【Member Papers】Low-Temperature Deposition of Polycrystalline ε-Ga₂O₃ for Deep Ultraviolet Perceptual Photodetection

      Researchers from Northeast Normal University have published a dissertation titled "Low-Temperature Deposition of Polycrystalline ε-Ga₂O₃ for Deep Ultraviolet Perceptual Photodetection" in The Journal of Physical Chemistry Letters.

Background

      ε-Ga₂O₃ is a metastable gallium oxide polymorph with intrinsic spontaneous polarization, which can realize both deep ultraviolet photodetection and neuromorphic memory, and has wide applications in secure UV communication and neuromorphic computing. Traditional β-Ga₂O₃ growth requires high temperature, which cannot be integrated with low-temperature flexible or silicon-based chips. Although ε-Ga₂O₃ has smaller lattice mismatch with sapphire, mixed α/β phases easily form during deposition. Disordered oxygen vacancy defects lead to uncontrollable carrier trapping, making it hard to balance fast photoresponse and synaptic plasticity. Previous studies only realize single-function devices, and low-temperature monolithic integration of dual-mode ε-Ga₂O₃ optoelectronic devices is rarely reported, forming an important research gap.

Abstract

      Metastable ε-Ga₂O₃ holds promise for integrated neuromorphic memory and photosensing owing to its spontaneous polarization and low-temperature deposition compatibility. However, realizing its optoelectronic perception function is hindered by poor control over crystalline allotropes and complex defect-mediated carrier trapping. Here, we address these issues by achieving low-temperature (350 °C) deposition of polycrystalline ε-Ga₂O₃ with tailored photocarrier dynamics using reactive oxygen plasma-enhanced atomic layer deposition. This low-temperature strategy suppresses undesired phase transformation and enables defect engineering. Microstructure analyses confirm (002)-oriented polycrystalline ε-Ga₂O₃ with a triple domain twinning architecture that yields macroscopic pseudohexagonal symmetry and reveal an orientation relationship of ε-Ga₂O₃(002)//α-Al₂O₃ (006) on c-plane sapphire. The ε-Ga₂O₃ deep ultraviolet photodetectors exhibit a rapid recovery time of 0.03 s and a high detectivity of 8 × 10¹¹ Jones under a low bias of 1 V. At biases exceeding 10 V, persistent photoconductivity emerges, attributed to bias-addressed carrier trapping at oxygen vacancy defects of different energy depths. In neuromorphic mode, key synaptic behaviors paired-pulse facilitation, excitatory postsynaptic current, and spike rate-dependent plasticity are emulated, and high-accuracy image recognition is achieved. This work establishes a low-temperature growth strategy for ε-Ga₂O₃ that integrates photodetection and neuromorphic visual functionality in a single material system.

Highlights

      Low-temperature (350 °C) PE-ALD growth of phase-pure (002)-oriented polycrystalline ε-Ga₂O₃on c-plane sapphire is realized.

      Triple-domain twinning microstructure of ε-Ga₂O₃and its epitaxial matching relationship with sapphire are systematically characterized.

      Bias-switchable dual-mode device functions including fast DUV detection and neuromorphic synaptic simulation are demonstrated.

      The fabricated device achieves 83.5% CIFAR-10 image recognition accuracy in simulated convolutional neural network.

Conclusion

      In summary, we demonstrate the low-temperature growth of ε-Ga₂O₃ films at 350 °C via PE-ALD. Cross-sectional TEM and FFT analyses confirm the formation of polycrystalline (002)-oriented ε-Ga₂O₃ with a triple-domain twinning architecture that yields macroscopically pseudohexagonal symmetry and reveal a well-defined orientation relationship of ε-Ga₂O₃(002)//α-Al₂O₃(006) with the c-plane sapphire substrate. The ε-Ga₂O₃ photodetectors exhibit bias-switchable dual-mode functionality. At a low bias of 1 V, the device delivers a fast photoresponse (0.03 s) and high detectivity (8 × 10¹¹ Jones). At an increased bias of 30 V, the enhanced electric field selectively addresses oxygen vacancy defects at different energy depths, amplifying persistent photoconductivity and transitioning the device into a neuromorphic computing mode, where synaptic behaviors (PPF, EPSC, and SRDP) and multistate conductance modulation are emulated. A simulated CNN based on device characteristics achieves 83.5% accuracy on CIFAR-10. These results establish that PE-ALD-grown ε-Ga₂O₃ is a promising candidate for integrated photoelectric sensing and neuromorphic computing platforms.

Project Support

      This work was supported by the National Key Research and Development Program (2021YFA0716404), the National Natural Science Foundation of China (12474164, 52302166, and 62404038), the 111 Project (B25030), and the Research Projects of the Education Office of Jilin Province (JJKH20261332KJ) and Changchun City (25GNYZ60).