Enhanced Performance of Metal‐Semiconductor‐Metal UV Photodetectors on Algan/Gan Hemt Structure via Periodic Nanohole Patterning

AlGaN/GaN High Electron Mobility Transistor (HEMT) structures offer superior electrical and material properties that make them ideal for the fabrication of high‐performance Ultraviolet photodetectors (UV PDs), especially using the metal‐semiconductor‐metal (MSM) configuration. However, the metal layout of the MSM design and crystal defects in multi‐stack HEMTs can reduce photocurrent and degrade device performance. Nano‐structuring of the AlGaN/GaN surface with different nanofeatures is a promising approach to improve light absorption efficiency and increase device response. In this work, AlGaN/GaN HEMT MSM UV photodetectors with enhanced performance parameters by engineering the surface with periodic nanohole arrays are demonstrated. Optical simulations are used to optimize the design of the nanoholes' periodicity and depth. Unpatterned and nanohole‐patterned devices with varying nanohole depths are fabricated, and their performance shows substantial enhancement with the incorporation of nanoholes. The device with 40 nm deep nanoholes and 230 nm array periodicity shows the highest performance in terms of photocurrent (0.15 mA), responsivity (1.4 × 105 A W−1), UV/visible rejection ratio (≈103), and specific detectivity (4.9  ×  1014 Jones). These findings present a HEMT‐compatible strategy to enhance UV photodetector performance for power optoelectronic applications, highlighting that nanohole patterning is a promising prospect for advancements in UV photodetection technology.


Introduction
[3][4][5] Additionally, UV photodetectors are gaining high importance in the rapidly expanding fields of the Internet of Things (IoT) and smart wearables. [6,7]In order to operate in extreme conditions, the active material of UV PDs should be able to withstand harsh environments and operate with high power efficiency.10] AlGaN/GaN HEMTs are particularly promising for UV PDs, due to their low-cost production, excellent quality, large-area availability, and the possibility of integration with Si-electronics. [11,12]The 2-D electron gas (2DEG) that forms in AlGaN/GaN HEMTs enables the realization of UV PDs with high responsivity.[18] However, the large lattice and thermal mismatches between (Al)GaN and Si (17% and 54%) can lead to a high density of dislocations in the epi-structure, which degrades the crystalline quality and surface smoothness. [19,20]This can limit the photogenerated carriers in the PD devices and significantly reduce the PD performance.Furthermore, to attain the wavelength selectivity of AlGaN/GaN UV PDs, the Al content in the Al x Ga 1-x N is increased to tune the bandgap to a higher value.[23] On the other hand, the metal electrodes in PDs with interdigitated MSM geometry occupy a large area and partially shadow the incoming lights, which reduces the PD sensitivity. [14,24]o realize improved performance in UV PDs, researchers have developed AlGaN/GaN PDs with a variety of (Al)GaN native as well as foreign nanostructures.The native nanofeatures including nanorods, [25] nanowires, [26] nanopillars, [27] nano-cones, [28] and nano dots, [29] have shown to outperform conventional devices.However, despite their high performance, these positive nanostructures, which can extend from the substrate into free space, are susceptible to mechanical degradation during fabrication and exposure to environmental conditions due to their free-standing nature. [30][33][34][35] Though, being foreign materials bring additional levels of complexity to the fabrication process.41] This study shows the fabrication and performance analysis of MSM UV PDs with and without periodic nanohole array patterns on AlGaN/GaN HEMT structure grown with a double GaN/AlN superlattice (SL) buffer stack.The SL-buffer is implemented to reduce structural defects in the AlGaN/GaN HEMT layers, while the nanohole patterns are designed to enhance photon capturing and achieve wavelength selectivity for the near UV (NUV) light of the PD devices.Prior to device fabrication, optical simulations were conducted to determine the optimal nanohole array period and depth for maximum absorption and higher selectivity in the UV bandwidth of interest.Structure quality and the device's optoelectronic properties are systematically analyzed by a set of structural, morphological, and electrical microscopy techniques.The results demonstrate that UV PDs with nanohole patterns outperform unpatterned devices in terms of responsivity, detectivity, and UV/visible rejection ratio.The suggested nanohole patterning of AlGaN/GaN HEMTs could provide a promising prospect for the seamless fabrication of UV PDs with enhanced responsivity.

