AN INTRODUCTION TO NANOCOMPUTING - Bio-Inspired and Nanoscale Integrated Computing - page 6

behavior represents a simple switch: the gate can allow or prevent current from
flowing through the wire. This is the switch abstraction. In practice, there are many
more abstractions placed on top of these two (for example, representing integers in
binary form with a series of 0's and 1's). However, the digital and switch
abstractions are particularly significant because they bridge between a physical
phenomenon, moving electrons and electric fields, to an entirely abstract world,
manipulating 0's and 1's with switches. This use of transistors is the cornerstone of
modern computing.
One particularly interesting achievement occurred in 1959, when both Robert
Noyce and Jack Kilby independently developed the integrated circuit. With
integrated circuits, one fabrication process simultaneously creates many transis-
tors, all of them integrated on a single crystalline structure such as silicon. As
fabrication techniques began to improve, it became possible to pack more
transistors together. By 1965 Gordon Moore, the co-founder of Intel, predicted
that the number of transistors that fit into a given area would double every 18
months due to continued improvements in the fabrication process. Following this
prediction known as Moore's Law, transistor size, speed, and power consumption
have exponentially improved for almost 50 years. Today it is possible to construct
hundreds of millions, even billions, of tiny transistors on a small piece of silicon
the size of a thumbnail (Fig. 1.4). In turn, it has become practical to create abstract
computers that use millions or billions of switches.
Because the fabrication process produces all transistors simultaneously, the cost
of fabricating these computers is largely independent of the number of transistors.
There is typically a large initial cost, and this initial cost can be amortized over
thousands or millions of processors, which can be produced cheaply. The economics
of this situation is staggering—with a smaller transistor, performance improves,
power consumption decreases, more abstract computation fits onto a single
processor, and all this happens as the price of each transistor decreases! With this
persistent exponential improvement, it is very easy to manipulate large amounts of
abstract information, and computers are used for a prolific number of applications
today. All of this has hinged on the fact that transistors continue to get smaller, and
this has led to the general trend that ''smaller is better.''
1.3.1. Difficulties with Transistors at the Nanometer Scale
Transistor sizes are already at the nanometer scale, and this causes many practical
difficulties. At the time of publication of this topic, many consumer products are
using a 45 nm fabrication process, and 32 nm technology has already been
demonstrated. At these small sizes, fundamental limitations have to be considered.
Entire topics have been written on the subject, and here we describe only a few
such challenges.
One primary example of these difficulties is a quantum phenomenon known as
tunneling, visualized in Figure 1.5. Due to the wave nature of particles, electrons
can ''jump,'' or tunnel, through barriers with some nonzero probability. This
probability increases exponentially as the size of the barrier decreases. The size of