12858 J. Phys. Chem. B, Vol. 106, No. 50, 2002
Letters
suggest that the SiH4 decomposition reaction is nucleated at
surface steps and/or other surface defects because the foils are
far more reactive toward SiH4 than is the flat Ni(111) surface.17
Similarly, on a Ni(977) surface with high step density, Si2H6
will likely decompose preferentially but not exclusively at step
edges. We infer from the above knowledge base that with our
experimental conditions Si2H6 decomposes and most of the
hydrogen is desorbed during annealing. (Note that, because of
the strong interaction between silicon and hydrogen, it is possible
that SiH species exist on the surface in trace quantities even
after annealing. Annealing at a higher temperature is precluded
by dissolution of silicon into the bulk.) After Si2H6 decomposes,
silicon atoms nucleate growth of nanowires. Various stages of
silicon nanowire growth following exposure of Si2H6 are
illustrated in Figure 2. The first panel demonstrates the initial
formation of a single-atom silicon bump on a 7.5 nm × 4.5 nm
template area, with subsequent panels showing the different
growth stages of a silicon nanowire. Since we were interested
specifically in early stage growth, exposures were limited such
that the typical dimensions for these silicon wires were
approximately 1 nm long by 0.25 nm wide; these dimensions
correspond to one atom in width and 4-5 atoms in length (the
atomic radius for silicon is ∼0.12 nm). Specifically, the average
width of the one-atom feature is 0.27 ( 0.05 nm while that for
the two-atom feature is 0.50 ( 0.06 nm. These correspond
therefore, to one and two-atom wide silicon nanowires. Under
higher Si2H6 exposures, the length of these nanowires can be
increased dramatically.
The width of the line, i.e., extension away from the step edges,
is calculated using the method described below (Figure 3a).
STM images measure the surface electron density of states rather
than the atomic geometry directly. Measuring silicon nanowire
width, therefore, requires deconvolution. Clean Ni(977) STM
images show uniform single-steps with widths of 1.65 nm. After
decorating a step edge with silicon, the surface electron density
in that local region changes dramatically. The edge-adsorbed
silicon effectively increases the width of the “upstairs” terrace
and equivalently decreases that of the “downstairs” terrace. It
is this change which is used to estimate the silicon nanowire
width using the following methodology: assume that the actual
silicon nanowire width in a local region is represented by x.
Recall, the total width of the two-step region is 3.3 nm (twice
the single-step width) on a Ni(977) surface with low Si2H6
exposure (Figure 3a). The widths of these two neighboring steps
can be represented by (1.65 nm + x) and (1.65 nm - x). Since
the ratio of these two widths can be obtained from the
experimental images, the silicon nanowire width x can be easily
calculated by solving a one-variable equation.
Figure 4. Different final Si nanowire distributions resulting from
various Si2H6 dosing exposures and rates. (a) 1.63 L exposure in 10
min, (b) 16.3 L exposure in 10 min, and (c) 16.3 L exposure dosage in
20 min. The bright, linear, and aligned regions seen in all three images
are the Si nanowires.
nucleation at the step edges: the experimental exposures we
use are much smaller than the amount needed to fully cover all
of the step edges. Hence, at higher dosing rates, more nucleation
sites are formed compared to the number observed at lower rates.
For a given total exposure less than that needed to fully titrate
the step edges, higher dosing rates will lead to nanowires having
shorter average length but with a larger number of such
structures. We fully expect that using higher Si2H6 exposures
will allow the step edges to be fully decorated and, ultimately,
for the entire terraces to be completely covered by silicon.
After the initial nucleation, the wire grows gradually in two
directions along the step edge (Figure 2b-d). During this
process, its height remains constant; however, its width can
change from one to two atoms (Figure 3b). By increasing the
Si2H6 exposure, a larger area becomes decorated with silicon
nanowires (Figure 4). Interestingly, in this intermediate regime,
the increase in coverage comes with a sacrifice in average wire
length. A 1.63 L Si2H6 exposure with a dosing rate of 0.163
L/min leads to ∼6% silicon coverage, with the average length
of silicon nanowires being 6.9 nm (Figure 4a). If the Si2H6
exposure is increased to 16.3 L, although the total silicon
coverage rises to ∼9%, the average silicon nanowire length
decreases to 3.4 nm (Figure 4b). A higher dosing rate (1.63
L/min), however, leads to shorter nanowires (3.4 nm) com-
pared with that measured (4.6 nm) using a lower dosing rate
(0.815 L/min). This can be explained by the rate of nanowire
STM has been employed to probe the reactive deposition of
silicon nanowires using Si2H6 as the source material. Initial
stages of Si2H6 decomposition on the stepped Ni(977) surface
and the subsequent formation of silicon nanowires have been
discussed. By varying the Si2H6 exposure and dosing rate, it is
possible to form different width and length distributions of the
nanowires. Moreover, recent advances in controlling the perfec-
tion of stepped structures15 suggests that hierarchical templates
of more complex structure may become available for such
purposes. This method of gas-phase reactive deposition on a
stepped metal surface, with the substrate acting as both catalyst
and nanotemplate with easily tunable width (by varying the
vicinal miscut angle) is a fruitful approach to creating massively
parallel nanoscale arrays of highly aligned structures.
Wang, Yi
