Thin Solid Films, 136 (1986) 167-172

ELECTRONICS AND OPTICS 167

I N F L U E N C E O F P R E P A R A T I O N PARAMETERS ON THE T U N N E L L I N G CHARACTERISTICS OF P b - B i - I n / O X I D E / P b - B i - I n J U N C T I O N S

K. OKUYAMA* AND K. H. GUNDLACH

lnstitut de Radio Astronomie Millimetrique, Domaine Universitaire de Grenoble, 38406 St. Martin d 'Heres (France)

(Received June 18, 1985; accepted September 18, 1985)

The quasiparticle millimetre wave mixer requires high quality tunnel junctions with a large gap voltage, a normal resistance R, of several tens of ohms and areas A of only a few square micrometres or less. We show that water vapour at the upper surface of the P b - B i - I n alloy junction can increase the product R,A by more than two orders of magnitude. However, indium at the upper surface can drastically decrease RnA but does so mostly at the expense of the quality of the junction. Appropriate annealing allows the normal resistance to be increased or decreased to close to the desired value.

I. INTRODUCTION

The superconductor/insulator/superconductor (SIS) quasiparticle mixer has been rapidly developed over the last 3 years 1. So far nearly all SIS mixers use lead alloy junctions: a Pb-In(Au) base electrode and a lead, Pb- In , Pb -Au or Pb-Bi counterelectrode. They are relatively easy to fabricate and give a sharp onset of the quasiparticle current at the gap voltage Vg = 2Ale. In addition, the capacitance of lead oxide and indium oxide barriers is relatively low, e.g. approximately three times lower than that of comparable niobium oxide barriers.

We have recently reported on all-Pb-Bi junctions 2. They have a larger gap voltage (2Ale = 3.45 mV) and a sharper onset of the quasiparticle current than other lead alloy junctions and are currently used in our receivers. This paper reports on the effect of small amounts of indium, between the counterelectrode and the oxide, on the normal resistance and the quality of the junction. We also show the drastic influence of the humidity on the tunnel barrier and present annealing data. The results are discussed in terms of the Schottky barrier model recently proposed by Baker and Magerlein 3.

* On leave from Yamagata University, 992 Yonezawa, Japan.

0040-6090/86/$3.50 © Elsevier Sequoia/Printed in The Netherlands

168 K. O K U Y A M A , K. H. G U N D L A C H

2 . EXPERIMENTAL DETAILS

The junctions discussed in this paper were fabricated in an evaporation system with an oil diffusion pump and a liquid nitrogen trap, capable of producing a vacuum of about 5 × 10-5 Pa. For the base electrode, first a layer 300 nm thick of pre-alloyed Pb-29wt.°/0Bi was evaporated onto a substrate of fused quartz held at room temperature. This was immediately followed by the deposition of an indium film, generally 30-35 nm thick. Current-heated tantalum boats were used for the evaporations. The film thickness was monitored using a quartz crystal oscillator. The indium layer at the base electrode is very important as it makes the all-Pb-Bi junction resistive to thermal cycling. The indium layer at the base electrode has no noticeable effect on the product RnA and the junction quality in the investigated thickness range 30-60 nm, contrary to the indium layer at the counterelectrode (to be discussed later). The P b - B i - I n base electrode was kept in the evaporation system at a pressure of about 5 × 10 5 Pa at room temperature for 1 h to allow interdiffusion. The sample was then taken out for about 1 h to interchange masks for the deposition of the counterelectrode. Except in one experiment the base electrodes were not oxidized on purpose, i.e. they were not, for example, exposed to pure oxygen or a plasma. Indeed, we found that this is not absolutely necessary. On the contrary, the method gives a lower R, A together with a good junction quality. The oxidation probably took place at the residual pressure of 5 × 10 5 Pa in the vacuum system and on exposure to the atmosphere. An indium layer (thickness, 0.4-60 nm) was evaporated immediately before the deposition of the Pb-29wt.~oBi counter- electrode approximately 350 nm thick. One to three junctions were fabricated in one run. What values of the junction normal resistance R n and area A are required for the SIS mixer? As a rough guideline, coR.C >> 1 is a necessary requirement to minimize harmonic conversion effects 4. Here ~ is the signal frequency and C the junction capacitance. If coRn C is too large a sophisticated tuning circuit is required to prevent the capacitance from shunting out the signal frequency and a small instantaneous input bandwidth generally results. Taking o R , C ~ 4 and for the lead alloy junction C/A = 4 p.F cm - 2, at co = 200n G H z we obtain R,A ~ 160 ~ pm 2. So, for example, if a good input signal power matching requires Rn = 80 ~, the area should be about 2 p.m z.

3. EFFECT OF W A T E R V A P O U R A N D INDIUM AT THE U P P E R SURFACE

It is known, but not widely reported in the literature, that the formation of a tunnel barrier is strongly affected by the humidity of the atmosphere. We therefore performed two sets of experiments, one in August-September when the humidity and the temperature were in the range 40~o-63~o and 24-28 °C respectively and the other in Oc tober -November when the relative humidity was 23~0-42~o and the temperature was in the range 21-26 °C.

As seen in Fig. 1,junctions prepared in the period of high humidity and without indium at the upper surface have an R,A up to a factor of 100 larger than those prepared in the period of low humidity. We checked if this effect can indeed be attributed to the higher humidity of the atmosphere. To this end two P b - B i - I n base

T U N N E L L I N G CHARACTERISTICS O F Pb-Bi-In/oxIDE/Pb-Bi-In J U N C T I O N S 1 6 9

I

10 ~ u l

10 o

E Z

E

&

\\ x\

\

×

×

, -#

o ~ ~ Z ~ o 7 ~b Thickness of indium in n m

Fig. 1. Product of the normal resistance and the area as a function of the thickness of the indium layer between the oxide and the counterelectrode. Curve A (a s = 9-20 g m -3) is obtained with a high atmospheric humidity and curve B (aw = 3.5-9 g m a) is obtained with a low atmospheric humidity, as and a, indicate the absolute humidity ranges in which the junctions were fabricated.

e lect rodes were s imul taneous ly evapo ra t ed at a low a tmospher i c humid i ty (relative humidi ty , 18.1~; T = 23.6 °C). Then one was placed for 40 min in a chamber with a high humid i ty (relative humidi ty , 55~o; T = 25.7 °C) while the o ther was kep t in the l a b o r a t o r y a tmosphere . Thereupon , after changing masks, the P b - B i top e lect rode (wi thout indium) was evapo ra t ed for bo th junct ions . The junc t ion whose base e lec t rode had been exposed to a high humid i ty had R.A = 7700 ~ lam 2 while the o ther gave R,A = 56 ~ i.tm 2. These two values are included in Fig. 1 as circles.

In the per iod of high humid i ty some base e lect rodes were r.f .-sputter c leaned in a rgon jus t before the evapo ra t i on of the P b - B i countere lec t rode. This p rocedure a lways led to values of R,A below 0.5 ~ lam 2.

The da t a shown in Fig. 1 refer to junc t ions in which the base e lec t rode was not exposed to pure oxygen or a p l a sma for oxida t ion . In a pe r iod in which we ob ta ined by this m e t h o d an R,A produc t in the range 50-150 f l lam 2, the base e lect rode was exposed immedia te ly after its fabr ica t ion to pure oxygen at a pressure of 2.7 x 104 Pa for 3 h. This resulted in an R,A produc t of 900 fl I.tm 2.

Figure 1 also d isp lays R,A as a function of the thickness of the ind ium film at the upper interface. In the per iod of high humid i ty (Fig. 1, curve A) the effect of ind ium is small up to a thickness of abou t 2 nm. Between 2 and 4 nm R,A decreases by more than two orders of magni tude . Above 4 nm the influence of ind ium saturates . In the per iod of low humid i ty (Fig. 1, curve B) a large scat ter ing of the da ta po in ts was observed. Nevertheless , a dras t ic decrease in R, A is c lear ly seen.

The ind ium not only decreases R , A but it a lso causes a de te r io ra t ion of the j unc t ion quali ty. In Fig. 2 the 1-V curves of a j unc t ion wi thout and of one with an

170 K. OKUYAMA, K. H. G U N D L A C H

P'l

5O

, o . . . .

