Double-function integrated phototransistor based on GaN-based heterojunction and preparation method of double-function integrated phototransistor
By using a GaN-based heterojunction-based bifunctional integrated phototransistor, a multilayer structure is grown on the substrate using an epitaxial growth method to achieve the switching of photodetection and photosynaptic modes. This solves the problems of data transfer delay and high power consumption in silicon-based chips in the prior art, improves the system's flexibility and adaptability, and is suitable for edge computing and biomimetic sensing applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTH CHINA NORMAL UNIV
- Filing Date
- 2025-12-31
- Publication Date
- 2026-05-05
AI Technical Summary
In the existing technology, silicon-based chips based on the von Neumann architecture suffer from data transfer delay and high power consumption when processing sensing signals and computing tasks. Discrete device solutions result in high system complexity and limited integration density, making it impossible to achieve high-density array integration and dynamic reconfiguration, thus limiting system flexibility and adaptability.
A bifunctional integrated phototransistor based on GaN heterojunction is adopted. A buffer layer, GaN barrier layer, AlN reflective layer, AlGaN barrier layer, GaN cap layer, metal oxide thin film layer and two-dimensional material layer are grown on the substrate by epitaxial growth method. The switching between photodetection mode and photosynaptic mode is controlled by bias voltage to realize the integration of sensing and computing functions.
It achieves flexible switching between photoelectric detection and photoelectric synapse modes in a single physical entity, reducing system complexity and power consumption, supporting dynamic working mode switching, and is suitable for edge computing and biomimetic sensing applications, simplifying the data processing flow.
Smart Images

Figure CN121985606A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of semiconductor photodetector technology, specifically to a bifunctional integrated phototransistor based on a GaN-based heterojunction and its fabrication method. Background Technology
[0002] With the rapid development of artificial intelligence, the Internet of Things, and edge computing, information processing systems are facing an urgent need for high-performance, low-power, and multifunctional integrated devices. Traditional silicon-based chips based on the von Neumann architecture suffer from the "memory wall" bottleneck and high power consumption caused by frequent data transfer between storage and processing units when processing sensing signals and computing tasks. Neuromorphic computing, inspired by biological neural systems, provides a revolutionary path to overcome these bottlenecks by mimicking the in-memory computing and parallel processing mechanisms of the human brain. Among these, hardware units capable of simultaneously realizing environmental perception (such as light signal detection) and brain-like computing (such as synaptic weight updates) are the core foundation for building efficient sensing-computing integrated systems.
[0003] Currently, achieving sensing and computing functions mainly relies on discrete device solutions, which use high-performance photodetectors to convert optical signals into electrical signals, which are then processed by back-end artificial synapses or neuronal circuits. However, discrete device solutions still have some drawbacks, such as high system complexity, high power consumption, additional delays and energy losses during signal transmission and conversion between discrete devices, limited integration density (discrete devices occupy a large area, making large-scale, high-density array integration difficult), and fixed device functions after fabrication, which cannot be dynamically reconfigured according to task requirements, limiting the system's flexibility and adaptability, thus restricting its application and development to some extent. Therefore, there is a need for a phototransistor that can simultaneously realize sensing and computing functions in a single physical entity. Summary of the Invention
[0004] To address the technical problems existing in the prior art, this invention provides a bifunctional integrated phototransistor based on GaN-based heterojunction and its fabrication method. The bifunctional integrated phototransistor has the physical basis for two high-performance operating modes to coexist in a single physical entity, and the switching between photodetection mode and photosynaptic mode of the bifunctional integrated phototransistor can be controlled by a low bias voltage electrical signal.
[0005] The first objective of this invention is to provide a bifunctional integrated optotransistor based on a GaN-based heterojunction.
[0006] The second objective of this invention is to provide a method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction.
[0007] The first objective of this invention can be achieved by adopting the following technical solution:
[0008] A bifunctional integrated phototransistor based on a GaN-based heterojunction includes a substrate layer. A buffer layer, a GaN barrier layer, an AlN reflective layer, an AlGaN barrier layer, a GaN cap layer, a metal oxide thin film layer, a two-dimensional material layer, and a metal electrode are sequentially grown on the substrate layer using an epitaxial growth method. The GaN barrier layer, AlN reflective layer, AlGaN barrier layer, and GaN cap layer together constitute a GaN / AlGaN heterojunction. Based on the synergistic effect of the two-dimensional material layer and the GaN / AlGaN heterojunction, the bifunctional integrated phototransistor switches between functional modes by controlling the bias voltage.
[0009] The fabrication method of bifunctional integrated phototransistors based on GaN-based heterojunctions includes:
[0010] Choose one of sapphire, SiC, or Si as the substrate, perform structuring on the substrate, and prepare patterns on the substrate surface to obtain a patterned substrate layer;
[0011] A buffer layer is deposited on top of the substrate using chemical vapor deposition.
[0012] A GaN barrier layer was deposited on top of an AlN buffer layer using chemical vapor deposition.
[0013] An AlN reflective layer was deposited on top of a GaN barrier layer using chemical vapor deposition.
[0014] An AlGaN barrier layer was deposited on top of the AlN reflective layer using chemical vapor deposition.
[0015] A GaN cap layer was deposited on top of the AlGaN barrier layer using chemical vapor deposition.
[0016] A metal oxide thin film layer is deposited on top of the GaN cap layer using ALD;
[0017] Two-dimensional material layers are grown on top of metal oxide thin films using experimental transfer or chemical vapor deposition / physical vapor deposition.
