Article by André DeHon appearing in

To date we have relied on the ``Law of Large Numbers'' below the device level
to guarantee deterministic device behavior (*e.g.* dopant ratios,
transition timing, electron state storage). However, at the
nanoscale, we hope to build devices with small numbers of atoms or
molecules (*e.g.* wires which are 3-10 atoms wide, diodes built from
1-10 molecules), and we hope to store state with small numbers of
electrons (*e.g.* 10's). If we are to build devices at these scales,
we will no longer be able to rely on the ``Law of Large Numbers''
**below** the device level. We must, instead, employ the ``Law of Large
Numbers'' **above** the device level in order to obtain predictable
behavior from atomic-scale phenomena which are statistical in nature. At
the same time, the ``Law of Large Numbers'' can also help us by providing
statistical differentiation at scales smaller than those we can pattern
directly or economically using lithography. In this chapter, we examine
various applications of the ``Law of Large Numbers'' above the device level
to build reliable and controllable systems from nanoscale devices and
processes that only have statistically predictable behavior.

- Section 7.6 talks about "architectures with choice" and just briefly mentions that this idea is also applicable to pterm-to-nanowire mapping. That idea is developed in detail and quantified in A Greedy Algorithm for Tolerating Defective Crosspoints in NanoPLA Design (ICFPT2004).