Appl. Phys.

A (2014) 117:1689–1696
DOI 10.1007/s00339-014-8631-9

Dielectric sensing by charging energy modulation in a

nano-granular metal
Michael Huth • Florian Kolb • Harald Plank

Received: 11 March 2014 / Accepted: 15 July 2014 / Published online: 29 July 2014
Ó Springer-Verlag Berlin Heidelberg 2014

Abstract Several sensing concepts using nanostructures 1 Introduction

prepared by focused-electron-beam-induced deposition
have been developed over the last years. Following work The potential for the development of highly miniaturized
on highly miniaturized Hall sensors for magnetic sensing sensor elements based on nanostructures fabricated by
with soft magnetic Co, strain and force sensing based on focused-electron-beam-induced deposition (FEBID) has
nano-granular platinum–carbon structures (Pt(C)) was been recognized rather recently. Sensitive Hall sensor
introduced. Very recently, the capability of nano-granular devices have been developed employing Co-based struc-
Pt(C) structures to detect the presence of adsorbate water tures obtained with the precursor Co2 (CO)8 with excellent
layers by conductance modulations was demonstrated. For sensitivity for local magnetic stray field detection [1, 2].
magnetic and strain sensing, the underlying physical The development of miniaturized strain and force sensors
mechanisms of the sensing effect have been analyzed in followed next which took advantage of the nano-granu-
detail and are now quite well understood. This is not the larity of Pt(C)-based FEBID structures obtained from the
case for the adsorbate layer-induced conductance modula- precursor (CH3)3CpCH3 Pt (Cp: cyclopentadienyl) [3] and
tion effect. Here, we provide a theoretical framework that proved to be highly interesting for atomic force microscopy
allows for a semi-quantitative understanding of the sensors [4]. The working principle for the exploited strain-
observed water-sensing effect. We show how the near- resistance effect was found to be rooted in the sensitive
interface renormalization of the Coulomb charging energy dependence of the hopping transport between the metallic
in the nano-granular metal caused by the dielectric Pt nano-grains of about 3 nm diameter through a carbo-
screening of the polarizable adsorbate layer leads to a naceous matrix [5]. Based on the same nano-granular FE-
conductance modulation. The model can account for the BID material, the most recent sensor development focused
conductance modulation observed in the water adsorbate on selective gas sensing using adsorbed water as a model
experiments and can also be applied to understand similar system [6]. In this work, a custom-made gas measurement
effects caused by near-interface dielectric anomalies of chamber was used to detect the conductance change of
ferroelectric thin films on top of nano-granular Pt(C). Our Pt(C) FEBID structures in the 5–60 nm thickness range
findings provide a pathway toward optimized nano-granu- under exposure to different relative humidity levels under
lar layer structures suitable for a wide range of dielectric or well-controlled conditions. For 5 nm devices, conductance
local potential sensing applications. modulations above 10 % were observed at humidity levels
above 40 %. Only a preliminary analysis of the physical
mechanism causing this conductance modulation, namely
M. Huth (&)
Physikalisches Institut, Goethe University, Max-von-Laue-Str. 1, the influence of the polarizable water adsorbate layer on the
60438 Frankfurt am Main, Germany inter-granular charge transport, was given in [6]. Here, we
e-mail: [email protected] present a theoretical framework, which is able to fully
account for the experimental findings. The theoretical
F. Kolb  H. Plank
Institute for Electron Microscopy and Nanoanalysis, Graz treatment is formulated on the mean-field level and can be
University of Technology, Steyrergasse 17, 8010 Graz, Austria applied to all transport regimes of nano-granular metals

123
1690 M. Huth et al.

from weak to strong inter-grain tunnel coupling. The for- medium. In this case, the coupling constant g would also be
malism is sufficiently general to be applicable to other modified. This effect is not considered in the approach
dielectric sensing scenarios based on three-layer hetero- followed here, because it will be rather small for the het-
structures with one nano-granular layer. It has already been erostructures considered here. If the matrix were to be
successfully applied to analyze recent experimental find- replaced fully by a highly polarizable medium, or even a
ings on the strong conductance modulations observed in ferroelectric material, polarization-induced changes of g
nano-granular Pt(C) structures in close proximity to thin can become relevant, as was recently shown in a theoretical
layers of an organic ferroelectricum [7]. We can anticipate study [9].
that our work will be of interest for the area of dielectric or
local potential sensing in general and may inspire future 2.1 Transport regimes of nano-granular metals
work on the basis of highly miniaturized nano-granular
sensing layers prepared by FEBID or other thin-film For as-grown Pt(C) nano-granular FEBID structures, the
techniques. temperature-dependent conductivity follows a stretched
exponential behavior of the form
h i
rðTÞ ¼ r00 exp ðT0 =TÞ1=2 ð1Þ
2 Modeling approach
which was experimentally shown in several works (see,
FEBID structures prepared with organometallic precursors e.g., [4, 12]). This behavior was only recently theoretically
often show a nano-granular microstructure, i.e., metallic understood as a variant of variable range hopping (corre-
grains of a few nm of diameter are embedded in a carbo- lated VRH) governed by sequential inelastic co-tunneling
naceous matrix. As long as the metallic grains do not between the grains that takes disorder-induced, local
directly touch to form a percolating path through the background charges and the charging energy of the grains
sample, charge transport is governed by thermally assisted into account [8]. Here, the activation temperature T0 is
tunneling of the electrons. Recent theoretical work has governed by the inelastic co-tunneling range n as follows
shown that different transport regimes have to be dis-
criminated depending on the strength of the inter-grain be2 4R
kB T0 ¼ and nðTÞ ¼  
tunnel coupling g and temperature T [8]. Here, the cou- 4p0 r n ln E =16pcin gðkB TÞ2
2
pling constant g is a dimensionless quantity measured in
ð2Þ
units of the spin-degenerate conductance quantum 2e2 =h.
For a material to qualify as a granular metal, the condition with E ¼ EC and cin ¼ 1 for simplicity (see [5] for details).
g  g0 has to be fulfilled. g0 denotes the dimensionless kB denotes the Boltzmann constant.
conductance within a grain, which is assumed to be in the By post-growth electron irradiation of Pt(C) nano-
diffusive limit [8]. Due to the small size of the metallic granular FEBID structures, the tunnel coupling strength
grains, their charge capacity C is very small. Assuming a between the Pt grains can be strongly increased [10, 11] up
spherical grain of diameter 2R (R: radius), the capacitance to the point of an insulator–metal transition at a critical
amounts to 4p0 r R, with 0 and r the dielectric constants coupling strength gc  0:3 [12]. Experimentally one
of vacuum and the surrounding material, respectively. In observes, for couplings above gc , a logarithmic temperature
the tunneling process, each tunnel event is associated with dependence of the conductivity from room temperature
a charging energy which, for one grain in a matrix, is given down to about 10 to 20 K where it crosses over to an
pffiffiffiffi
by EC ¼ e2 =2C. It is the interplay of this charging energy, approximate T behavior [12]. Both of these temperature
possible finite size effects for very small grains (smaller dependences have also been theoretically predicted for a
than about 1 nm) and the microstructural disorder in the nano-granular metal modeled as a regular, three-dimen-
granular metals that leads to the different transport regimes sional array of metallic grains in the strong-coupling limit.
[8]. According to theory, the logarithmic temperature-depen-
In order to develop a model that can account for the dent correction to the diffusive conductivity r0 is given by
change of the conductivity of a thin nano-granular metal Eq. 3 (three-dimensional case) [8]
prepared in contact with a polarizable medium, such as an   
1 gEc
adsorbed water layer or a dielectric thin film, the influence rðTÞ ¼ r0 1  ln ð3Þ
6pg kB T
of the polarizable medium on the charging energy has to be
considered. In principle, also the attenuation length of the It is important to note that within the critical region itself
electronic wave function at the surface of the grains could (g  gc ), theory is not able to provide a conclusive answer
be influenced by the presence of the nearby polarizable to the temperature dependence of the conductivity.

123
Dielectric sensing by charging energy modulation in a nano-granular metal 1691

Fig. 1 Geometric
representation of the three-layer
heterostructure scenarios (A)
and (B). a The nano-granular
metal layer (2) is sandwiched
between the dielectric layers 1
and 3. b The nano-granular
metal layer forms the bottom
layer (1) on top of which a
polarizable adsorbate layer (2)
and a nonpolarizable layer or
weakly polarizable layer (3) are
assumed. c One grain of the
nano-granular metal together
with six point sources at
positions rQ;i (at half-radius
positions) and collocation points
rC;i (on surface). See text for
details. The lower part of
c represents the geometry for
scenario (B) in side view

2.2 Layer geometry of modeling approach the grain capacitance, if the thickness of layer 2 changes.
This scenario is used here to study the effects of an
In order to make progress in understanding the observed adsorbed layer, such as water, on the conductance of the
conductance changes of the nano-granular metal in prox- nano-granular metal. At this point, one may criticize that
imity to an electrically polarizable medium, we introduce the second scenario does not take the finite thickness of the
the three-layer model structure depicted in Fig. 1. It con- nano-granular layer into account which is itself deposited
sists of a layer with index 2 of thickness d2 and dielectric on a different dielectric half-space (substrate). This is a
constant 2 embedded in two layers of infinite thickness valid point. However, as will be shown next, the three-layer
with dielectric constants 1 and 3 . We wish to calculate the approach followed here allows for a transparent analytic
change of the capacitance of a spherical metal grain of solution of the capacitance problem and is appropriate for
diameter 2R embedded in layer 2 or 1 (see Fig. 1a or b). describing the adsorbate effects without introducing too
The spherical grain is one out of a regular array of grains much computational complexity. Also, the conceptual
for which we assume the same diameter. A nano-granular steps in this derivation can be extended to more complex
metal layer of given thickness dG is now formed by layer structures.
stacking several of these arrays in such a way that all
neighbored grains are coupled by the same coupling con- 2.3 Capacitance calculation for single grain
stant g. We discriminate two scenarios: (A) The nano-
granular metal layer is surrounded by two dielectric half We now turn to the capacitance calculation for a spherical
spaces (Fig. 1a). Half-space 1 represents the substrate and grain embedded either in layer 1 or layer 2. Here, two steps
half-space 3 a highly polarizable layer of rather large are needed. In the first step, we wish to get a closed
thickness. In this case, any change of the thickness of layer expression for the Coulomb potential at any given point
3 will not lead to changes in the capacitance of a grain in r ¼ ðx; y; zÞ in layer 1 or 2, if a point charge Q is placed at
layer 2. (B) The nano-granular metal layer represents layer any other point r0 in the same layer. Without loss of gen-
1 and is covered by two additional layers. We assume that erality, we set r0 ¼ ð0; 0; z0 Þ. In the next step, we use this
layer 2 is strongly polarizable, whereas layer 3 could be result to calculate the capacitance of a spherical grain
vacuum or air. In this case, we are interested in changes of employing the source point collocation method [13].

123
1692 M. Huth et al.

2.3.1 Coulomb potential of point charge in three-layer In order to ensure an equipotential surface on the spherical
medium grain, we replace the sphere by n point charges at prede-
fined fixed positions frQ;i g within the sphere volume. This
For the three-layer medium, the method of image charges is exemplarily shown for n ¼ 6 in Fig. 1c. The point
can be used to calculate the Coulomb potential in a charges Qi have to be determined such that the potential /0
straightforward way, if a\z0 \a and a\z\a (see on the n collocation points frC;i g on the surface of the
Fig. 1a, c) [14]. This is also possible for the case z0 \  a sphere is constant (see also Fig. 1c). From this, the fol-
or z0 [ a, as was shown in [15]. A substantial reduction of lowing system of linear equations results ðj ¼ 1; . . .nÞ
complexity can be reached by Fourier transforming the X n
1 Qi
expressions for the Coulomb potential in x and y /0 ¼ ð8Þ
4p0 r jrQ;i  rC;j j
(R ¼ ðx; yÞ). Here, we collect the expressions relevant in i¼1
the present case following [15]: from which the unknown point charges Qi are obtained by
Scenario (A): a\z0 \a; a\z\a matrix inversion. The sought for capacitance then follows
Q 2p h kjzz0 j 0 directly from
Vðk; z; z0 Þ ¼ e  DfL12 ekjzþz þDj Pn
4p0 2 k
i Qi
0
þ L32 ekjzþz Dj  2L12 L32 cosh ðkjz  z0 jÞg C ¼ i¼1 ð9Þ
/0
ð4Þ In Fig. 2, we show the result of capacitance calculations
0
Scenario (B): z \  a; z\  a within scenario (B) performed with n ¼ 6 and with varying
thickness of layer 2, simulating an adsorbate layer of water.
Q 2 22 2p h kjzz0 j The simulation parameters used are already chosen with a
Vðk; z; z0 Þ ¼ e
4p0 2 þ 3 2 þ 1 k ð5Þ view to using the results for modeling the experimentally
0 0
i
 ekðjzjþjz jÞ þ Dekðjzjþjz jÞ observed conductance modulations of Pt(C) FEBID struc-
tures exposed to water under controlled relative humidity
with levels, which will be presented in the discussion section.
The simulation parameters are listed in Table 1. In all
i  2 1
Li2  and D ð6Þ cases, we assume a dielectric constant of 2 ¼ A ¼ 80 for
i þ 2 1  L12 L32 e2kD
the adsorbate layer, like for bulk liquid water [18]. We
and D ¼ 2a. comment on this simplifying assumption in the discussion
By back transforming into R-space, a closed integral section. With regard to the relevant thickness range, we
expression for the electrostatic potential is obtained take from literature results from scanning probe micros-
Z copy (SPM) and dynamic force spectroscopy (DFS)
1 1
VðR; z; z0 Þ ¼ kJ0 ðkRÞVðk; z; z0 Þdk ð7Þ experiments on the coverage of different surfaces with
2p 0
water films under different conditions of relative humidity
J0 ðkRÞ denotes the zero-order Bessel function of the first [19, 20]. From these experiments, we identify the suitable
kind. This integral has to be evaluated numerically and thickness range for dA to be from 0.25 nm, corresponding
yields the electrostatic potential for a given point charge to one monolayer of water, to 5 nm, as is observed at room
distribution, which is needed in the capacitance calculation temperature for relative humidity levels up to 90 %. Since
described next. the thickness of the water film depends on the substrate
material, we took values which were found for graphite
2.3.2 Capacitance calculation using source point [20]. This should be representative for the Pt(C) FEBID
collocation method structures employed here, if we assume that the surface
termination layer of the nano-granular metal is amorphous
The source point collocation method for capacitance cal- carbon. From these calculations, we deduce as the most
culations of arbitrarily shaped metallic electrodes in a relevant results: (1) with increasing adsorbate layer thick-
dielectric medium uses the fact that the surface of a metal ness, the capacitance of a spherical grain close to the
electrode is an equipotential surface. In the present case, interface between layer 1 and 2 increases. This effect
the metallic electrode is a spherical grain and the dielectric grows with increasing adsorbate layer thickness and tends
medium is formed by the surrounding granular metal, to saturate for larger thicknesses. (2) The capacitance
which is treated within an effective medium theory increase drops off for larger distances to the interface and
according to Maxwell and Garnett [16, 17]. This is tends to saturate for distances above about 20 nm. From
appropriate within the mean-field treatment of the problem. this, we can conclude that nano-granular metal layers are

123
Dielectric sensing by charging energy modulation in a nano-granular metal 1693

theoretically predicted by Gor’kov and Eliashberg [21],

critically discussed by Straessler et al. [22] and is now
investigated in connection with interrupted metal strands
of ultrafine particles [23]. In the present case, for particle
sizes of a few nm, the average level spacing at the Fermi
level is in the 1 meV range or below, so that in the
discussed temperature range, the level spectrum can be
approximated as continuous [8]. In addition, the grains are
sufficiently well separated from each other by the
dielectric carbonaceous matrix, so that metal strand for-
mation can be excluded.

2.4 Conductance calculation

In order to calculate the conductance r of a nano-granular

metal layer of thickness dG we employ a simple parallel
Fig. 2 Capacitance of spherical grain as a function of the distance resistor model. For any given thickness dG we consider the
from the interface between layer 1 and 2. The nano-granular metal is layer as a parallel circuit of the conductance of m single
in layer 1. The different curves refer to different adsorbed water layer layers of identical spherical grain arrays. For each grain
thicknesses of layer 2, as indicated. See Table 1 for the used
array, positioned at a certain distance zi form the layer 1/
simulation parameters
layer 2 interface, we use the capacitance of a single grain
calculated for this distance. From this, we obtain the cor-
Table 1 Table of parameters used for the simulations shown in Figs. responding single-layer conductance rðzi Þ and the overall
P
2, 3 and 4 conductance r ¼ i rðzi Þ. In all cases, we assume that the
Figs. 1 2 3 g T/K R/nm dA /nm dG /nm first layer of grains (next to the interface) is at a distance
z1 ¼ R þ 0:5 nm. The distance between subsequent single
2 5 80 1 – – 1.5 Variable –
layers is assumed to be 2R. This corresponds approxi-
3 5 80 1 0.01 298 1.5 Variable Variable
mately to a close-packed grain arrangement with a 0.5 nm
4 5 80 1 Variable 298 1.5 2 22
distance (tunnel barriers) between the surfaces of neigh-
For the capacitance calculations n ¼ 6 was chosen. The fixed source boring grains. The calculated conductance values given
point positions were set to the sphere half-radius value along the three below are all stated as relative changes of the conductance
orthogonal axes from the center of the sphere, as is schematically
indicated in Fig. 1c
for different adsorbate layer thicknesses
rd  rdA;ref
Dr ¼ A ð10Þ
rdA;ref
most sensitive to this capacitance renormalization effects They do not depend sensitively on the chosen geometric
for thicknesses up to the 20 nm range. model for the microstructural arrangement of the spherical
The accuracy of the capacitance calculation based on the grains.
source point collocation method can easily be improved by At room temperature, this model is adequate since the
using n ¼ 8 or n ¼ 12. We did comparative capacitance average residence time of an excess electron on a single
calculations with these larger n-values and found a very grain is small due to the thermally assisted tunneling. In
good correspondence with the n ¼ 6 calculation within an this case, nonhomogenous current distributions due to local
error margin of less than 4 %. In the following, we use the Coulomb blockade effects can be neglected. Also, in the
obtained capacitance values and calculate the correspond- strong-coupling limit g [ gc the parallel resistor model is
ing changes in the conductance of a nano-granular metal appropriate. However, additional complexity arises at low
due to the associated reduction of the charging energy temperatures in the weak-coupling limit. In this case, a
EC / 1=C. distribution of charges in the system of nano-grains has to
Concluding this subsection, we would like to note that be calculated such that the overall electrostatic energy of
possible enhancements of the electrical polarizability of the system is at its minimum. This has been done as a
the nano-grains due to their discrete level spectrum, for function of the applied bias voltage in the limiting case of
temperatures sufficiently low so that no thermally induced next-neighbor tunneling only [24]. Within the correlated
level mixing can occur, are not relevant in the present VRH scenario, no corresponding theoretical analysis is
case. Such a polarization enhancement has been available.

123
1694 M. Huth et al.

Fig. 3 a Conductance change of nano-granular metal layers of

different thickness as function of the water adsorbate layer thickness.
As a reference value, a two-monolayer coverage, i.e., dA ¼ 0:5 nm, Fig. 4 Conductance change of a nano-granular metal with fixed
was assumed. b Conductance change at fixed adsorbate thickness as values for dG and dA as function of the inter-granular coupling
function of the nano-granular layer thickness. See Table 1 for the strength g within the weak- (left) and strong-coupling regime (right).
used simulation parameters The critical region around g ¼ gc is shown by the gray area. See
Table 1 for the used simulation parameters

We now discuss the consequences of the increased

capacitance due to the adsorbate layer for the weak- and modulation of several 10 % for a nano-granular metal
strong-coupling regime. consisting of only two grain layers (dG ¼ 7 nm) as the
adsorbate layer thickness increases to 5 nm, and a fast drop
2.4.1 Weak-coupling regime of the conductance changes with increasing thickness of
the nano-granular metal. The parameter values for dA and
In the weak-coupling regime, following the correlated dG were selected with a view to previous experimental
VRH transport behavior, the renormalized capacitance of a observations of the water adsorbate effect shown in Fig. 5.
grain close to the interface leads to two modifications in the A comparison of our model calculations with these
respective activation temperature T0 . As can be seen from experimental results follows in the discussion section.
Eq. 2, T0 depends on the inelastic co-tunneling range
which on the other hand depends on the effective dielectric 2.4.2 Strong-coupling regime
constant of the nano-granular metal r and the charging
energy EC . We use the calculated capacitance Cðzi Þ of a The model calculations in the strong-coupling regime are
single grain within a grain array at a given distance to the governed by the logarithmic correction of the temperature
interface to obtain the renormalized dielectric constant dependence of the conductance (see Eq. 3). They use the
r ðzi Þ and charging energy EC ðzi Þ as follows renormalized values for the charging energy EC ðzi Þ, as
Cðzi Þ e2 already given in Eq. 11. One would expect a rather weak
r ðzi Þ ¼ 1 and EC ðzi Þ ¼ ð11Þ response of the conductance on an adsorbate layer, since in
CG;0 2Cðzi Þ
the strong-coupling regime the charging energy is not very
with CG;0 representing the grain capacitance in vacuum, relevant anymore. This is confirmed by the results of the
i.e., CG;0 ¼ 4p0 R. From this, the corresponding layer- model calculations shown in Fig. 4 which presents the
specific activation temperature T0 ðzi Þ is obtained which dependence of the conductance modulation on the coupling
yields via Eq. 1 the single-layer conductance. strength in the weak- (left) and strong-coupling regime
In Fig. 3a, we show the result of a calculation of the (right). Within the critical region, the theory is not appli-
relative conductance change of nano-granular layers of cable. In this respect, the apparent increase of the con-
different thickness as a function of the adsorbate layer ductance modulation as the critical region is approached
thickness dA . In Fig. 3b, the dependence of the relative from the weak-coupling side should be considered with
conductance change on the layer thickness dG is depicted caution. In the next section, we discuss the results of the
for a fixed adsorbate layer thickness of 2 nm. Apparent model calculation with a view to the experimental
qualitative trends are a rather strong conductance observations.

123
Dielectric sensing by charging energy modulation in a nano-granular metal 1695

an order of magnitude between hydrophilic and hydro-

phobic surfaces.
Depending on the assumed relation between the final
adsorbate layer thickness and the relative humidity level,
the model calculations can account on a qualitative or even
semi-quantitative level for the experimental observations.
We note, in particular, that the experimentally observed
drop by about a factor of 2.9 (refer to dashed line in
Fig. 5b) of the conductance going from dG ¼ 5 to 20 nm is
very well reproduced by the model calculations, which
yield a factor of 2.4 (cmp. values for dG ¼ 7 and 22 nm).
We are therefore confident that our treatment on the mean-
field level covers the essential aspects of the physics
involved. Nevertheless, we would like to stress again that
the exact relationship between the adsorbate layer thick-
Fig. 5 a Experimentally observed conductance change of nano-
granular Pt(C) FEBID layer of different thickness as function of the ness and the relative humidity level is not known. Fur-
relative humidity level, i.e., water adsorbate layer thickness. As a thermore, we would expect that the dielectric properties of
reference value, the measured conductance in the presence of a water the adsorbate layer will differ from the bulk properties in
adsorbate layer at 5 % humidity level was used. The temperature was
the few monolayer regime. This has not been accounted for
stabilized at 298 K. b Experimentally observed conductance change
at fixed relative humidity level of 75 % as function of the nano- in our model, although this could be remedied by a more
granular layer thickness. The dashed line represents a guide to the eye sophisticated calculation in which the adsorbate layer itself
is modeled with a z-dependent dielectric function 2 ðzÞ. In
this case, the capacitance calculations become much more
3 Discussion involved and one may need to resort to a full numerical
treatment of the ensuing boundary value problem.
We now compare the results from the modeling section
with the experimental observations previously presented in
[6]. In Fig. 5a, we show the experimental results for the 4 Conclusions
conductance modulation of FEBID-based Pt(C) nano-
granular metal layers of different thickness on the relative We have presented a mean-field modeling approach that
humidity level. Figure 5b depicts the relative conductance describes the influence of an electrically polarizable med-
change dependence on the nano-granular layer thickness at ium on the temperature-dependent conductance of a nano-
a fixed relative humidity level. The representation of the granular metal in the weak and strong inter-granular cou-
data in this form allows for a direct comparison with the pling regime. Based on a three-layer heterostructure
model calculations shown in Fig. 3. The experiments were geometry, two scenarios have been discriminated. In sce-
performed under well-controlled temperature (298 K) and nario (A), the nano-granular metal is assumed to be sand-
relative humidity level (5–80 %) conditions, as described wiched between two dielectric half spaces. This scenario
in detail in reference [6]. A quantitative comparison of our proved to be successful in describing the strong conduc-
calculations with the experimental data is not readily tance modulation observed for a nano-granular metal at the
possible since, to the best of our knowledge, the relation paraelectric–ferroelectric phase transition of an organic
between the adsorbate layer thickness and the relative ferroelectric thin film on top of a nano-granular Pt(C) FE-
humidity level is not known. What can be stated from BID structure [7]. Scenario (B) takes the nano-granular
independent research using scanning probe [19] and metal layer as the bottom layer of the three-layer hetero-
dynamic force spectroscopy [20] is the following: (1) The structure, the second layer is assumed to be a polarizable
water adsorbate layer thickness changes from one mono- adsorbate and the third layer represents vacuum or any
layer (0.25 nm) to 5 nm as the humidity level grows from other nonpolarizable or weakly polarizable medium. This
about 0 (low vacuum conditions) to more than 90 % under scenario was successfully applied to describe the conduc-
ambient temperature conditions. (2) The actual layer tance modulations observed for Pt(C) FEBID layers of 5 to
thickness for a given humidity level may show transient 60 nm thickness covered with adsorbed water under well-
behavior, in particular, a crossover from droplet formation controlled relative humidity levels. We can envision a
(about 5 nm height) to closed layers (about 2 nm thick- rather wide range of additional sensor applications that use
ness) over several hours. (3) The adsorbate layer thickness the influence of the polarization properties of different
depends on the substrate material and can vary by almost media in contact with thin nano-granular metal layers

123
1696 M. Huth et al.

prepared by FEBID. Due to the excellent down-scaling 6. F. Kolb, K. Schmoltner, M. Huth, A. Hohenau, J. Krenn, A. Klug,
capabilities of FEBID, spatially resolved sensor structures E.J.W. List, H. Plank, Nanotechnology 24, 305501 (2013)
7. M. Huth, A. Rippert, R. Sachser, L. Keller, submitted to Nano-
can be realized. Our findings suggest that the correlated technology (2014), arXiv:1404.7669
variable range hopping, i.e., weak-coupling, regime is most 8. I.S. Beloborodov, A.V. Lopatin, V.M. Vinokur, K.B. Efetov,
sensitive to dielectric modulation effects. Assuming a Rev. Mod. Phys. 79, 469 (2007)
typical inelastic co-tunneling range of about 20 nm, the 9. O.G. Udalov, N.M. Chtchelkatchev, A. Glatz, I.S. Beloborodov,
Phys. Rev. B 89, 054203 (2014)
working principle introduced here for nano-granular sensor 10. F. Porrati, R. Sachser, C.H. Schwalb, A. Frangakis, M. Huth, J.
structures should scale down gracefully to the sub-50 nm Appl. Phys. 109, 063715 (2011)
range. 11. H. Plank, G. Kothleitner, F. Hofer, S.G. Michelitsch, C. Gspan,
A. Hohenau, J. Krenn, J. Vac. Sci. Technol. A 29, 051801 (2011)
Acknowledgments M. H. thanks the Deutsche Forschungsgeme- 12. R. Sachser, F. Porrati, ChH Schwalb, M. Huth, Phys. Rev. Lett.
inschaft for financial support through the Collaborative Research 107, 206803 (2011)
Centre SFB/TR 49. 13. C. Wasshuber, Computational Single-Electronics, 139ff
(Springer, Wien, 2001)
14. N. Anderson, Am. J. Phys. 38, 1483 (1970)
15. R.G. Barrera, O. Guzman, B. Balaguer, Am. J. Phys. 46, 1172
References (1978)
16. J.C. Garnett, Philos. Trans. R. Soc. Lond. 203, 385 (1904); 205,
1. G. Boero, I. Utke, T. Bret, N. Quack, M. Todorova, S. Mouaziz, 237 (1906)
P. Kejik, J. Brugger, R.S. Popovic, P. Hoffmann, Appl. Phys. 17. X.C. Zeng, D.J. Bergmann, P.M. Hui, D. Stroud, Phys. Rev. B 38,
Lett. 86, 042503 (2005) 10970 (1988)
2. M. Gabureac, L. Bernau, I. Utke, G. Boero, Nanotechnology 21, 18. M. Uematsu, E.U. Frank, J. Phys. Chem. Ref. Data 9, 1291 (1980)
115503 (2010) 19. A. Gil, J. Colchero, M. Luna, J. Gomez-Herrero, A.M. Baro,
3. C.H. Schwalb, C. Grimm, M. Baranowski, R. Sachser, F. Porrati, Langmuir 16, 5086 (2000)
H. Reith, P. Das, J. Mller, F. Vlklein, A. Kaya, M. Huth, Sensors 20. A. Opitz, M. Scherge, S.I.-U. Ahmed, J.A. Schaefer, J. Appl.
10, 9847 (2010) Phys. 101, 064310 (2007)
4. M. Huth, F. Porrati, C.H. Schwalb, M. Winhold, R. Sachser, M. 21. L.P. Gor’kov, G.M. Eliashberg, Sov. Phys. JETP 21, 940 (1965)
Dukic, J. Adams, G. Fantner, Beilstein J. Nanotechnol. 3, 597 22. M. Strässler, M.J. Rice, P. Wyder, Phys. Rev. B 6, 2575 (1972)
(2012) 23. S.K. Saha, Phys. Rev. B 69, 125416 (2004)
5. M. Huth, J. Appl. Phys. 107, 113709 (2010) 24. A.A. Middleton, N.S. Wingreen, Phys. Rev. Lett. 71, 3198 (1993)

123