Home Search Collections Journals About Contact us My IOPscience

Cluster beam deposition: a tool for nanoscale science and technology

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2006 J. Phys. D: Appl. Phys. 39 R439

(http://iopscience.iop.org/0022-3727/39/22/R02)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 134.151.40.2
This content was downloaded on 30/01/2014 at 06:20

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 39 (2006) R439–R459 doi:10.1088/0022-3727/39/22/R02

TOPICAL REVIEW

Cluster beam deposition: a tool for

nanoscale science and technology
K Wegner1,2 , P Piseri3 , H Vahedi Tafreshi4 and P Milani3,5
1
Tethis S.r.l., Via Russoli 3, 20151 Milano, Italy
2
Particle Technology Laboratory, Institute of Process Engineering, Department of
Mechanical and Process Engineering, ETH Zurich, CH-8092 Zurich, Switzerland
3
CIMAINA and Dipartimento di Fisica, Università di Milano, Via Celoria 16, 20133 Milano,
Italy
4
NCRC, North Carolina State University, 2401 Research Dr., Raleigh, NC 27695-8301, USA
E-mail: pmilani@mi.infn.it

Received 21 June 2006, in final form 7 August 2006

Published 3 November 2006
Online at stacks.iop.org/JPhysD/39/R439

Abstract
Gas phase nanoparticle production, manipulation and deposition is of
primary importance for the synthesis of nanostructured materials and for the
development of industrial processes based on nanotechnology. In this
review we present and discuss this approach, introducing cluster sources,
nanoparticle formation and growth mechanisms and the use of aerodynamic
focusing methods that are coupled with supersonic expansions to obtain
high intensity cluster beams with a control on nanoparticle mass and spatial
distribution. The implication of this technique for the synthesis of
nanostructured materials is also presented and applications are highlighted.
(Some figures in this article are in colour only in the electronic version)

1. Introduction reduction as a general benchmark. Other important parameters

are for instance, the possibility of improvement in performance
The presence of nanoparticles in industrial processes dates per unit cost and the compatibility with other production
back well before the advent of nanotechnology [1, 2]: carbon technologies (for example, in the semiconductor industry) [8].
nanoparticles as rubber additives for tyres [3] or titania The design and production of devices containing
nanoparticles for pigmentary applications [4, 5] represent nanoparticles and nanomaterials require the capability of
a paradigm of mass production of nanoscale objects. In integrating different components at different length scales in
the last decade, the increasing understanding and control of terms of structure, chemical composition, packaging, etc. It is
the fundamental properties of nanoparticles has stimulated thus necessary to determine the structure and chemical status
the interest in their use as building blocks of devices and of nanoparticles and to transfer them onto suitable substrates
systems with sophisticated properties and functionalities [6]. or into matrices with a very precise control on positioning
The ultimate goal of the nanotechnological approach is to in an exactly defined area. Once deposited in a device, the
provide new paradigms of production; however, it is reasonable nanoparticles should retain their properties and individuality.
to consider that the success of nanotechnology will rely on A crucial point to be solved for a real technological
its capability of complementing actual technologies before breakthrough is the possibility of manipulating nanoscale
substituting them [7]. objects [9]. The general term ‘manipulation’ is used here with
special reference to the following meanings: (i) the ability to
The success of manufacturing approaches based on
sort the objects in terms of a size or geometry classification;
the assembling of nanoparticles is largely determined by
(ii) the ability to control the position and (iii) the ability
the established infrastructure and customer base with cost
to modify them physically or chemically. As most of the
5 Author to whom any correspondence should be addressed. available nanoparticle synthetic routes are not able to produce

0022-3727/06/220439+21$30.00 © 2006 IOP Publishing Ltd Printed in the UK R439

Topical Review

a perfectly monodisperse population, a refinement of the fluxes and deposition rates [24]. A further improvement
particle population is often necessary depending on the specific in aerodynamic manipulation techniques for different gas
application. The second and third requisites are naturally phase synthesis methods is opening new perspectives for the
connected to the fabrication of a class of systems exploiting integration of gas phase nanoparticle production into the large
the extraordinary properties of the nanoscale objects. and well-consolidated arena of physical vapour deposition
Among different processes for the production of technologies [23].
nanoparticles, gas-phase routes are very popular both for In this review we will present and discuss features
large scale production and for fundamental studies [2]. of gas-phase cluster production and deposition relevant for
A number of extraordinary qualities make the gas-phase the fabrication of nanostructured systems. In particular,
approach a very interesting technique for the fabrication of we will concentrate on those fundamental aspects affecting
nanostructured systems. Gas-phase synthesis is an established the performance of different gas phase production and
and well-developed process able to produce large scale manipulation methods with their advantages and bottlenecks
quantities of nanoparticles [10–12] with a high level of control in view of possible technological uses. After this introduction,
on particle physico-chemical properties such as phase and section 2 is devoted to the basic principles of nanoparticle
composition [13]. Effective post-synthesis treatment such as formation and growth in the gas phase, while section 3
high temperature annealing [14, 15] and coating [16] has been describes methods for manipulation and handling of free
demonstrated, as well as the compatibility of aerosol methods nanoparticles. Cluster sources are described in section 4;
with the high purity standards of the semiconductor industry section 5 discusses patterning and coupling to planar
[17]. Positioning of gas phase particles on a surface with a technologies. An overview of applications and conclusions
resolution in the 100 nm range and below has been shown [18] are given in sections 6 and 7, respectively.
as well as the possibility of micropatterning particle-assembled
thin films [19]. 2. Gas phase nanoparticle formation and growth
Among different gas phase approaches to nanofabrication,
the deposition of clusters from supersonic beams is gaining Formation and growth processes of objects relevant for CBD
increasing attention extending the interest for this field from follow the same physical and chemical mechanisms as any
basic to applied research [20]. Aggregates ranging from a gas-phase particle synthesis process. These mechanisms have
few atoms to a few thousands of atoms, known as clusters, been extensively studied in aerosol synthesis, the gas-phase
are produced and carried in supersonic expansions and have manufacture of nanoparticles at atmospheric pressure [28],
been the playground for the characterization of the transition and can also be applied to the cluster sources used in CBD
from atomic behaviour to solid-state collective features. The processes.
role of the electronic structure and of the lattice structure in In gas phase synthesis, nanoparticles are made by
determining the stability of the clusters has been recognized ‘building’ them from individual atoms or molecules up to the
since 30 years and has been widely studied [21]. However, desired size. Cluster embryos are formed either by physical
the spectroscopic characterization of isolated clusters remains means such as condensation of a supersaturated vapour or
the holy grail for scientists since the high dilution of the vast by chemical reaction of gaseous precursors. Examples
majority of clusters prevents the use of standard physico- include inert gas condensation [29, 30], plasma [31] and flame
chemical characterization techniques. processes [32]. Depending on the embryo concentration,
Cluster beam deposition (CBD) has great potential for system temperature and pressure, these clusters continue
the production of nanostructured and nanocomposite films to grow to larger entities by coagulation and coalescence
although several technical limitations have hampered its use and/or surface growth. Particle dynamics [33, 34] aim at a
as one of the bottom-up approaches to the synthesis of mathematical description of these growth processes, taking the
nanomaterials [20]. Supersonic expansions have several entire nanoparticle population into account.
advantages for cluster manipulation over effusive beams that
make this approach very powerful for the deposition of
2.1. Particle formation
nanostructured films and the coupling with microfabrication
techniques. This is due to the fact that supersonic CBD Formation of particles in the gas phase takes place either
favours the manipulation and positioning of nanoparticles by by homogeneous nucleation or by coagulation (collision)
the exploitation of nanoparticle inertial properties [22, 23]. processes. The starting material can be vaporized from a hot
Many theoretical and experimental approaches have been source into a low density inert gas employing Joule heating,
developed to solve the problem of neutral nanoparticle thermal plasma or laser ablation. Cooling of the vapour rapidly
manipulation in the gas phase [24]. The merging of leads to supersaturation followed by homogeneous nucleation
solutions and models developed for aerosols and for supersonic and the formation of first product clusters [35].
expansions has stimulated a novel and interdisciplinary route Often, a chemical reaction is the first step in the
to this problem [24, 25]. The solutions proposed and tested in nanoparticle formation process. This is the case when the
these last years have shown that the synthesis of nanostructured background gas in the evaporation process is not inert but
materials with tailored structural and functional properties can reacts with the precursor vapour to form product molecules,
be obtained by exploiting aerodynamical focusing in seeded for instance, in the synthesis of aluminium nitride by the
supersonic beams. In particular the use of aerodynamic reaction of aluminium vapour with ammonia in an aerosol flow
lenses [26,27] allows an unprecedented control on nanoparticle reactor [36]. Another prominent and the industrially most
spatial and mass distribution, while keeping very high relevant example of gas-phase particle formation involving

R440
 Topical Review

chemical reactions is the introduction of a volatile precursor case with clusters, the coagulation theory in the free molecular
such as a chloride or organometallic compound into a high regime has to be applied. Neglecting again the spread of the
temperature environment, for instance, a flame [1, 33]. Here, particle distribution and the morphology:
precursor decomposition, oxidation, hydrolysis or, depending

on the environment, any other reaction lead to the formation 5/2 5/2 10 6kB T 1/2
dP = dP,0 + Vvol t, (3)
of product molecules. π ρP
The thermodynamics of the system determine whether
where Vvol is the total volume of particles per unit volume of
these product molecules grow to clusters by homogeneous gas and ρP is the density of the particles.
nucleation or by coagulation. A criterion to determine the Particles that grow by Brownian coagulation typically
formation path of a cluster is its thermodynamically critical reach asymptotic distributions, the so-called self-preserving
diameter, the Kelvin diameter d1,C : size distributions [28]. These distributions remain invariable
4γ Vmol,1 with time when the particle size and concentration are scaled
d1,C = . (1) with the average particle volume. For an initially monodisperse
kB T ln(S)
aerosol, it takes time tSPSD to reach the self-preserving size
Here, γ and Vmol,1 are the surface tension and the molecular distribution. In the free molecular regime, tSPSD is [39]:
volume of the cluster, S is the dimensionless saturation ratio 4
at temperature T and kB is the Boltzmann constant. A tSPSD =  1/2  3Vvol 1/6 . (4)
6kB T 5/6
nucleus is stable when its critical diameter is much smaller ρP
N0 4π
than the diameter of a single molecule of the product species.
In this case, there will be no growth or shrinkage by The geometric standard deviation for the number self
condensation or evaporation and the particle will form by preserving size distribution is about 1.46 while that for the mass
coagulation [37]. In the case of unstable clusters, particles are is about 1.3 [39, 40]. These asymptotic distributions place a
formed by homogeneous nucleation: balanced condensation lower limit on the reduction of the polydispersity of particles
and evaporation of molecules to and from clusters of the grown by coagulation.
product species. The Gibbs–Kelvin equation (1) should If the entire particle size distribution rather than a
be used with caution, though, as it relies on macroscopic monodisperse aerosol is to be considered for particle growth
particle properties like surface tension. For ceramics, the by coagulation, the population balance equation in the
critical cluster diameter is typically smaller than the molecular free molecular and continuum regimes has to solved using
diameter and particle formation takes place by coagulation. computer codes.
Metal nanoparticles on the other hand are usually formed by It should be noted that the particle morphology may affect
homogeneous nucleation, for instance, in the production of the rate of particle growth by coagulation. In the free molecular
iron nanoparticles in a plasma reactor [31] or the synthesis of limit, agglomerates collide much faster than spheres of the
bismuth nanoparticles by evaporation–condensation [38]. same mass since they have more surface area for collisions.
In the continuum limit, the enhanced collision surface of
agglomerates does not play a major role because agglomerates
2.2. Particle growth
experience enhanced drag by the medium.
The newly formed particles continue to grow either by surface Surface growth consists of a first step of molecule or
growth (addition of atoms or molecules to the particle) or atom transport to the surface of an already-formed particle
by coagulation (inelastic particle–particle collisions) which is and a second step involving a chemical reaction or a phase
usually followed by coalescence. change at the particle surface. Especially during the first stages
Coagulation describes particle–particle collisions due to of particle formation from supersaturated vapour, surface
Brownian motion or other mechanisms such as shear or growth can be significant as the initially formed clusters act as
electrostatic forces. Typically, it is applied to particle growth condensation seeds for the remaining vapour. By controlling
in combination with coalescence. For the description of the supersaturation at a low level, particle nucleation can
particle growth by coagulation, one has to distinguish between be slowed down while the rate of surface growth by vapour
collisions in the free molecular regime (particle diameter deposition can be increased [34]. In systems with chemical
dp smaller than the mean free path of the gas λ) and in reaction, species transported to the particle surface either
the continuum regime (dP
λ). Neglecting the particle react with the particle thus changing its chemistry or react
morphology and the spread of the particle size distribution, at the particle surface with other gaseous species with the
the classical theory for Brownian coagulation of monodisperse reaction product being deposited on the surface. Surface
spheres in the continuum regime at temperature T is used to growth can contribute to the entire particle growth process in
calculate dp [28]: systems where vapour or gas molecules are present in sufficient
quantity. In general, the rate of change of the particle diameter

