micellar block copolymer templated galvanic displacement for epitaxial nanowire device integration
TRANSCRIPT
Dynamic Article LinksC<Journal ofMaterials Chemistry
Cite this: J. Mater. Chem., 2011, 21, 8807
www.rsc.org/materials PAPER
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online / Journal Homepage / Table of Contents for this issue
Micellar block copolymer templated galvanic displacement for epitaxialnanowire device integration†
Gregory S. Doerk,‡a Charles Dhong,xa Christine Politi,xb Ian Laboriante,a Carlo Carraroa
and Roya Maboudian*a
Received 16th February 2011, Accepted 13th April 2011
DOI: 10.1039/c1jm10693g
New strategies for catalyst nanoparticle placement in arbitrary patterns and non-planar geometries will
likely accelerate the large scale device integration of epitaxial semiconductor nanowires (NWs) grown
through the vapor-liquid-solid (VLS) process. Herein we report a technique for rational metal catalyst
nanoparticle deposition based on galvanic displacement onto semiconductor surfaces through block
copolymer micelle templates. Nanoparticle volumes and areal densities are controlled by the time in the
plating solution and by mixing homopolymer with the micelle suspension, respectively. Above
a minimal nanoparticle diameter, the mean diameters of epitaxial VLS-grown Si NWs scale directly
with mean sizes of the template deposited nanoparticle seeds from which they are grown. The substrate
selectivity and conformality of galvanic displacement makes possible two-level micro/nano-patterning
in a variety of geometries by applying micellar templates over photolithographically patterned masks;
the growth of single sub-50 nm diameter Si NWs in 600–700 nm diameter wells demonstrates a feature
size reduction greater than one order of magnitude. Two-point electrical measurements across single
NWs or a few NWs epitaxially bridging silicon-on-insulator electrodes after ex situ doping demonstrate
the viability of this approach for epitaxial NW device integration.
1 Introduction
The high crystal quality, extensive range of synthetically avail-
able materials,1 and capacity for controlled formation of axial,2
core–shell,3 and branched4,5 semiconductor NW heterostructures
made possible through VLS synthesis portends great promise for
its use in future technologies. Indeed single NW devices have
been demonstrated in numerous application areas including
nanoelectronics,6 photonics,7 and sensing.8 However, the full
potential of NW-based devices cannot be realized without the
development of large scale, high yield device integration strate-
gies with accurate spatial registry. For NWs synthesized by the
VLS method in general, devices are fabricated either on NWs
that have been transferred from the growth substrate to
aDepartment of Chemical and Biomolecular Engineering, University ofCalifornia, Berkeley, California, 94720, USA. E-mail: [email protected]; Fax: +1 510 642 4778bDepartment of Chemical and Biological Engineering, University ofCol-orado, Boulder, Colorado, 80309, USA
† Electronic supplementary information (ESI) available: Auger electronspectra of substrates at different points in the templating process,a plot showing mean Au NP sizes versus galvanic displacementimmersion time for low and high areal densities, and a XPS spectrumof a plain Si substrate. See DOI: 10.1039/c1jm10693g
‡ Present address: IBMAlmaden Research Center, 650 Harry Road, SanJose, CA 95120-6099, USA.
x These authors contributed equally to this work.
This journal is ª The Royal Society of Chemistry 2011
a separate device substrate, or from NWs that have been grown
epitaxially in the location and geometry desired for the device, as
directed by the relation between the orientation of the substrate
surface and the NW preferred growth orientation. While post-
growth assembly techniques have advanced impressively towards
achieving large scale NW device integration,9,10 this integration
problem remains acute for devices composed of epitaxially
grown semiconductor NWs.
This severely hinders the development of technology utilizing
the unique advantages of epitaxial VLS-grown NWs. For
instance, highly aligned vertical NW arrays can be fabricated by
growing the NWs epitaxially on flat substrates if the crystallo-
graphic orientation orthogonal to the surface coincides with the
preferred NW crystal growth orientation. Such arrays may serve
as the basis for cellular biomolecular delivery platforms,11
photovoltaics,12 and three-dimensional integrated circuits.13,14 In
ideal cases, the position and size of each vertical nanowire is
determined by the same characteristics in the predeposited
metallic nanoparticle (NP) that serves as the growth catalyst.
Electron beam lithography may be used to control catalyst
particle placement for epitaxial NW growth,15,16 but its high cost
and serial nature make it unsuitable for large scale processing.
Colloidal metal NPs offer a possible lower cost solution, and
a number of techniques for positioning Au colloids (and hence
the resulting NWs) have emerged,17–20 even to the extent of single
NP resolution.19 However, the range of particle densities per unit
J. Mater. Chem., 2011, 21, 8807–8815 | 8807
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
area may be limited depending on the technique, and accurate
placement of a small number of catalyst NPs (1–3) still requires
high resolution nanolithography (< �200 nm) that will likely
drive up costs. Moreover, these patterning strategies may not be
easily applied to substrates with varied topography (i.e.
nonplanar). This is critical in case of fabricating epitaxially
bridging VLS-grown NWs in microtrenches, where features such
as rigid mechanical clamping21 and ultra-low electrical contact
resistances22 promote their use in potential applications such as
rapid microfluidic biosensors,23 directly grown logic gates and
photovoltaic devices,24 horizontal wrap-gate transistors,25 and
nanomechanical resonant NW mass sensors.26 Nevertheless, to
date the placement of Au colloids on trench sidewalls is still
essentially random.
