COMPUTING WITH QUANTUMDOT CELLULAR AUTOMATA - Bio-Inspired and Nanoscale Integrated Computing - page 143

large area of the tunnel junctions, the devices require operating temperatures
lower than 15mK. To achieve such low temperatures in the experiments, the
devices required cooling via a dilution refrigerator. At these temperatures the
charging energy is much higher than the thermal ambient energy k B T facilitating
the precise control over single electron charging of the metal island. Under the
appropriate bias, a small change in the input can cause single electron transfer
between islands, and this change in charge will cause a similar effect in the
neighboring double-dot. However, due to the precise biasing requirements, each
cell has several external connections and it is therefore difficult to scale this
implementation of QCA to much more complex circuits.
The metal islands that make up the quantum dots are tightly coupled to the
substrate or heat bath. As a result, the time evolution of the cell is not quantum
mechanically coherent and can be modeled using semiclassical descriptions. Unlike
the semiconductor implementation, the metal islands are filled with many conduc-
tion band electrons, and QCA operation is verified by observing the change of
charge rather than the total charge at a particular site. Electron tunneling is the only
nonclassical phenomenon in the metal-island implementation.
4.12.4. Molecular QCA
Developing cells using molecules represents a very promising approach to
implementing QCA. Molecular QCA has the potential for room temperature
operation, high device density, and high operating speed [27]. Room temperature
operation is possible as a result of the high kink energies of molecular cells [53]. In
theory, molecular implementations could also be fabricated with much higher
uniformity than those achievable with semiconductor or metallic-island imple-
mentations [31]. Estimates using 1 nm 2 QCA cells indicate that molecular QCA
could reach densities of 10 14 devices/cm 2 [29]. However, one has to be careful
when making a direct comparison of the functional capability of a QCA circuit
with this density to a transistor circuit of the same density since the contribution of
each QCA cell to the overall functionality is far lower than that of a transistor
circuit as a majority of the cells are consumed in the interconnect network between
logic gates. Although the cell density can be very high, recent research on QCA
clocking is progressing towards relatively low density circuits as a result of the
large area consumed in the crossovers [31]. Quantum mechanical analysis of QCA
dynamics has predicted cell switching times as low as 2 ps per cell resulting in the
possibility of THz operation [51]. It is not clear if the clocking network itself could
operate at such high speeds and the limits to QCA operating speed could end up
being determined by the clocking implementation rather than the cell dynamics.
Initial research into molecular QCA has focused on bistable mixed valence
compounds similar to those first developed by Aviram [83] and Hush et al. [84].
These molecules consist of multiple redox centers, where each center can be
reduced (gain an electron) or can be oxidized (lose an electron) [32]. The Aviram
molecule consists of two allyl groups connected by a butyl bridging ligand. One of
the allyl groups is a neutral radical, while the other is a cation. The unpaired