Information Technology Reference
In-Depth Information
provides one energy minimum, wherein the contact resistance between the two
layers is zero, and the switches (the contacts between the two sets of ortho-
gonal nanowires) are ''off''. When the tubes are transiently charged to produce
attractive electrostatic forces, the suspended tubes flex to meet the tubes
directly below them, and a contact is made, representing the ''on'' state. The
''on''/''off '' state could be read by measuring the resistance at each junction, and
could be switched by applying voltage pulses at the correct electrodes. This
theory was tested by mechanically placing two sets of nanotube bundles in a
crossed mode and measuring the I(V) characteristics when the switch was ''off '' or
''on.'' Based on this nonvolatile RAM design, Zhang and Jha proposed a
reconfigurable architecture that can achieve much better performance,
reconfigurability, and density than existing field-programmable gate arrays
(FPGA) [70].
While Lieber employed the SWNTs themselves as switches, Stoddart and
Heath employed nanoimprinting and the SNAP method to fabricate large and fine
nanowire crossbar arrays [32, 68]. These large and fine crossbar arrays (400 by 400
array and 33 nm pitch) have a monolayer of bistable [2]rotaxane molecules
between the two orthogonal arrays [28, 29]. The monolayer is formed using the
LB process. Stoddard and Heath intended the 400 by 400 crossbar array as a high-
density memory unit. The promise of extremely high density memory units by
nanowire crossbar arrays has also inspired sophisticated circuit and architecture
designs based on nanowire crossbar memory units [71, 72]. However, none of
these designs have been realized.
11.3.2.3. Challenges. The crossbar array architecture faces several hurdles
before it can be useful. First, addressing an individual nanowire or crosspoint is
difficult due to their miniature size. For example, although fine and large crossbar
arrays have been demonstrated [32, 68], it is still difficult to address an individual
crosspoint or make a contact with a single nanowire. For example, in [32],
electrode contacts for crosspoint characterization can only be lithographically
made to two to four adjacent nanowires. As a result, a small array of crosspoints
had to be measured to infer the properties of a single crosspoint as a memory unit.
The fundamental challenge is to select a single nanowire in a dense array through
microscale metal wires. Most proposed solutions rely on lithographically over-
laying an orthogonal array of microscale metal wire on top of the nanowire array
to build a decoder or demultiplexer.
Williams and Kuekes patented a proposal that randomly deposits gold
particles between the metal wire array and the nanowire array so that a random
set of crosspoints becomes conductive when gold particles are deposited therein
[73]. DeHon et al. employed modulation-doped nanowires to interface with
microscale metal wires [74]. A modulation-doped nanowire has different dopings
along its length, which can be controlled during growth. On the contrary, Savage
et al. employed axial-doped nanowires, which have different dopings along the
radius [75]. Both solutions seek to distinguish a nanowire from its neighbors in an
array based on its unique doping heterogeneity sensed by the microscale array of
 
Search WWH ::




Custom Search