Specialized metal deposition and templating chemistry may
help bridge the gap between the resolution achieved at low cost
with photolithography and the high resolution required to ach-
ieve devices consisting of single or only a few NWs, while
potentially reducing process complexity. One simple yet attrac-
tive way to deposit metals with nanoscale control is through
galvanic displacement, in which metal ions dissolved in (usually
aqueous) solutions are spontaneously reduced to elemental form
by electrons provided through oxidation of the underlying
substrate. Typically metals with redox potentials above hydrogen
may be deposited, while substrates must be able to oxidize
readily; these criteria still leave available a wide variety of
important metals and semiconductor substrates.27 The require-
ment for substrate oxidation makes the deposition inherently
selective, permitting the use of common masking materials such
as SiO2, SiNx, or some polymers. Furthermore, deposition is
conformal and thickness may be controlled by immersion time
and metal ion concentration. No seed layers are necessary,
though in most cases an additional solute like hydrofluoric acid
must be added to remove the oxidized substrate as it develops.
Gao et al. employed galvanic displacement from water-in-oil
microemulsions with micelles containing an aqueous solution of
a Au salt and hydrofluoric acid to deposit Au NPs selectively
onto Si microtrench sidewalls and flat Si substrates with
controlled nanoscale diameters to seed the subsequent growth of
epitaxial Si nanowires.28 Aizawa and Buriak demonstrated the
use of diblock copolymer micelles to template the deposition of
various size-controlled noble metal NPs on Si, Ge, InP and
GaAs.29,30 Unfortunately, both methods give rise to high NP
areal densities that may only be viable for the growth of dense
NW arrays. However, an intriguing aspect to surface patterning
with block copolymer micellar templates is the fact that the
micelles are often trapped in a non-equilibrium state; they are
then amenable to density modulation by the addition of inert
micelles,31 empty micelles, or homopolymers.32
In this report we demonstrate a general strategy for the
epitaxial integration of semiconductor NWs based on the
controlled deposition of catalyst NPs via galvanic displacement
through micellar diblock copolymer templates. The chemically
rate-controlled Au deposition affords direct control over catalyst
NP size, while the addition of polystyrene homopolymer enables
rational reduction of NP areal densities. Thus, NPs with a range
of sizes may be deposited using a single micellar suspension. The
mean diameters of epitaxial Si nanowires follow the mean sizes of
the as-deposited NPs from which they are grown. Using dip
8808 | J. Mater. Chem., 2011, 21, 8807–8815
coating, we have applied this templating route to both planar and
microtrench substrates patterned with Si ‘‘windows’’ in SiO2
where Au NPs can be deposited selectively. Single Si NWs with
diameters of approximately �50 nm may be grown in windows >
600 nm in diameter, indicative of an order of magnitude reduc-
tion in feature size. Electrical measurements across two Si
electrodes connected by single or few-numbered epitaxially
bridging Si NWs exhibit ohmic responses and specific contact
resistances that compare favorably with lowest values for other
contact methods reported in the literature. The catalyst place-
ment strategy presented here may be a promising step towards
achieving high spatial registry at low cost for the integration of
epitaxial NWs into future devices.
2 Experimental
2.1 Materials
Si(111) dice 10 mm� 11 mm (p-type, r¼ 5–10 mU cm) were used
for basic spin- and dip-coating experiments, as well as for
unpatterned nanowire (NW) growth studies. Two types of
patterned substrates were used. Si(111) substrates covered by an
SiO2 film approximately 400 nm thick and patterned with wells of
different shapes and sizes ranging from < �0.5 mm2 to > 20 mm2
that exposed the underlying Si were fabricated for vertical growth
studies. For epitaxial bridging NW growth, silicon-on-insulator
substrates with an approximately 500 nm thick buried oxide
layer and an approximately 1 mm thick highly boron doped
(> 1018 cm�3) device layer were used. Details about the fabrication
and patterning of these samples has been provided in a previous
publication.33 In all cases, before block copolymer templating,
samples were cleaned by sequential sonication in acetone,
isopropyl alcohol, and deionized (DI) water for 10 min. each,
followedby treatmentwith aUVozone cleaner for 10min. (Jelight
UVO-Cleaner Model No. 42). Two polymers were used in
templating. The block copolymer was polystyrene-block-
poly(2-vinylpyridine) (PS-P2VP, fP2VP ¼ 0.28; total
Mn¼ 46 kg mol�1, Mw/Mn¼ 1.18), and the homopolymer was PS
(Mn ¼ 6 kg mol�1, Mw/Mn ¼ 1.5). Both were synthesized using
anionic polymerization, following standard methods.34
2.2 Block copolymer templating and gold deposition
Micelles were formed by dissolving the PS-P2VP in toluene at
a mass concentration of 5 mg mL�1, and heating the solution
to > 70 �C for more than 10 min., during which time micelles
formed. The solution was then cooled to room temperature.
Homopolymer solutions were made by dissolving PS in toluene
at room temperature with a mass concentration of 5 mg mL�1,
and block copolymer/homopolymer mixtures were prepared with
the desired mass ratio by mixing the PS-P2VP micelle suspension
with the PS solution in the same volume ratio. All mixtures were
stored under refrigeration, but allowed to warm to room
temperature before use, and could be reused repeatedly over the
course of months.
Block copolymer templates were applied to substrates by spin-
or dip-coating. Before use, the liquid mixture was stirred using
a magnetic stir-bar for at least 30 min. For spin coating, 15 mL of
a polymer mixture was placed on a sample, which was then
spin-coated in air at 4000 rpm. Dip-coating was accomplished by
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
pulling the sample out of the polymer mixture at a rate of
80–90 mm/min.
Metal deposition was accomplished via galvanic displacement.
Plating solutions were made by adding 100 mL of a 10 mM
aqueous solution of potassium tetrachloroaurate (KAuCl4)
along with 10 mL of 48–50% (w/w) hydrofluoric acid (HF) to 10
mL of DI water. Particles of different sizes were deposited
through the micelle template by immersing samples in the plating
solution for 10 min. or less, with negligible effect on the particle
density.† After metal deposition, samples were sequentially
immersed in at least 2 beakers of clean DI water followed by
drying under a nitrogen stream.
