Aluminum thin films have been in common use as interconnections on Si devices since the early 1960’s. However, during the past decade alloying additions have been made to Al to improve some aspect of interconnection performance. For example, Si additions minimize metal penetration at ohmic contacts during heat treatment1; Cu additions greatly reduce the mass transport of Al due to electromigration.2It is widely recognized that Al interconnections will corrode in the field if not adequately protected by hermetic packages or a glass passivation system.3However, there has been little discussion of the corrosion of Al films during the fabrication of devices (i.e., in‐process) when there is typically little protection for the film, yet much exposure to corrosive reagents and environments. It is the purpose of this paper to focus on in‐process corrosion.The corrosion of Al and Al alloys in bulk metallurgical form has been studied for many years and is now reasonably well understood.4For example, it is known that the Cu‐bearing Al alloys are more subject to intergranular and pitting corrosion than pure Al. However, one very important difference between bulk and thin film observations is that superficial surface corrosion in bulk material can be equivalent to total destruction of a thin film which is typically 1 μm in thickness.When Al–Cu alloy (4.5% Cu, balance Al) was first introduced to the manufacturing of integrated circuits, the periodic presence of black spots on the interconnection patterns was noted. The level of the problem was low; less than 1% of the integrated circuits on a wafer would exhibit one or more black spots. At first, it was thought that the spots were of cosmetic importance, a consequence of poor photoresist removal or contamination of the surface. However, the scanning electron microscope soon revealed that the problem was one of ’’missing aluminum’’ in which a large part of the metallic cross section was absent. Therefore, the phenomenon was of great importance with regard to long term reliability of the product. The defect was too fine and infrequent to effectively screen with visual inspection.The use of electron microscopy and diffraction techniques showed the problem to be intergranular corrosion occurring in the processing sequence after subtractive etching, but before sputtered SiO2deposition. The corrosion was traced from the surfaces of the conductor along grain boundary paths. The subsequent use of ultrasonic cleaning magnified the defect by causing the separation of granular masses of Al. Typically, the associated corrosion products were chlorides, oxychlorides, and oxides of Al and Cu.The prime source of corrosion was found to be chlorinated hydrocarbons (Freon and trichloroethylene) which had decomposed in wafer cleaning baths. The decomposition came from the unintentional contamination of the chlorinated solvents with water and/or alcohol which caused the generation of free chloride ions in the bath. Chloride ions act as strong corrosion accelerators for reactive metals such as Al.5Characterization of alloy Al–Cu films showed large compositional inhomogeneities in the as‐deposited films, with anomalously large concentrations of Cu near the bottom of the film. After the heat treatment required to form good ohmic contacts, the Al–Cu becomes metallurgically overaged. Large particles of ϑ‐phase (CuAl2) deplete the Al alloy of Cu adjacent to grain boundaries. The Al‐rich regions in contact with the higher Cu regions establish microscopic galvanic couples which accelerate the dissolution of the Al‐rich regions. The result is intergranular and pitting corrosion.6Some effort was made to optimize the structure of the alloy to minimize corrosion. A more uniform distribution of Cu in the vertical profile, a reduction in tensile residual stress, and the sandwiching of the alloy between thin sacrificial layers of pure Al (thin film Alclad) were all steps in the right direction. However, optimization of the film structure could only modify the mode of the corrosion process, not eliminate it.The removal of chlorinated solvents, wherever possible, was the most effective means of eliminating ’’missing Al’’ even in nonhomogeneous, overaged alloy.However, once sensitivity to the problem was established, other promoters of the corrosion were also identified: fluoride ions and highly polar alcohols.It had been recognized, even with pure Al films, that the buffered HF used to etch via holes in sputtered SiO2passivation will rapidly attack Al when the etchant is greatly diluted. The danger period is not immediately at the end point of the etch cycle when the buffered HF mixture impinges directly on Al in the bottom of the via, but in the subsequent wash cycle in deionized water. As the HF is diluted with water, it passes through a low concentration level at which the etch rate of Al is maximized.7The postetch washing process must pass through this peak quickly to avoid corrosion.Molded CTFE (chlorotrifluorethylene) wafer carriers having trace impurities of free fluoride, a result of polymer decomposition during molding, were also found to be a cause of corrosion.During the recent petrochemical shortage, a much more subtle corrodant was discovered. The substitution of the more readily obtained methanol for isopropyl alcohol triggered a mild outbreak of in‐process corrosion in Al–Cu. It was then confirmed that a highly polar alcohol such as methanol can also be mildly corrosive to Al and should be avoided in thin film processing.