Polarization-sensitive photo-synapses based on anisotropic β-Ga₂O₃ for dynamic visual perception

Abstract

Polarization, as a fundamental property of light, carries abundant environmental and target-specific information that is invisible to the human eye but crucial for advanced vision. In biological visual systems, such polarization information is effectively encoded, processed, and integrated to enhance scene perception and target recognition. Inspired by this biological paradigm, polarization sensitive optoelectronic synapses provide a promising pathway toward information perception and neuromorphic computing. Benefiting from its low-symmetry monoclinic lattice, β-Ga₂O₃ inherently exhibits strong optical absorption anisotropy, which can be used to enable photo-synapses. Here, the structural anisotropy of β-Ga₂O₃ single crystals with (100), (010), and (001) orientations was systematically investigated through atomic arrangement analysis, polarization-resolved Raman spectroscopy, polarization-dependent absorption, and XPS characterization. Compared with the (010) and (001) orientations, the (100) β-Ga₂O₃ crystal exhibited stronger optical absorption anisotropy and a higher density of oxygen vacancies, making it a favorable candidate for constructing polarization-sensitive neuromorphic synapses. A solar-blind polarization-sensitive optoelectronic synapse was developed in this work, demonstrating polarization-dependent excitatory postsynaptic currents (EPSC), paired-pulse facilitation (PPF), and learning-forgetting-relearning functionalities. Furthermore, the optoelectronic synapse enabled the construction of a neuromorphic visual system (NVS), achieving noise suppression, enhanced image quality, and handwritten digit recognition accuracy up to 98.74%. Beyond static recognition, integration of the arrays into a reservoir computing system allowed efficient motion direction perception with accuracies approaching 99.7%. These results highlight the potential of anisotropic β-Ga₂O₃ as a material foundation for next-generation solar-blind polarization-sensitive neuromorphic vision systems.

Graphical abstract

Keywords

polarization / β-Ga₂O₃ / anisotropy / optoelectronic synapses / solar-blind

1 Introduction

Polarization is a fundamental property of light that encodes rich environmental and target specific information beyond intensity and wavelength [1−3]. In natural environments, many insects and marine animals have evolved visual systems capable of sensing polarization information, which play a crucial role in navigation, target detection, and communication. In contrast, the human visual system lacks sensitivity to polarization states, resulting in the loss of this important dimension of visual information. Inspired by biological synapses, optoelectronic synaptic devices have attracted extensive attention due to their ability to emulate key synaptic functions under optical stimulation, including excitatory postsynaptic currents (EPSC), paired-pulse facilitation (PPF), and learning-forgetting behaviors. To date, a variety of materials, such as Ga₂O₃ [4], CeO₂ [5] and 2D material MoS₂ [6] have been widely explored for optoelectronic synaptic applications [7−9]. However, most reported devices still suffer from challenges related to large-area uniformity, device-to-device variability, and interface instability, which severely limit long-term reliability and scalable integration. More importantly, the majority of optoelectronic synapses respond only to non-polarized optical stimuli and generally lack polarization sensitivity, making it difficult to perceive and utilize polarization information [10−12].

Conventional polarization-sensitive devices typically rely on external polarizers to achieve selective responses to polarized light. However, such approaches inevitably increase device complexity and size, hindering integration and miniaturization. With the continuous advancement of semiconductor technology, it has become increasingly evident that polarization detection and manipulation can be intrinsically realized by exploiting the crystallographic anisotropy of low-symmetry semiconductors, eliminating the dependence on external optical components [13−15]. Many low-dimensional materials exhibit pronounced in-plane anisotropy, such as ReS₂ and GeAs, which makes it highly attractive for polarization-resolved photodetection. Recent studies have demonstrated that polarization-sensitive photodetectors based on such materials can operate effectively across a broad spectral range, spanning from the infrared to the visible region [16−20]. However, extending polarization detection into the solar-blind deep-ultraviolet (DUV) regime remains challenging. The DUV component of solar radiation is almost completely absorbed by the Earth’s atmosphere during propagation, resulting in a near-zero background radiation level at the ground. Therefore, device operating in this spectral region can intrinsically achieve low-noise and high signal-to-noise ratio without the need for complex optical filtering. These unique characteristics make DUV light highly attractive for next-generation optoelectronic and neuromorphic systems, particularly in secure communication, environmental monitoring, and artificial vision applications [21−26].

Benefiting from its ultra-wide bandgap (~4.9 eV) and inherently low-symmetry monoclinic crystal structure, β-Ga₂O₃ exhibits intrinsic optical anisotropy that can be exploited for polarization-sensitive photodetection in the DUV region. Recent studies have confirmed that the anisotropy of β-Ga₂O₃ can significantly enhance the polarization selectivity of optoelectronic devices [27−30]. For example, Wei et al. theoretically elucidated the anisotropic solar-blind UV polarization absorption mechanism in β-Ga₂O₃ by analyzing electronic transition probabilities, and further realized (100)-oriented β-Ga₂O₃ nanobelts for solar-blind polarization photodetection without external polarizers. Owing to morphology-enhanced anisotropy, the device exhibited a high polarization ratio under 254 nm linearly polarized illumination [31,32]. In our previous work, we proposed a lattice-engineering-based anisotropy regulation strategy to address the common issue of isotropic features in β-Ga₂O₃ epitaxial films caused by the symmetry transfer effect during conventional heteroepitaxy [33]. By constructing a synergistic optimization system between substrate structure and growth kinetics, we successfully broke the symmetry limitation of heteroepitaxy and achieved significant enhancement of anisotropy in β-Ga₂O₃ epitaxial films. However, despite the intrinsic optical anisotropy of β-Ga₂O₃, polarization-sensitive optoelectronic synapses based on this material have been rarely reported.

In this study, the crystallographic anisotropy of β-Ga₂O₃ single crystals with (100), (010), and (001) orientations is systematically analyzed by combining atomic structure analysis with polarized Raman spectroscopy. Polarization-resolved absorption spectroscopy and X-ray photoelectron spectroscopy (XPS) are further employed to investigate the orientation-dependent optical absorption and oxygen-vacancy distribution. The results reveal that the (100)-oriented β-Ga₂O₃ crystal exhibits stronger absorption anisotropy and a higher oxygen-vacancy concentration than the (010) and (001) counterparts, indicating its greater potential for polarization-sensitive neuromorphic applications. Based on anisotropic β-Ga₂O₃ crystals with favorable optical and defect characteristics, a solar-blind polarization-sensitive optoelectronic synapse is demonstrated. The device exhibits polarization-dependent excitatory EPSC, paired-pulse facilitation, and dynamic learning-forgetting-relearning behaviors. Furthermore, by integrating a β-Ga₂O₃ synaptic array, a neuromorphic vision system is constructed to realize noise suppression and image quality enhancement. The feasibility of the system is validated through handwritten digit recognition using a convolutional neural network (CNN) as well as motion trajectory perception within a reservoir computing framework, demonstrating the potential of anisotropic β-Ga₂O₃ for polarization-sensitive neuromorphic vision.

2 Method

β-Ga₂O₃ single crystals with (100), (001), and (010) orientations (5 mm × 5 mm × 0.5 mm, undoped, Wuxi Jingdian Semiconductor Materials Co., Ltd.) were used. Ti/Au electrodes (50 nm) with 200 μm spacing and 2800 μm channel length were deposited by magnetron sputtering to form interdigital MSM devices. Polarization-dependent Raman spectra were collected at room temperature using a 532 nm excitation source (Horiba). Optical absorption anisotropy was measured with a UV-Vis spectrophotometer. Device electrical and synaptic properties were tested using Keithley 2400/6487 analyzers and a probe station. A 254 nm UV LED served as the illumination source, calibrated with a UV power meter. In all polarization-dependent experiments, the reported polarization angles are defined relative to the crystallographic orientation of the β-Ga₂O₃ crystal (Fig. S1, Supporting information).

3 Results and discussion

In nature, many insects and marine organisms rely on polarization for navigation, predation, and communication, thereby achieving visual capabilities far beyond those of humans. For example, the visual system of cuttlefish exhibits remarkable detection and recognition abilities. As shown in Fig. 1(a), the visual neurons of cuttlefish possess relatively short axons, in contrast to the long axons commonly found in the human eye. This structural feature significantly shortens the signal transmission path between neurons, thereby improving the efficiency of information processing. In addition, the retinal cell layer of cuttlefish is regularly arranged, a unique configuration that may allow these cells to exhibit direction-dependent absorption and response to polarized light. In complex underwater environments, polarization information is far less affected by medium scattering and absorption than color or intensity information. As a result, cuttlefish can exploit polarization vision to gain significant survival advantages in predator avoidance, defense, and prey capture.

