Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Article

pubs.acs.org/crystal

Metal-Seeded Growth Mechanism of ZnO Nanowires


Heike Simon, Tobias Krekeler, Gunnar Schaan, and Werner Mader*
Institute of Inorganic Chemistry, University of Bonn, Roemerstrasse 164, D-53117 Bonn, Germany
*
S Supporting Information

ABSTRACT: The widely applied metal-catalyzed growth


mechanism of ZnO nanowires (NWs) is investigated by
advanced methods of transmission electron microscopy and is
discussed with respect to thermodynamic growth conditions.
Au catalyst particles do not contain a substantial amount of Zn
proving a solid Au catalyst at 1173 K growth temperature. This
result is owed to the high equilibrium Zn partial pressure over
Au−Zn alloys which in turn leads to a very low sticking
coefficient of Zn from vapor and prevents alloying. Growth
rates of ZnO NWs were measured between 5.5 nm s−1 and 36
nm s−1 as a function of oxygen partial pressure. The enhanced
growth rate at higher oxygen partial pressures is explained by
an increased sticking coefficient of Zn atoms at the Au catalyst. A growth mechanism is proposed which is quite different from
the classic vapor−liquid−solid (VLS) mechanism: Zn alloys only in a thin surface layer at the catalyst and diffuses to the vapor−
catalyst−NW triple phase line. There, together with oxygen, ZnO ledges nucleate which grow laterally to inner regions of the
ZnO−Au heterointerface where Zn and oxygen can diffuse and finally promote NW growth in a rather kinetically controlled
process. The geometry of the ZnO−Au interface  planar or stepped  and the associated diffusional transport properties are
shown to be determined by the orientation relationship between Au and ZnO and hence by the atomic structure of the interface.

1. INTRODUCTION supplied with the inert carrier gas. This is a very efficient and
One-dimensional nanostructures exhibit distinctive and novel fairly simple method to apply. Only a few reports exist which
properties that may differ from those of bulk and thin-film systematically focus on the growth conditions of such as oxygen
materials.1,2 Hence, semiconductor nanowires (NWs) have partial pressure,24,25 growth temperature,26,27 the influence of
been extensively studied for fundamental science as well as for the substrate,28 and details of the nucleation.29 In these studies
potential applications in electronics, photonics, and as sensors.3 it is assumed that the growth mechanism follows the VLS
One of the most widely used growth methods of NWs is the so- concept where a liquid seed particle is involved. However,
called vapor−liquid−solid (VLS) process, because it offers growth of ZnO NWs should be different from VLS growth
excellent control over size, shape, and location of growth,4−7 because oxygen has virtually no solubility in gold, neither in the
which are most important issues for the integration of one- solid nor in the liquid state. Hence, oxygen cannot reach the
dimensional nanostructures into functional devices.8,9 In the catalyst−ZnO interface (growth interface) via diffusion through
VLS process, first proposed by Wagner and Ellis,10 atoms or the catalyst.
precursor molecules are supplied by the gas phase (vapor). The On the other hand, mechanisms of nucleation and growth
vapor species decompose and stick at a metallic seed particle have been intensively studied for NWs of elemental and III−V
acting as a catalyst. Soluble atoms alloy with the particle, and semiconductors including thermodynamic and kinetic aspects
finally a supersaturated eutectic liquid is formed. At the contact at the seed particle.11−14,30−35 In studies of catalyzed GaAs and
face with a substrate a solid NW nucleates and grows as long as InAs NW growth it was observed that the solubility and
vapor species are supplied. The VLS process has become a content of arsenic in the Au catalyst is very low, and it was
widely used method for fabricating one-dimensional nanostruc- concluded that the NWs grow via a solid-phase diffusion
tures from a rich variety of inorganic materials that include
mechanism, and hence with growth different from VLS.32,35
elemental semiconductors, compound semiconductors, sulfides,
and oxides.11−14 These results indicate that NW growth of compounds can be
Owing to unique semiconducting and piezoelectric proper- quite different from standard VLS growth of elemental
ties of ZnO15,16 as well as due to simplicity of fabrication, a semiconductors. Thermodynamic and kinetic aspects may
wealth of work has been published on the preparation, lead to alternative models of catalyzed NW growth including
structural characterization, and chemical and physical properties
of ZnO nanostructures grown by the VLS process in the past Received: August 10, 2012
decade.4,5,17−23 In nearly all of these studies, zinc vapor is Revised: December 12, 2012
generated by the reduction of ZnO with carbon, and oxygen is Published: December 19, 2012

© 2012 American Chemical Society 572 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design Article

surface diffusion models and have been discussed and growth experiments. For very low partial pressures of oxygen
proposed.36 (pO2 < 10−10 Pa), realized by setting the electrolysis cell voltage
In the literature, growth of ZnO NWs via VLS appears to be to zero, growth of structures of any type was not observed.
generally accepted. However, there is no proven evidence that Nanostructures grown with oxygen contents lower than 20
catalyzed growth of ZnO NWs has to follow the known ppm oxygen have a low density on the substrate, exhibit less-
mechanisms of VLS. Hence, a better understanding of the defined shapes, and partial formation of ZnO films is observed.
growth mechanisms is certainly desirable and may be helpful Oxygen contents between 40 ppm (pO2 = 4 Pa) and 100 ppm
for catalyzed growth of ZnO NWs together with dopants. In
this contribution we present a detailed and systematic study of (pO2 = 10 Pa) in the carrier gas resulted in well-defined
Au-catalyzed growth of ZnO NWs. The carbothermal method nanorods as shown in Figure 1. Oxygen contents higher than
enables a precise monitoring of the supply of zinc and oxygen
in the vapor. Methods of transmission electron microscopy
(TEM) are applied to clarify the growth situation at a
microscopic scale focusing (i) on structural details at the
catalyst surface and the catalyst−NW interface, and (ii) on the
chemical composition of the catalyst particle. Experimental
findings as well as thermodynamic considerations confirm that
alloying of the Au catalyst with Zn is not possible and that the
catalyst interior is solid. A kinetic transport growth model is
developed which is based on surface and interface diffusion and
hence is quite different from classical VLS growth.

2. EXPERIMENTAL SECTION
ZnO nanowires were grown in a furnace with two heating zones by a
vapor phase transport process. As source material, an equimolar
mixture of high purity ZnO powder (Sigma Aldrich 99.999%) and
graphite (Alfa Aesar 99.9995%) was placed in an alumina boat located
at heating zone I held at 1000 °C. Fused silica substrates coated with
Au were placed vertically at heating zone II at variable temperatures. Figure 1. SEM images of ZnO nanowires grown at different oxygen
The sputter-deposited Au particles, originally having diameters of 5− partial pressures: 15 ppm (a), 40 ppm (b), and 80 ppm (c, d).
10 nm, increased in size to several 10 nm by heating in zone II. The
vapor produced in zone I was transported to the substrate by an Ar gas
flow of 500 sccm at ambient pressure. The dried 5 N Ar gas is purified ca. 2000 ppm resulted in oxidation of the graphite in the source
with a titanium furnace built into the Ar feeding pipe. Oxygen is mixture, and growth of nanostructures was not observed.
generated by electrolysis of concentrated H2SO4 where the amount of Earlier reports on ZnO NW growth with “pure” argon as carrier
oxygen and hence the partial pressure can be precisely adjusted with gas can be concluded to be the result of oxygen impurities in
the current through the cell. The complete setup of the thermal CVD
growth system is shown in Figure S1 (Supporting Information). For
the argon gas bottle.
rapid cooling in some of the growth experiments a wire was fixed to Comparison of the lengths of NWs shown in Figure S2
the substrate to quickly (ca. 0.5 s) pull it out of the furnace into a cup (Supporting Information) proves the growth rate to clearly
with petroleum. depend on the supply of oxygen in the carrier gas. After
Nanowire structures on the substrate were examined by field reaction time of 20 min and at oxygen contents of 40 and 80
emission scanning electron microscopy (FESEM, JEOL JSM 6400F ppm, the NWs are ca. 6.5 and 42 μm in length, respectively. For
and LEO Supra 55). Transmission electron microscopy (TEM) and length determination ca. 10 of the longest NWs were measured,
electron diffraction studies were conducted with a Philips CM30ST because they can be presumed to have nucleated at the very
transmission electron microscope at 300 kV, and for chemical beginning of the growth experiment. This assumption is
microanalysis by energy-dispersive X-ray (EDX) spectroscopy a Si(Li)
X-ray detector fitted to the electron microscope was used. High
supported by the result of length measurements for 5, 10, and
resolution TEM (HRTEM) and electron energy-loss spectroscopy 20 min reaction time t shown in Figure 2. The lengths l as
(EELS) were performed in a Philips CM300UT field-emission
electron microscope (FEG-TEM) with postcolumn imaging filter
(Gatan Imaging Filter, GIF) operated at 297 kV. For TEM studies the
NWs were separated from the substrate by ultrasonic treatment in
methanol. A drop of the suspension on a holey carbon film supported
by a copper grid was dried in air and used for the TEM studies. To
study isolated Au particles by TEM and EDX spectroscopy the ZnO
NWs were dissolved in dilute (7%) HCl solution. The remaining Au
particles were dispersed in ethanol, and the suspension was dropped
and dried onto a carbon film supported by an aluminum TEM grid.

3. RESULTS
To investigate the optimum growth conditions of ZnO NWs,
the oxygen content of the Ar carrier gas and the temperature of
deposition were varied. Best conditions for ZnO NW growth Figure 2. Lengths of longest ZnO NWs measured after growth for 5,
were observed between deposition temperatures TII of 870 and 10, and 20 min. Linear fit results in a growth rate of 5.6 nm s−1 for 40
940 °C. In the following, TII = 900 °C was chosen for all ppm O2 and 36 nm s−1 for 80 ppm O2.

573 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580


Crystal Growth & Design Article

function of time are on a straight line with slopes equal to the signal at E = 8.49 keV interferes with the Zn line. Hence, the
growth rate which meet at l = 0, t = 0 proving that the longest lines of the Zn_L emission (Zn_Lα1: E = 1.01 keV) are more
NWs started growing at the beginning of the growth appropriate for the detection of Zn. Neither by visual
experiment. With 40 ppm oxygen content NWs grow with a inspection nor by quantification of the spectra with the EDX
rate of ca. 5.5 nm s−1, with 80 ppm oxygen content the growth analysis software an X-ray signal of Zn could be detected, and
rate is ca. 36 nm s−1. A volume of 1 nm3 ZnO contains 42 the detection limit is ca. 0.3 atom %. This means that Zn is
atoms of Zn and O. Hence, at a growth rate of 5.5 nm s−1 (36 virtually not present in the volume of the catalyst during growth
nms−1) 231 atoms (1.512 atoms) of Zn and O are deposited of the NW.
and attached to a 1 nm2 area of a growing ZnO NW. Imaging of ZnO NWs in the electron microscope yields
Furthermore, it is worth mentioning that the growth rate is further information on the processes at the Au catalyst particles.
virtually independent of the NW diameter and that all Electron microdiffraction was used to determine the absolute
nanostructures are grown via a VLS mechanism proven by growth direction which is [0001] of the polar ZnO NWs (see
Au particles on top of the wires. Figure S3, Supporting Information).40 The NWs grow in the
An important issue is the yield of Zn used for the growth of positive direction of the ZnO c axis, and hence the Au catalyst
NWs as part of the amount of Zn vapor delivered with the particles are on top of the zinc-terminated face of ZnO NWs.15
carrier gas. For that purpose the mass loss of the source mixture Au particles of moderately cooled samples are clearly faceted,
is determined, which is ca. 25 mg/h. The products of the whereas the particles of quenched samples are either roundish
carbothermal reaction of the source mixture, in shape or they exhibit small facets and round edges as shown
ZnO(s) + C(s) → Zn(g) + CO(g) (1) in Figure 4. At many NWs a specific geometry at the interface
between the growing ZnO NW and the Au catalyst is observed.
are essentially Zn and CO in equal molar quantity, and the Such configurations are shown in Figure 5, where the ZnO
mass loss of ZnO is calculated to be 17.5 mg/h = 7.5 × 10−8
mol/s. The amount of Ar fed into the tube is 3.7 × 10−4 mol/s.
From that, the partial pressure of Zn and CO is determined to
pZn = pCO = 20 Pa. Kinetic gas theory gives the number z of
atoms impacting per area and time as z = n/4v,̅ where n is the
number of atoms/molecules in a standard volume and v ̅ is the
mean velocity of the gas species. In a gas at 1 bar the number n
of gas species producing 20 Pa partial pressure corresponds to n
= 5 × 10−6 nm−3, and Zn vapor at 900 °C yields v ̅ = 616 m/s =
6.16 × 1011 nm/s. As a result, z = 770.000 Zn atoms are hitting
a surface area of 1 nm2 in one second.
The process of VLS growth of semiconductor nanostructures
was originally explained by supersaturation of the liquid catalyst
with the growth material.10,37−39 The growth material is alloyed
and has to be present in the catalyst particle also after cooling.
Hence, the chemical composition of the catalyst particle gives
valuable information on the processes associated with the
catalyst function. We have probed the content of Zn by EDX
analysis in the transmission electron microscope in ca. 50 Au
catalyst particles. A particle isolated by dissolving the entire
ZnO NW is shown in Figure 3 together with the corresponding
EDX spectrum. The Zn_Kα line at E = 8.64 keV would be
most appropriate for EDX analysis; however, a nearby Au_Lλ

