Investigating Self-Heating Effects in Ferroelectric FinFETs for Reliable In-Memory Computing

Ferroelectric (Fe) FET has emerged as a promising candidate for efficient in-memory computing due to its properties, such as non-volatility and low power. However, scaled 3D devices such as Fe-FinFET suffer from significant self-heating effects (SHE) and process variations. These issues cause inconsistent performance and reduce reliability, limiting their applicability in high-performance applications like ternary content addressable memory (TCAM) and Hyperdimensional computing (HDC). In this paper, we explore the impact of SHE on 14 nm Fe-FinFETs using a cross-layer framework, analyzing how these effects and associated variations affect both circuit-level (TCAM cells) and system-level (HDC) performance. Our results reveal an increased error probability in Hamming distance (HD) calculations through the TCAM array when SHE and variations are present. Additionally, we demonstrate how SHE and variations influence the inference accuracy of the HDC framework.