AlGaN/GaN HEMT Structure Characterization
A 5 μm depletion mode (D-mode) AlGaN/GaN HEMT structure was grown on a 200 mm Si (111) wafer by metal-organic chemical vapor deposition (MOCVD).An illustration schematic of the grown structure is presented in Figure 1a.The growth rate of each layer is depicted in Table S1 (Supporting Information).The top AlGaN/GaN HEMT structure was grown at temperature 1100 °C using high-grade precursors, which are trimethylaluminum (TMAl) as Al precursor, trimethylgallium (TMGa) as Ga precursor, and ammonia (NH3) as N 2 precursor, with V/III ratio in the GaN buffer and GaN channel maintained at 50 and 199, respectively.Figure 1b displays a scanning transmission electron microscope (STEM) image of a sample from the structure that was cut using a focused-ion beam (FIB) tool.The STEM image clearly shows the layer stack, which comprises of 223 nm AlN nucleation layer, 597 nm GaN/AlN short-period superlattice (SL1), 1.4 μm thick lightly carbon doped AlGaN buffer, 1.5 μm thick GaN/AlN long-period superlattice (SL2), 1 μm carbon-doped GaN buffer, 300 nm unintentionally doped (UID) GaN channel, 1 nm AlN spacer, and 18.2 nm Al 0.23 GaN barrier.The Al mole fraction in the AlGaN barrier was selected to be 23% to tune the bandgap for the 320-325 nm light absorption.The presence of vertical threading dislocations (TDs) is apparent in the lowermost layers of the structure, caused by the mismatch between AlN and Si.The purpose of AlN nucleation is to reduce this mismatch for the subsequent GaN epitaxial layers.The propagation of TDs continues through the SL1 layer which provides stress compensation through the GaN/AlN, causing bending of the TDs at the SL1/AlGaN buffer interface, into the bulk region of the thick Al-GaN buffer layer.The TD density is even lower in SL2, and further bending of the TD lines takes place at the SL2/GaN buffer interface, resulting in a GaN channel region almost free of the TD lines.Figure 1c displays high-resolution STEM images of SL1 and SL2, showing that they exhibit sharp interfaces between GaN and AlN with a high contrast.The top AlGaN/GaN HEMT stack is of excellent quality, as evident from the high magnification STEM image in Figure 1d.The image showcases a GaN channel that is free of dislocations and shows well-defined interfaces among the GaN channel, AlN spacer, and AlGaN barrier layers.
The surface quality of the structure is evaluated using tappingmode atomic force microscopy (AFM), as depicted in Figure 1e, which shows a 5 × 5 μm scan of the AlGaN surface.The scan reveals a root-mean-square roughness (RMS) value of ≈0.4 nm, indicating a smooth surface, with visible growth steps (average step height is ≈200 pm). Figure 1f is a 1 × 1 μm AFM scan, that is used to count the surface pit density, and it is found to be ≈14 × 10 8 cm −2 .
To get insight into the crystal quality, a triple axis 2- highresolution X-ray diffraction (HRXRD) measurement along the (002) plane was conducted.The results of the diffraction scan are shown in Figure 1g.The image displays a narrow central GaN peak, another peak with the second highest intensity that represents the AlGaN buffer.The satellite peaks on both sides of the central GaN peak represent the superlattice periods.The presence of these satellite peaks in the HRXRD scan indicates sharp SL periods in the structure, which is consistent with STEM images in Figure 1c.The insets in Figure 1g 102) HRXRD scans can be utilized to determine the density of edge (in-plane) and screw (out-of-plane) dislocations within the structure by using the following equations: [42 ] where b screw and b edge are the Burger vectors of screw and edge dislocations in GaN (0.51 and 0.32 nm).The calculated screw and edge TD densities are found to be 3.3 × 10 8 cm −2 and 4.1 × 10 9 cm −2 , respectively.This shows that edge TD density is ≈12 times larger than the screw TD density, so the edge type dislocations dominate in this structure.To further validate the structure quality, asymmetric reciprocal space mapping (RSM) along the (114) plane was conducted, and results are plotted in Figure 1h.The fringes in the RSM graph correspond to the GaN buffer, AlGaN buffer, AlGaN barrier, SLs, and AlN nucleation (from bottom to top).The vertical and inclined yellow dashed lines represent the fully strained and fully relaxed directions of the layers, with respect to the GaN fringe.The superlattice fringes follow the strain line, while the AlGaN buffer follows the relaxation line, demonstrating the stress compensation achieved through the alternation between superlattices, AlGaN, and GaN buffers.
The combination of HR-STEM, AFM, RSM, and HRXRD measurements provided a comprehensive understanding of the material content, interface, crystal, and surface quality, which qualified this structure for the fabrication of UV PD devices.
To further study the crystal quality of the structure, Figure 1i shows the visible (488 nm) μ-Raman spectrum of the Full HEMT stack.Since the 488 nm is able to probe the entire HEMT stack down to the Si substrate, the Si Raman peak (520.7 cm −1 ) could be distinguished.Additionally, the GaN E 2 (High) and the longitudinal optical (LO) GaN A 1 phonon modes are present in the Raman spectrum, as well as the AlN E 2 phonon mode.The E 2 (High) Raman peak is found at position 567.8 cm −1 , with FWHM of 5.1 cm −1 , indicating for the high crystal quality of the GaN layer in the structure.The 325 nm UV Raman spectrum was collected from the structure and presented in Figure 1j, to probe the top AlGaN/GaN HEMT layers quality.The GaN E 2 peak is present in the spectrum, similar to that obtained from the visible Raman spectrum, but with lower intensity due to the shallower penetration of the UV light through the structure compared to visible light, resulting in less light-material interaction.The GaN A 1 (LO) mode could not be identified in the UV Raman spectrum, which is attributed to the high-density 2DEG present in the structures that screens this phonon mode.Additionally, the UV Raman spectra shows AlGaN A 1 (LO) phonon at 751.2 cm −1 and two interface phonon peaks IF1 and IF2 at 605.6 cm −1 and 725.7 cm −1 , respectively.IF1 and IF2 peaks emerge during the growth of the barrier layer and could be linked to GaN and Al-GaN alloy disorder, which -if their Raman signal is dominant -causes a high probability of electron scattering off the 2DEG channel and degrades its density.In this measured structure, IF1 and IF2 signals are low, which means they have less impact on the quality of the 2DEG channel in the studied structure.The UV photoluminescence spectrum (PL) depicted in Figure 1k shows high near-band edge (NBE) emission from the GaN channel layer at 3.41 eV, which corresponds to the wavelength of 363.1 nm.
To address the electrical properties of the 2DEG channel in the grown structure, Hall measurement was conducted on 1 cm × 1 cm samples after depositing 1 mm indium pads at the sample corners -in a square geometry -annealed at 500 °C in ambient atmosphere.The 2DEG resistivity (R s ), carrier mobility (μ), and carrier concentration (N s ) were extracted from the measurements to be 397 Ω/sq, 1550 cm 2 /V.s, and 1.08 × 10 13 cm −2 , respectively.To assure the quality of the 2DEG channel across different locations on the grown structure, the Hall measurement results were collected from three more samples, and the results -in Figure 1l -show very good homogeneity of the 2DEG R s , μ, and N s .

