DIELECTROPHORETIC ARCHITECTURES - Bio-Inspired and Nanoscale Integrated Computing - page 160

(99.7%) and B 2 H 6 dopant (0.3%) in He carrier gas (100 ppm) at 450 torr and
450 C. Growth was performed for 10-60min to achieve nanowire lengths of

10-60 m m. The nanowire growth wafer was sonicated lightly in isopropanol for
1min. The suspension was vacuum filtered using a 12- m m mesh (Millipore
Isopore) in order to remove unnucleated Au catalyst particles and short nano-
wires. The filter mesh was sonicated in isopropanol, and the suspension was again
filtered. The second filter mesh was sonicated in benzyl alcohol for 2min and the
suspension was used for trapping experiments. Benzyl alcohol was selected as a
viscous, low-vapor-pressure solvent [23] for reconfiguration in order to damp
Brownian motion, minimize toxicity [24], and allow ambient operation. Addi-
tionally, its static relative permittivity is slightly smaller than that of bulk Si (11.9
versus 12.1, respectively) [23], reducing van der Waals interactions at low
frequencies and favouring dielectrophoretic trapping of conductive structures
[6]. Heavily doped silicon nanowires were selected as interconnects to demonstrate
potential compatibility of our technique with the assembly of more complex
semiconducting nanostructures, such as axial heterostructures [25].
Trapping experiments were performed with 100-nm Au electrodes (5 nm Cr
wetting layer) to avoid oxidative damage, on a Si wafer with a 200-nm oxide to
prevent shorts. Thicker electrodes, with reduced fringing fields, were found to
better allow nanowires to migrate along their edges toward the trapping region.
Thinner electrodes tended to permanently pin nanowires to the top electrode faces
wherever they were first

1

trapped. The electrodes were defined by e-beam

lithography with a 10 taper angle and a 1- m m tip radius of curvature.

The nanowire suspension was pipetted onto the electrode chip to form a
250- m m-thick reservoir, as shown schematically in Figure 5.4a. For trapping,
electrode pairs were biased at 10 kHz to minimize both solvent electrolysis and
parasitic capacitance. The bias was modulated into 10-ms bursts at 110 V RMS with
a period of 100ms, which allowed migration of nanowires toward the trapping
region in controlled steps. The time between bursts was manually increased to
1000ms as nanowires approached the inter-electrode region, and the bursts were
halted when the desired number of nanowires had been trapped. Movies of
nanowire motion were recorded at 4 fps.

1

Figure 5.4. Dielectrophoretically trapped nanowires. (a) Schematic illustration of
nanowire trapping process. (b) Light microscope image of multiple nanowires
stably trapped between electrodes separated by 40 mm. Scale bar is 40 mm.