Au-catalized self-assembly of III-V semiconductor nanowires by metalorganic vapour phase epitaxy

II 
Vapour phase technologies for self-assembly of  
quasi-1D semiconductor nanostructures 
 
 
 
2.1 Self-assembly of quasi-1D nanostructures 
Semiconductor nanodevices based on quasi-1D nanostructures call for new 
methods of nano-scale manufacturing, as “top-down” processes 
1,2,3,4
, which are 
dominant in the current ULSI technologies of semiconductors, cannot be easily 
(sophisticated equipment or intrinsic limitations in spatial resolution) or 
economically (high production costs) applied to the new class of devices.  
Self-assembly of semiconductor nanostructures is an alternative “bottom-
up” approach that has many potentialities. Nanowires with designed dimensions 
and properties can be fabricated more easily by self-assembling methods because 
in such methods individual building blocks (atoms or molecules) interact each-
others in pre-defined ways, resulting in the spontaneous self-organization of size-
selected nanostructures 
5
.  
Any self-assembly nanotechnology process has the following advantages 
6 
1) it is a parallel process, a feature very important at the nano-scale to guarantee 
high productivity; 
2) it can make structures with sub-nanometer precision (typically in the 1-100 nm 
range); 
3) it is able to generate three-dimensional nano-scale architectures; 
4) the process can be easily altered by external forces and geometrical constraints 
to dynamically re-assemble/re-configure on-demand any nano-scale object. 
A classical example of a self-assembly process is the quantum dots (QDs) 
formation on semiconductor surfaces by metalorganic vapour phase epitaxy 
(MOVPE) or molecular beam epitaxy (MBE) using the Stranski-Krastanov growth 
mode 
7
. In this process a very thin layer of the desired material (a few nanometers) 
is grown epitaxially on a substrate with a different lattice constant and the strain in
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
25
the deposited layer gives rise to a self-reorganization, converting the continuous 
epilayer into a random array of nanoislands. Using this method, the self-organized 
growth of SiGe QDs in Si and InGaAs QDs in GaAs has been successfully studied, 
obtaining QDs diameters ranging from ten to a few hundred nanometers; by 
varying the growth parameters it is possible to tune the material self-assembly on 
the substrate controlling the dots size and distribution and the average spacing 
between adjacent dots 
7,8,9
. 
In the last few years several vapour phase methods have been exploited for 
growing quasi-1D nanostructures of semiconductor materials with different control 
over dimensions, shape and uniformity. These strategies are summarized in fig. 2.1 
and include 
10
 
1) the use of the intrinsically anisotropic crystallographic structure of a solid; 
2) the introduction of a liquid-solid interface to reduce a symmetry of a seed 
(vapour-liquid-solid process); 
3) the use of templates with 1D morphologies to directly grow 1D structures; 
4) quasi-1D self-assembly of 0D nanostructures. 
The first approach (fig. 2.1a) is suggested by the observation that many solid 
materials naturally grow into 1D nanostructures due to their highly anisotropic 
crystallographic structure (e.g. asbestos); it has been applied to the growth of 
chalcogens (mainly Se and Te) and chalcogenides 
10
.
 
The second one (fig. 2.1b), which uses a liquid-solid interface to reduce the 
symmetry of a crystal, is the most developed vapour phase approach for the growth 
of both elemental and compound semiconductors nanowires and will be described 
in more details in the following sections.  
Template-directed synthesis (fig. 2.1c) use a template (e.g. channels within a 
porous material, step edges present on the surfaces of a solid substrate, etc.) as a 
scaffold within (or around) which a different material is generated and shaped into 
a nanostructure with a morphology complementary to that of the template; these 
methods have been successfully applied to both metal and semiconductor nanowire 
growth 
10
. 
Four different approaches to grow semiconductor nanowires using a 
template are reported in fig. 2.2. Nanowires can be prepared by shadow evaporation 
against an array of V-grooves etched on (100) Si substrates 
11
 (fig. 2.2a). In another 
procedure, by supplying the precursor species at normal incidence using
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
26
 
 
Figure 2.1  Strategies for self-assembly of quasi-1D nanocrystals from the vapour: a) dictation 
by anisotropic crystallographic structure of a solid; b) confinement by a liquid droplet (vapour-
liquid-solid process); c) use of a template; d) self-assembly through hierarchical clustering of 0D 
nanostructures (Ref. 10). 
 