Optical Simulation
To validate the impact of nanohole arrays on the optical absorption of the AlGaN/GaN HEMT structure and determine the op-timal nanohole array period and radius for the fabrication process, optical simulations were carried out using the finite element method approach with COMSOL Multiphysics -electromagnetic waves module.Figure 2a shows a schematic of the simulated structure, which consists of 300 nm GaN, 1 nm AlN, and 18 nm AlGaN layers.To enhance the realism of the simulation, 200 nm air layers were inserted at the top and bottom of the structure, and 200 nm blocks of perfectly matched layers (PML) were added to both vertical ends of the structure.The simulation edges were set to periodic boundary conditions to mimic the periodic nanostructure.The mesh was set to a finer mesh with minimum and maximum element size limits of 2 to 50 nm.The interior port of the top PML was designated as the source port (S1) and was used to calculate the reflectance (R) with a wavelength range of 250 to 500 nm, while the interior port of the bottom PML was set as the transmission port (S2) and was used to calculate the transmittance (T).Finally, the absorptance (A) was calculated using the equation A = 1 -R -T.Absorptance spectra were collected at different nanohole array periods (P = 200-500 nm) and diameter (D) / P ratios (0.1-0.9), with the nanohole depth fixed at 12 nm (r is nanohole radius).
Figure 2b displays a contour map that highlights the distribution of peak absorption wavelengths at nanohole arrays with different P and D/P values.The results show that the 320 nm wavelength exhibits peak absorption at all P values when D/P is between 0.1 and ≈0.3.As the nanohole period increases, there is a noticeable red shift toward longer absorbed wavelengths.This shift can be attributed to the understanding that when the distance between nanoholes is small, their pitch is also small, causing shorter wavelengths to scatter at the hole walls while longer wavelengths are absorbed by the gap region, resulting in improved coupling of short wavelength photons.As the nanohole period increases, there is space for longer wavelengths to also scatter at the hole walls, and the peak absorption redshifts. [43]The contour map also indicates that by adjusting the hole size there is a range of wavelengths that can be coupled into the nanoholes rather than a single wavelength.To further confirm this behavior, the absorptance in a single nanohole unit cell was studied using the same simulation design, and the simulated absorptance is shown in Figure S1 (Supporting Information).The results presented in Figure S1 (Supporting Information) demonstrate that the nanohole period is the only effective factor in wavelength selection.The absorptance spectra exhibit a single peak at  = 325 nm when P = 230 nm, and the peak wavelength shifts towards longer wavelengths as P increases up to 400 nm (in agreement with Figure 2b).However, for P values greater than 400 nm, the main absorptance peak experiences a blue shift towards 270 nm at P = 500 nm, and additional absorptance peaks emerge, providing extra absorption of the visiblewavelengths band.These simulations aided in choosing the best nanohole array for a specific wavelength.In our study, we selected the nanohole array with a period of 230 nm to investigate the selectivity of 320-325 nm light.
To examine the impact of nanohole etching depth on light absorption into the AlGaN/GaN HEMT structure, periodic simulations were conducted on nanohole unit cells with the selected periodicity of 230 nm and three different etching depths: 12, 18, and 40 nm.The 12 nm depth represents very shallow nanoholes on the AlGaN surface, while the 18 nm depth represents complete AlGaN barrier etching within the nanohole region.The 40 nm depth represents nanoholes that extend into the GaN channel.Figure 2c,d depict the electric field distribution of 325 nm light inside the nanoholes, as seen from a top view and cross-sectional view, respectively.As seen in the top view representation, deeper nanoholes exhibit more concentrated electric fields, with even higher concentration at the nanohole side walls.The cross-sectional representation in Figure 2d confirms that the existence of nanoholes in the simulated structure leads to light being coupled toward the bottom of the nanohole and confined within the nanohole dimensions.As the nanohole depth increases, more light confinement occurs, leading to enhanced light coupling.The resulting simulated A, R, and T spectra of both unpatterned and nanohole-patterned structures are plotted in Figure 2e-g.The spectra indicate overall enhanced absorptance in the nanohole patterned structure compared to the unpatterned structure in the UV wavelength range of 250-360 nm.They also show an increase in absorptance as the nanohole depth increases.The enhancement in absorptance is attributed to decreased reflectance in the aforementioned wavelength range.From Figure 2g, it can be observed that the visible wavelength range of 370-500 nm has high transmittance, resulting in a UV/Visible absorptance ratio of 9/5.

