Formation of lithographic metallic hetero-contacts

Formation of lithographic metallic hetero-contacts

PHYSICA[ ELSEVIER Physica B 218 (1996) 97-100 Formation of lithographic metallic hetero-contacts N . N . G r i b o v a'b, J. Caro a'*, S. Radelaar a...

522KB Sizes 0 Downloads 13 Views


Physica B 218 (1996) 97-100

Formation of lithographic metallic hetero-contacts N . N . G r i b o v a'b, J. Caro a'*, S. Radelaar a aDelft Institute of Microelectronics and Submicron Technology (DIMES), Delft Universi O' of Technology, Lorentzweg 1, 2628 CJ Delft, The Netherlands b B . I . Verkin Institute of Low Temperature Physics and Engineering, Academv of Scienses of Ukraine, 47 Lenin Avenue, Kharkiv 310164, Ukraine

Abstract We have made an electron-microscopy study of nanoholes in membranes in successive stages of one-sided deposition of Au and Ag, as is relevant for hetero-contacts. Key results are: (i) the holes are not filled during deposition and (ii) closing of the holes is by lateral growth of the film on the membrane.

1. Introduction In recent years it has been demonstrated that welldefined and very stable metallic point contacts can be realised with nano-fabrication techniques [1,2]. The contacts are made by deposition of metal on a thin membrane with a hole. In this way the hole is filled with metal and electrodes are formed (see inset in Fig. 1). With nanofabricated contacts several new, subtle transport phenomena have been discovered, e.g. two-level resistance fluctuations due to defect motion in the constriction [3, 4] and conductance fluctuations involving electron scattering at remote defects [5, 6]. So far these devices were almost exclusively fabricated as homo-contacts, i.e. contacts made of one and the same (very pure) metal. However, it should also be possible to fabricate and measure non-homogeneous contacts such as hetero-contacts or contacts with a layer of impurities inserted in the constriction. Indeed, hetero-contacts can readily be made with the existing fabrication scheme by depositing different metals.

* Corresponding author.

This is demonstrated in Fig. 1, which shows point-contact spectra of two types of Ag/AI hetero-contacts. For type A contacts 200 n m Ag was deposited on side 1 of the membrane and 200 n m A1 on side 2. For type B, after deposition of 200 n m Ag on side 1, a bilayer of 10 n m Ag/200 nm AI or 20 n m Ag/200 n m AI was deposited on side 2, to move the Ag/A1 interface away from the constriction. As expected, the spectrum of a type A device has clear p h o n o n peaks of both Ag and A1. For a type B contact the 35 mV peak of A1 is still present in case of 10 nm Ag, but is absent in case of 20 nm Ag. Transport through a hetero-contact or a contact with a layer of impurities at a homo-interface in the constriction depends sensitively on the interface roughness and interface purity [7]. Before trying to control these quantities, one should first understand the way of formation of hetero-contacts, which so far is unknown. In this paper we make the first step in this direction.

2. Microscopy of nanoholes in successive stages of metal deposition Samples for microscopy were prepared using our fabrication process for metallic point contacts [2]. Holes are

0921-4526/96/$15.00 ,~'~' 1996 Elsevier Science B.V. All rights reserved SSDI 0 9 2 1 - 4 5 2 6 ( 9 5 ) 0 0 5 6 8 - 4


N.N. Gribov et al. / Physica B 218 (1996.) 97 100 I





2OO nm metal

Si 200 nm metal


10 nm A g / 2 U 0 nm AI


~ >

50 nm

4o AAg ~ . t . , / ~ 0 0

2o 0

nm AI














V (mY) Fig. I. Spectra of the Ag/Al hetero-contacts described in the text. For a 20 nm thick Ag interlayer phonon peaks from AI are absent. Contact diameters are in the range 10-25 nm. The inset shows a schematic cross-section through a point contact.