 
techniques based on vapour phase (e.g. MBE) it is possible to produce nanowires at 
the bottom of each V-groove 
12,13,14
 (fig. 2.2b). 
The cross-sections of multilayer films grown by MBE have also been 
exploited as template to grow patterns of quantum structures for many 
semiconductors 
15
 (fig. 2.2c). In this technique, called cleaved-edge overgrowth, a 
superlattice (e.g. of AlGaAs/GaAs) is grown by MBE, cleaved in situ to produce an 
atomically clean surface; then, epitaxials layers are grown by MBE on selected 
regions of the exposed surface producing quantum wires. Finally, layer-by-layer 
growth mode can be used to grow quantum wire structures through nucleation on 
step edges on the surface of a solid substrate (2.2d) 
16,17
.
 
Monodispersed colloids (fig. 2.1d) can act as building blocks for the 
formation of wire-like structures through self- or externally-manipulated assembly. 
Scanning probe microscopy is used to manipulate the nanoparticles; in addition, 
lateral capillary forces, optical tweezers, magnetic fields and electrostatic 
interactions have been also studied to direct the self-assembly of nanoparticles. 
However, the low speed of these method limits its application to large-scale 
production 
10
.
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
27
 
 
Figure 2.2  Template-directed approaches to the growth of semiconductor 1D nanostructures by 
shadow evaporation (a), reconstruction at the bottom of V-grooves (b), cleaved-edge overgrowth 
on the cross-section of a multilayer film (c) and templating against step edges on the surface of 
a solid substrate (Ref. 10).
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
28
2.2 The Vapour-Liquid-Solid method 
Among all vapour-based self-assembly technologies, methods using a 
nanometer-sized solid-liquid interface are the most successful for growing 
nanowires with single-crystalline structures and in large quantities. The first 
demonstration of one dimensional growth was made by Wagner et al.
18,19
 and as-
grown Si structures (which were in the micrometer range) were called “whiskers”, 
while the growth mechanism was named “vapour-liquid-solid” (VLS). This bottom-
up growth mechanism involves the vapour, a liquid and a solid phase and is based 
on the role of solvent of a metal-catalyst which is able to dissolve some or all of the 
atomic species composing the semiconductor. The metal solvent also named “the 
catalyst” forms with the crystalline material a liquid solution that becomes a 
preferred site for deposition; precipitation from the supersaturated liquid induces a 
unidirectional (i.e. perpendicularly to the solid-liquid interface) growth of the 
crystal
19
.
 
The process is illustrated in fig. 2.3 for the Si nanowires growth. When a 
liquid metal catalyst droplet is in contact with the crystal components the following 
phenomena may occur: if the crystal material is soluble into the liquid, the liquid 
supersaturates with material supplied from the vapour and the crystal grows by 
precipitation at the solid-liquid interface like in a LPE system. The anisotropy in the 
 
 
 
Figure 2.3 Steps of VLS mechanism for grow silicon nanowhiskers. (a) Deposition from the 
vapour on a liquid solution (Au-Si alloy) and (b) precipitation from the supersaturated liquid 
solution at the liquid-solid interface (producing 1D whisker growth) (Ref. 19).
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
29
solid-liquid interfacial energy consequently produces the unidirectional growth of 
the crystal. The VLS growth mode is therefore a two steps process: 
1) deposition from the vapour directly on a liquid solution in a vapour-liquid system 
and 
2) precipitation from the supersaturated liquid solution at the liquid-solid interface 
in a liquid-solid system. 
Gas kinetics allows to estimate the growth rate J (atoms/cm
2
s) of an 
extended crystal surface for a condensation reaction as: 
19
 