Nanohole Patterned AlGaN/GaN HEMT MSM UV PD Devices
To advance the study, four MSM UV PD devices were fabricated on the AlGaN/GaN HEMT structure, labeled as PD-A, PD-B, PD-C, and PD-D, representing unpatterned, 12, 18, and 40 nm deep nanohole arrays, respectively.A comprehensive schematic of the fabrication process flow as well as an optical microscope image of the fabricated devices are presented in Figure 3, and a detailed description of the fabrication process can be found in the experimental section in this report.To verify the patterning of the nanohole arrays, scanning electron microscopy (SEM) and AFM scans were used to analyze the surface morphology and depth profiles of the devices.The SEM image in Figure 4a demonstrates the successful patterning of periodic nanohole arrays, while a higher magnification SEM image in Figure 4b is used to measure the nanohole dimensions and calculate the array lattice constant (the period P).The results indicate a nanohole diameter of 151 nm and a pitch of 79.8 nm, which resulting in a period of ≈230 nm.Additionally, Figure 4c,d     The ideality factors (n) and Schottky barrier heights (SBH) of the four devices are extracted from the forward bias branch of both dark and photocurrents using Shockley's diode equation. [44,45]Under dark conditions, the n value for PD-A, PD-B, PD-C, and PD-D are 1.45, 1.49, 1.46, and 1.33, respectively, while the SBH values are 0.61 V, 0.59 V, 0.68 V, and 0.67 V, respectively.The SBH values agree well with previously reported SBH of Ni/Au Schottky contact with AlGaN/GaN at 300 K. [46] The change in SBH and n values from PD-A to PD-D could be deemed to the breakage of the 2DEG under deeper nanohole patterning, since the fabricated devices share the same structure, ge-ometry and fabrication process, and differ only in the patterned nanohole depths.Under illumination, n values were reduced to become 1.23, 1.26, 1.34, and 1.20, for PD-A to PD-D, respectively.The reduction in n is a result of the reduction of SBH of the devices and hence the increased number of free carriers available for conduction.The extracted SBH values for PD-A to PD-D under illumination are 0.56 V, 0.51 V, 0.53 V, and 0.47 V, respectively.
Figure 5e displays the photo-to-dark (I ph /I d ) current ratio for all devices.For all devices, this ratio displays a plateau behavior with increasing the applied voltage, with PD-D exhibiting the highest ratio of ≈10 4 .This indicates that PD-D shows the most effective photodetection performance among all the devices, demonstrating the benefits of deeper etching and stronger light coupling in improving the photo-to-dark current ratio.It is noticed that the photo-to-dark current ratio in PD-D shows a kink when the applied voltage is < 0.8 V.This kink could also be observed in Figure 5d, not only in the photocurrent but also in the dark current characteristics.This could be explained as when the applied voltage is below 0.8 V the electric field across the MSM in device PD-D is not very high to overcome the high-density trap states due to deeper etching, which traps the electrons and prevents them from being collected.After the applied voltage is increased, the electric field becomes stronger and de-traps the trapped electrons, which causes a sudden increase in the device current, creating the observed kink.
In order to investigate the relationship between photocurrent and illumination power, I-V measurements were conducted at a bias voltage of 5 V under 325 nm illumination with varying power levels ranging from 0.04 to 1.3 mW.The summary of the results is presented in Figure 5f, with individual device results made available in the supplementary information (Figure S2, Supporting  Information).The four devices exhibit a linear increase in photocurrent at low power levels until 0.1 mW, after which the linearity drops and the photocurrent tends to reach saturation.The relationship between photocurrent (I ph ) and incident power (P) is described by a power law, which governs the behavior of the photodetector in response to changing illumination power: [47] where x is the power coefficient.The fitted I ph − P plots according to Equation ( 3) are shown in the supplementary information (Figure S3, Supporting Information).The plots for all devices display a characteristic in which x is greater than 1 in the linear regime, and x is less than 1 in the saturation regime.Previous reports have suggested that in 2DEG structures when x is greater than 1, the photocurrent is influenced by an internal gain mechanism.On the other hand, when x is less than 1, the gain factor decreases under high power.The normalized gain has been calculated and plotted in Figure S3e (Supporting Information), which shows that the gain decreases and reaches saturation at high power levels beyond 0.1 mW, thereby explaining the linear and saturation regions of the photocurrent at different illumination powers.The value of x is found to be nearly the same for all devices in the linear region, while in the saturation region, PD-C displays the lowest value of 0.2, which is correlated to its lower photocurrent.Responsivity (R), noise-equivalent-power (NEP), specific detectivity (D*) and normalized photo-to-dark current ratio (NPDR) are calculated from the following Equations: where A is the device active area, and e is the electron charge.
The performance results of the optoelectronic characterization of the unpatterned and patterned MSM UV PDs under 325 nm illumination and 5 V bias voltage are summarized in Table 1.
Figure 5g,h shows the change in responsivity and detectivity with the increase in incident power.The responsivity values, which indicate the sensitivity of the devices, exceed unity at all power levels, confirming the presence of an internal gain mechanism in the devices.As the incident power increases, both R and D * decrease and eventually saturate.The linearity in the low-power regime and the saturation in the high-power regime are in agreement with the photocurrent behavior under different powers.
The results demonstrate that the nanopatterned devices show improved performance compared to the unpatterned devices, with the highest R, D, and NPDR values observed for PD-D.The behavior of photocurrent, responsivity, and detectivity of all devices under different powers provide further insight into the underlying physics of the optoelectronic properties of the MSM UV photodetectors.
The spectral responsivity (SR) of the four devices was measured in the spectral range of 250-500 nm at a low power density of 14.7 μW/cm 2 and a bias voltage of 5 V.The results, presented in Figures 6a-d, show that all the devices are highly sensitive to UV radiation in the range of 250-365 nm, even at this very low power density.The nanopatterned devices, PD-B, PD-C, and PD-D, with a nanohole array period of 230 nm demonstrate enhanced SR levels compared to the unpatterned device PD-A.The unpatterned PD-A device has a steady SR of 3.6 × 10 2 A/W in the range of 250-295 nm, which then increases to 5.7 × 10 2 A/W in the range of 300-325 nm, before dropping back to 3.6 × 10 2 A/W in the range of 330-365 nm.The nanopatterned devices, on the contrary, show an exponential increase in SR, with a peak SR value at 325 nm.The peak SR values for PD-A, PD-B, PD-C, and PD-D are 55.7 × 10 2 A/W, 5.3 × 10 3 A/W, 8.9 × 10 2 A/W, and 1.4 × 10 5 A/W, respectively.The results align with the optical simulations and highlight the selectivity of the nanopatterned devices at 325 nm wavelengths as well as the improved performance compared to the conventional unpatterned device.The PD-D device exhibits high performance when compared to the other devices, with a high UV/Visible rejection ratio of 10 3 .Furthermore, the specific detectivity of PD-D is calculated to be 4.90 × 10 14 Jones.Such a high detectivity of PD-D at low power levels indicates low-noise operation and high performance for low-light detection applications, making it a suitable choice for applications where low-light sensitivity and low-noise performance are crucial.The improved performance of the nanohole patterned devices -specially PD-D -over the unpatterned device is attributed to the increased confinement of the electric field component of light wave and enhanced absorption, leading to the enhanced performance.To provide a comparison among various PDs, Table 2 compares the performance of conventional and nanostructured AlGaN/GaN UV photodetectors previously reported in the literature with this investigation.Figure 6e presents the plot of normalized SR for all the devices, showing a −3 dB cut-off wavelength of 365 nm, which aligns with the energy bandgap of the GaN channel.
To further investigate the impact of nanohole array periodicity on wavelength selection, three additional devices were fabricated with nanohole periods of 300, 400, and 600 nm.SEM images of the nanoholes and SR spectra of these three devices can be found in Figure S4 (Supporting Information).The spectra reveal that the SR peak wavelength can be tuned to 340 nm (for P = 300 nm), 350 nm (for P = 400 nm), and exhibit triple absorption peaks at 250, 330, and 360 nm (for P = 600 nm).The good matching between the experimentally obtained results and the simulated optical absorbance results further validates the proposed design approach and provides a guideline for the development of highperformance AlGaN/GaN UV photodetectors.
To assess the operating speed of the devices, the transient photocurrent was measured under 325 nm illumination with Table 2. Comparison of performances between nanohole patterned AlGaN/GaN HEMT-based UV PDs in these work and previous reported conventional devices and devices with different nanostructures, in terms of wavelength (), Responsivity (R), photo-to-dark current ratio (I ph /I d ), Detectivity (D).

