Thin, 100-nm films of first silver and then copper were deposited consecutively onto inert substrates by magnetron sputter deposition. Constant anodic current densities were applied at room temperature to dissolve the outer copper film to varying depths. The 50Cu/50Ag interface, derived from the Auger Electron Spectroscopic concentration-depth profile, initially moved into the copper toward the outer dissolving surface, indicating enhanced diffusion of copper into silver. After longer times at all anodic Current densities, the interface reversed and moved back toward the underlying silver-rich layer, indicating that eventually diffusion of silver into copper predominated. The reversal time was inversely proportional to the anodic Current density. These effects are explained by anodic formation of subsurface vacancies which migrate as divacancies to the copper/ silver interface where they affect interface movements by the well known Kirkendall mechanism Calculated diffusivities up to 10-12 cm2/sec at maximum anodic current densities of 900 µA/cm2 are dramatically above any that are normally observed at room temperature.
Vacancies may be formed during anodic dissolution and corrosion. Figure I shows the mechanism suggested by Pickering and Wagner1, in which vacancies are formed when subsurface atoms jump into surface kinks or ledges on a dissolving crystal surface. Forty2 originally suggested that vacancies produced by dezincification may interact with dislocations to produce stress corrosion cracking in brass. Pickering and Wagnerl further implicated vacancies as the cause of dealloying in brasses. Revie and Uhlig3 postulated anodically generated vacancies to explain enhanced creep during anodic dissolution of copper. One of the present authors4 proposed that the processes causing increased creep may also play a role in the mechanism of stress corrosion cracking (SCC) and corrosion fatigue. Meletis5 suggests that corrosion-generated vacancies are at1racted to dislocations, modify dislocation configuration, and produce embrittlement by SCC. Magnin6 and Aaltonen et al7 also have receot1y suggested a role fur corrosion-generated vacancies in the mechanism of SCC.
The present authors reported initial experimental evidence of anodically generated vacancies in Cu/Ag thin film diffusion couples8 and more recently presented subsequent results along with preliminary observations on data analysis and mechanism9. This paper describes the completed analysis of the diffusion data along with further theoretical interpretations.
EXPERIMENTAL PROCEDURES
The Cu/Ag diffusion couples were prepared using magnetron sputter deposition on commercially available polished Si (111) wafers which provided a convenient reproducible substrate fur deposition with no apparent further role in subsequent processes. In the evacuated deposition chamber, vertical Si wafers were rotated horizontally past 0.999 pure Ag and Cu magnetron sources, ensuring that layer thicknesses were uniform over each sample surface and from sample to sample. independent measurements showed maximum errors of 5% in calculated layer thicknesses. A 100-nm base layer of Ag was deposited first on each Si-wafer substrate, followed by a 100-nm Cu layer on the Ag base layer. X-ray diffraction characterization indicated that the Cu and Ag films grew epitaxia1ly in a (111) orientation on the Si substrate. Preliminary examinations revealed a grain size in the Ag and Cu layers of about 1 -µm.