[0018] Metal electrodes are fabricated on a two-dimensional material layer using electron beam evaporation.
[0019] Specifically, the pattern is a periodic array of nanopillars, a micrometer-scale groove, or an inverted pyramid structure, used to improve light absorption, reduce dislocation density, and enhance heat dissipation.
[0020] Specifically, the buffer layer is AlN, GaN or AlGaN, and the thickness of the buffer layer is 10 nm to 1000 nm.
[0021] Specifically, the GaN barrier layer is p-GaN, n-GaN, or i-GaN, and the thickness of the GaN barrier layer is 1 nm to 200 nm.
[0022] Specifically, the AlGaN barrier layer is p-AlGaN, n-AlGaN, or i-AlGaN, with an Al composition of 0.1 to 0.8 and a thickness of 1 nm to 500 nm.
[0023] Specifically, the GaN cap layer is p-GaN, n-GaN, or i-GaN, and the thickness of the GaN cap layer is 0.1 nm to 10 nm.
[0024] Specifically, the metal oxide thin film layer is colorless and transparent, including Al2O3 or ZnO, and the thickness of the metal oxide thin film layer is 1 nm to 100 nm.
[0025] Specifically, the two-dimensional material layer is Graphene, MoSe2, or MoS2, and the thickness of the two-dimensional material layer is 2 nm to 400 nm.
[0026] Specifically, the metal electrode is Au, Cr / Au, Ti / Au, or Ti / Al / Ni / Au, and the thickness of the metal electrode is 5 nm to 20 μm.
[0027] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0028] This invention provides a GaN-based heterojunction-based dual-function integrated phototransistor and its fabrication method. The GaN-based heterojunction-based dual-function integrated phototransistor includes a substrate layer. Using epitaxial growth, a buffer layer, a GaN barrier layer, an AlN reflective layer, an AlGaN barrier layer, a GaN cap layer, a metal oxide thin film layer, a two-dimensional material layer, and a metal electrode are sequentially grown on the substrate layer. The GaN barrier layer, AlN reflective layer, AlGaN barrier layer, and GaN cap layer together constitute a GaN / AlGaN heterojunction. Based on the synergistic effect of the two-dimensional material layer and the GaN / AlGaN heterojunction, the dual-function integrated phototransistor switches functional modes by controlling the bias voltage. By utilizing the cross-dimensional heterojunction structure, a physical basis for the coexistence of two high-performance operating modes is achieved within a single physical entity. The dual-function mode switching is realized by using a simple electrical signal—low bias voltage—as a "function selector." This provides a new technical solution for future multi-functional fusion chips for edge computing, biomimetic sensing, and other applications. Attached Figure Description
[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0030] Figure 1 This is a schematic diagram of a bifunctional integrated phototransistor structure based on a GaN-based heterojunction in an embodiment of the present invention.
[0031] Figure 2 This is a schematic diagram of dual-mode bias switching of a dual-function integrated phototransistor in an embodiment of the present invention;
[0032] Figure 3 This is a schematic diagram of the formation mechanism of a dual-function integrated phototransistor in an embodiment of the present invention;
[0033] Figure 4 This is a schematic diagram of the photosynaptic mode of the dual-function integrated phototransistor in an embodiment of the present invention;
[0034] Figure 5 This is a schematic diagram of the dual-function integrated phototransistor photodetector mode in an embodiment of the present invention.
[0035] The labels in the figure are: 1-substrate layer, 2-buffer layer, 3-GaN barrier layer, 4-AlN reflective layer, 5-AlGaN barrier layer, 6-GaN cap layer, 7-metal oxide thin film layer, 8-two-dimensional material layer, 9-metal electrode. Detailed Implementation
[0036] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments, and the implementation of the present invention is not limited thereto. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] Example 1:
[0038] This embodiment provides a dual-function integrated phototransistor based on an AlGaN / GaN heterojunction, enabling flexible switching between optical signal sensing and information processing by introducing a dynamic bias switching mode. The innovation of this device lies in its composite structure of a metal oxide thin film layer and a two-dimensional material layer, which optimizes the interface contact characteristics. Simultaneously, it utilizes the synergistic effect of the two-dimensional material and the GaN / AlGaN heterojunction to achieve effective switching between photodetection and photosynaptic modes.
[0039] like Figure 1 The diagram shows a schematic of a GaN-based heterojunction-based dual-function integrated phototransistor. The GaN-based heterojunction-based dual-function integrated phototransistor of this invention utilizes epitaxial growth to sequentially grow an AlN buffer layer 2, a GaN barrier layer 3, an AlN reflective layer 4, an AlGaN barrier layer 5, a GaN cap layer 6, a metal oxide thin film layer 7, a two-dimensional material layer 8, and a metal electrode 9 on a substrate layer 1. The metal oxide thin film layer 7 optimizes the contact surface between the GaN layer 6 and the two-dimensional material layer 8. The GaN barrier layer 3, AlN reflective layer 4, AlGaN barrier layer 5, and GaN cap layer 6 together constitute a GaN / AlGaN heterojunction. Based on the synergistic effect of the two-dimensional material layer 8 and the GaN / AlGaN heterojunction, the dual-function integrated phototransistor switches functional modes by controlling the bias voltage through the combination of the two-dimensional material layer 8 and the GaN / AlGaN heterojunction.
