visiting imec from LJMUOf the MOSFET degradation mechanisms, the variability of Time-Dependent Dielectric Breakdown (TDDB) and Bias Temperature Instability (BTI) in deeply scaled devices has been described with a degree of success [1,2]. In comparison to these cases, Channel Hot Carrier (CHC) degradation is inherently more complex [3,4]. CHC stress is non-uniform, with large currents present, and the resulting degradation is typically localized at the drain, thus offering a wider range of variability sources, as well as potential measurement artifacts. Variability of CHC degradation has been reported previously, mainly in planar devices [5][6][7][8][9]. Here we reexamine this topic in two generations of nFinFET devices and observe that CHC variability is generally higher than that of BTI, flagging it as the main contributor to time-dependent variability of FinFETs [10]. We examine in detail several intrinsic (random) and extrinsic (process-induced, systematic) sources of this increased variability. To aid this procedure, we demonstrate that matching pairs [7] can be used to eliminate extrinsic (process-related) time-dependent variability sources, analogously to time-0 variation [11]. We discuss the intrinsic variability sources in the defect centric framework [2,12] and provide a blueprint for projecting CHC variability to operating conditions and lifetimes. A population of nFinFET devices with (as drawn) gate lengths L G ranging from 130 nm down to 28 nm, fin height ~30 nm, highk gates, fabricated in two imec technologies, distributed across wafer, have been subjected to either CHC or PBTI stress at either RT or 125 o C. An example of the increased spread of CHC degradation wrt BTI for the same mean threshold voltage shift ΔV th is in Fig. 1. In the following we quantify the variability in terms of average charged trap impact η = σ ΔVth 2 /(2ΔV th ) (eq. 1), a crucial defect-centric parameter [2,13]. We first identify several potential sources of this increased CHC variability (Fig. 2). Since the CHC energy distribution is an intricate function of the electric field distribution in the device body, a dependence on device length (Fig. 2a) and biases (Fig. 2b) variations can be expected. From CHC ΔV th vs L g dependence (not shown) we can readily extract the effect's sensitivity to channel L g variation to be (technology dependent) ~2.5 mV/nm. With this in mind and assuming σ Leff = 2 nm, η = 0.13 mV (cf. Eq. 1), suggesting this mechanism is not the main contributor to CHC variability. Fig. 3 illustrates the sensitivity of CHC energy distribution at the drain to the drain bias V D . Again from measurements of CHC ΔV th vs. V D (not shown) we can conclude that η is negligible 0.07 mV for σ VD = 10 mV, however, η is considerable 2.2 mV for σ VD = 50 mV. Setting aside circuit implications of varying V D for the moment, we conclude that variations in source/drain series resistance (Fig. 2b) could be responsible for the increased CHC variability. We, however, observe no correlation between channel current d...