Adjustable direct bandgaps of nitrides, typically in the AlGaN alloy, range from 3.4 to 6.2 eV [1,2]. Electron transitions in AlGaN induce the light absorption edge at wavelengths of 200–364.7 nm, rendering it a suitable material system for manufacturing ultraviolet devices. To date, AlGaN-based photosensitive devices have exhibited significant potential in diverse applications such as secure optical communication, missile flame warnings, and outdoor fire monitoring. This is attributed to their advantageous characteristics including a wide bandgap, exceptional chemical/thermal stability, high radiation resistance, and well-established growth techniques [3–5]. However, the relatively low conductivity and carrier mobility of AlGaN impose limitations on the photocarrier collection efficiency within a confined carrier lifetime. For a basic photodetector employing a bare AlGaN layer with a lateral metal-semiconductor-metal (MSM) device configuration, the photosensitive area is typically restricted to several hundred square microns, resulting in responsivity at the magnitude of mA/W [6,7]. Otherwise, the vertical AlGaN/GaN junction offers a pathway to achieve a high-gain photodetector through interfacial two-dimensional electron gas (2DEG) [8–10]. However, the generation of a highly concentrated 2DEG requires the direct growth of a thick AlGaN layer with elevated Al content on the GaN epilayer. Nevertheless, attaining a high-quality epitaxial interface remains challenging due to their difference in growth temperature and lattice constant [11]. Additionally, the stress relaxation accompanied with an increase in epilayer thickness diminishes the piezoelectric polarization effect, thereby consuming the concentration of 2DEG [12].

To overcome the inherent limitations of AlGaN materials, heterojunction structures were constructed by incorporating other conductive materials, mainly including metal nanowire [13,14], indium tin oxide (ITO) [15–17], and 2D graphene [18,19], onto the AlGaN epilayer. This device configuration possesses two additional functionalities: one is promotion of the heterojunction built-in electric field for efficient separation of photocarriers; the other is enhanced carrier collection efficiency by the conductive channel over the AlGaN layer [20]. Subsequently, the designed AlGaN heterojunction devices exhibit improved response capabilities. Despite the low sheet resistance and high optical transparency of the metal nanowire, their widespread use of copper or Ag metals is hindered by oxidation in air, leading to inadequate long-term stability while working [21,22]. Generally, ITO exhibits chemical inertness to oxygen and moisture, but it presents significant optical absorption in the ultraviolet region, thereby impeding the light utilization of the AlGaN layer underneath [23]. 2D graphene, with atomic thickness, has shown uniformly high transmittance across a wide range of wavelengths from near-ultraviolet to infrared. Furthermore, it also exhibits ultrahigh carrier mobility, good environmental stability, and favorable mechanical flexibility [24]. Due to the absence of dangling bonds on the 2D graphene surface, the interface effect between graphene and AlGaN is governed by a weak van der Waals (vdWs) force, resulting in an optimal junction interface. These unique characteristics of the graphene conductive layer render it a favorable partner for the fabrication of AlGaN-based heterojunction photodetectors.

There are two prevalent device configurations for the ultraviolet photodetector based on 2D graphene/3D AlGaN junctions [25]. The lateral structure is characterized by a pair of metal electrodes deposited on the graphene, which serve as a photocarrier collection channel and are driven by the entire bias voltage [26,27]. Due to the high conductivity of the graphene channel, it is possible to significantly increase the distance between paired electrodes. Tian et al. reported a hybrid graphene/GaN phototransistor, which has a photosensitive area of up to 15.2mm2 and obtains a responsivity of 361 mA/W at a bias voltage of 10 V [28]. In addition, the other vertical structure is formed by the deposition of two separate metal electrodes on graphene and AlGaN, with a bias voltage applied to the 2D graphene/3D AlGaN vdWs junction [29]. However, the vertical structure typically necessitates the formation of 3D AlGaN mesas, thus requiring intricate dry etching, or separation techniques to obtain free-standing nitride epilayers [30,31]. Aiming to mitigate these fabrication challenges, a dielectric layer of HfO2 or h-BN is employed as an inserting layer to spatially isolate graphene and AlGaN, which facilitates the efficient separation and collection of photogenerated electron-hole pairs [32,33]. Hence, Li et al. demonstrated the high-sensitivity ultraviolet photodetector based on a vertical graphene/nanoporous GaN heterojunction, with SiO2 serving as the inserting layer. The device exhibits a responsivity of ∼101A/W and detectivity of 1017 Jones at a bias voltage of −1.5V [34]. However, the operation capacity for vertical graphene/AlGaN photodetectors at zero bias is rarely addressed thus far, despite its crucial effect in the self-powered detection. Besides, the comprehensive investigation and comparison between these two device configurations are still lacking, leaving their potential applications unclear.