~unm iag

patterned by single-pixel exposure with a Leica e-beam writer of P M M A resist on a Si3N 4 membrane, and the resist is developed. Holes in resist are transferred to the nitride by dry etching in an SF6/He plasma. Etching is stopped when the resist is completely consumed and a nitride thickness of 25 nm results. Then, in the deposition chamber, the membrane is cleaned in an O2 glow discharge and metal is deposited by thermal evaporation. The chamber pressure, substrate temperature and deposition rate are 5 x 10 7 Torr, 300 K and 0.3-0.5 nm/s, respectively. First we made scanning electron microscope (SEM) inspections to determine the shape of the holes and the way they are filled and/or covered with metal during one-sided deposition of a thick (200 rim) Au or Ag layer on substrates oriented perpendicular to the evaporation beam. We used arrays of 750 x 750 holes placed on an 80 nm period square grid, which were etched as holes for point contacts. In this case, however, we did not use membranes but a nitride layer supported by Si, to facilitate later breaking of the wafer through a "line of holes". The micrographs in Figs. 2(a), (b) give resulting crosssections. As can be seen the nitride surrounding a hole is severely underetched, which results from the high isotropic etch rate of Si in the SF6/He plasma. So, effectively each hole is in a (small) membrane, as a real contact hole. The profile of a hole is rounded. This arises from transferring the resist profile to the nitride layer and the subsequent overetching. The lower diameter of a hole is 25 nm. Further, the step coverage of the Ag film is very bad (see Fig. 2(b)): the film covers the hole, but does not fill it, contrary to the assumption made in Ref. [1]. This finding agrees with the fact that thermal evaporation of metal is not suitable to fill vias in microelectronics devices,

inte rface

Fig. 2. SEM micrographs of cross sections through 25 nm diameter holes, before deposition (a) and after deposition of 200 nm Ag (b). The voids result from etching in the SF6/He plasma. At the bottom of the voids in (b) some Ag is deposited. Small disturbances of the cross-sections may exist as a result of breaking. Dots guide the eye in following interfaces/surfaces.

contrary to chemical vapour deposition [-8]. Inspection of the outer Ag surface (not visible in Fig. 2(a)) revealed an array of shallow pits above the holes. These pits result from missing atoms which have passed through the holes before closing. In case of a 200 nm Au layer deposited on array samples we made very similar observations. As a further step we followed with transmission electron microscopy (TEM) the growth of Au and Ag in successive stages of deposition on membranes with an array of 7 x 5 holes of diameters in the range 30-70 nm. Fig. 3 shows T E M micrographs in three stages of the deposition of Au. F r o m these micrographs and similar ones in other deposition stages we make the following observations: 1. In the initial deposition stage ( ~ 2 nm on thickness monitor, i.e. 1.2x 1016atoms/cm 2, Fig. 3(a)) there is a certain density of grains on the unpatterned part of the


N.N. Gribov et al. / Physica B 218 (1996) 97 100 I


60 40







AU -!

20 0 0

40 80 120 Film thickness [nm]

Fig. 4. Remaining hole diameter (average over membranes and directions) as a function of Au thickness deposited on one side of the membrane, for four initial hole diameters. Error bars result from differences between membranes and measurement directions. Dashed segments are extrapolations.

Fig. 3. TEM micrographs of membranes with holes after deposition of 2, 5 and 20 nm Au {a, b and c, respectively). Closing of the hole is by lateral growth of the continuous film. A necklace of grains in the narrow part of the hole is discernible in (a) and (b). The bar marker in tc) is the same for (a) and (b).

membrane and the rounded wall of the hole. However, the steep region near the inner edge of the holes (close to the narrowest part) is decorated with a necklace of smaller grains, which have a higher density. 2. Grain coalescence on the membrane and on the rounded walls (2.9 x 1016 atoms/cm 2, Fig. 3(b)) occurs in the usual way for this type of amorphous substrate. Grains of the necklace grow slower than grains on the rounded wall, so that they cannot coalesce with the regular grains in the broader section of the hole. 3. In the stage of formation of a closed poly-crystalline film on the membrane (11.8 x 1016 atoms/cm 2, Fig. 3(c)) closing of the holes occurs by lateral growth of the film.

Also for these deposition experiments the observations for Au and Ag were very similar, albeit that Ag exposure required to reach the stages of nucleus growth and grain coalescence are approximately twice those of Au. For four initial hole diameters we measured in T E M micrographs the remaining hole size after deposition of different metal thicknesses. Fig. 4 shows the resulting dependencies. The rapid initial decrease is due to formation of the necklace. Further size decrease proceeds slowly. F r o m the figure it can be seen that for an initial diameter of 30 nm, one should deposit about 70 nm Au to close the hole, i.e. a layer more than twice as thick as the hole diameter. The metal that covers the hole will acts as a microsubstrate during deposition on the other side. T E M inspection after deposition of 20 nm Ag on side two (200 nm Ag was deposited on side one) showed a dark region inside the hole, surrounded by a light halo, while it was not clear that growth had started on the side-wall. This suggests preferential growth on the metal, which after prolonged deposition might lead to a (poly-)crystal bulging out of the hole. If this is true, it may explain the absence of the LA peak in Fig. 1 for the Ag/A1 heterocontact with the 20 nm interlayer, which is rather thin compared to the expected depth to which the electron phonon interaction can be probed.