0 1 2 3 4 5 Vol tage V in mV

Fig. 2. C o m p a r i s o n of the current vol tage charac ter i s t ics of a junc t ion with an ind ium layer 7.4 nm thick between the oxide and the countere lec t rode (curve 1) and a junc t ion wi thou t such a layer (curve 2). In both

cases R n ~ 50 f~ and T ~ 4.2 K.

indium film 7.4 nm thick at the upper interface are compared. Both have nearly the same value of R n A. The indium causes a larger subgap current which apparently sets in at half the gap voltage. This deterioration was always observed when indium was added to the upper surface, and it was generally less pronounced for thin indium films than for thick indium films.

4. STORAGE AND ANNEALING BEHAVIOUR

Keeping the unprotected junction in the air results in an increase in its normal resistance. This effect strongly differs from junction to junction. Some exhibit an increase in R, by a factor of up to 10 in 1 month while for others R, increased only by 30% in the same period of time. If the junction quality is poor after fabrication, the increase in R, is accompanied by a distinct improvement in quality. One can accelerate the resistance increase in the air by simply placing a drop of water on a separate support a few millimetres away from the junction.

In one experiment three junctions covered with an SiO film about 1 Ixm thick and one without any protection were stored for 25 days in the air. For the latter, R, increased by 30% while for the other three Rn increased by 5%, 20% and 100%. For junctions stored at a temperature of - 10 °C for 3 months there was only a very small change in the I-V curve which could also have resulted from thermal cycling. The resistance R, can be reduced by annealing in the air at a temperature up to 80 °C. Annealing the unprotected junction for 2 h at 70 °C generally reduced Rn by a factor of 1.5-2,5. Junctions with an SiO coating need a much longer annealing time.

5. DISCUSSION

Recently the understanding of the tunnel barrier of lead alloy junctions was advanced by the work of Baker and Magerlein 3. While the trapezoidal barrier model appears to be a reasonable approximation for some materials, such as AI/A1203/AI 4, P b / P b O / P b 5 and Sn/SnOx/Sn 6, these researchers strongly suggest that tunnelling in P b - l n - A u / o x i d e / P b - A u occurs through a Schottky barrier

TUNNELLING CHARACTERISTICS OF Pb-Bi-In/oxIDE/Pb-Bi-In JUNCTIONS 171

provided the oxide consists mainly of indium oxide, which is known to be an n-type semiconductor, doped with an excess of indium as high as 5 × 1020 cm-3. Schottky barriers are thought to exist at the base electrode and at the counterelectrode interface. But tunnelling was found to be controlled by the counterelectrode interface: the product R , A of the normal resistance and the area of the junction is hardly affected by the composition at the base electrode interface but is strongly affected by the composition at the counterelectrode interface. The Schottky barrier at the base electrode is probably thin because of the large donor concentration due to the proximity of the indium metal in the base electrode. The barrier at the upper interface becomes large when a very thin layer of PbO is formed between the indium oxide and the counterelectrode and becomes small when indium is added to the counterelectrode.

The results from our fabrication procedure strongly suggest that the oxide on the base electrode mainly consists (before the evaporation of the counterelectrode) of indium oxide. The upper surface of the oxide is covered with molecules such as oxygen, water etc. Consequently, not only a Schottky barrier of indium oxide is formed. It is likely that in addition a thin layer of lead oxide and bismuth oxide forms on top of the indium oxide barrier. Water molecules promote the formation of this additional barrier and thus lead to a larger R . A . Sputter cleaning in argon just before the evaporation of the counterelectrode cleans the upper surface of the indium oxide and thus prevents the formation of the additional barrier; as a result, R, A is very low. Broom et al. ~ reported that the junction normal resistance increases with increasing partial pressure of oxygen in the system during evaporation of the first monolayers of the counterelectrode and suggested that PbO is formed at the interface between the indium oxide and the counterelectrode by the reaction of lead with adsorbed oxygen. The thin film of indium deposited just before the evaporation of the counterelectrode can produce two effects: (a) it increases the donor concentration and thus causes a thinner Schottky barrier; (b) it reacts with the oxygen and water molecules, which are trapped on the surface of the oxide layer, to form indium oxide and hence reduces or prevents the formation of a layer of lead oxide and bismuth oxide on top of the indium oxide during the evaporation of the Pb/Bi counterelectrode, Both effects increase the tunnelling probability. Water molecules at the upper surface promote the formation of the additional barrier and hence reduce the effectiveness of the indium in reducing R, A.