Figure 3. EDX spectrum and TEM image of an isolated Au catalyst


particle. The Zn content of the catalyst particle is below the detection Figure 4. TEM images of ZnO NWs with faceted Au catalyst particles
limit of ca. 0.3 atom % as shown by the missing Zn_L line at ionization of moderately cooled (a) and nonfaceted particles of quenched
energy E = 1.01 keV (see inset). samples (b).

574 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580


Crystal Growth & Design Article

Figure 5. TEM image of frequently observed interface geometry


between ZnO NW and Au catalyst where outer regions have grown
further than inner regions of the NW (a, b). Steps at the interface are Figure 6. HRTEM image of interface region between ZnO NW and
observed in the highly magnified image (c) of the boxed area in (b). Au catalyst quenched to RT (a). Fourier transforms of ZnO NW (b),
Note that ZnO is sensitive to electron radiation visible at the NW interface region (c), and Au catalyst (d) indicate crystal overlap in
close to the interface. projection as illustrated schematically in (e).

NWs have grown further at outer regions of the reaction face


than in the center of the wire, and the catalyst appears to be
sunk into the NW. This is also shown and illustrated in Figure 6
where epitaxially grown ZnO can be detected in lattice images
around the Au particle even some 10 nm away from the bottom
face of the Au catalyst. The high-resolution TEM image,
considered as a projection of the crystal structures, shows
overlapping images of Au and ZnO which confirms the
geometry shown in Figure 6e.
A further observation at Au particles  rapidly quenched or
moderately cooled to RT  is a noncrystalline surface layer
with thickness in the range of 1−2 nm (Figure 7). The layer
thickness does not depend on the volume of the Au particle
shown at catalyst particles with different sizes. The chemical
composition of the surface layer was determined by EELS using
an electron probe of ca. 5 nm diameter at 20 catalyst particles.
The accumulated EEL spectrum of seven single spectra is
shown in Figure 8. The quantification of the intensities of the
Zn_L2,3 edge vs that of the O_K edge yields the composition of
ZnO. Hence, the ZnO surface layer is likely to be formed by
oxidation of some Zn content at or near the surface of the Au
catalysts upon removal of the substrate out of the furnace into
the air. Figure 7. (a−c) Gold catalyst particles of different sizes and volume
exhibit a surface layer of virtually the same thickness (1−2 nm)
4. DISCUSSION proving the layer to be generated by a surface process.
The conditions for the growth of ZnO NWs in our
experiments, i.e., growth temperature, oxygen content of the literature.4−6,17,20 Note that in earlier studies oxygen was not
carrier gas, and Zn vapor supply by carbothermal reduction of added to the carrier gas; there it was just an impurity in
ZnO, are very similar to those of many previous studies in the commercially supplied argon gas. With equilibrium thermody-
575 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design Article

concentration xZn = 12.5 atom % (Figure S4, Supporting


Information). The equilibrium Zn partial pressure pZn(T) of
that liquid is given by
0
pZn (T ) = a Zn(T ) ·pZn (T ) (2)
where p0Zn(T) is the partial pressure of pure Zn and aZn(T) is
the activity of Zn in the liquid alloy. Since NW growth is close
to the boiling temperature 1184 K of Zn, the equilibrium
pressure over liquid Zn is high, i.e., p0Zn(1173 K) = 8.7 × 104
Pa.43 In Figure S5 (Supporting Information) the Zn activity is
extrapolated for x = 12.5 atom % from the activities given by
Ipser et al. resulting in aZn(1173 K) = 2.4 × 10−4.44 The
equilibrium Zn pressure is determined from eq 2 as pZn(1173
K) = 21 Pa. It is worth noting that the activity and hence the Zn
pressure of a Au−Zn liquid with xZn = 18 atom % is already 60
Pa, and Zn would even evaporate from that liquid alloy to
constitute its equilibrium partial pressure.
In thermodynamics of crystal growth, the Gibbs energy of
formation ΔGf is the driving force for the reaction of the
constituents, which is for ZnO NWs
Zn(Au, l) + 1/2 O2 (g) → ZnO(s), T = 1173 K (3)
The standard Gibbs energy of formation of ZnO(s) at 1173 K
is ΔG0f = −217 kJ/mol, calculated for liquid Zn.45 For a Au−Zn
liquid alloy with xZn = 12.5 atom % this value has to be
corrected for the Gibbs energy of mixing, ΔGm = −12 kJ/mol,
which has to be taken as a positive value for Zn
exsolution.44,46,47 Hence, the Gibbs energy of formation of
the reaction 3 is ΔGf = −205 kJ/mol, which proves a very high
driving force for ZnO NW formation. A comparison with III−V
Figure 8. EEL spectrum accumulated from seven individual spectra at semiconductors can be made with the standard enthalpy of
Au catalyst surfaces proving that the surface layers consist of ZnO (a). formation ΔH0f (ZnO) = −350.5 kJ/mol versus ΔH0f (GaN) =
TEM image of one catalyst with electron probe position indicated at −110.5 kJ/mol and ΔH0f (GaP) = −88 kJ/mol.45 The energies
surface region (b). associated with the precipitation of Si (and Ge) wires from a
eutectic melt with a metal in the VLS process are even lower.48
Hence, the resulting growth rates are much lower than for ZnO
namic calculations using the code CVTRANS we can show the NW growth.
products of the carbothermal reaction (eq 1) to be essentially The oxygen partial pressure was shown to strongly influence
Zn vapor and CO over solid ZnO and carbon.41 This is in full
the ZnO NW growth rate. The increase of pO2 from 4 to 8 Pa
agreement with the well-known Boudouard equilibrium of C,
CO, and CO2 at high temperatures. The content of CO2 is results in an increase of the growth rate by a factor of ca. 7,
smaller than that of CO by a factor of 1.000 and hence is a whereas the Zn vapor supplied by the carrier gas is kept
minor product not essential for the growth process. virtually constant. In fact, the higher growth rate of the ZnO
Thermodynamic Considerations of ZnO NW Growth. NW requires more of the impacting Zn atoms to stick at the
In our study, the Zn content in the carrier gas could be well surface of the Au catalyst which, at first sight, seems
estimated from the mass loss of the source material. The high contradictory in view of a higher pO2 as that might suppress
number of impacting Zn atoms (770.000 nm−2 s−1) onto any the deposition rate or the surface diffusion.49 The effect of the
surface in the reaction tube and hence on the Au catalyst seems oxygen content on the growth rate can, however, be
to be in contradiction to the relatively small number of Zn understood having a view on the associated kinetics and
atoms (231 and 1.512 atoms nm−2 s−1 at 40 and 80 ppm O2, thermodynamics at the Au catalyst: The Au−ZnO interface is
respectively) used for ZnO NW growth. Here the larger surface an effective sink for the Zn delivered via the Au catalyst and the
of the Au catalyst compared to the growth interface is not even Zn is finally consumed for ZnO NW growth. (i) A higher
considered. Additionally, it is known that the sticking growth rate implies an enhanced consumption of Zn from the
coefficient, i.e., the number ratio between impacting atoms Au catalyst which is very likely to reduce the Zn concentration
from the gas phase and atoms attached to the surface, is close to there. (ii) With the same thermodynamic arguments as used
1 on a liquid surface. The small number of Zn atoms effectively above, the equilibrium Zn partial pressure over the Au−Zn
sticking and being chemisorbed at the surface of the Au catalyst catalyst is lowered. (iii) As a result, the Zn vapor supplied by
can be explained well by thermodynamic considerations: the the gas phase exceeds the equilibrium pressure over the catalyst,
equilibrium Zn partial pressure over a Au−Zn melt has to be and the system will react to increasing the Zn content which is
compared with the Zn partial pressure of ca. 20 Pa supplied by realized by a higher number of Zn atoms sticking to the Au
the carbothermal reaction. The Au−Zn liquid with lowest Zn catalyst. (iv) The amount of Zn transferred by the Au catalyst
content is at the concentration of the liquidus boundary in the to the Au−ZnO heterointerface will increase until steady state
Au−Zn phase diagram.42 At 1173 K this is a liquid with Zn growth conditions, i.e., equal rates of supply of Zn and
576 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design Article

consumption for ZnO formation, are reached. Thereby, the state conditions must exist at the catalysts surface demanding
amount of Zn at the Au catalyst is regulated by the oxygen equal rates of Zn incorporated from vapor and Zn transported
content in the carrier gas, and the mechanism is kinetically away to be consumed for ZnO crystallization and NW growth.
controlled over a wide range of pO2. If the rate of Zn uptake were higher than that of Zn
State of Catalyst Particles. In this context, the state of the consumption, excess Zn would alloy with the catalyst. It seems
Au catalyst during growth is of general interest for the path of that the surface layer takes up a limited amount of Zn and
Zn atoms migrating to the growing Au−ZnO interface. readily compensates for consumption losses by a higher
According to the Au−Zn phase diagram,42 the Zn content incorporation rate of Zn from vapor. On the basis of the
has to be at least 12.5 atom % for a completely liquid catalyst experimental results we may discuss the transport kinetics of
droplet. For a solid catalyst with a liquid surface layer it would the Zn atoms to the reaction front as well as the growth kinetics
be sufficient that the overall composition is in the two-phase of the ZnO NWs on a microscopic scale. During growth, the
field between solidus and liquidus, i.e., between xZn = 8.5 atom Zn concentration gradient or difference between the Au catalyst
% and xZn = 12.5 atom %, respectively. The catalyst surface surface and the reaction interface is the driving force for
where the Zn is supplied will exhibit the liquidus composition diffusional transport of Zn atoms. They may be transported (i)
and the interior has the solidus composition (Figure S4, via volume diffusion through the solid catalyst or (ii) along the
catalysts surface along a thin liquid alloy film and by surface
Supporting Information). However, as proven by experiment,
diffusion. The difference in diffusion coefficients in the two
the Au particles essentially do not contain Zn which has two
regions will be decisive for the kind of path the Zn atoms
consequences: (i) the interior of the catalyst particles has to be
prefer. For Zn diffusion in solid Au the diffusion energy Q =
solid during growth at 1173 K, and (ii) there is no thermal
1.64 eV and the pre-exponential factor D0 = 0.082 × 10−4 m2
equilibrium established at the catalyst particles. This leads to
s−1 are known which yields a diffusion coefficient of D(1173 K)
the conclusion that the solid Au particles feature a thin liquid
= 7.4 × 10−13 m2 s−1.55,56 There are no diffusion data available
alloy film at the surface by impacting Zn atoms supplied by the
for Zn in a liquid Au−Zn alloy or for Zn surface diffusion on a
gas phase. This is also reasonable since a liquid alloy film Au surface. However, typical diffusion coefficients in liquids are
produces isotropic surface tensions and will not lead to faceting in the range of 10−8 m2 s−1.57 Apart of the diffusion constant,
which was observed at quenched catalyst particles. However, the transport rate depends on the length and the cross section
annealing of previously grown ZnO NWs at 1173 K on air and of the diffusion path. The paths through the interior and
subsequent quenching to RT also did not produce faceting of around the catalyst particle are similar, independent of the
Au catalyst particles. Obviously, the surface energy anisotropy starting point of Zn atoms: If we consider the Au particle to be
at that temperature is substantially decreased and is not a cube with edge length a having a surface film with thickness b,
sufficient to prove a liquid film at the catalyst surface.50 the cross section for diffusion through the interior is a2 and the
A further reason for the existence of a liquid film on Au one along the surface film is 4 ab, which results in a/4b as the
catalyst particles may be premelting or surface melting. This ratio of the two cross sections. For a = 50 nm and b = 2 nm the
phenomenon, known since the 1950s and observed at ratio is ca. 6, and even for a large particle with a = 100 nm and a
nanoparticles of various metals, may occur well below the thin film with b = 1 nm the cross-section ratio is only 25. The
bulk melting temperature as a result of extreme surface large difference in diffusion constants indicates that the
curvatures.51−54 Premelting of entire Au particles with preferred path of Zn atoms is along a liquid alloy film at the
diameters of 10 nm at 1225 K, more than 100 K below catalyst particles surface which is consistent with the
Tm, Au = 1337.6 K, was observed by Buffat and Borel.52 Frenken observation of the catalyst interior being virtually free of Zn.
et al. observed surface melting of ca. 5 monolayers even on a The length increase of zinc oxide nanowires should be
flat [110] surface of a Pb crystal 150 K below Tm.53 explained by nucleation and growth on a molecular or atomic
Qualitatively, premelting of an alloy may be understood as scale where the reacting constituents or monomers are Zn
temperature reduction of solidus and liquidus lines which is atoms delivered by catalyst surface regions and oxygen in the
displayed schematically for the Au−Zn system in Figure S4 vapor. The activities of the monomers are constant everywhere;
(Supporting Information). Assuming premelting at 60 K below however, the existing ZnO nanowire provides appropriate sites
Tm,Au, a Au alloy particle containing somewhat more than 5 for ZnO nucleation. It was reported that ZnO nanostructures
atom % Zn may exhibit a molten surface layer at 1173 K. which were grown via VLS along [110̅ 0] thicken during growth
Therefore, in addition to alloying with Zn, surface melting may and finally grow to nanosails.28 At ZnO NWs grown in [0001],
further enhance the formation of a liquid layer of Au(Zn) on the prism side faces, however, do not thicken at all. The fast
the catalyst particles. It is worth noting that NWs of other metal growing zinc-terminated [0001] face in contact with the Au
oxides may well grow via VLS with a liquid alloy catalyst catalyst offers favorable nucleation sites, and it is there, at the
particle. We have produced In2O3 NWs which grow with a Au− junction of the three phases  the vapor, the liquid (catalyst
In catalyst with In contents up to 30 atom %, and which are surface) and the solid (ZnO)  where nucleation of new layers
definitely liquid at 1173 K according to the phase diagram. This starts: at this triple phase junction the activities of Zn and
observation can be explained by the equilibrium partial pressure oxygen are higher than at regions further inside the Au−ZnO
of In over liquid In which is lower by ca. 105 than the Zn partial interface. The function of the gold may be 2-fold: besides
pressure at the same temperature (1100 K).43 Hence, the In trapping Zn from the vapor, it may also act as a catalyst
content supplied by the gas phase is much higher than the weakening the double bond of oxygen molecules and hence to
equilibrium pressure whereby the In is alloyed in the Au facilitate the dissociation of O2. Then, activated oxygen
catalyst particles. molecules or atoms together with Zn atoms may diffuse
Growth Mechanism. It has been shown that Zn does not along the Au−ZnO interface to react and nucleate to ZnO at
alloy in the catalyst interior and is present only at surface appropriate sites. These sites will be steps at the interface or
regions of the catalyst particles. As mentioned above, steady even kinks where two steps meet, similar as at free surfaces
577 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design Article

where ad-atoms preferentially attach. Scenarios of the ZnO NW as zinc and oxygen. (ii) Some of the ZnO NWs have an
growth mechanism are illustrated in Figure 9, starting with the interface shaped as a bowl with the catalyst as counterpart (c.f.
Figure 5). This geometry is generated when lateral growth of
ledges is not yet completed and new ledges already nucleate on
top of the existing ones at the triple phase junction. The reason
for this may be a hindered diffusional transport of species at the
interface. At conditions where nucleation only occurs at the
outer rim of the catalyst particle it is possible to even grow ZnO
nanotubes.19 The formation of the bowl-shaped and faceted
interface geometry is likely to be caused by the special
crystallographic orientation relationship between ZnO NW and
Au catalyst resulting in interfaces having different transport
properties than general interfaces. A high resolution TEM
micrograph of the NW in Figure 5b is shown in Figure 10,

Figure 9. Schematic of zinc and oxygen paths at Au catalyst (a).


Nucleation at triple phase junction and lateral growth of ZnO ledges
by diffusion of species along the Au−ZnO interface and attachment to
steps. Schematics of the diffusion path of oxygen and Zn species along
the Au−ZnO interface (b).

nucleation of a new ZnO ledge around the edge of the NW.


The minimum ledge height may have the size of a ZnO4
tetrahedron equivalent to c/2 of the ZnO unit cell.
Essentially, two different shapes of the catalyst−ZnO
interface were observed, and hence different growth modes
may be assumed. (i) Many of the ZnO NWs exhibit virtually
flat interfaces with the Au catalyst particle. Formation of this
geometry requires that the time for completion of a ZnO layer
by lateral ledge growth is shorter than the time scale for
nucleation of a new ledge at the edge. Rapid ledge growth Figure 10. High-resolution TEM micrograph of region between Au
requires fast diffusional transport of species along the Au−ZnO catalyst and ZnO NW exhibiting steps and ledges at the interface. The
interface. Similar as in polycrystalline materials where grain special orientation relationship of crystals is proven by Fourier
boundary diffusion is much faster than diffusion in the lattice, transforms of regions 1 and 2.
diffusion of Zn and O atoms is preferred at the Au−ZnO
interface for the following reasons: interfaces between noble
metals and oxides exhibit weak chemical bonding, characterized where the lattices of Au in ⟨110⟩ and ZnO in a axis orientation
by high interfacial energies and small work of adhesion known ⟨211̅ 0̅ ⟩ are imaged. The orientation relationship is special, given
from contact angle measurements.58,59 Furthermore, interfacial by
energy is a function of the interface geometry or orientation {111}Au (0001)ZnO and ⟨110⟩Au ⟨2 1̅ 1̅ 0⟩ZnO
relation and hence of the atomic structure.58,60 General
heterophase interfaces are created when the Au crystals are The interface is formed by a {111} plane of Au and the (0001)
not in a special orientation with the underlying ZnO crystal face of the ZnO crystal, which both are closed-packed planes.
which is the statistically most probable situation. Owing to this, Despite a lattice mismatch of 12%, the interface exhibits a high
high-indexed planes of the metal with only a small number of density of coincidence sites and appears dense as illustrated in
atoms are in contact with the oxide crystal as schematically Figure S6b. As a consequence, diffusional transport along this
shown in Figure S6a (Supporting Information). However, interface can be imagined to be much more difficult as along a
general interfaces provide good diffusion paths for species such general interface which is more “open”. This observation may
578 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design Article

serve as proof for the effect of interface geometry and atomic (3) The growth rate of ZnO NWs is hardly influenced by the
structure to transport properties. amount of Zn vapor since it usually does not exceed the partial
A similar observation of step formation and ledge growth was pressure over Au−Zn alloy catalysts. However, the oxygen
made by in situ TEM experiments at Si NWs growing by a content regulates the consumption of the Zn for ZnO NW
vapor−solid−solid (VSS) mechanism where the catalyst growth: increase of oxygen results in a higher consumption of
particles are solid.30,31 Different from the ZnO−Au system Zn which decreases the Zn concentration at the catalyst surface
the supply of Si atoms occurs over the complete heterointerface and leads to a higher uptake of Zn from the vapor and finally to
by supersaturation of the catalyst. In this context, the absolute a higher growth rate of the NWs.
values of growth rates of Si and Ge NWs provide a further (4) The transport of Zn from the vapor to the reaction zone
argument for a liquid phase being involved in catalyzed growth at the catalyst−NW heterointerface is concluded to take place
of ZnO NWs. While Ge NWs grow from a liquid AuGe droplet at the Au catalyst surface via a liquid Au−Zn surface layer
(VLS) with a rate of 0.11 nm s−1, the rate drops to 0.013 nm followed by diffusion along the Au−ZnO interface. Obviously,
s−1 when the catalyst solidifies (VSS) at virtually the same oxygen does not react with Zn at the catalyst surface. It may be
temperature.12,31 The authors suggest that the reduced concluded that O2 entering the Au−ZnO interface dissociates
transport kinetics as well as the reduced sticking probability to oxygen atoms which then facilitates the diffusional transport.
at the catalyst surface is responsible for slowing down growth in
The preferred nucleation site of new ZnO unit cells is the outer
the VSS process. If we consider the high growth rates of ZnO
periphery of the catalyst−NW heterointerface where the
NWs ranging between 5 to 36 nm s−1, it is difficult to imagine
oxygen activity is highest, and, as a start, ZnO ledges are
that there should be no liquid phase involved. A further issue is
the formation of a reduced diameter or “neck” of the NW at the formed there. Growth of the ledges toward inner interface
triple phase junction often observed at III−V semiconductor regions proceeds by diffusion of oxygen and zinc along the
systems.32 The neck appears to be important in forming heterointerface. The interface geometry, planar or bowl-shaped
geometry with contact angles where the three energies γSL of and stepped, is the result of the crystallographic orientation
the solid−liquid interface, γLV of the liquid−vapor interface, and relationship between ZnO NW and Au catalyst resulting in
γSV of the solid−vapor interface are close to equilibrium. differing atomic structures of the interfaces, which then possess
Whereas VLS growth rates of III−V semiconductor NWs are as different transport properties.
low as of Si and Ge NWs and obviously allow formation of (5) Catalyzed growth of NWs of any metal oxide must be
equilibrium geometries at the triple phase junction, the quite different from classical VLS growth where the
situation will be certainly different in ZnO NW growth. The components are soluble in the liquid catalyst. Since oxygen is
high growth rates prove a rapid delivery of Zn and oxygen right not soluble in Au or is soluble in very small amounts in other
to the triple phase junction, and such kinetically dominated metals, it can be supplied only at the triple phase line where
growth conditions will prevent formation of equilibrium new metal oxide units nucleate. Growth proceeds by a ledge
geometries at most of the NW−catalyst interfaces. mechanism, whereas the constituents of the oxide have to
Cheyssac et al. discussed different growth mechanisms of diffuse along the catalyst−NW interface. In the case of NWs
nanowires together with related issues where one is a surface where the constituent metal is soluble in the catalyst, the
diffusion model inspired by the observations of Persson et al. particle may be liquid at growth temperature. Nevertheless,
where solid Au catalyst particles on GaAs NWs were grown by NW growth will not be homogeneous because of the gradient
chemical vapor deposition.32,36 The As concentration in the of oxygen activity along the catalyst−NW interface. This
catalyst particle was shown to be very low, and transport of As conclusion may be generalized to any system where one
was suggested to proceed along the catalyst surface and through component of the NW is not soluble in the catalyst.


