.%/id-.%Ur E/echonic8 Vol. 24, No. 8. pp. 717-718. I981 Printed in Great Britain. 0038-1l0l/8l/aW717-ot#n.OOM Pergmon Press Ltd. l/f NOISE IN GaAs MESFETS C. H. SUH, A. VAN DER ZIEL and R. P. JINDAL Electrical Engineering Department, University of Minnesota, Minneapolis, MN 55455,U.S.A. (Received 27 September 1980;in reoised form 3 January 1981) Abstract-An attempt was made to discriminate between number fluctuation and mobility fluctuation I/f noise in GaAs MESFETs. It was found, that both models could explain the data, even though the mobility fluctuation I/f model seems more likely. We report here on some I/f noise measurements on NEC GaAs MESFETs, made in an attempt to discriminate between the number fluctuation model[l] and the mobility fluctuation model[2] of l/f noise. In a previous paper[3] we showed that these devices had a spectrum of the form (AU) [l - (%T) tan-‘(oT,)l (1) where A and 70 were constants that depend somewhat on the gate bias condition. We interpreted these results in terms of the following distribution in time constants T dT/r -for To<T<T ‘(T) dr = ln(T1/TO) = o otherwise. versus In f at both drain voltages; here IO = S,,(f)/2q) is the equivalent saturated diode current of the noise and q is the electron charge. Since To does not depend on V,+, the dependence of I., upon Vd must come from the dependence of the parameter A upon Vd. Figure 2 shows In I,, versus In Vd at V, = -1.50 V and f - 1 kHz. At low drain bias the value of I., varies as V:, as expected theoretically (see below). At higher values of V,J the value of I., passes through a maximum at Vd = 0.65 V, a minimum at 1.5OV and levels off to a constant value above 2.0 V. According to Park et al. [8], the drain noise spectrum for the number fluctuation model is (2) Such a distribution is usually invoked to explain the occurrence of a l/f spectrum[4-71; it gives a spectrum of the form (1) for o > l/71 and a l/f spectrum for l/71 < 0 < l/T,. We measured the drain noise spectrum S,(f) at V, = - 1.50V at very low drain bias ( Vd = 0.10 V) and at saturation (V, = l.SOV) and found that the constant TV depended hardly on V,,, whereas the parameter A did. This is shown in Fig. 1, which gives a plot of InI,, I%(: x)l.ff dx (3) whereas for the mobility fluctuation model S,,(f) =qh =LJcdxMEo, xl dx. fL I0 (4) Here w is the device width, L the device length, q the lo4 Go As mset NEC24483 lot. 99-3C47) IO3 4 IO’ $ IO Fig. 1. fCq= Sr,,f_f)/Zqas vd=o.lv; V,=-1.5V, a function of frequency for V, = OV, V,,=t,.tV and V,=-1.5V, vd= 1.5v. Pia. 2. I., as a function of Vdat V, = -1.5 V and at V, = -0.5 V atf=lkHz. 717 C. H. SrIH et al. 718 electron charge, f the frequency, E,(x) the d.c. field strength at x, U&X) the drift velocity at x and Id the drain current at the drain voltage Vd. Furthermore [&(E~,x)]~R is the effective trap density at the Fermi level at x, E a tunneling parameter and a(E,,x) the Hooge parameter at x at the field strength Eo(x). The characteristic 4( V,,) increased monotonically with increasing V,, reaching its maximum at saturation, as expected. Since for low drain bias Vd the drain current Id and the drift velocity Ud(x) at x are both proportional to V,, we see that Ies at low Vd should be proportional to V:, in agreement with our data. Now in GaAs the drift velocity Ud(x), considered as a function of EJx) first varies linearly with increasing E,, passes through a maximum at a critical field strength, decreases with further increase of E,, passes through a minimum and finally increases again at very large E,; this phenomenon is caused by inter valley transfer of carriers. In elemental semiconductor devices such as silicon MOSFETs, [,&(E,&R decreases with increasing V, - V, (Fig. 1) and hence [&(E,, X&R increases with increasing Vd at a given x and with increasing x at a given Vd. Putting it all together for GaAs MESFETs, we see that eq. (3) can qualitatively explain the curve shown in Fig. 2. That is, the number fluctuation model can explain the data. In elemental semiconductors like Si, a(E,, x) decreases with increasing E,. But in GaAs one would expect first a decrease, then an increase, followed by a possible second decrease due to inter valley transfer effects. That is, the mobility fluctuation model can also qualitatively explain our data. Hence our measurements do not permit to discriminate between the two models. To make such a discrimination possible, much more should be known about [NT(E,,x)].s and a(E,,x) than we can surmise at present. The mobility fluctuation model seems to be more likely, however. Perhaps this information could have been obtained by carrying out many more measurements on the GaAs MESFETs under study, but this is not certain. Time did not permit us to carry out such a program since one of us (C. H. Suh) had to return to Korea. We therefore decided to publish the available data. The above qualitative discussion can be made more quantitative by writing for the mobility CL. cL=PcLt+U-Pb2 (5) Here p1 and ~2 are the mobilities in the upper and lower valley, respectively, and /3(E,) is the relative occupancy of the lower valley. Hence from which the dependence of ud upon E, follows. In the same way, for fluctuations Sp, and Spz in IL, and w2, respectively, ijp = p&L1 + (1 -Pm from which follows But according to Bosman et al. [9] S,,W/d= 40, xYIPfN[l +(PIoEIus)~I~ = /.LE”= /$.L&, + (1 - P)/&> (6) (9) where cr(0,x) is the low-field value of Hooge’s parameter (Y,plo and pzo the low-field values of the mobilities :L~ and ~2, respectively, and pLs the velocity of sound. Hence a(Eo, x) = 40, xl ( +P(PIIl.# (1- P)(l.L*Il# 1 (/~,oE/ud~+ I+ bzoE/us)’ >’ (10) This explains the qualitative discussion of a(E,, x). Since in eqn (6) the carriers can obtain a limiting velocity of a few times 10’ cm/set., and us is only of the order of 106cm/sec., it is most likely that the mobility fluctuation effect will predominate over the number fluctuation effect. The mobility fluctuation model may thus give the more likely explanation of Fig. 2. Acknowledgemenb-This work was performedunder an Army Research Office grant. Prof. C. H. Suh was on leave of absence from Hong-Ik University at Seoul, Korea, and was supported by a Korean government scholarship. REFERENCES 1. A. van der Ziel, Advances in Electronics and Electron Physics 2, 3. 4. 5. ;: 8 9. ud (7) (Edited by L. Martin) pp. 225-297.Academic Press, New York (1979). F. N. Hooge, Phys. Len. A29, 139 (1%9); Physica (Utrecht), 83B, 14 (1976). C.H. Suh and A. van der Ziel, Appl. Phys. Lett. 37,565 (1980). M. J. Surdin, J. Phys. Radium IO, 188(1939); 12,778 (1951). A. van der Ziel, Physica (Utrecht) 16. 359 (1950). F. K. du Pre. Phvs. Rev. 78. 615 (1950). A. L. McWhortei, l/f Noise and Related Surface .E,fects in Germanium. MIT Lincoln Lab Report No. 80, May 1955;also in Semiconduclor Surface Physics. University of Pennsylvania Press (1957). H. S. Park, A. van der Ziel, R. J. J. Zijlstra and S. T. Liu, J. Applied Physics. To be published. I. Bosman, R. J. J. Zijlstra and A. van Rheenen, To be published.
© Copyright 2026 Paperzz