Garno 8 fabricated Pb/oxide/Pb junctions with a conductance range of 104 ~ - 1 cm-2 by oxidizing lead tunnel barriers in a humidity-controlled oxygen- regulated atmosphere. The resistivity of the junction changed from 5 x 103 f~ i,tm z for 50~o humidity (oxidation time, 50 min) to 8.3 x l 0 6 ~ igm 2 for 98~o humidity. A Pb-Bi- In /oxide/Pb-Bi junct ion fabricated in our laboratory for 55~o humidity has a resistivity of 7.7 × 103 ~ ~tm z. This value is close to Garn. o's result (oxygen of 50~o humidity). However, it should be pointed out that the role of humidity or water molecules for Pb-ox ide -Pb junctions is different than that for Pb-Bi-In/oxide/ Pb-Bi junctions. For the former, water molecules contained in the oxygen atmosphere can promote the growth of a lead oxide layer which indeed acts as a tunnel barrier. For the latter, water molecules adsorbed onto the surface of the indium oxide are important; these are considered to enhance the formation of a thin

172 K. OKUYAMA, K. H. GUNDLACH

Pb-Bi oxide layer between the indium oxide and the upper Pb-Bi electrode. This explanation is based on the fact that the current density of indium oxide barrier junctions is relatively independent of the physical thickness of the oxide for thicknesses rang!ng from 3 to 6 nm (see ref. 3).

The pronounced current rise at half the gap voltage V = Ale in Fig. 2 is qualitatively similar to the result for other lead alloy tunnel junctions (generally with a high current density) without an evaporated indium layer at the upper interface 9. This has been explained in terms of two-particle tunnelling in a "patchy" barrier. It is believed that indium induces spots with high tunnelling probability in the oxide, favouring two-particle tunnelling. Our result in Fig. 2 indicates that the effect can easily be created by including indium at the upper interface.

The increase in the normal resistance in the air together with the improvement of the junction quality is probably due to penetration of water vapour into the barrier as was similarly suggested for other lead alloy junctions 7. Clearly, the SiO layer l rtm thick may reduce, but cannot prevent, penetration of water vapour into the junction. It is known that the indium oxide barrier is more transparent than, for example, lead oxide. The decrease in the junction resistance by annealing at elevated temperatures could be due to the growth of indium oxide at the expense of other oxides. But it is obviously also related to the presence of water molecules in the barrier since a protective SiO layer enhances the annealing time remarkably.

REFEREN CES

1 T .G . Phillips and D. P. Woody, Ann. Rev. Astron. Astrophys., 20 (1982) 285. R. Blundell, K. H. Gundlach and E. J. Blum, Electron. Lett., 19 (1983) 498. E. C. Sutton, IEEE Trans. Microwave Theory Tech., 31 (1983) 473. L. Olsson, S. Rudner, E. Kollberg and C, A. Lindstr6m, Int. J. ln/~'ared Millimeter Waves, 4 (1983) 847. S. K. Pan, M. J. Feldman, A. R. Kerr and P. Timbie, Appl. Phys. Lett., 43 (1983) 786.

2 K . H . Gundlach, S. Takada, M. Zahn and H. J. Hartfuss, Appl. Phys. Lett., 41 (1982) 294. 3 J .M. Baker and H. J. Magerlein, J. Appl. Phys., 54 (1983) 2556.

H. J. Magerlein, J. Appl. Phys., 54 (1983) 2569. 4 K . H . Gundlach and J. H61zl, Surf Sei., 27 (1971) 125. 5 S. Basavaiah, J. M. Eldridge and J. Matisoo, J. Appl. Phys,, 45 (1974) 457. 6 H. Wehr and K. Knorr, Z. Phys. B, 33 (1979) 21. 7 R .F . Broom, R. Jaggi, Th. O. Mohr and A. Oosenbrug, IBM J. Res. Dev., 24 (1980) 206. 8 J .P. Garno, J. Appl. Phys., 48 (1977) 4627. 9 P . W . Epperlein, in H. D. Hahlbohm and H. Lfibbig (eds.), Proc. Int. Con/] on Superconducting

Quantum Devices, Berlin, May 6 9, 1980, de Gruyter, Berlin, 1980, p, 131.