 dp at temperature T by vapour deposition in the free molecular
4kB T N0 1/3
dP = dP,0 1 + t , (2) regime is given by [28]:
3µ
ddP 2v1 (p1 − pd )
where dP,0 and N0 are the initial particle diameter and = F1 , (5)
dt (2π mkB T )1/2
concentration, respectively, µ is the dynamic viscosity of the
gas and t is the residence time. If the particle diameter is much where v1 is the volume of the transported molecule in the
smaller than the mean free path of the gas, as is usually the particle phase, p1 is the partial pressure of the gas or vapour

R441
Topical Review

far from the particle and pd is the partial pressure at the time and space is by a population balance equation:


particle surface obtained from the equilibrium vapour pressure ∂n(v) ∂ dv
over a flat surface using the Kelvin equation (1). F1 is the = −∇n u + ∇D∇n + n − ∇ cn
∂t ∂v dt
Fuchs–Sutugin factor for bridging the free molecular with the (a) (b) (c) (d)
continuum regime [41]. In the continuum regime, the rate of 
1 v
change of the particle diameter is [28]: + β(ṽ, v − ṽ)n(ṽ)n(v − ṽ) dṽ
2 0
 ∞
ddP 4Dv1 (p1 − pd )
= F2 , (6) − β(v, ṽ)n(v)n(ṽ) dṽ
dt dP k B T 0
(e)
where D is the diffusion coefficient of the gas or vapour and F2  ∞
is the Fuchs–Sutugin factor for bridging the continuum with − S(v)n(v) + γ (v, ṽ)Sn(ṽ) dṽ (7)
v
the free molecular regime.
(f)
If the nucleation burst produces particles from size d0
to d0 + d, the upper and lower bound sizes will increase In this general example of a population balance, the variation of
due to vapour deposition but the size range will not change. the particle size distribution function n with time t is influenced
Thus, the relative range of particle sizes d/d, will decrease by convection (a), diffusion (b), surface growth (c), external
as the particles grow, leading to a narrower distribution than forces (d), coagulation (e) and fragmentation (f). Thereby,
u is the velocity, D the diffusion coefficient, v the particle
that obtained by growth through coagulation, as observed by
volume, c the particle velocity, β the collision function, S
Okuyama et al [42]. This has been used for instance by Bowles
the rate of fragmentation and γ the fragment size distribution
et al [43] who generated clusters of controlled size and then
(see Friedlander [28]). The collision function β describes
grew them by vapour deposition, leading to a very narrow
coagulation due to Brownian motion, shear or electrical
particle size distribution. charges. Furthermore, particle morphology, expressed for
The classical coagulation theory assumes rapid and full instance by the fractal dimension, as well as information on
coalescence of particles after a collision event, resulting coagulation in the free molecular, continuum or transition
in spherical product particles. Full coalescence, however, regime are considered in the collision function. The form
only takes place at sufficiently high temperature at which of the population balance and its terms have to be adopted
particles are present as low viscosity liquid droplets. At to the physically relevant growth processes of the investigated
lower temperature, the two or more colliding particles will system. Several examples are given by Friedlander [28].
stick to each other and fuse by sintering, often resulting As the population balance equation is a nonlinear, partial
in agglomerates rather than spherical product nanoparticles. integrodifferential equation, numerical solutions are usually
Thus, a sintering rate has to be introduced into the description required. A discrete solution accounting for all particle sizes
of particle growth. The sintering kinetics depend strongly leads to more than 109 differential equations to be solved
not only on material and temperature but also on particle simultaneously if the entire growth spectrum below 1 µm
size. Driven by surface energy minimization, coalescence is considered [48]. Therefore, the solution of the discrete
or sintering of two particles can occur via viscous flow or population balance is usually only carried out for a limited
solid state diffusion mechanisms. Typically, however, the size fraction, describing for instance the early stages of
application of macroscopic material properties to nanoparticles growth. Computationally less intensive but also less accurate
are moment, sectional or monodisperse models that rely on
does not describe the observed coalescence or sintering
simplifying assumptions. Moment models make assumptions
rates. The significantly different thermodynamic properties
regarding the shape of the particle size distribution [49],
of nanoparticles with diameters below about 10 nm that are
sectional models divide the size distribution into size classes
for instance evident from the melting point decrease of gold
and sections, which are characterized by a mean value of
nanoparticles with decreasing particle size [44] can result in a particle characteristic [50] while monodispersity can be
large errors in the predicted coalescence rate [45]. Thus, considered for rough estimates.
models must be applied to more accurately describe the
observed coalescence rates of nanoparticles [46, 47].
2.4. Control of particle synthesis
Depending on the governing growth mechanisms and process
2.3. The population balance equation conditions, product nanoparticles are individual, dense par-
ticles or fractal-like agglomerates consisting of at least two
The particle formation and growth processes described in subunits, the so-called ‘primary particles’. Depending on the
the previous paragraphs usually take place simultaneously. strength and nature of the bonds connecting the primary par-
Depending on the system investigated, additional effects such ticles, an agglomerate is termed hard or soft [34]. In hard ag-
as electrical charges or intraparticle chemical conversion might glomerates, primary particles are strongly bound together as a
play a role in the particle growth process. Furthermore, in result of sintering necks while in soft agglomerates the inter-
an accurate mathematical description the entire polydisperse particle bonds are of physical nature and in most cases van der
particle population should be considered rather than an Waals forces. Soft agglomerates are formed at temperatures
idealized monodisperse aerosol. The general mathematical too low for significant coalescence and unlike hard agglom-
description for the evolution of a particle size distribution in erates they may be separated into their component primary

R442
 Topical Review

Figure 1. Mapping the degree of silica nanoparticle agglomeration in flame aerosol reactors expressed as the ratio of the collision diameter
of the hard agglomerate and the primary particle diameter (dcH /dpH ) in the process parameter space of the inlet SiCl4 -precursor molar fraction
(φ) and the cooling rate for reactor residence times of 2 ms at 2100 and 2200 K maximum process temperature (Tmax ). Isopleths of the
primary particle diameters (dpH ’s, nm) are shown also (white lines). Increasing Tmax reduces the degree of agglomeration and increases the
primary particle diameter. Pictures reproduced with permission from Langmuir 2004 20 5933–9 [51]. ©2004, American Chemical Society.

particles with relative ease. The temperature field, reactor The production of particle beams is based on the expansion
residence time and chemical additives affect the particle size of a particle-gas mixture through a nozzle, generating a sonic
and the extent of agglomeration and, consequently, the par- or supersonic gas stream [20]. The basic concept of an effusive
ticle morphology. If the growth conditions are such that the beam source is very simple: it is an orifice in a very thin wall of
rate of coalescence is faster than that of coagulation, spherical a reservoir where the gas or vapour is in thermal equilibrium.
particles are obtained. However, as particles grow, the sinter- The opening is small enough so that the outgoing flow will not
ing rate decreases due to increasing particle size and typically affect the equilibrium in the reservoir. If the pressure in the
due to decreasing reactor temperature so that particle coales- reservoir is low enough, the outgoing flow will be molecular
cence usually becomes slower than coagulation. As a result, so that the effusion rate and both the angular and velocity
irregularly shaped agglomerate particles are formed [32]. distributions of the formed beam can be calculated on the basis
Especially in ceramic nanoparticle production full of the gas kinetic theory without any assumption [53].
coalescence is only observed at very high temperatures. A supersonic expansion can be obtained by imposing a
Accounting for simultaneous gas phase chemical reaction, pressure ratio less than Pb /P0 = 0.478 across a convergent
coagulation, and sintering during the formation and growth nozzle driving an isentropic flow expansion where P0 and Pb
of silica (SiO2 ) nanoparticles, Tsantilis and Pratsinis [51] are the stagnation and the background pressures, respectively.
computed maps showing the degree of agglomeration as a The expansion through a convergent nozzle will always take
function of the reactor temperature, precursor concentration place in a subsonic regime regardless of the amount of the
and cooling rate. Figure 1 shows how an increased process applied pressure ratio. Outside the converging nozzle, depend-
temperature and high cooling rates of 105 to 106 K s−1 can ing on the pressure ratio, the flow will supersonically expand
avoid the formation of hard agglomerates even at rather high to pressures even much lower than the background. A normal
concentrations. This was observed experimentally by Wegner shock, known as Mach disc, matches the pressure inside the jet
et al [52] for titania nanoparticles when rapidly quenching a to the background and closes an area called the zone of silence.
synthesis flame by expansion through a critical flow nozzle. The location of the Mach disc has been empirically determined
as being only a function of the pressure ratio and independent
3. Manipulation and handling in the gas phase of the fluid nature and nozzle geometry [54].
The sudden free expansion of the flow produces a high
As outlined in the previous section, gas phase synthesis usually outward radial velocity at the beginning of the free jet close
does not give monodisperse particles but a size distribution, the to the nozzle outlet. This radial velocity (drag) greatly
width of which mainly depends on the synthesis conditions varies with the radial position both in the jet and at the
in the particle source. For many applications, however, a nozzle outlet. It is weak at the centreline and very strong
refinement of the particle population is desired. The ability to close to the nozzle wall at the outlet. Hence, in contrast to
sort nanoparticles in the gas phase directly after their synthesis the particles located in the central regions, those far from
in terms of a size or geometry thus is of major importance. the axis are exposed to a strong radial drag. Consequently,
For device fabrication involving controlled particle deposition the trajectories of particles concentrated on the centreline can
on substrates, a second requisite is of importance: the ability remain approximately unchanged, while particles far from the
to deposit nanoparticles with very high lateral resolution. In axis may strongly diverge in the free expansion. It should
CBD, this can be obtained by controlling the shape of the be also noted that the response of the particles to this radial
particle beam and the particle velocities. drag is a function of the particle size. The above mentioned

R443
Topical Review

Figure 2. Simulation of the gas flow (a) and trajectories of 15 nm particles (b) in an aerodynamic lens system according to Gidwani and
Di Fonzo et al [112, 149]. The particles are concentrated at the centre axis of the lens system, forming a beam. (c) Basic principle of
aerodynamic size-selection and focusing. The inertia of large particles (St
1) exceeds the drag action of the gas, separating the particle
trajectories from the gas stream lines and leading to wall deposition. Very small particles (St  1) closely follow the flow. Intermediate size
particles have trajectories only slightly decoupled from the stream lines, which can be exploited to concentrate them at given positions in the
flow-field [29, 66].