[0040] Specifically, the functional modes of the dual-function integrated phototransistor include photodetector mode and photosynaptic mode. In photodetector mode, the dual-function integrated phototransistor functions as a high-performance photoconductor or phototransistor. When illuminated, photons are absorbed not only by the GaN-based material to generate photogenerated carriers but also by the upper two-dimensional material layer. These photogenerated carriers alter the Fermi level of the two-dimensional material layer and the carrier concentration in the lower AlGaN / GaN heterojunction 2DEG channel, which is sensitively detected as a significant change in source-drain current. The presence of the AlN reflective layer enhances light absorption, resulting in high responsivity and detectivity. This is a fast, linear photoelectric conversion process, corresponding to the sensing of optical signals. In photosynaptic mode, relying on the synergistic effect of the two-dimensional material layer and the GaN / AlGaN heterojunction, a bias voltage can be applied to the top two-dimensional material layer and the bottom heterojunction, respectively, enabling the phototransistor to perform neural synaptic functions. By controlling these bias conditions, the device can be guided to operate primarily in a fast photoelectric response (photodetection mode) state, or in a nonlinear dynamic response (photosynaptic mode) state with memory and computational capabilities.
[0041] Example 2
[0042] This embodiment also provides a method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction, specifically including the following steps:
[0043] S1. Select one of sapphire, SiC, or Si as the substrate material, perform structuring treatment on the substrate, and prepare patterns on the substrate surface to obtain a patterned substrate layer.
[0044] Specifically, the patterns include, but are not limited to, periodic nanopillar arrays, micron-scale grooves, or inverted pyramid structures, used to enhance light absorption, reduce dislocation density, and improve heat dissipation. The structured pattern of the substrate acts as a light trap, improving the light absorption efficiency of the device and enhancing the responsivity of the photodetector mode. The induced lateral epitaxial effect can effectively release stress and significantly reduce the dislocation density of the epitaxial layer. This reduces the dark current noise of the detector and provides a high-quality semiconductor channel foundation for the precise and stable control of weights in the photodetector synaptic mode.
[0045] S2. Deposit a buffer layer on top of the substrate using chemical vapor deposition (MOCVD).
[0046] Specifically, the buffer layer includes, but is not limited to, AlN buffer layer, GaN buffer layer, and AlGaN buffer layer, with a thickness of 10 nm to 1000 nm. It can reduce dislocation density, provide a single-crystal template, and prevent epitaxial layer cracking or warping. The generated AlN buffer layer provides a crucial growth template for subsequent high-quality nitride epitaxial layers, effectively alleviating the significant lattice and thermal mismatch between the substrate (such as sapphire or SiC) and the GaN-based epitaxial layer. This significantly reduces the penetration dislocation density, fundamentally improving the crystal quality of the active region. It lays a low-defect-density material foundation for forming high-performance two-dimensional electron gas channels, ensuring high carrier mobility and long lifetime. It also significantly reduces device leakage current and noise levels, ensuring the prerequisites and structural pillars for the device's photoelectric detection sensitivity, synaptic weight modulation stability, and overall operational reliability.
[0047] S3. Deposit a GaN barrier layer on top of the AlN buffer layer using chemical vapor deposition (MOCVD).
[0048] Specifically, the GaN barrier layer includes, but is not limited to, p-GaN, n-GaN, and i-GaN barrier layers, with a thickness of 1 nm to 200 nm. It is a core component in forming the two-dimensional electron gas (2DEG). The key to generating an extremely high concentration and high mobility 2DEG lies in the piezoelectric polarization and spontaneous polarization effects of the two materials. The GaN barrier layer, together with the AlGaN barrier layer or AlN reflective layer, forms a heterojunction. Its narrow bandgap and the adjacent wide bandgap material create a significant conduction band step, thus forming a sharp quantum well at the interface, effectively confining the high-concentration 2DEG within it. This provides a high-mobility, low-noise fast carrier channel for photoelectric detection modes, greatly improving the device's responsivity and response speed. Furthermore, it provides an ideal physical carrier for the precise and dynamic adjustment of the weights (conductivity) of simulated biological synapses in artificial synapse modes.
[0049] S4. An AlN reflective layer is deposited on top of the GaN barrier layer using chemical vapor deposition (MOCVD).
[0050] Specifically, the AlN reflective layer has a thickness of 0.1 nm to 10 nm, which can reduce lattice mismatch, lower interface defect density, significantly enhance the polarization effect at the interface, reduce the impact of alloy disorder scattering on electron mobility, and improve the mobility of the two-dimensional electron gas in the GaN / AlGaN heterojunction. With this structure, the AlN reflective layer modulates the two-dimensional electron gas channel located between the AlGaN barrier layer and the GaN channel layer, achieving enhanced polarization and atomic-level interface optimization. This significantly enhances the total polarization electric field at the heterojunction interface and increases the areal density of the two-dimensional electron gas. Simultaneously, its excellent lattice matching characteristics with the upper and lower layers provide an ideal heterojunction interface with atomically flat surface and extremely low defect density, effectively suppressing alloy disorder scattering and ensuring high mobility of channel carriers (two-dimensional electron gas).
[0051] S5. Deposit an AlGaN barrier layer on top of the AlN reflective layer using chemical vapor deposition (MOCVD).