Device Description
Peak Si-doped Al 0.45 Ga 0.55 N nanorods [ 49] 250-276 0.115 --Bi-crystalline GaN nanowires [ 50] 360 1.74 × 10 7 203 2.82 × 10 14 WO 3 gated AlGaN/GaN [51] 200-395 1.67 × 10 4 1.1 -ZnO nanorod-gated AlGaN/GaN HEMT [ 34]  Al-plasmonic based-GaN [ 8] 355 670 -1.48 × 10 15 AlGaN/GaN HEMT phototransistors [ 52] 365 1 × 10 6 2.5 × 10 8 1.8 × 10 17 Al 0.18 Ga 0.8 2N/GaN HEMT [53] 365 800 10 3 1.28 × 10 14 Semipolar AlGaN on AlN/sapphire [ 54] 270 1.84 × 10 3 1.61 × 10 3 5.77 × 10 10 AlGaN nanopillar/superlattice avalanche PD [ 27] 324 1.46 --AlGaN/GaN HEMT [ 48] 365 1670  (Supporting Information).Figure 6f summarizes the rise time ( rise ) and fall time ( fall ) values of the devices.Among all the devices, the unpatterned PD-A device exhibits the fastest rise and fall times.The nanopatterned devices, with voids in the structure due to the nanoholes etching, have a longer path for photogenerated carriers to travel during the ON and OFF states of the illumination, leading to a longer response time compared to the unpatterned device.The relatively slow fall time in all devices is attributed to the persistent photoconductive effect, where energetic electrons are injected into deep impurity levels within the conduction band, which delays the recombination process. [48]he schematic representation and band structure of the patterned and unpatterned UV photodetection (PD) devices are shown in Figure 7 to illustrate their detection mechanism and optical performance enhancement.In the PD-A device in Figure 7a, UV photons are absorbed into the AlGaN barrier, which results in the generation of photocarriers.These photocarriers are then collected by the Schottky metal, producing a photocurrent.The UV light penetrates deep through the thin AlGaN barrier layer and into the GaN channel layer, generating additional photocarriers.Under an applied bias, photogenerated electrons from the GaN layer drift towards the 2DEG channel and increase its density, which contributes more to the photocurrent of the device.This is why there are two SR regimes at 325 nm and 365 nm in Figure 6a, which match the bandgaps of Al 0.23 GaN and GaN layers, respectively.The shallow 12 nm nanoholes in PD-B, as depicted in Figure 7b, are designed to effectively trap more of the incoming UV photons, reducing their reflection on the surface.This increases the absorption of the UV photons into the AlGaN barrier layer and generates a higher number of photocarriers.In addition to the increased 2DEG density in a manner similar to PD-A, the nanoholes also feature trap states on their sidewalls due to the etching process that captures the photogenerated holes resulting from the interaction between UV photons and the nanohole sidewalls.These trapped photogenerated holes serve as fixed positive charge centers, which facilitate the electron's motion toward the 2DEG, further increasing its density and leading to enhancement in photocurrent in PD-B compared to PD-A. Figure 7c illustrates the PD-C device, where the nanohole etching process has completely removed the AlGaN barrier at the locations of the nanoholes in the array, creating discontinuities in the 2DEG channel.Although these discontinuities reduce the dark current level in PD-C, they also result in a reduction of the device's photocurrent.The prolonged path of the incident UV photons into the 18 nm nanoholes allows the photons to be coupled to the GaN layer, creating more photocarriers in the process, which makes the photocurrent level more similar to that of PD-A.The deeper 42 nm nanoholes in the PD-D device (Figure 7d) allow for capturing more of the incident UV photons, which increases the photocarrier generation.The deeper sidewalls of the nanoholes enhance the scattering of the UV photons and prolong their path further into the nanoholes, leading to better coupling within the GaN channel layer.This is evident from the increased responsivity at 365 nm as shown in Figure 6d, which matches the GaN layer bandgap energy.Despite the discontinuities in the 2DEG in PD-D, the high density of hole trap states in the nanohole sidewalls allow more electrons to accumulate just below the bottom of the nanoholes and at their sidewalls.This further increases the 2DEG density and creates a prolonged path for electrons around the nanohole geometry in the GaN layer, as shown in Figure 7d.This results in enhanced photocurrent in the PD-D device compared to all the other devices but at the expense of slowing down the operation speed of the device.A summarized schematic on the mechanism of action of etching trap states on the photogenerated carriers is illustrated in Figure 7e.Since nanoholes increase the effective surface area of the material, the light absorption is enhanced and the generation of more electron-hole pairs occurs.UV absorption in GaN and AlGaN layers creates a vertical electric field that -on one hand -causes photo-generated holes to be trapped at nanohole entrance trap states as well as at etched sidewall traps, while -the other hand -the photogenerated holes in GaN channel layer drift toward the buffer side of the HEMT stack due to valence band discontinuity in AlGaN/GaN heterostructure. [55]Hence, under illumination and applied electric field, electrons are more likely to travel through the material, while holes can be easily trapped in the etching trap states, which increases the electron collection efficiency.
To validate the concept of the 2DEG discontinuity and the impact of nanoholes on enhancing device performance, we measured the transient photocurrent for all devices under 0 V bias conditions (shown in Figure S6, Supporting Information).The results revealed a clear transition from the dark current to photocurrent when the light source was turned ON, with photocurrent values of 2.8 nA and 4.3 nA for PD-A and PD-B, respectively, and − 11.4 nA and − 120 nA for PD-C and PD-D, respectively.The shift from positive photocurrent in PD-A and PD-B to negative photocurrent in PD-C and PD-D is attributed to the discontinuity in the 2DEG in the latter two devices, where photocarriers move in the opposite direction to compensate for the channel discontinuity and contribute to the collected current only under the influence of incident light.This also demonstrates the "self-powered" operation of all devices, a crucial property for integration into energy-efficient wearable devices.

Conclusion
This study has successfully demonstrated the fabrication of UV photodetector (PD) devices with patterned and unpatterned nanohole arrays.The devices were fabricated through a combination of photolithography and dry etching techniques, and optical simulations were performed to gain insight into the device operation and optimize the nanohole array geometry for better device performance.The I-V and SR results indicated that the nanohole patterning in the PD devices had a significant impact on the device performance, where nanopatterned devices exhibited higher responsivity and photocurrent compared to the unpatterned one, achieving SR levels as high as 1.4 × 10 5 A/W in the deeper nanohole patterned device.The mechanism behind the performance enhancement due to nanohole patterning was attributed to the understanding that the deeper nanoholes capture more of the incident UV photons, prolonging their path into the GaN layer, and increasing the photocarrier generation.The nanohole etching and trap states also played a critical role in the improved performance of the PD devices, by trapping photogenerated holes and increasing the electron accumulation.The deeper nanoholes resulted in a longer path for electrons around the nanohole geometry in the GaN layer, leading to enhanced photocurrent, but slower carrier recombination when the illumination light was turned OFF.Furthermore, the devices demonstrated remarkable sensitivity to weak signals, as evidenced by their high detectivity levels at extremely low levels of illumination power.This capability enables them to accurately detect weak signals while filtering out any unnecessary background noise, resulting in a clearer and more precise signal output.Additionally, both the nanopatterned and unpatterned devices displayed selfpowered action in the absence of any external applied voltage.Overall, this study provides valuable insights into the simulation, fabrication, performance, and operation mechanism of UV PD devices, and highlights the importance of nanohole patterning in optimizing their performance.The findings can guide the design and development of future UV PD devices that could be integrated into power electronic systems, with improved detection capabilities.

Experimental Section
Structure Characterization: The AlGaN/GaN HEMT structure employed in this study was fabricated using metal-organic chemical vapor deposition (MOCVD) on a 200 mm Si (111) wafer.To determine the quality of the structure, a combination of structural and morphological characterization techniques was used.A vertical cross-section of the sample was obtained through a focused-ion-beam (FIB) using FEI Helios -NanoLab 600 system, and the FIB sample was imaged using JOEL JEM-2100 highresolution scanning transmission electron microscopy (STEM).The surface morphology was studied through tapping mode in Bruker Dimension ICON atomic force microscopy system (AFM), while the crystal quality was analyzed through high-resolution X-Ray diffraction (HRXRD) and reciprocal space mapping (RSM) in PANalytical X'Pert PRO system.
Device Fabrication: The fabrication of four devices, PD-A, PD-B, PD-C, and PD-D, was carried out in a single run using conventional photolithography techniques (as shown in Figure 3). 1 × 1 cm 2 samples were first cut out from the full wafer, and a photomask was utilized to define the active area of the devices, which was 100 × 105 μm 2 .The device isolation was achieved through an inductively coupled plasma (ICP)/reactive ion etching (RIE) mesa etching process.To ensure thorough device isolation, the mesa depth was etched to 300 nm through the use of a BCl 3 :Cl 2 gas mixture in the UNAXIS SLR-7701 ICP/RIE reactor.Next, another photomask was used to define the MSM interdigitated metal fingers, and semi-transparent Ni/Au (10/10 nm) Schottky metal was deposited on the active area through Denton Explorer e-beam evaporation system.The MSM interdigitated geometry consisted of 11 metal fingers, with dimensions of × W × S = (85 × 5 × 5) μm 3 , where L, W, and S refer to length, width, and spacing, respectively.Finally, the electrical probing contact pads were deposited to complete the device fabrication.
Nanohole Patterning: After the devices were successfully fabricated, a 230 nm period nanohole arrays were designed and transferred to the Elionix ELS-7000 electron beam lithography (EBL) system.Devices PD-B, PD-C, and PD-D were spin-coated with 300 nm thick polymethyl methacrylate (PMMA) to act as an EBL resist and baked at 180 °C for 2 min.After EBL exposure, the devices were removed and developed in methyl isobutyl ketone (MIKB) solution for 90 s.Nanohole etching was achieved inside the ICR/RIE plasma chamber using BCl 3 :Cl 2 gas mixture, followed by SF 6 gas flow at the same plasma conditions.The RIE power and exposure time were changed to obtain different nanohole depths in the devices.
Device Characterization and Performance Tests: The morphologies of the nanohole patterns were observed using the JOEL JSM6700 scanning electron microscope (SEM) and Bruker Dimension ICON AFM.The optoelectronic properties of the devices were evaluated by probing them using a Keithley source-meter and illuminating the active area with a 325 nm He-Cd laser.Spectral responsivity measurements were conducted using a broadband light source (250-1100 nm) connected to a Bentham monochromator while probing the samples using Keithley source-meter.

