MOLECULAR COMPUTING: INTEGRATION OF MOLECULES FOR NANOCOMPUTING - Bio-Inspired and Nanoscale Integrated Computing - page 310

metal wires. Like the gold particle-based solution, they are inherently probabilistic
due to the inability to distinguish nanowires with different doping patterns during
assembly or accurately position any given nanowire. Therefore, all three solutions
can lead to low utilization of the nanowire array. So far fabrication has not yet
been reported for any of these structures.
Lieber fabricated a small decoder based on chemically modifying lithogra-
phically selected crosspoints in the nanowire crossbar array [76]. The input
nanowires interface with microscale metal wires through lithographically made
contacts. Apparently, the decoder is limited by lithography. It does not really
allow the selection of a nanowire from an array with a pitch beyond the resolution
of lithography. Heath later demonstrated a decoder that selects a nanowire
from an array of 150 silicon nanowires with a pitch of 34 nm [48]. The design is
based on forming a binary tree pattern in dielectric between the microscale
metal array and the nanowire array. The finest pattern is determined by the pitch
of the nanowire array and hence e-beam lithography has to be employed. So far,
no truly nanoscale addressing solution has yet enjoyed cost-effective, scalable
fabrication.
Furthermore, signal strength degrades as it travels along the nanowires. Gain
is typically introduced into circuits by the use of active devices, such as transistors.
However, placing a transistor at each crosspoint is an untenable solution because
doing so will eliminate the size advantage of the molecular-based system. Like-
wise, in the absence of a transistor at each cross point in the crossbar array,
molecules with very large ''on'' : ''off '' ratios will needed. For instance, if a switch
with a 10:1 ''on'' : ''off '' ratio was used, then 10 switches in the ''off '' state would
appear as an ''on'' switch. Hence, isolation of the signal via a transistor is
essential, but presently the only solution for the transistors' introduction would be
for a large solid-state gate below each cross point, again defeating the purpose for
the small molecules.
Additionally, if SWNTs are to be used as the crossbars, connection of
molecular switches via covalent bonds introduces sp 3 -hybridized carbon atom
linkages at each junction, disturbing the electronic nature of the SWNT and
possibly obviating the very reason to use the SWNTs in the first place.
Noncovalent bonding of the device molecule to the SWNT will probably not
provide the conductance necessary for the circuit to operate. Therefore, continued
work is being done to devise and construct crossbar architectures that address
these challenges.
11.4. MEET-IN-BETWEEN PARADIGM
The traditional silicon MOSFET fabrication is top-down, relying on lithography
to accurately position each individual transistor, as specified by design. However,
due to process variations, such a what-you-design-is-what-you-get assumption is
no longer true for present deep submicron MOSFET fabrication, let alone truly
nanoscale structures such as silicon nanowires. It is extremely hard, if at all