In this study, two representative photodetectors based on the graphene/AlGaN (Gr-AlGaN) vdWs junction in the typical lateral and vertical device configurations are simultaneously fabricated and compared. The paired metal electrodes are designed into the ringlike shape, facilitating the efficient photocarrier collection and enabling the direct fabrication of the vertical photodetector with no need for a dielectric inserting layer. The disparity in photodetection mechanisms between these two Gr-AlGaN photodetectors is presented, proving the photocarrier transport behavior is influenced by the application of bias voltage on the junction electric field whether or not. Due to the ultrahigh carrier mobility of the graphene channel, the high gain of the lateral photodetector contributes to a maximum responsivity of 1.27×104A/W and detectivity of 3.88×1012 Jones, but it suffers from a prolonged response time of several seconds. In contrast, although the vertical photodetector shows lower responsivity and detectivity (2.61×10−2A/W and 1.34×1011 Jones), it presents a significantly improved response speed and self-powered detection capability. The imaging of the corresponding photodetector array proves that the vertical Gr-AlGaN device is satisfied in specialized working environments, such as deep space. Conversely, the vertical Gr-AlGaN device has exhibited a consistently enhanced photocurrent along with light illumination time, indicating its great potential for biomimetic visual imaging.

The growth of graphene on copper foils was achieved through chemical vapor deposition (CVD). Prior to graphene growth, the copper foils were pre-treated by a diluent HCl solution and high-temperature annealing (1060°C for 3 h) in H2. These treatments aim to eliminate the surface oxidation layer and enhance the crystalline quality of copper foils, respectively. During the graphene growth, the temperature of the CVD system was raised to 1000°C with a flow rate of 5 sccm H2 and 100 sccm Ar. Subsequently, a carbon source consisting of 1–3 sccm CH4 was input for a duration of 20 min. After that, the CVD system was subjected to a gradual cooling process until it decreased to room temperature. The unintentionally doped (UID) AlGaN was epitaxially grown using metal organic chemical vapor deposition (MOCVD), while a 200 nm annealed AlN template on a sapphire substrate was employed. The high-temperature AlN layer with a thickness of 450 nm and AlN/AlGaN superlattice of 130 nm in 10 loops was initially grown using MOCVD to ensure superior crystalline quality of AlGaN in the subsequent growth. TMAl, TMGa, and NH3 were employed as precursor materials. Then, an AlGaN epilayer with a thickness of 450 nm was grown at a temperature of 1125°C for 30 min, and the flow rates of TMAl, TMGa, and NH3 were set as 120 sccm, 20 sccm, and 3450 sccm.

The Raman spectra and intensity mapping of a typical band for the graphene and AlGaN layer were acquired using a Raman spectrophotometer (LabRAM HR Evolution, HORIBA) with a 532 nm laser at an intensity of 25%. For Raman mapping, the scanning step was set to 0.5 μm. The photoluminescence (PL) spectra of AlGaN were measured using a home-made system with a 266 nm laser as the excitation source, while other main components were purchased from Zolix Co., Ltd. The X-ray diffraction (XRD) instrument (D8 Discover, Bruker) was utilized to test the crystalline properties of AlGaN. The device configurations were examined by the optical microscope (DSRi2, Nikon).