[0052] The AlGaN barrier layer includes, but is not limited to, p-AlGaN, n-AlGaN, and i-AlGaN, with an Al composition of 0.1–0.8% and a thickness of 1 nm–500 nm. It is a core component in forming the two-dimensional electron gas, confining charge carriers within the active region and improving luminescent recombination efficiency. As a core absorption layer, changing the Al composition enables detection in the UVA–UVC bands. This structure, with its wide and tunable bandgap, makes it a highly efficient ultraviolet absorber, generating electron-hole pairs through intrinsic absorption, providing an initial source of photogenerated charge carriers for photoelectric detection. The significant polarization difference and band shift between it and the GaN channel layer together construct a strong built-in electric field and a steep quantum well. This not only provides the fundamental driving force for generating high-concentration, high-mobility two-dimensional electron gases under undoped conditions but also lays the physical foundation for simulating weight changes by externally controlling the interface charge distribution in artificial synaptic modes.
[0053] S6. Deposit a GaN cap layer on top of the AlGaN barrier layer using chemical vapor deposition (MOCVD).
[0054] Specifically, the GaN cap layer includes, but is not limited to, p-GaN, n-GaN, and i-GaN, with a thickness of 0.1 nm to 10 nm. It can passivate surface states, suppress the interaction between the AlGaN surface and environmental substances, suppress surface effects, suppress current collapse, reduce dynamic on-resistance, and enhance device reliability. From a surface passivation perspective, this structure effectively suppresses the sensitivity of AlGaN surface states to environmental gases and surface defects, thereby significantly reducing current collapse effects. From an interface engineering perspective, it provides a more stable Schottky contact interface for the subsequent metal gate, significantly reducing gate leakage current. From a stability perspective, it acts as a robust atomic-level protective shell, isolating the sensitive high-aluminum composition barrier layer below from the external environment, ensuring the stability of the device's electrical parameters during dynamic operation and long-term operation.
[0055] S7. Deposit a metal oxide thin film layer on top of the GaN cap layer using ALD.
[0056] Specifically, the metal oxide thin film layer is colorless and transparent, including but not limited to Al2O3 and ZnO, with a thickness of 1 nm to 100 nm. The metal oxide thin film forms a tunneling layer, which can reduce defects between the two-dimensional material layer and the GaN layer. Utilizing the rectifying properties of the metal oxide layer, it improves the contact quality between the metal electrode and the metal oxide layer, significantly reducing dark current and enhancing light absorption and efficient carrier injection and confinement. With this structure, in terms of electrical performance, its high dielectric constant effectively shields Coulomb scattering in the channel, improving carrier mobility, while significantly reducing gate leakage current through high-quality dielectric properties. In terms of photoelectric conversion, its appropriate band structure can serve as a selective transport layer for photogenerated carriers, promoting efficient separation and collection of electron-hole pairs. In terms of interface modification, its dense amorphous structure can perfectly passivate the surface dangling bonds and defect states of the underlying material, significantly suppressing current collapse effects. In terms of functional expansion, its abundant defect states and oxygen vacancies can serve as controllable charge trapping centers, providing a physical basis for realizing multi-valued storage and simulating the plastic behavior of neural synapses. Therefore, the metal oxide layer serves not only as a performance enhancement layer but also as a functional extension layer, creating key structural conditions for multifunctional reconfigurable devices to achieve high performance, high stability, and new functional characteristics.
[0057] S8. A two-dimensional material layer on top of a metal oxide thin film layer by a growth method or a transfer method.
[0058] Specifically, the two-dimensional material layer includes, but is not limited to, Graphene, MoSe2, and MoS2, with a thickness of 2 nm to 400 nm, and can be patterned. Its growth methods include, but are not limited to, chemical vapor deposition and physical vapor deposition, and its transfer methods include, but are not limited to, wet transfer and dry transfer. Van der Waals forces are used to bond the two-dimensional material with the metal oxide thin film to form a heterojunction. The good conductivity of the two-dimensional material enhances the output efficiency of photogenerated carriers. Simultaneously, the two-dimensional material can form unique cross-dimensional Type-I or Type-II heterojunctions with the underlying material, modulating the behavior of the original 2DEG and generating new transport channels or optical transitions at the interface, making it suitable for high-performance sensors, photodetectors, or ultra-low-power logic devices. With this structure, from the perspective of photoelectric detection, its atomically thin thickness and strong light-matter interaction make it a highly efficient light absorber and carrier generation layer, significantly broadening the effective sensing spectrum and improving responsivity. From the perspective of artificial synapse mode, its rich interface states, tunable defect energy levels, and extremely sensitive carrier transport characteristics to external fields (light and electricity) provide a natural physical carrier for simulating the short-term plasticity and long-term enhancement / inhibition dynamic weight update behavior of biological synapses. At the same time, the cross-dimensional van der Waals interface formed by this layer material and the underlying nitride heterojunction not only effectively passivates the surface and suppresses current collapse, but its unique band alignment and low-dimensional characteristics also become a functional switching hub for flexibly controlling the dominant physical processes (photoelectric conversion or charge trapping / release) through low bias voltage.
[0059] S9. Metal electrodes are prepared on a two-dimensional material layer by electron beam evaporation to obtain a dual-function integrated phototransistor.
[0060] Specifically, positive and negative electrodes of a symmetrical electrode are prepared by electron beam evaporation of 10 nm Cr and 50 nm Au metal electrodes on a two-dimensional material layer, with growth temperatures including but not limited to 100 ℃ ~ 2000 ℃. The metal electrodes are Au, Cr / Au, Ti / Au, or Ti / Al / Ni / Au, and the thickness of the metal electrodes can be 5 nm ~ 20 μm. The substrate with the two-dimensional material is firmly fixed on the sample stage in a vacuum chamber. Under high vacuum, the metal source is bombarded with an electron beam, causing it to melt and evaporate. The vapor condenses into a film on the low-temperature two-dimensional material substrate. The deposition rate and thickness of the film are monitored to finally obtain the metal electrode. From an electrical function perspective, this structure, acting as a low-resistance ohmic contact window, provides an efficient carrier injection and collection path for the two-dimensional electron gas in the channel, ensuring the device's current-driven capability at high speeds. From a photoelectric signal conversion perspective, its optimized work function and specific geometry can form a built-in electric field with the semiconductor layer, efficiently separating and rapidly extracting photogenerated carriers, thereby significantly improving the responsivity and speed of photoelectric detection. From a dynamic function switching perspective, as the direct application terminal of external bias, it serves as the physical interface for precise and rapid reconfiguration of the device between operating modes (detector / synapse). By adjusting the applied voltage amplitude and timing, the charge distribution and trapping state at the interface can be actively manipulated. From a device integration and performance optimization perspective, as part of the heat sink, it effectively dissipates Joule heat, ensuring operational stability at high power densities. Simultaneously, its specific interdigitated or gridded design can reduce series resistance and enhance local field effects without significantly blocking incident light.
[0061] The GaN-based heterojunction-based dual-function integrated phototransistor provided in this embodiment supports dynamic operating mode switching. By adjusting the gate bias voltage, it can seamlessly switch between sensing mode (high gain, fast response) and cognitive mode (memory, learning). For example, when high-sensitivity detection is required, the device can be biased in sensing mode; while when pattern recognition or signal processing is required, it can be switched to cognitive mode by adjusting the bias voltage. This flexibility allows a single device to adapt to multiple application scenarios, reducing system complexity and power consumption. In artificial vision systems, traditional solutions require transmitting the light signals collected by the sensor to the processor for separate processing, resulting in significant latency and power consumption. This device, however, can directly perform preliminary processing of the light signals at the sensor end, such as edge detection and feature extraction, significantly reducing data transmission volume. By arraying and integrating such devices, intelligent vision chips with real-time processing capabilities can be constructed for applications such as autonomous driving and robot navigation. In the fields of optical communication and the Internet of Things, the device can serve as both a high-speed photodetector and perform simple analysis and recognition of light signals. For example, in visible light communication systems, the device can simultaneously achieve signal reception and modulation recognition, simplifying receiver design. In IoT nodes, this integrated device can reduce system complexity and extend battery life.
[0062] In a specific implementation, the fabrication method of a bifunctional integrated phototransistor based on a GaN-based heterojunction includes:
[0063] First, a 2-inch c-plane sapphire was selected as the substrate layer 1. A periodic nanopillar array with a period of 500 nm and a depth of 200 nm was prepared on the substrate surface using photolithography combined with inductively coupled plasma etching technology to improve the crystal quality of the subsequent epitaxial layer and enhance the light absorption efficiency.
[0064] Subsequently, the patterned substrate was placed in a metal-organic chemical vapor deposition apparatus, and the temperature was raised to 1050°C in a hydrogen atmosphere to first grow a 100 nm thick AlN buffer layer 2 to effectively alleviate the lattice mismatch between sapphire and GaN.
[0065] Subsequently, in the same reaction chamber, the temperature was adjusted to 1000°C, and the following layers were grown sequentially: a 50 nm thick undoped GaN barrier layer 3; a 1 nm thick AlN reflective layer 4; a 25 nm thick AlGaN barrier layer 5 with an aluminum composition of 0.25; and a 2 nm thick GaN cap layer 6. The entire epitaxial growth process was carried out under low pressure to ensure atomic-level flatness of the interfaces between each layer. After epitaxial growth, the epitaxial wafer was removed and transferred to an atomic layer deposition (ALD) apparatus to grow a metal oxide thin film. Using trimethylaluminum and water as precursors, a 10 nm thick Al₂O₃ thin film was deposited on the GaN cap layer 6 at 200°C as the metal oxide layer 7. This layer was used for interface passivation and served as an ideal substrate for subsequent two-dimensional materials.
[0066] Subsequently, a monolayer of graphene can be prepared on copper foil as a two-dimensional material layer 8 using chemical vapor deposition (CVD). The two-dimensional material grown by CVD is completely transferred from its metal growth substrate (copper foil) to the target functional substrate using standard PMMA (polymethyl methacrylate) wet transfer technology. Alternatively, a PMMA solution can be spin-coated onto the MoS2 / copper foil to form a PMMA support layer, followed by etching away the copper foil with ammonium persulfate solution, transferring the remaining PMMA / MoS2 film to the surface of the sample with deposited Al2O3. Finally, the PMMA support layer is dissolved and removed with acetone, completing the integration of the two-dimensional material layer. The MoS2 layer is patterned using electron beam lithography and oxygen plasma etching to define the two-dimensional material layer.
[0067] Metal electrode patterns were defined on a two-dimensional material layer using electron beam lithography. A 5 nm thick chromium adhesion layer and a 50 nm thick gold layer were then sequentially deposited using electron beam evaporation to form metal electrode 9. A subsequent lift-off process was then performed to ultimately form a symmetrical interdigitated electrode structure.
[0068] like Figure 2 As shown, this is a schematic diagram of the dual-mode bias switching of a dual-function integrated phototransistor. The embodiment verifies the device's performance in two operating modes through electrical and photoelectric tests. All tests were conducted at room temperature with different bias voltages for mode switching. Within the bias range of 0–0.6 V, the device exhibits photodetector characteristics, with rise and fall times in the millisecond range, indicating rapid separation and extraction of photogenerated carriers. When the bias voltage rises to 0.6–1.0 V, the device switches to photosynaptic mode, with a significantly enhanced fall time of 20 s. Its excitatory postsynaptic current (EPSC) response exhibits typical neuromorphic behavior.
[0069] like Figure 3The diagram illustrates the formation mechanism of a bifunctional integrated phototransistor. In the absence of light and bias, due to the polarization effect of the GaN / AlGaN heterojunction and the confinement effect of the 1 nm AlN and 1 nm GaN layers, two-dimensional electron gases and two-dimensional hole gases are formed in the AlN and GaN layers, respectively. Due to the presence of the two-dimensional graphene material layer, a small number of holes in the AlGaN / GaN heterojunction have the opportunity to tunnel directly through the Al2O3 layer and be captured by the graphene.
[0070] Under zero bias conditions, the strong polarization electric field at the AlGaN / GaN heterojunction interface induces the formation of a high-density two-dimensional electron gas, which dominates the rapid separation and transport of charge carriers. Photogenerated electron-hole pairs are rapidly separated under the influence of the built-in electric field. At this point, photogenerated electron-hole pairs generated by the AlGaN / GaN bulk material separate; photogenerated electrons drift into the AlGaN / GaN bulk material, while holes accumulate in Gr, resulting in a transverse potential gradient parallel to the device surface. The resulting two-dimensional electron gas channels create an extremely strong built-in electric field in the AlGaN barrier layer. This strong built-in electric field of the AlGaN / GaN heterojunction rapidly separates photogenerated electrons and holes. Photogenerated electrons are swept towards the heterojunction interface, while photogenerated holes are pushed towards the device surface or electrodes. Photogenerated electrons swept to the interface are immediately injected into the 2DEG channels. Due to the extremely high electron mobility and extremely low resistance of the 2DEG, these photogenerated electrons can be rapidly transported to the electrodes with almost no loss, forming a photocurrent. This avoids the probability of electrons being scattered or recombine by defects or impurities during transmission, thus significantly improving the collection efficiency of photogenerated carriers, which is equivalent to improving the device's response speed and responsivity. Simultaneously, the alumina layer selectively filters out low-energy hot electrons through quantum tunneling (barrier height ~3.1 eV), reducing the dark current to ~10 eV. -12 It achieves a response on the order of A while maintaining a high responsivity of 69 A / W (285 nm, 0 V, 30 uW / cm2). This process exhibits a millisecond-level transient response, consistent with the carrier dynamics characteristics of traditional photodetectors.
[0071] When a bias voltage is applied, the external electric field begins to significantly modulate the band structure of the heterojunction. The bias voltage lowers the Schottky barrier at the graphene / GaN interface. This allows more photogenerated carriers (e.g., holes injected from GaN into the graphene) to overcome the lowered alumina barrier via the Fowler-Nordheim tunneling mechanism and be injected into the heterojunction interface or the graphene on the other side. However, not all injected carriers are immediately collected. Numerous defect states exist at the heterojunction interface, and these traps capture these injected carriers. For example, electrons trapped by interface states form localized positive charge centers. These trapped charges do not immediately recombine or disappear but reside for a long time. This is equivalent to "storing" the information from the light stimulus in the interface. From a device perspective, these trapped charges modulate the barrier at the interface, resulting in a device conductivity that remains higher than the initial dark-state conductivity even after the light is turned off. This perfectly simulates the short-term memory (STM) and long-term memory (LTM) functions of photoelectric synapses in biological synapses, where synaptic weights are enhanced by stimulation. When the light stimulation stops, the trapped charge carriers are slowly released through thermal excitation or tunneling. This release process leads to a gradual decrease in stored charge, a slow recovery of the interfacial barrier, and a slow decay of the device's conductivity back to its original dark-state level. This slow decay process corresponds precisely to the signal intensity decay and forgetting mechanism in biological synapses.
[0072] Meanwhile, the applied bias voltage directly determines the efficiency of carrier injection and the degree of trap filling. A higher bias voltage lowers the potential barrier more, resulting in higher carrier injection efficiency, more traps being filled, greater conductance changes, and a longer memory effect (duration). This is equivalent to adjusting the "intensity" and "memory time" of synaptic learning. Different bias voltage polarities can control whether electrons or holes become the primary injection and trapping carriers, thus achieving more precise control over synaptic behavior.
[0073] In two-dimensional (MoS2) heterojunction devices, the charge distribution on the material surface can be modulated by a brief period of low gate voltage, thereby changing the intensity of the built-in electric field and controlling the photocurrent. By driving the migration of ions in the metal oxide layer (7) or changing the distribution of interface charge through bias voltage, the conductivity of the underlying AlGaN / GaN channel can be non-volatilely modulated. This change in conductivity can be regarded as an update of "synaptic weights". By applying light pulses (simulating neural pulses) of different intensities, numbers, or frequencies in combination with electrical pulses, various biological synaptic functions can be simulated, such as: short-term plasticity: brief weight changes used for instantaneous information filtering. Long-term enhancement / inhibition: persistent weight changes, which are the basis of learning and memory. Pulse time-dependent plasticity: adjusting weights according to the order of input pulses, which is a key rule for realizing unsupervised learning.