The polymer template was removed by immersing samples into
neat toluene heated to temperatures above greater than 70 �C for
at least 10 min. Unpatterned samples were then sonicated in the
same solution for 10 min. In patterned samples sonication
sometimes resulted in a significant loss in the number of Au
nanoparticles (NPs) and so it was not always done. Auger elec-
tron spectra confirm the removal of most of the polymer film
with sonication in heated toluene.†
2.3 Synthesis and boron doping of silicon nanowires
Si NWs were synthesized via vapor-liquid-solid (VLS) growth
from the template-deposited Au NPs. Samples were placed in
a quartz tube, hot-wall, atmospheric pressure chemical vapor
deposition reactor, with NW growth occurring at 835 �C for 5–
10 min. SiCl4 was transported into the reactor by bubbling
10 sccm of carrier gas (10%H2 in Ar) through liquid SiCl4 held at
0 �C, while 100 sccm of the carrier gas flowed directly to the
reactor. For the growth of Si NWs using Au catalyst NPs with
diameters above �20–30 nm, the thermodynamically preferred
growth direction is the <111> direction; thus a large proportion
of Si NWs grown from Au epitaxially from Si(111) surfaces will
stand perpendicular to the surface.5
Ex situ boron doping was performed using BBr3 in the same
reactor used in Si NW growth but with a different quartz tube.
The process has been described in a previous publication,35 but
here we briefly summarize it with details pertinent to this work.
Prior to doping, the sample was immersed in an iodine/potassium
iodide Au etchant solution (4 : 1 : 40 KI:I2:H2O) for more than
30 min. to remove the Au catalyst material, followed by rinsing
with isopropyl alcohol and drying under a nitrogen stream. The
sample was then cleaned by UV ozone treatment for 30 min. and
inserted into the reactor for doping, which was performed in two
steps. In the first step boron was introduced to the Si NWs at
700 �C by bubbling 6 sccm of the carrier gas (again 10%H2 in Ar)
through liquid BBr3 held at 0 �C for 1 min., while 270 sccm of the
carrier gas flowed directly to the reactor. At this stage, BBr3 is
believed to be reduced by H2 to deposit boron on the Si NW
surfaces. The BBr3 flow was then cut off and the carrier gas was
switched to Ar gas only. The sample was annealed under this Ar
carrier gas at the same temperature for 60 min. to drive-in and
activate the boron as a dopant.
2.4 Characterization
Scanning electron microscope (SEM) images were obtained using
either an Agilent 8500 field-emission SEM or a Leo 1550
This journal is ª The Royal Society of Chemistry 2011
Schottky field-emission SEM. Atomic force microscopy (AFM)
measurements were performed in air with a Digital Instruments
Nanoscope IIIa system in tapping mode. Mean NP volumes were
determined by flooding analysis using WSxM Scanning Probe
Microscopy Software,36 and the error bars presented in the
report represent the standard deviation in the mean NP volume
over several images. For aesthetic presentation, Gwyddion
scanning probe microscopy software was used to create topo-
graphic images. Au NP counts and areal densities were deter-
mined from SEM and AFM images. Error bars in such cases
represent standard deviations in the mean values obtained from
at least three images. All images used for measurements were
taken from the same samples, though repeated experimental
conditions resulted in highly consistent results. Surface cleanli-
ness and composition were characterized by Auger electron
spectroscopy (AES; PHI-Perkin-Elmer model 10-155) and X-ray
photoelectron spectroscopy (XPS; Omicrometer analyzer,
EA 125).
Two-point electrical measurements were performed using
a probe station with a HP4145B Semiconductor Parameter
Analyzer. Since a number of electrodes are interconnected by
epitaxially bridging Si NWs, and these in turn occasionally
bridge to a third ‘‘gate’’ electrode nearby, a small degree of
leakage current was observed in some measurements. We cor-
rected for this by measuring not just the resistance across the
NWs under test (‘‘source to drain’’, RSD) but also through the
other connected conduction paths (‘‘source to gate’’, RSG, and
‘‘gate to drain’’, RGD). The actual NW resistance (RNWs) is
obtained by the following equation:
RNWs ¼ RSDðRSG þ RGDÞRSG þ RGD � RSD
(1)
The measurements of RSG and RGD most likely underestimate
their true values due to the parallel conduction path through the
wire; therefore this measurement may overestimate the NW
resistance. The maximum negative error is included in Fig. 6(d);
however, a large potential barrier to electrical conduction by the
leakage pathsmakes this negative error small inmost cases for the
voltage range over which resistance is measured (�1 V to 1 V).
3 Results and discussion
A schematic for the micellar templating strategy is presented in
Fig. 1. The block copolymer used was polystyrene-block-
poly(2-vinylpyridine) (PS-P2VP; Mn ¼ 46 kg mol�1, Mw/Mn ¼1.18). In bulk form, the P2VP volume fraction (fP2VP ¼ 0.28)
implies that the polymer should microphase separate to cylin-
drical P2VP domains in a PS matrix;37 when dissolved in toluene
(5 mg mL�1) and heated to 70 �C the preferential interactions
between toluene and polystyrene result in the formation of
spherical micelles with P2VP cores and PS coronas (Fig. 1(a))
that are preserved when the solution is cooled to room temper-
ature.29 This micellar suspension is then mixed with a solution of
PS homopolymer (Mn ¼ 6 kg mol�1, Mw/Mn ¼ 1.5) in toluene at
the samemass concentration to obtain a newmixture with a mass
ratio of PS homopolymer to block copolymer that is equivalent
to the volume ratio of the two mixtures (Fig. 1(b)). A micellar
template film is applied to a silicon substrate by either spin or dip
coating. After the sample is visibly dry, templated samples are
J. Mater. Chem., 2011, 21, 8807–8815 | 8809
Fig. 1 Block copolymer templating scheme. (a) Micellization of
PS-P2VP occurs in toluene above room temperature, with P2VP cores
(red) and PS coronas (blue). (b) After being cooled to room temperature,
the micelle suspension is mixed with a solution of PS in toluene having the
same mass concentration in the desired mass proportion. (c) The micelle
template is applied to a semiconductor surface by spin- or dip-coating,
where the PS serves as a spacer between micelles. The sample is then
exposed to the aqueous galvanic displacement bath, where the P2VP
micelle cores allow selective access to the surface for metal deposition. (d)
The polymer template is removed, leaving an array of metal NPs.
Fig. 2 Results of the templating process with no added PS. SEM images
of (a) the initial micellar template, (b) after Au galvanic displacement for
10 min, and (c) after polymer removal. Scale bar ¼ 1 mm for all images.
(d) Au region of the XPS spectrum from the sample shown in (c).
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
immersed in a Au plating solution consisting of potassium
tetrachloroaurate (KAuCl4) and excess hydrofluoric acid (HF) in
deionized (DI) water for variable times, during which Au is
deposited via galvanic displacement only through the P2VP cores
that give aqueous solutions access to the underlying Si surface, as
depicted in Fig. 1(c). The PS homopolymer acts as a spacer on
the surface, decreasing the surface areal density of the micelles.
Finally, the polymer is removed through immersion in hot
toluene (> 70 �C), followed by sonication (which may be skipped
if particle adhesion is poor) leaving an array of Au NPs with
controlled size and areal density (Fig. 1(d)). The effectiveness of
the polymer removal is verified using Auger electron
spectroscopy.†
The results of this templating approach (by spin coating) with
no PS homopolymer added are displayed in Fig. 2. Fig. 2(a)
shows a scanning electron microscope (SEM) image of the initial
polymer template, while Fig. 2(b) and (c) show SEM images after
10 min of galvanic displacement and subsequent polymer
removal, respectively. There are fewer particles (� 40–50 mm�2)
after galvanic displacement than there are apparent micelles
(> 60 mm�2), though the final number of particles after polymer
removal is approximately the same after galvanic displacement.
8810 | J. Mater. Chem., 2011, 21, 8807–8815
The imperfect formation of particles from the micellar template
may be attributed to the presence of a PS homopolymer layer
directly above (or underneath) particular micelles that prevent
direct access of the aqueous solution from the substrate. The Au
region of an X-ray photoelectron spectrum (Fig. 2(d)) taken
from the sample shown in Fig. 2(c) indicates that the deposited
material is indeed elemental Au. The small hump at �90 eV is
from the Si substrate itself.†
A primary reason for using block-copolymer micellar
templating is the capability to tune the particle areal density over
a wide range. We do this by adding PS homopolymer, the results
of which are summarized in Fig. 3. SEM images in Fig. 3(a–c)
depict as-deposited Au NPs after polymer removal for three
different mass fractions of PS-P2VP block-copolymer with PS
homopolymer as the other binary component (the total mass
concentration of the mixture was held constant). As the block-
copolymer mass fraction is increased from�0.03 (a) to�0.33 (b),
and finally up to 1.0 (c; undiluted block copolymer), the Au NP
areal density increases proportionally. Since the areal density (ra)
is proportional to D�2, where D is the domain size, reduction in
areal density is also possible by using higher molecular weight
block copolymers. In the case of PS-P2VP which may be
considered amphiphilic, or at least strongly segregated (at the
molecular weight used in this report or above), D may be expected
to scale with the degree of polymerization N as D f N2/3�1.38
As a result, polymers used to achieve this reduction in ra must
possess molecular weights at least �5–10 times larger, and the
weak scaling of micelle center-to-center distance with degree of
polymerization for metallic NPs deposited through PS-P2VP
micellar templates in the previous work by Aizawa and Buriak
corresponds more closely to the strongly segregated limit (i.e.
requiring molecular weights at least 10 times larger).30 On the
other hand, as demonstrated here extensive areal density tuning
may be achieved in a rational way through the mixture of only
a single block copolymer and a single homopolymer. Though the
Au NP size is controlled separately by galvanic displacement
This journal is ª The Royal Society of Chemistry 2011
Fig. 3 NP areal density control by homopolymer addition. (a–c) SEM
images of samples produced with block copolymer mass fractions of (a)
0.03 (1 min galvanic displacement), (b) 0.33 (5 min galvanic displace-
ment), and (c) 1.0 (10 min galvanic displacement). All images were taken
after polymer removal, and the scale bar ¼ 1 mm for each image.
Immersion times for each image were chosen to provide optimal NP
clarity and contrast, due to the decreased average Au NP deposition rate
with increased micelle areal density.† (d) NP areal density as a function of
the block copolymer mass fraction for dip-coated samples. The dashed
line depicts expected areal densities assuming it is proportional to the
volume fraction of block copolymer in the mixture. The error bars
represent standard deviations in mean values obtained over several
images.
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
plating time (discussed later in this report), the surface micelle
areal density does have an notable impact. Specifically, NP
growth rate is reduced as ra is increased, a fact that we attribute
to increased competition for reactant. For this reason, lower
immersion times were necessary to achieve optimal NP clarity
and contrast in the SEM images depicted in Fig. 3(a–c). A graph
comparing Au NP growth at > 40 mm�2 and < 2 mm�2 is con-
tained in the ESI.†
Fig. 3(d) shows a plot of particle areal density on dip-coated
samples versus the block-copolymer mass fraction in the mixture.
Undiluted block copolymer results in particle areal densities
�42 mm�2, whereas particle areal densities less than 2 mm�2 are
readily obtained for block copolymer mass fractions
< 0.05 (homopolymer to block-copolymer mass ratios greater
than 20 to 1). Similar results were obtained for spin-coating. If
one assumes that the volume fractions of block copolymer and
homopolymer in the thin film micellar template are approxi-
mately identical with their volume fractions in the toluene
mixture, then the areal density of NPs may be considered directly
proportional to the block copolymer volume fraction, fBC, rep-
resented by a dashed blue line in Fig. 3 (The maximum NP areal
density is set to the experimentally measured value and volume
This journal is ª The Royal Society of Chemistry 2011
fraction and mass fraction of block copolymer are nearly iden-
tical, given the equivalent molar volume39 and nearly equivalent
molecular weights for the PS and P2VP monomer units). This
simple model provides a fair guide for expected NP areal densi-
ties, though there is a noticeable discrepancy at a mass fraction of
�0.09. This may be a consequence of the polymer deposition
method since no further steps that are commonly used in block
copolymer template processing like thermal or solvent annealing
have been applied. It may also be due to higher NP deposition in
the micellar template. The connections between processing
conditions and areal density will be investigated more thoroughly
in future work.
Various proposed applications for epitaxially VLS-grown
semiconductor NWs demand that their diameters be precisely
controlled. Since VLS-grown NW diameters are determined
primarily by the size of the metal NP used as a catalyst, this is the
most critical parameter to control. In the method proposed here,
catalyst NP size is regulated by the galvanic deposition rate
(encompassing chemical kinetic and mass transport factors) and
the time of deposition. In order to link NP size with NW diam-
eter, both were measured from the same sample before and after
growth. Furthermore, in order to minimize particle agglomera-
tion due to the high mobility of Au on Si at the elevated
temperatures of our growth reactor (> 800 �C)5,40,41 that wouldstrongly affect the nanowire diameter distribution, we used
samples with low areal densities (< 2 mm�2; template applied by
spin-coating). Fig. 4(a) and (b) show atomic force microscopy
(AFM) topography images of Au NPs on Si(111) deposited
through the method described in this report for immersion times
of 1 min. and 10 min., respectively, and Fig. 4(c) and (d) show
SEM images of the resulting Si NWs grown from the samples in
(a) and (b), demonstrating the clear increase in mean diameter
for NWs grown from the larger particles. The epitaxial nature of
the as-grown Si NWs is evidenced by the high degree of vertical
growth.5
If one assumes that the Au droplet on top of the NW is
approximately hemispherical and the NW radius is nearly
equivalent to the droplet radius, then the NW diameter is
expected to be proportional to the cube root of the Au NP
volume from which it is grown. The exact proportionality is
determined by the eutectic droplet contact angle on Si and the
equilibrium Si mole fraction in the droplet at the growth
temperature, as well as the Si supersaturation during growth and
the ellipticity of the droplet. We ignore the portion of the AuNPs
recessed in the Si substrate though this is offset to some extent by
the convolution of tip size with the NP size measurement.
Therefore, we focus on the proportional scaling between NW
diameter with Au NP size as determined by immersion time.
Fig. 4(e) shows the cube root of the mean Au NP volume as well
as the mean NW diameter versus immersion time. Both increase
monotonically with increasing immersion time, with a much
faster growth rate characterizing the initial NP nucleation stage
(< �1 min.). For times above one minute the mean NW diam-
eters scale directly with the NP size, as expected. A larger
difference between the mean NW diameter and the cube root of
the NP volume is seen for the 30 s. sample. In this case, fewer
NWs grow from smaller Au clusters, a fact that is commonly
attributed to the Gibbs-Thomson effect in which the supersatu-
ration in the droplet is energetically limited by the increased
J. Mater. Chem., 2011, 21, 8807–8815 | 8811
Fig. 4 Control over NP size by immersion time. (a–b) AFM topography
images of Au NPs deposited on Si(111) by galvanic displacement through
�2 mm�2 micellar templates for (a) 1 min. and (b) 10 min. The squares are
both 5 mm � 5 mm and the vertical scale is the same for both images
(500 nm). (c–d) Si NWs epitaxially grown from the Au NPs shown in (a)
and (b), respectively. The scale bars are 5 mm in both images. (e) Cubic
root of the mean NP volume and mean diameters of Si NWs grown from
these particles versus immersion time in the aqueous galvanic displace-
ment solution. For 1 min. and above the NW diameters and NP volume
cube roots scale with each other, while for less than 1 min. NWnucleation
barriers skew the measured mean NW diameter to a higher value. Con-
necting Lines are guides to the eye and the error bars represent standard
deviations.
Fig. 5 Block copolymer templating and Si NW growth on Si(111)
samples with patterned wells in a SiO2 hard mask. The PS:PS-P2VP mass
ratio is 10 : 1. (a) Si NWs grown in �100 mm2 wells (scale bar ¼ 10 mm).
(b) Si NWs grown in circular wells 600–700 nm in diameter (scale bar ¼10 mm). (c) Higher magnification of individual Si NWs grown in the wells
shown in (b) (scale bar¼ 1 mm). (d) Number of AuNPs per well area. The
solid line represents expected values based on measurements of areal
densities on unpatterned samples.
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
droplet surface energy relative to chemical energy in the droplet
volume as the droplet size is reduced.42 Since the VLS NW
growth rate is proportional to supersaturation, smaller NWs
grow much more slowly or not at all; the overall NW density is
then reduced and the mean NW diameter is skewed to a higher
value than expected based on the geometric argument given
above. Similar Au NP growth behavior is observed for higher
areal density micellar templates, though the overall NP growth
rate is reduced.†
A significant advantage to galvanic displacement is its
substrate selectivity, which permits highly controlled conformal
metal deposition in patterns without additional lift-off steps.
Therefore, we applied a block-copolymer template (PS:PS-P2VP
ratio¼ 10 : 1) to Si(111) substrates with a patterned silicon oxide
hard mask by dip coating, and deposited Au NPs via galvanic
displacement for 1 min. The selective deposition of Au NPs
means that Si NWs can only be grown in the wells exposing the Si
8812 | J. Mater. Chem., 2011, 21, 8807–8815
substrate, as shown in Fig. 5(a)–(c). Critically, in Fig. 5(c) it is
demonstrated that individual NWs less than �50 nm in diameter
may be grown in wells larger than 600 nm in diameter, achieving
a reduction in feature size greater than an order of magnitude
with the same approximate pattern registry. For uniform
coverage of the micellar template, the number of NPs and
subsequently grown NWs in a well will be proportional to the
well area, which is consistent with experimental data shown in
Fig. 5(d). Note that the solid line is not a linear fit to the data, but
rather the expected number of particles based on the AuNP areal
densities measured on unpatterned samples. Subsequently grown
NWs track the Au NP density well.
The conformality of galvanic displacement suggests that this
templating technique may also be applied to appropriately
patterned substrates with trenches having the correct orientation
of sidewalls (namely Si(111) for the Au/Si epitaxial NW growth
system studied in this report5,28) so that epitaxial Si NWs most
often grow perpendicularly from them for the fabrication of
devices based on in-plane epitaxial NWs. We primarily focus on
NWs grown from the 30 : 1 template (ra < 2 mm�2) since this is
the most likely template to enable the fabrication of devices
based on single or few NWs using only micron-scale lithography.
Au NPs were deposited using a 1 min. immersion time in the Au
deposition bath. Fig. 6(a) shows Si NWs grown in a �2 mm�2
wide trench to epitaxially bridge two electrically isolated single-
crystal Si electrodes on a buried silicon oxide layer. The top of the
electrodes were also covered with silicon oxide, preventing Au
cluster deposition on the top surface as well. Selective growth of
Si NWs can be enhanced if the trench sidewalls are also covered
with an oxide mask and a second lithography step is used to
This journal is ª The Royal Society of Chemistry 2011
Fig. 6 Epitaxial Si NW growth from templated Au NP seeds (ra < 2 mm�2) on silicon-on-insulator trench sidewalls. (a) SEM image of Si NWs grown in
a 2 mmwide trench (scale bar¼ 2 mm) (b) SEM image (bottom) of a single Si NW grown in a photolithographically defined Si window depicted in the top
schematic diagram (scale bar ¼ 2 mm). (c) Current–voltage response for the Si NW shown in (b) after ex situ boron doping. (d) Resistance versus the
physical scaling parameter inside parentheses in eqn (2) for 11 single Si NW or few-NW epitaxially bridging device structures on the same sample. The
red line is a linear fit to the data.
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
define micron-scale windows of exposed Si, as for example in
Fig. 6(b). In this case a single Si NW epitaxially bridges the two
highly-doped single-crystal Si electrodes near a third electrode
that may be used for further device fabrication if desired (the
capacitive coupling with the NW is too small to yield any gate
dependent response in the current configuration). After Si NW
growth, the sample was doped with boron through an ex situ
method reported previously35 to achieve an active dopant
concentration greater than 10�19 cm�3. After doping, the top
layer of silicon oxide covering the Si electrodes was removed and
two-point current–voltage measurements were performed on
various single-NW or few-numbered NW device structures.
Fig. 6(c) depicts the measured current–voltage curve for the
single Si NW in (b) evidencing an ohmic response that was
typical of these device structures. The two-point current–voltage
responses for 11 such structures on the same die were measured.
The high doping level minimizes the surface carrier depletion
width so that we may assume that the full NW diameters are
involved in electronic conduction. For a given two point
measurement, the NWs which conduct current may possess
different diameters (D), but are all approximately the same
length (L). In addition, if we assume that the electrode and probe
tip resistances are negligible, and that the doping is uniform
This journal is ª The Royal Society of Chemistry 2011
across the substrate, the resistance of the NWs (RNWs) is given
by:
RNW s ¼ 2RC þ r
4L
pPN
i D2i
!(2)
where RC is the contact resistance and r is the resistivity of the
NWs. Plotting measured resistance versus the size parameter
contained in parentheses in eqn (2) (Fig. 6(d)), we derive a resis-
tivity value of 10� 5 mU cm and an average contact resistance of
7 � 3 kU. Based on the average cross sectional area of bridging
NW device structures, the estimated specific contact resistance is
(1.0� 0.5)� 10�6 U cm2. This value is superior to specific contact
resistances reported for Si NWs using deposited metal films,43,44
and is comparable or lower than previous values for epitaxial
NW contacts,22,45 though NiSi contacts to Si NWs have been
shown to exhibit lower values.46 Nonetheless, the electrical data
presented here reveal the great potential for fabricating devices
based on epitaxially integrated NWs with catalyst deposition
achieved via micellar block-copolymer templated galvanic
displacement.
Since the micellar template is disordered, the registry for
particle deposition with the larger photolithographic pattern is
J. Mater. Chem., 2011, 21, 8807–8815 | 8813
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
statistical and thus imperfect; hexagonal ordering might signifi-
cantly improve upon this. Using polymers with more uniform
chain lengths would likely enhance the hexagonal packing to
a small degree. We avoided thermal annealing since the micellar
template is a trapped morphology that is not favored in thin film
or bulk phases and would likely transfer to an in-plane cylin-
drical morphology (for the given P2VP volume fraction in the
PS-P2VP) if thermally annealed.47 On the other hand, solvent
annealing typically induces non-equilibrium morphologies and
has been shown to improve hexagonal ordering of spherical
micelles.48,49 It may be possible to apply solvent annealing to
block-copolymer micelle/homopolymer systems like the one
described here, though we are not aware of any preceding work
on solvent annealing of block copolymer micelles with low areal
densities or homopolymer molecular spacers. Besides micelles,
equilibrated block copolymer thin films may exhibit excellent
ordering and uniformity with the help of chemical epitaxy50 or
graphoepitaxy,51 though significant effort has been focused on
pattern density multiplication,50,51 opposing the effort of this
work. Nevertheless, Stuen et al. have recently shown that the
domain spacing of polystyrene-block-poly(methyl methacrylate)
(PS-PMMA) block copolymer templates consisting of perpen-
dicular cylindrical minority domains (PMMA) may be adjusted
by a factor of 3 by concurrent PS and PMMA homopolymer
addition, though a substantial reduction in domain size unifor-
mity was observed and the pattern areal densities were still well
above 100 mm�2 (assuming hexagonal packing).52 Papalia et al.
have also recently demonstrated areal density tuning of spherical
minority polyisoprene domains in polystyrene-block-poly-
isoprene (PS-PI) over a very broad range by the addition of PS
homopolymer and controlled film thickness, though hexagonal
ordering was not apparent and the minimal areal density
reported was > �20 mm�2.53 In any case, we conjecture that the
evolving developments in ordering block copolymer thin film or
micellar templates may be leveraged with galvanic displacement
for the epitaxial integration of NW devices via precise substrate
and pattern registry in various geometries.
4 Conclusions
In conclusion, we have demonstrated a facile technique to inte-
grate epitaxially grown semiconductor NWs into device struc-
tures based on galvanic displacement of catalyst seed NPs
through PS-P2VP block copolymer micelle templates. Areal
density tuning from > �40 mm�2 to < �2 mm�2 is achieved
through mixture with PS homopolymer, and Au NP size is
directly controlled by immersion time in the aqueous plating
solution. The inherent substrate selectivity and conformality of
galvanic displacement enables epitaxial NW registry with much
larger photolithographically defined patterns in both horizontal
and vertical geometries, and the growth of single NWs with
diameters of�50 nm in wells with diameters larger than�600 nm
indicate pattern size reductions greater than one order of
magnitude. Electrical testing of ex situ doped epitaxially bridging
Si NW device structures further establishes the viability of this
approach.
In this report, we have concentrated on the galvanic
displacement of Au through NP templates as this is the most
commonly used catalyst metal for VLS NW growth. However,
8814 | J. Mater. Chem., 2011, 21, 8807–8815
the likelihood of internal54 or surface40 Au contamination that
may severely impair the performance of devices based on VLS-
grown Si NWs has generated considerable interest in other
catalysts that may be more CMOS compatible.55 Fortunately, as
noted earlier in this report galvanic displacement may be used to
deposit a number of different metals on various semiconductor
surfaces.27 In principle, all these metals may be directed to
deposit as NPs through the templating method described here.
Beyond NW integration, the capability to deposit noble metal
NPs selectively on semiconductor surfaces with controlled
diameters, areal densities and pattern registry may prove valu-
able in plasmonic applications such as enhancing light absorp-
tion in thin film photovoltaics56 or for plasmonic chemical or
biological sensing.57
5 Acknowledgements
We acknowledge the support of the National Science Founda-
tion, Grants No. EEC-0832819 (through the Center of Inte-
grated Nanomechanical Systems) and DMR-0804646. We also
thank Marta Fernandez, Dr Alvaro San Paulo, and Dr Noel
Arellano for the use of patterned substrates, as well as Megan
Hoarfrost and Prof. Rachel Segalman for providing the
polymers.
References
1 M. Law, J. Goldberger and P. Yang,Annu. Rev.Mater. Res., 2004, 34,83–122.
2 M. Gudiksen, L. Lauhon, J. Wang, D. Smith and C. Lieber, Nature,2002, 415, 617–620.
3 L. Lauhon, M. Gudiksen, C. Wang and C. Lieber, Nature, 2002, 420,57–61.
4 K. Dick, K. Deppert, M. Larsson, T. Martensson, W. Seifert,L. Wallenberg and L. Samuelson, Nat. Mater., 2004, 3, 380–384.
5 G. Doerk, N. Ferralis, C. Carraro and R. Maboudian, J. Mater.Chem., 2008, 18, 5376–5381.
6 J. Xiang, W. Lu, Y. Hu, Y. Wu, H. Yan and C. Lieber, Nature, 2006,441, 489–493.
7 L. Cao, J. Park, P. Fan, B. Clemens and M. Brongersma, Nano Lett.,2010, 10, 1229–1233.
8 Y. Cui, Q. Wei, H. Park and C. Lieber, Science, 2001, 293, 1289–1292.9 M. Li, R. Bhilvadvala, T. Morrow, J. Sioss, K. Lew, J. Redwing,C. Keating and T. Mayer, Nat. Nanotechnol., 2008, 3, 88–92.
10 E. Freer, O. Grachev, X. Duan, S. Martin and D. Stumbo, Nat.Nanotechnol., 2010, 5, 525–530.
11 A. Shalek, J. Robinson, E. Karp, J. Lee, D. Ahn,M. Yoon, A. Sutton,M. Jorgolli, R. Gertner, T. Gujral, G. MacBeath, E. Yang andH. Park, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 1870–1875.
12 M. Kelzenberg, S. Boettcher, J. Petykiewicz, D. Turner-Evans,M. Putnam, E. Warren, J. Spurgeon, R. Briggs, N. Lewis andH. Atwater, Nat. Mater., 2010, 9, 239–244.
13 T. Bryllert, L. Wernersson, L. Froberg and L. Samuelson, IEEEElectron Device Lett., 2006, 27, 323–325.
14 S. Hu, P. Leu, A. Marshall and P. McIntyre,Nat. Nanotechnol., 2010,4, 649–653.
15 T. Martensson, M. Borgstrom, W. Seifert, B. Ohlsson andL. Samuelson, Nanotechnology, 2003, 14, 1255–1258.
16 H. Schmid, M. Bjork, J. Knoch, H. Riel, W. Riess, P. Rice andT. Topuria, J. Appl. Phys., 2008, 103, 024304.
17 A. Hochbaum, R. Fan, R. He and P. Yang, Nano Lett., 2005, 5, 457–460.
18 J. Huang, A. Tao, S. Connor, R. He and P. Yang,Nano Lett., 2006, 6,524–529.
19 T. Kraus, L. Malaquin, H. Schmid, W. Riess, N. Spencer andH. Wolf, Nat. Nanotechnol., 2007, 2, 570–576.
20 P. Manandhar, E. Akhadov, C. Tracy and S. Picraux, Nano Lett.,2010, 10, 2126–2132.
This journal is ª The Royal Society of Chemistry 2011
Publ
ishe
d on
12
May
201
1. D
ownl
oade
d by
Lom
onos
ov M
osco
w S
tate
Uni
vers
ity o
n 06
/09/
2013
14:
02:0
6.
View Article Online
21 A. S. Paulo, J. Bokor, R. Howe, R. He, P. Yang, D. Gao, C. Carraroand R. Maboudian, Appl. Phys. Lett., 2005, 87, 053111.
22 A. Chaudhry, V. Ramamurthi, E. Fong and M. Islam, Nano Lett.,2007, 7, 1536–1541.
23 D. Kim and X. Zheng, Nano Lett., 2008, 8, 3233–3237.24 D. Kim, C. Lee and X. Zheng, Nano Lett., 2010, 10, 1050–1054.25 E. Garnett, Y. Tseng, D. Khanal, J. Wu, J. Bokor and P. Yang, Nat.
Nanotechnol., 2009, 4, 311–314.26 E. Gil-Santos, D. Ramos, J. Martinez, M. Regulez-Fernandez,
R. Garcia, A. S. Paulo, M. Calleja and J. Tamayo, Nat.Nanotechnol., 2010, 5, 641–645.
27 C. Carraro, R. Maboudian and L. Magagnin, Surf. Sci. Rep., 2007,62, 499–525.
28 D. Gao, R. He, C. Carraro, R. Howe, P. Yang and R. Maboudian, J.Am. Chem. Soc., 2005, 127, 4574–4575.
29 M. Aizawa and J. Buriak, J. Am. Chem. Soc., 2005, 127, 8932–8933.30 M. Aizawa and J. Buriak, Chem. Mater., 2007, 19, 5090–5101.31 W. Hwang, J. Choi, T. Kim, J. Sung, J. Myoung, D. Choi, B. Sohn,
S. Lee, D. Kim and C. Park, Chem. Mater., 2008, 20, 6041–6047.32 R. Bennett, A. Miller, N. Kohen, P. Hammond, D. Irvine and
R. Cohen, Macromolecules, 2005, 38, 10728–10735.33 A. S. Paulo, N. Arellano, J. Plaza, R. He, C. Carraro, R. Maboudian,
R. Howe, J. Bokor and P. Yang, Nano Lett., 2007, 7, 1100–1104.34 H. Yokoyama, T. Mates and E. Kramer, Macromolecules, 2000, 33,
1888–1898.35 G. Doerk, G. Lestari, F. Liu, C. Carraro and R. Maboudian,
Nanoscale, 2010, 2, 1165–1170.36 I. Horcas, R. Fernandez, J. Gomez-Rodriguez, J. Colchero,
J. Gomez-Herrero and A. Baro, Rev. Sci. Instrum., 2007, 78, 013705.37 R. Segalman, Mater. Sci. Eng., R, 2005, 48, 191–226.38 S. Forster, M. Zisenis, E. Wenz and M. Antonietti, J. Chem. Phys.,
1996, 104, 9956–9970.39 H.Eitouni andN.Balsara, inPhysicalProperties ofPolymersHandbook,
Second Edition, ed. J. Mark, Springer, New York, 2007, pp. 339–356.
This journal is ª The Royal Society of Chemistry 2011
40 J. Hannon, S. Kodambaka, F. Ross and R. Tromp,Nature, 2006, 440,69–71.
41 N. Ferralis, F. E. Gabaly, A. Schmid, R. Maboudian and C. Carraro,Phys. Rev. Lett., 2009, 103, 256102.
42 E. Givargizov, J. Cryst. Growth, 1975, 31, 20–30.43 S. Mohney, Y. Wang, M. Cabassi, K. Lew, S. Dey, J. Redwing and
T. Mayer, Solid-State Electron., 2005, 49, 227–232.44 H. Tong, S. Chen, W. van der Wiel, E. Carlen and A. van den Berg,
Nano Lett., 2009, 9, 1015–1022.45 A. Lugstein, M. Steinmair, C. Henkel and E. Bertagnolli, Nano Lett.,
2009, 9, 1830–1834.46 S. Habicht, Q. Zhao, S. Feste, L. Knoll, S. Trellenkamp, B. Ghyselen
and S. Mantl, Nanotechnology, 2010, 21, 105701.47 J. Chai and J. Buriak, ACS Nano, 2008, 2, 489–501.48 S. Yun, S. I. Yoo, J. Jung, W. Zin and B. Sohn, Chem. Mater., 2006,
18, 5646–5648.49 T. Kim, J. huh, J. Hwang, H. Kim, S. Kim, B. Sohn and C. Park,
Macromolecules, 2009, 42, 6688–6697.50 R. Ruiz, H. Kang, F. Detcheverry, E. Dobisz, D. Kercher,
T. Albrecht, J. de Pablo and P. Nealey, Science, 2008, 321, 936–939.
51 I. Bita, J. Yang, Y. Jung, C. Ross, E. Thomas and K. Berggren,Science, 2008, 321, 939–943.
52 K. Stuen, C. Thomas, G. Liu, N. Ferrier and P. Nealey,Macromolecules, 2009, 42, 5139–5145.
53 J. P. A. Marencic, D. Adamson, P. Chaikin and R. Register,Macromolecules, 2010, 43, 6946–6949.
54 J. Allen, E. Hemesath, D. Perea, J. Lensch-Falk, Z. Li, F. Yin,M. Gass, P. Wang, A. Bleloch, R. Palmer and L. Lauhon, Nat.Nanotechnol., 2008, 3, 168–173.
55 V. Schmidt, J. Wittemann and U. Gosele,Chem. Rev., 2010, 110, 361–388.
56 H. Atwater and A. Polman, Nat. Mater., 2010, 9, 205–213.57 J. Homola, Chem. Rev., 2008, 108, 462–493.
J. Mater. Chem., 2011, 21, 8807–8815 | 8815