3. Discussion Our observations for growth of Au and Ag films on unpatterned parts of the membranes and on the rounded edges of the holes agree with the accepted picture for growth on amorphous substrates and need no further discussion. The main point to be addressed is that the


N.N. Gribov et al. / Physica B 218 (1996) 97 100

holes do not fill with metal, apart from the inner edge region, where the necklace is formed. For this we suggest the following explanation. The edge of the hole is a substrate portion with both positive curvature (rounded profile, see Fig. 2) and negative curvature (due to the circular shape, as seen in top view). F r o m studies of the metallization of vias in integrated circuits it is known [9] that areas of dominating negative curvature are preferential sites for nucleation and growth. In our case this leads to nuclei close to the narrowest part of the hole and subsequently to the necklace of small grains. Further, the hole is a missing substrate area, so that atoms are lost. This reduces the number of adatoms collected by nuclei near the hole, so that these nuclei grow slower than those away from the hole. This effect is even enhanced because of the angle of the wall of the hole with the incoming beam of atoms (increase of area). We believe that the avoidance of the inner edge of the hole by coalesced grains (see Fig. 3(b)) and the slower growth of necklace grains arise from this mechanism. The results for one-sided deposition tempt us to speculate about the fabrication of homo-contacts, which are usually formed while rotating the substrate in the beam emerging from the melt. We expect that such a two-sided deposition will not noticeably suppress the tendency of the wall of a hole to remain uncovered. However, when a hole starts to close from two sides, edges become available to grow on (as seen from the inside), so that filling starts. Simultaneously, the interior of the hole becomes less accessible, so that filling becomes harder and may not be completed in some cases (void formation). Clearly, this would limit the yield of the fabrication process. In summary, we have made a study of the closing of nanoholes in membranes in successive stages of onesided deposition of Au and Ag, as it occurs in heterocontact formation. It is found that the holes do not fill with metal in a one-sided deposition. The holes are closed

by lateral growth of the film on the membrane when its thickness is more than twice the hole diameter. Metal closing one side of a hole serves as a micro-substrate for growth of metal deposited from the other side. This growth seems to occur preferentially on the metal and not on the side-walls.

Acknowledgements N.N. Gribov acknowledges the "Nederlandse Organisatie voor Wetenscbappelijk Onderzoek (NWO)" for a grant received in a program for scientists of the former Soviet U n i o n (Ref. no. 714-033). We thank C.D. de Haan of the National Centre of H R E M of T U D e l f t for T E M analyses.

References [-13 K.S. Rails, R.A. Buhrman and R.C. Tiberio, Appl. Phys. Lett. 55 (1989) 2459. I-2] P.A.M. Holweg, J. Caro, A.H. Verbruggen and S. Radelaar, Microelectronic Eng. l 1 (1990) 27. [3] K.S. Rails and R.A. Buhrman, Phys. Rev. Lett. 60 (1988) 2443. [4] P.A.M. Holweg, J. Caro, A.H. Verbruggen and S. Radelaar, Phys. Rev. B 45 (1992) 9311. 1-5] P.A.M. Holweg, J.A. Kokkedee, J. Caro, A.H. Verbruggen, S. Radelaar, A.G.M. Jansen and P. Wyder, Phys. Rev. Lett. 67 (1991) 2549. [6] P.A.M. Holweg, J. Caro, A.H. Verbruggen and S. Radelaar, Phys. Rev. B 48 (1993) 2479. 1-7] G.E.W. Bauer, Phys. Rev. B 51 (July 15 issue, 1995). [8] N.N. Gribov, J. Caro, T.G.M. Oosterlaken and S. Radelaar, Physica B 218 (1996) 101. [9] T. Smy, S.K. Dew, M.J. Brett, Presented at a course in conjunction with the Advanced Metalisation Conf. in Texas, 1994.