[0074] like Figure 4 As shown in the schematic diagram of the photosynaptic mode, the conductivity state of the device can be dynamically and reversibly adjusted gradually according to the amplitude, width, and number of voltage pulses, successfully simulating the short-term plasticity and other learning behaviors of biological synapses. The photosynaptic mode can be applied to neuromorphic computing to construct an integrated sensor-memory-computing system. To verify the performance of the photosynaptic mode, this invention sets several variables in the photosynaptic mode (at a 1V bias voltage): different light wavelengths, different light powers, and different numbers of light pulses. For example... Figure 4 As shown in a and d, at 254 nm, the size of the photosynapse increases with increasing optical power. This means that at low optical power, carriers are mainly trapped by shallow energy levels (polarization charges in the graphene), forming short-term memory (STM), while high optical power can excite the filling of deep trapped states (GaN / AlGaN polarization charges), promoting the formation of long-term memory (LTM), demonstrating the photosynaptic performance of this invention under the synergistic effect of bias voltage at the 254 nm band. Simultaneously, the experiment also showed the same performance at 285 nm and bias voltage. Figure 4 As shown in figures b and e, experiments were conducted at 254 nm and 285 nm, with a bias voltage of 1 V. The pulse width remained constant at 1 second, while the interval between consecutive pulses varied from 1 second to 10 seconds, demonstrating the excitatory postsynaptic current (EPSC) recorded at different pulse frequencies. At 254 nm and 285 nm, the EPSC value increased significantly proportionally with the increase in the number of pulses, indicating that the device is capable of simulating biological synaptic behavior, particularly long-term potential (LTP) in sharp pulse-dependent plasticity (SRDP). The experimental data effectively demonstrate that, under this mechanism, the EPSC amplitude increases with increasing pulse frequency, simulating the reinforcement process of biological synapses under repeated stimulation. This verifies that the device can achieve the function of a biological synapse under the combined action of UVC, UVB, and a 1 V bias voltage. Figure 4As shown in c and f, the present invention tested photosynapses at 254 nm and 285 nm, respectively. The results show that 10 cycles are required for the first learning, while only 4 cycles are needed for the second learning to achieve the same effect as the first. The observed shortening of the learning cycle (from 10 cycles to 4 cycles) indicates that a stable conductive path or defect state distribution is formed inside the device during the initial photostimulation training. These structural changes lower the energy barrier for subsequent learning. Specifically, under the action of high-energy photons at 254 nm, the local electric field of deep-level defects promotes the migration and recombination of acceptor ions, forming a low-resistance path; while under near-band-edge excitation at 285 nm, the polarization electric field of the GaN / AlGaN interface is enhanced, resulting in a 38% increase in carrier injection efficiency. This photoinduced structural memory effect reduces the external stimulus intensity required for synaptic weight updates, significantly improving neural plasticity and providing key experimental evidence for the development of adaptive neuromorphic optoelectronic systems.
[0075] like Figure 5 As shown, a schematic diagram of the photodetector mode of a dual-function integrated phototransistor is presented. Applying a bias voltage of 0 V to 0.5 V to the dual-function integrated phototransistor causes the device to enter photodetector mode, exhibiting excellent photoelectric response. When irradiated with 285 nm or 254 nm ultraviolet light pulses, the light illuminates the two-dimensional material layer (8) and the AlGaN / GaN heterojunction, generating photogenerated carriers. Under the action of the channel electric field, the photocurrent is rapidly amplified and readout, and the device generates a fast and stable photocurrent response with a dark current as low as 10. -14 The self-driven PDCR (photoluminescence-darkness ratio) reaches a maximum of 3534, the responsivity reaches a maximum of 689 mA / W, the external quantum efficiency (EQE) reaches 337%, and the detectivity (D*) reaches 5.75 × 10⁻⁶. 13 It can simulate the instantaneous sensing function of the retina, exhibiting the characteristics of a high-performance photodetector. The photodetector mode is suitable for high-speed optical communication and ultraviolet imaging.
[0076] like Figure 5 As shown in a and f, in photodetector mode (0 V bias), the IT characteristic curves indicate that the invention has corresponding performance at both 254 nm (UVC) and 285 nm (UVB). The device relies entirely on the built-in electric field to drive carrier separation, and can operate without an external power supply, significantly reducing system energy consumption. Figure 5 As shown in b and g, the photocurrent increases with increasing optical power. Figure 5 As shown in c and h, the response time is 0.471 s at 254 nm and 0.300 s at 285 nm, demonstrating a fast response to UVB and UVC without applying a bias voltage. Figure 5As shown in d and i, at 254 nm, the responsivity (R) and detectivity (D*) of the device gradually decrease with increasing optical power. This is because at 254 nm, when the optical power increases, the concentration of photogenerated carriers increases, leading to a significant increase in the Auger recombination probability between high-density carriers (∝n). 3 The saturation of trapped states significantly shortens the effective carrier lifetime and reduces the responsivity (R∝τ). Simultaneously, the saturation of trapped states prevents shallow-level defects from extending carrier transport time, accelerating recombination. Furthermore, the space charge effect partially shields the built-in electric field, weakening carrier separation efficiency. However, at 285 nm, the responsivity (R) and detectivity (D*) of the photodetector increase linearly with increasing optical power under 285 nm light excitation. This is because under 285 nm light excitation, the device's intrinsic interband transitions dominate light absorption, the optical power is below the Auger recombination threshold (P < 10 mW / cm²), and the AlGaN / GaN polarized built-in electric field effectively suppresses the space charge effect, resulting in stable carrier lifetime (τ) and separation efficiency, leading to lower responsivity (R) and detectivity (D*). ∗ It increases linearly with optical power.
[0077] like Figure 5 As shown in e and g, at 254 nm, the external quantum efficiency (EQE) is greater than 150%, with a maximum quantum efficiency of approximately 330%, and decreases with increasing incident light intensity. At 285 nm, the EQE ranges from 5% to 35%, and increases with increasing incident light intensity. The wavelength-dependent quantum efficiency variation observed in this invention can be explained by the coupling mechanism between the material bandgap and carrier dynamics. Under high-energy photon excitation at 254 nm, the photon energy is higher than the defect levels (oxygen vacancies or AlGaN interface states) of AlGaN, triggering a defect-assisted carrier multiplication effect—a single photon generates multiple electron-hole pairs through cascade ionization or tunneling injection, resulting in an EQE exceeding 100%. As light intensity increases, defect states are gradually occupied by photogenerated carriers, Auger recombination dominates the loss, leading to a decrease in the gain factor. At 285 nm near-band edge excitation, the photon energy is close to the bandgap of the GaN-dominant region. Photogenerated carriers are mainly generated through interband transitions, and their concentration increases linearly with light intensity, enhancing the separation efficiency of the built-in electric field for carriers. Simultaneously, high injection conditions saturate the Shockley-Read-Hall recombination centers, reducing recombination losses. Therefore, the EQE increases with light intensity. This wavelength-selective response essentially reflects the competition mechanism between material defect states and interband transitions for photoexcitation modes.
[0078] In summary, this invention provides a phototransistor based on an AlGaN / GaN heterojunction integrated with a two-dimensional material / metal oxide thin film. Through the optimization of the interface by the metal oxide thin film layer 7 and the innovative combination of the two-dimensional material layer 8 and the GaN / AlGaN heterojunction, high-performance photoelectric sensing capabilities and tunable neuromorphic cognitive functions are integrated into a single device. This novel phototransistor can dynamically switch its operating mode through low bias voltage, providing a practical technical solution to the energy efficiency bottleneck of separating sensing and computing.
[0079] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.
Claims
1. A bifunctional integrated phototransistor based on a GaN-based heterojunction, characterized in that, The substrate (1) includes a buffer layer (2), a GaN barrier layer (3), an AlN reflective layer (4), an AlGaN barrier layer (5), a GaN cap layer (6), a metal oxide thin film layer (7), a two-dimensional material layer (8), and a metal electrode (9) grown sequentially on the substrate (1) using an epitaxial growth method. The GaN barrier layer (3), the AlN reflective layer (4), the AlGaN barrier layer (5), and the GaN cap layer (6) together constitute a GaN / AlGaN heterojunction. Based on the synergistic effect of the two-dimensional material layer (8) and the GaN / AlGaN heterojunction, the dual-function integrated phototransistor is switched in function mode by controlling the bias voltage.
2. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 1, characterized in that, Including the following steps: Choose one of sapphire, SiC, or Si as the substrate, perform structuring on the substrate, and prepare patterns on the substrate surface to obtain a patterned substrate layer; A buffer layer is deposited on top of the substrate using chemical vapor deposition. A GaN barrier layer was deposited on top of an AlN buffer layer using chemical vapor deposition. An AlN reflective layer was deposited on top of a GaN barrier layer using chemical vapor deposition. An AlGaN barrier layer was deposited on top of the AlN reflective layer using chemical vapor deposition. A GaN cap layer was deposited on top of the AlGaN barrier layer using chemical vapor deposition. A metal oxide thin film layer is deposited on top of the GaN cap layer using ALD; Two-dimensional material layers are grown on top of metal oxide thin films using experimental transfer or chemical vapor deposition / physical vapor deposition. Metal electrodes are fabricated on a two-dimensional material layer using electron beam evaporation.
3. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The pattern is a periodic array of nanopillars, a micrometer-scale groove, or an inverted pyramid structure, used to improve light absorption, reduce dislocation density, and enhance heat dissipation.
4. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The buffer layer is AlN, GaN or AlGaN, and the thickness of the buffer layer is 10 nm to 1000 nm.
5. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The GaN barrier layer is p-GaN, n-GaN, or i-GaN, and the thickness of the GaN barrier layer is 1 nm to 200 nm.
6. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The AlGaN barrier layer is p-AlGaN, n-AlGaN, or i-AlGaN, with an Al composition of 0.1 to 0.8 and a thickness of 1 nm to 500 nm.
7. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The GaN cap layer is p-GaN, n-GaN, or i-GaN, and the thickness of the GaN cap layer is 0.1 nm to 10 nm.
8. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The metal oxide thin film is colorless and transparent, including Al2O3 or ZnO, and the thickness of the metal oxide thin film is 1 nm to 100 nm.
9. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The two-dimensional material layer is Graphene, MoSe2, or MoS2, and the thickness of the two-dimensional material layer is 2 nm to 400 nm.
10. The method for fabricating a bifunctional integrated phototransistor based on a GaN-based heterojunction according to claim 2, characterized in that, The metal electrode is Au, Cr / Au, Ti / Au, or Ti / Al / Ni / Au, and the thickness of the metal electrode is 5 nm to 20 μm.