the interface between catalyst and GaAs NW.32 This scenario
resembles very much the situation of Zn and O in ZnO NW ASSOCIATED CONTENT
growth. Furthermore, in a recent study by Bao et al.29 with a
focus on nucleation and kinking mechanisms of ZnO NWs, *
S Supporting Information

elemental mapping using EDX shows a very low signal of Zn Figure S1. Schematic of the thermal CVD growth system.
inside the Au catalyst which confirms our findings of catalysts Figure S2. SEM images of ZnO nanowires grown for different
consisting of virtually pure Au. The Zn signal as well as the O times. Figure S3. TEM image, electron diffraction pattern,
signal29 is likely to originate from the ZnO scale which we selected area electron diffraction (SAED) pattern, micro-
observe on all of the Au catalyst particles. diffraction pattern. Figure S4. Au-Zn phase diagram. Figure
S5. Activity of Zn in the liquid Au-Zn phase. Figure S6. Atomic
5. CONCLUSIONS configurations of Au crystals on a (0001) face of ZnO,
(1) The carbothermal reduction of ZnO essentially produces unrelaxed models. Figure S7. Examples of planar interfaces
CO and Zn vapor. The formation of ZnO NWs by the generated by general orientations between Au catalyst and ZnO
oxidation of Zn vapor requires a supply of additional oxygen NW. This material is available free of charge via the Internet at
which can be precisely added to the carrier gas by electrolysis. http://pubs.acs.org.
(2) The catalyzed growth of ZnO NWs is characterized by
the high partial pressure of Zn over Au−Zn alloys at typical
growth temperatures which prevents alloying of the Au catalyst.
■ AUTHOR INFORMATION
Corresponding Author
Therefore, the catalyst particles do not melt at the growth
*E-mail: mader@uni-bonn.de.
temperature; only thin surface regions may become a liquid
alloy, shown by a zinc free catalyst and by a uniformly thin ZnO Notes
scale after removal of the furnace. The authors declare no competing financial interest.
579 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580
Crystal Growth & Design


