Future Trends in Microelectronics: From Nanophotonics to Sensors to Energy

1/f Noise: The Funeral is Cancelled (or Postponed)

M. E. Levinshtein and S. L. Rumyantsev

Ioffe Physical-Technical Institute of Russian Academy of Sciences St. Petersburg, Russia

1.   Introduction

Some years ago, a talk proposing the end of the 1/f noise era was presented at the “Future Trends in Microelectronics: the Nano Millennium” workshop on Ile de Bendor (2001). This talk was based on numerous papers where disappearance of the 1/f noise was predicted in devices “of the next generation”. It was stated that l/f noise would vanish as the size of “future” semiconductor devices decreased.

It looked like this prediction was based on a very firm foundation. Indeed, it is widely believed that 1/f noise in semiconductors (and metals) appears as a superposition of random processes with very wide distribution of characteristic times τ_i. However, if the device is extremely small, there can be only one or two active defects causing these random processes. In this case, the noise in the time domain is of the random telegraph signal (RTS) form. The spectrum of this noise is single Lorentzian (one defect) or superposition of two Lorentzians (two defects).

Figure 1 shows an example of the noise spectra for three small MOSFETs’ with the gate area A = 0.4 μm². Since downscaling the device size is the mainstay of semiconductor technology, it was expected that the 1/f noise would vanish in future, smaller semiconductor devices. The “future” is now the present and, surprisingly, the prediction ...