【International Papers】Domain engineered gallium oxide phototransistors enabling intelligent ultraviolet vision with dynamic gating

      Researchers from the Seoul National University have published a dissertation titled "Domain engineered gallium oxide phototransistors enabling intelligent ultraviolet vision with dynamic gating" in InfoMat.

Background

      With the rapid development of machine vision, increasing demands are being placed on real-time performance and energy efficiency in image acquisition and recognition. However, conventional von Neumann architectures suffer from high power consumption and latency due to the physical separation of sensing, memory, and computing units, which makes them particularly unsuitable for edge scenarios where sensing and computation must occur simultaneously. As a result, neuromorphic image sensors that integrate sensing and computation at the device level have attracted growing interest as a promising route toward fast and energy-efficient visual perception. For neuromorphic vision systems, high-selectivity and high–signal-to-noise photodetection is a critical prerequisite. Deep-ultraviolet (DUV) solar-blind detection can inherently suppress interference from visible and ambient light, offering intrinsically high signal-to-noise ratios and enabling reliable operation under complex illumination conditions. Consequently, DUV detection has become an important building block for neuromorphic vision platforms. Among candidate materials, β-Ga₂O₃, an ultra-wide-bandgap semiconductor, stands out due to its wide bandgap, high breakdown field, strong DUV absorption, and excellent thermal stability. Despite these advantages, β-Ga₂O₃-based devices are limited by low carrier mobility, deep trap states, and crystallographic domain boundaries formed during heteroepitaxial growth, which act as scattering centers and trap sites. These intrinsic material and structural issues lead to carrier confinement and persistent photoconductivity (PPC), significantly degrading charge transport and temporal response. Previous studies have mainly relied on circuit-level approaches, such as gate-pulse modulation, to mitigate PPC; however, such strategies do not address the root cause at the material level and often increase system complexity and power consumption. Therefore, there is a clear need for a materials-level solution combined with device and circuit optimization to fundamentally overcome the transport limitations of β-Ga₂O₃. Addressing crystallographic and defect-related issues at the source is essential for realizing high-speed, high-sensitivity, and low-power deep-ultraviolet photodetectors and neuromorphic vision systems.

Abstract

      Domain-induced carrier scattering and persistent photoconductivity (PPC) remain major bottlenecks in β-Ga₂O₃-based deep-ultraviolet (DUV) photodetectors, limiting their use in high-speed imaging, optical communication, and neuromorphic sensing systems. Here, we propose a materials–circuits co-design strategy that integrates crystallographic domain engineering with pulsed-gate modulation to overcome these challenges. Si-doped β-Ga₂O₃ thin films were epitaxially grown on 6° off-axis sapphire substrates, where atomic step edges induced single-domain alignment, resulting in improved lateral carrier transport and reduced leakage current compared with on-axis counterparts. Under comparable pulsed conditions, single-domain devices exhibited faster recovery than previously reported β-Ga₂O₃ phototransistors, highlighting the synergistic interplay between domain-engineered transport and dynamic gating. As a result, the devices achieved a high detectivity (D*) of 9.17 × 10¹⁵ Jones and a fast photoresponse of 0.7 ms under reset conditions, while maintaining stable and energy-efficient operation under sub-volt bias. Beyond individual devices, the optimized phototransistors were integrated into a 24 × 24 active-pixel array for system-level DUV imaging. Coupled with convolutional neural networks (CNNs), the array achieved accurate pattern recognition and image reconstruction, while synaptic depression and active reset processes enabled rapid afterimage suppression and image recovery. Overall, this work establishes a domain-engineered, pulse-modulated β-Ga₂O₃ phototransistor platform that unifies materials-, device-, and system-level innovations, providing a scalable and energy-efficient route toward intelligent ultraviolet vision.

Conclusion

      In this study, we implemented a complementary strategy that integrates materials- and circuit-level approaches to address two major bottlenecks in β-Ga₂O₃-based DUV phototransistors: domain-induced carrier scattering and PPC. Epitaxial growth on off-axis sapphire substrates effectively suppressed rotational domain boundaries, thereby enhancing lateral carrier transport and reducing leakage currents relative to conventional on-axis counterparts. Gate-pulse modulation accelerated carrier recombination and effectively mitigated PPC, leading to sub-millisecond response. When integrated into a 24 × 24 active pixel phototransistor array, the optimized devices enabled high-contrast DUV imaging and accurate CNN-based pattern recognition under sequential illumination. Beyond achieving a D* of 9.17 × 10¹⁵ Jones and a 0.7 ms response time, this work underscores that domain engineering and dynamic gate modulation together provide an effective framework for scalable and intelligent UWBG optoelectronics, providing a practical foundation for real-time neuromorphic ultraviolet vision. Importantly, synaptic depression further allowed clear image reconstruction under sub-volt operation, underscoring the feasibility of low-power neuromorphic sensing. Furthermore, the coexistence of reset and residual-depression functionalities underscores the adaptive nature of domain-engineered β-Ga₂O₃ phototransistors, which can flexibly adapt between rapid-reset operation for fast recovery and residual-depression operation for energy-efficient recognition, depending on the system requirements .

      Building upon these results, we further explored how dynamic gate control can be extended toward more compact array-level operation through a bipolar photogate pulsing scheme. Owing to the wide bandgap and highly insulating nature of the domain-engineered β-Ga₂O₃, the I_ph can be effectively modulated even without auxiliary switching transistors, demonstrating an inherent material-driven advantage that enables stable on/off control and simplified circuit design. Alternating positive and negative pulses stabilized I_ph dynamics, with positive pulses suppressing PPC and negative pulses enabling selective pixel access through current modulation, thereby offering clear benefits in integration density, energy efficiency, and process scalability. These advances underscore the synergistic value of combining material- and circuit-level design, positioning β-Ga₂O₃ phototransistors not only as high-performance DUV detectors but also as versatile building blocks for neuromorphic sensing. Taken together, our findings establish domain-engineered, pulse-driven β-Ga₂O₃ phototransistors as a scalable and energy-efficient platform for next-generation ultraviolet vision, paving the way for integration into multifunctional neuromorphic and optoelectronic systems.