Inspired by both the structural organization and functional superiority of cuttlefish vision, significant efforts have been devoted to developing artificial vision-enhancement systems based on polarization-sensitive optoelectronic synapses. Unlike conventional polarization detection schemes that rely on external polarizers, these emerging systems aim to directly detect and process polarized light by utilizing the intrinsic anisotropy of semiconductor materials. β-Ga₂O₃ crystallizes in a monoclinic structure with space group C2/m and inherently exhibits anisotropic electrical and optical properties owing to its low-symmetry lattice. The lattice parameters are a ≈ 12.23 Å, b ≈ 3.04 Å, c ≈ 5.80 Å, with a monoclinic angle β ≈ 103.7°. Gallium atoms occupy two distinct coordination environments, with Ga(I) located in GaO₄ tetrahedra and Ga(II) in GaO₆ octahedra, while oxygen atoms are distributed over three inequivalent sites, O(I), O(II), and O(III), accompanied by pronounced distortions in bond lengths and bond angles. This complex atomic configuration collectively gives rise to the intrinsic structural anisotropy of β-Ga₂O₃. Consequently, the isotropy of the electronic band structure and optical transition probabilities is broken, leading to strong orientation dependence of optical absorption, carrier transport, and defect formation energies. The atomic packing characteristics differ significantly among distinct crystallographic planes [Fig. 1(b)]. The (100) plane is composed of alternating layers of GaO₄ tetrahedral chains and GaO₆ octahedral chains. Both octahedral and tetrahedral chains extend continuously along the b-axis, forming a seamless connection. GaO₆ octahedra are linked end-to-end through edge-sharing and tetrahedral corner-sharing, creating highly extended chain-like structures. These straight chains produce dense and regular atomic arrangements with the smallest interatomic spacing along the b-axis, thereby enhancing electrical and thermal conduction in this direction. Along the c-axis, however, adjacent octahedral chains are bridged by GaO₄ tetrahedra, forming an interlayer-coupled rather than continuous-chain structure. The interatomic spacing is larger and the packing sparser than along the b-axis, resulting in poorer electron and phonon transport efficiency. On the (010) plane, along the a-axis, octahedral and tetrahedral chains alternate irregularly. Because of the large lattice parameter, periodic layered repetition occurs, and atomic chains are discontinuous, yielding weaker transport properties. Along the c-axis, in contrast, GaO₆ octahedra and GaO₄ tetrahedra are alternately and tightly arranged. Given the shorter lattice parameter, interatomic spacing is smaller and packing denser, thereby exhibiting stronger anisotropy. On the (001) plane, along the a-axis, GaO₆ octahedra and GaO₄ tetrahedra are alternately distributed but the chains are discontinuous. Along the b-axis, however, octahedral/tetrahedral chains extend continuously, again forming a seamless connection. To further investigate the anisotropic properties of β-Ga₂O₃, ARPR spectroscopy was employed on single crystals cut along different orientations. Figures 1(c)−(e) and S2 presents the ARPR spectra collected from the (100), (010), and (001) planes over polarization angles spanning 0°−180°, allowing a systematic evaluation of in-plane anisotropy. In the monoclinic lattice of β-Ga₂O₃, two GaO₆ octahedra and two GaO₄ tetrahedra link together along the b-axis to form a characteristic double-chain framework. The primitive cell contains 10 atoms, which give rise to 30 vibrational modes, including 27 optical phonons. At the Brillouin zone center, these phonons split into Raman-active modes (Ag, Bg) and infrared-active modes (Au, Bu), as summarized in Eq. (1):

(1){\mathrm{\Gamma }}_{opt}=10A_g+{4A}_u+5B_g+8B_u.

The Raman-active phonons of β-Ga₂O₃ can be broadly classified into three characteristic spectral ranges. Modes appearing below 300 cm⁻¹ primarily originate from lattice translational motions and collective vibrations of the GaO₆−GaO₄ chains. The region between 300 and 600 cm⁻¹ is dominated by distortive vibrations within GaO₆ octahedra. In contrast, signals observed above 600 cm⁻¹ are mainly attributed to symmetric stretching and bending motions of GaO₄ tetrahedra. Across the 100−1000 cm⁻¹ spectral range, twelve reproducible Raman features were detected in all examined orientations [34−36]. As predicted by selection rules, Bg modes, which involve vibrations perpendicular to the b-plane, vanish in the (010)-oriented spectra. Comparative examination further highlights marked intensity variations among the planes. For instance, at 0° polarization, the A_g³ mode near 200 cm⁻¹, originating from GaO4 vibrations, is pronounced in (010) and (001) samples but is much weaker in (100). Conversely, the A_g¹⁰ mode at 764 cm⁻¹ appears strongest in the (100) orientation and is nearly absent in (001). Overall, the polarization-dependent modulation of Raman intensities offers clear experimental evidence of the intrinsic structural anisotropy of β-Ga₂O₃.

The polarization-dependent absorption spectra of β-Ga₂O₃ single crystals with different planes are presented in Figs. 2(a)−(c). For all the samples, the absorption intensity varied significantly with polarization angle, especially in the deep-UV range of 240−300 nm, indicating pronounced optical anisotropy. Such anisotropic behavior originates from the intrinsically low-symmetry monoclinic crystal structure of β-Ga₂O₃. The XPS survey spectrum of all the β-Ga₂O₃ samples confirms the elemental composition and chemical states of the sample surface. As shown in Fig. S3, distinct peaks corresponding to Ga and O are observed, while no additional impurity signals are detected within the detection limit, indicating high material purity. Photo-Synapses exploit the persistent photoconductivity (PPC) effect to emulate essential synaptic functions. In oxide semiconductors, oxygen vacancy (VO) defects serve as the dominant origin of PPC. In the dark state, most VOs in β-Ga₂O₃ exist in a neutral configuration (VO⁰), while a small fraction remains in the ionized state (VO²⁺) [37−40]. Upon illumination with photons of appropriate energy, neutral vacancies can be ionized into charged states, simultaneously releasing free electrons into the conduction band and giving rise to a photoresponse (Fig. S5). It is worth noting that oxide semiconductors are strongly ionic in nature, and variations in the surrounding charge environment can induce lattice relaxation. Such relaxation modifies the potential landscape, and when the local charge environment attempts to recover, an energy barrier arises. Consequently, once the light stimulus is removed, the ionized vacancies cannot immediately return to their neutral state. Instead, their recovery is hindered by this deionization barrier, leading to a slow recombination process. During this period, the released carriers continue to contribute to conduction, resulting in the widely observed PPC phenomenon. Therefore, the presence of a sufficient concentration of oxygen vacancies is crucial for enabling the synaptic functionality of oxide-based devices. The O 1s XPS spectra of the samples obtained directly from the as-prepared sample surface, are provided in the Supporting Information (Fig. S4). To minimize the influence of surface adsorbates and ambient exposure, the sample surface was subsequently etched prior [Figs. 2(d)−(f)]. The O 1s spectra were deconvoluted into two components, which O_Ⅰ corresponding to lattice oxygen bonded with Ga, and O_Ⅱ attributed to oxygen-vacancy-related states. The calculated O_Ⅱ/O_total ratios were 46.82%, 24.51%, and 34.34% for the (100), (010), and (001) planes, respectively. The results show that the oxygen vacancy concentration on the (100) surface is the highest, while the defect density on the (010) surface is the lowest.

In biological systems, thousands of neurons are interconnected through synapses to enable complex signaling. As shown in Fig. 3(a), when an action potential reaches the presynaptic neuron, the opening of voltage-gated calcium channels triggers vesicular exocytosis of neurotransmitters. These neurotransmitters diffuse across the synaptic cleft and bind to specific receptors on the postsynaptic membrane, inducing conformational changes in ligand-gated ion channels and ultimately generating an EPSC. Inspired by this mechanism, electrodes were deposited on different β-Ga₂O₃ to emulate biological functions such as learning and memory, and to investigate the dependence of synaptic behavior on crystallographic orientation. Here, DUV light irradiation generates carriers that act as neurotransmitter analogs, thereby eliciting EPSC in the device. As shown in Fig. 3(b), the β-Ga₂O₃ device with a (100) crystal orientation exhibits clear EPSC responses under 270 nm light pulse stimulation with increasing power densities (from 80 to 300 μW cm⁻²), measured at a bias voltage of 6.5 V with a fixed pulse interval of 2 s. The peak EPSC increases from 560 to 660 nA as the illumination intensity is enhanced, indicating effective optical modulation of synaptic weight. At low light intensities, the induced EPSC decays rapidly to its baseline level, corresponding to short-term memory (STM) behavior dominated by transient photo-generated carriers and shallow trap states. In contrast, as the light intensity increases, the EPSC exhibits a much slower decay with a pronounced residual current, suggesting a transition from STM to long-term memory (LTM), which can be attributed to the increased participation of deep trap states and the accumulation of persistent photocarriers. The device based on the (001)-oriented β-Ga₂O₃ crystal exhibits a similar intensity-dependent evolution of synaptic behavior; however, the magnitude of the EPSC modulation is noticeably smaller [Fig. 3(c)]. In contrast, the device fabricated from the (010)-oriented crystal shows negligible synaptic response, indicating an almost complete absence of synaptic functionality (Fig. S6). In addition to light intensity, repeated stimulation likewise influences synaptic plasticity. As shown in Fig. 3(d), increasing the number of optical pulses from 20 to 40 on the (100) orientation successfully drives the transition from STM to LTM. By contrast, as shown in Fig. 3(e), the (001) orientation exhibits negligible changes under the same conditions. This behavior is likely associated with a saturation effect of oxygen vacancy traps. During the initial stimulation cycles, oxygen vacancies are progressively filled, limiting the ability of subsequent photogenerated carriers to be effectively trapped and stabilized. Consequently, the enhancement of EPSCs is suppressed, hindering further plasticity modulation. Paired-pulse facilitation (PPF) represents a form of short-term synaptic plasticity, wherein a second stimulus pulse (A2), delivered shortly after the first (A1), evokes an enhanced postsynaptic response relative to the initial pulse [Fig. 3(f)]. This behavior is quantitatively expressed as PPF = (A2/A1) × 100%. The (100) orientation exhibits markedly higher PPF values (180%) compared to other crystallographic orientations. With increasing pulse intervals, the PPF indices for all four surfaces exhibit an exponential decay, which can be accurately described by a biexponential fitting function:

(2)PP{F}_{\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}}=1+C_1\exp \left({\displaystyle \frac{-\mathrm{\Delta }t}{{\tau }_1}}\right)+C_2\exp \left({\displaystyle \frac{-\mathrm{\Delta }t}{{\tau }_2}}\right),

where C₁ and C₂ represent the initial facilitation amplitudes, while {\tau }_1 and {\tau }_2 denote the two characteristic decay times. {\tau }_1 is attributed to fast relaxation processes, such as shallow-trap-assisted carrier recombination and rapid detrapping near the surface or interface, while {\tau }_2 corresponds to slower relaxation associated with deep trap states and oxygen-vacancy-related lattice relaxation. The fitted decay times for the (100) orientation are 1.64 s and 2.02 s [Fig. 3(g)]. As shown in Fig. 3(h), the EPSC exhibits a progressive increase with longer illumination times. Notably, the EPSC exhibited a pronounced angular dependence, with the response under 0° polarized illumination being markedly higher than that observed at 90°, thereby confirming its strong polarization sensitivity [Fig. 3(i)].

Biological retinas possess intrinsic memory capabilities that play a pivotal role in the preprocessing of visual information. To emulate this retinal memory function, β-Ga₂O₃ crystals with different orientations were illuminated with a 270 nm optical image at a power density of 300 μW/cm² for 2 s, followed by a 20 s dark relaxation period [Fig. 4(a)]. Figures 4(b)−(d) present the synaptic current responses of the (100), (001), and (010) crystal orientations during illumination and subsequent relaxation. Notably, the (100) and (001) orientations maintain pronounced image contrast even after 20 s of decay, highlighting their superior memory retention capability. In comparison, the (010) orientation almost completely lost image details, highlighting the exceptional visual memory capability of β-Ga₂O₃ (100) orientation based device. The biological retina not only possesses memory capability but also performs preprocessing functions, which provide a solid foundation for subsequent recognition in the brain and significantly enhance recognition accuracy. Inspired by this mechanism, we emulate a neuromorphic visual system for image perception and preprocessing, as shown in Fig. 4(e). The system comprises a β-Ga₂O₃ synaptic array for preprocessing, integrated with a CNN. The β-Ga₂O₃ synaptic array is used to simulate the perceptual and preprocessing functions of biological systems. By introducing sampling points during the current decay process, it can effectively distinguish different illumination conditions. Since signals and noise exhibit significantly different decay characteristics, with high-intensity signals generally decaying slowly while noise decays rapidly, this difference allows the suppression of noise while retaining key information. The preprocessing of the β-Ga₂O₃ synaptic array significantly improves image quality and provides a reliable basis for subsequent data processing. CNN is a deep learning algorithm that has outstanding advantages in image recognition and is used to train and recognize the modified handwritten digits from the Modified National Institute of Standards and Technology (MNIST). The network consists of a 28 × 28 input, followed by two convolution and pooling operations, then transformed into a 256-dimensional vector, where the ReLU function is applied to convert neurons, and the output neurons represent the digits 0−9 together with each weight and bias. As shown in Fig. 4(f), after 20 training epochs, images preprocessed by the (100)-oriented β-Ga₂O₃ array achieved an accuracy of 98.74%. Figures 4(g)−(i) demonstrate the confusion matrices of the test set for the (100), (001), and (010) plane, respectively. It is evident that images preprocessed by the β-Ga₂O₃ synaptic array on the (100) plane achieve superior classification performance, with predicted labels closely matching the ground truth. These results provide compelling evidence that NVS preprocessing enables effective feature extraction and significantly enhances recognition accuracy.

The perception of motion direction is critical importance in biological visual systems. To emulate this capability, we developed a reservoir computing (RC) system based on a β-Ga₂O₃ synaptic array for motion perception. As a class of recurrent neural networks, RC systems demonstrate remarkable proficiency in processing spatiotemporal information. As shown in Fig. 5(a), the system primarily consists of three components: an input layer, a reservoir layer, and an output layer. The input layer is responsible for receiving incoming signals, while the reservoir layer comprises a set of storage nodes that encode the inputs into internal reservoir states. The output layer is typically a linear layer trained for classification using the reservoir states generated by the reservoir. Subsequently, as shown in Figs. 5(b)−(e), we employed the β-Ga₂O₃ synaptic array to obtain a dataset of images representing four motion directions, using an airplane as an example. Each image contains historical frame information, where residual traces left on the synaptic array by preceding frames combine with subsequent frame data to form a clear motion trajectory. Compared with conventional image sensors, this approach enhances the accuracy of motion direction prediction. For the four motion types, training datasets comprising 2000 images each were established, and the RC system was trained over 20 epochs. As shown in Fig. 5(f), the recognition accuracy for the four motion directions reached 99.69%, exhibiting very low training loss [Fig. 5(g)]. The confusion matrix for 800 test images is shown in Fig. 5(h), indicating high classification precision across all four motion directions. These results demonstrate that the constructed β-Ga₂O₃ synaptic array possesses strong potential for motion direction perception, providing a novel solution for motion sensing tasks.

