By far the most research and development on materials for MCM dielectric layers has gone into polyimides, and most existing applications are based on polymers of this family. Great strides have been made in achieving the demanding property mix required through careful tailoring of the monomer chemistry. Improved adhesion, lower dielectric constant, reduced sensitivity to moisture, higher thermal stability, and other properties have been improved greatly. The in-plane coefficient of thermal expansion was reduced and adjusted to the range of silicon, metals, and ceramics. Most major electronics companies manufacture MCMs based on polyimides.

In spite of the extent of commitment to polyimides, it has proved difficult to achieve all the desired properties in a given composition. Other polymer dielectrics are in use, and new materials are under consideration. For example, commercial MCMs are manufactured by one electronics systems provider based on a proprietary epoxy-acrylate-triazine polymer that is photodefinable. Sample MCMs have been produced based on a benzocyclobutene (BCB) polymer dielectric. In spite of the large experience base with the polyimide materials, the newer polymers have advantages and offer attractive alternatives. All of the candidates are glassy polymers. The dielectric constants may be compared as follows:

alumina

9

glass ceramics

4-5

fused silica

4

polyimides

3-4

triazine

2.8

BCB

2.7

In the final analysis, the choice of materials will be based on the sum of property advantages and processing practicality. Polymers offer the lowest dielectric constants and the thinnest "wires."

Lithographic processes and associated technologies have advanced to the point that semiconductor device cells and conductor lines (i.e., the on-chip "wires") are so small (less than 1 µm) and the switching times are so fast that the continual increase in performance traditionally derived from a combination of improvements in device structure and reduction of device dimensions cannot be fully realized. This is owing to the fact that the propagation of signals through the wiring on the chip (and in the module) is becoming the dominant limitation on processor cycle time.

The velocity of pulse propagation in these structures is inversely proportional to the square root of the dielectric constant of the medium. Hence, reductions in the dielectric constant translate directly into improvements in processor cycle time, in part because of the speed of propagation. In addition, the distance between signal lines is dictated by noise issues or "cross-talk" that results from induced current in conductors adjacent to active signal lines. A reduction of the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement