Search for: All records

Abstract Memristive crossbar architectures are promising as efficient, low-power inference engines for edge AI applications. However, inputs with minor differences often yield similar outputs, requiring additional processing methods such as confidence scoring, feedback mechanisms, crossbar redundancy, or hybrid analog-digital approaches to resolve. These methods can be impractical for resource-limited edge devices. In contrast, three-terminal memtransistors can dynamically tune conductance via gate control, effectively resolving similar outputs and enhancing separability without retraining. Here, we present dense, large-scale crossbar array architectures incorporating up to 2048 MoS₂memtransistors per array, achieving >92% yield across multiple arrays while individual memtransistors exhibit write energies as low as ~0.2 fJ, maintain read margins up to 10⁵, and offer a projected retention exceeding three years. These architectures demonstrate the ability to resolve inference ambiguities through gate modulation without the need for costly retraining or reprogramming. We also validate their performance by successfully classifying handwritten digits from the MNIST database. Finally, we benchmark the performance of MoS₂memtransistors against other 2D material-based architectures and project their potential compared to state-of-the-art AI accelerators. We believe that this work furthers the ongoing development of in-memory processors for decentralized edge applications and that future studies aimed at reducing device-to-device variation and improving long-term non-volatile memory would only enhance inference capabilities. 

Gallium nitride (GaN)-based high electron mobility transistors (HEMTs) are essential components in modern radio frequency power amplifiers. In order to improve both the device electrical and thermal performance (e.g., higher current density operation and better heat dissipation), researchers are introducing AlN into the GaN HEMT structure. The knowledge of thermal properties of the constituent layers, substrates, and interfaces is crucial for designing and optimizing GaN HEMTs that incorporate AlN into the device structure as the barrier layer, buffer layer, and/or the substrate material. This study employs a multi-frequency/spot-size time-domain thermoreflectance approach to measure the anisotropic thermal conductivity of (i) AlN and GaN epitaxial films, (ii) AlN and SiC substrates, and (iii) the thermal boundary conductance for GaN/AlN, AlN/SiC, and GaN/SiC interfaces (as a function of temperature) by characterizing GaN-on-SiC, GaN-on-AlN, and AlN-on-SiC epitaxial wafers. The thermal conductivity of both AlN and GaN films exhibits an anisotropy ratio of ∼1.3, where the in-plane thermal conductivity of a ∼1.35 μm thick high quality GaN layer (∼223 W m−1 K−1) is comparable to that of bulk GaN. A ∼1 μm thick AlN film grown by metalorganic chemical vapor deposition possesses a higher thermal conductivity than a thicker (∼1.4 μm) GaN film. The thermal boundary conductance values for a GaN/AlN interface (∼490 MW m-2 K−1) and AlN/SiC interface (∼470 MW m−2 K−1) are found to be higher than that of a GaN/SiC interface (∼305 MW m−2 K−1). This work provides thermophysical property data that are essential for optimizing the thermal design of AlN-incorporated GaN HEMT devices.