Highly Sensitive Ultraviolet Photodetector Based on an AlGaN/GaN HEMT with Graphene‐On‐p‐GaN Mesa Structure

The advantageous role of 2D electron gas presence at the AlGaN/GaN interface attracts huge interest in the field of GaN‐based ultraviolet photodetector technology. However, the presence of high dark current deteriorates the photodetector performance by diminishing several figures of merit. In this work, enhanced figures of merit are demonstrated by employing interdigitated p‐GaN finger structure on the top of the AlGaN/GaN heterostructure. The commonly present high dark current in p‐GaN/AlGaN/GaN planar photodetector is largely reduced (from ≈µA to few pA) by etching the p‐GaN, excluding the electrode region. Furthermore, by using a graphene transparent electrode along with the p‐GaN interdigitated fingers on AlGaN/GaN heterostructure, ultraviolet photodetectors with superior sensitivity (3.55 × 106) and ultrahigh detectivity (1.91 × 1014 cm Hz1/2 W−1) are realized at 360 nm. A comparison of graphene/p‐GaN and Ni/Au/p‐GaN interdigitated fingers and planar p‐GaN (with interdigitated graphene contacts) all on AlGaN/GaN heterostructure allows to understand the dominant roles of electrode transparency and the heterojunction structure. The simple and high electron mobility transistor‐compatible fabrication process of UV detectors provides a unique application in the field of UV sensing technology.

in highly responsive UV photodetectors. [23][24][25][26][27][28][29][30] The formation of 2-DEG with high carrier density and electron mobility due to the piezoelectric and spontaneous polarization without any intentional doping has frequently been referred to as a potential material for the next generation high power and high-frequency devices. [22,23] The addition of a p-GaN layer to the AlGaN/GaN changes the resulting p-GaN/AlGaN/GaN HEMT (enhancement-mode) from normally-on to normally-off through band bending and suppression of 2-DEG at zero-bias, which leads to low dark current. [31][32][33][34][35] In previously reported enhancementmode p-GaN/AlGaN/GaN photodetectors, the p-GaN acted as the gate by the means of photo induced carriers. However, there are only a very few studies on the utilization of p-GaN/ AlGaN/GaN structure, and most of them focused on phototransistor mode of operation, where the area of the p-GaN optical gate is limited by the transistor geometry. [35][36][37] A widened p-GaN region on the photodetector would broadly and effectively suppress the 2-DEG, which could help in enhancing the photosensitivity through lowering the dark current. Moreover, the detailed operation mechanism and the effect of electrode transmittance on the photodetector performance still need to be explored further.
In this study, we propose a highly photosensitive UV photodetector based on AlGaN/GaN with graphene finger electrodes on a p-GaN mesa structure. Unlike the typical p-GaN layer, which acts as an optical gate in p-GaN/AlGaN/GaN HEMT-based UV photodetectors, the p-GaN mesa-finger structure being largely formed on AlGaN/GaN heterostructure, widely suppresses the 2-DEG, which leads to substantially less dark current. Meanwhile, the high transparency of the graphene finger electrodes boosted the amount of light in photodetector active region, resulting in greater photocurrent by effective 2-DEG formation. To demonstrate the superiority of the proposed structure, we systematically compared it to two different photodetectors: p-GaN/AlGaN/GaN with graphene finger electrodes (transparent) (comparative sample #1) and AlGaN/GaN with Ni/Au finger electrodes (semitransparent) on p-GaN mesa structure (comparative sample #2). The graphene electrodes and p-GaN mesa structure of the proposed photodetector design produce higher figures of merit than those of other two comparator devices. The superior performance characteristics of the novel device structure are analyzed and described by kelvin probe force microscopy (KPFM) analysis and energy band diagram. The proposed structure may offer intriguing opportunities in highly photosensitive UV photodetectors. Figure 1a shows the schematic diagrams of the AlGaN/GaN photodetector with a graphene finger electrodes on a p-GaN mesa structure. Ni/Au (30/120 nm) was deposited on both ends www.advmatinterfaces.de of the p-GaN mesa structure as a probing pad. It is evident that the p-GaN mesa structure is largely formed on the AlGaN/GaN heterostructure even at underneath of probing pads for effective 2-DEG suppression at AlGaN/GaN interface. An enlarged view of the p-GaN finger can be seen in the middle and a corresponding scanning electron microscopy (SEM) image of the p-GaN mesa finger can be seen to the right of Figure 1a. The cross-sectional high-resolution transmission electron microscopy (HR-TEM) image and energy-dispersive X-ray spectroscopy (EDS) mapping of the device structure show the distinct p-GaN, AlGaN, and GaN layers on the Si substrate (Figure 1b,c). The thickness of each layer is ≈105 (p-GaN), 20 (AlGaN), and 300 nm (GaN), respectively. The right panel in Figure 1b shows the high-magnification TEM image of the interface region of the photodetector, which confirms that the p-GaN/AlGaN/GaN photodetector has a high structural integrity and effectively suppressed defect or dislocation density. Figure 1c (quantitative EDS mapping of photodetector) reveals that Ga, N, and Al predominantly compose each layer. The N atoms were uniformly distributed throughout the entire region of p-GaN/AlGaN/GaN. However, Ga mapping showed a quantitative decrease in the AlGaN region, which is regarded as a phenomenon that occurs when Ga 3+ site is substituted by Al 3+ ( Figure S1, Supporting Information).