To fabricate photodetectors based on the 2D graphene/3D AlGaN junction, the graphene grown on copper foils was initially transferred onto the AlGaN epilayer by using a previously reported wet transfer strategy, wherein S1805G polymer was employed as a supporting mediator [35]. After the first-time lithography created a patterned polymer mask on graphene, plasma with air flow was used for graphene etching under the conditions of 50% power intensity for 5 min. After that, the polymer mask was removed in 85°C hot acetone. The second-time lithography was applied at a specific position to determine the paired metal electrodes, followed by a traditional lift-off process. In the present case, the paired metal electrodes were designed into a ringlike shape, with sequential deposition of 30 nm nickel and 30 nm gold by electron beam deposition under a basic chamber pressure of 5×10−4Pa and growth rate of 1 Å/s.

The current-voltage (I–V) curves of the photodetector were measured by using a semiconductor analyzer (Metatest E2, MetaTest). A xenon lamp was employed as the light source, and the light power was calibrated with a dynamometer. For the measurement of the current-voltage-time (I–V–T) curve, a shutter in front of xenon lamp controlled the light switch. For the spectral response measurement, the light wavelength was selected by a monochromator. The target picture for 3×3 device array imaging was generated by the light illumination through a defined optical mask, and the I–V–T output of the photodetector array was collected one by one.

In addition to the designed lateral Gr-AlGaN and vertical Gr-AlGaN photodetectors, the Gr-only device is also fabricated as a reference to verify the contribution of graphene to the light response of Gr-AlGaN junctions, as schematically shown in Figs. 1(a)–1(c). The paired metal electrodes are arranged into a ringlike shape, with the outer electrode circling the inner electrode, thereby ensuring an optimized distribution of bias voltage within the photosensitive region. For the Gr-only photodetector, the graphene is transferred onto the bare sapphire with paired metal electrodes both deposited on it, as depicted in Fig. 1(a). The electrode position resembles that of the lateral Gr-AlGaN device, except in this case, the graphene is prepared on an AlGaN epilayer instead of sapphire as shown in Fig. 1(b). The vertical Gr-AlGaN device is achieved by depositing the inner metal electrode on the patterned graphene, while the outer is deposited on the AlGaN epilayer. Thanks to the ringlike electrode design, there is no need for a dielectric inserting layer at the interface of the Gr-AlGaN junction, as illustrated in Fig. 1(c). The corresponding locally enlarged optical microscope images of these three types of devices are shown in Figs. 1(a)–1(c), revealing the distribution of graphene between paired metal electrodes.

The relative position of graphene and AlGaN in the above three devices is further evaluated by Raman mapping in Figs. 1(d)–1(f). The distributions for the typical E2 band intensity of AlGaN (red color) and 2D band intensity of graphene (green color) are simultaneously measured within a defined region, indicated by their schematic cross-sectional structures and optical microscope images in Figs. 1(a)–1(c). The Raman 2D band intensity of graphene in the Gr-only photodetector exhibits a continuous distribution, while it is absent in the typical AlGaN, as shown in Fig. 1(d). For the lateral Gr-AlGaN photodetector in Fig. 1(e), both graphene and AlGaN are observed. In contrast, the edge of patterned graphene within paired metal electrodes is clearly shown in the vertical Gr-AlGaN photodetector, proving the successful patterning of graphene by air plasma, as depicted by Fig. 1(f). Notably, the Raman mapping of graphene has several bright spots that correspond to the nucleation sites of the graphene adlayer as it is initially grown on copper foils.

The Raman spectra presented in Fig. 2(a) provide a detailed confirmation of the distribution and properties of AlGaN and graphene. The typical E2 band of AlGaN is located at a wavenumber of 650.8cm−1, exclusively appearing in Gr-AlGaN junctions. The typical D, G, and 2D bands of graphene are observed. Notably, the intensity of the defect-related D band is extremely weaker compared to that of the G and 2D bands, demonstrating the exceptional graphene crystalline quality. Besides, the monolayer nature of graphene could be revealed by the intensity ratio of 2D to G band, which is close to 2. As for the patterned graphene within Gr-AlGaN junctions, the negligible intensity of the D band confirms that the patterning process minimally induces additional defects in graphene. The Raman bands associated with graphene are not observed at vertical Gr-AlGaN 2, as shown in Raman mapping, further indicating the effective removal of graphene by air plasma.

The PL spectra of AlGaN are presented in Fig. 2(b), revealing a prominent emission peak at 293.5 nm, which corresponds to an Al content of ∼0.43. The emission peak exhibits a full width at half-maximum (FWHM) of ∼10.9nm, indicating a high level of homogeneity in the Al content [36]. Meanwhile, the excitation source of the 266 nm laser also appears in the PL spectra. As depicted in Fig. 2(c), the XRD rocking curves (RCs) of AlGaN along (002) and (102) orientations demonstrate its crystalline quality. The FWHMs for the (002) RC and (102) RC are measured as 783.2 and 963.3 arc sec, in which the total dislocation density is calculated to be 6.1×109cm−2 [37].

The I–V curves of three photodetectors based on Gr-only, lateral Gr-AlGaN, and vertical Gr-AlGaN junctions are plotted in Figs. 3(a)–3(c); the black and red lines are measured under dark (Idark) and light (Ilight) conditions (polychromatic light from xenon lamp with power density of 421.6μW/cm2). The green lines represent the photocurrent (Iph) plotted against the bias voltage, which is determined by subtracting the Idark from Ilight. Due to the high conductivity of graphene, the Gr-only photodetector exhibits a maximum current of 21.7 mA at the bias voltage of 2 V, as depicted in Fig. 3(a). Additionally, the I–V curves exhibit high symmetry with respect to zero bias, which is in line with the symmetrical configuration of source (S) and drain (D) metal electrodes on the graphene, and the linear increase in current further confirms the ohmic contact characteristic. The photocurrent of the Gr-only photodetector is merely 0.71 mA (at 2 V), indicating a low Iph to Idark ratio (Iph/Idark) of 0.033, which can be attributed to the limited light absorption efficiency of the Gr-only device. The features of I–V curves for the lateral Gr-AlGaN photodetector in Fig. 3(b) show similarity to those of the Gr-only device in Fig. 3(a). However, an obvious enhancement in photocurrent at 2 V is observed, reaching 4.5 mA and resulting in an increased Iph/Idark of 0.184. The enhanced response of the Gr-AlGaN device is attributed to the higher light utilization efficiency achieved by the thicker 3D AlGaN material, which would be further elaborated on later. To be different, the dark current of the vertical Gr-AlGaN photodetector in Fig. 3(c) decreases to 2.2×10−7mA, which exhibits a reduction of over 108 lower compared to the lateral Gr-AlGaN one, thereby confirming that the entire bias voltage applied on the graphene channel results in a relatively high dark current. The asymmetrical I–V curve of the vertical Gr-AlGaN device exhibits typical rectifying characteristics, and a remarkably high Iph/Idark of 1.59×105 is achieved at zero bias voltage, demonstrating its suitability for self-powered detection.

The rise time (τr) and fall time (τf) of these devices are determined based on their I–V–T curves in Figs. 3(d)–3(f), corresponding to the transition between 10% and 90% of the maximum current signal. For Gr-only and lateral Gr-AlGaN photodetectors, the bias voltage of 2 V is applied, while the vertical Gr-AlGaN photodetector is measured under zero bias to assess its self-powered detection capability. The τr/τf of Gr-only (4.46 s/3.99 s) and lateral Gr-AlGaN (3.54 s/1.31 s) photodetectors exhibit much longer response time compared to the vertical Gr-AlGaN device (70 ms/200 ms), indicating that the light response mechanism is distinct for Gr-AlGaN junctions in different configurations.

