Search

Convolutional Neural Networks (CNNs) are pivotal in modern digital computing, particularly for tasks like image classification, inspired by the receptive fields of the human brain. Nevertheless, CNNs implemented on conventional digital computers face significant limitations due to inflexible kernels that cannot adjust to dynamic inputs, and the von Neumann architecture, which leads to inefficient data transfer between memory and processing units. This research presents a hardware‐software co‐designed solution, a Dynamic Convolutional Neural Network (dCNN), empowered by three‐terminal adaptive two‐dimensional (2D) memristors. These memristors consist of a vertical heterostructure integrating silver, an atomically thin insulator (CrPS4), and graphene as a semimetal. This configuration allows for the dynamic tuning of conductive filament properties, emulating the heterosynaptic plasticity observed in biological neural systems. The three‐terminal memristor design permits the dCNN to actively adjust kernel weights in its attention layer according to the input stimuli. The empirical tests demonstrate that image classification accuracy using our adaptive 2D memristor‐enhanced dVGG reaches up to 94% on the CIFAR‐10 dataset, which exceeds the performance of static VGG. Furthermore, the energy efficiency of our dVGG significantly outperforms that of GPUs, aligning more closely with the energy dynamics of the human brain in terms of both consumption and classification accuracy. [ABSTRACT FROM AUTHOR]

Photodetectors that detect near-infrared (NIR) light serve as important components in contemporary energy-efficient optoelectronic devices. However, detecting the low-energy photons of the NIR light has long been challenging since the ease of photoexcitation inevitably involves increasing the background current in the dark. Herein, we report the atomic-scale interface modification in SrRuO 3 /LaAlO 3 /Nb-doped SrTiO 3 (SRO/LAO/Nb:STO) heterostructures for NIR photodetection. The interfacial band alignment by a polar monolayer LAO allows precise tuning of the Schottky barrier to achieve a specific energy band profile suitable for the NIR photodetection. The SRO/LAO/Nb:STO heterojunctions show a high photoresponsivity up to ∼1.1 mA/W under NIR light irradiation (λ = 850 nm), while keeping the pA-scale dark current. The increase in the responsivity by interface modification is evaluated at a maximum of 1371%. Based on the enhanced NIR photoresponsivity, as a proof of concept, we demonstrate the spatial imaging of NIR signals using a conceptual array of SRO/LAO/Nb:STO heterojunctions. In addition, the experimental-data-based simulation verifies that the array device can implement pulse-number-dependent plasticity, which is based on the characteristic persistent photoconductivity. This study suggests that atomic-scale interface modification is a facile and powerful method for optimizing the photoresponsive properties of complex-oxide-based heterojunctions.