Results and Discussion
The photoluminescence (PL) spectroscopic measurements of the p-GaN/AlGaN/GaN structure are presented in Figure 1d.
The observed sharp peak with full width at half maximum of 7.2 nm due to band-edge-related emission at ≈364 nm corresponds to 3.4 eV, which correlates to the band gap of hexagonal wurtzite GaN crystal. [38] The defect, vacancy, and other complex related yellow luminescence peaks observed at longer wavelengths than the GaN peak are found much less intense implying that the materials are highly crystalline. [39,40] To investigate the properties of the graphene and its integrity, Raman spectroscopic measurement was performed after its transfer to the device substrate. Figure 1e shows the signature peak of graphene, that is, the G band near 1584 cm −1 , corresponding to the high-frequency optical E 2g phonon at the center of the Brillouin region. The band due to A 1g breathing mode, that is, the characteristic D band that originates due to translational symmetry breaking or defects appeared at around 1340 cm −1 . The second order overtone of the D band, that is, the 2D band observed at 2668 cm −1 corresponds with the peak that occurs during the momentum conservation process by two phonons with opposite wave vectors. The calculated 2D to G peak intensity ratio is higher than unity, indicating the monolayer nature of the graphene, and the lower intensity of defect induced disorder band (D peak) suggests a minimal amount of defects after the completion of the transfer process. [41] Figure 1f shows the transmittance spectra of Ni/Au film (5/5 nm). It is well obvious that the transmittance is higher in the visible region (≈75% at 500 nm) but it starts to drop to lower values in www.advmatinterfaces.de the UV region (55% at 360 nm and 28% at 310 nm) due to the enhanced absorption at the lower wavelength regime.
Atomic force microscopy (AFM) and KPFM analyses were performed to demonstrate the role of the p-GaN mesa structure in our photodetector. Figures 2a,b show the 3D and 2D AFM images of the surface of the p-GaN mesa structure on the AlGaN, respectively, and indicate that the p-GaN finger shape with a mesa structure were clearly patterned. From the line scan profiles, an etch depth of about 107 nm is obtained, as shown in Figure 2c. Here, KPFM images of the p-GaN mesa structure are also obtained on the same area, as presented in Figure 2d,e. A significant difference in the surface potential between two regions is observed, indicating the presence of p-GaN and AlGaN on the top surface and implying that an appropriate etching depth had been reached. First, we calibrate the cantilever work function with a highly ordered pyrolytic graphite (HOPG, 4.6 eV), which is commonly used method for KPFM calibration before measuring the desired sample for correct value of contact surface potential. Then, we obtained the nanoscale AFM mapping images of the surface potential of a selected area of the sample with the calibrated cantilever, and the results showed the p-GaN area as ≈360 mV and AlGaN area ≈540 mV yielding the estimated CPD value of 180 mV ( Figure 2f). [42] This potential difference suppressed 2-DEG at the AlGaN/GaN interface underneath the large p-GaN area by lifting up the AlGaN conduction band and resulting flattened GaN conduction band. An energy band diagram of the p-GaN/ AlGaN/GaN heterostructure under dark and illumination condition is presented in Figures 2g,h, respectively. The addition of the p-GaN mesa structure to AlGaN/GaN changes the ensuing p-GaN/AlGaN/GaN device from normally-on to normallyoff state by suppressing 2-DEG at the AlGaN/GaN interface under dark conditions, which would lower the dark current significantly. Similarly, under illumination with UV light, the conduction band of the p-GaN is expected to lower, which in turn would result in the formation of 2-DEG at the AlGaN/GaN interface. Figure 2h,i is the energy band diagram and crosssectional schematic diagrams of the device under UV illuminated conditions, which clearly shows the formation of 2-DEG under UV illumination. The photogenerated electron-hole pairs under UV illumination are separated and move in the opposite direction which weakens the built-in electric field and shrinks the depletion region induced by p-GaN layer and the 2-DEG is restored. The high amount of photocurrent in the graphene transparent electrode in comparison to semitransparent Ni/Au electrode might be related to the formation of high 2-DEG carrier density at the interface as shown in Figure S2, Supporting Information. As a whole, the p-GaN mesa structure in the proposed photodetector suppresses and forms 2-DEG at the AlGaN/GaN interface, respectively, in dark and illuminated conditions.
To evaluate the optoelectrical properties of the proposed photodetector, we fabricated and investigated the UV photodetector of the proposed structure as well as two additional UV photodetectors with different structures. Figure 3a provides schematic illustrations of type-A (comparative sample #1), type-B (comparative sample #2), and type-C (proposed structure) photodetectors. Type-A graphene finger electrode (transparent) on p-GaN layer without mesa structure. Type-B employed Ni/Au elec-trodes (semi-transparent), a p-GaN mesa structure, AlGaN, and GaN. The I-V characteristics of each type of UV photodetector under dark and UV illuminated conditions were assessed, and a comparison is drawn up in Figure 3b,c. Unlike the dark current (I dark ) of the type-A photodetector that used p-GaN as the electron path, the I dark of type-B and type-C photodetectors are extremely low (J dark ≈ 1.04 × 10 −8 A cm −2 ) and relatively flat as the 2-DEG is effectively suppressed by the p-GaN mesa structure formed on the AlGaN/GaN heterostructure. Under light at a wavelength of 360 nm, the photocurrent from the type-A and type-C photodetectors is clearly higher than that of the type-B. This is because the transmittance of Ni/Au electrodes is lower than that of the graphene electrodes (Figure 1f). [43] The extracted I dark and I ph under the applied voltage of 20 V are presented in Figure 3c. One can expect that the reduction of finger width and spacing may lead to enhancing the photocurrent; however, a further decrease in the finger width may lead to the normally-on HEMT-like characteristics that can increase the dark current.
The type-C photodetector, which possesses the advantages of both type-A (transparent electrodes) and type-B (mesa structure), showed excellent performance across the figures of merit applicable to photodetectors (Figure 3d-g). The photosensitivity, defined by the equation S ph = I ph /I dark , is highest for type-C among the tested photodetectors (Figure 3d). Photoresponsivity (R ph ) is calculated by using the equation: R ph = I ph /(P inc × A ch ), where P inc denotes the incident power density and A ch denotes the illuminated area of the channel material. Under the applied voltage of −20 V, the highest R ph (≈21 A W −1 ) is also obtained from type-C device (Figure 3e). The quantum efficiency, calculated using η = h R ph / , where h is the Planck's constant, is the light frequency, and is the charge of the electron, at the applied voltage of −20 V of each of the three photodetectors is presented in Figure 3f. As the quantum efficiency is a function of photoresponsivity, the results follow the same trend as that observed in our photoresponsivity tests. Finally, the specific detectivity (D*) of the three devices, perhaps the meaningful performance parameter for a photodetector, is calculated using D* = / 2 ch D A R qI . As shown in Figure 3g, type-C device demonstrates the highest detectivity of 1.91 × 10 14 cm Hz 1/2 W −1 . Although the R ph of the type-A device is adequate, the higher dark current associated with the p-GaN channel, leads to a low D. The maximum values applicable to each figure of merit for each of the three photodetectors are listed in Table 1. The proposed type-C photodetector achieved the highest values for all figures of merit. The performance of the photodetector is compared with various previous works, as presented in Table 2. Figure 4a shows the temporal photoresponse of each of the three types of photodetectors under irradiation with pulsed UV light having a wavelength of ex = 360 nm and P in = 1.96 mW cm −2 at applied bias voltages of 1, 5, 10, 15, and 20 V. In the case of the proposed type-C device, current at the time of light ON gradually increases with the applied voltage, resulting in the highest on-off current ratio at light ON and OFF among the photodetectors studied. Figure 4b,c compares the rise time and fall time for each of the tested photodetectors at an applied bias voltage of 20 V. The rise time and the fall time were respectively extracted from 20% to 80% and from 80% to 20% of the peak current value. [44] In short, while the type-C device www.advmatinterfaces.de outperforms other tested photodetectors in metrics associated with the figures of merit, its response speed is still relatively slow. Although the response speed is faster than the several recent reports on UV Photodetectors, still it is not good enough and needs to be reduced for ultrafast applications. [10,45] The dry etching-induced defects act as the trap on the surface causing a higher response time in the photodetector with a p-GaN mesa structure. [45] In addition, under 365 nm illumination, the photogenerated carrier should pass through the AlGaN barrier which hinders the speed of photogenerated carrier collection and thus the rise time. [28,46] At the same time, the presence of an AlGaN barrier control over the carrier collection should lead to a faster decay time, however, the presence of persistent photoconductivity may rise to a longer fall time. [47][48][49][50] Graphene/p-GaN mesa structure may further suffer from the contact characteristics resulting in the lower hole injection further slowing down the 2-DEG formation and hence the response time. [51] It was reported that various factors such as increasing illumination power, bias voltage, AlGaN band edge illumination (i.e., shorter illumination wavelength), reducing the persistent photoconductivity effect, and post treatment can improve the response speed of photodetector. [25,28,45,52,53]

Conclusions
We reported a highly photosensitive UV photodetector composed of AlGaN/GaN with graphene finger electrodes on p-GaN mesa structure. The proposed design of p-GaN mesa finger Figure 3. a) Schematic diagram of three photodetectors with graphene finger electrodes (type-A), Ni/Au finger mesa electrodes (type-B), and graphene finger mesa electrodes (type-C); b) I-V curves of the three UV photodetectors under dark and UV illuminated conditions ( ex = 360 nm and P in = 1.96 mW cm −2 ); c) Comparison of I photo and I dark , d) photosensitivity, e) photoresponsivity, f) quantum efficiency, and g) detectivity calculated from each of the three types of UV photodetectors at an applied voltage of −20 V. Adv. Mater. Interfaces 2023, 10, 2202379 www.advmatinterfaces.de structures with transparent graphene electrode is systematically investigated and shown to be capable of achieving superior figures of merit better than those associated with two alternative photodetector structures. The photodetector showed a high sensitivity of the order of 10 6 and detectivity as high as 1.91 × 10 14 cm Hz 1/2 W −1 at 360 nm. The mechanism responsible for the photodetector's high photosensitivity is explored by KPFM analysis and energy band diagram. Our findings show that the enhanced performance is a result of effective 2-DEG suppression and formation by the p-GaN mesa structure widely located on the AlGaN/GaN heterostructure and improvement on light absorption by the transparency of graphene electrodes. Ultimately, this novel structure may prove to be highly useful as a high-performance photodetector.

Experimental Section
Epitaxial Growth: A p-GaN/Al 0.2 Ga 0.8 N/GaN heterostructure grown on a 6 in. p-Si (111) substrate using metal organic chemical vapor deposition was used in this study. AlN grown on the Si substrate was followed by the growth of standard graded AlGaN layer, a GaN channel layer, an AlGaN barrier layer, and a p-GaN top layer. EDS mapping of the sample along Si substrate presented in Figure S3, Supporting Information, clearly indicated that the Al mole fraction of the graded AlGaN layer toward the top layer was reduced gradually in order to realized the lattice matched GaN channel layer. The electron mobility (1400 cm 2 V −1 s −1 ) and sheet carrier density (n s >7.5 × 10 12 cm −2 ) were estimated by conducting Hall effect measurement on the AlGaN/GaN heterostructure without p-GaN. 105 nm p-GaN was used to ensure that there was an adequate thickness to suppress the n s at AlGaN/GaN interface estimated as   at the AlGaN/GaN interface which ensured that the low dark current was equivalent to the OFF state of normally-off HEMT, where, σ pol is the net polarization induced charge density, q is the electron charge, ε is the dielectric constant, d is the thickness of AlGaN layer, φ b is the surface barrier height, E f is the fermi level position with respect to GaN conduction band edge at the AlGaN/GaN interface, and ∆E c is the conduction band discontinuity between GaN and AlGaN. [59,23,60] Device Fabrication: p-GaN/AlGaN/GaN grown on Si samples were cleaned by acetone and isopropyl alcohol on ultrasonic bath, Piranha solution (H 2 SO 4 :H 2 O 2 , 1:1), and buffered oxide etchant for 5 min each sequentially. The interdigitated finger pattern of the p-GaN was made by inductively coupled plasma reactive ion etching in conjunction with a conventional photolithographic technique. Further, the CVD grown graphene on Cu foil was transferred to both a planar and patterned p-GaN/AlGaN/GaN sample in order to realize samples A and C, respectively. The details of the PMMA assisted wet transfer method employed in this study can be found elsewhere. [61] 5/5 nm of Ni/Au was deposited on the prepatterned sample by e-beam evaporation and lift-off followed by annealing at 550 °C for 100 s (for sample B). Afterward, 30/120 nm of Ni/Au probing metal pad was deposited to samples A, B, and C. Finally, the unwanted graphene from sample A and C was removed by O 2 plasma etching, again in conjunction with a photolithographic technique. The schematic of complete fabrication process is presented in Figure S4, Supporting Information. In total eight pairs of fingers with a length of 300 µm, width and spacing of 10 µm, and effective illumination area of 310 × 310 µm 2 were considered for this study.
Measurement Setup: The current-voltage characteristics of the UV photodetectors were measured by a Keithley 2400 source meter in a completely dark room equipped with a probe station at ambient room temperature. The light source was an optical fiber connected to a Bentham SSM150Xe switching monochromator. The illumination power was calibrated by using a standard New Port photodetector and a power-meter.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.