 
0
ασp
J=
2 πmkT
                                                (2.1) 
 
where α is the accommodation coefficient, 
0
0
p-p
σ=
p
 is the supersaturation of the 
vapour in which p and p
0
 are the vapour pressure and the equilibrium vapour 
pressure of the solid respectively, at temperature T. α is the fraction of impinging 
atoms accommodated on the growing surface, and depends on the state of surface 
and on σ; thus, the predicted growth rate J (see fig. 2.4) is different for an imperfect 
crystal surface (containing dislocations), a perfect crystal surface and a clean liquid 
surface (having unit accommodation coefficient). Below a critical supersaturation 
( σ
1
 or σ
2
, respectively) the growth rate is zero for the cases of crystalline surfaces 
and increases to the ideal value with increasing σ. On the contrary the deposition 
on clean liquid surfaces has an ideal trend also at low supersaturations.  
When a vapour deposition happens on a perfect crystal surface with a local 
area covered by a liquid solution (e.g. formed by alloying a suitable agent with the 
crystal substrate), for σ< σ
2
 the deposition is possible on the liquid (with an ideal 
rate) while on crystal surface no crystal growth occurs (see fig. 2.4). The liquid 
covered surface of the substrate is thus a preferred site for the deposition within 
the range σ< σ
3
. Similar considerations can be made for an imperfect crystal 
surface. These predictions were made for evaporation-condensation and for 
homogeneous vapour phase reactions; for heterogeneous reactions the liquid 
droplet in the deposition region can act as a catalyst for the process, enhancing the 
deposition rate on the liquid. During the VLS process a flux of precursors reaches 
the liquid solution from its surface and the resulting concentration gradient in the 
solution provides the driving force for diffusion of material components from the 
vapour-liquid interface to the liquid-solid one. Diffusion controls the material
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
30
 
 
Figure 2.4 Growth rates of a crystal on a clean liquid (unit accomodation), an imperfect crystal 
and a perfect crystal (Ref. 19). 
 
 
transport in the liquid under isothermal conditions. The size of the catalyst droplet 
determines the nanowire diameter, while the vapour concentration of precursors 
and growth temperature influence the growth rate. 
During the VLS growth, the liquid solution must be stable imposing a 
limitation on the minimum diameter of a VLS crystal. The degree of 
supersaturation σ influences the stability of a liquid droplet of curvature r. The 
minimum critical radius has the following expression 
19
 
 
LV L
min
2 σ V
r=
RTln σ
                                                  (2.2) 
 
where σ
LV
 is the liquid-vapour interfacial energy, V
L
 is the liquid molar volume and 
R and T are the ideal gas constant and absolute temperature, respectively. Most 
materials have r
min
≈10
-5
 cm for σ=1.02. In order to obtain sub-micron dimensions of 
the whiskers, it is necessary to increase the supersaturation of the material 
components in the liquid catalyst.
II – Vapour phase technologies for 1D self-assembly 
__________________________________________________________________________________ 
31
2.2.1 Criteria for metal catalyst selection 
An agent (impurity) suitable for VLS growth must 
19 
1) form a liquid solution at the growth temperature with some of metal elements 
composing the semiconductor nanowire; 
2) have a liquid-solid distribution coefficient at the growth temperature k=C
s
/C
l
<<1, 
where C
s
 and C
l
 are the metal solubilities in the solid (i.e. the semiconductor 
crystal) and the liquid alloy, respectively; 
3) have a small equilibrium vapour pressure over the liquid alloy in order to avoid 
volume and, therefore, cross-section of the whisker; 
4) be inert to chemical reaction products; 
5) have a small contact angle with the underlying substrate material; this non-
wetting characteristic strongly influences the shape of as-grown nanowires. 
Au 
19
 is the most used catalyst for VLS growth of most elemental and 
compound semiconductors as it meets all of the above criteria; furthermore it is an 
inert material, it does not oxidize prior to or during the process, and can form 
eutectic compounds with several atomic constituents of semiconductors. The T-x 
phase diagram of the Au-Si system is reported in fig. 2.5 showing the occurrence of 
the eutectic point at a temperature of 363°C; other metals used as catalyst for the 
VLS growth are Fe 
20,21
, Mn 
22
, Ti 
23
 and Al 
24
. 
In the VLS growth of compound semiconductors, such as GaAs or SiC, the 
process can be also self-catalysed by one of the semiconductor components, i.e. Ga 
or Si. However, Au is also used in most cases for the VLS growth of III-V, II-VI and 
III-N compound semiconductors. 
 
 
 
Figure 2.5 T-x phase diagram of the Au-Si system (Ref. 21).