PhD Defense of TATARIDOU Angeliki

Despite the advancements in the semiconductor device research and development, with the incorporation of new materials and architectures, as well as the downsizing of the geometrical dimensions, which lead to superior device performance and speed, the Low Frequency Noise, LFN, has become a major concern for micro- and nanoscale transistor components, as its impact on both device and circuit level is more important than ever. First of all, it should be noted that, in this dissertation, when referring to LFN we mean the internal type of noise due to trapping/detrapping or/and scattering of free carriers. For 1/f-like spectra, the power spectral density of the flat-band voltage, SVfb, is inversely proportional to the area, and thus going from micro- to nanoscale devices the LFN level is increased. In addition, with the miniaturization of the transistor area, a different type of noise called random telegraph noise, RTN, appears and becomes the main contribution instead of 1/f, as individual trap behavior becomes visible. On top of that, the introduction of new materials and architectures in the CMOS technology results to the appearance of peculiar noise behavior like the generation-recombination, GR, noise, which show a Lorentzian PSD instead of 1/f-like. As a result, next-generation electronic components will be governed by instabilities arising from their intrinsic noise sources. It is therefore essential to change the methods of characterization and simulation of LFN/RTN to allow the technology improvement. This is why a thorough theoretical and experimental study of all noise sources in emerging components becomes indispensable in this field of research in microelectronics.
In this dissertation, various devices which meet the ITRS specifications under the demand of “More Moore” and “More than Moore” technology roadmaps (i.e. FinFETs, TriGate NW FETs, CoolCube 3DSI FETs), have been characterized in terms of LFN. Through this study, the physical phenomena that induce the intrinsic device noise have been identified, an information useful not only for the device itself, but also for the accurate noise modeling and therefore for the design facilitation of the associated circuits. On top of that, LFN measurements were utilized as a diagnostic tool for the identification of defective zones giving information on the quality of the fabricated transistors. The latter is essential for the optimization of fabrication steps. In addition, although the theory of LFN, as well as its corresponding models, are well-established and successfully used for years, issues existing in aggressively scaled down devices might on one hand, hinder the reliable extraction of noise parameters and on the other hand, they can be utilized for traps’ effects decoupling and identification. Consequently, a revised version of the “carrier number with correlated mobility fluctuations” (CNF/CMF) model has been proposed so as the impact of series resistance on noise parameters to be eliminated, and a new measurement methodology has been introduced, utilizing an inhomogeneous carrier distribution inside the channel that allows for the maximum RTN-inducing trap detection. Finally, since the appearance of Lorentzian spectra is more and more frequent in the advanced FET technologies, LFN frequency domain models need to be revised. This is because the existing frequency domain models are limited to the typical 1/f behavior which, as we demonstrate in this thesis, can compromise the nominal operation of circuits. Hence, we present a method for the implementation of Lorentzian noise spectra through Verilog-A. Once this method is validated, some circuit noise application examples are examined, in order to showcase how non-1/f noise can affect a circuit’s performance.

• Gérard GHIBAUDO, RESEARCH DIRECTOR, CNRS DELEGATION ALPS : Supervisor
• Panagiota MORFOULI, PROFESSOR of UNIVERSITIES, GRENOBLE INP : Examiner
• Matthias BUCHER, ASSOCIATE PROFESSOR, TECHNICAL UNIVERSITY of CRETE : Examiner
• Bogdan CRETU, ASSOCIATE PROFESSOR  HDR , NORMANDIE UNIVERSITY : Reviewer
• Fabien PASCAL, PROFESSEUR of UNIVERSITIES, UNIVERSITY of  MONTPELLIER : Reviewer