Figure 1 .
Figure 1.a) schematic cross-section illustration of AlGaN/GaN HEMT structure on Si (111) substrate showing the arrangement of different layers.b) FIB-cut cross-sectional STEM image of the whole structure, with annotated layer thicknesses.c) High magnification STEM image of short and long superlattice structures (SL1 and SL2).d) high magnification STEM image of the top AlGaN/GaN HEMT structure.e) and f) top-view 5 × 5 μm and 1 × 1 μm AFM scans of the AlGaN/GaN HEMT surface.g) Triple axis 2 −  HRXRD scan of (002) plane, centered around GaN (002) peak.The two insets show a double axis -scan of (002) plane (left) and (102) plane (right).h) RSM image of the full structure, with each corresponding peak annotated.i-k) are visible Raman, UV Raman, and UV PL spectra of the AlGaN/GaN HEMT structure.l) Hall measurement results on 4 different locations on the grown structure.
display double axis -scan HRXRD along the (002) and (102) planes, showing the peak positions of GaN (002) and GaN (102) with full width at half maximum (FWHM) values of 0.11 o and 0.27 o , respectively.The narrow FWHM values indicate for the excellent crystal quality of the GaN layer in the AlGaN/GaN HEMT structure on Si.The (002) and (

Figure 2 .
Figure 2. a) schematic of the UV PD structure used for the optical simulations, with a zoom-in on the nanohole cell defining its period (P) and radius (r).b) Contour map representation of the peak wavelength collected from different simulated absorption spectra as a function of nanohole array period and diameter/period ratio.c) and d) are top view and cross-section views of electric field component |E x | 2 distribution inside unpatterned structure and the nanohole unit cell.e,f), and g) are simulated absorptance, reflectance, and transmittance spectra of unpatterned structure and nanoholes with period P = 230 nm and depths of 12, 18, and 40 nm.
,e display the top and bottom 3D views as well as the depth profiles obtained from AFM scans on the nanoholes of devices PD-B, PD-C, and PD-D, respectively.The top and bottom 3D AFM views reveal a high degree of uniformity of the etched nanoholes in the three devices, showing that they exhibit well-defined sidewalls.The depth profiles are used to calculate the nanoholes depths and are found to be 14.2, 18.5, and 42.3 nm, for PD-B, PD-C, and PD-D, respectively.
Figure 5a-d shows the results of optoelectronic characterization of unpatterned and patterned MSM UV PDs.The asymmetrical current-voltage (I-V) plots observed in deeper etched nanohole devices PD-C and PD-D can be attributed to the different interface between the Ni/Au Schottky metal and AlGaN or nanoholes underneath it.The level of dark current (I d ) for PD-A and PD-B are nearly equivalent, while PD-C and PD-D exhibit almost one order of magnitude lower dark current.This variation can be attributed to the underlying D-mode HEMT structure, where the presence of the 2DEG in unpatterned PD-A and shallowly patterned PD-B results in the dark current due to the normal ON operation of the device.However, the localized partial removal of the AlGaN barrier layer in PD-C and the deeper etching in PD-D in the nanohole regions leads to discontinuities

Figure 3 .
Figure 3. Schematic of the full MSM UV PD fabrication process flow for both unpatterned and patterned devices.The optical microscope image is given at the right side showing the fabricated AlGaN/GaN HEMT MSM UV PD devices.

Figure 4 .
Figure 4. a) SEM image of the nanohole array after etching.b) Higher magnification SEM image of the nanohole array, annotated with the hole's diameter and pitch values.(c), (d), and (e) are AFM 3D top view (left), bottom view (middle), and depth profile of nanoholes (right) for devices PD-B, PD-C, and PD-D, respectively.

Figure 5 .
Figure 5. Semi-log scale of current-voltage characteristics of devices a) PD-A, b) PD-B, c) PD-C, and d) PD-D, under dark and 325 nm light conditions.e) Photocurrent (I ph ) to dark current (I d ) ratio versus applied voltage for all devices.Dependence of f) photocurrent, g) responsivity, and h) detectivity of all devices under 325 nm incident light power.

Table 1 .
Dark current (I d ), photocurrent (I ph ), and photo-to-dark current ratio (I ph /I d ) at 5 V applied bias, and Responsivity (R), Detectivity (D), and normalized photo-to-dark current ratio (NPDR) values of the fabricated MSM UV PDs at 5 V applied bias at 1.3 and 0.1 mW illumination powers.Device I d (nA) I ph (μA) I ph /I d R (A/W)

Figure 6 .
Figure 6.Spectral responsivity (SR) of devices a) PD-A, b) PD-B, c) PD-C, and d) PD-D, measured at the applied voltage of 5 V and illumination power density of 14.7 μW/cm 2 .e) Normalized spectral responsivity (N-SR) at 5 V applied voltage.f) Bar plot showing rise time ( rise ) and fall time ( fall ) values of all devices.

Figure 7 .
Figure 7. Schematic and band structure illustrations of the operation mechanisms of a) PD-A, b) PD-B, c) PD-C, and d) PD-D.e) Summarized schematic on the mechanism of action of etching trap states on the photogenerated carriers.