Article

REFERENCES (36) Cheyssac, P.; Sacilotti, M.; Patriarche, G. J. Appl. Phys. 2006,
100, 44315.
(1) Xia, Y.; Yang, P.; Sun, Y.; Wu, Y.; Mayers, B.; Gates, B.; Yin, Y.; (37) Westwater, J.; Gosain, D. P.; Tomiya, S.; Usui, S.; Ruda, H. J.
Kim, F.; Yan, H. Adv. Mater. 2003, 15, 353−389. Vac. Sci. Technol. B 1997, 15, 554−557.
(2) Li, Y.; Qian, F.; Xiang, J.; Lieber, C. M. Mater. Today 2006, 9, (38) Wu, Y.; Yang, P. Chem. Mater. 2000, 12, 605−607.
18−27. (39) Zhang, Y.; Zhang, Q.; Wang, N.; Yan, Y.; Zhou, H.; Zhu, J. J.
(3) Yang, P.; Yan, R.; Fardy, M. Nano Lett. 2010, 10, 1529−1536. Cryst. Growth 2001, 226, 185−191.
(4) Fan, Z.; Lu, J. G. J. Nanosci. Nanotechnol. 2005, 5, 1561−1573. (40) Mader, W.; Rečnik, A. Phys. Status Solidi A 1998, 166, 381−395.
(5) Fan, H. J.; Lee, W.; Huaschuild, R.; Alexe, M.; Le Rhun, G.; (41) Binnewies, M.; Glaum, R.; Schmidt, M.; Schmidt, P. Chemical
Scholz, R.; Dadgar, A.; Nielsch, K.; Kalt, H.; Krost, A.; Zacharias, M.; Vapor Transport Reactions; Walter de Gruyter: Berlin, 2012.
Gösele, U. Small 2006, 2, 561−568. (42) Okamoto, H.; Massalski, T. B. Bull. Alloy Phase Diagrams 1989,
(6) Zacharias, M.; Subannajui, K.; Menzel, A.; Yang, Y. Phys. Status 10, 59−69.
Solidi B 2010, 247, 2305−2314. (43) Barin, I.; Knacke, O. Thermochemical Properties of Inorganic
(7) Tomioka, K.; Ikejiri, K.; Tanaka, T.; Motohisa, J.; Hara, S.; Substances; Springer; Verlag Stahleisen: Berlin, 1973.
Hiruma, K.; Fukui, T. J. Mater. Res. 2011, 26, 2127−2141. (44) Ipser, H.; Krachler, R.; Komarek, K. L. Z. Metallkunde 1988, 79,
(8) Cui, Y.; Lieber, C. M. Science 2001, 291, 851−853. 725−734.
(9) Lu, W.; Lieber, C. M. Nat. Mater. 2007, 6, 841−850. (45) Lide, D. R. CRC Handbook of Chemistry and Physics: A Ready-
(10) Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4, 89−90. Reference Book of Chemical and Physical Data, 75th ed. (1994−95 ed.).;
(11) Kodambaka, S.; Tersoff, J.; Reuter, M. C.; Ross, F. M. Phys. Rev. CRC Press: Boca Raton, FL, 1994.
Lett. 2006, 96, 96105. (46) Gerling, U.; Predel, B. Z. Metallkunde 1980, 71, 79−84.
(12) Kodambaka, S.; Tersoff, J.; Reuter, M. C.; Ross, F. M. Science (47) Prasad, R.; Bienzle, M.; Sommer, F. J. Alloys Compd. 1993, 200,
2007, 316, 729−732. 69−74.
(13) Schwarz, K. W.; Tersoff, J. Phys. Rev. Lett. 2009, 102, 206101. (48) Schmidt, V.; Wittemann, J. V.; Gösele, U. Chem. Rev. 2010, 110,
(14) Dick, K. A. Prog. Cryst. Growth Charact. Mater. 2008, 54, 138− 361−388.
173. (49) Craciun, V.; Amirhaghi, S.; Craciun, D.; Elders, J.; Gardeniers, J.
(15) Ö zgür, Ü .; Alivov, Y. I.; Liu, C.; Teke, A.; Reshchikov, M. A.; G. E.; Boyd, I. W. Appl. Surf. Sci. 1995, 86, 99−106.
Doğan, S.; Avrutin, V.; Cho, S. J.; Morkoç, H. J. Appl. Phys. 2005, 98, (50) (a) Heyraud, J. C.; Metois, J. J. Acta Metall. 1980, 28, 1789−
41301. 1797. (b) Heyraud, J. C.; Métois, J. J. Surf. Sci. 1983, 128, 334−342.
(16) Klingshirn, C. F. Zinc Oxide: From Fundamental Properties (51) Takagi, M. J. Phys. Soc. Jpn. 1954, 9, 359−363.
Towards Novel Applications; Springer: Heidelberg; London, 2010. (52) Buffat, P.; Borel, J. P. Phys. Rev. A 1976, 13, 2287−2298.
(17) Huang, M. H.; Wu, Y.; Feick, H.; Tran, N.; Weber, E.; Yang, P. (53) Frenken, J. W. M.; Marée, P. M. J.; van der Veen, J. F. Phys. Rev.
Adv. Mater. 2001, 13, 113−116. B 1986, 34, 7506−7516.
(18) Huang, M. H.; Mao, S.; Feick, H.; Yan, H.; Wu, Y.; Kind, H.; (54) Lereah, Y.; Kofman, R.; Pénisson, J. M.; Deutscher, G.;
Weber, E.; Russo, R.; Yang, P. Science 2001, 292, 1897−1899. Cheyssac, P.; David, T. B.; Bourret, A. Phil. Mag. B 2001, 81, 1801−
(19) Kong, X.; Sun, X.; Li, X.; Li, Y. Mater. Chem. Phys. 2003, 82, 1819.
997−1001. (55) Neumann, G.; Tölle, V. Phil. Mag. A 1995, 71, 231−247.
(20) Wang, Z. L. J. Phys.: Condens. Matter 2004, 16, R829−R858. (56) Cardis, D. Der Einfluss eines elektrischen Gleichfeldes auf den
(21) Heo, Y. W.; Norton, D. P.; Tien, L. C.; Kwon, Y.; Kang, B. S.; Materietransport substitutionell gelöster Fremdatome in Gold. Ph.D.
Ren, F.; Pearton, S. J.; LaRoche, J. R. Mater. Sci. Eng., R 2004, 47, 1− Thesis; Westfälische Wilhelms-Universität Münster: Münster (Ger-
47. many), 1977.
(22) Mergenthaler, K.; Gottschalch, V.; Bauer, J.; Paetzelt, H.; (57) Cussler, E. L. Diffusion: Mass Transfer in Fluid Systems, 3rd ed.,
Wagner, G. J. Cryst. Growth 2008, 310, 5134−5138. reprinted; Cambridge University Press: Cambridge, 2009.
(23) Kim, H. W.; Kebede, M. A.; Kim, H. S.; Srinivasa, B.; Kim, D. Y.; (58) Sutton, A. P.; Balluffi, R. W. Interfaces in Crystalline Materials;
Park, J. Y.; Kim, S. S. Curr. Appl Phys. 2010, 10, 52−56. Clarendon Press; Oxford University Press: Oxford, 1995.
(24) Borchers, C.; Müller, S.; Stichtenoth, D.; Schwen, D.; Ronning, (59) Chatain, D.; Coudurier, L.; Eustathopoulos, N. Rev. Phys. Appl.
C. J. Phys. Chem. B 2006, 110, 1656−1660. (Paris) 1988, 23, 1055−1064.
(25) Ramgir, N. S.; Subannajui, K.; Yang, Y.; Grimm, R.; Michiels, R.; (60) Ernst, F. Mater. Sci. Eng., R 1995, 14, 97−156.
Zacharias, M. J. Phys. Chem. C 2010, 114, 10323−10329.
(26) Subannajui, K.; Ramgir, N.; Grimm, R.; Michiels, R.; Yang, Y.;
Müller, S.; Zacharias, M. Cryst. Growth Des. 2010, 10, 1585−1589.
(27) Wongchoosuk, C.; Subannajui, K.; Menzel, A.; Burshtein, I. A.;
Tamir, S.; Lifshitz, Y.; Zacharias, M. J. Phys. Chem. C 2011, 115, 757−
761.
(28) Jebril, S.; Kuhlmann, H.; Müller, S.; Ronning, C.; Kienle, L.;
Duppel, V.; Mishra, Y. K.; Adelung, R. Cryst. Growth Des. 2010, 10,
2842−2846.
(29) Bao, P.; Zheng, R.; Du, S.; Li, L.; Yeoh, W. K.; Cui, X.; Ringer, S.
P. Cryst. Growth Des. 2012, 12, 3153−3157.
(30) Wen, C. Y.; Reuter, M. C.; Tersoff, J.; Stach, E. A.; Ross, F. M.
Nano Lett. 2010, 10, 514−519.
(31) Ross, F. M. Rep. Prog. Phys. 2010, 73, 114501.
(32) Persson, A. I.; Larsson, M. W.; Stenström, S.; Ohlsson, B. J.;
Samuelson, L.; Wallenberg, L. R. Nat. Mater. 2004, 3, 677−681.
(33) Dick, K. A.; Deppert, K.; Mårtensson, T.; Mandl, B.; Samuelson,
L.; Seifert, W. Nano Lett. 2005, 5, 761−764.
(34) Kim, B. J.; Tersoff, J.; Kodambaka, S.; Reuter, M. C.; Stach, E.
A.; Ross, F. M. Science 2008, 322, 1070−1073.
(35) Dubrovskii, V. G.; Cirlin, G. E.; Sibirev, N. V.; Jabeen, F.;
Harmand, J. C.; Werner, P. Nano Lett. 2011, 11, 1247−1253.

580 dx.doi.org/10.1021/cg301640v | Cryst. Growth Des. 2013, 13, 572−580

You might also like