Page 385 - Sami Franssila Introduction to Microfabrication

364 Introduction to Microfabrication

vacuum tube but that it can do things that the vacuum every year. In 1965, Gordon Moore spoke about this
tube could never do!” (from reference Ross). pattern. The observation was based on few data points,
Many MEMS and nanodevices today are minia- but the conclusion became famous. Later, the prediction
turized versions of existing devices. Sometimes, a was revised to doubling every 18 months, and this
smaller size is useful because it results in, for example, version has been especially long lasting. It has been
smaller power consumption or higher speed. However, dubbed Moore’s law, even though it is only an empirical
it is equally important to look for new applications pattern without fundamental justification (Table 38.1).
in which new physical phenomena, new combina- Moore’s 1965 prediction extended till 1975 and
tions of speed and power can be utilized, or where his extrapolation was quite accurate. The trend has
macroscopic counterparts do not exist, or where the continued approximately at the predicted speed, give or
scale economies of microfabrication have not yet take some fluctuations. At the turn of the millennium,
been utilized. the pace has been even faster than that predicted by
The whole can be more than the sum of its parts. The Moore’s law. DRAM memory chips are best suited for
very concept of integration seems to have escaped the Moore’s law studies because the law is about production
attention of the supporters of argument number four. economics: chip size and cost minimization. Processors
Integrated circuits paved the way for more powerful are governed by quite different laws: they are design-
electronic systems. And the savings in assembly costs heavy, rather than manufacturing-driven, and proprietary
quickly more than compensated the higher cost of ICs. architectures are not subject to ultimate cost reductions.
Argument five was mathematically valid, but it was One gigabit DRAM circuits were unveiled with 0.18 µm
based on the technology of its day, and it did not geometries as predicted, and 4 Gbit DRAM memory
anticipate the tremendous strides in microfabrication circuits with 0.10 µm dimensions are being made in
technologies. The success of ICs has been dependent on 2003. It should be borne in mind that sometimes the
the fact that in spite of continuous miniaturization and product demonstration date is used (when the first
complexity of the manufacturing process, the yield of fully functional chips are fabricated), sometimes the
individual transistors on ICs has improved dramatically. production start date is used and sometimes the peak
In 1960, the yield of 50% for individual devices production year is stated.
resulted in a 3% yield for a five-transistor IC (today, Shrink versions make the situation more complex: the
integrated microfluidic systems face a similar situation: first functional 1 Gbit DRAMs were demonstrated using
while pumps, valves and mixers may have reasonable 0.18 µm technology, but production versions have been
yields, systems consisting of many such devices have made at smaller linewidths: 0.13 to 0.10 µm. Add to
rather low yields). In the year 2000, 64 Mbit DRAMs this the minor differences between companies, and it is
with some 130 million transistors and capacitors were
fair to accept discrepancies of a few years in Moore’s
manufactured with ca. 90% yields, which translates to
law data.
practically a 100% yield for individual devices, for
Moore’s law was originally proposed in the era of
0.25 µm devices, compared to the ca. 25 µm devices of
bipolar transistors and it has held well in the era of
1960. PMOS, NMOS and CMOS, and it seems to hold for
The early proponents of the IC had to balance the next decade of strained silicon and SOI-CMOS and
between two options:
other evolutionary MOS technologies. Moore’s law is
about device-packing density and cost, and not about
1. Suitable only for price-insensitive applications like any particular technology. There have been a number
military or space technology. of dubious extensions of Moore’s law: it has been
2. Will be cheap in the future once technology matures. said to apply to computing power, which is not true
because computer architecture is not part of Moore’s
Early growth was, of course, along the first argument law. Despite its non-fundamental nature, it is one of the
because somebody had to pay for the chips but at the end few predictions about future technology that has held
of the 1960s the second argument was finally realized, for 40 years.
and the IC became a household term. Linewidth scaling has been very predictable. Junction
depths were scaled with linewidths as L/5 for decades,
38.2 MOORE’S LAW but more recently it has been difficult to scale x j down
as aggressively as linewidth. Gate oxide thickness used
The development of ICs seemed to follow a regular to scale as L/45 for a long time, but with oxide thickness
pattern: doubling the number of devices on the chip now approaching one nanometre EOT, it is not possible