Search

Abstract Advancements in AI led to the emergence of in-memory-computing architectures as a promising solution for the associated computing and memory challenges. This study introduces a novel in-memory-computing (IMC) crossbar macro utilizing a multi-level ferroelectric field-effect transistor (FeFET) cell for multi-bit multiply and accumulate (MAC) operations. The proposed 1FeFET-1R cell design stores multi-bit information while minimizing device variability effects on accuracy. Experimental validation was performed using 28 nm HKMG technology-based FeFET devices. Unlike traditional resistive memory-based analog computing, our approach leverages the electrical characteristics of stored data within the memory cell to derive MAC operation results encoded in activation time and accumulated current. Remarkably, our design achieves 96.6% accuracy for handwriting recognition and 91.5% accuracy for image classification without extra training. Furthermore, it demonstrates exceptional performance, achieving 885.4 TOPS/W–nearly double that of existing designs. This study represents the first successful implementation of an in-memory macro using a multi-state FeFET cell for complete MAC operations, preserving crossbar density without additional structural overhead.

Abstract Fully-printed temperature sensor arrays—based on a flexible substrate and featuring a high spatial-temperature resolution—are immensely advantageous across a host of disciplines. These range from healthcare, quality and environmental monitoring to emerging technologies, such as artificial skins in soft robotics. Other noteworthy applications extend to the fields of power electronics and microelectronics, particularly thermal management for multi-core processor chips. However, the scope of temperature sensors is currently hindered by costly and complex manufacturing processes. Meanwhile, printed versions are rife with challenges pertaining to array size and sensor density. In this paper, we present a passive matrix sensor design consisting of two separate silver electrodes that sandwich one layer of sensing material, composed of poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS). This results in appreciably high sensor densities of 100 sensor pixels per cm $$^2$$ 2 for spatial-temperature readings, while a small array size is maintained. Thus, a major impediment to the expansive application of these sensors is efficiently resolved. To realize fast and accurate interpretation of the sensor data, a neural network (NN) is trained and employed for temperature predictions. This successfully accounts for potential crosstalk between adjacent sensors. The spatial-temperature resolution is investigated with a specially-printed silver micro-heater structure. Ultimately, a fairly high spatial temperature prediction accuracy of 1.22 °C is attained.