4 Conclusion

In summary, polarization-sensitive neuromorphic synapses based on anisotropic β-Ga₂O₃ were successfully demonstrated, addressing the intrinsic limitation of the human visual system and conventional optoelectronic synapses in perceiving polarization information. Structural anisotropy and oxygen vacancy distributions in β-Ga₂O₃ crystals were systematically correlated with polarization-dependent optoelectronic properties, with the (100) orientation exhibiting the most favorable characteristics. Using these properties, β-Ga₂O₃ synaptic devices realized solar-blind polarization-sensitive EPSC, PPF, and advanced synaptic functionalities, enabling both static and dynamic neuromorphic vision tasks. The constructed NVS achieved superior handwritten digit recognition accuracy, while reservoir computing based on β-Ga₂O₃ arrays provided highly accurate motion perception. These findings establish anisotropic β-Ga₂O₃ as a candidate for next-generation polarization-sensitive optoelectronic synapses and pave the way toward integrated neuromorphic vision systems capable of multidimensional information processing.

References

[1] J. Xiong , Q. Yu , X. Hou , B. Liu , S. Li , H. Deng , Z. Yang , J. Leng , S. Zhu , Y. Sun , Z. Jiang , N. Huo , J. Wu , and P. Zhou , Short‐wave infrared photodetectors based on β‐In₂Se₃/Te heterojunctions for optical communication and polarimetric imaging applications, Adv. Funct. Mater. 34(26), 2314972 (2024)

[2] M. Dai , C. Wang , B. Qiang , F. Wang , M. Ye , S. Han , Y. Luo , and Q. J. Wang , On-chip mid-infrared photothermoelectric detectors for full-Stokes detection, Nat. Commun. 13(1), 4560 (2022)

[3] X. B. Yan , X. Han , Z. L. Fang , Z. Zhao , Z. X. Zhang , J. M. Sun , Y. D. Shao , Y. X. Zhang , L. L. Wang , S. Q. Sun , Z. Q. Guo , X. T. Jia , Y. P. Zhang , Z. Y. Guan , and T. Shi , Reconfigurable memristor based on SrTiO₃ thin-film for neuromorphic computing, Front. Phys. (Beijing) 18(6), 63301 (2023)

[4] J. Yuan , C. Wu , S. Wang , F. Wu , C. K. Tan , and D. Guo , Enhancing plasticity in optoelectronic artificial synapses: A pathway to efficient neuromorphic computing, Appl. Phys. Lett. 124(2), 021101 (2024)

[5] Y. Lin , W. Wang , R. Li , J. Kim , C. Zhang , H. Kan , and Y. Li , Multifunctional optoelectronic memristor based on CeO₂/MoS₂ heterojunction for advanced artificial synapses and bionic visual system with nociceptive sensing, Nano Energy 121, 109267 (2024)

[6] N. Li , C. He , Q. Wang , J. Tang , Q. Zhang , C. Shen , J. Tang , H. Huang , S. Wang , J. Li , B. Huang , Z. Wei , Y. Guo , J. Yuan , W. Yang , R. Yang , D. Shi , and G. Zhang , Gate-tunable large-scale flexible monolayer MoS₂ devices for photodetectors and optoelectronic synapses, Nano Res. 15(6), 5418 (2022)

[7] S. Lee , J. Kim , and S. Kim , Self-aligned TiO_x-based 3D vertical memristor for a high-density synaptic array, Front. Phys. (Beijing) 19(6), 63203 (2024)

[8] C. Yoo,T. J. Ko,M. G. Kaium,R. Martinez,M. M. Islam,H. Li,J. H. Kim,J. Cao,M. Acharya,T. Roy,Y. Jung, A minireview on 2D materials-enabled optoelectronic artificial synaptic devices, APL Mater. 10(7), 070702 (2022)

[9] W. B. Niu , G. L. Ding , Z. Q. Jia , X. Q. Ma , J. Y. Zhao , K. Zhou , S. T. Han , C. C. Kuo , and Y. Zhou , Recent advances in memristors based on two-dimensional ferroelectric materials, Front. Phys. (Beijing) 19(1), 13402 (2024)

[10] W. Wang , S. Gao , Y. Li , W. Yue , H. Kan , C. Zhang , Z. Lou , L. Wang , and G. Shen , Artificial optoelectronic synapses based on TiN_xO_2–x/MoS₂ heterojunction for neuromorphic computing and visual system, Adv. Funct. Mater. 31(34), 2101201 (2021)

