Advanced semiconductor device interconnects are on the brink of a dram
atic change, With ever-increasing demand for faster logic devices and
higher capacity memories, semiconductor designers will maintain the cu
rrent trend toward shrinking design ground rules and increased interco
nnect packing densities. As interconnect lines shrink and move closer
together, the resistance of aluminum/ copper and tungsten conductors a
nd the capacitance of structures with SiO2-based dielectrics limit fur
ther increases in clock speed, Already, interconnect RC delays approxi
mately equal delays due to transistor gate length. In addition, reduce
d interconnect line cross sections translate into increased current de
nsities for logic devices, Line failure due to electromigration become
s a serious concern for aluminum/copper interconnect lines with cross
sections significantly below those used in state-of-the-art semiconduc
tor devices. Many semiconductor manufacturers envision switching to lo
wer-resistance, electromigration-stable copper conductors and lower pe
rmittivity (low-k) organic dielectrics. This revolutionary path is com
plicated by the absence of manufacturable processes and integration kn
owledge for copper and low-k dielectrics. Evolutionary process enhance
ments will thus drive current materials as far as possible until the a
dvantages of new materials outweigh their risks. This article discusse
s alternative interconnect structures for future generation devices. I
t will point out where fundamental limitations will likely demand new
process and material sets, and where the advantages of new materials a
nd processes may compel their use.