Elsevier

Microelectronics Reliability

Volume 54, Issues 9–10, September–October 2014, Pages 1749-1752
Microelectronics Reliability

RTN distribution comparison for bulk, FDSOI and FinFETs devices

https://doi.org/10.1016/j.microrel.2014.07.013Get rights and content

Abstract

In this paper we investigate the sensitivity of RTN noise spectra to statistical variability alone and in combination with variability in the traps properties, such as trap level and trap activation energy. By means of 3D statistical simulation, we demonstrate the latter to be mostly responsible for noise density spectra dispersion, due to its large impact on the RTN characteristic time. As a result FinFETs devices are shown to be slightly more sensitive to RTN than FDSOI devices. In comparison bulk MOSFETs are strongly disadvantaged by the statistical variability associated with high channel doping.

Introduction

In the context of downscaling, Statistical Variability (SV) had become a major concern in respect of device matching and its impact on circuits performances [1], [2]. Indeed discreteness of charge and matter and intrinsic lithography limitations lead to a significant dispersion of device performance and therefore to a strong need of proper approaches to predict the dispersion in devices parameters [1], [3]. This has led to the introduction of device architectures, such as Fully Depleted Silicon on Insulator (FDSOI) and Fin Field Effect Transistor (FET) devices, which tolerate low channel doping and therefore reduce the SV. Nevertheless the devices scaling has changed also the expression of reliability problems; discrete charge trapping and detrapping as well as traps interaction with SV has transformed oxide reliability into a time dependent variability [4]. However traps properties, determining traps characteristic times, are also widely distributed [5]. This raises the legitimate question whether device to device SV or traps properties dispersion dominates the dispersion of time dependent reliability impact.

In this paper we focus on Random Telegraph Noise (RTN) dispersion for bulk, FDSOI and FinFET MOSFETs. After studying the electrostatic impact of a single charged trap, we follow the methodology presented in [6], to analyse RTN noise density spectra, initially without neither SV nor traps properties dispersion and then with both of them separately and in combination.

Section snippets

Methodology

We use a sample of 100 devices for each architecture designed to meet the requirements of the 20/16/14 nm CMOS technology. The planar devices are 22 nm wide; FDSOI features a 6 nm deep Silicon layer on a 10 nm buried oxide box, whereas the FinFET buried oxide is 20 nm deep and its aspect ratio is 25 nm high for 10 nm wide. Every architectures Equivalent Oxide thickness is 0.9 nm. Those SV suffering devices are illustrated in Fig. 1. In this work we account for Random Dopants Fluctuations (RDF), Metal

Results

Fig. 2 presents single trap impact on current at threshold voltage as a function of its position along the channel. Whereas in the uniform devices case they clearly follow the source to drain potential barrier, in presence of SV these impacts are widely dispersed for bulk transistors suffering from their source to drain percolative behaviour. On the contrary dopant free channel devices robustness toward SV is demonstrated. Multi-traps impact distributions are exhaustively analysed in [7]; note

Conclusion

In this paper we have investigated different MOSFETs architectures robustness towards RTN, considering not only traps interactions with SV but also traps properties dispersion impact on noise spectrum density distributions. Even if it is of paramount importance to take SV into account to evaluate trap impact on current, we demonstrated in this work that SV is not the main source of RTN spectral power density dispersion. Indeed the characteristic RTN times dispersion is mainly related to the

Acknowledgment

The authors would like to thank Amanda Smith for her constant help in arranging conferences accommodations.

References (9)

There are more references available in the full text version of this article.

Cited by (9)

  • Tsallis non-extensive statistics and multifractal analysis of the dynamics of a fully-depleted MOSFET nano-device

    2019, Physica A: Statistical Mechanics and its Applications
    Citation Excerpt :

    It is apparent, that noise analysis in such devices is an important issue that must be studied for the proper characterization of the device, in terms of its reliability when operating in various circuit designs. To this direction analysis of the low-frequency noise (LFN) in MOSFET devices appears in [4–8], where a study in the frequency domain and the co-existence of flicker and Lorentzian-like noise was revealed. Averaging of thousands of measurements, as implied in the previously mentioned studies, could hide or diminish the device’s real dynamics or any critical phenomena in noise.

  • Intermittency-induced criticality in the random telegraph noise of nanoscale UTBB FD-SOI MOSFETs

    2019, Microelectronic Engineering
    Citation Excerpt :

    It is apparent that averaging of thousands of measurements (implied in those studies) could hide or diminish dynamic or critical phenomena in noise. Consequently, the necessity to analyze the original signals in the time domain emerges [3–8]. To this direction, the deterministic chaotic nature of the complex random telegraph noise in UTBB FD-SOI MOSFET devices was revealed and evaluated in [9].

View all citing articles on Scopus
View full text