[11] K. Wang , J. Wu , M. Wang , F. Zhang , X. Li , M. Xu , D. Zhu , J. Han , J. Liu , Z. Liu , W. Huang , and A. Biodegradable , Stretchable, healable, and self-powered optoelectronic synapse based on ionic gelatins for neuromorphic vision system, Small 20(44), 2404566 (2024)

[12] S. Song , C. Choi , J. Ahn , J. J. Lee , J. Jang , B. S. Yu , J. P. Hong , Y. S. Ryu , Y. H. Kim , and D. K. Hwang , Artificial optoelectronic synapse based on spatiotemporal irradiation to source-sharing circuitry of synaptic phototransistors, InfoMat 6(2), e12479 (2024)

[13] Y. Chen , C. Chen , R. Kealhofer , H. Liu , Z. Yuan , L. Jiang , J. Suh , J. Park , C. Ko , H. S. Choe , J. Avila , M. Zhong , Z. Wei , J. Li , S. Li , H. Gao , Y. Liu , J. Analytis , Q. Xia , M. C. Asensio , and J. Wu , Black arsenic: A layered semiconductor with extreme in-plane anisotropy, Adv. Mater. 30(30), 1800754 (2018)

[14] P. K. Venuthurumilli , P. D. Ye , and X. Xu , Plasmonic resonance enhanced polarization-sensitive photodetection by black phosphorus in near infrared, ACS Nano 12(5), 4861 (2018)

[15] S. Wang , X. Gong , J. Wei , L. Du , F. Wang , S. Lai , X. H. Shao , W. B. Gao , and C. Wang , Low-dimensional tellurium for electronics, optoelectronics, quantum devices and beyond, Front. Phys. (Beijing) 20(3), 034401 (2025)

[16] Z. Zhou , M. Long , L. Pan , X. Wang , M. Zhong , M. Blei , J. Wang , J. Fang , S. Tongay , W. Hu , J. Li , and Z. Wei , Perpendicular optical reversal of the linear dichroism and polarized photodetection in 2D GeAs, ACS Nano 12(12), 12416 (2018)

[17] S. Liu , W. Xiao , M. Zhong , L. Pan , X. Wang , H. X. Deng , J. Liu , J. Li , and Z. Wei , Highly polarization sensitive photodetectors based on quasi-1D titanium trisulfide (TiS3), Nanotechnology 29(18), 184002 (2018)

[18] L. Li,W. Wang,P. Gong,X. Zhu,B. Deng,X. Shi,G. Gao,H. Li,T. Zhai, 2D GeP: An unexploited low-symmetry semiconductor with strong in-plane anisotropy, Adv. Mater. 30(14), 1706771 (2018)

[19] M. Zhong , Q. Xia , L. Pan , Y. Liu , Y. Chen , H. X. Deng , J. Li , and Z. Wei , Thickness-dependent carrier transport characteristics of a new 2D elemental semiconductor: Black arsenic, Adv. Funct. Mater. 28(43), 1802581 (2018)

[20] W. Luo , Y. Cao , P. Hu , K. Cai , Q. Feng , F. Yan , T. Yan , X. Zhang , and K. Wang , Gate tuning of high-performance InSe-based photodetectors using graphene electrodes, Adv. Opt. Mater. 3(10), 1418 (2015)

[21] S. Oh , C. K. Kim , and J. Kim , High responsivity β-Ga₂O₃ metal–semiconductor–metal solar-blind photodetectors with ultraviolet transparent graphene electrodes, ACS Photonics 5(3), 1123 (2018)

[22] Y. Qin , L. H. Li , Z. Yu , F. Wu , D. Dong , W. Guo , Z. Zhang , J. H. Yuan , K. H. Xue , X. Miao , and S. Long , Ultra-high performance amorphous Ga₂O₃ photodetector arrays for solar-blind imaging, Adv. Sci. (Weinh. ) 8(20), 2101106 (2021)

[23] C. Wu , T. Zhao , H. He , H. Hu , Z. Liu , S. Wang , F. Zhang , Q. Wang , A. Liu , F. Wu , and D. Guo , Enhanced performance of gallium-based wide bandgap oxide semiconductor heterojunction photodetector for solar-blind optical communication via oxygen vacancy electrical activity modulation, Adv. Opt. Mater. 12(10), 2302294 (2024)

[24] H. Zhang , R. Zhu , Z. Mei , F. ChiChung·Ling , Y. Zhao , and S. Su , Amorphous gallium oxide-based crosstalk-suppressing solar-blind imaging array prepared by one-step method, J. Phys. Chem. Lett. 15(28), 7272 (2024)

[25] D. Han , Y. Yang , L. Meng , S. Hu , K. Liu , H. Lin , N. Liu , W. Zhang , and J. Ye , High-performance self-powered solar-blind ultraviolet photodetector based on a 4H-SiC/ZnGa₂O₄ heterojunction and its application in optical communication, Appl. Phys. Lett. 123(20), 201104 (2023)

[26] J. Guo , P. Zhang , L. Zhang , L. Tao , Y. Sui , Q. Fu , X. Wang , and B. Song , High-performance multifunctional solar-blind photodetector based on Gd-doped perovskite rare-earth nickelate heterojunctions, Appl. Phys. Lett. 126(26), 262102 (2025)

[27] A. Mock , R. Korlacki , C. Briley , V. Darakchieva , B. Monemar , Y. Kumagai , K. Goto , M. Higashiwaki , and M. Schubert , Band-to-band transitions, selection rules, effective mass, and excitonic contributions in monoclinic β‐Ga₂O₃, Phys. Rev. B 96(24), 245205 (2017)

[28] F. Litimein,D. Rached,R. Khenata,H. Baltache, FPLAPW study of the structural, electronic, and optical properties of Ga₂O₃: Monoclinic and hexagonal phases, J. Alloys Compd. 488(1), 148 (2009)

[29] D. Verma , M. M. R. Adnan , S. Dhara , C. Sturm , S. Rajan , and R. C. Myers , Anisotropic excitonic photocurrent in β‐Ga₂O₃, Phys. Rev. Mater. 7(6), L061601 (2023)

[30] X. Chen , W. Mu , Y. Xu , B. Fu , Z. Jia , F. F. Ren , S. Gu , R. Zhang , Y. Zheng , X. Tao , and J. Ye , Highly narrow-band polarization-sensitive solar-blind photodetectors based on β-Ga₂O₃ single crystals, ACS Appl. Mater. Interfaces 11(7), 7131 (2019)

[31] K. Zhao,J. Yang,P. Wang,Z. Zhou,H. Long,K. Xin,C. Liu,Z. Han,K. Liu,Z. Wei, β‐Ga₂O₃ nanoribbon with ultra-high solar-blind ultraviolet polarization ratio, Adv. Mater. 36(46), 2406559 (2024)

[32] H. Long , T. Xiong , J. Hu , K. He , Y. Liu , J. Yang , and Z. Wei , Solar-blind ultraviolet anisotropic polarization detection by β-Ga₂O₃ based Schottky photodiodes, IEEE Electron Device Lett. 45(7), 1153 (2024)

[33] C. Wu , G. Zhang , J. Jia , H. Hu , F. Wu , S. Wang , and D. Guo , Highly polarization-deep-ultraviolet-sensitive β-Ga₂O₃ epitaxial films by disrupting rotational symmetry and encrypted solar-blind optical communication application, J. Phys. Chem. Lett. 15(14), 3828 (2024)

[34] E. Ohba , T. Kobayashi , T. Taishi , and K. Hoshikawa , Growth of (100), (010) and (001) β-Ga₂O₃ single crystals by vertical Bridgman method, J. Cryst. Growth 556, 125990 (2021)

[35] K. Upadhyaya , V. K. Gudelli , W. Ogieglo , H. Alamoudi , F. Alreshidi , and I. S. Roqan , Identifying orientation-dependent optical properties of single-crystalline β-Ga₂O₃ films, AIP Adv. 14(7), 075204 (2024)

[36] C. Schmidt and D. R. T. Zahn , Effect of impurities on the Raman spectra of spray-coated β-Ga₂O₃ thin films, J. Vac. Sci. Technol. A 40(4), 043404 (2022)

[37] L. Hu , J. Yang , J. Wang , P. Cheng , L. O. Chua , and F. Zhuge , All-optically controlled memristor for optoelectronic neuromorphic computing, Adv. Funct. Mater. 31(4), 2005582 (2021)

[38] S. Jeon , S. E. Ahn , I. Song , C. J. Kim , U. I. Chung , E. Lee , I. Yoo , A. Nathan , S. Lee , K. Ghaffarzadeh , J. Robertson , and K. Kim , Gated three-terminal device architecture to eliminate persistent photoconductivity in oxide semiconductor photosensor arrays, Nat. Mater. 11(4), 301 (2012)

[39] J. Yang , L. Hu , L. Shen , J. Wang , P. Cheng , H. Lu , F. Zhuge , and Z. Ye , Optically driven intelligent computing with ZnO memristor, Fundam. Res. (Beijing) 4(1), 158 (2024)

[40] J. Xue , X. Wang , G. Xu , X. Tao , T. Pan , Z. Chen , Q. Cai , P. Shao , G. Yang , Z. Huang , T. Zhi , K. Wang , B. Liu , D. Chen , R. Zhang , and J. Wang , Multiple exciton generation boosting over 100% quantum efficiency photoelectrochemical photodetection, Nat. Commun. 16(1), 5275 (2025)

RIGHTS & PERMISSIONS

Higher Education Press