Jihyun Kim

Neuromorphic optical sensors that merge sensing and in‐pixel preprocessing within a single device are emerging as a key building block for next‐generation artificial vision. Here, we present a WSe ₂ /ZrO _2‐x van der Waals (vdW) heterostructure that realizes a light‐induced field‐tunneling synapse (LIFTS) for visual preprocessing. Exploiting the thin vdW heterostructure with the gate–drain overlap of a single‐gate transistor, two distinct operation modes can be reconfigured by simply adjusting the applied bias condition. In global gating mode, the device operates as a conventional p ‐type transistor with fast photo‐switching, whereas in local gating mode, it exhibits diode‐like rectification mediated by unique field‐tunneling carrier dynamics. Notably, this carrier dynamics enables intensity‐dependent photo‐synaptic operation—remaining inactive under dim light but activated under bright illumination, mimicking human photopic adaptation. At the pixel level, this intensity‐dependent photoresponse is spatially adaptive: dim regions retain a fast positive photoresponse, whereas bright regions activate the field‐tunneling pathway and trigger a negative post‐synaptic response, enabling selective suppression of saturated pixels at the point of detection. Building on this pixel‐level functionality, LIFTS arrays serve as scalable neuromorphic preprocessing layers that progressively recover visual features from glare‐saturated inputs, including realistic street scenes, via edge enhancement and bright‐region attenuation.

Artificial crystals, composed of artificially assembled 2D atomic layers, represent a versatile platform for tailoring material properties beyond the limitations of naturally occurring layered systems. The development of scalable fabrication techniques has accelerated their transition toward practical applications in electronics, optoelectronics, and quantum devices. Beyond assembly strategies, the precise modulation of the van der Waals gap (vdWG) between adjacent layers has been recognized as a critical approach for tailoring the fundamental properties of artificial crystals. vdWG engineering─achieved through methods such as intercalation, mechanical compression, and chemical functionalization─enables systematic tuning of interlayer coupling, band alignment, charge transport, excitonic behavior, and structural phase. These advances provide a versatile foundation for integrating vdWG-engineered artificial crystals into high-performance transistors, photodetectors, superconducting devices, and spintronic platforms. In this perspective, we comprehensively examine fabrication methodologies for artificial crystals and categorize existing vdWG engineering strategies. Then, the discussion expands to the influences of the controlled interlayer spacing toward material properties. Their potential for next-generation technological applications is examined while addressing critical challenges to fully realize the potential of vdWG-engineered artificial crystals in real-world device applications.

Sensor-based computing minimizes latency and energy consumption by processing data at the capture point, thereby eliminating extensive data transfer and enabling real-time decision-making. Here, we present a breakthrough in in-sensor computing via circularly polarized light detectors that integrate cholesteric liquid crystal reflectors with two-dimensional van der Waals p-n heterostructures. Our device exhibits a high dissymmetry factor (1.90), allowing effective separation of mixed circularly polarized images, along with a rapid photoresponse (4 μs) and wide linear dynamic range (up to 114.1 dB), suitable for analog multiply-and-accumulate operations in convolution-based in-sensor computing. Harnessing these detectors, we propose mixed-multimodal in-sensor computing using the chiral state of circularly polarized light to dynamically control responsivity, which enables the blending of two arbitrary image processing modes within a single, non-reconfigurable circuit. By effectively integrating polarization-sensitive detectors into the in-sensor computing framework, the proposed architecture preserves kernel optimization capabilities while simplifying circuit complexity.

MXenes have garnered significant attention due to their atomically thin two-dimensional structure with metallic electronic properties. However, it has not yet been fully achieved to discover semiconducting MXenes to implement them into gate-tunable electronics such as field-effect transistors and phototransistors. Here, a semiconducting Ti₄N₃T_x MXene synthesized by using a modified oxygen-assisted molten salt etching method under ambient conditions, is reported. The oxygen-rich synthesis environment significantly enhances the etching reaction rate and selectivity of Al from a Ti₄AlN₃ MAX phase, resulting in well-delaminated and highly crystalline Ti₄N₃T_x MXene with minimal defects and high content of F and O, which led to its improved hydrophobicity and thermal stability. Notably, the synthesized Ti₄N₃T_x MXene exhibited p-type semiconducting characteristics, including gate-tunable electrical conductivity, with a current on-off ratio of 5 × 10³ and a hole mobility of ∼0.008 cm² V^-1 s^-1 at 243 K. The semiconducting property crucial for thin-film transistor applications is evidently associated with the surface terminations and the partial substitution of oxygen in the nitrogen lattice, as corroborated by density functional theory (DFT) calculations. Furthermore, the synthesized Ti₄N₃T_x exhibits strong light absorption characteristics and photocurrent generation. These findings highlight the delaminated Ti₄N₃T_x as an emerging two-dimensional semiconducting material for potential electronic and optoelectronic applications.

The significance of metal-semiconductor interfaces and their impact on electronic device performance have gained increasing attention, with a particular focus on investigating the contact metal. However, another avenue of exploration involves substituting the contact metal at the metal-semiconductor interface of field-effect transistors with semiconducting layers to introduce additional functionalities to the devices. Here, a scalable approach for fabricating metal-oxide-semiconductor (channel)-semiconductor (interfacial layer) field-effect transistors is proposed by utilizing solution-processed semiconductors, specifically semiconducting single-walled carbon nanotubes and molybdenum disulfide, as the channel and interfacial semiconducting layers, respectively. The work function of the interfacial MoS₂ is modulated by controlling the sulfur vacancy concentration through chemical treatment, which results in distinctive energy band alignments within a single device configuration. The resulting band alignments lead to multiple functionalities, including multivalued transistor characteristics and multibit nonvolatile memory (NVM) behavior. Moreover, leveraging the stable NVM properties, we demonstrate artificial synaptic devices with 88.9% accuracy of MNIST image recognition.

2D transition metal dichalcogenides (TMDCs) have shown promise as a class of optoelectronic material for flexible photosynaptic devices. However, widespread use of 2D TMDCs has been limited because producing high-quality, large-scale films on flexible substrates has still been challenging. Herein, we propose a facile solution processing strategy to realize large-scale arrays of flexible photosynaptic devices based on 2D TMDCs. Bulk molybdenum disulfide (MoS ₂ ) crystals were first electrochemically exfoliated into few-layer nanosheets using molecular intercalants, followed by dispersion in isopropyl alcohol for spin coating on flexible substrates. Large lateral size of the nanosheets along with good wettability of the dispersion enabled the formation of continuous MoS ₂ nanosheet networks for scalable device fabrication. Sulfur vacancies within the MoS ₂ channel served as trapping sites for photogenerated carriers and supported photocurrent retention characteristics needed to mimic synaptic functions. The device array successfully demonstrated reinforcement learning capabilities for image recognition under repeated optical pulse signals. Finally, the optoelectronic performance was fairly stable under mechanical bending because devices did not crack or delaminate from the substrate.