radial drag can effectively separate the particles of different limits of application to nanometre-sized particles and heavy
sizes especially if they are not concentrated on the centreline. molecules [60]. Regarding an exploitation of these effects
Small particles can follow the expanding carrier gas, while for the production of high intensity molecular beams, only
large particles persist on their original trajectories [55, 56]. very little work was performed in the early times, even though
If the particles can be concentrated on the nozzle Fenn already recognized in 1963 that the gas mixture in a
centreline, only Brownian diffusion (and lift-force effects in seeded supersonic beam source is nothing but an aerosol [61].
case of non-spherical particles [26,27]) can perturb/spread the A probable reason for this neglect is in the complexity and
configuration obtained by aerodynamic focusing. ambiguity of the results of the early experiments, where the
background gas was playing a major role [61, 62].
Dirac suggested during World War II that inertial effects
The first attempts to focus particles through supersonic
in a gas flow can be used for mass separation [57]. One decade
expansions were made by Murphy and Sears [63] and Israel and
later, the experimental evidence of such an effect was obtained
Friedlander [64]. Dahneke and Flachsbart [65] increased the
by Becker and co-workers [58]. Although the first observations particle concentration in the core of an aerosol free jet using an
were done in the jet expansion of a supersonic molecular beam extra stream, sheath air, that confines the boundaries of the core
source, the subsequent work by Becker was aimed at uranium jet downstream of the nozzle. In the recent literature, the major
enrichment in a cascade of aerodynamic separation stages (the development in particle focusing dates back to the work of Liu
‘separation nozzle’) the product of which was not a beam of et al [26,27]. They were the first to produce an enriched stream
the selected species [59]. of particles using only aerodynamic effects induced by nozzles.
In aerosol science, particle separation effects have long This was achieved with a system of so-called aerodynamic
been exploited for particulate sampling with impactors. lenses consisting of successive axisymmetric contractions–
Significant advances in this field have been made pushing the enlargements of the aerosol flow passage (figures 2 (a) and (b)).

R444
 Topical Review

The work of Liu et al [26, 27] was inspired by the

pioneering research carried out by Fernandez de la Mora
[66,67] and co-workers who revealed the possibility of particle
focusing and the existence of a common focal point for the
near-axis particles when expanding an aerosol through a thin-
plate orifice (figure 2(c)). Unfortunately, particle beams may
diverge after a sharp focal point downstream of the nozzle
because the gas streamlines diverge due to the sudden gas
expansion. The over-exposed images reported by Fuerstenau
et al [68] visualized the above divergence in aerosol jets
expanded through thin-plate orifices. The novelty of the work
of Liu et al [26, 27] is that they employed thin plate orifices
in a confined passage to manipulate the spatial distribution of
particles prior to the nozzle and the subsequent expansion in
the free jet.
In 1999, Mallina et al [69] demonstrated that the beams
produced by expansion through capillaries have lower angular
spread than those formed by expansion through conical
nozzles. This is due to the focalization of particle beams
at a point downstream of the conical nozzle and beam
divergence afterwards, similar to thin-plate orifices. In
capillaries, however, particles asymptotically converge to the
focal point, which appears to be inside the capillary, and do not
Figure 3. Cross sectional view of the focusing nozzle assembly (a),
significantly diverge afterwards as the capillary walls confine
to scale. The narrow gap between the two pieces, where the gas flow
the gas streamlines. is directed towards the nozzle axis can be seen on the left of the
Goo [70] simulated the aerosol concentration at figure; (b) is a perspective view of the same two pieces (exploded).
atmospheric pressure in a cascade of aerodynamic slit lenses The focusing mechanism is shown in (c); at a given pressure only
followed by a virtual impactor. Soon afterwards, Lee et al particles with proper size have trajectories leading to particle
concentration close to the nozzle axis. Reprinted from Piseri P,
[71] reported on experiments and numerical simulation of Podestà A, Barborini E, Milani P 2001 Rev. Sci. Instrum. 72
particle focusing at atmospheric pressure. Zhang et al [72] 2261–7 [56]. ©2001 American Institute of Physics.
repeated the simulations of Liu et al [26] with less restrictive
assumptions, allowing them to study a wider range of particle
focusing effect of their lens but can broaden the focused
sizes. Specifically, they considered the compressibility of
beam to some degree (figure 4) [75]. Middha and Wexler
the continuous phase as well as particle loss due to wall
[76] recently proposed an aerodynamic focuser with so-called
impact. Zhang et al [72] observed that the maximum particle
capped-cone geometry that has revealed a certain improvement
displacement, as well as particle loss, occurs at a particle Stokes
over the previous designs [73, 74].
number near unity.
Since the invention of aerodynamic lenses by Liu et al
Piseri et al [56] developed a new type of aerodynamic
[26,27], this system has been used in a variety of applications.
lens system. By placing an obstacle (a flat plate hereafter
One example is single particle mass spectrometry for aerosol
called a focuser) upstream of a capillary nozzle, they forced
sampling. The ability of aerodynamic lenses to concentrate the
the flow to undergo two 90◦ turns to reach the nozzle inlet.
beam on a very narrow near-axis region is of crucial importance
The spacing between the focuser and the nozzle inlet is the
for the performance of these instruments where a pulsed laser
controlling parameter for selecting particles of the desired size
for particle ionization/vaporization is triggered by the single
(figure 3). In comparison to the aerodynamic lens system of
particle itself.
Liu et al [26, 27], particles approach the orifice with a more
uniform velocity and direction and experience a more uniform Mallina et al [77] have reported on the capability of
and higher acceleration. Furthermore, this new design is less variable pressure inlets at the entrance of the lens system to
sensitive to the upstream position of the particles (for instance produce beams of selected size ranges. By changing the nozzle
in the cluster source or synthesis chamber). Thus, a broader source pressure, their design obviates the need for sizing the
size range of particles originating from distances farther from components in order to aerodynamically focus a special range
the centreline can be focused. The device is very compact, has of particles. Fernandez de la Mora [78] used aerodynamic
few components that are easy to machine, and is not subject lenses in a variable-pressure impactor to improve the resolution
to critical alignment requirements. The main drawback of this of an aerosol size spectrometer. Recently, McMurry and co-
design is the high rate of particle deposition to the wall. workers have proposed an aerodynamic lens system to focus
Vahedi Tafreshi et al [73] have simulated the performance nanoparticles smaller than 30 nm [25].
of the above aerodynamic nozzle in its full three-dimensional
geometry. They also considered an axisymmetric model and 4. Cluster sources
studied the effects of different operating conditions, spacing
and the initial position of particles [74]. They reported that the Gas-phase cluster formation processes are characterized by
Brownian diffusion cannot drastically affect the aerodynamic critical parameters such as the number of collisions between

R445
Topical Review

or pulsed, effusive or supersonic. At a first glance, continuous

production methods coupled to continuous gas flow regimes
seem to guarantee an easier and a more efficient production and
control of cluster parameters. Actually this can be the case only
when very huge gas loads can be handled for ambient pressure
aerosol techniques (i.e. flame pyrolysis) [32]. For CBD under
vacuum conditions a continuous gas flow must be compatible
with stringent vacuum requirements and hence only effusive
regimes are of practical interest. Vacuum requirements also
affect the pressure attainable in the cluster source that usually
should not exceed a few Torr [20]. The critical parameter
to control the number of collisions then is the cluster source
dimension and geometry.
The realization and operation of pulsed cluster sources
appear to be more complicated compared with continuous
ones; moreover, they are characterized by a low duty cycle [79].
On the other hand, the reduced gas load has the advantage
of allowing the use of a supersonic expansion regime and
the compatibility with HV and UHV standards. Moreover,
it is possible to control the gas pressure in the source region,
where cluster formation takes place, over a very wide range.
These aspects are of fundamental importance for applications
and in particular for the compatibility of CBD processes with
microfabrication and planar technologies [23].
The structure and operation of cluster sources is also
determined by the methods used for precursor production.
The reader can find many informative reviews describing in
detail sources for CBD based on different production methods
[80]. Here we will consider cluster sources as possible
working tools for the fabrication of nanostructured systems
and we will present the aspects related to the interconnection
of different parameters (precursor vaporization methods, gas
flow, cluster extraction regime, etc) and their importance
Figure 4. Trajectories of 1.5 nm diameter spherical particles of for the realization of the source hardware for the following
uniform density inside the PMCS that are released from the cathode methods: (i) joule heating, (ii) sputtering, (iii) laser ablation,
surface. The simulations show that without (a) and (b) and with
(iv) pulsed microplasma cluster source (PMCS) and (v) arc
Brownian motion accounted for, the focuser (b) and (d) is able to
increase the particle number density on the axis. Reprinted from discharge.
Vahedi Tafreshi H, Piseri P, Barborini E, Benedek G and Milani P
2002 J. Nanopart. Res. 4 511–24, figure 6 [75]. ©2002 with kind
4.1. Joule heating
permission of Springer Science and Business Media.
Cluster sources based on joule heating are conceptually simple:
aggregating species, collisions between aggregating species in many respects their principle of operation is very similar
and the thermalizing gas and by the method to produce, in to that of an atomic beam source [81]. They are based
a defined volume, the cluster precursors (see section 2.2). on a reservoir with well-defined exit opening in which a
Cluster formation takes place in a buffer gas acting as certain vapour pressure of the precursor material must be
a thermal bath and at the same time as a carrier of the sustained. The control on the vapour pressure and temperature
nanoparticles. These two aspects are intimately related so necessary to induce nucleation is usually realized by mixing
that the pressure inside the particle source determines the the precursors with an inert gas.
condensation efficiency, whereas the pressure gradient affects Vaporization of the precursors is obtained by joule heating
the cluster extraction and manipulation. of high-temperature crucibles as in the case of molecular beam
The pressure gradient depends on the pressure of the epitaxy (MBE) [81, 82]. However, high intensity sources
backing gas, the source and nozzle geometry and the for the production of cluster beams have more stringent
ratio between the pressures inside and outside the source requirements. In particular, the vapour pressures are typically
pint /pext . The different technical solutions adopted to control about two orders of magnitude higher than those used in MBE
these aspects determine the operation principles of different which are roughly 10−3 -10−4 mbar; hence they operate at
cluster sources and influence their performance in terms of significantly higher temperatures.
cluster mass distribution, cluster production and extraction Sources based on joule heating can be used to produce
efficiency [20]. clusters from materials with a low temperature of vaporization.
Cluster sources can be catalogued by considering the Though, even in this case a complicated construction scheme
regimes governing gas introduction and extraction: continuous should be adopted in order to keep the hot oven apart from the

R446
 Topical Review

rest of the source. Moreover, the large amount of thermalizing Plasma sputtering offers a method for vaporizing refrac-
gas necessary to reach condensation conditions makes this tory materials without involving the complications of the target
kind of source impractical for many applications since the heating process. The combination of plasma sputtering with
operation is in a continuous regime and results in huge gas gas condensation was reported in 1986 and then developed by
loads (figure 5) [83]. Haberland and co-workers through the use of magnetron sput-
tering [86, 87]. This source is relatively easy to operate and,
4.2. Sputtering in principle, it allows the production of intense cluster beams;
however, to date, applications to nano- and microfabrication
Plasma discharges are of particular relevance for cluster are scarce.
formation. Schematically a discharge consists of a voltage In figure 6 two different configurations of a plasma-
supply that drives current through a low pressure gas between gas condensation cluster beam apparatus are depicted. The
two conducting plates or electrodes [84]. The gas breaks general and common features of these systems are: a sputtering
down to form a weakly ionized plasma. Charged particles chamber, a growth chamber and a deposition chamber. The
in the plasma acquire high kinetic energies and collide with sputtering of the target material can be obtained with different
the neutrals in the gas, causing the formation of very reactive discharge modes and configurations: dc or rf modes, hollow
species. Material sputtering from the plates or electrodes takes cathode and magnetron sputtering. The cluster size can be
place; in this way the electrode material acts as a feedstock of controlled by adjusting the sputter yield, the gas pressure and
particles injected in the plasma [85]. the volume of the cluster growth region [86,88–90].
The gas pressure inside the sputtering chamber influences
the discharge while gas inlet and extraction are critical to
determine the nucleation and condensation processes. In order
to control the cluster mass distribution, the distance between
the sputtering region and the extraction nozzle can be varied.
To favour condensation, the source region can be cooled by
liquid nitrogen, which however imposes constraints on the
source dimensions and geometry [91].
During cluster production, a typical gas pressure in the
condensation chamber is 1 mbar, with a gas flow rate of a few
hundreds of a standard cm3 per minute. In order to achieve a
high vacuum level of 10−7 mbar in the deposition chamber,
a high throughput pump is installed in each section of the
apparatus. The base pressure of the source is 10−8 mbar, while
typical pressures during cluster production in the three sections
are 10−3 mbar, 10−4 mbar and 10−6 mbar, respectively. This
requires a high pressure level, which is found to pose problems
to the sputtering operation. Conventional magnetron guns are
Figure 5. Cluster source based on evaporation by Joule heating. designed to operate at 10−2 mbar, whereas the typical operating
Metal vapour and inert gas expand from the heated high pressure pressure in a cluster source is 1 mbar. At such high pressure
reservoir (1, 2350 K, 500 kPa) out of a conical supersonic nozzle (2). the plasma is unstable and sputtering on the inside of the
Clusters are formed in the expanding flow. The cluster beam is magnetron gun can take place [90].
geometrically confined by a skimmer (3) and a collimator (4).
Reprinted from Gatz P and Hagena O F 1995 Surf. Sci. 91 Another important parameter is the geometry of the gas
169–74 [83]. ©2005 with permission from Elsevier. inlet as it affects the pressure in the region where the sputtering

Figure 6. Single (a) and double source (b) plasma-gas condensation cluster beam apparatus. Reprinted from Sumiyama K, Hihara T,
Peng D L and Katoh R 2004 Sci. Technol. Adv. Mater. 6 18–26. [91]. ©2004 with permission from Elsevier.

R447
Topical Review

Figure 8. Schematic representation of a LVCS. Clusters are

generated by focusing a pulsed laser beam through the light entrance
on a target rod of the desired material. An inert carrier gas is pulsed
over the rod. Metal clusters are formed and thermalized in the
near-sonic flow of the carrier gas as it passes down a thermalization
channel before it is expanded through a nozzle. Picture after Milani
and deHeer [93].

Since the vaporization volume in a LVCS is very

Figure 7. Two configurations of a plasma-gas-condensation type small, the dimensions of the source are significantly smaller
cluster deposition apparatus reproduced from (a) Hihara and compared with hot oven or plasma sources. In figure 8
Sumiyama [88] and (b) Peng et al [135]. ‘CM’, ‘TA ’, ‘TMP’, a schematic representation of a LVCS is shown where the
‘MBP’, and ‘CMP’ represent capacitance manometers, sputtering critical parameters affecting the source performance can be
targets, turbomolecular pumps, mechanical booster pumps and recognized.
compound molecular pumps, respectively.
The light of a high intensity pulsed laser (usually with a
pulse length in the order of tens of nanoseconds) is focused onto
takes place and hence the cluster formation in terms of a target vaporizing a small amount of material into a flow of an
cluster mass distribution. Sumiyama and co-workers have inert carrier gas. The inert gas quenches the plasma and cluster
characterized this aspect in detail [88]. Figure 7 shows two condensation is promoted. The mixture is then expanded into
different inlet configurations. vacuum and forms a cluster beam. When a pulsed vaporization
takes place it is convenient to operate the LVCS with a pulsed
4.3. Laser vaporization valve for carrier gas introduction. Different geometries have
been developed for target mounting and to favour cluster
Laser vaporization can generate a high density vapour of formation and growth prior to the extraction. Originally,
virtually any material in a short time interval and in a Smalley and co-workers proposed a scheme where the inert
well-localized volume. By rapid quenching of the plasma, gas flow was directed through a narrow channel (typically a
clusters and nanoparticles can be produced [92, 93]. This few mm in diameter). Perpendicular to this channel a hole was
technique was originally applied to the production of clusters drilled to allow the laser to hit a target rod. This geometry does
in molecular beams by Smalley and co-workers [94]. Since not favour cluster formation, moreover the interaction of the
this pioneering work, laser vaporization has become one of plasma plume with the walls of the source causes instabilities in
the most common techniques for generating cluster beams the operation [94]. Cluster formation can be improved by using
especially of refractory materials [20]. long channels or pre-expansion extensions of the nozzle [95].
The kinetics of aggregation of the plasma produced by However, a long channel length allows the metal vapour and
laser ablation and the characteristics of the resulting aggregates clusters to diffuse to the walls and hence be lost.
(density, size distribution, structure, etc) as well as the intensity To overcome these shortcomings, a cavity can be
and stability of the laser vaporization cluster source (LVCS) introduced where the vaporization takes place before the
are influenced by the quantity and type of ablated material, channel. This cavity should be properly dimensioned in order
the plasma—buffer gas interaction, the plasma—source wall to minimize the interaction of the plume with the walls, thereby
interaction, and the cluster residence time prior to expansion. reducing material deposition, and optimizing heat transfer to
High power pulsed solid-state or excimer lasers are used for the cavity walls by the inert gas. Furthermore, the processes
the vaporization; the pulsed nature of the precursor production involved in vaporization, cluster formation and thermalization
makes this approach particularly suited for pulsed gas regimes. are effectively decoupled from the expansion process [93].

R448
 Topical Review

The characteristics of the cluster population are controlled

by the local gas pressure during plasma production and the
residence time of the particles in the source body. The plasma
gas interaction affects not only the final cluster distribution
but also the subsequent expansion and beam formation. By
monitoring the pressure evolution in a LVCS, it has been shown
that vaporization in a low pressure environment produces a
large amount of monomers. Increasing the pressure during the
ablation results in a shift of the cluster distribution towards
larger masses [96].
Different geometries have been used depending on
the target shape: when the target is a disc (as for many
semiconductor materials), a gear mechanism inducing rotation
must be used in order to allow a uniform consumption of the
surface of the target. In particular, a hypocycloidal planetary Figure 9. Schematic representation of a PMCS according to
gear assembly has been used to burn out a spirographic path on Barborini et al [99]. Inert gas is introduced through a pulsed valve
the target [97]. The disc configuration requires the solution of into a cavity containing the rotating target cathode. When a high
several problems such as the sealing of the disc-source block voltage is applied between anode and cathode, the material is
sputtered from the cathode rod. The precursor vapour condenses
interface while keeping the disc moving. into clusters which grow in the thermalization cavity before they are
The laser pulse characteristics influence the quantity and extracted from the source through an expansion nozzle.
the state of the matter removed from the target. Pellarin et al
[98] have shown that the use of a Ti : sapphire laser (790 nm, cylindrical cavity (figure 9). The channel holds the target
30 ns pulse width) can produce intense transition metal cluster to be vaporized which typically is connected to the negative
beams over a wide mass range. Deposition rates of 0.2 nm s−1 pole of a power source, thus acting as a cathode. The anode
for transition metals have been reported. can be placed at some location inside the cavity or can also
LVCS are very flexible and allow the production of be introduced through the channel, opposing the cathode. A
metallic, oxide and alloy clusters with a large variety of solenoid pulsed valve for introduction of inert carrier gas closes
structures and stoichiometric combinations [80]. The low one side of the cavity while a nozzle is located in the opposite
gas load allows the coupling of LVCS to UHV deposition cavity wall. The valve, backed with a high gas pressure
equipment [80]. On the other hand, the deposition rates are (typically up to 50 bars), delivers inert gas pulses to the source
rather low and typically only small areas are covered. This is cavity with an opening time of a few hundreds of microseconds.
a serious bottleneck in the use of LVCS for the synthesis of If the pulsed valve is closed, the source cavity is at the same
nanostructured materials in view of its applications. pressure as the first vacuum chamber. Once the valve opens, the
large pressure difference causes the formation of a supersonic
4.4. Pulsed microplasma cluster source (PMCS) gas jet directed against the cathode. A pulsed voltage (typical
duration: 50 µs) applied between the electrodes, ionizes the gas
In sputtering-based gas aggregation sources, plasma-assisted and generates a plasma. This is accelerated against the cathode
vaporization and cluster growth take place in two different and ablates the material that thermalizes and condenses to
regions but in direct succession. Though, stable cluster beams form clusters [99, 100]. Due to the formation of a strong
are produced, this class of sources requires huge pumping pressure gradient close to the cathode surface, an aerodynamic
systems and it is not well suited for supersonic expansions. confinement of the plasma is produced in the source cavity
In pulsed laser vaporization sources (PLVS) the target ablation and the sputtering process is restricted to a cathode area of less
is confined in a very small volume, thus allowing stability of than 1 mm2 , as has been demonstrated by computational fluid
operation. However beam intensity is the major problem [20]. dynamics (figure 10) [100].
One of the key factors for improving the performance of These simulations showed how a hypersonic helium jet
cluster sources is the confinement of the ablation plasma in a develops inside the source at the time of the electric discharge,
well-defined region of the target in order to achieve material causing the confinement of the ablation plasma, improving
vaporization with high efficiency and to control cluster growth the sputtering yield and favouring cluster seed condensation
by the inert gas pressure in that region. Source geometry and by creating a high pressure region that coincides with the
dimensions should also be optimized to produce intense and one of ablation. This is crucial to assure a PMCS operation
stable cluster beams, while using a relatively low carrier-gas stability and reproducibility that is superior to other cluster
load. sources [56,101]. Since the volume of the source cavity is very
PMCSs represent a combination of different elements small (∼ 2 cm3 ), the source mean pressure rises rapidly after
typical of sputtering sources and LVCSs [99]. The working each pulse. The pressure difference across the nozzle drives
principle of PMCSs is based on spatially confined pulsed the aerosol expansion into the adjacent expansion chamber in
plasma discharge ablation of a target placed in a condensation the form of a supersonic beam.
chamber. The vaporized species are quenched by a pulse of In general, it is very difficult to obtain experimental
inert gas and condense to form clusters [100]. information on the temperature and pressure fields inside a
Schematically the source consists of a ceramic body with PMCS due to its transient flow features, the high temperature
a channel drilled through to perpendicularly intersect a larger and short duration of the discharge and the geometrical

R449
Topical Review

Figure 11. Scanning electron micrograph of a thick film formed by

carbon nanoparticle deposition using CBD with PMCS.

chemical species, mixed clusters can be obtained in order to

deposit nanocomposite films at high deposition rates over a
large area. PMCSs have been used for the production of
devices such as supercapacitors [103], sensor arrays [23, 104]
or high-throughput screening arrays for biological applications
[105]. PMCSs can be equipped with focusing devices as
described in section 3 in order to further improve cluster beam
intensity and patterning capabilities [24].

4.5. Arc discharge

A pulsed source of highly ionized metal plasma is the
electric arc, consisting of a discharge between two conductive
electrodes. The arc regime is characterized by a relatively
Figure 10. Velocity (a) and Mach number (b) fields in a PMCS by low cathode potential fall in contrast to a glow discharge
computational fluid dynamics simulations. It can be seen how a where the potential drop is several hundred volts [84]. This
hypersonic helium jet develops inside the source at the time of the
electric discharge causing the confinement of the ablation plasma to is due to different mechanisms underlying cathode emissions:
the region of the jet stagnation point on the cathode rod. Reprinted arc cathodes emit a large electron current as a result of
with permission from Vahedi Tafreshi H, Piseri P, Benedek G and thermionic, field electron and field emission [84]. Arc
Milani P 2006 J. Nanosci. Nanotechnol. 6 1140–9. ©2006 discharges are characterized by large currents (1–105 A),
American Scientific Publishers, http://www.aspbs.com [100]. causing a considerable heating either of the entire cathode or of
a small area for short time intervals, thus inducing vaporization
compactness and complexity of the source cavity. Thus, of the electrode material. The vaporized material, ejected in
computational fluid dynamics simulations backed up by various forms and forming a plasma between the electrodes,
experimental data, for instance at the source inlet, and outlet can undergo condensation into clusters. The presence of a
can play a significant role in understanding the fluid and buffer gas surrounding the electrodes influences the plasma
particle growth dynamics inside the source and can assist expansion from the electrode spots and also changes the
in improving the design of such systems. Carrying out behaviour of the arc and the role of the electrodes.
such computations, Vahedi Tafreshi et al [73–75] thoroughly The application of an arc discharge to the production of
studied the flow inside and exiting a PMCS and revealed the supersonic cluster beams was originally proposed by Meiwes-
importance of aerodynamic effects in controlling the spatial Broer and co-workers in 1990 [106, 107]. The pulsed arc
distribution of particulate phase in the cluster-laden supersonic cluster ion source (PACIS) was conceived as an alternative
flow exiting the PMCS. to laser vaporization sources: the vaporization of the precursor
For a graphite cathode, a lognormal cluster mass electrode being obtained by a pulsed high current discharge
distribution in a range from few tens to several thousands of between two electrodes. The plasma produced in this way
atoms per cluster with an average size at about 1000 atoms is mixed and thermalized by an inert gas jet between the
per cluster is obtained with the PMCS [56]. By operating electrodes that is introduced through a pulsed valve. A
the PMCS with a pulse frequency of 5 Hz, deposition rates of schematic representation of the PACIS is shown in figure 12.
100 h−1 over a surface of 1 cm2 placed at 500 mm from the Two cylindrical electrodes of a few millimetres in diameter are
exit of the nozzle can be achieved. Thus, even thick layers placed in an insulator block (usually Macor or boron nitride)
can be deposited by supersonic CBD at room temperature on with 1–2 mm spacing in between them. A pulsed valve that can
inorganic and polymeric substrates (figure 11) [56, 101]. be synchronized with the firing of the discharge supplies pure
PMCSs can be used for the production of refractory or mixed gases. The gas reaches the electrode gap through
material clusters and in particular of transition metal-oxides a channel of variable diameter and length (in the design of
and nitrides [101, 102]. Using cathodes consisting of various Siekmann et al [107] the channel has a diameter of 1 mm and

R450
 Topical Review

chemical deposition techniques are used for the patterning

of semiconductor, metallic and polymeric films and for the
production of dot arrays. These techniques require putting
resists or stamps in contact with a substrate that undergoes
different pre-or post-deposition etching or thermal treatments.
The effect of such treatments on cluster-assembled
materials has not been investigated systematically; however,
several problems due to the porosity and ‘granularity’ of
nanostructured materials can be expected. For nanostructured
films, a non-contact patterning would thus be highly desirable,
and, in principle, it would be viable e.g. by depositing
particle beams on a substrate through a stencil mask. To
be useful for practical applications this process must fulfil
several requirements such as high lateral resolution, large
area deposition, high deposition rate, surface finish and
compatibility as well as low temperature processing.
Supersonic cluster beams are characterized by high
directionality, collimation and intensity and thus are well suited
for patterned deposition of films through stencil masks that
Figure 12. Principle of the PACIS: Two metal rods are mounted are placed into the beam in front of the substrate (figure 13)
opposite to each other in a ceramic block. A pulsed arc discharge [23,110]. A remarkably high step sharpness could be obtained
produces a plasma which is flushed by a carrier gas pulse through a as reported by Barborini et al [110]: a width of 450 nm using a
nozzle into high vacuum. No additional ionizing agent is needed to mask placed at 0.33 mm from the substrate. The profile of the
charge the clusters. Figure after Ganteför et al [106].
mask is reproduced even in its sub-micrometre features. By
the use of micromachined masks patterns with micrometric
a length of 10 mm). The plasma produced by the discharge pitch can be reproduced in the deposited material. Patterning
and mixed with the carrier gas travels through a larger channel over larger areas can be realized by source rastering at the
(2 mm diameter, 10 mm length) and it expands into vacuum cost of a little reduction in the attainable resolution. Rastering
or reaches a subsequent ‘mixing chamber’ characterized by produces as a matter of fact an effective broadening of the
a variable volume and shape [108, 109], which acts as a source extension thus leading to larger penumbra regions; as
thermalization zone in a way similar to the one discussed for the typical spot size of a focused beam is about 6 mm at 300 mm
LVCSs. distance from the source, the maximum virtual extension of the
Several parameters influence the PACIS performance in source is limited to this size. Moreover, the use of alignment
terms of produced species, stability, and intensity: nucleation stages controlled by piezoelectric actuators should also allow
and neutralization processes are controlled by the gas dynamics the fabrication of structures with high definition and controlled
in the source body and by the nature of the discharge. The shapes.
geometry of a pulsed arc source can be considered in a Structures with high aspect ratio and controlled shape,
way similar to that of LVCSs. In particular, channels and arranged in ordered arrays can be created on any kind of
volumes inside the source should be designed to avoid any substrate. For instance, Milani and co-workers deposited
interference between the plasma and the source walls. For an patterns on silicon, aluminium, copper, stainless steel and
arc source, a further requirement is the electrical insulation of polyethylene [111]. Structures with heights up to several tens
the electrodes from the rest of the source body. The relevant of micrometres could be grown on metallic and polymeric
gas pulse characteristics are duration and intensity: the duration substrates while heights of several micrometres were realized
determines the quantity of gas introduced into the source on silicon. The lateral resolution was in the sub-micrometre
chamber and the pressure inside the source. In principle, by range [56]. Thus, micrometre-size objects with a structure
varying the delay between the valve opening and the discharge, and porosity on the nanoscale could be created, as the low
it is possible to change the gas dynamics between the electrodes deposition energy in CBD from supersonic beams typically
and hence to influence cluster growth. does not lead to compaction or fracture of particles upon impact
A continuous version of the arc source has been proposed: on the substrate.
it is based on an arc discharge in a hollow cathode [80]. The The lateral resolution does not depend so much on the
produced clusters are mainly ions. With this source, clusters sharpness of the mask edges as on the beam collimation.
of magnetic materials have been studied after size selected Supersonic expansions have a higher collimation compared
deposition in UHV [80]. with effusive beams; however, clusters in supersonic beams
can still have a non-negligible thermal divergence and a broad
5. Patterning and coupling to planar technologies mass distribution [56]. In order to study the effect of the
transverse cluster velocities on the step sharpness, films were
The fabrication of functional devices based on nanostructured deposited at different supersonic conditions by masking part
materials requires the ability to assemble nanoparticles in of the substrate with a straight edge mask. Assuming the film
micrometre and sub-micrometre patterns with high precision thickness as a measure for the beam flux, integrated over the
and compatibility with planar technology. Physical and deposition time, the profile of the step obtained with the mask

R451
Topical Review

Figure 14. (a) Array of 210 titania gas sensors made by CBD onto
an alumina wafer with interdigitated gold electrodes (b) employing
hard mask patterning [23]. The layer thickness represented by the
grayscale in figure 14(c) varies between the sensing elements
leading to individual different gas sensing performance.

Other patterning mechanisms have been developed for

deposition of gas phase nanoparticles; we review them briefly
here for a comparison of the resulting systems, which show
strong similarities with those obtainable by CBD, although the
deposition regime is completely different. These mechanisms
rely on low-range attractive forces, i.e. thermophoretic and
electrostatic forces effective in the moderate velocity regime
and mostly in the presence of diffusion. Electrostatic
forces are used in designing electrostatic precipitators which
capture charged particles in the gas flow through a collector
Figure 13. Schematic showing the deposition of patterned films (a) chamber. In a simple electrostatic precipitator, a high voltage
by placing a hard mask into the particle beam at some distance from difference is applied to two plates; the plate connected to
the substrate. The scanning electron micrograph (b) shows a carbon the positive pole of the power supply charges the particles
film obtained by this method with CBD. Adopted from Barborini
et al [110].
while the negative grid attracts them. An electrostatic
precipitator was used for example by Camata et al [116] for
the deposition of nanoparticles smaller than 6 nm. Jacobs and
is affected by the distribution of cluster transverse velocities. Whitesides [117] have produced submicron patterns of trapped
With a standard supersonic beam the step is very noisy and not charges on thin-films of electret polymers (polymers which
well defined. In order to increase the step sharpness one should have the ability to retain electrical charge or polarization)
increase the Mach number of the beam but this will require a using flexible electrically conductive electrodes. They could
huge pumping system [56]. The use of a focusing nozzle can pattern areas as large as one square centimetre with trapped
circumvent this problem by concentrating the clusters on the charges at a resolution better than 150 nm in less than 20 s
centre beam, keeping intact the other expansion parameters. which showed a promising application for high-resolution
Di Fonzo et al [112], for instance, used an aerodynamic lens charge-based printing. Creating charge patterns using
assembly for the deposition of patterned microstructure films. (polydimethylsiloxane) PDMS stamps, Krinke et al [18, 19]
Mazza et al [23] fabricated micropatterned sensing arrays developed a parallel process for deposition of monodisperse
by supersonic CBD. By one-step CBD a total of 210 sensors singly charged nanoparticles on oxidized silicon surfaces.
was manufactured on an alumina wafer where a 14 × 15 With their technique, they deposited gas-phase nanoparticles
array of interdigitated gold electrodes had been previously on a substrate to form nanowires with diameters as narrow as
microfabricated (figure 14). A custom-made stencil mask was 100 nm and spaced 2000 nm apart. A numerical simulation
superimposed to the substrate in order to deposit the film only of the deposition of charged nanoparticles in the size range
in the area of the interdigitated electrodes. below 100 nm in a homogeneous electric field on flat substrate
The biocompatibility of cluster-assembled titania films surfaces was presented by Krinke et al [118]. They calculated
[105] suggests patterned deposition by CBD and stencil masks the nanoparticle trajectories by solving the Langevin equation
in the fabrication of microarrays for proteomics, genomics and of motion taking into account the influence of the electric
post-genomic applications. field, the particle charge and van der Waals interactions as
If arrays of individual clusters rather than nanostructured well as Brownian diffusion. Trajectory calculations were
films are to be produced, these nanoparticles have to be pinned compared with experimental results obtained from scanning
to the surface to prevent particle diffusion and subsequent electron microscopy analysis of deposition patterns. Krinke
aggregation. This can be achieved if the impact energy of et al [118] concluded that in the case of uncharged particles on
the cluster is high enough to displace at least one surface atom the substrate, any deposited particle attracts incoming particles,
of the substrate, thus creating a reactive site that binds to and which leads to an enhanced agglomeration on the substrate
immobilizes the cluster [113–115]. rather than a random distribution. In the case of charged

R452
 Topical Review

particles on the substrate, the agglomeration on the substrate

can be reduced.
Thermophoretic forces act on particles in the gas phase
when a temperature gradient is present that results in a
momentum imbalance between the warmer and colder side
of the particle. Such forces can be used for nanoparticle
deposition and patterning. Ostraat et al [17], for example,
designed a two-stage aerosol reactor with deposition chamber
in order to integrate Si/SiO2 nanoparticles into memory
devices. Nanoparticle deposition was based on thermophoretic
forces. Their deposition chamber was designed to produce
a controllable particle density profile along the substrate
where particles were deposited uniformly. Ostraat et al [17]
also simulated the flow of aerosol nanoparticles subjected to
thermophoretic forces in their deposition chamber using a
commercial software.

6. Applications
Figure 15. (a) Schematic view of the gas-flow reactor where SiH4 is
Clusters and nanoparticles of metals oxides, carbon and decomposed by continuous or pulsed CO laser irradiation followed
semiconductors have been extensively studied in the gas phase by expansion into a molecular beam apparatus. The beam of
with particular attention to their electronic properties. Using photoluminescent silicon quantum dots was deposited though a
mask with the letters ‘MPI’. Figure 15 (b) shows the 3 mm high
CBD, nanoparticles have been deposited on various substrates letters under normal (top) and UV-light (bottom) illumination.
to study the influence of flux, kinetic energy, mass distribution Figures reproduced from Huisken et al [122].
and other parameters on the particle–surface interaction and
on the post-deposition particle organization. Apart from
confinement [123]. Voigt et al [124] investigated the growth
electronic properties, applications of cluster-assembled films
of ∼78% porous thin films of size-selected Si nanocrystals
as well as individual clusters in optics, magnetism, gas
in detail and compared it with Monte Carlo computer
sensors, catalysis and biotechnology are briefly presented in
simulations. Simultaneously with film growth they recorded
this section.
charge transport in a coplanar contact arrangement that showed
three regimes: (i) negligible contribution at the beginning of
6.1. Microelectronics, optoelectronics and optics film growth due to missing connections, (ii) a percolation type
Semiconductor-based nanocrystalline materials are the subject of superlinear increase of current with exponent 3/2 and (iii)
of considerable interest driven by their optical and electronic linear increase with further film thickness. Furthermore, Voigt
properties [6]. Group IV-based structures, for example et al [124] found that clusters of agglomerated Si nanocrystals
consisting of Ge and Si nanocrystals embedded in matrices in the film conserve the electronic properties of the single
or free-standing, exhibit visible photoluminescence that is nanocrystals.
sensitively dependent on the particle size and size distributions The production of silicon nanoparticle-based floating gate
[119]. Since CBD allows a good control over particle sizes metal oxide field effect transistors was the focus of Ostraat et al
and shapes it may offer some advantages in the preparation [17] who managed to design a gas phase nanoparticle synthesis
of nanocrystalline samples over more conventional techniques reactor system that is compatible with silicon manufacturing
for the preparation of porous silicon which are based on technology (figure 16). Silicon nanoparticles were first
electrochemical etching or plasma enhanced chemical vapour synthesized by thermal decomposition of silane and then
deposition from an rf discharge of silane [119]. passivated by surface oxidation in an ultraclean (transition
An intense source of silicon particles that were studied metal contamination <1010 atoms cm−2 ) two-stage aerosol
for their photoluminescence has been realized by Ehbrecht process reactor. Product particles were directly deposited on a
and co-workers [120]. Silicon cluster molecular beams 200 mm silicon wafer.
were produced by using a gas-flow reactor where SiH4 is Ostraat et al [17] managed to overcome any concerns
decomposed by continuous or pulsed CO laser irradiation regarding particle contamination in microelectronic device
followed by expansion into a molecular beam apparatus. Thin fabrication, as particles typically create defects leading to
nanostructured films of size-selected Si nanocrystals could be device failure and introduce transition metal contamination
grown, for example, on sapphire substrates by combining the that degrades performance. They successfully demonstrated
CBD technique with a velocity selector, namely a fast spinning that gas phase nanoparticle and cluster production can
chopper [121]. Deposition through hard masks has also been be integrated with silicon-technology based microelectronic
exploited to shape the beam and manufacture structured thin device fabrication.
films of silicon quantum dots (figure 15) [122]. Nanostructured carbon systems have been proposed as
Strong visible photoluminescence of the Si nanocrystals promising materials for electron field emission applications
upon ultraviolet radiation was found to be dependent on such as for the production of stable and inexpensive cold
cluster size and could be explained on the basis of quantum flat cathodes, as they might be superior to conventional

R453
Topical Review

Figure 17. Schematic view of a double-layer supercapacitor (left)

and picture of a prototype based on nanostructured carbon deposited
by CBD (right). Dimensions: 30 × 20 × 4 mm3 , weight: 6.3 g,
capacitance: 200 mF, equivalent series resistance: 24 . Reprinted
from Bongiorno G et al 2006 J. Mater. Sci. Mater. Electron. 17
427–41, figure 17 [126]. ©2006 with kind permission of Springer
Science and Business Media.

luminescence spectrum of pure particles with BCC structure

for diameters as low as 2.8 nm was reported to be very different
from that of bulk material and showed a broad peak at around
625 nm. As with larger nanoparticles, gap widening with the
decrease of particle size was observed for clusters with a mean
size of 3.2 nm and 2.8 nm. Crystal field symmetry breaking
induced by the surface was suggested to be a major mechanism
controlling the light emission spectrum. Contamination from
air was found to modify the luminescence of the Eu3+ ions
due to hydroxide formation, suggesting the requirement of a
passivating coating of the clusters.
Figure 16. (a) Experimental set-up showing the ultraclean aerosol Kreibig and Vollmer [128] give a comprehensive report on
reactor and deposition chamber used by Ostraat et al [17]. The
deposition chamber (b) is housed within a class 100 clean room and the optical properties of metallic clusters deposited on surfaces
allows controlled thermophoretic deposition of nanoparticles onto a or embedded in matrices. Noble metal clusters and noble
200 mm silicon wafer. Pictures reproduced with permission from metal cluster alloys embedded in transparent matrices have
Ostraat et al 2001 J. Electrochem. Soc. 148 265–70 [17]. ©2001 been prepared by CBD to study the dependence of their optical
The Electrochemical Society. properties on cluster size and stoichiometry. UV-vis optical
properties were investigated for copper, gold and silver clusters
materials in terms of extraction fields, current densities and (diameter range 1–5 nm) produced by a laser vaporization
site emission densities [125]. Using supersonic CBD, Ducati sources and deposited in an alumina matrix. Also, embedded
et al [125] produced such carbon films on double polished mixed transition metal/noble metal clusters were characterized
n-doped single crystal silicon substrates. They observed [129–131].
that samples heated during deposition showed a stable and Whangbo and co-workers [132] deposited 120 nm thin
reproducible emission, a threshold field as low as 4 V mm−1 ZnO films on Si (100) by reactive ionized CBD. Using
and a saturation current density of the order of 0.5 mA cm−2 . an evaporation source, Zn clusters were deposited at room
It was found that the presence of graphitic particles in the temperature on the substrate and oxidized by purging oxygen
films substantially enhances the field emission performance. over the film surface. Electrical and optical film properties
Working on patterned films (arrays of micrometric dots made were analysed in view of possible applications as surface
by deposition through hard masks) in the ball-tip anode acoustic wave bandpass filters, optical waveguides, laser
configuration, a localized effective conditioning could be deflectors, transparent conducting oxide coatings, gas sensors
obtained which led to an emission site density of 5×104 cm−2 . and varistors. The resistivity of the films was below
Nanostructured carbon films made by supersonic CBD 10−2 cm, the optical transmittance was about 90% in the
over large areas were applied also for energy storage in visible region, and the band-gap energy was 3.23 eV. Similarly,
supercapacitors (figure 17) [103, 126]. The carbon film Zhao et al [133] fabricated nanocluster-assembled ZnO
electrodes showed a high power density of 500 kW kg−1 and thin films by nanocluster-beam-deposition and characterized
can be deposited at room temperature on thin foils, thus their optical properties. ZnO nanoparticles were produced
ensuring compatibility with actual manufacturing processes with a magnetron sputtering gas aggregation source. The
for supercapacitors. average particle size was around 6 nm with the corresponding
Masenelli et al [127] investigated luminescence and other absorption edge located at 3.84 eV and UV emission at
fundamental properties of films resulting from the deposition 3.32 eV, respectively. After the films were annealed in air for
of Eu3+ doped Gd2 O3 clusters by low energy CBD. The 2 h, the shoulder in the absorbance spectra became steeper

R454
 Topical Review

and the intensity of UV emission in the photoluminescence

spectra became stronger due to the enhanced crystallinity.
Redshifts were observed in the evolution of absorption edge
and UV emission with increasing annealing temperature—
being ascribed to the quantum-size effect.
The formation of iron silicide thin films by CBD to
be used as thermoelectric materials was investigated by
Kizaki et al [134]. Therefore, Fe and Si precursors were
vaporized into inert gas using a DC electric arc heat source.
The aerosol formed in the colder regions of the source
chamber was expanded through a nozzle into the deposition
chamber into which a controlled stream of oxygen could be
introduced. Alternatively, two separate sources (Fe and SiO)
were employed, the beams of which were focused on the
substrate. After post-deposition annealing in Ar, the films
produced by the latter method exhibited a Seebeck coefficient
and electric conductivity very close or superior to those of the
corresponding bulk material. Figure 18. (a) Schematic illustration of the integrated system for
Aiming at the development of new multifunctional deposition of magnetic nanoclusters with magnetron sputtering
source, infrared heater for recrystallization and co-sputtering guns
devices, Peng et al [135] managed to produce nanostructured for deposition of a thin carbon overcoat on the deposited magnetic
films that simultaneously exhibit high optical transmittance, layer (b). Pictures reproduced from Qui et al [137].
low resistivity and room temperature ferromagnetism. This
was achieved by combining beam deposition of magnetic Fe
with integrated in-flight nanoparticle manipulation such as size
clusters with sputtering of indium–tin oxide (ITO) in one
selection, annealing, focusing and positioning. Thus, existing
apparatus. Thereby, a ∼ 10 nm Fe cluster layer was embedded
technologies have been advanced and new processing routes
into the ITO matrix. The material had about 90% optical
have entered the field, such as combined flame synthesis and
transmittance, a resistivity in the order of 10−4 cm and
direct thermophoretic deposition of highly crystalline Pt-doped
showed ferromagnetism at room temperature.
tin oxide nanoparticles on sensor substrates [139]. The process
resulted in 10–40 µm thick films of high (98%) porosity and
6.2. Magnetic applications 10 nm grain size which were applied in CO sensing.
The large body of work carried out in this field was recently Kennedy et al [138,140] coupled a gas phase evaporation–
reviewed in this journal by Binns and coworkers [136] and by condensation process for tin oxide nanoparticles with an in-
Sumiyama and coworkers [91]. Thus, we refrain here from flight annealing zone and a differential mobility analyser
a new review of this topic and will only point out the recent to deposit size selected nanoparticles of 10–35 nm on
work of Qui et al [137] as it shows how CBD could enter interdigitated sensor substrates by electrostatic precipitation.
mass production of thin films. Using CBD, they developed CBD was employed to fabricate 1–10 µm nanostructured
a process for fabrication of ultra-high density recording carbon films for humidity sensing (figure 19) [104, 126]. The
tape media. In this field, new cost-effective fabrication capacitive-type sensors were tested at relative humidity of
techniques are needed to increase the cost advantage of 10–70% and favourably compared with commercial devices.
magnetic tapes over disc-type storage media. In the approach Mazza et al [23] produced libraries of gas sensors for volatile
of Qui et al [137], magnetic CoPt nanoparticles with a organic carbon based on titania films deposited by CBD on
mean diameter of 5.2 nm were generated through gas-phase microfabricated substrates using stencil masks. A total of
aggregation in a magnetron sputtering source. Nanoparticles 210 sensors were made in one-step on an alumina wafer with
were then expanded into a conditioning chamber equipped a 14x15 array of interdigitated gold electrodes. Gradients
with an infrared heater for recrystallization before they in the anatase/rutile phase composition across the sensor
were deposited at room temperature on polymer substrates. array resulted in individual sensing elements with different
Using co-sputtering, a thin carbon overcoat was deposited performances. By using a combinatorial approach, such an
following the nanoparticle deposition, which stabilized the array can give more reliable information than a single sensor.
nanoparticles both mechanically and chemically. The so-
formed nanocomposite films were then applicable as tape 6.4. Catalysis
media (figure 18).
Monosize clusters deposited on inert surfaces constitute an
ideal model system for studying heterogeneously catalyzed
6.3. Gas sensors
chemical reactions and for gaining fundamental insight in
Apart from wet chemistry methods, sputtering, laser ablation the dependence of product selectivity on the catalyst size.
and gas phase condensation routes have been widely applied in CBD has been used frequently to obtain such model catalysts.
the production of solid-state gas sensing devices, as is apparent A typical experimental set-up has been reported by Heiz
from the tabulation of Kennedy et al [138] for tin oxide sensors. et al [97] who generated clusters by high frequency laser
Today, the focus is on combined processes for synthesis and vaporization and subsequent supersonic expansion of the metal
direct deposition, ideally without thermal after-treatment, but vapour. Cluster generation was followed by mass selection

R455
Topical Review

Figure 19. Sketch of the layers composing a capacitive humidity sensor (left) and picture of a capacitive sensor based on nanostructured
carbon (right). Two serial capacitors are realized by first evaporating two rectangular Au electrodes (rear electrodes) on a glass substrate; the
deposition of the ns-C film has then been carried out onto the rear Au electrodes to produce the dielectric layer, and a thin and permeable Au
layer electrode was evaporated on top. A gold thermoresistance sensor is buried under the ns-C film in order to measure the temperature of
the device. Reprinted from Bongiorno et al 2006 J. Mater. Sci. Mater. Electron. 17 427–41, figure 19 [126]. ©2006 with kind permission
of Springer Science and Business Media.

using a quadrupole mass spectrometer and UHV deposition

on thin oxide films grown on metal single crystals. The
equipment of Heiz et al [97] was equipped with in situ FTIR,
temperature desorption and Auger electron spectroscopy and
was used to study CO on Ni clusters on thin MgO films.
Abbet et al [141] used a similar source for generating
size selected clusters of Ni, Si and Au to be used as
monodisperse model catalysts. Specifically, they investigated
the adsorption of CO on Pd atoms and CO oxidation on Au
clusters. Aizawa et al [142] used this type of equipment to
Figure 20. Patterned nanostructured titanium oxide thin film
investigate deposition dynamics and chemical properties of deposited through a hard mask onto a standard microscope glass
size-selected Ir clusters on TiO2 , focusing on CO adsorption slide for application in microarray technology.
and desorption. Noble metal clusters of Pd [143] and bimetallic
Pd–Pt [144] were deposited on graphite and alumina substrates, of a microscope, representing the ultimate limit in biochip
respectively, employing laser vaporization and CBD. Both sensitivity.
cluster types were employed as model catalysts for studying Carbone et al [105] characterized the biocompatibility
the hydrogenation of unsaturated hydrocarbons. of nanostructured TiO2 films produced by the deposition of
Anpo and co-workers intensively studied cluster beam a supersonic beam of TiOx clusters. The cluster-assembled
methods to produce various photocatalysts: Ti–Si binary oxide film supported normal growth and adhesion of primary and
thin film photocatalysts for decomposition of NO into N2 and cancer cells with no need for coating with extracellular
O2 under UV light irradiation [145], Pt/TiO2 for photocatalytic matrix proteins. Physical analysis showed that the films
oxidation of acetaldehyde with O2 under UV light irradiation possess a nanoscale granularity and porosity mimicking
[146] and combined CBD–metal ion implantation for doped that of typical extracellular matrix structures and adsorption
TiO2 photocatalysts [147]. A layer of V-ion-implanted TiO2 properties that could allow surface functionalization with
on quartz substrates showed photocatalytic decomposition of different macromolecules such as DNA proteins and peptides.
formic acid even under visible light radiation [147]. Films of nanostructured titania are proposed as an optimal
substrate for different applications in cell-based assays,
6.5. Biotechnological applications biosensors, bioactive orthopedic, dental and vascular implants
or microfabricated medical devices. Such films may be
Palmer and co-workers [114, 148] immobilized proteins realized even at a commercial production scale as CBD using
(GroEL chaperonin molecule from Escherichia Coli, pHAT- a PMCS can be developed into a continuous high-throughput
GFP and Human Oncostatin M protein molecules) on coating technique (figure 20).
individual Au clusters that were pinned to a graphite substrate.
The clusters were created with a magnetron sputtering, gas 7. Conclusions and outlook
aggregation cluster beam source and size-selected with a time-
of-flight mass spectrometer. Immersion of the Au-cluster The use of beams of clusters and nanoparticles in the gas
decorated substrate in a buffer solution containing protein phase has reached a degree of complexity and maturity that
molecules resulted in preferential binding of the proteins to the makes this approach very promising for the production of
Au clusters, as was witnessed by AFM imaging. Thus, well- nanostructured materials and systems. The transition from
separated arrays of immobilized proteins could be produced. the use of CBD as a tool for basic research to a method
This technique would permit single molecule optical studies for the fabrication of nanostructured systems is driven by
in which only one protein was present in the field of view the combination of supersonic expansions with aerodynamic

R456
 Topical Review

focusing. A careful control on the aerodynamic conditions [26] Liu P, Ziemann P J, Kittelson D B and McMurry P H 1995
during cluster production, extraction and deposition is of Aerosol Sci. Technol. 22 293–313
[27] Liu P, Ziemann P J, Kittelson D B and McMurry P H 1995
fundamental importance to meet the requisites necessary for Aerosol Sci. Technol. 22 314–24
applications: high intensity, stability, repeatability, high lateral [28] Friedlander S K 2000 Smoke, Dust and Haze (New York:
resolution and parallel processing. Oxford University Press)
The cluster beam approach is inherently compatible with [29] Hagena O F and Obert W 1972 J. Chem. Phys. 56 1793–802
[30] Siegel R W 1991 Annu. Rev. Mater. Sci. 21 559–78
planar technologies typical of thin film deposition and it should
[31] Girshick S L, Chiu C P, Muno R, Wu Y U, Yang L, Singh S K
be further developed in order to improve the compatibility with and McMurry P H 1993 J. Aerosol Sci. 24 367–82
silicon-based microfabrication technologies. The achievement [32] Pratsinis S E 1998 Prog. Energy Combust. Sci. 24 197–219
of this goal will assure CBD a pivotal position in the efforts [33] Bandyopadhyaya R, Lall A A and Friedlander S K 2004
towards the merging of nanofabrication with microfabrication. Powder Technol. 139 193–9
[34] Flagan R C and Lunden M M 1995 Mater. Sci. Eng. A
204 113–24
[35] Granqvist C G and Buhrman R A 1976 J. Appl. Phys.
Acknowledgments 47 2200–19
[36] Pratsinis S E, Wang G, Panda S, Guiton T and Weimer A 1995
We thank G Bongiorno for his help in the preparation of the J. Mater. Res. 10 512–20
manuscript. [37] Ulrich G D 1971 Combust. Sci. Technol. 4 47–57
[38] Wegner K, Walker B, Tsantilis S and Pratsinis S E 2002 Chem.
Eng. Sci. 57 1753–62
References [39] Landgrebe J D and Pratsinis S E 1990 J. Colloid Interface Sci.
139 63–86
[1] Ulrich G D 1984 Chem. Eng. News 62 22–9 [40] Xiong Y and Pratsinis S E 1993 J. Aerosol Sci. 24 283–300
[2] Stark W J and Pratsinis S E 2002 Powder Technol. 126 103–8 [41] Fuchs N A and Sutugin A G 1971 Topics in Current Aerosol
[3] Donnet J-B, Bansal R C and Wang M-J (ed) 1993 Carbon Research ed G M Hidy and J R Brock (New York:
Black (New York: Dekker) Pergamon)
[4] Braun J H, Baidins A and Marganski R E 1997 Prog. Org. [42] Okuyama K, Kousaka Y, Tohge N, Yamamoto S, Wu J J,
Coat. 20 105–38 Flagan R C and Seinfeld J H 1986 AIChE J. 32 2010–19
[5] Diebold U 2003 Surf. Sci. Rep. 48 53–229 [43] Bowles R S, Park S B and Andres R P 1983 J. Mol. Catal.
[6] Nalwa H S (ed) 2004 Encyclopedia of Nanoscience and 20 279–87
Nanotechnology (Stevenson Ranch, CA: American [44] Buffat P and Borel J P 1976 Phys. Rev. A 13 2287–98
Scientific Publishers) [45] Wu M K, Windeler R S, Steiner C K R, Börs T and
[7] Roco M 2005 J. Nanopart. Res. 7 707–12 Friedlander S K 1993 Aerosol Sci. Technol. 19 527–48
[8] Clavin R K, Zhirnov V V, Bourianoff G I, Hutchby J A, [46] Xing Y and Rosner D E 1999 J. Nanopart. Res. 1 277–91
Herr D J C, Hosack H H, Joyner W H and Wooldridge T A [47] Tsantilis S, Briesen H and Pratsinis S E 2001 Aerosol Sci.
2005 J. Nanopart. Res. 7 573–86 Technol. 34 237–46
[9] Siegel R W, Hu E and Roco M C (ed) 1999 Nanostructure [48] Kruis F E, Maisels A and Fissan H 2000 AIChE J. 46 1735–42
Science and Technology: A Worldwide Study (Loyola [49] Pratsinis S E 1988 J. Colloid Interface Sci. 124 416–27
College, Maryland: National Science and Technology [50] Hounslow M J, Ryall R L and Marshall V R 1988 AIChE J.
Council (NSTC)) available from: http://www.wtec.org/ 34 1821–32
loyola/nano/final/ [51] Tsantilis S and Pratsinis S E 2004 Langmuir 20 5933–9
[10] Wegner K and Pratsinis S E 2003 Chem. Eng. Sci. 58 4581–9 [52] Wegner K, Stark W J and Pratsinis S E 2002 Mater. Lett.
[11] Kammler H K, Mädler L and Pratsinis S E 2001 Chem. Eng. 55 318–21
Technol. 24 583–96 [53] Ramsey N F 1955 Molecular Beams (Oxford: Oxford
[12] Singh Y, Javier J R N, Ehrman S H, Magnusson M H and University Press)
Deppert K 2002 J. Aerosol Sci. 33 1309–25 [54] Hagena O F 1974 Molecular Beams and Low Density Gas
[13] Wegner K and Pratsinis S E 2003 AIChE J. 49 1667–75 Dynamics ed P P Wegener (New York: Dekker)
[14] Nakaso K, Shimada M, Okuyama K and Deppert K 2002 J. [55] Beijering H and Verster N 1981 Physica C 11 327–52
Aerosol. Sci. 33 1061–74 [56] Piseri P, Podestà A, Barborini E and Milani P 2001 Rev. Sci.
[15] Nanda K K, Kruis F E, Fissan H and Acet H 2002 J. Appl. Instrum. 72 2261–7
Phys. 91 2315–21 [57] 1946 Oxford Report #36 UKAEA, Harwell, BR 694, p 2
[16] Karlsson M N A, Deppert K, Magnusson M H, Karlsson L S [58] Becker E W and Bier K 1954 Z. Naturf. 9 a 975–86
and Malm J-O 2004 Aerosol Sci. Technol. 38 948–54 [59] Becker E W 1979 Top. Appl. Phys. 35 245–68
[17] Ostraat M L, De Blauwe J W, Green M L, Bell L D, [60] Ude S and Fernández de la Mora 2003 J. Aerosol Sci.
Atwater H A and Flagan R C 2001 J. Electrochem. Soc. 34 1245–66
148 G265–70 [61] Reis V H and Fenn J B 1963 J. Chem. Phys. 39 3240–50
[18] Krinke T J, Deppert K, Magnusson M H and Fissan H 2002 [62] Waterman P C and Stern S A 1959 J. Chem. Phys. 31 405–19
Part. Part. Syst. Charact. 19 321–6 [63] Murphy W K and Sears G W 1964 J. Appl. Phys. 85 1986–7
[19] Krinke T J, Fissan H, Deppert K, Magnusson M H and [64] Israel G W and Friedlander S K 1967 J. Colloid Interface Sci.
Samuelson L 2001 Appl. Phys. Lett. 78 3708–10 24 330–7
[20] Milani P and Iannotta S 1999 Cluster Beam Synthesis of [65] Dahneke B E and Flachsbart H 1972 J. Aerosol Sci. 3 345–9
Nanostructured Materials (Berlin: Springer) [66] Fernández de la Mora J and Riesco-Chueca P 1988 J. Fluid
[21] de Heer W A 1993 Rev. Mod. Phys. 65 611–76 Mech. 195 1–21
[22] Milani P, Piseri P, Barborini E, Podestà A and Lenardi C 2001 [67] Fernández de la Mora J and Rosell-Llompart J 1989 J. Chem.
J. Vac. Sci. Technol. A 19 2025–33 Phys. 91 2603–15
[23] Mazza T et al 2005 Appl. Phys. Lett. 87 103108 [68] Fuerstenau S, Gomez A and Fernández de la Mora J 1994 J.
[24] Piseri P, Vahedi Tafreshi H and Milani P 2004 Curr. Opin. Aerosol Sci. 25 165–73
Solid State Mater. Sci. 8 195–202 [69] Mallina R V, Wexler A S and Johnston M V 1999 J. Aerosol
[25] Wang X, Kruis, F E and McMurry P H 2005 Aerosol Sci. Sci. 30 719–38
Technol. 39 611–23 [70] Goo J 2002 J. Aerosol Sci. 33 1493–507

R457
Topical Review

[71] Lee J W, Yi M Y and Lee S M 2003 J. Aerosol Sci. 34 211–24 [111] Milani P, Piseri P, Barborini E, Lenardi C, Castelnovo C and
[72] Zhang X F, Smith K A, Worsnop D R, Jimenez J, Jayne J T Podestà A 2001 Nanostructured Carbon for Advanced
and Kolb C E 2002 Aerosol Sci. Technol. 36 617–31 Applications (NATO Advanced Study Institute Series II:
[73] Vahedi Tafreshi H, Benedek G, Piseri P, Vinati S, Barborini E Mathematics, Physics and Chemistry) ed G Benedek et al
and Milani P 2001 Eur. Phys. J. Appl. Phys. 16 149–56 (Dordrecht: Kluwer)
[74] Vahedi Tafreshi H, Benedek G, Piseri P, Vinati S, Barborini E [112] Di Fonzo F, Gidwani A, Fan M H, Neumann D,
and Milani P 2002 Aerosol Sci. Technol. 36 593–606 Iordanoglou D I, Heberlein J V R, Tymiak N, Gerberich W
[75] Vahedi Tafreshi H, Piseri P, Barborini E, Benedek G and W and Rao N P 2000 Appl. Phys. Lett. 77 910–2
Milani P 2002 J. Nanopart. Res. 4 511–24 [113] Carroll S J, Pratontep S, Streun M, Palmer R E, Hobday S
[76] Middha P and Wexler AS 2003 Aerosol Sci. Technol. and Smith R 2000 J. Chem. Phys. 113 7723–7
37 907–15 [114] Palmer R E, Pratontep S and Boyen H-G 2003 Nature Mater.
[77] Mallina R V, Wexler A S, Rhoads K P and Johnston M V 2000 2 443–8
Aerosol Sci. Technol. 33 87–104 [115] Melinon P, Hannour A, Prevel B, Bardotti L, Bernstein E,
[78] Fernández de la Mora J 1996 Chem. Eng. Commun. 151 Perez A, Gierak J, Bourhis E and Mailly D 2005 J. Crystal
101–24 Growth 275 317–25
[79] Gentry W R 1988 Atomic and Molecular Beam Methods [116] Camata R P, Atwater H A, Vahala K J and Flagan R C 1996
ed G Scoles (Oxford: Oxford University Press) Appl. Phys. Lett. 68 3162–4
[80] Bansmann J et al 2005 Surf. Sci. Rep. 56 189–275 [117] Jacobs H O and Whitesides G M 2001 Science
[81] Ross K J and Sonntag B 1995 Rev. Sci. Instrum. 66 4409–33 291 1763–6
[82] Sattler K, Muehlbach J and Recknagel E 1980 Phys. Rev. Lett. [118] Krinke T J, Deppert K, Magnusson M H, Schmidt F and
45 821–4 Fissan H 2002 J. Aerosol Sci. 33 1341–59
[83] Gatz P and Hagena O F 1995 Appl. Surf. Sci. 91 169–74 [119] Ehbrecht M, Kohn B, Huisken F, Laguna M A and Paillard V
[84] Raizer Y P 1991 Gas Discharge Physics (Berlin: Springer) 1997 Phys. Rev. B 56 6958–64
[85] Lieberman M A and Lichtemberg A J 1994 Principles of [120] Ehbrecht M, Ferkel H and Huisken F 1997 Z. Phys. D
Plasma Discharges and Materials Processing (New York: 40 88–92
Wiley) [121] Laguna M A, Paillard V, Kohn B, Ehbrecht M, Huisken F,
[86] Haberland H, Karrais M, Mall M and Thurner Y 1992 J. Vac. Ledoux G, Papoular R and Hofmeister H 1999 J. Lumin.
Sci. Technol. A 10 3266–71 80 223–8
[87] Haberland H, Mall M, Moseler M, Qiang Y, Reiners T and [122] Huisken F, Kohn B and Paillard V 1999 Appl. Phys. Lett.
Thurner Y 1994 J. Vac. Sci. Technol. A 12 2925–30 74 3776–8
[88] Hihara T and Sumiyama K 1998 J. Appl. Phys. 84 5270–6 [123] Huisken F, Ledoux G, Guillois O and Reynaud C 2002 Adv.
[89] Yamamuro S, Sumiyama K and Suzuki K 1999 J. Appl. Phys. Mater. 14 1861–5
85 483–9 [124] Voigt F, Brüggemann R, Unold T, Huisken F and Bauer G H
[90] Pratontep S, Carroll S J, Xirouchaki C, Streun M and 2005 Mater. Sci. Eng. C 25 584–9
Palmer R E 2005 Rev. Sci. Instrum. 76 045103 [125] Ducati C, Barborini E, Piseri P, Milani P and Robertson J
[91] Sumiyama K, Hihara T, Peng D L and Katoh R 2005 Sci. 2002 J. Appl. Phys. 92 5482–9
Technol. Adv. Mater. 6 18–26 [126] Bongiorno G, Podestà A, Ravagnan L, Piseri P, Milani P,
[92] Maruyama S, Anderson L R and Smalley R E 1990 Rev. Sci. Lenardi C, Miglio S, Bruzzi M and Ducati C 2006 J. Mater.
Instrum. 61 3686–93 Sci.: Mater. Electron. 17 427–41
[93] Milani P and deHeer W A 1990 Rev. Sci. Instrum. 61 1835–38 [127] Masenelli B et al 2005 Eur. Phys. J. D 34 139–43
[94] Dietz T G, Duncan M A, Powers D E and Smalley R E 1981 J. [128] Kreibig U and Vollmer M 1995 Optical Properties of Metal
Chem. Phys. 74 6511–2 Clusters (Berlin: Springer)
[95] Geusic M E, Morse M D, O’Brien S C and Smalley R E 1985 [129] Gaudry M, Lerme J, Cottancin E, Pellarin M, Vialle J L,
Rev. Sci. Instrum. 56 2123–30 Broyer M, Prevel B, Treilleux M and Melinon P 2001 Phys.
[96] Woenckhaus J and Becker J A 1994 Rev. Sci. Instrum. Rev. B 64 085407
65 2019–22 [130] Gaudry M et al 2003 Phys. Rev. B 67 155409
[97] Heiz U, Vanolli F, Trento L and Schneider W-D 1997 Rev. Sci. [131] Celep G et al 2004 Phys. Rev. B 70 165409
Instrum. 68 1986–94 [132] Whangbo S W, Jang H K, Kim S G, Cho M H, Jeong K and
[98] Pellarin M et al 1994 Chem. Phys. Lett. 224 338–44 Whang C N 2000 J. Korean Phys. Soc. 37 456–60
[99] Barborini E, Piseri P and Milani P 1999 J. Phys. D: Appl. [133] Zhao Z W, Tay B K, Chen J S, Hu J F, Sun X W and Tan S T
Phys. 32 L105–9 2005 Appl. Phys. Lett. 87 251912
[100] Vahedi Tafreshi H, Piseri P, Benedek G and Milani P 2006 J. [134] Kizaki Y, Osada H and Aoki H 1993 Japan. J. Appl. Phys.
Nanosci. Nanotechnol. 6 1140–9 32 5778–90
[101] Bongiorno G et al 2005 J. Nanosci. Nanotechnol. 5 1072–80 [135] Peng D L, Sumiyama K, Nozawa N, Hihara T and
[102] Bongiorno G et al 2005 Carbon 43 1460–9 Morikawa H 2005 IEEE Trans.Magnetics 41 3406–8
[103] Diederich L, Barborini E, Piseri P, Podestà A, Milani P, [136] Binns C et al 2005 J. Phys. D: Appl. Phys. 38 R357–79
Schneuwly A and Gallay R 1999 Appl. Phys. Lett. [137] Qui J-M, Xu Y-H, Judy J H and Wang J-P 2005 J. Appl. Phys.
75 2662–4 97 10P704
[104] Bruzzi M, Miglio S, Scaringella M, Bongiorno G, Piseri P, [138] Kennedy M K, Kruis F E, Fissan H, Mehta B R, Stappert S
Podestà A and Milani P 2004 Sensors Actuators B and Dumpich G 2003 J. Appl. Phys. 93 551–60
100 173–6 [139] Mädler L, Roessler A, Pratsinis S E, Sahm T, Gurlo A,
[105] Carbone R et al 2006 Biomaterials. 27 3221–9 Barsan N and Weimar U 2006 Sensors Actuators B
[106] Ganteför G, Siekmann H R, Lutz H O and 114 283–95
Meiwes-Broer K H 1990 Chem. Phys. Lett. 165 293–6 [140] Kennedy M K, Kruis F E, Fissan H, Nienhaus H, Lorke A
[107] Siekmann H R, Lüder C, Faehrmann J, Lutz H O and and Metzger T H 2005 Sensors Actuators B 108 62–9
Meiwes-Broer K H 1991 Z. Phys. D 20 417–20 [141] Abbet S, Judai K, Klinger L and Heiz U 2002 Pure Appl.
[108] Siekmann H R, Holub-Krappe E, Wrenger B, Pettenkofer C Chem. 74 1527–35
and Meiwes-Broer K H 1993 Z. Phys. B 90 201–6 [142] Aizawa M, Lee S and Anderson S L 2003 Surf. Sci.
[109] Cha C Y, Ganteför G and Eberhardt W 1992 Rev. Sci. 542 253–75
Instrum. 63 5661–6 [143] Cadete Santos Aires F J, Sautet P, Rousset J-L, Fuchs G
[110] Barborini E, Piseri P, Podestà A and Milani P 2000 Appl. and Mélion P 1994 J. Vacuum. Sci. Technol. B
Phys. Lett. 77 1059–61 12 1776–9

R458
 Topical Review

[144] Rousset J L, Cadete Santos Aires F J, Bornette F, Cattenot M, [147] Zhou J, Takeuchi M, Zhao X S, Ray A K and
Pellarin M, Stievano L and Renouprez A J 2000 Appl. Surf. Anpo M 2006 Catal. Lett. 106 67–70
Sci. 164 163–8 [148] Collins J A, Xirouchaki C, Palmer R E,
[145] Takeuchi M, Dohshi S, Eura T and Anpo M 2003 J.Phys. Heath J K and Jones C H 2004 Appl. Surf. Sci.
Chem. B 107 14278–82 226 197–208
[146] Takeuchi M, Tsujimaru K, Sakamoto K, Matsuoka M, [149] Gidwani A 2003 PhD Thesis Department of
Yamashita H and Anpo M 2003 Res. Chem. Intermediates Mechanical Engineering, University of Minnesota,
29 619–29 Minneapolis 55455, USA

R459