Exploring the Floating Body Effect in Bulk nFET Devices for Enhanced Physical Reservoir Computing

Abstract

The floating body effect in partially depleted silicon-on-insulator (SOI) technology significantly influences drain current and threshold voltage, leading to undershoot and overshoot transients. This effect is also present in standard bulk nFET devices when the body is intentionally left floating. In this work, we establish the potential of the floating body effect for physical reservoir computing, attributed to its inherent nonlinearity and recovery characteristics. By performing gait authentication experiments using foundry-fabricated 28-nm nFET devices with floating body effect, we achieve a 2.9% EER for ten subjects without the need for postprocessing. Moreover, a shorter gate length will provide a shorter time constant in the floating body effect, thereby enhancing the flexibility of reservoir computing. This flexibility is crucial because different applications require specific reservoir time constants to achieve optimal performance. By adjusting the gate length, one can effectively tune these time constants to meet the demands of various applications. Last but not least, depending on whether the body is grounded or left floating, negative bias temperature instabilities (NBTIs) and the floating body effect can coexist in standard bulk devices, and both can be used for physical reservoir computing. We provide criteria for choosing between these two effects based on user cases.