JOURNAL OF APPLIED PHYSICS 106, 044311 2009

Catalyst proximity effects on the growth rate of Si nanowires

S. T. Boles,1,a E. A. Fitzgerald,1 C. V. Thompson,1 C. K. F. Ho,2 and K. L. Pey2
1
Department of Materials Science and Engineering, Massachusetts Institute of Technology, 77 Massachusetts
Ave., Cambridge, Massachusetts 02139, USA
2
School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798,
Singapore
Received 3 July 2009; accepted 22 July 2009; published online 31 August 2009
Si nanowires grown by the vapor-liquid-solid VLS mechanism were fabricated using Au-catalyst
nanoparticles and silane SiH4 gas on Si substrates. Au was deposited on the substrate surface both
by electron-beam evaporation and Au-colloid deposition. Both kinking defects and vertical
nanowire epitaxy on Si 111 substrates were found to be directly related to SiH4 flow rate. A
correlation between Au-colloid dilution and the nanowire growth rate was also observed, with the
growth rate increasing with increasing concentrations of Au-catalyst particles on the wafer surface.
Systematic experiments relating the nanowire growth rate to the proximity of nearest-neighbor Au
particles and Au reservoirs were carried out, and the results were found to be in good agreement
with a SiH4 reaction model, which associates decomposition to form SiH2 with higher nanowire
growth rates. Implications toward the realization of VLS-grown single nanowire transistors are
discussed. 2009 American Institute of Physics. DOI: 10.1063/1.3207821

I. INTRODUCTION II. EXPERIMENTS AND RESULTS

Silicon nanowires have been studied extensively in the
past decade for a host of potential applications including
sensors,1 transistors,24 and battery anodes.5,6 Despite the
great deal of interest in the properties and performance of
single-wire devices, many aspects of nanowire integration
into larger scale systems are still unclear. While a top-down
approach is essentially limited by lithography, the vapor-
liquid-solid VLS mechanism7 for Si nanowire growth
should theoretically be limited only by catalyst placement. A
variety of techniques focused on the manipulation of Au-
catalyst particles for VLS growth have been explored. Some
groups have pursued densely packed vertical wire arrays8,9 as
a means of leveraging the properties of individual nanowires,
while others have undertaken more complex silicon-on-
insulator configurations to incorporate wires into functional
systems.1012
This work is focused primarily on the relationship be-
tween Au-catalyst density and the Si nanowire growth rate.
The ability to control the precise placement of the Au-
catalyst particles and subsequent nanowire growth is funda-
mental to future nanowire integration. Borgstrm et al.13
suggested that the rate of III-V nanowire growth is affected
by interparticle distance. Our preliminary experiments also
show that isolated catalysts grow at very slow rates. Through
a combination of reaction modeling and experimental design,
we have identified a fundamental aspect of growth of Si
nanowires using the VLS mechanism and a silane precursor
in a cold-wall chemical vapor deposition CVD system,
where the precursor decomposition is selective to the catalyst
surface. This observation has implications for device integra-
tion in the wide range of applications for which nanowires
are potentially useful.
a
Electronic mail: steven.boles@gmail.com.
Gold Au catalyst nanoparticles used in our experiments
were made through two routes: 1 e-beam deposition of dis-
continuous Au films type 1 and 2 application and drying
of Au-colloid solutions type 2. In order to optimize the
recipe for vertical nanowire growth for our CVD chamber,
we have used samples of type 1. Processing of type 1 can be
easily controlled to yield an oxide-free Au-catalyst template
whereas the drying of Au-colloid solutions can leave residue
of organics and oxide on the Au catalysts. Samples approxi-
mately 1 2 cm2 were cut from lightly doped, n-type
0.01 cm, As Si111 wafers. The wafer pieces were
dipped in 50:1 HF: H2O for 20 s to remove the native oxide
and leave a hydrogen terminated surface. The samples for Au
thin-film deposition were then immediately loaded into a
Perkin Elmer e-beam evaporator and pumped down to 5
109 torr. Using a quartz crystal monitor for control, 0.6
nm of Au was deposited at 0.1 nm/min at room temperature.
Au colloids were deposited by mixing various dilutions of
H2O with a 50 nm aqueous Au-colloid solution from Ted
Pella Inc. Droplets of the solutions were placed in the center
of each sample piece immediately purged with N2. This pro-
cess ensures both a clean Au-colloidSi interface for epitaxy
and simultaneously mitigates potential contamination from a
poly-l-lysine adhesion layer.14 Samples with Au pads serving
as reservoirs were made on lightly doped, n-type
0.01 cm, As Si100 wafer pieces. Photolithography,
e-beam evaporation, and lift-off were used to create patterns
of pads. Au-colloid particles were deposited on these
samples in the same way described previously.
Samples of both type 1 and type 2 were subjected to a
second HF dip immediately before growth to ensure the re-
moval of the silicon-oxide shell covering the metal catalyst.
Omission of this step can result in minimal or no nanowire
growth, as seen in Fig. 1.

0021-8979/2009/1064/044311/9/$25.00 106, 044311-1 2009 American Institute of Physics

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-2 Boles et al. J. Appl. Phys. 106, 044311 2009

FIG. 2. Top-down SEM image of Si nanowires grown on a Si 111 sub-

strate using an e-beam evaporated Au catalyst. The silane flow rate in this
FIG. 1. Tilted SEM image of Si nanowires grown on a Si 111 substrate
case was 15 SCCM, which results in a large number of kinking defects.
using an e-beam evaporated Au catalyst. The native oxide in this case was
Scale bar is 500 nm.
not removed prior to growth, preventing smaller Au particles from facilitat-
ing wire growth. Scale bar is 500 nm.
Following success in establishing controlled vertical
Following oxide removal, the samples were loaded into nanowire epitaxy using e-beam films, Au-colloid solutions
a Thomas Swan atmospheric-pressure metal-organic chemi- were dispersed onto Si111 wafers for nanowire growth. The
cal vapor deposition MOCVD reactor. This reactor has a samples were loaded into the reactor along with a piece of
e-beam Au on Si111 as a control. Nanowire growth was
custom designed quartz tube to ensure laminar flow of feed
carried out to give the best vertical epitaxy for e-beam Au
gas over a graphite susceptor and is traditionally used for
films on Si111 using the same recipe as the one used to
III-V thin film growth. However, the double-dilution silane
produce Fig. 4. The contrast between the wires grown from
dopant line allows us to grow Si nanowires with a wide
Au colloids and e-beam films is clear Figs. 5a and 5b.
range of dilutions 200.0002 SCCM SCCM denotes cubic
The wires from the Au-colloid sample appear to be both full
centimeter per minute at STP. The samples were purged
of kinking-type defects and are significantly shorter than the
first in 15 lpm of H2 for a few minutes before the reactor
wires grown from e-beam films.
temperature was brought up to 700 C where it was held for
An additional run was done to test Si nanowire growth
1 min. The temperature was then lowered to 550 C for Si
from Au-colloid particles. In this run, a drop of undiluted
nanowire growth. This two-step pregrowth anneal and
Au-colloid solution was dispersed onto a piece of Si111
growth method has been adopted by several others and has
and annealed in N2 at 600 C for 10 min. The samples were
been shown to improve nanowire epitaxy despite the fact that
then HF dipped and loaded into the reactor. The sample was
the exact cause is not certain.15,16 The growth temperature of
first annealed in H2 at 700 C for 10 min before being
550 C had been previously determined to be optimal for
brought to 550 C. The silane flow was increased for this
VLS silicon nanowire growth in our system. At temperatures
above 550 C a noticeable layer of a-Si can be seen on the
quartz tube of the reactor, suggesting that amorphous silicon
a-Si is also likely to be depositing on the Si substrates. No
Si nanowire growth was observed at temperatures below
500 C.
The flow rate of SiH4 was optimized to minimize nano-
wire kinking. Setting the SiH4 flow to 15 SCCM and higher
resulted in heavily kinked nanowires as seen in Fig. 2. Dial-
ing back the flow rate below 15 SCCM eliminated the kink-
ing phenomenon. However, the wires still did not exhibit
strict vertical epitaxy on Si111. Instead, the silicon nano-
wires appear to grow in each of the four 111 directions
one normal 111 and three 111 out of plane orientations
with respect to the substrate as seen in Fig. 3. Further re-
duction in the SiH4 flow rate from 10 to 0.18 SCCM was
found to dramatically increase nanowire growth-direction se-
lectivity Fig. 4. The correlation between the decrease in FIG. 3. Tilted SEM image of Si nanowires grown on a Si 111 substrate
using a 0.6-nm-thick e-beam evaporated Au catalyst. The flow rate of silane
silane partial pressure in the reactor and the corresponding was 0.50 SCCM. Note that although the majority of the wires grow in the
increase in vertical nanowire epitaxy is consistent with ear- direction of the 111 normal to the substrate, a significant number of wires
lier findings.17,18 grow in other 111 directions. Scale bar is 500 nm.

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-3 Boles et al.
FIG. 4. Tilted SEM image of Si nanowires grown on a Si 111 substrate
using a 0.6-nm-thick e-beam evaporated Au catalyst. The growth conditions
here are identical to those of Fig. 3, except that the flow rate of silane is
0.18 SCCM. This difference appears to significantly improve vertical
nanowire epitaxy. Scale bar is 500 nm.
growth from to 0.18 SCCM to 10.0 SCCM to eliminate any
unforeseen artifacts related to diluted silane flow.
After growth the sample was inspected both at the edge
of the Au-colloid solution droplet and in the center region of
the droplet. Images from the two regions are seen in Figs.
6a and 6b. The average length of nanowires near the
droplet edge is approximately 6.8 m, whereas the aver-
age length of wires from the center region is only 1.0 m.
Wire growth was inspected around the circumference of the
droplet and found to be fairly consistent. This helps rule out
the possibility that precursor depletion associated with flow
across the sample is affecting our results. The number of
wires per m2 is significantly greater in the region near the
edge of the droplet, where the Au particle density is higher.
This difference in Au-colloid particle density is due to the
coffee-stain effect where particles in solution preferentially
decorate the outer edge of a liquid droplet as it dries.19 With
everything else held constant in this run, the increase in the
growth rate of wires at the outer edge of the droplet can be
directly correlated with the increase in density of Au-colloid
particles. The link between Au particle density on the surface
FIG. 5. Tilted SEM images of Si nanowires grown from e-beam evaporated
Au catalysts a and Au-colloid solutions b on Si 111 substrates. The
wires grown in a grow predominately in the 111 direction of the sub-
strate and are approximately 500 nm in length. The wires in b exhibit poor
epitaxy, as well as poor directionality, and vary between 50 and 500 nm in
total length. Scale bars are 500 nm.
Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
J. Appl. Phys. 106, 044311 2009
FIG. 6. Cross-sectional SEM images of Si nanowires grown from Au-
colloid solutions. Image a was taken near the perimeter of the original
droplet where the Au-colloid concentration is relatively high. Image b was
taken in the region near the center of the original Au-colloid droplet. The
Au-colloid concentration here is significantly lower. Note that the length of
the wires in a are considerably longer than the wires in b. Scale bars are
2.0 m.
of the sample and the growth rate of Si nanowires suggests
that a synergistic effect between neighboring wires is
present, as is the case for III-V wires as shown by Borgstrm
et al.13
A series of growths was carried out with varying dilu-
tions of Au colloids to explore the relationship between the
Au particle spacing and the Si nanowire growth rate. Dilu-
tions of 0:1, 5:1, 50:1, and 100:1 were made by mixing
deionized DI water and Au-colloid solution. The Au col-
loids in this case were dispersed onto Si100 and Si111
pieces and dried in flowing N2 so as to prevent the coffee-
stain pattern and give a more consistent spacing between
particles. With a high rate of evaporation of the Au-colloid
solution on the Si substrate, there is an inadequate amount of
time to allow the nanoparticles to be driven toward the con-
tact line by capillarity effects.19 Silicon nanowires were then
grown from these particles in the same manner as discussed
previously with the initial anneal in N2 at 600 C from the
previous growths omitted.
A series of scanning electron microscope SEM images
was made at multiple locations on each sample. The numbers
of Au-catalyst particles per unit area were counted and the
lengths of the corresponding Si nanowires were measured
and corrected for tilt and the 111 growth direction on the
Si100 substrates. A plot of nanowire growth rate versus
interparticle distance is shown in Fig. 7. The overall shape of
the plot indicates a significant increase in the growth rate of
Si nanowires around a spacing of 1.0 m between neighbor-
ing particles.
Samples with Au-colloid particles interspersed among
large patterned Au reservoirs were fabricated to understand
the role of exposed Au surfaces as they relate to nanowire
growth. Photolithography, e-beam evaporation, and lift-off
were used to create Au patterns Fig. 8 of 30, 60, and 120
nm thicknesses on Si100. Growth on these samples was
carried out using the method described above where the
samples were annealed for 10 min at 700 C followed by 90
min of growth at 10 SCCM of silane.
A SEM image series starting at 500 m away from the
pattern edge to 100 m between the Au reservoirs is shown
in Fig. 9a. Representative images from this series are
044311-4 Boles et al. J. Appl. Phys. 106, 044311 2009

FIG. 7. Plot comparing the Si nanowire growth rate vs the average interpar-
ticle spacing. Error bars in the y-direction were determined from the stan-
dard deviation of wire lengths in each image taken at the corresponding
spacing.

shown in Fig. 9b. The length of each nanowire in the field

of view of each image was subsequently measured and cor-
rected for tilt. A summary of the nanowire lengths and the
respective distance from the Au reservoirs is shown in Fig.
10.
The shape of the curve in Fig. 10 is very similar to that
of the curve in Fig. 7. However, the relevant length scales in
this case are much greater than with Au-colloid particles
without nearby reservoirs. From the data it appears that Au
reservoirs as far as 300 m away from the Au-colloid par-
ticles may be affecting the growth of Si nanowires.
The nanowires growing in proximity to 120 nm pads
seem to exhibit a slightly higher growth rate than the wires
growing the same distance from 30 nm thick pads, even
though the relative particle density is comparable. This sug-
gests that the 10% increase in surface area of the 120 nm
pads can have an effect on the growth of nanowires in the
near vicinity.

FIG. 9. Image a shows the line along which images bg were taken.
Images bg show the change in wire length as the colloidal particles are
progressively further from the reservoirs. Image b was taken between the
Au reservoirs and g was taken 200 m away from the Au reservoirs.

III. MODEL
FIG. 8. Top-down SEM image of 60-nm-thick Au pads that serve as Au
reservoirs after Si nanowire growth. Au-colloid particles were dispersed
To explain the correlation between enhanced growth rate
evenly across the surface of the wafer. Note that the reservoirs toward the
center of the structure appear to have enhanced wire growth compared to the and the local concentration of Au, we consider a model for
reservoirs along the outer edge of the structure. Scale bar is 50 m. SiH4 decomposition and subsequent wire growth. First, we

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-5 Boles et al. J. Appl. Phys. 106, 044311 2009

r2H = f 2H 2H , 10
2 2 2

where rx represents the rate of each reaction and f x is the

reaction rate constant for reaction x. In addition, we recog-
nize that in the steady state the growth rate, or Eq. 9 in this
case, must be equal to Eq. 10, since neither can exceed the
other Eq. 11. In addition, the adsorption rate of silane
Eq. 7 must be equal to the sum of Eqs. 8 and 9, giving
Eq. 12.
r1 = r2H , 11
2

rSiH4 = r1 + rSiH . 12
4

Using Eqs. 612 we find the following form for silicon

nanowire growth rate in the steady state:
f r1 f SiH4 pSiH4

FIG. 10. Color online Plot of the nanowire growth rate vs distance from r1 = . 13
Au reservoirs. Each color indicates a different set of Au reservoir thickness f r1
and colloid dilution. Negative values in distance indicate that the wires are f SiH + f r1 + 1 + f SiH4 pSiH4
4 f 2H
growing between the channels of the Au reservoirs. Pink triangles30-nm- 2
thick Au, 50:1 dilution; black squares60-nm-thick Au, 0:1 dilution; light
blue diamonds30-nm-thick AuPd, 0:1 dilution; red circles30-nm-thick Previous experiments not presented here have shown a lack
Au, 0:1 dilution; dark blue triangles120-nm-thick Au, 0:1 dilution; and of a dependence of the growth rate on temperature or silane
green triangles120-nm-thick Au, 0:1 dilution. flow rate. This contradicts the growth rate dependencies in
Eq. 13. Increasing the Au concentration could, in principle,
consider a simple reaction for silane decomposition on the affect the probability in the reaction rate constant prefactor
catalyst surface.20 This can be described by Eq. 1 if we of the silane adsorption term, f SiH . However, this only is
4
assume direct decomposition of silane to form Si on the sur- relevant in cases where pSiH4 is small and thus we would also
face of the catalyst. This makes the silicon nanowire growth expect a dependence on pSiH4,
rate directly proportional to rate of reaction
f r1 f SiH4 pSiH4
SiH4g Sis + 2H2g. 1 r1 . 14
f SiH + f r1
4
Next we consider the physical process behind this reaction
leading to Si nanowire growth. Reaction 1 can be broken We do not find this in experiments.
down into four steps: In the case where more Au in the vicinity of a growing
nanowire might increase the prefactor of the Si incorporation
SiH4g SiH4 , 2 reaction, f r1, we find a dependence that should theoretically
asymptote as the probability of this reaction occurring is in-
SiH4 SiH4g, 3 creased,
f r1 f 2H
SiH4 Sis + 2H2 , 4 r1 2
. 15
f 2H + f r1
2
2H2 2H2g, 5
This is also negated by the data in Figs. 7 and 10 as the
where an asterisk is used to indicate a surface-adsorbed spe- growth rate continues to increase as the amount of Au in the
cies. Given a limited number of equivalent surface sites near vicinity is increased.
available for each reaction to take place, we denote each type The growth regimes allowed in the model described by
of site independently as a fraction of the total sites available. Eqs. 1 and 13 do not seem to be supported by the experi-
mental evidence from Si nanowire growth using Au-colloid
empty + SiH + 2H = 1. 6 particles with and without relatively large volumes of pat-
4 2
terned Au, or reservoirs, nearby. To understand this, a more
Assuming first order reaction rates for Eqs. 25, we de-
complex model was then evaluated for which it was assumed
fine the following:
that an initial decomposition step from silane to silylene
rSiH4 = f SiH4 pSiH4empty , 7 SiH2, shown as follows, is followed by silylene reaction to
form adsorbed Si and wire growth:
rSiH = f SiH SiH , 8 SiH4g SiH2g + H2g Sis + 2H2g. 16
4 4 4

This is often assumed to be the likely decomposition path for

r1 = f r1SiH , 9
4 silane.21,22

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-6 Boles et al. J. Appl. Phys. 106, 044311 2009

Similarly to reaction 1, reaction 16 can be expanded into r1 = f r1SiH , 28

4
eight steps:
SiH4g SiH4 , 17 rH = f H H , 29
21 21 21

SiH4 SiH4g, 18 rSiH2 = f SiH2 pSiH2empty , 30

SiH4 SiH2 + H2 , 19 rSiH = f SiH SiH , 31

1
2 2 2

H2 H2g, 20 r2 = f r2SiH , 32
1
2

SiH2g SiH2 , 21 rH = f H H . 33
22 22 22

SiH2 SiH2g, 22 Note that although silylene is only produced by reaction

19, both silane and silylene must compete for the same
SiH2 Sis + H2 , 23 vacant surface sites. It is also assumed here that H2 does not
2
readsorb to the surface once it is in the gas phase.
H2 H2g. 24 In the steady state, the following conditions must hold:
2

Again, the surface sites are defined by the fraction of the r1 = rH , 34

21
total number of sites occupied by the various adsorbed spe-
cies r2 = rH , 35
22
empty + SiH + SiH + H + H = 1. 25
4 2 21 22
rSiH4 = r1 + rSiH , 36
4
The reaction rates are defined by their respective reaction
rate constants and site availability. rSiH2 = r2 + rSiH . 37
2
rSiH4 = f SiH4 pSiH4empty , 26
Solving the system of Eqs. 2537 for r2 in terms of the
reaction rate constants and partial pressures of the reactants
rSiH = f SiH SiH , 27
4 4 4 yields

1
r2 = . 38

f SiH4

1+
f SiH
1 1 f r1
2
1+ f SiH4 + + pSiH4
1 1 f r2 f H f r2 f SiH4 pSiH4

21
+ +
f H f r2 f SiH
22 4
f SiH2 1 + pSiH2
f r1

In this analysis it is assumed that some fraction of the incoming silane will be converted to silylene. This fraction remains
relatively constant during nanowire growth for a given condition in steady state. We introduce a parameter, , which represents
fraction

pSiH4
= . 39
pSiH2

Substituting Eq. 39 into Eq. 38 gives

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-7 Boles et al. J. Appl. Phys. 106, 044311 2009

1
r2 = . 40

f SiH

4
1+
f SiH
1 1 f r1
1+ 2
f SiH4 + +
1 1 f r2 f H f r2 f SiH4 pSiH4

21
+ +
f H f r2 f SiH
22 4
f SiH2 1 +
f r1

Note that when = 1, all silane is converted into silylene. should decrease as well. This is in good agreement with the
When = , silylene production becomes the bottleneck and experimental results of Figs. 7 and 10 above.
r2 is quenched, which stops Si nanowire growth. The form of r2 in Eq. 40 follows a reciprocal function
with respect to .

IV. DISCUSSION x X
r2 = . 41
y 1+Y
To examine the form of r2, a term proportional to the 1+
pSiH4
wire growth rate, all reaction rate constants were set to a
fixed value and pSiH4 and were chosen as input variables Using X and Y as fitting parameters Eq. 41, we fit a curve
that describe the gas concentration within the boundary layer. to the data in Fig. 7, yielding values of X and Y of 95 and
A plot of r2 versus pSiH4 for several values of is shown in 1.53, respectively. The Y-value should be dependent on a
Fig. 11. Clearly r2 has little dependence on pSiH4 except at number of factors including many of the reaction rate con-
extremely low values of pSiH4. This is supported by previous stants; thus its numerical value cannot be calculated. How-
experiments where an increase in silane flow did not increase ever, the value of 95 nm/h for X should simply describe the
the growth rate of silicon nanowires. Next, r2 is plotted as a shape of the curve and not the horizontal shift. Next, an
function of for several values of pSiH4, as seen in Fig. 12. empirical value of Y was determined for our growth condi-
As is increased from 1 to the nanowire growth rate drops tions and parameters from the e-beam thin film samples. For
off tremendously. For a fixed value of approximately this determination, is assumed to be equal to 1, pSiH4 is
40 the nanowire growth rate is insignificantly increased as 1.2 105, and r2 is approximately 500 nm/h. Solving for Y
pSiH4 is increased. in this case gives us a value of 1.428 105. This allows us
Lastly, we consider the connection between our experi- to extrapolate values of for the curve in Fig. 13 by using
mental results and the variable from Eq. 39. The growth the pSiH4 value of 6.66 104 and the X- and Y-values above.
conditions for the experiments were in a regime where the A table and plot of average interparticle spacing, , and 1 /
growth of amorphous and poly-Si is minimized while the are shown in Figs. 14 and 15, respectively.
growth rate of VLS Si is maximized. In order to do this, a The approximate values of 1 / suggest that there is con-
temperature was chosen where the only silane decomposition siderable room for enhancing the growth rate of silicon nano-
that takes place is on Au surfaces. If reaction 17 can only wires using a more reactive gas species, such as SiH2 or
occur on a Au surface, then an increase in the amount of Si2H6. This is consistent with the findings of Ref. 23, in
exposed Au surface area should decrease . In other words,
as the average distance between Au islands is decreased, G rowth R ate
A.U.

0.04
G rowth R ate
A.U.
0.25

Increas ing 0.03

0.20 Increas ing P S iH4

0.15 0.02

0.10
0.01

0.05

20 40 60 80 100
P S iH 4
A.U.
20 40 60 80 100
FIG. 12. Color online Plot of r2 vs for several values of pSiH4 ranging
FIG. 11. Color online Plot of r2 vs pSiH4 for several values of , ranging from 0.1 to 100. As is decreased, the growth rate is enhanced dramatically.
from 1 to 100. The growth rate quickly asymptotes to a fixed value regard- An order of magnitude difference in pSiH4 can be negated by increasing by
less of pSiH4 for a fixed value of . a factor of 2 or less in the low pSiH4 regime.

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
044311-8 Boles et al.
FIG. 13. Color online Data from Fig. 7 fit with a curve of the form given
by Eq. 41. The value a = 0.095 m / h is used to estimate values of and Figs. 13 and 14. 1 / is the approximate percentage of SiH4, which is con-
1 / Fig. 14.
which it is demonstrated that extremely high Si nanowire
growth rates are obtained using a Si2H6 precursor. To further
substantiate this idea, in our reactor we are able to produce
growth rates that are more than two orders of magnitude
faster for Au catalysts that are dispersed sparsely i.e., inter-
particle distance 1 m when an overhanging perforated
gold mesh is used. With an increased 1 / afforded by the
proximity structure as shown in Fig. 12, the growth rate can
be controlled regardless of the interparticle distance.
Furthermore, the strongly suppressed growth rates of Si
nanowires from highly dilute Au-colloid solutions may im-
pede the ability to create low density Si nanowires for dis-
crete transistor or sensor applications. Many of the potential
applications including single nanowire transistors and sen-
sors may require relatively low nanowire densities for actual
devices. The ability to grow the nanowire component in
FIG. 14. Plot of vs average interparticle spacing for the data in Fig. 13. As
the average interparticle spacing increases, also increases. Figure 12 indi-
cates that regardless of pSiH4, will dominate in determining the nanowire
growth rate.
Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
J. Appl. Phys. 106, 044311 2009
FIG. 15. Plot of 1 / vs average interparticle spacing taken from the data in
verted to SiH2 during steady state nanowire growth. According to this plot,
only a very small fraction of incoming silane is being converted to silylene.
This accounts for the suppressed growth rate of Si nanowires grown from a
low density of Au colloids, despite the relatively high pSiH4 compared to Si
nanowires grown with densely spaced colloids, as shown in Fig. 4.
place using silane might involve either a process by which an
excess of nanowires are grown simultaneously and then
some are eliminated, or the addition of precracking agent is
used to convert SiH4 to SiH2 before reaching the metal cata-
lyst particle for nanowire growth.
V. CONCLUSIONS
In summary, silicon nanowires have been grown using
Au-catalyst particles and silane. The particles were deposited
by using subcontinuous e-beam thin films or Au-colloid par-
ticles deposited from solution. The growth rate of silicon
nanowires was found to be dependent on the interparticle
spacing of the Au-catalyst particles. A model for two-step
silane decomposition into silylene and then silicon was used
to explain the growth rate dependency on Au surface cover-
age.
ACKNOWLEDGMENTS
The authors would like to thank the staff of the CMSE
Electron Microscopy facilities for their assistance with char-
acterization of these experiments. They would also like to
acknowledge the financial support of the Singapore-MIT Al-
liances program in Advanced Materials for Micro- and
Nano-systems AMM&NS.
1
Y. Cui, Q. Wei, H. Park, and C. M. Lieber, Science 293, 1289 2001.
2
Y. Cui, Z. Zhong, D. Wang, W. U. Wang and C. M. Lieber, Nano Lett. 3,
149 2003.
3
N. Singh, A. Agarwal, L. K. Bera, T. Y. Liow, R. Yang, S. C. Rustagi, C.
H. Tung, R. Kumar, G. Q. Lo, N. Balasubramanian, and D. Kwong, IEEE
Electron Device Lett. 27, 383 2006.
4
Y. Huang, X. Duan, Y. Cui, L. J. Lauhon, K. Kim, and C. M. Lieber,
Science 294, 1313 2001.
5
C. K. Chan, H. Peng, G. Liu, K. McIlwrath, X. F. Zhang, R. A. Huggins,
and Y. Cui, Nat. Nanotechnol. 3, 31 2008.
6
K. Peng, J. Jie, W. Zhang, and S. Lee, Appl. Phys. Lett. 93, 033105
044311-9 Boles et al. J. Appl. Phys. 106, 044311 2009

2008. 16
A. B. Greytak, L. J. Lauhon, M. S. Gudiksen, and C. M. Lieber, Appl.
7
R. S. Wagner and W. C. Ellis, Appl. Phys. Lett. 4, 89 1964. Phys. Lett. 84, 4176 2004.
8
A. I. Hochbaum, R. Fan, R. He, and P. Yang, Nano Lett. 5, 457 2005. 17
J. Westwater, D. P. Gosain, S. Tomiya, S. Usui, and H. Ruda, J. Vac. Sci.
9
T. Shimizu, T. Xie, J. Nishikawa, S. Shingubara, S. Senz, and U. Gsele, Technol. B 15, 554 1997.
Adv. Mater. Weinheim, Ger. 19, 917 2007. 18
H. Schmid, M. T. Bjork, J. Knoch, H. Riel, W. Riess, P. Rice, and T.
10
M. Saif Islam, S. Sharma, T. I. Kamins, and R. Stanley Williams, Appl. Topuria, J. Appl. Phys. 103, 024304 2008.
Phys. A: Mater. Sci. Process. 80, 1133 2005. 19
11 R. D. Deegan, O. Bakajin, T. F. Dupont, G. Huber, S. R. Nagel, and T. A.
R. He and P. Yang, Nat. Nanotechnol. 1, 42 2006.
12 Witten, Nature London 389, 827 1997.
N. J. Quitoriano and T. I. Kamins, J. Appl. Phys. 102, 044311 2007. 20
13
M. T. Borgstrom, G. Immink, B. Ketelaars, R. Algra, and P. A. M. Bak- R. F. C. Farrow, J. Electrochem. Soc. 121, 899 1974.
21
kersErik, Nat. Nanotechnol. 2, 541 2007. C. G. Newman, H. E. Oneal, M. A. Ring, F. Leska, and N. Shipley, Int. J.
14
A. Mikkelsen, J. Eriksson, E. Lundgren, J. N. Andersen, J. Weissenreider, Chem. Kinet. 11, 1167 1979.
22
and W. Seifert, Nanotechnology 16, 2354 2005. B. S. Meyerson and J. M. Jasinski, J. Appl. Phys. 61, 785 1987.
15 23
H. Adhikari, A. F. Marshall, C. E. D. Chidsey, and P. C. McIntyre, Nano W. I. Park, G. Zheng, X. Jiang, B. Tian, and C. M. Lieber, Nano Lett. 8,
Lett. 6, 318 2006. 3004 2008.

Downloaded 02 Sep 2009 to 18.82.0.212. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp