Multiple atomic scale solid surface interconnects for atom circuits and molecule logic gates

Abstract

The scientific and technical challenges involved in building the planar electrical connection of an atomic scale circuit to N electrodes (N > 2) are discussed. The practical, laboratory scale approach explored today to assemble a multi-access atomic scale precision interconnection machine is presented. Depending on the surface electronic properties of the targeted substrates, two types of machines are considered: on moderate surface band gap materials, scanning tunneling microscopy can be combined with scanning electron microscopy to provide an efficient navigation system, while on wide surface band gap materials, atomic force microscopy can be used in conjunction with optical microscopy. The size of the planar part of the circuit should be minimized on moderate band gap surfaces to avoid current leakage, while this requirement does not apply to wide band gap surfaces. These constraints impose different methods of connection, which are thoroughly discussed, in particular regarding the recent progress in single atom and molecule manipulations on a surface.

Full text

IOP PUBLISHING JOURNAL OF PHYSICS: CONDENSED MATTER
J. Phys.: Condens. Matter 22 (2010) 084025 (18pp) doi:10.1088/0953-8984/22/8/084025
Multiple atomic scale solid surface
interconnects for atom circuits and
molecule logic gates
C Joachim1,2,DMartrou
1, M Rezeq2, C Troadec2,DengJie
2,
N Chandrasekhar2and S Gauthier1
1Centre d’Elaboration de Mat´eriaux et d’Etudes Structurales (CEMES-CNRS) 29, rue Jeanne
Marvig, BP 94347, 31055 Toulouse Cedex 4, France
2Institute of Materials Research and Engineering, A*STAR (Agency for Science, Technology
and Research), 3 Research Link, 117602, Singapore
Received 5 June 2009, in final form 12 September 2009
Published 5 February 2010
Online at stacks.iop.org/JPhysCM/22/084025
Abstract
The scientific and technical challenges involved in building the planar electrical connection of
an atomic scale circuit to Nelectrodes (N>2) are discussed. The practical, laboratory scale
approach explored today to assemble a multi-access atomic scale precision interconnection
machine is presented. Depending on the surface electronic properties of the targeted substrates,
two types of machines are considered: on moderate surface band gap materials, scanning
tunneling microscopy can be combined with scanning electron microscopy to provide an
efficient navigation system, while on wide surface band gap materials, atomic force microscopy
can be used in conjunction with optical microscopy. The size of the planar part of the circuit
should be minimized on moderate band gap surfaces to avoid current leakage, while this
requirement does not apply to wide band gap surfaces. These constraints impose different
methods of connection, which are thoroughly discussed, in particular regarding the recent
progress in single atom and molecule manipulations on a surface.
(Some figures in this article are in colour only in the electronic version)
1. Introduction
Single molecule mechanics [1], mono-molecular electronics [2]
and multi-probe experiments on atomic scale constructed
devices [3] or machinery (memory, transducers) [4] require the
development of a specific surface interconnection technology
able to provide a large number of information and energy
access channels to the atomic (or molecular) scale machinery
constructed on a surface (see figure 1) with atomic precision
and cleanliness [5]. Atomic precision is compulsory for
all these devices and machinery because when the end
of a molecule electronically interacts with its metallic
interconnection pads, an optimum contact conductance is
obtained for inter-atomic distances between the 0.2 nm
chemisorption and the 0.3 nm van der Waals distance
ranges. For a conjugated molecular wire connected to copper
electrodes, the optimum was demonstrated to be around
0.25 nm [6]. A 0.2 nm adsorption distance would lead to
a strong electronic coupling between the molecule πsystem
and the pads, changing the pad metallic density of state
locally. This change increases the reflection coefficient of
the metallic Bloch waves on the pad–molecule–pad junction
and therefore decreases its conductance [6]. An adsorption
distance above 0.3 nm would be too large for an overlap
between the πelectronic system of the molecule and the pad
yet large enough to lead to efficient electron transfer events.
The required 0.05 nm precision in distance implies a perfect
control of each atom position on the pads and of the atomic
composition of the corresponding contact surface where the
molecular electronic cloud will overlap with the metallic pads.
Therefore, atom control and atomic precision are the keys
to success for perfectly reproducible experiments on a single
molecule for devices or machine applications, as illustrated in
figure 1. Mastering such a precision will avoid the statistical
approach that is commonly used, for example in molecular
wire conductance measurements on ill-defined systems [7].
0953-8984/10/084025+18$30.00 ©2010 IOP Publishing Ltd Printed in the UK1

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 1. A single five wings molecule motor [1] positioned between
a 4 Au nanopad junction constructed on a Si(100)-Hsurface. The
motor is driven by the tunneling electrons sequentially transferred
through the motor wings. The 4 black wires coming out of the
surface are indicative of the interconnection step 3, discussed in the
text, depending on the electronic gap of the supporting surface.
At the end of the 1980s, it was expected that e-
beam nanolithography would be able to provide such an
interconnection technology [8]. But with its resist based
approach (be it a grown, a deposited or a contamination
resist), e-beam nanolithography can not make it in a planar
approach [9], because it is not able to respect simultaneously
the atomic scale precision, cleanliness and the expected large
number, N, of access channels to the atomic or molecular
scale machinery [10]. Alternative new surface nanolithography
techniques, such as nano-imprint [11] or nanostencil [12], are
neither adapted to encompass all the interconnection stages
from the macroscopic to the atomic scale, nor clean enough
down to the atomic scale. At the turn of the century, this
problem associated with new targets in atomic scale machinery
such as molecule logic gates [13], molecule motors [1]
and surface atomic scale electronic circuits [14] triggered
a new approach starting from the bottom; that is from the
fundamentals of surface science.
This review describes the new laboratory scale approaches
explored today to solve this multi-access atomic scale precision
interconnection problem for electrical interconnects. The same
question can be put forward for photonic multi-access to the
atomic scale in a planar approach [15]. In section 2,the
general principles of the few UHV interconnection machines
in construction are described. We distinguish 2 types of
machines, depending on the electronic gap of the surface,
where the atomic scale devices and machinery are supposed
to run. Section 3describes an atomic scale interconnection
machine devoted to wide gap semiconductor and insulator
materials. Section 4describes a machine devoted to moderate
gap semiconductor surfaces. In each case, the surface
science problems experimented, solved or yet to be solved are
described. In conclusion, we discuss the large Nlimit of this
atomic scale interconnection approach with electrons and how
to pass from laboratory scale UHV interconnection machines
to a more industrial-like approach. The work reported here
is one output of the EU ICT integrated project Pico-Inside
in Toulouse and of the VIP A*STAR Atom Tech program in
Singapore.
2. Atomically precise electrical interconnection
machine
An atomic scale precision multiple access interconnection
machine must provide Nconducting wires converging toward
a very small surface area where an active machine (see figure 1
for a N=4 example) has been constructed or assembled
with atomic scale precision. These Ninterconnects are
positioned somewhere on a large handleable wafer surface.
As a consequence, a very efficient navigation system must
be designed to locate this very small active area from
a macroscopic perspective while keeping the local atomic
precision of the interconnection construction. The solution
to this navigation requirement is to combine two types of
microscopy: a far field one (optical microscopy, scanning
electron microscopy (SEM)) for large scale navigation and a
near field one (scanning tunneling microscopy (STM), atomic
force microscopy (AFM)) for the atomic scale part, with a
compulsory overlap between these 2 types of microscopy.
In air, the combination of an optical microscope with a
standard AFM to interconnect a single wall carbon nanotube
on metallic nanoelectrodes made by e-beam nanolithography
has already shown the power of this combination of two
microscopes, associated with a dedicated way to electrically
contact the nanopads [16]. The contacts were taken by
introducing, under the AFM head, a comb of metallic
microcantilever electrodes to minimize the overall length of the
electrical circuit fabricated on the wafer surface [16]. While
not atomic scale, this proof of concept demonstrates all the
ingredients that are at the basis of the next generation UHV
and atomic scale interconnection machines, as schematically
presented in figure 2.
A UHV atomic scale interconnection machine is designed
to follow a dedicated interconnection sequence.Onan
atomically clean well-prepared surface, an atomic scale
circuitry is fabricated (A). This circuit must have a minimal
lateral extension to make possible its connection to a large
number Nof metallic nanopads (B) positioned around it.
In the example of figure 2, a molecule is connected to
these nanopads by atomic metallic wires. Depending on
the electronic surface gap of the supporting material, the
nanopads (B) have to be contacted from the top by a
series of Natomically sharp metallic tips (C1) or by a
series of Nnanoscale wires (C2) up to the point where
mesoscopic metallic wiring or microelectrodes (D) can be
surface fabricated and contacted by Nmicroscale metallic
cantilevers (E) also from the top of the wafer. What determines
the choice of the interconnection technology between C1 and
C2 (and after the need for the D and E interconnection steps
in figure 2(b)) is the electronic gap of the surface, which in
turn dictates the kind of far field microscopy to be used for
navigation over the wafer surface.
For a large valence band–conduction band (VB–CB)
electronic surface gap (more than a few eV up to 12 eV
for good insulators), SEM is difficult to use because the
electron beam will charge the surface. In this case, an
optical microscope must be used. This microscope determines
the minimum length of metallic surface wiring that must be
2

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Table 1. Some of the most studied wide band gap semiconductors and insulators.
Material KBr MgO Al2O3CaF24H-SiC 2H-GaN 2H-AlN
Bulk band gap
(eV)
6.6 [88]7.8[89]8[90] 12.1 [91]3.07[92]3.47[92]6.28[92]
Figure 2. Scheme of the atomic scale interconnection machines for
(b) wide and (c) moderate surface band gap substrates. (a) and A: the
atomic scale circuitry embedded into the surface, B: contacting
metallic nanopads, C1: ultra-sharp metallic tips, C2: nanowires,
D: microelectrodes, E: metallic microcantilevers.
fabricated starting from the nanopads (B) toward the next
contact stage based on metallic microcantilevers (figure 2(b)).
Fortunately, with a large surface gap, the surface area of those
interconnects can be expanded laterally without significant
leakage currents between the different electrodes. This is
the basis of the UHV interconnection machine described in
section 3where a low temperature approach is not compulsory.
For a moderate VB–CB electronic surface gap (around a
few eV), it is not possible to use very long surface metallic
circuitry due to the leakage currents that will appear between
the surface electrodes. In this case, one solution is to use ultra-
sharp STM-like tips positioned from the top on the nanopads
(figure 2(c)). In this case, the bulk of the tip is not in contact
with the supporting surface and one goes continuously from
the tip apex radius of curvature of a few nanometer up to
the 100 μm or more of a section of the tip body. In this
case, navigation on the surface can be performed using an
UHV-SEM (figure 2(c)) by grounding the sample during SEM
imaging to avoid charging. This is the basis of the UHV
interconnection machine described in section 4. Here, a low
temperature approach is compulsory to limit leakage currents
due to the low electronic gap of the surface of the supporting
material.
3. Large electronic surface gap supporting materials
As presented in figure 2(b), a large surface electronic gap
for the material supporting the interconnection structure leads
to 5 levels of interconnects. Levels (A) and (B) depend
on our mastering of atomic scale surface science phenomena
while level (C), (D) and (E) rely on mesoscopic scale
surface phenomena such as surface wetting, associated with
developments in UHV instrumentation. The starting point of a
multi-access interconnection technology on a large electronic
gap surface material is the careful selection and surface
preparation of the supporting material. In this section, the
progress and the problems that remain to be solved for these 5
levels are presented up to the description of the interconnection
machine developed in Toulouse.
3.1. Atomic scale circuitry (level A)
3.1.1. Surface selection and preparation. For natural as well
as for artificial mineral crystalline compounds, a wide band
gap is more the rule than the exception. But, in this huge class
of materials, only a few have been investigated in detail by
surface science techniques, mainly because of the inadequacy
of techniques using electron beams to study them. Stringent
conditions should be satisfied to make possiblethe construction
of the first atomic scale level (A) of the interconnection
sequence presented in figure 2(b): the surface should be (1)
atomically clean, with a well-controlled stoichiometry, (2)
atomically flat, with large enough terraces and (3) should
present few defects. In these conditions, a typical 50 nm ×
50 nm ‘perfect’ area, suitable to accommodate the atomic
scale device, becomes readily accessible. A selection of wide
band gap materials that have been studied in detail, either for
their simplicity or for their applications, and that satisfy these
requirements, is presented in table 1.
KBr is representative of the family of alkali halides, the
archetype of ionic compounds. The most stable surface,
prepared by cleavage, is the non-polar (001) surface. Large
atomically clean terraces are easily obtained. MgO is one
of the simpler ionic oxides, used for instance as a model
catalyst. Cleavage allows a clean surface to be obtained
with the nominal stoichiometry of the material. The surfaces
of the different phases of alumina have been studied for a
3

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
long time, but despite their technological interest, their atomic
structure remains largely unknown. They are difficult to
prepare and usually present complex phase diagrams, where
the surface stoichiometry plays a determining role. CaF2is an
ionic insulator that has been considered for a long time as a
good candidate for an insulating material in microelectronics
due to its small lattice mismatch with Si. It is nowadays
investigated for its optical properties for the fabrication of
lenses for deep UV optical lithography. The CaF2(111)surface
can be obtained by cleavage. It is atomically flat over large
areas. SiC is one of the most studied wide band gap material,
finding applications in such diverse area as power electronics,
microelectromechanical systems and as a supporting material
for graphene. It exists in different polytypes, which display
a rich variety of surface reconstruction, some of which have
been investigated in detail by STM [17]. More recently,
group-III nitride semiconductors (GaN, AlN) have attracted
much interest for their applications as short wavelength optical
emitters and detectors, as well as for high-speed electronics. A
large variety of surface structures is obtained depending on the
growth method and conditions. Some of these surfaces have
been studied in detail, in particular by STM [18].
For all these materials, the surface band gap should be
distinguished from the bulk band gap: for instance, at the
MgO(001) surface, the band gap is reduced from the bulk value
of 7.8 eV to 6.2 eV, due to an unoccupied surface state [19]. An
extreme case is that of the metal-rich terminationof GaN(0001)
and AlN(00001) faces, which is metallic [18].
Some of these materials have also been studied in the
form of ultra-thin insulating layers adsorbed on a conductive
substrate, in particular by STM. (For a recent review, see [20].)
Useful information can be extracted from these studies, but
caution should be exercised because they cannot always be
extrapolated to the thicker films that are needed for building
the interconnection described here to limit the leakage current
between the interconnection pads.
3.1.2. Atomic scale imaging: NC-AFM. After the surface
preparation, STM atomic scale surface imaging is practical
only for SiC, GaN, and the metal-rich faces of AlN, among
the compounds listed in table 1. Fortunately, recent progress
in AFM imaging with the so-called non-contact (NC) or
frequency modulation (FM) mode [21,22] has opened the way
to atomic and molecular resolution on large gap and insulating
surfaces. Notice that this NC-AFM can be efficiently combined
with optical microscopy if suitable markers in the micron range
are fabricated on the substrate, as discussed at this end of this
section.
In NC-AFM, the cantilever is embedded in a positive
feedback loop that oscillates at the cantilever resonance
frequency while another loop maintains its oscillation
amplitude at a pre-set value. In contrast to the amplitude
modulation (AM) or ‘tapping’ [23] method, where the
frequency is externally fixed, the resonance frequency of the
cantilever in the FM method varies under the influence of
the tip–sample forces. A ‘topographic’ image is obtained by
scanning the surface while adjusting the tip–substrate distance
required to maintain this frequency shift at a pre-set value. The
Figure 3. Atomically resolved NC-AFM image of KBr(001) surfaces
showing a monoatomic step and atomic defects.
specificity of this method is that very high quality factors force
sensors, which lead to increased force sensitivity, can be used
without penalizing the acquisition time, as in AM-AFM, as
demonstrated in the seminal paper on FM-AFM by Albrecht
et al [24].
True atomic resolution was first obtained in NC-AFM
by Giessibl [25] and Kitamura and Iwatsuki [26] in 1995.
It has now been achieved on a wide variety of surfaces
(metals, covalent or ionic semiconductors, covalent or ionic
insulators) and the technique, after being used in UHV for
a long time, is now adapted to ambient conditions and to
liquids, especially for applications in biology [27]. Referring
to table 1, true atomic resolution has been obtained on
KBr(001) (figure 3)[28] and many other alkali halides [29],
MgO(001) [30], Al2O3(0001)[31], CaF2(111)[32] but not yet
on SiC, GaN and AlN.
Even less is reported on the imaging of molecules on
the surface of the materials of table 1by NC-AFM. Most of
the reported work was performed on alkali halide surfaces,
such as KBr(001), and at room temperature [33]. In these
conditions, for low coverage, most molecules diffuse, due
to their weak interaction with these surfaces, making the
observation of single isolated molecules impossible, except
when adsorbed on a defect, for instance a step edge. Using
structured surfaces [34] allows a more efficient trapping
of the molecules [35]. For higher coverage, monolayer
islands followed by multilayer 3D crystallites are often
observed (PTCDA on NaCl(001) [36] and on KCl(001) [37]).
Molecular lines are also formed in certain cases (DiMe-
PTCDI/KBr(001) [38]).
It is clear that NC-AFM will boost this rapidly developing
new domain of surface science. With the advent of low
temperature NC-AFM heads, these problems of mobility of
molecules on insulating surfaces will become less limiting.
Nevertheless, a large effort is needed to synthesize molecules
adapted to room temperature experiments in the longer-term
vision of devices that should ideally work at this temperature.
3.1.3. Atomic scale surface interconnection circuit. The next
step in the interconnection strategy presented in figure 2(b)
4

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 4. The 2 step self-assembly process for the fabrication of a
long metallo-organic molecular wire stabilizing metal atomic wires
at the surface of an insulator. (A) 2 types (gray and black) of Lander
molecule self-assembled at the surface of the insulator in a very long
molecular chain. (B) The metallo-organic molecular wire is
assembled by surface diffusion during (or after) the molecule Lander
building block sublimation (adapted from The Pico-Inside European
project final report. www.phantom.com/pico-inside).
is to build conductive atomic wires on the surface of the
wide gap material to connect the atomic scale circuit to the
metallic nanopads, as presented in figure 2(b). Therefore,
it is not enough to use, for example, surface step edges to
self-stabilize metal atomic wires following a peculiar surface
crystalline direction. These atomic wires have to converge
toward a specific surface location whose lateral dimension
will be around 10 nm and this is generally not compatible
with surface crystallographic directions. Two strategies have
been envisioned to construct these atomic wires: single
atomic manipulation or a self-assembly process via molecular
molding.
Theoretical works suggest that it should be possible to
manipulate single atoms [39,40] and molecules [41]on
insulating surfaces. The simulated manipulation process
generally requires a very precise positioning of the tip
that is becoming achievable, at least with the recent
low temperature NC-AFM. The first well-controlled atom
manipulation experiments using NC-AFM were reported very
recently on semiconducting surfaces in the vertical [42]and
in the lateral mode [43]. Poorly controlled manipulation of
an unknown species was also reported on CaF2(111)[44].
It is thus likely that controlled manipulations of single metal
atoms and molecules on the surface of an insulating material
will become possible in the near future. Notice that it is not
sufficient to build a conductive line. It must be stable as
such. It turns out that metallic cohesion does not help much
in this regard since there is generally a mismatch between
the inter-atomic distance that 2 metallic atoms must adopt
on an insulating surface for the complete atomic wire to be
conductive and the distance for 2 metallic atoms to remain
stable at their adsorption site selected by atom manipulation.
On NaCl(001), the distance between two gold atoms in a
chain should not be less than 0.7 nm for the chain to maintain
its stability [40]. Such a distance is too large for the
conductance of this atomic wire to reach at least one quantum
of conductance [45].
To circumvent this problem, one solution is to create
atomic scale defects at the surface of the wide gap material by
atom extraction. This is a not so well explored area and only
e-beam destructive pits on alkali halides have been made to test
how metal atoms can be successfully stabilized at the edge of
those pits [46]. The other solution is to play with the recently
discovered ability of certain well designed organic molecules
to mold metallic atoms [47] and even to surface extrude a single
well shaped atomic wire from a step edge [48]. This molecular
property was discovered on plain metal surfaces using LT-
UHV-STM single molecule manipulation. It is now explored
in a more systematic way by designing Lander molecules that
can self-assemble in long molecular chains, first on a metal
surface [49],andthenonaninsulatingsurface[50]. It is
expected that the co-evaporation of metal atoms with these
Lander molecules on an insulating surface self-assemble into
long metallo-organic molecular wires with a good conductance
(figure 4). After such a self-assembly, it remains to manipulate
these molecular wires one at a time to position them with
respect to the interconnection architecture. It may also turn out
that surface chemistry, where monomers of low gap molecular
wires are sublimated on a surface, can help in a surface
reaction, where under an increase of the surface temperature,
a fully covalently bonded molecular wire is synthesized at
the surface of an insulator, ready to be AFM manipulated in
the suitable surface interconnection configuration [44]. These
surface chemistry processes have been already experimented
with on metal surfaces [51] under the STM and are ready to
be transferred under a NC-AFM for wide gap surfaces. In this
case, the dynamic nanostencil technique (see section 3.3) will
help in delimiting the part of the surface on which the chemical
reaction will occur.
3.2. Metallic nanopad fabrication and re-configuration (level
B)
The metallic nanopads (B) of the figure 2(b) interconnects
have 2 functions: first, they serve to bridge the atomic scale
circuit (A) to the next mesoscopic level of interconnection (C).
Second, they provide the metallic density of states necessary in
a decoherence like process to superpose the billions of through
bond electron transfers per second that are occurring between
a couple of voltage bias nanopads (B) in the atomic scale
circuit (A). This superposition results in the tunneling current
intensity which can be measured with a standard ammeter [52].
The pad lateral dimension must be compatible with the actual
precision of the dynamic nanostencil technique (described in
section 3.3), which is of the order of a few tens of nanometers.
These nanopads should also be not too high for the AFM tip to
be able to image the surface where the central atomic scale
circuit is located and should be crystalline to present well-
defined facets without which imaging a molecule becomes very
difficult.
The first classification of growth modes of a deposit on
a surface was proposed by Bauer [53], who distinguished 3
basic modes: (i) the Frank–van der Merwe two-dimensional
growth, where atomic layers are formed one after each
other, (ii) the island Volmer–Weber growth, where three-
dimensional clusters are formed and (iii) the Stranski–
Krastanov mode, where two-dimensional growth is followed
by cluster formation. A fourth mode was introduced later:
the reactive mode, where an interphase is formed between
5

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 5. (a) AFM image of an alumina surface where 4 ML of copper have been deposited at room temperature. (b) Profile corresponding to
the line plotted on the image.
Figure 6. Mg metallic nano-island growth on the GaN(0001) surface. Left: the RHEED oscillations at different substrate temperatures during
the growth. Right: UHV-STM image of a 0.6 ML deposit showing Mg monolayer islands with very small inter-island distances. Adapted
from [55].
the substrate and the deposit as a consequence of a chemical
reaction or inter-diffusion at the interface.
It turns out that, in most cases, metals grow on the
materials listed in table 1in the Volmer–Weber mode, that is
in the form of low aspect ratio clusters. This is particularly
the case for alkali halides and most oxide surfaces, while much
less is known on the other materials. An example is shown in
figure 5, where 4 monolayers (ML) of copper were deposited
at room temperature on a Al2O3(0001)surface and imaged
with AFM: compact islands with sizes ranging from 2 to 6 nm
are observed. This general behavior can be traced back to
the fact that the surface free energy of insulating materials is
quite generally much lower than that of metals. Of course,
many other parameters influence the growth mode, such as
the lattice mismatch between the two materials or the nature
of the interface (reactive or not) as well as the experimental
conditions during the growth, which can drive the system far
from thermodynamic equilibrium. We refer for a detailed
description to the book of Noguera [54].
From a practical point of view, this behavior constitutes
a serious problem to make suitable nanopads. Fortunately, it
was recently demonstrated that Mg grows on the N-rich face
of GaN(0001) in the layer-by-layer mode [55], making this
system promising for our purpose. Figure 6shows the RHEED
oscillations during the growth at different temperatures and a
STM picture showing monolayer (height ∼0.3 nm) epitaxial
islands. This behavior is favored by the small lattice parameter
mismatch of 0.3% between the substrate and the deposit. Note
that STM imaging can be practiced on GaN, due to its relatively
small band gap (table 1). In contrast, studies to find a metal
that would grow in the layer-by-layer mode on AlN, which has
a larger gap (table 1), require the use of NC-AFM.
Aside from continuing the exploration of the growth
conditions of metal nano-islands on large gap semiconducting
and insulating surfaces, for example by playing with kinetic
effects to better adapt the metal nano-island shape, it is also
important to consider how to arrange them on the surface. In
figures 5and 6, they are randomly distributed over the surface,
which may be very good to find 2 nanopad contacts as indicated
in figure 6with sometimes very small inter-nanopad distances
below 1 nm. Such small gaps are impossible to fabricate
with conventional e-beam nanolithography techniques. For
a larger number Nof interconnects, the nanopads will have
to be arranged in a specific order and orientation on the
insulating surface. At present, the only way to construct such
an interconnection structures, with for example Nnanopads
converging in a circle toward the same surface spot, is to
manipulate them one-by-one using the tip of a NC-AFM. This
puts forward the need to study manipulation by NC-AFM, not
only of single atoms or molecules, but also of larger nano-
objects made of millions of identical atoms. This had already
been demonstrated on a small gap semiconductor surface with
6

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 7. (a) Schematic view of the geometrical configuration for the
Toulouse AFM nanostencil machine. (b) Scanning electron
microscopy (SEM) image showing a recessed box in the rear face of
an AFM tip, which was first thinned down to 80 nm by FIB. Adapted
from [59].
STM (see section 4) and is now required for larger electronic
gap surfaces.
3.3. The interconnections from the nano- to macroworld
(levels C2, D and E).
According to the interconnection scheme illustrated in
figure 2(b), and after having fabricated well faceted metallic
nano-islands (nanopads), the next steps are the realization of
the three interconnection levels C2, D and E with increasing
width from the nanometer to the macroscopic scale. To
realize these electrodes on the surface, it is not possible to use
either a resist based nanolithography technique or a stamping
technique because the atomic cleanness of the working area
where the atomic scale circuitry A and the nanopads B
should be built should be preserved. The only remaining
possibility is to use different stencil techniques under UHV to
grow the nanoscale wires (C2) and the microelectrodes (D),
and then to use metallic microcantilevers (E) to contact the
microelectrodes. The order in which these different operations
have to be performed does not follow the order in which
they were presented in section 2, but depends on technical
constraints related to the necessity of successive alignments of
the different structures. The operations performed to realize the
interconnection levels C2, D and E, described in this section
are combined with levels A and B described in sections 3.1
and 3.2 according to the following sequence:
(1) Surface preparation, for instance growth of AlN.
(2) Growth and possibly manipulation of the metallic
nanopads (level B).
(3) Growth of the microelectrodes by static stenciling (level
D).
(4) Nanostencil deposition of the nanowires between the
metallic nanopads and the end of the microelectrodes
(level C2).
(5) Building of the atomic scale circuitry by NC-AFM (level
A).
(6) Electrical measurement of the molecule properties with
the help of metallic microcantilevers (level E).
3.3.1. The stencil techniques. Two main stencil techniques
exist: the static technique, in which the mask is placed in
contact with the substrate and the pattern is directly evaporated
on its surface, and the dynamic technique [56–60], in which
the pattern is drawn on the substrate by moving the mask. This
last mode is generally implemented on an AFM microscope.
Indeed an AFM cantilever can be patterned using the focused
ion beam (FIB) technique and used as the shadow mask.
The principle of the nanostencil technique based on an AFM
cantilever is presented in figure 7.
Figure 7(a) shows the geometrical configuration of the
dynamic stencil on the Toulouse machine. Si3N4cantilevers,
with hollow tips, are used. The stencil pattern is made in the
rear side of the tip, using a focused gallium beam. It should
be positioned as close as possible to the substrate in order to
minimize the geometrical enlargement of the deposited pattern,
which varies as the ratio between the pattern–surface distance
and the evaporation source–surface distance. To reach small
mask pattern sizes, it is also necessary to thin down an area
of the tip side from about 800 to 80 nm (the box displayed in
figure 7(b)) before drilling the patterns in this membrane.
The high positioning precision of the AFM combined
to the nanometric size of the apertures made in the AFM
cantilever allows one to obtain nanowires. The dynamic
nanostencil is a UHV in situ technique with the additional
advantage of allowing the imaging of the surface at the
different stages of the fabrication process and therefore directly
aligning the mask relative to the nanopads.
As demonstrated in different works, one of the difficulties
of the stencil techniques is that the apertures tend to clog
rapidly [60–62]. This well-known phenomenon is the main
disadvantage of the nanostencil process. One solution is to
use self-assembled monolayers (SAM) to passivate the surface
of the cantilever. It has been demonstrated that alkyl or
perfluoroalkyl terminated SAMs delay the clogging of the
stencil by reducing the adhesion of the deposited metal inside
the nanostencil apertures [61,63]. The use of wet etching
to eliminate the clogging of stencils after gold or aluminum
evaporation has been reported as well [64,65].
3.3.2. Growth of the microelectrodes by the static stencil
technique. The microelectrodes are deposited in UHV by
the static stencil technique. Specific microelectrodes were
designed in such a way that they are protected by the shadow
of the AFM cantilever used during the next steps to grow the
nanowires.
Figure 8(a) presents the microelectrodes stencil mask,
observed by optical microscopy. One microelectrodes device
constitutes the bridge between level C2 (1 μm) and E
(10 μm): the pads used for the connection with the metallic
microcantilevers are 6 μm in width separated by 6 μm. So the
total device is 62 μm×20 μm. It is easily seen with an optical
microscope, but also with the wide scan range AFM available
with the Microclean Room developed in Toulouse, described
in section 3.3.5.A75
μm×75 μm NC-AFM image of one
device, made by depositing copper on SiO2[60] is displayed
in figure 8(b).
7

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 8. (a) Optical microscope image of the microelectrodes aperture made in a static stencil membrane (b) 75 μm×75 μmAFMimagein
the non-contact mode showing one device [60].
3.3.3. Growth of the nanoelectrodes by the dynamic
nanostencil technique. The nanopatterning of the inter-
microelectrodes area by the nanostencil technique, described
previously, relies on the precise determination of the actual
location of the deposited nanopattern with respect to the
nanostencil AFM tip. Consequently, a preliminary calibration
step is needed for each new nanostencil AFM cantilever. In a
typical calibration experiment, the cantilever is approached to
a bare substrate in either non-contact or contact AFM mode. A
deposit is then made using the dedicated effusion cell. Large
scale AFM images are then obtained by scanning the X–Y
table and the relative position of the deposited pattern and the
AFM tip can be precisely measured. Subsequently, the bare
substrate is exchanged with the substrate with the deposited
microelectrodes. Hence, knowing the location of one set of
microelectrodes and the relative position of the deposit with
respect to the AFM tip apex, the nanopositioning table is
moved so as to deposit the nanopatterns precisely between
the microelectrodes. Figure 9shows an NC-AFM image of
5 nm thick copper nanoelectrodes deposited between the 4
microelectrodes.
This example shows the difficulties met for precise
alignment: the nanoelectrodes pattern is correctly aligned in
the ydirection but has a misaligning of around 100 nm in
xdirection and a defect angle of 10◦. The error in the
xand ydirection is related to the mechanical drift of the
piezoelectric tube during the experiment. The angle error
comes from addition of error angles during the FIB drilling and
the gluing of the cantilever on its holder. We are confident that
these errors can be minimized and plan to improve the overall
accuracy to 20 nm, which is the limit imposed by our wide
range (XY) table (see section 3.3.5).
3.3.4. Electric measurements with the microcantilevers.
The micropads at the end of the microelectrodes (figure 8)
should be connected to the macroscopic world. Metallic
microcantilevers mounted on a specific UHV compatible
printed circuit board (PCB) connected to the outside of the
UHV chamber are used for that purpose. They should have
good mechanical and electrical properties in order to establish
low resistance electrical contacts with the micropads. The
Figure 9. Non-contact AFM image (4 μm×4μm) of Cu
nanoelectrodes deposited between the Pd microelectrodes made by a
static stencil on a Si substrate; (inset) SEM image of the FIB drilled
pattern distorted to correct the geometrical distortions due to the
evaporation geometry. Adapted from [59].
first models of these metallic microcantilevers used in air
were made entirely with gold and have shown bad mechanical
properties: they are easily plastically deformed. In more recent
designs, 1 μm thick SiC layers were used to support metallic
wires. To avoid contact problems, it was also necessary to
induce a curvature of several μm of the SiC cantilevers [66].
Furthermore, the size should be correctly chosen for the
microcantilevers to be easily seen with the optical microscope.
In the particular case of the Microclean Room, the size of the
microcombs is 6 μm in width for SiC, 4 μm in width for each
metallic wire, 100 μm in length and 24 μm in periodicity.
Figure 10 shows a microcomb mounted on its dedicated PCB.
3.3.5. The Toulouse setup. To realize under UHV
the 5 levels of interconnect described in figure 2(b), the
deposition of molecules, their observation by NC-AFM and
the measurement of their electrical properties, the GNS group
has designed and realized a dedicated UHV equipment called
DUF (DiNaMo UHV Factory). This equipment allows the
transfer of samples under UHV between five complementary
UHV chambers:
8

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 10. (left) SEM image of the end of a SiC based cantilever with the Au wire; (middle) SEM image of the microcombs made of a 10 SiC
based cantilever; (right) microcombs chip mounted on its specific PCB. The SiC based microcombs were made by the LPN (CNRS UPR20) at
Marcoussis during the DiNaMo project (French ANR founding No ANR-05-NANO-014).
Figure 11. The DUF (DiNaMo UHV Factory) equipment allows the transfer of samples between 5 complementary UHV chambers in order to
realize the 5 levels of interconnect on wide band gap semiconductors (GaN, AlN). The different chambers are described in the text.
(1) MBE growth dedicated to nitride semiconductor growth,
metallic nanopad growth and stencil evaporation for
microelectrodes.
(2) Mass spectrometer transformed in a molecular ion source.
(3) Preparation chamber for clean substrates, STM tips and
AFM cantilevers.
(4) Room temperature AFM/STM for surface characterization
by STM and NC-AFM.
(5) The microclean room: AFM/STM modified for nanosten-
ciling experiments and electrical measurements.
Figure 11 shows this equipment in its actual state.
To realize the steps 4 and 6 described in the previous
section, a UHV Omicron Nanotechnology VT STM/AFM
head has been modified to accommodate different tools,
namely [60]:
(1) A flexural-hinge-guided (XY) piezoceramic driven nano-
positioner stage (100 μm×100 μm, repeatability 5 nm)
with a closed loop control based on capacitive sensors,
(2) An evaporation system highly collimated on the cantilever
to perform nanostencil deposition,
(3) A (XYZ) piezo-table for positioning the metallic
microcombs,
(4) An optical microscope to control the positioning of the
microcombs.
These modifications were designed by the mechanical
department of the laboratory. Figure 12 shows a scheme (a)
and a picture (b) of the modified AFM.
The main advantage of using a commercial UHV
AFM/STM is to benefit from its good characteristics for SPM
imaging. But the piezoelectric tube used to scan has a range of
afewμm only. The addition of an XY table to move the sample
offers the possibility to perform wide range scanning, up to
80 μm SPM images, while keeping the possibility to realize
atomic scale imaging with the piezoelectric tube.
One of the disadvantages is the small accessible space
around the SPM head. Indeed, it is not possible to place an
optical microscope with normal incidence with respect to the
substrate, and an atomic source for the nanostencil experiments
with normal incidence with respect to the AFM cantilever.
In our case, the image obtained by the optical microscope
is reflected by a mirror that makes an angle of 30◦with the
substrate plane. This gives distorted images, with a loss of
resolution: only 3 μm instead of 1 μm in normal incidence.
The effusion cell is fixed on a port of the UHV chamber at
an angle of 33◦with the horizontal plane, and another at an
angle of 28◦between the two vertical planes passing through
the evaporation beam and the central axis of the cantilever
(figure 7(a)). This orientation of the atom beam induces
distortion, which should be taken into account in the design
9

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 12. (a) Schematic view of the functionalities implemented inside the microclean room and (b) the modified AFM/STM head.
of the nanopattern to be drilled into the pyramidal tip of the
cantilever (figure 9)[60].
4. Small VB–CB electronic gap surfaces
As presented in figure 2(c), a small electronic gap (below a
few eV) of the surface material supporting the interconnection
structure leads to only 3 levels of interconnects. As in
section 3, levels (A) and (B) depend on the mastering of atomic
scale surface science phenomena while level (C1) relies on the
ability to fabricate multiple ultra-sharp metallic tips to contact
each metallic nano-island of level (B). The constraint with a
multi-access interconnection technology on a small electronic
gap surface material is to avoid surface leakage currents
between the electrodes contacting the surface metallic nano-
islands. Therefore, it is compulsory to contact these nano-
islands from the top. This constraint has a lot of consequences
for the structure of the interconnection machine depicted in
figure 2(c). In particular, a scanning electron microscope is
required instead of the figure 2(b) optical microscope because
the lateral size of the nano-islands is necessarily very small
to avoid inter-nano-island leakage currents. In this section,
progress and problems that remain to be solved for these 3
levels of interconnections are presented up to the description
of the interconnection machine in construction at IMRE in
Singapore, which is devoted to these small gap supporting
materials.
4.1. Semiconductor surface preparation
The preparation of a semiconductor surface usually begins
by degassing and destroying the oxide layer. The surface is
then annealed by passing through the sample a current whose
intensity depends on the doping level of the bulk material [68].
These preparation conditions have to be carefully optimized
to get the necessary large atomically flat terraces, in the
50 nm width range and with a homogeneous reconstruction
all along a terrace. A problem is that such surfaces are
usually very reactive, due to the presence of dangling bonds.
Even under very strict UHV conditions, they get contaminated
on the timescale needed to build the atomic scale circuitry
and its connections, which is at least of the order of a few
hours. Saturating these surfaces by adsorbing hydrogen,
chlorine, sulfur or small molecules is an efficient solution
to preserve their integrity on this timescale [69]. In many
cases, this modification has the additional benefit of increasing
the electronic surface band gap. For example the native
Si(100) surface gap of 1.2 eV increases to 2.1 eV with an
hydrogen saturation in its 2 ×1 reconstruction phase [67]. It
was demonstrated recently [75] that this saturation layer can
also decouple electronically a conjugated molecule from the
Si(100) surface similarly to what was observed for an ultra-
thin ionic insulator on a metal surface [76]. This electronic
decoupling is very important because, on such a surface, the
only way to interconnect a molecule with an atomic precision
is to fabricate surface atomic wires (see section 4.2 hereafter)
and to STM manipulate the molecule between them, restoring
the coupling of the molecule to both ends of these wires as
proposed in figure 2for the atomic circuit A. A limitation
of this approach for the next processing steps is that these
saturation layers have a limited thermal stability (for instance
300–400 ◦C for Si(100) 2 ×1-H [70]).
There are a few semiconductor surfaces where this
saturation can be considered as part of the material structure.
This is the case of lamellar semiconductors like MoS2.In
this case, the sulfur overlayer is hardly bonded to the Mo
atoms underneath. The MoS2(0001)surface is rather easy
to prepare by a fast UHV cleaning followed by a standard
surface preparation temperature in the 300 ◦C range [71]. The
extension of the atomically flat surface of lamellar compounds
is usually rather large, up to a few microns in lateral extension
in some cases. MoS2is one of the most interesting materials
of the lamellar compound series with a bulk gap around 1.3 eV
and a surface gap in the 1.0 eV range [72]. The MoS2surface
remains atomically perfect and flat up to 1100 K. Above
1200 K, a reconstruction of the top layers of MoS2to Mo2S3
is observed, leading to one-dimensional atomic double rows
on the reconstructed surface [73]. Such temperature stability
is very important for the fabrication of surface atomic wires by
STM extracting rows of sulfur atoms. The drawback is that it is
10

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 13. STM images of a Si(100) 2 ×1 and a MoS2(0001)surface showing atomic resolution before the start of any construction or
fabrication steps. As described in this section (and in section 3for large gap surface materials), these construction and fabrication procedures
must be carefully tested in such a way that atomic resolution is kept up to the multi-probe I–Vmeasurements step. (a) The Si(100) surface
was prepared with the objective of increasing the terrace size before the hydrogenation step. (b) The MoS2surface was prepared to determine
the experimental conditions to extract the sulfur surface atoms one-by-one.
extremely difficult to extract these surface atoms as compared
to the surface atoms of the Si(100) 2×1-H surface [74]. There
is a crucial need for a material stable enough in temperature
and whose surface atom chemical stability is between Si(100)-
HandMoS
2.
STM images of the 2 principal surfaces used at IMRE
in Singapore to explore atomic scale interconnection are
presented in figure 13, each showing a clear atomic resolution.
One imaging problem comes from the softness of lamellar
semiconductor materials in the direction perpendicular to the
surface. The sample is then easily deformed by the force that
the tip exerts on the surface during imaging. In this situation,
adsorbates or metallic nano-islands can be easily brushed away
by the tip because the tip apex to surface distance can be very
small during STM imaging [71].
4.2. Atomic scale surface interconnection circuit
At the passivated surface of a semiconductor, the STM
fabrication of atomic wires is facilitated because the created or
the native dangling bond atom saturation can be manipulated
atom by atom to create a line of surface dangling bonds [14].
It was theoretically predicted that a perfect row of surface
dangling bonds introduces at least one conduction band in the
electronic surface gap induced by the passivation layer [77].
The effective conductance of this dangling bonds electronic
path will certainly depend on the doping of the substrate and
around the surface Fermi level, its dispersion is generally much
less that 1 eV, making the corresponding carrier effective mass
quite large. The detailed molecular orbital composition of
such dangling bonds lines is generally complex because it
involves the liberated atomic orbital per surface atom plus the
contribution of the nearby atoms whose orbital hybridization
may have been changed too [72]. Even with a conduction
band width smaller than 1 eV, a ballistic channel of conduction
requires the fabrication of a long line of dangling bonds. If the
line is too short between the molecule to be contacted and the
metallic pad, it remains a quantum box with its characteristic
discrete electronic resonances, on average one or two per
extracted atom [72,77]. Mastering the fabrication of long
lines of dangling bonds is also very important for the inter-
metallic nanopad (B) distances to be large enough to minimize
the surface tunneling leakage current intensity between them.
Surface atoms saturating the dangling bonds of a
semiconductor surface have a bonding energy around a few
eV [14]. Therefore, a voltage pulse of a few volts applied
between the tip and the surface is able to destabilize a given
surface atom [14,74]. This is a vertical manipulation process
since, in many cases, the extracted surface atom is desorbed
from the surface and reaches the apex and, eventually, by
subsequent tip surface diffusion, the body of the tip. It depends
on the surface material but also whether or not an electronic
resonant state of the surface is accessible for the selected tip–
surface bias voltage. In many cases, the main desorption
process requires the resonant excitation of the anti-bonding
state of the surface atom [14]. Another mechanism, active for
large tunneling currents, involves the vibrational excitation of
the atom surface bond by inelastic tunneling.
Using this vertical manipulation technique, writing lines
of surface dangling bonds at the atomic scale has been
practiced by a few groups up to the point where atomic scale
circuits could be constructed atom-by-atom [14]. The first
single atom desorption experiments were reported on a non-
saturated Si(111) surface [78], then on a hydrogen passivated
Si(100) surface [79] and then on a MoS2surface [74]. What
is not yet known is the number of atoms the same tip apex
can extract before being chemically transformed and becoming
inactive for the vertical manipulation process. The drawback
of this vertical manipulation technique is that some extracted
atoms already adsorbed on the tip apex can be transferred again
and re-saturate some dangling bonds of the already fabricated
line. Here, there is a clear need for a better understanding
11

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 14. SEM image of the distribution of Au nano-islands on
MoS2(0001)after the fabrication process detailed in section 4.3.
Many islands are 30 nm in lateral size. Notice also the very sharp
summits of these triangles.
of these extraction and re-saturation processes together with
a good understanding of the lateral surface diffusion processes
occurring on the surface while a newly created dangling bond
is re-saturated by a nearby saturating atom jumping over.
4.3. Metallic nanopads fabrication and re-configuration
The metallic nanopads (B) in figure 2(c) are necessary to pass
from the surface atomic wires made of dangling bond lines to
the metallic tip apex (C1). This very local metallization must
be practiced directly on the passivated semiconductor surface.
It turns out that there are not a lot of passivated surfaces
whose passivation overlayer can sustain the temperature
required for metallization [80]. These temperatures, which
are usually higher that a few hundred degrees, locally destroy
the passivation layer. One solution to this problem is to
start with a robust semiconductor surface like MoS2and then
to transfer the metallic nano-islands onto another surface by
contact printing in UHV [81].
On the MoS2(0001)surface, gold atoms form triangular
nano-islands whose lateral dimension can go down to 5 nm
for a thickness of a few nm [71]. An optimum of 30 nm
lateral and 10 nm in height can be easily obtained by tuning
the surface temperature during the metal evaporation [71]. In
a standard preparation, MoS2wafers are fabricated from bulk
molybdenite stones directly mined in Australia. The slides are
cut into 10 ×10 mm2pieces. The top surface of a MoS2slide
is cleaved and immediately loaded into a thermal evaporator.
After the sample is outgassed at 400 ◦Cfor3h,Auisthermally
evaporated onto the MoS2surface at 400 ◦C with a deposition
rate of 0.02 nm s−1. The total thickness of evaporated Au,
monitored by a quartz microbalance, can be kept as low as
1 nm. After Au deposition, the sample is maintained at 400 ◦C
for another 1 h to facilitate the formation of the nano-islands.
During this process, the pressure in the chamber must be
below 2 ×10−5Pa. The nano-island growth conditions can
be optimized to achieve a majority of nano-islands being of
equilateral triangular shape with a 20–30 nm lateral size (see
figure 14).
Such a fabrication and self-assembly process usually leads
to a rather homogeneous distribution of Au islands. In some
cases, the apices of 2 islands can be face-to-face at distances
as low as a few nanometers, constituting an ultra-clean and
atomically well-ordered nano-junction to interconnect a single
molecule since the background surface has generally kept
its atomic scale corrugation. Unfortunately, the MoS2(001)
electronic surface band gap is too small for the ‘island-surface-
island’ conductance to be much smaller than the one of a
conjugated molecule to be interconnected. Therefore, instead
of searching on the surface for a multi-pad configuration
adapted to such conductance measurements with the good
inter-pad distance, we have learnt the conditions to manipulate
these nano-islands one-by-one on the MoS2surface [71].
There are 2 modes of manipulation. In the soft one, the tip apex
does not mechanically touch the nano-island but is charging
it. A large number of scans is necessary to charge a nano-
island enough for the electrostatic force between the STM tip
apex and the nano-island to be large enough for the nano-island
to move away, but with a precision better than 0.1 nm [71].
In the mechanical mode, the tip apex is simply pushing on a
facet of the nano-island. This mode of manipulation is less
precise (1 nm or more) but very fast [71]. Figure 15 presents
a few examples of Au nanopad nanostructures constructed by
manipulating the nano-islands one after the other.
Starting from a flat MoS2surface, these nanostructures
cannot be transferred as such by a direct contact printing
between this surface and a larger gap semiconductor surface.
Firstly, the construction will be deformed during the printing.
Secondly, for interconnecting a surface dangling bond
interconnecting circuit, it is better to manipulate the nano-
islands after the fabrication of the surface atomic wires. In
this way, a given metallic nano-island can be manipulated
step-by-step over the corresponding atomic wire up to the
point where the conductance is larger than the surface leakage
conductance. We have recently demonstrated that Au nano-
islands can be transferred to an Si(100)-H surface after a
specific MoS2surface microscale structuration [81]. The MoS2
surface must be structured in micronscale pillars creating a
matrix of stamps. We have shown that Au nano-islands can
be fabricated on the surface of each pillar after changing the
characteristic fabrication temperature. A 10% overall transfer
rate to Si(100)-H was obtained for 10–20 μm lateral size
pillars. The experimental conditions to manipulate these nano-
islands on the Si(100) 2 ×1-H surface, as it was done on the
MoS2surface, remain to be determined. Notice also that the
hydrostatic pressure during transfer printing mustbe optimized
to preserve the hydrogen passivation layer under the nano-
islands.
4.4. Tip apex fabrication
Fabrication of atomically clean nano-probes in a highly
controlled manner with diameters in the range of 2–5 nm is
crucial for contacting metallic nano-islands positioned in close
proximity on the surface of a semiconductor, as presented in
figure 15. Furthermore, these ultra-small tips must have a very
high aspect ratio to be able to increase the number of tips
12

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 15. Example of contacting nanostructures constructed by manipulating the nano-islands one after the other in a lateral STM
manipulation mode on the MoS2(0001)surface. (a) More than 50 nano-islandshave been manipulated around to assemble the central 4
contacting nanostructure. (b) Zoom on a construction of 4 nano-islands manipulated in a square. (c) Zoom on a construction of a 6
nano-islands interconnection structure converging toward a 12 nm central active area where the atomic circuit is supposed to have been
atomically fabricated before this construction. (a)–(c) are UHV-STM images with V=0.2VandI=10 pA.
Figure 16. Two optical microscope images with different magnifications (a) and (b) and a SM image (c) of a very sharp tungsten tip
electrochemically prepared before being FIM imaged. The presented FIM image (d) shows circular arrangements of W atoms corresponding
to specific crystallographic orientations at the end of the tip apex.
that should converge towards the same nanoscale area of the
surface, as for example the 6 nanopads configuration presented
in figure 15. To ensure a rigorous cleanliness of these tips
and to precisely determine the tip size and shape, a field ion
microscope (FIM) is used to shape atom-by-atom the tip apex
using the spatially controlled electric field assisted technique
and at the same time to be able to image with an atomic scale
precision the fabricated tip apex.
Invented by Muller in 1956 [82], the FIM is based on
placing, in a UHV chamber, a sharp tip (generally in tungsten),
with a typical radius of curvature of 100 nm, in front of a
phosphor screen. Such 100 nm tip apices are usually prepared
by AC electrochemical etching as illustrated in figure 16.
When a high voltage, in the range of 5–20 kV is applied, a high
electric field in the order of 5 V ˚
A−1is generated at the tip apex.
This field is adequate to ionize an inert atom gas (for instance
He) introduced in the FIM chamber. Once a He atom is ionized
at the tip apex, it is expelled towards the screen where it is
detected as a bright spot. As these events take place all over the
tip apex surface, a stereographic projection of the tip surface
atomic structure appears on the screen. The optimum image
contrast is attained when the tip is cooled to liquid nitrogen
temperature, since the ionization probability and atom surface
atom stability are improved, as presented in figure 16(d).
The tip radius of a standard clean tip is calculated by iden-
tifying two atomic poles and counting the number of atomic
rings between the centers of consecutive planes. For the tung-
sten tip of figure 16, this gives a 12.8 nm tip radius of curvature.
13

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 17. Schematic illustration of the electric field assisted
technique where N2is adsorbed on the tip body, decomposed, and
during the decomposition expels metal atoms from the tip apex [85].
To contact a 30 nm nanopad (figure 15), the tip apex radius
needs to be reduced below 5 nm. Under an SEM microscope,
this will ease the alignment of each tip apex with a given nano-
island before contacting it. Such a reduction of the tip radius
of curvature can be achieved by using the surface electric
field assisted technique [84]. This technique depends on the
introduction of a small amount of N2gas into the FIM chamber
just after the preparation of an atomically clean broad tip as
presented in figure 16. Notice that nitrogen has a smaller
ionization barrier (14.5 eV) than helium (24.5 eV). Therefore,
nitrogen atoms cannot reach the apex where the electric field
is maintained at the best imaging conditions for ionizing He
because they will be ionized before that. In these conditions,
N2prefers to settle down on the shank where the field is much
lower. On a W surface, it is well known that N2dissociates
and subsequently diffuses into the top atomic layer [85]. This
results in the creation of metal atoms protruding out of the
surface and therefore in a local enhancement of the tip surface
electric field [84]. This electric field is high enough to ionize
and then field evaporate these protruding metal atoms. This
etching process can be controlled by a careful adjustment of
the applied voltage in order to maintain the apex field at the
imaging threshold value. One reaches very small tip apex
radii of curvature down to a single atom tip apex [86]. This
surface diffusion completed by a field evaporation process is
schematically illustrated in figure 17. The sequence of FIM
images in figure 18 represents the tip apex evolution during
the sharpening process. The decrease in the apex area and
the increase in the spot size are direct indications of the tip
sharpening as the FIM magnification increases [86].
Notice that when the tip radius is reduced to 1 or 2 nm,
a few atomic layers (rings) appear on the FIM image. In this
regime, the concept of atomic planesor poles is no longer valid
and hence the conventional method of radius estimation cannot
be used. In this case, the most appropriate way to evaluate the
tip radius with good approximation is to build a bcc atomic
model that can reproduce an atomic feature similar to the FIM
image [83].
With this surface electric field assisted technique, it is
possible to fabricate an ultra-clean tip apex with radii of
curvature adapted to the lateral size of the nano-islands to
be contacted. The advantage of the associated FIM image is
that the tip apex atomic structure can be followed during the
field assisted etching process. This opens the way to fabricate
many identical tips having similar apex atomic structure and
crystallographic orientation. This is very important since
the tip apex work function depends on its crystallographic
orientation. It is essential that all the tips converging toward
a given nanostructure get the same contact conductance, that
is the same work function. Notice that this local work
function can be measured by changing the polarity of the
tip bias voltage, transforming the FIM into a field emission
microscope [86].
Figure 18. A sequence of FIM images following the step-by-step fabrication of a very sharp STM tip apex using the N2field assisted
technique. (a) The initial FIM image of the tip apex with its characteristics facets. (b)–(d) a progressive increase of the electric field to reach a
very sharp tip apex with only 2 W atoms at the tip apex end.
14

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 19. (a) STM image of marks A to C formed by sweeping out many Au nano-islands using the STM brushing mode, (b) SEM image of
the same A to C marks after transferring the MoS2wafer from the UHV-STM to the SEM and searching for the A, B and C marks with a
moderate SEM magnification, (c) SEM image of the nanostructure located at the mark A indicated by a black circle in (b). (d) The original
STM image of the four-metallic-pad nanostructure constructed by STM manipulating single nano-islands before the transfer of the wafer from
the UHV-STM to the SEM.
4.5. Navigation and the structure of the interconnection
machine
The STM constructed nanostructure now exists ‘somewhere’
on the semiconductor surface. To align it with the field
assisted fabricated ultra-sharp tips, described in section 4.4,
a navigation system is required. It must be based on the
determination of a fixed reference to pass from the nanoscale of
the contacting nanostructure (∼10 nm) to the mesoscopic scale
of the body of the tip (∼100 μm). Only a scanning electron
microscope (SEM) is able to access such a range of scales, as
indicated in figure 2(c).
Before transferring the sample from the UHV-STM to
the UHV-SEM chamber and to ease the navigation on the
wafer surface, the addition of appropriate marks is necessary
to re-locate the contacting nanostructure over the millimeter
wafer size now positioned under the UHV-SEM. This marking
is performed with the STM just after the construction of
the contacting nanostructure. These marks can be a simple
bar code formed by brushing out many nano-islands of the
semiconductor surface with a series of micronscale squares.
It was also proposed to crash the UHV-STM tip, after the
nano-island manipulation, near the constructed interconnection
nanostructure. Whatever the marking technique, these marks
must be easily imaged in a medium range magnification UHV-
SEM [71].
Figure 19(a) displays the STM images of 3 marks (A,
B and C) fabricated after the construction of a four metallic
nanopad nanostructure. The lateral size of the marks is close
to the maximum scanning ability of a standard UHV-STM but
can be imaged with low SEM magnification. After transferring
the sample to the UHV-SEM, these A, B and C patterns are first
located on the wafer surface by the UHV-SEM (figure 19(b)).
Navigating from the marks, the contacting nanostructure can
be easily located in absolute direction and distance. The UHV-
SEM image of the corresponding nanostructure is presented
in figure 19(c), where all nano-islands are found in the
same position as in the initial UHV-STM image presented in
figure 19(d), recorded just before the wafer transfer from the
STM to the SEM.
Following the figure 2(c) principle, the contacting
machine on a semiconductor surface is made of an assembly
of 3 UHV chambers. One chamber is dedicated to the wafer
preparation and metallic nano-island deposition and transfer
(when necessary). A second chamber is equipped with a LT-
UHV-STM for fabricating the atomic wires, manipulating the
molecule logic gate and positioning the metallic nano-islands,
one on each termination of an atomic wire. The third chamber
15

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
Figure 20. The final design of the IMRE (Singapore) interconnection
machine with its 3 main chambers: (A) the preparation chamber with
the FIM, (B) the LT-UHV-STM from Omicron Nanotechnology and
(C) the nano-probe system from Omicron Nanotechnology with its
UHV-SEM (3 nm resolution) and the 4 integrated STM. (D) is the LT
transfer from the LT-UHV-STM to the UHV-SEM.
is equipped with an UHV-SEM with a resolution better than
5 nm to be able to image the contacting metallic nano-islands
and to position a tip apex on each of these nano-islands. The
details of the IMRE Singapore interconnection machine are
presented in figure 20.
Under the UHV-SEM, the contact between a nano-island
and its tip apex is controlled by the feedback loop system
of an STM to avoid the local destruction of the sample.
Consequently, each tip has to be positioned on a STM
mechanical system driven by its own control electronics.
A SEM image of 4 nano-islands manipulated on purpose
inarowonaMoS
2surface is presented in figure 21
together where the 4 tip apices approaching for the contact.
Electrical measurements are now in progress to record the
‘nanopad—MoS2surface—nanopad’ I–Vcharacteristics for
inter-nanopad distances lower than 10 nm.
5. Discussion
As presented in sections 3and 4of this review, it is
now technically possible to construct atomic scale precision
interconnection machines dedicated to contact electronically
an atomic scale or a single molecule circuit as proposed in
figure 2. Depending on the electronic surface gap of the
supporting material, solutions have been found to navigate
on the surface and to locate the active nano-area respecting
the atomic precision of this surface and the atom-by-atom (or
molecule-per-molecule) construction. The solutions are based
on the use of two different types of microscopy at the same
time: a far and a near field one. For large gap surfaces,
an optical microscope is looking after a NC-AFM one. For
moderate gap surfaces, an SEM is looking after an STM
one. Contacting the atomic scale structure relies on the same
strategy for the two types of surfaces: bridging and positioning
metallic nano-islands on the surface. These nanopads are
contacted using a specific nanostencil technique on large gap
Figure 21. SEM image of the tip apex of 4 chemically etched
tungsten tips converging toward 4 Au nano-islands that have been
manipulated one-by-one to construct a 4 point line to perform surface
conductance measurements. This image has been recorded on a
multi-probe system from Zyvex.
surfaces or multiple STM tips landing from the top of the
surface on low gap surfaces. What remains to be tested is how
these nanopads interact with the surface atomic wires in charge
of contacting the central molecular machinery.
One drawback of these UHV interconnection machines
is that each step uses a specific UHV equipment. Therefore,
the sample must travel frequently in a large UHV system from
equipment to equipment. A future direction would be to reduce
this traveling distance by integrating a maximum number of
steps in the same UHV equipment. This improvement would
also significantly simplify the navigation problem.
The future of atomic scale interconnection machines will
face two very difficult technological problems: the packaging
and the unavoidable increase of the number Nof interconnects
to increase the molecule machinery complexity. The two
types of interconnection machines presented in this review are
laboratory equipment. When a molecule device is constructed,
interconnected and characterized in UHV, there is no way
yet of carrying it intact outside the UHV chamber and in
certain cases increasing the substrate temperature up to room
temperature. The packaging problems carried out by atomic
scale technologies is almost absent in the literature and only
one patent was recently issued on this problem [87].
The number of possible interconnects converging towards
a given nanoscale machinery is intimately related to packaging,
since, for example, it seems to be very difficult to greatly
multiply the number of ultra-sharp STM tips converging
toward a nanoscale area. At the surface of an atomically
mastered wafer, we do not yet know practically how many
metallic interconnects Ncan be constructed to converge
towards a nano-spot of diameter Das a function of the atomic
wire width δand for a given inter-atomic wire distance dat the
interconnection point. Of course, the simple geometric answer
to this question is:
N=πD(d+δ)−1.
Using wires with a section of a single metallic atom for the
interconnects, the minimum value for δis simply twice the van
16

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
der Waals radius of the corresponding metal atom, or simply
the effective lateral expansion of a surface dangling bond that
is δ>0.5 nm. If we suppose that very long atomic wires
can be fabricated with the methods discussed in sections 3
and 4, the limit on Nfor a given Dis imposed by the leakage
tunneling current that can be tolerated on the surface of the
wafer for the performance of the atomic or molecular circuits to
be measurable. Taking a standard large gap insulating surface,
a leakage current intensity lower than a fA leads to d>1nm
at the interconnection side between the molecule circuit and
the atomic wires for a bias voltage lower than a volt. For a
low gap semiconductor surface, the fA leakage current regime
will be reached for d>10 nm and at a much lower bias
voltage. These numbers are putting an extreme constraint on
the size of the molecule (or of the atomic scale circuit) that
would be accessible with Nelectrodes. For a very complex
molecule of effective lateral size of about D=5nm,N<10
on an insulator and N<2 on a semiconductor surface. The
number Nof required interconnects is limited very much by
the surface electronic characteristics and care must be taken
to optimize the computing power of the molecule for a given
lateral extension D.
In a standard circuit design based on the Kirchhoff laws,
and aside from the fabrication and the physical limits on δ
and d, there is also an important architectural consideration
which fortunately limits Nfor a given D. Already in
microelectronics, the empirical law N=kMp(k=2.24 and
p=0.5) limits the minimum number ofinterconnects required
on average by a designer to interconnect a microprocessor logic
circuit made of a assembly of Minterconnected transistors [5].
For the atomic scale, if the diameter Dof the molecule is too
small, there will be not enough space in the molecule or in the
atomic circuit to support, in a classical circuit design, a given
computing power (measured for example by the complexity
of the logic gate implanted in the molecule circuit). In this
case, a complex logic circuit will require a very large molecule,
maybe in the D=10 nm range and the number Nof required
interconnects will be rather limited. On the contrary, if a large
computing power can be implanted in a molecule, for example
by a quantum approach [13], difficulties may arise since it
will be impossible to make enough Natomic wires converging
towards this molecule to benefit from the full computing power
of the molecule in a spatial interconnection architecture. In
this case, multiplexing of the data will be required, forcing the
interconnection machine to use photons and not electrons to
exchange data with a single molecule. This will introduce a
new range of surface science studies for dielectric or plasmonic
wave guides converging at the atomic scale towards a single
complex molecule.
Acknowledgments
This work has been supported by the French National
Research Agency DiNaMo project n◦ANR-05-PNANO-014.
Partial support from the European Commission within the
project Picolnside (Contract n◦IST-015847) is gratefully
acknowledged. The authors wish to thank the European
Commission, the CNRS and A*STAR for their continuous
support during the definition, the design and now the
construction of atomic scale interconnection machines. We are
grateful to E Dujardin, T Ondarc¸uhu and J S Yang for their
contribution to this work.
References
[1] Rapenne G, Launay J-P and Joachim C 2006 J. Phys.:
Condens. Matter 18 S1797
[2] Joachim C, Gimzeswki J K and Aviram A 2000 Nature
408 541
[3] Wada Y 1996 Microelectron. Eng. 30 375
[4] Ample F, Ami S, Joachim C, Thieman F and Rapenne G 2007
Chem. Phys. Lett. 434 280
[5] Joachim C 2002 Nanotechnology 13 R1
[6] Stojkovic S, Joachim C, Grill L and Moresco F 2005 Chem.
Phys. Lett. 408 134
[7] Ramachandran G K, Hopson T J, Rawlett A M, Nagahara L A,
Primak A and Linsay S M 2003 Science 300 1413
[8] Itoua S, Joachim C, Rousset B and Fabre N 1992
Nanotechnology 310
[9] Saifullah M S M, Ondarc¸uhu T, Koltsov D F, Joachim C and
Welland M 2002 Nanotechnology 13 659
[10] Cacciollati O, Joachim C, Martinez J-P and Carsenac F 2004
Int. J. Nanosci. 3233
[11] Chou S Y, Krauss P R and Renstrom P J 1996 Science 272 85
[12] Thet N T, Lwin M H, Kim H H, Chandrasekhar N and
Joachim C 2007 Nanotechnology 18 335301
[13] Duchemin I, Renaud N and Joachim C 2008 Chem. Phys. Lett.
452 269
[14] Soukiassian L, Mayne A J, Carbone M and Dujardin G 2003
Surf. Sci. 528 121
[15] Lin S, Li M, Dujardin E, Girard C and Mann S 2005 Adv.
Mater. 17 2553
[16] Ondarc¸uhu T, Nicu L, Cholet S, Bergaud C, Gerdes S and
Joachim C 2000 Rev. Sci. Instrum. 71 2987
[17] Heinz K, Bernhardt J, Schardt J and Starke U 2004 J. Phys.:
Condens. Matter 16 S1705
[18] Feenstraa R M, Dong Y and Lee D C 2005 J. Vac. Sci. Technol.
B23 1174
[19] Sch¨onberger U and Aryasetiawan F 1995 Phys. Rev. B
52 8788–93
[20] Schintke S and Schneider W-D 2004 J. Phys.: Condens. Matter
16 R49
[21] Giessibl F J 2003 Rev. Mod. Phys. 75 949
[22] Morita S, Wiesendanger R and Meyer E 2002 Noncontact
Atomic Force Microscopy (Berlin: Springer)
[23] Zhong Q, Innis D, Kjoller K and Ellings V B 1993 Surf. Sci.
290 L688
[24] Albrecht T R, Grutter P, Horne H K and Rugar D 1991 J. Appl.
Phys. 69 668
[25] Giessibl F J 1995 Science 267 68
[26] Kitamura S and Iwatsuki M 1995 Japan. J. Appl. Phys. 2
34 L145
[27] Fukuma T, Higgins M J and Jarvis S P 2007 Phys. Rev. Lett.
98 106101
[28] Venegas de la Cerda M A, Abad J, Madgavkar A,
Martrou D and Gauthier S 2008 Nanotechnology 19 045503
[29] Bammerlin M, L¨uthi R, Meyer E, Baratoff A, L ¨uJ,
Guggisberg M, Loppacher C, Gerber C and
G¨untherodt H-J 1998 Appl. Phys. A66 S293
[30] Barth C and Henry C R 2003 Phys. Rev. Lett. 91 196102
[31] Barth C and Reichling M 2001 Nature 414 54
[32] Barth C, Foster A S, Reichling M and Shluger A L 2001
J. Phys.: Condens. Matter 13 2061
[33] Pfeiffer 0, Gnecco E, Zimmerli L, Maier S, Meyer E, Nony L,
Bennewitz R, Diedrich F, Fang H and Bonifazi D 2005
J. Phys.: Conf. Ser. 19 166
17

J. Phys.: Condens. Matter 22 (2010) 084025 C Joachim et al
[34] Bennewitz R 2006 J. Phys.: Condens. Matter 18 R417
[35] Nony L, Gnecco E, Baratoff A, Alkauskas A, Bennewitz R,
Pfeiffer O, Maier S, Wetzel A, Meyer E and Gerber C 2004
Nano Lett. 11 2185–9
[36] Burke S A, Ji W, Mativetsky J M, Topple J M, Fostner S,
Gao H-J, Guo H and Gr¨utter P 2008 Phys.Rev.Lett.
100 186104
[37] Dienel T, Loppacher C, Mannsfeld S C B, Forker R and
Fritz T 2008 Adv. Mater. 20 959
[38] Fendrich M and Kunstmann T 2007 Appl. Phys. Lett.
91 023101
[39] Trevethan T, Kantorovich L N, Polesel-Maris J, Gauthier S and
Shluger A L 2007 Phys. Rev. B76 085414
[40] Bouju X, Joachim C and Girard C 1994 Phys. Rev. B50 7893
[41] Sushko M L, Gal A Y, Watkins M and Shluger A L 2006
Nanotechnology 17 2062
[42] Oyabu N, Custance O, Yi I, Sugawara Y and Morita S 2003
Phys. Rev. Lett. 90 176102
[43] Sugimoto Y, Abe M, Hirayama S, Oyabu N, Custance O and
Morita S 2005 Nat. Mater. 4156
[44] Hirth S, Ostendorf F and Reichling M 2006 Nanotechnology
17 S148
[45] Joachim C, Bouju X and Girard C 1993 Atomic and Nanometer
Scale Modification of Materials: Fundamentals and
Applications (NATO ASI Series, Series E: Applied Science)
vol 239 (London: Kluwer Academic)
[46] Meyer E 2009 personal communication
[47] Gross L, Rieder K H, Moresco F, Stojkovic S, Gourdon A and
Joachim C 2005 Nat. Mater. 4892
[48] Rosei F, Schunack M, Jiang P, Gourdon A, Joachim C and
Besenbacher F 2002 Science 296 328
[49] Baratin R and Gourdon A 2009 Eur. J. Org. Chem. 71022
[50] Bombis C, Kalashnyk N, Xu W, Laegsgaard E,
Besenbacher F and Linderoth T R 2009 Small 52177
[51] Lafferentz L, Ample F, Yu H, Hercht S, Joachim C and
Grill L 2009 Science 323 1193
[52] Joachim C and Ratner M A 2005 Proc. Natl Acad. Sci.
102 8801
[53] Bauer E 1958 Z. Kristall. 110 372
[54] Noguera C 1996 Physics and Chemistry at Oxide Surfaces
(Cambridge: Cambridge University Press)
[55] Pezzagna S, V´ezian S, Brault J and Massies J 2008 Appl. Phys.
Lett. 92 233111
[56] L¨uthi R, Schittler R R, Brugger J, Vettiger P, Welland M E and
Gimzewski J K 1999 Appl. Phys. Lett. 75 1314
[57] Zahl P, Bammerlin M, Meyer G and Schlittler R R 2005 Rev.
Sci. Instrum. 76 023707
[58] Egger S, Llie A, Fu T, Chongsathien J, Kang D and
Welland M E 2005 Nano Lett. 515
[59] Guo H, Martrou D, Zambelli T, Polesel-Maris J, Piednoir A,
Dujardin E, Gauthier S, van den Boogaart M A F,
Doeswijk L M and Brugger J 2007 Appl. Phys. Lett.
90 093113
[60] Guo H, Martrou D, Zambelli T, Dujardin E and
Gauthier S 2008 Rev. Sci. Instrum. 79 103904
[61] Kolbel M, Tjerkstra R W, Brugger J, van Rijn C J M,
Nijdam W, Huskens J and Reinhoudt D N 2002 Nano Lett.
21339
[62] Yan X M, Contreras A M, Koebel M M, Liddle J A and
Somorjai G A 2005 Nano Lett. 51129
[63] K¨olbel M, Tjerkstra R W, Kim G, Brugger J, van Rijn C J M,
Nijdam W, Huskens J and Reinhoudt D N 2003 Adv. Funct.
Mater. 13 219
[64] Tun T N, Lwin M H T, Kim H H, Chandrasekhar N and
Joachim C 2007 Nanotechnology 18 335301
[65] Vazquez-Mena O, Villanueva G, van den Boogaart M A F,
Savu V and Brugger J 2008 Microelectron. Eng. 85 1237
[66] Coutrot A-L, Roblin C, Lafosse X, David C, Madouri A,
Laloo R and Martrou D 2009 Microelectron. Eng. 86 119
[67] Akremi A, Lacharme P and S´ebenne A 1997 Surf. Sci. 192 377
[68] Bellec A, Riedel D, Dujardin G, Rompotis N and
Kantorovich L 2008 Phys. Rev. B78 165302
[69] Hofer W A, Fisher A J, Lopinski G P and Wolkow R A 2001
Phys. Rev. B63 085314
[70] Bolland J J 1991 Phys. Rev. B44 1383
[71] Yang J S, Deng J, Chandrasekhar N and Joachim C 2007
J. Vac. Sci. Technol. 25 1694
[72] Yong K S, Oltavaro D M, Duchemin I, Saeys M and
Joachim C 2008 Phys. Rev. B77 205429
[73] Kiwari R K, Yang J S, Saeys M and Joachim C 2008 Surf. Sci.
602 2628
[74] Hosoki S, Hosaka S and Hasegawa T 1992 Appl. Surf. Sci.
60/61 643
[75] Bellec A, Ample F, Riedel D, Dujardin G and Joachim C 2009
Nano Lett. 9144
[76] Villagomez C, Zambelli T, Gauthier S, Gourdon A, Barthes C,
Stojkovic S and Joachim C 2007 Chem. Phys. Lett. 450 107
[77] Doumergue P, Pizzagalli L, Joachim C, Altibelli A and
Baratoff A 1999 Phys. Rev. B59 15910
[78] Huang D H, Uchida H and Aono M 1992 Japan. J. Appl. Phys.
31 4501
[79] Lyding J W, Chen J C, Hubacek J S, Tucker J R and
Abeln C C 1994 Appl. Phys. Lett. 64 2010
[80] Tanaka M, Chu F, Shimojo M, Takegushi M, Mitsuishi K and
Furuya K 2006 J. Mater. Sci. 41 2667
[81] Deng J, Troadec C, Kim H H and Joachim C 2009
Nanotechnology submitted
[82] Muller E W and Tsong T T 1969 Field Ion Microscopy:
Principles and Applications (New York: Elsevier)
[83] Rezeq M, Joachim C and Chandrasekhar N 2009
Microelectron. Eng. 86 996
[84] Rezeq M, Pitters J and Wolkow R 2007 J. Scanning Probe
Microsc. 21
[85] Rezeq M, Pitters J and Wolkow R 2006 J. Chem. Phys.
124 204716
[86] Rezeq M, Joachim C and Chandrasekhar N 2009 Surf. Sci.
603 697
[87] Thet N T, Joachim C and Chandrasekhar N 2009 Patent No
WO 2009/022982 19 Feburary, Int. Patent classification:
H01L 21/768 (2006.01) B82B 3/00 (2006.01)
[88] Mei W N, Boyer L L, Ossowski M M and Stokes H T 2000
Phys. Rev. B61 11425
[89] Roessler D M and Walker W C 1967 Phys. Rev. 159 733–8
[90] French R H 1990 J. Am. Ceram. Soc. 73 477
[91] Rubloff G W 1972 Phys. Rev. B5662
[92] Cimalla V, Pezoldt J and Ambacher O 2007 J. Phys. D: Appl.
Phys. 40 6386
18