In the next decade, the width of some metal lines inside computer chips will decrease to approximately 50nm, which is comparable to the mean free path of conduction electrons in both Al and Cu at room temperature. At those small dimensions, electron scattering from metal surfaces will play an important role in electron transport.

      The well-established way to characterize the surface scattering is to study electrical properties of thin metal films. To the best of our knowledge, all previous work related to Al and Cu has been done on polycrystalline films deposited on SiO₂ [1,2]. However, the temperature-dependent measurements of similar films have suggested [3] that the observed thickness dependence of resistivity may be caused by grain-boundary scattering, as the grain size typically decreases with film thickness. To avoid this masking effect, it is advantageous to work with single-crystalline metal films. In addition, it is very convenient for resistivity measurements to have an insulating substrate. Epitaxy has been previously reported for both Al and Cu on CaF₂(111) [4,5], but the initial stages of their growth have not been addressed. In this work, we have studied thin (<20nm) Al and Cu films grown on CaF₂/Si(111) substrates by molecular beam epitaxy (MBE).

      The hydrogen-terminated Si(111) wafers were first outgassed at 250°C for one hour. One monolayer of CaF₂ was then deposited, followed by 750°C anneal to react the CaF₂ with Si. In this way, we made sure that the C atoms inevitably present on Si surface remain buried as the isolated impurities, instead of migrating along the surface to form SiC crystallites. The 30nm-thick epitaxial CaF₂ was then grown at 750°C, in order to form an insulating layer sufficiently thick to resist mechanical punch-through during subsequent resistivity measurements of Al. The atomic force microscopy (AFM) showed that the resulting surface had atomic terrace width of about 50nm and rms roughness of 0.2nm on the 500nm-wide scan. In most of the structures, additional 1.5 monolayers of CaF₂ were deposited at nominally room temperature (perhaps the surface was heated somewhat by radiation from the source) to create high concentration of atomic steps. For this surface, the lateral size of small islands was below 5nm, since it could not be resolved by the AFM tip with radius of approximately 15nm.

      The reflection high energy electron diffraction (RHEED) patterns from Al on CaF₂ confirmed epitaxy, as no polycrystalline rings were visible and streak spacing was consistent with that of the Al(111) surface. For Cu on CaF₂, RHEED showed formation of the polycrystal at several temperatures and deposition rates, contrary to ref. 5. However, we have recently succeeded in growing epitaxial Cu by depositing it on 3nm-thick Al seed layer. In this case, RHEED showed only streaks, with the spacing indicative of the Cu(111) surface. All corresponding RHEED patterns are shown on Fig. 1.

      From the AFM we learned that Al initially forms islands when deposited on CaF₂. To grow thin continuous films, it is necessary to make those islands coalesce as early as possible, which is accomplished by increasing their nucleation density. This is in turn achieved by using low growth temperatures, high deposition rates, and substrates with high concentration of special nucleation sites. In agreement with this approach, for growth of 10nm-thick films, we observed that the Al rms roughness is the smallest (0.26nm on the 500nm-wide scan) when Al is grown at nominally room temperature, at the rate of 2nm/s, and on highly stepped CaF₂ surface.

      To learn more about the structure of the Al films, we performed the scanning tunneling microscopy (STM) measurements on a locally designed instrument coupled to the MBE chamber. In Fig. 2, we present an STM image from the 12nm-thick Al film.

    Having obtained thin well-connected epitaxial Al films on an insulating substrate, we proceeded with measurements of resistivity dependence on thickness. The Al was deposited through a shadow mask, and the sheat resistance of the resulting test structures was measured ex-situ by a linear 4-point probe. The film thickness was obtained with precision of ±0.5nm by performing a large-area AFM scan at the shadow edge of Al and averaging the apparent step height along the step line. To arrive at the true thickness of conducting Al, we then subtracted 1.2nm from the measured value to take into account Al oxidation. The resistivity of Al was calculated as a product of sheat resistance and thickness.

    On Figure 3, the resistivity data from the 5 samples grown so far are shown along with the calculation by the Fuchs-Sondheimer model [6] assuming bulk aluminum resistivity of 2.7microOhm·cm, electron mean-free path of 40nm, and totally diffuse scattering at metal surfaces.

    In the end, we want to stress again that these data are valuable because of the absence of grain-boundary scattering effects in them.

[1] J. Gogl, J. Vancea, H. Hoffmann, J. Phys. : Condens. Matter 2 (1990) 1795.
[2] E. Dobierzewska-Mozrzymas, F. Warkusz, Thin Solid Films 43 (1977) 267.
[3] J.W.C. de Vries, Thin Solid Films 167 (1988) 25.
[4] C.-C. Cho, H.-Y. Liu, H.-L. Tsai, Appl. Phys. Lett. 61 (1992) 270.
[5] N. Mattoso, D.H. Mosca, I. Mazzaro, S.R. Teixeira, W.H. Schreiner, J. Appl. Phys. 77 (1995) 2831
[6] E.H. Sondheimer, Adv. Phys. 1 (1952) 1.