Limitations on high-temperature processing of aluminum-copper metallization

Abstract

type in Fig. 2(c), all such systems being fed in parallel. (This is a special case of (S).) Similarly for (lo), except that now the coefficients ci, replace the y;, in Fig. 2(c) and generate sensitivity system poles which cancel one pair of zeros in Y. Hence this method is not suitable in general for non-minimum-phase zeros, though one pair of such zeros may be accommodated in the last stage of the cascade, the sensitivities being obtained without the need for an additional model, as in Fig. 2(a).In view of the simplicity of Fig. 2(a) for sensitivities associated with numerator coefficients, it may be of advantage in general to use a scheme in which only the denominator polynomial is factored, the numerator being generated by means of a fully parallel feedforward structure, at the expense of increased sensitivities. Nevertheless, even if a cascaded numerator is used, the overall number of sensitivity models is only one for each polynomial factor in the transfer function, except for the last numerator factor in the cascade, for which the sensitivities remain available on the original system by the method of Fig. 2(a). REFERENCES (11 R. Tomovic, Scnsitiaity Analysis of Dynamic Sysfnu (transl.). New York: [Z] P. V. Kokotovic. 'Method of sensitivity points in the investigation and o p

Evolution and Current Status of Aluminum Metallization

“…In addition, the adherence, etching, step coverage, bonding, and contact resistance properties for the above alloys remain essentially the same as for aluminum while resistivity is increased only about 15%. Care must be exercised in postmetallization heat-treatments because of the reduced eutectic temperatures of the alloys (255). With one possible exception, it may therefore be stated that the employment of aluminum alloy metallization has allowed the requirements for metallization on silicon to be met and has greatly extended the usefulness of a basically aluminum process.…”