Low Frequency Noise in Small MOS Devices
- PDF / 856,952 Bytes
- 12 Pages / 420.48 x 639 pts Page_size
- 42 Downloads / 209 Views
LOW FREQUENCY NOISE IN SMALL MOS DEVICES
MICHAEL J. UREN Defence Research Agency, RSRE Malvern, St. Andrew's Road, Malvern, Worcestershire WR14 3PS, United Kingdom
Great
ABSTRACT Random telegraph signals can be seen in the conductivity of submicron MOSFETs due to capture and emission from individual Si0 2 defects. They show activated capture times, temperature dependent energy levels as well as a complex scattering behaviour. It is shown how 1/f noise in large MOSFETs arises naturally from the properties of these defects and how models based on this understanding can be constructed. 1/f noise is shown to be a poor tool for the study of the traps due to the loss of information associated with averaging over many active signals. INTRODUCTION 1/f noise in MOS devices has been the subject of intensive study for over 25 years and is still somewhat controversial. There have been two main models propounded to explain the fact that a 1/f noise spectrum has been observed over at least 10 decades in frequency. The first was due to McWhorter[l], who suggested that the noise is due to free electrons in the inversion layer tunnelling to and from defects located within the gate oxide of the device. The huge range in time constants results from the exponential dependence of tunnelling time on distance into the oxide, with the change in conductivity arising from a fluctuation in the number density of electrons. Hooge proposed that the noise is the result of a mobility fluctuation rather than a number fluctuation[2]. This model is essentially empirical, but in the case of MOSFETs he suggested that the noise is associated with bulk phonons. In 1984, a new approach to the rather sterile debate that had developed, became possible through the work of Ralls et al[3] on deep submicron MOSFETs. They showed that in such small samples it is perfectly possible to enter the mesoscopic regime where there are only one or two active defects present in the device. These are observed as a discrete two-level fluctuation (known as a random telegraph signal, RTS) in the conductivity of the device as a single electron is captured or emitted from a trap. This paper will give a description of the electrical properties of the traps responsible for the RTSs, review the evidence that they are responsible for the 1/f noise, and finally show how models of 1/f noise can be constructed that build on this microscopic understanding. Reviews of 1/f noise[4,5] and RTSs[6,7] are available that cover aspects of the material in more detail. INDIVIDUAL DEFECTS A schematic cross-section and band diagram of a n-channel MOSFET (Metal-Oxide-Silicon-Field-Effect-Transistor) is shown in figure 1, illustrating the principle features of the device. The degenerately doped poly-silicon gate electrode is isolated from the Mat. Res. Soc. Symp. Proc. Vol. 262. 01992 Materials Research Society
752
(a)
(b)
Metal
.................
Source Gate Si02
p Silicon c. b.
............... E
Drain
nn+ET-
v.b.
p silicon
1
E
Figure 1. (a) Schematic cross-section of a n-channel
Data Loading...
