An In-Depth Study on Device Variability in 22 nm FD-SOI NMOSFETs Under X-Ray Exposure

The effect of total-ionizing-dose (TID) on device-to-device variability in 22 nm FD-SOI NMOS transistors was experimentally quantified. Across ten <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">W/L</i> geometries, two device types (RVT, LVT), and three irradiation-bias conditions (Off-Stress, Work-Mode, Power-Off), ensembles of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">N</i> = 80 devices per geometry, 4800 transistors in total were irradiated to 300 krad(SiO<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub>) and subsequently characterized. Pelgrom plots of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V_th</i> and the current-factor β were analyzed. TID consistently shifted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">V</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">_th</sub> negative and broadened its distribution while preserving the 1/ √ <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">WL</i> law; the extracted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A_V_th</i> increased by about 9–47% depending on bias/type, with a robust ordering <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Off-Stress</i> > <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Work-Mode</i> > <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Power-Off</i>. In contrast, β exhibited near-zero mean drift and only modest variance growth (typically ≤20%), indicating a much weaker mobility-driven response. Size-segregated fits confirmed an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">area-dominated</i> mechanism: small-area devices exhibited the largest variance increase, whereas large-area devices approached the measurement floor. A compact dose-aware Pelgrom model was established, adding a TID-dependent term to <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A_V_th</i> that links the Poisson-distributed BOX trapped charge and the bias-dependent electrostatic coupling, and captures the observed bias ordering and size scaling.