A}_{d}}{2\sqrt{2}\pi {L}^{2}},\) with \(\sigma =5.67\times {10}^{-12}W\cdot {cm}^{-2}\cdot {K}^{-4}\). Here, Tb, T0, Ab, Ad, and L represent the blackbody temperature, room temperature, exit area of the blackbody, photosensitive area, and distance from the detector to the blackbody, respectively. The detectivity D* can be calculated with \({D}^{*}=\frac{\sqrt{{A}_{d}{\Delta }f}}{NEP}.\) \(NEP=\frac{{I}_{noise}}{{\text{R}}_{bb}}\) and\({ I}_{noise}=\sqrt{2qI\varDelta f+\frac{4kT}{{R}_{shunt}}\varDelta f}\), where q is the elementary charge, I is the average dark current, k is the Boltzmann constant, T is the temperature and Rshunt the shunt resistance, \(\varDelta f\) is the bandwidth. The calculated device noise is about 8 fA⋅Hz−1/2which aligns with the noise current measured in Fig. 4b. The measurement system schematic is detailed in Supplementary Figure S13. For accurate measurements, we utilized a custom HgCdTe detector with a similar photosensitive area, operating in the MWIR for noise and responsivity calibration. Notably, both detectivity and responsivity increase as the blackbody temperature rises, due to the central wavelength moving into the mid-wavelength regime and an increase in total radiation energy from 700K to 1150K.
We also measured the absorption spectra of the devices using a home-built MWIR monochromator spectroscopy system; a schematic of the testing equipment is presented in Supplementary Figure S14. The noise equivalent photon (NEPh) is a crucial metric for evaluating the device’s ability to detect weak signals. As shown in Fig. 4d, we calculated the responsivity and NEPh of the heterojunction across different wavelengths using the formula: \(NEPh=\frac{\lambda *NEP*{T}_{int}}{h*c}\), where λ is the incident laser wavelength and 𝑇int is the integration time, selected as the falling edge of the device response at 446 µs. At 3.5 µm, the NEP reaches the femtowatt order of magnitude, indicating that the device can detect a minimum of approximately 50 photons, demonstrating exceptional sensitivity.
In Fig. 4e, we compare the peak detectivity and dark current density performance of our device with reported works and commercial products, illustrating that our device achieves the highest detectivity and lowest dark current density among MWIR detectors. This underscores that the light-induced band inversion effect significantly enhances the detection capabilities to ultimate sensitivity levels.
Finally, as shown in Fig. 4f, the BLIP curve calculated for a field of view of 2π at a background temperature of 300K reveals that our device achieves record-high detectivity, touching the BLIP limit in the infrared range (3460 to 4000 nm) where energy band inversion occurs. The peak detectivity is an impressive 9.81×1011 cm⋅Hz1/2⋅W− 1 at 3460 nm. It is noteworthy that detectivity rapidly increases past λR. Consistent with prior analyses, minimizing photogenerated carrier interfacial recombination effectively enhances the detectivity of the device. This achievement sets a new benchmark in the field, showcasing our heterojunction's potential to attain ultra-limit sensitivity unmatched by conventional photoelectric or photothermal materials.
Given the wavelength-dependent energy band inversion characteristic of our device, it demonstrates a natural resilience to background interference. This attribute allows the device to exhibit an interference-resistant optical response, capable of delivering both positive and negative outputs across different wavelengths. Such dual photoresponse capability is particularly advantageous in environments subjected to light pollution or sunlight, enabling the acquisition of more comprehensive infrared radiation information about the target through vectorial responsivity. This is a significant enhancement over conventional photovoltaic devices that are limited to capturing intensity information alone.
The time-resolved photocurrent measurements at 4000 nm under dark conditions (Fig. 5a) and amidst variations in background visible light (Fig. 5b) demonstrate stable negative and positive responses, respectively, illustrating the device's robust performance under fluctuating light conditions.
To assess the practical imaging capabilities of our BP/Ge heterojunction, we conducted passive imaging tests at room temperature and zero bias using a home-built scanning imaging system (details in Supplementary Figure S15). Preliminary tests included partially obscuring a carbon heat pipe with a four-inch silicon wafer. The commercial Si CCD image shown in Fig. 5c serves as a reference for comparative analysis. The infrared passive images captured by the BP/Ge detector, displayed in Fig. 5d, clearly delineate the grain structure of the carbon tube. These images underscore the high-resolution imaging capabilities of the device, confirming its potential to resolve thermal details with remarkable clarity."
To further explore the wavelength-selective imaging capabilities of our heterojunction, we utilized a 2-2.5 µm bandpass filter to isolate the detector from extraneous wavelength interferences. The positive photoresponse imaging results, shown in Fig. 5e, precisely delineate the contours of the observed objects. These results highlight the precision and effectiveness of the heterojunction in capturing detailed thermal images within this specific wavelength range.
For MWIR passive imaging, we utilized a 3–5 µm bandpass filter, as illustrated in Fig. 5f. Notably, the negative photoresponse was slightly more pronounced than the positive, attributed to the higher radiation power emanating from an 800 K target within the 3-4.2 µm bandwidth (details in Supplementary Figure S16). This distinctive behavior further underscores the unique capabilities of our device, demonstrating its advanced response characteristics across different infrared bands and its potential for sophisticated thermal imaging applications.