Pergaman Solid State Communications, Vol. 98, No. 11, pp. 961-963, 1996 Copyright 0 1996 Published by Elsevier Science Ltd Printed in Great Britain. All rights reserved 0038-1098/96 $12.00 -I- .OO zyxwvutsr PII SOO38-1098(96)00186-X On the relationship between the bulk recombination lifetime and the excess l/f zyxwvutsrqponmlkjihgfedcbaZYX noise in silicon p-n junction diodes E. Simoen, J. Vanhellemont and zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPO C. Claeys IMEC, Kapeldreef 75, B-3001 Leuven, Belgium (Accepted 6 M arch 1996 by D. van Dyck) Abstract In this paper, the relationship between the bulk recombination lifetime and the excess l/f noise in Si n+p junction diodes is investigated. Two types of behaviour can be distinguished: for low-lifetime substrates (z c 10 ps), the low-frequency noise spectral density is found experimentally to obey a power law of the kind SI - $18t-O.O5. For high lifetime diodes on the other hand, not such a relationship is observed, indicating a different origin of the underlying noise sources. These findings will be compared and discussed with previously proposed theoretical and empirical relationships. Copyright 0 1996 Published by Elsevier Science Ltd 1. INTRODUCTION Excess flicker noise showing a l/f-like power spectrum at low frequencies has been studied extensively in Si p-n junction diodesle4. Depending on the source of the underlying current fluctuations, a so-called “surface origin” l-3, or a mobility- related bulk origin4 has been invoked. In the latter case, for long, ideal junction diodes, a ‘mobility-fluctuations’ based model has been derived, resulting in the following expression for the current noise spectral density4: with f the frequency, z the bulk recombination lifetime and IF the forward current. In eq. (1) a is the dimensionless empirical Hooge parameter which takes up values typically in the range 10-5-10-3. On the other hand, it has been proposed on many occasions that there exists a direct correlation between the bulk recombination lifetime and the excess l/f noise in p-n junction diodes and bipolar transistors5-7. In fact, Lukyanchikova7 proposes a linear relationship between the Hooge parameter and the recombination lifetime z : a=pz (2) with @300 the best empirical fit to the available literature data. Other reports find no relationship at all between SI and 78. Therefore, the aim of this paper is to investigate the nature of this relationship, by studying the low-frequency noise in a large set of nip junction diodes, fabricated on different Si substrates with different carrier recombination properties. It is shown that there exists a correlation between the two parameters at stake, for diodes having sufficiently low T. For substrates with a high bulk recombination lifetime on the other hand, the excess l/f noise is dominated by another source, most likely associated with the diode periphery. 2. EXPERIMENTAL Junction diodes have been fabricated on different Si substrates and with different pretreatments as explained previouslyg. The recombination lifetime of these diodes and substrates has been extensively studied with a variety of techniques, showing a span of more than 4 orders of magnitude in 2 between the best (Float Zone : FZ) and the “worst” (high oxygen Czochralski : HO Cz) substrateslo. Initial LF noise results have also been reported recently1 1, showing a clear impact of the substrate quality on the noise magnitude. Here, the LF noise is measured in forward operation on a large set of square 900 pmx900 j.tm diodes, for IF=1 PA and 10 PA. The noise spectral density at 1 Hz is obtained from a least-squares fit to the experimental points, after subtraction of the shot noise (=2qIF) and the back-ground system noise. The noise spectra are generally MY-like with a frequency index y varying between 0.7 and 1.3 12. The recombination lifetime for the diodes is derived from forward current measurements, using a recently proposed practical analysis of the forward current intercept valuelo, with corrections for the diode non- ideality and the perimeter contribution. For not too extreme values, a good match, within the expected accuracy is observed between these Z’S and the values derived from more sophisticated measurement techniques. 3. RESULTS AND DISCUSSION In Fig. 1, SI at 1 and 10 PA and a frequency f=l Hz is 961