In our previous studies, we have demonstrated that the as-fabricated graphene exhibits p-type doping [35,37], thereby resulting in hole conduction within the graphene channel. Under the light illumination, the scattering effect between the original free carrier and photocarrier in graphene decreases its conductivity [38], representing the underlying response mechanism for the Gr-only photodetector, as depicted in Fig. 3(g). The negative photoconductivity (NPC) effect (Ilight<Idark) shown in Fig. 3(a) is consistent with the proposed response mechanism [39]. On the other hand, due to the limited light absorption of monolayer graphene (0.23% at 532 nm), the Gr-only photodetector exhibits a relatively low photocurrent. As for Gr-AlGaN junctions, the integration of graphene with AlGaN immediately establishes a junction electric field (E) at their interface, as depicted in Fig. 3(h). Considering that the photocurrent of the lateral Gr-AlGaN photodetector is almost one order of magnification higher than that of the Gr-only one, AlGaN plays a pivotal role in enhancing light response. Apart from the scattering effect in graphene, the photon absorption of AlGaN generates electron-hole pairs that are subsequently separated by the junction electric field at the junction, with electrons being driven towards graphene. In addition, the 3D AlGaN with a thickness of 450 nm facilitates efficient light absorption in the ultraviolet region, which leads to abundant electron-hole pair generation. The recombination between original majority holes in graphene with the drifted electrons further decreases the conductivity, thereby reinforcing the NPC effect with a much higher photocurrent of 4.5 mA in Fig. 3(b). The presence of trap states within the Gr-AlGaN junction results in the enhanced photocarrier lifetime, followed by the persistent photoconductive effect and high gain. Moreover, this phenomenon exhibits a response time that can extend up to several seconds [40]. However, the application of proper bias voltage across the graphene channel is necessary for electrical signal collection in this device structure, rendering the self-powered detection unattainable for this lateral Gr-AlGaN photodetector.

With another device configuration, the vertical Gr-AlGaN photodetector is characterized by separately depositing S and D metal electrodes on graphene and AlGaN, as illustrated in Fig. 3(i). Same as the lateral Gr-AlGaN device, photons with energy higher than the AlGaN bandgap are absorbed, leading to electron transition from the valence band to conduction band and remaining holes in the valence band. This dominant response process is more favored over the weak light absorption of graphene. Specifically, the photogenerated electron-hole pair is separated by the junction electric field and subsequently drifted to the graphene (electron) and AlGaN (hole), respectively. When the external circuit is open, the nonequilibrium carrier accumulation contributes to a photogenerated electric field that opposes the electric field at the Gr-AlGaN junction, resulting in an open-circuit voltage of −0.21V, measuring between S and D electrodes, as indicated by I–V curves in Fig. 3(c). If the vertical Gr-AlGaN photodetector is converted into a closed circuit externally, the photocarriers would traverse this path to generate photocurrent. In this process, electrons flow back into the AlGaN, aiming to restore the potential equilibrium between the graphene and AlGaN. Hence, the vertical Gr-AlGaN photodetector demonstrates its capability for self-powered detection at zero bias voltage. Additionally, it is worth noting that the high-resistivity AlGaN crucially participates in the photocarrier transportation, resulting in an extremely low dark current compared to the lateral Gr-AlGaN device. Furthermore, the limited carrier mobility of AlGaN also hinders the persistent photoconductive effect observed in lateral Gr-AlGaN photodetectors, and it is responsible for the reduced photocurrent and enhanced response speed.

Upon the limited response of graphene contributing to the Gr-AlGaN devices, the characteristics of lateral Gr-AlGaN and vertical Gr-AlGaN photodetectors are primarily studied in the following sections, which operate at the bias voltage of 2 V and 0 V. The spectral response for these two photodetectors, as depicted in Fig. 4(a), exhibits a peak photocurrent at around 300 nm, which closely aligns with the PL emission peak of AlGaN in Fig. 2(b). Along with the increase in incident light wavelength from 300 nm to 330 nm, there is a rapid decline in the photocurrent, resulting in a large ratio of 23.4 (2.57 mA to 0.11 mA) and 169.8 (1.8×10−5mA to 1.06×10−7mA) for the variation of the photocurrent in lateral Gr-AlGaN and vertical Gr-AlGaN devices. The above results confirm that the light response in Gr-AlGaN devices primarily comes from AlGaN, irrespective of device configurations.

The responsivity (R) and detectivity (D*) of the photodetector are calculated by the following equations: R=IphP×A,(1)D*=R×A1/2(2q×Idark)1/2,(2)where P is the power density of incident light, A is the photosensitive area (8.46×10−4cm2), and q is the elementary charge. The responsivity and detectivity as a function of light power density are depicted in Fig. 4(b), wherein the response performance exhibits a decline with increasing light power density. This phenomenon is attributed to the attenuation of the junction electric field caused by the drift of photogenerated electrons and holes, consequently leading to the reduced separation efficiency for the subsequent photocarriers [40]. With the lowest light illumination power density of 78.9μW/cm2, the lateral Gr-AlGaN photodetector obtains a maximum responsivity of 1.27×104A/W and detectivity of 3.88×1012 Jones, which are nearly six orders and one order of magnitude higher than that of the vertical Gr-AlGaN device (2.61×10−2A/W and 1.34×1011 Jones). As depicted in Fig. 3(h), the significantly improved device response performance is attributed to the high photocarrier gain on the lateral graphene channel. The comparison of the device response performance with recently reported results is presented in Table 1, and the configuration design of the Gr-AlGaN van der Waals junction in this work achieves better detection ability than previous similar photodetectors based on Gr-(Al)GaN junctions.Table 1.

Comparison of Photodetector Performance for Those Based on Gr-(Al)GaN Junctions

As depicted in Fig. 5(a), the 3×3 photodetector array is fabricated and employed for imaging. To generate the target symbol of “+”, an optical mask was positioned in front of light source. The states of each pixel in the photodetector array, whether dark or light, are depicted in Fig. 5(b), which is consistent with the “+” symbol of the optical mask in Fig. 5(a). To determine the potential application scenarios, the vertical Gr-AlGaN photodetector array was operated at zero bias voltage, while the lateral Gr-AlGaN device works at the bias voltage of 2 V based on their respective advantages illustrated in Fig. 3. The schematic diagrams of external circuit connections for these two devices are depicted in Fig. 5(c).

The vertical Gr-AlGaN photodetector array in Fig. 5(d) demonstrates a distinct symbol of “+” with high contrast in its real imaging. The photocurrent of illuminated pixels is distributed uniformly at the magnitude of 10−6mA, ranging from 4.2×10−6mA to 9.6×10−6mA, which remains stable throughout the whole duration of light illumination. The self-powered detection capability devoid of any external driving power demonstrates its immense potential for light detection in some specialized environments, such as deep space and other untraversed regions. As depicted in Fig. 5(e), the lateral Gr-AlGaN photodetector array also achieves clear imaging of “+” symbol with the photocurrent reaching a few mA through external power driving. Meanwhile, the response strength continuously increases with the prolonged light illumination time. In detail, the photocurrent is measured as 1.62 mA at 15 s, and it is enhanced by ∼31% (2.12 mA) at 40 s. This detection feature renders the lateral Gr-AlGaN photodetector suitable for steady object recognition, thereby satisfying low light emission targets through delayed acquisition time. Moreover, the progressive image sharpness exhibits similarity to the cognitive process of a human being, which is featured with initial recognition (for the first few seconds, dramatic current onrush) and long-term memory (for the longer time, slow current variation to be stable), offering potential applications in biomimetic visual imaging [41–43].

In conclusion, the impact of 2D graphene/3D AlGaN junction configuration design on the device response characteristics of Gr-AlGaN ultraviolet photodetectors is comprehensively investigated. The paired metal electrodes with ringlike shape ensure efficient collection of photocarriers, also enabling the fabrication of a vertical Gr-AlGaN photodetector without the need of any dielectric inserting layer. It is proved that the electron transition in the AlGaN through photon absorption contributes to the primary response of Gr-AlGaN devices. The lateral Gr-AlGaN photodetector achieves a high responsivity of 1.27×104A/W and detectivity of 3.88×1012 Jones, which is attributed to the persistent photoconductive effect observed in a lateral graphene channel within paired electrodes. Since the paired electrodes are irrelevant to the junction electric field in this device configuration, it is imperative to apply a bias voltage on the graphene channel for the collection of the photocarrier. In the case of a vertical Gr-AlGaN photodetector, where the paired metal electrodes are separately deposited on the graphene and AlGaN, and then the photogenerated electron-hole pairs are efficiently separated by the junction electric field and accumulate at the opposite sides of space charge region, this results in the generation of an open-circuit voltage. When the external circuit is connected, the photocurrent can be induced at zero bias. Although the response performance is inferior to that of the lateral Gr-AlGaN device, it shows the unique capability of self-powered detection. Based on the imaging measurement, the potential applications of these two Gr-AlGaN photodetectors are suggested. This study has significance for advancing the practical implementation of photosensitive devices with 2D graphene/3D AlGaN vdWs junctions.