Sei sulla pagina 1di 66

Contrast inversion on the h-BN nanomesh

Masterthesis
University of Basel
Markus Langer

Prof. Ernst Meyer


Supervisors: Thilo Glatzel

Juli 2010
”Erfolg braucht Leistung”
”Präzision braucht Leidenschaft”
Markus Langer

i
Abstract
In this work an advance analysis of the h-BN nanomesh is given. The
often seen contrast inversion is discussed in detail and possible expla-
nations are shown. After the analysis, a height dependent contrast
became most probable. In addition, a domain analysis is perfomed,
which shows a low domain boundary mobility and no preferred domain
orientation. Atomic resolution with the torsional oscillations fT R was
achieved for large scan areas. In all measurements bimodal dynamic
force microscopy was used and the theoretical background is discussed.
Further, the torsional oscillation was used for imaging and 2D force
spectroscopy to achieve lateral force information on the nanomesh.

ii
Contents
1 Introduction 1

2 Theory 2
2.1 Interaction forces in DFM measurements . . . . . . . . . . . . 2
2.1.1 Electrostatic forces . . . . . . . . . . . . . . . . . . . . 2
2.1.2 Van der Waals force . . . . . . . . . . . . . . . . . . . 3
2.1.3 Short range forces . . . . . . . . . . . . . . . . . . . . 4
2.2 Cantilever dynamics . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.1 Bimodal detection . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Lateral force detection . . . . . . . . . . . . . . . . . . 7
2.3 Theoretical predictions and properties of h-BN nanomesh . . 9
2.3.1 Theory of the electronic structure and work function . 10

3 Experimental 12
3.1 Dynamic force microscope . . . . . . . . . . . . . . . . . . . . 12
3.2 Cantilever preparation . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Preparation of Rhodium(111) films . . . . . . . . . . . . . . . 13
3.4 Nanomesh preparation . . . . . . . . . . . . . . . . . . . . . . 14

4 Measurements 16
4.1 Mesh stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Contrast inversion . . . . . . . . . . . . . . . . . . . . . . . . 18
4.2.1 Tip apex changes . . . . . . . . . . . . . . . . . . . . . 19
4.2.2 Double tip . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.3 Tip geometry . . . . . . . . . . . . . . . . . . . . . . . 25
4.2.4 Height dependent contrast inversion . . . . . . . . . . 29
4.2.5 2D spectroscopy . . . . . . . . . . . . . . . . . . . . . 30
4.2.6 Detailed spectroscopy analysis . . . . . . . . . . . . . . 33
4.2.7 2D bimodal spectroscopy . . . . . . . . . . . . . . . . 36
4.3 Domain orientation . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Atomic contrast measured by torsional resonance . . . . . . . 46

5 Conclusion 49

6 Acknowledgment 50

iii
1 Introduction
Since the technique of feeling atoms by a frequency modulation dynamic force
microscope was established in 1995 [1], with a wide range of applications in
science. Applications from human medicine [2], biology [3], applied surface
science [4] to fundamental surface analysis [5] in physics. The frequency
modulated dynamic force microscopy as it is used for surface characteriza-
tion allows insights into the forces with high signal resolution as well as high
spacial resolution. For fundamental surface analysis this is a credent fea-
ture for revealing unseen surface properties. A major step towards the high
resolution data in this work was the use of bimodal excitation of different
resonance frequencies. Beside the fundamental flexural oscillation, a second
oscillation can be superimposed, like the second flexural or the first torsional
[20] [21]. Every oscillation is controlled separately, allowing high resolution
information acquisition. With the torsional oscillation we gain direct access
to the lateral forces [21].
This powerful techniques were used to bring clarity in the analysis of a sp2
monolayer on a metal surface [15]. In recent years, studies on sp2 monolayer
have become significant. One of the well known representatives is graphene
[6, 7, 8], a monolayer of carbon. Due to its properties graphene is a fas-
cinating material showing an increase of electron mobility, spin transport,
increased thermal conductivity and features different mechanical properties.
On the other hand, boron nitride forms a sp2 monolayer as well [23]. Evap-
oration on a rhodium(111) surface forms a corrugated monolayer with a
honeycomb superstructure, better known as the h-BN nanomesh. Unlike
graphene, the electronic structure makes it an insulator. The topic of this
master thesis is the study on this corrugated single layer as well as the ex-
planation of often observed features, like the contrast inversion.

1
2 Theory
By measuring the forces perpendicular to the surface the analysis of fun-
damental structures as well as the electronic properties are available. The
interacting forces between two objects have a large variety of origins and
will be discussed in the order of their decay length. Exact measurements
of forces require high sensitivity as well as precision. Our equipment for
surface characterization is a homebuilt atomic force microscope(AFM) [16].
The design and fundamental mode of operation of a nc-AFM is introduced
shortly. Cantilever dynamics and the advantage of bimodal dynamic force
microscopy (DFM) incorporate with the detected signal will be covered in
more detail.

2.1 Interaction forces in DFM measurements


The forces acting between two separated objects has rarely only one com-
ponent. In extreme cases different forces can have a dominant contribution.
For example, earth makes his circles around the sun, due to the attractive
gravitational force, because every other force has a smaller decay length and
is weak compared to gravitation. On the other hand, two electrons repel
each other at very small distances at room temperature, due to repulsive
coulomb forces. Here the strong repulsive force is dominant compared to
any other force.
Systems in the nanoscale range consist of many different forces. Analyzing
and characterizing surfaces with a dynamic force microscope is challenging,
due to the sum of interacting forces between tip and sample surface, which
are combined in one signal. Nevertheless it is possible to separate those
forces and their origins. First of all the decay length is a way for a clas-
sification. Two classes are defined, the forces with a long range character,
like the electrostatic forces, see section:2.1.1 and the van der Waals forces,
see section:2.1.2 and forces with a short range character, like the chemical
interaction, which is formulated either with the Lennard Jones potential, see
section:2.8 or the Morse potential Eq.:2.10.

2.1.1 Electrostatic forces


A force affects on two electric charges. This force is repulsive, if the charges
have the same sign. On the other hand the force is attractive for differently
signed charges. This electrostatic interaction is described by Coulomb’s law,
which specifies the strength and the decay length of the force acting on two
objects:

1 Q1 Q2
FCoulomb (z) = (2.1)
4π0 z 2

2
Here, 0 is the permittivity of the vacuum and Q1 and Q2 are the sums over
all charges in each object with the distance z between them. DFM allows
only the detection of localized charges within a sample with the tip above.
In the most cases localized charges are formed due to surface preparation,
like ultra high vacuum (UHV) cleaving of ionic crystals, changing abruptly
the bulk formation. The bulk-vacuum interface forms uncompensated local
charges, due to missing atoms. Ion sputtering as well as plasma deposition
can cause local charges in the surface. Those local charges can last for several
hours up to days under UHV conditions.
Electrostatic interaction does not need only defects or adatoms within the
surface. An electrostatic variation can arise due to charge transfer between
different materials, which causes a contact potential difference, resulting in
an electrostatic force.
In a DFM setup, tip and sample, can be assumed as a distance dependent
capacitor, with a capacity C. The force is given by:

1 ∂C
Fel (z) = (VBias − VCP D )2 . (2.2)
2 ∂z
The electrostatic force is minimal, when the applied bias voltage Vbias com-
pensates the contact potential VCP D . The contact potential arises due to
the different work functions of the materials.
The capacity gradient ∂C/∂z can be approximated by assuming the tip as
a cone with a half sphere on top defined by radius R. The dominant term
can be [14, 45] reduced to:

R
Fel (z) = −π0 (VBias − VCP D )2 . (2.3)
z

2.1.2 Van der Waals force


The van der Waals interaction between two neutral atoms is a quantum me-
chanical consequence. The expectation value for an electric dipole moment
vanishes for neutral atoms. This expectation value is subject to quantum
mechanical and thermal fluctuations. Those fluctuations trigger a dipole
moment even in neutral atoms. A spontaneous triggered dipole moment in
the first atom creates an electric field with a decay rate of 1/~r3 . A second
atom within the range of the decay rate gets electrically polarized. A dipole
moment is induced by d2 ∼ α2 /~r3 . α2 is the polarizability of the second
atom. The interaction of those dipole moments is proportional to −d1 d2 /~r3 .
The electric field of the second atom influences the dipole moment of the first
atom d1 ∼ α1 /~r3 . The resulting potential from the van der Waals interaction
is described by [11]:

d1 d2 α1 α2
V (~r) ∼ − 3
=− 6 . (2.4)
~r ~r

3
For the case of an integration over a sphere in the distance z above a surface
the potential adds up to a distance dependence which is proportional to
VvdW (z) ∝ −1/z. The force of the interaction is resulting in:

HR
FvdW (z) = (2.5)
6z 2
with the Hamaker constant H [12], tip radius R and distance z.The inter-
action of two dipole moments is fragile to any perturbation. The influence
of a third dipole moment is not additive, hence this theoretical approach is
only qualitative. One reason is the polarizability of an ensemble is not the
sum over the polarizabilities of every atom [11].

2.1.3 Short range forces


Short range force or chemical interaction arises, if the wave functions of
atom nuclei overlap. Those atom nuclei experience an attractive force on
each other, if the overlap causes a lower energy state of both systems. The
overlap can become also repulsive, if the Pauli exclusion principle is violated
(pauli repulsion) or due to coulomb repulsion of two nuclei. These forces can
be described via model potentials, which have been fitted on experimental
data and have no derivations.
The Lennard-Jones potential Fig.:1 describes the interaction of two neutral
atoms and the Morse potential Fig.:1 describes the interaction of a diatomic
molecule.
The Lennard Jones potential is given by:
  
σ 6  σ 12
VLJ (z) = −4 − . (2.6)
z z
 describes the effective height of the potential wall and σ the effective sep-
aration of the atoms. The Lennard Jones potential has a global minimum,
which defines the equilibrium bond distance z0 = 21/6 σ. The term of the
order of six describes the attractive van der Waals interaction, while the
term of the order of twelve specifies the repulsive interaction [19]. With the
equilibrium bond distance z0 we get:
   
z0 6  z0 12
VLJ (z) = − 2 − . (2.7)
z z
The corresponding force of the Lennard Jones potential is:
 
  z0 7  z0 13
FLJ (z) = −12 − . (2.8)
z0 z z
The Morse potential gives an approximation of a diatomic molecule. This
empirically derived potential was fitted on measurements, but can also be
justified due to the solution of the Schrödinger’s equation of a H2+ ions. It is

4
fully described by the parameter of the equilibrium bond distance z0 , binding
energy  and the effective range κ [19]:
 2
VM (z) =  1 − e−a(z−z0 ) . (2.9)
p
a is defined as κ/2, which defines the potential width and with this the
bond dissociation energy. The derivation −∂V /∂z gives:

∂VM  2
FM (z) = − = −2a e−a(z−z0 ) − ae−2a(z−z0 ) . (2.10)
∂z
Both potentials can be used to characterize the short range interaction be-
tween tip and sample.

Figure 1: Lennard Jones and Morse potential. Parameters:  = 1eV binding


energy, z0 = 0.2nm equilibrium bond distance and a = 20eV bond dissocia-
tion energy.

Figure 1 shows both potentials. Here we see, that the Morse potential is
repulsive for z ≤ 0.28nm. However, the Lennard Jones potential shows the
repulsive interaction at lower equilibrium bond distances z ≤ 0.2nm. On the
other hand the latter shows a larger binding energy of ∼ −12nN for inter
atomic interactions.

2.2 Cantilever dynamics


In non contact dynamic force microscopy, a silicon cantilever prong is excited.
This cantilever will oscillate free at the resonance fi for large tip-sample

5
separation, for decreasing tip-sample separation the interaction between tip
and sample will cause a shift in oscillation frequency ∆fi , in the imaging
mode this shift is kept constant to control the tip-sample distance. The free
oscillation becomes perturbated by the interaction, as long as the shift in
frequency ∆fi is small compared to the resonance frequency fi , the system
can be assumed by an slightly perturbed harmonic oscillator. For this system
the equation of motion is:

mi z̈ = −ki z + Aexci cos(ωi t + ϕi ) − γ ż + F (z). (2.11)


mi is the effective mass of the oscillating cantilever. Due to the small effective
mass of the cantilever gravitation can be neglected, hence the selection of
the z axis is free. ki is the effective stiffness of the i-th resonance and F (z)
the interaction force. γ defines the damping of the oscillation, Aexci is the
driving amplitude of the i-th resonance with the frequency of ωi = 2πfi .
The oscillation amplitude has a phase shift of ϕ = 90◦ in respect to the
driving amplitude Aexc . For a constant oscillation amplitude, one assumes
that the damping and the excitation are compensating each other. Thus,
the equation reduces to:

mi z̈ = −ki z + F (z). (2.12)


The amplitude Ai is usually larger than the decay length λ of the interaction
forces Ai  λ, therefore most of the forces will contribute at the lower
turning point of the oscillation. The cantilever trajectory is assumed to
be harmonic. This justifies the substitution of z = z0 + Ai cos(ωi t) in the
equation of motion 2.12.

−mi Ai ωi2 cos(ωi t) = −ki z0 − kAi cos(ωi t) + F (z0 + Ai cos(ωi t)). (2.13)
The integration over one period of the resonance frequency fi integrates all
forces, which contribute to the extension of the period.

  Z 2π
ωi 1 ωi
Ai ki 1− = F (z0 + Ai cos(ωi t))cos(ωi t)dt (2.14)
ωi0 2π 0

Where ωi0 is the angular frequency of the unpertubated system and ωi the
effective angular frequency. If both are comparable, the force interaction
averaged over one period is the experimental determined shift in frequency
∆fi divided by the resonance frequency fi . With Θi = ωi t
Z 1
∆fi 1 fi t
Ai ki = F (z0 + Ai cos(Θi ))cos(Θi )dΘi (2.15)
fi 2π 0
This frequency shift dependence has been shown at first by Giessibl [18] and
is valid for any resonance frequency, which is excited alone.

6
2.2.1 Bimodal detection
In this mode the cantilever is excited at two different resonance frequen-
cies simultaneously [20]. The first resonance frequency is used to control
the tip-sample distance, while the second resonance detects various surface
properties, like surface mechanics, magnetic structure and local contact po-
tential of heterogeneous materials. Bimodal DFM imaging was applied for
high resolution atomic-scale imaging, due to stable distance control of the
fundamental frequency and the enhanced sensitivity of higher order oscilla-
tion frequencies to short-range forces [20] [21].
The two resonance frequencies are incommensurate. Hence, we have to ac-
knowledge that the trajectory of one oscillation cycle differs from the next
oscillation cycle, because fi+1 /fi is not an integer, therefore we need to av-
erage over many oscillation cycles. Due to the limited, system-dependent
measurement bandwidth of only several Bm ≈ 100Hz and fi  Bm the tip
will perform a large number ni of oscillation cycles on every recording point.
Equation 2.15 changes to:
Z 2πni
∆fi 1
Ai ki = F (Tip(Z(t)))cos(Θi )dΘi (2.16)
fi 2πni 0
Tip(Z(t)) denotes the Tip apex position at the time t, while the z position
is adequately described by Z(t) = z0 + Ai cos(Θi ) + Aj cos(Θj ), here z0 is the
equilibrium position and Θi = 2πfi t the phase of the ith mode.
Equation 2.16 is valid for small shifts in frequency ∆fi of any resonance,
meaning ∆fi /fi  1. For i = 1 the cantilever is prevented of jump-to-
contacts for large A1st ; the condition i = 2 can be fulfilled even for very
small A2nd , due to the higher effective stiffness k2nd . f1st and f2nd are
incommensurate, except for the oscillating force component at fi , which
makes a finite contribution to the integral equation 2.16. So we expect the
same for A1st = A2nd for the right hand side. For the case, that A1st 
λ > A2nd we expect ∆f1 to show the same behaviour as in normal DFM. In
the limit f2nd  f1st this implies that the integration of Θ2 over successive
cycles of the oscillation at f1st is equivalent to a dense sampling over a single
cycle.
Z 2π
∆f2nd 1
k2nd =− F 0 (z0 + A1st cos(Θ1st ))dΘ1st (2.17)
f2nd 4π 0
F 0 = dF/dz varies quicker than F itself and ∆f2nd with the same A1st has a
stronger dependence on the short-range forces than ∆f1st in normal DFM.

2.2.2 Lateral force detection


In a usual DFM system, the cantilever oscillates in the vertical direction with
respect to the surface, hence only force variations in the out of plane direc-

7
tion FZ are detectable. This gives an access to a large set of information,
mainly the topography, local work function and the stiffness of materials,
etc.. But, due to the direction of the cantilever oscillation, we do not have
access to any information about a lateral force FX,Y .
The excitation of a lateral oscillation mode is possible for PPP-NCL at a
frequency of around fT R ∼ 1.5M Hz. This allows to detect lateral force
components FX [21]. Site-dependent long-range interaction cause no lateral
forces, but the strongly distance dependent short-range interactions[21]. So,
we expect extremely high spatial resolution in the frequency shift signal of
the torsional mode ∆fT R . The torsional amplitude of oscillation is very
small compared to the cantilever dimensions, therefore the tip apex is vi-
brating parallel to the surface. This allows an adequate sense of the lateral
interaction forces FX between tip and sample.
In bimodal DFM the flexural resonance fi for a stable height control is ex-
cited simultaneously to the torsional resonance fT R for lateral resolution.
This can be treated in a similar way with respect to the excitation of two
flexural resonances. The trajectory of the tip apex during one oscillation
cycle of the flexural resonance differs to the next, due to the incommen-
surable resonance frequencies. In every point of the oscillation, the short-
range interaction can be split into its components along Z and X. While
the site-dependent long range force only contributes to the out of plane force
component FZ . Assuming that all force components weakly perturbate the
harmonic oscillation, an expression for the resonance shift can be given as:
Z 2πni
∆fi 1
Ai ki = Fj (Tip(X, Z)))cos(Θi )dΘi (2.18)
fi 2πni 0
Here we denote Fj for the interaction force component in direction j and
Tip(X, Z) give the tip apex position in the X-Z plane, in the sense of:

Tip(X, Z) = (AT R cos(ΘT R ), z0 + Ai cosΘi ) (2.19)


with AT R for the torsional oscillation amplitude and ΘT R = 2πfT R t. High
resolution in lateral force detection is achieved, if the force is not averaged.
This means the corresponding amplitude of the lateral oscillation AT R must
be set smaller compared to the lattice constant. Like in bimodal DFM with
two flexural resonance modes, the shift ∆fT R is given, for the conditions
that A1st  λ > AT R :
Z 2π
∆fT R 1
kT R =− FX0 (z0 + A1st cos(Θ1st ))dΘ1st (2.20)
fT R 4π 0
here FX0 is the lateral interaction force gradient in the X direction. The
flexural amplitude has to be large compared to the decay length, to neglect
any tip apex deformation, due to the lateral interaction force.

8
Comparing the detection sensitivities of both bimodal DFM modes, the sec-
ond flexural ∆f2nd and the torsional ∆fT R with a large A1st . We notice,
that the difference between Equation 2.17 and 2.20 is the direction of the
force gradient. The torsional frequency shift arises from the site depen-
dent interactions, hence the macroscopic shape of tip apex should not affect
the lateral force detection, as long as stable imaging remains possible. The
lateral sensitivity is higher and better compared to a flexural mode with
ultra-small amplitudes. One of the reasons is the better mechanical quality
factor of the torsional mode QT R , which is typically around ∼ 140000 and
so, about 10 times higher than the quality factor of the second flexural mode
Q2nd ∼ 14000.

2.3 Theoretical predictions and properties of h-BN nanomesh


The hexagonal boron nitride single layer grows on a transition metal sur-
face, due to the good catalytic activity. Transition metals allow a perfect
layer growth by means of vapor deposition [31]. The single layers are sp2
hybridized honeycomb networks and form strong in plane σ and weaker π
bonds to the substrate and the adsorbates.
A corrugated single layer enhances the binding energy by a lateral binding
component for small molecules. These additional binding energies exceed the
thermal energy and makes adsorbates thermally stable up to several 100K
above RT . The nanomesh is a sp2 hybridized layer, which is very robust
to environment conditions so it can be immersed e.g. into liquids without
structural changes [44].
For developments in nanotechnology it is useful to have single layer systems
which are inert and remain clean at ambient conditions and are stable up to
high temperatures.
The h-BN nanomesh is an outstanding example for such a sp2 single layer
system, but has a highly reactive surface to bind adsorbates. Compared to
graphene, which is another sp2 hybridized layer, the nanomesh, grown on a
transition metal is an insulator, while graphene is metallic.
The atomic structure and the corresponding lattice constant of the sp2 layer
plays a key role in the understanding of the corrugated topography, depend-
ing on the strong in plane σ bonds of the honeycomb lattice and related
to these, the weak out of plane π bonds to the substrate. The π bonding
depends on the registry to the substrate atoms, where the layer - substrate
hybridization causes a tendency for lateral lock-in of the overlayer atoms to
the substrate atoms. If the over layer has the same lattice symmetry as the
substrate, a lattice mismatch M is defined as:
asurf ace − abulk
M=
abulk
where asurf ace and abulk are the lattice constant of the surface layers,respectively.

9
The sign of M defines a compressive stress of the surface layer (+) or a ten-
sile stress (−). If the lattice mismatch M of a laterally rigid sp2 exceeds a
critical value, corrugated super structures with a large lattice constant are
formed, which lead to a template function.
The theory of Greber et al. [15] predicts an importance of occupancy of
the d-band of the substrate by the nitrogen atoms. Since it is energetically
unfavorable, if the nitrogen atoms are moved in plane, away form the top
sites, it is expected that the energy will scale the adsorption energy. For
a rigid sp2 layer without a lateral lock-in energy available, no corrugation
is expected, but instead a moiré type pattern. But with a lock-in energy
gain for preferred bonding sites a dislocation of the surface layer is expected.
This involves the formation of commensurate coincidence lattices between
the substrate and the surface layer with regular dislocations.
For h-BN on rhodium(111) the lattice mismatch is M = −6.7% and the
13 × 13 BN units coincide with 12 × 12 Rh units Fig.2a. This leads to a
residual compression of the 13 BN units by 0.9%. The corrugated super
structure Fig.2b, called h-BN nanomesh, has a regular pattern with a peri-
odicity of 3.2nm. The pattern consists of a honeycomb mesh with ”wires”
and ”holes” or ”pores”.
It turned out, that the special super structure is only a corrugated single
layer of boron nitride. The height difference of the corrugated layer is cal-
culated to be 0.05nm. This is sufficient to produce a distinct functionality,
like trapping molecules [24]. The reason is, that the super structure has two
electronically distinct regions, which are related to the topography Fig. 2c
[15].

2.3.1 Theory of the electronic structure and work function


The work function is a material dependent constant. It is the minimum
energy to remove a electron from the Fermi to the vacuum level. For a dis-
cussion of the electric field near the surface, we have to recall the Helmholtz
equation [25] that relates classically the work function Φ of a flat surface
with vertical electric dipol:
e
Φ= Na · p, (2.21)
0
e denotes the elementary charge, 0 the permittivity and Na the areal density
of dipoles p, if the dipoles are assigned to the atoms.
If an insulating single layer is placed on top of a metal surface, the verti-
cal dipole field of the metal surface will polarize the insulating single layer,
this polarization will decrease the work function Φ by ∆Φs = e0 Na · pind by
screening Fig.2c. The induced dipole pind is proportional to the out of plane
electric field of the surface E⊥ : pind = α · E⊥ , with polarizability α. The
electric field has a strong vertical distance dependence of the surface dipole

10
layer, this involves a correlation with the screening induced work function
shift ∆Φs . In the case of the corrugated h-BN surface layer we will expect a
different local work function for a ”wire” and a ”hole” site, due to the height
difference. Hence, from the Smoluchowski effect [13], the corrugation also
induces a non-uniform surface charge density and thus lateral electric fields.
In the case of the h-BN nanomesh the two different electro negativities of
boron and nitride cause a local charge transfer from the boron to the nitride
atoms. If the h-BN is on a metal the screening of the ionic charges increases
slightly a site dependent charge transfer, resulting in a net charge displace-
ment of 0.06e− per atom towards the substrate for the hole sites [26]. A
higher work function is expected for the wires compared to the holes, as well
as lateral electric potential variations Φholes < Φwires .

Figure 2: (a) is a simplified side view of the nanomesh, which show the
dislocation of the 13 boron or nitride atoms on top of the 12 rhodium atoms.
Figure (b) is a DFT calculation done by Laskowski et al. [27], it shows the
top view of the structure and denotes the corrugation height with respect to
the rhodium surface by the color scale. Figure (c) is a schematic view of the
energy levels above the nanomesh. The levels show a variation, which has
its minimum within the region of the holes. This level variation is expected
to be extended far above ∆z ∼ 10nm the surface of the nanomesh.

11
3 Experimental
The atomic force microscope (AFM) [10] can be operated in several modes,
depending on the the cantilever dynamic and the imaging mode. The can-
tilever dynamic is separated in two classes the static and the dynamic. Static
deflection AFM or nc-AFM uses the static bending of the cantilever prong
for force detection. In the dynamic mode of operation, the cantilever prong
gets actuated in the fundamental flexural resonance f1st , here the force in-
teraction is measured by the shift in the resonance frequency ∆f . For the
measurements shown in this work, the non contact dynamic mode is used
and is referred as non contact dynamic force microscopy (DFM).

3.1 Dynamic force microscope


All measurements were made with an in house designed beam deflection
DFM Fig.3 [16] operating at room temperature under ultrahigh vacuum
(UHV) conditions (pAnalysis < 1 · 10−10 mbar). The advantage of the DFM-
system lies in the detection bandwidth of ∼ 3M Hz. This allows simul-
taneous excitation of multiple cantilever resonances with a large separa-
tion in their frequencies. Signal readout is done with two NANONIS PLL
(Phase-Lock-Loop) controllers, a Real time-Computer (RT-Computer) and
a NANONIS SPM Controller realised in a GUI software.

3.2 Cantilever preparation


Here, highly n+ doped silicon cantilever from Nanosensors PPP-NCL are
used. The Cantilever is l ≈ 200µm long, w ≈ 40µm width and t ≈ 7µm
thick. On top of the cantilever prong is a conical shaped tip with an height
of h ≈ 20µm. The cantilever support is actuated with an piezoelectric
crystal. The typical fundamental resonance frequency is in the range of
f1st ≈ 150 ± 10kHz with a quality factor of Q1st ∼ 30000. The second flex-
ural resonance frequency lies at f2nd ≈ 970 ± 20kHz at Q2nd ∼ 15000. In
order to use these cantilevers under UHV conditions, the cantilevers have to
be glued on Omicron Nanotechnology holders. Therefore, the support of the
cantilever is aligned on the holder and glued with a two component Epotek
glue and the 4-point-glueing-technique, where the cantilever is glued only at
its four edges on the cantilever holder plate. This technique has achieved
the highest Q-values for the first and second resonance frequency.
The glue must be baked in an oven at T ≈ 393K for ∆t ≈ 120min to harden.
The cantilever with the holder is then introduced into the UHV-chamber for
further preparation. Inside the vacuum, the cantilever is baked again at
T ≈ 393K to evaporate the absorbed water. When the preparation chamber
reaches normal pressure values of pprep = (4 − 7) · 10−10 mbar the baking is
stopped. After the cool down to T = RT the cantilever is sputtered with ar-

12
Figure 3: scheme of a ”beam deflection” Atomic Force Microscope

gon ions to remove remaining adsorbates and the silicon dioxide(SiO2 ) layer.
For the argon sputtering a filament current of I ≈ 10mA is set for ionizing
the argon and applying a bias voltage of Vsputter = 680V to accelerate the
ions to the cantilever. The procedure takes t = 1min, with a longer sputter
process, double or blunt tips become more likely.

3.3 Preparation of Rhodium(111) films


The rhodium thin film has been grown epitaxial on a sapphire crystal (Al2 O3 ).
This crystal was orientated in (0001) direction to have a hexagonal surface
structure. The lattice mismatch of rhodium(111) can be neglected, hence a
rhodium film with a thickness of 150nm has no remaining stress, strain and
is nearly defect free. Rhodium(111) has a hexagonal surface structure, due
to the growth mode on sapphire.
The thin rhodium film has to be cleaned by many cleaning cycles, because
of the contamination of C, O, H2 O. On the other hand, are the rhodium
terraces broadened by these cycles. One cycle consists of:

• Argon sputtering at room temperature T = RT , sputter pressure of


pprep = 3 · 10−6 mbar, sputtering voltage Vsputter = 800V for about
∆t1 = 45min. This surface bombardment breaks up the surface.

13
• The rhodium is heated with a ramp ∆T% = 60K/min to T1 = 1023K.

• Temperature T1 is held for ∆t2 = 60min. At this temperature con-


taminations diffuse toward the surface and the thin film forms large
terraces.

• The sample is cooled down to T2 = 723K with ∆T& = 20K/min.

• At the temperature T2 the surface is exposed to oxygen O2 at a chamber


pressure of pprep = 4 · 10−8 mbar for ∆t3 = 10min. The temperature
ensures a sufficient initial energy, for oxidation the contaminations, to
form clean terraces of 100nm width.

For the rhodium around 10 − 15 cycles were performed. A check was done
with a DFM measurement. After the preparation of the rhodium, we have
set up the borazine (B3 H6 N3 ) to evaporate it and form the h-BN nanomesh.

3.4 Nanomesh preparation


In this work all measurements were done on a rhodium(111) layer covered
by a monolayer of boron nitride, called the h-BN nanomesh. To form the h-
BN nanomesh on the clean Rh(111) surface, borazine is used as a precursor
molecule.
Borazine is a cyclic inorganic compound with the molecular formula B3 H6 N3 .
The units BH and N H alternate in the compound. Borazine is a special
compound, because it is isoelectronic and isostructural with benzene C6 H6 .
Hence it is called the ”inorganic benzene”. The borazine synthesis was re-
ported first in 1926 [28]. It is a reaction with diborane B2 H6 and ammonia
N H3 in the ratio 1 : 2 at 520 − 570K.

3B2 H6 + 6N H3 → 2B3 H6 N3 + 12H2


Borazine is a colourless liquid with an aromatic smell. It is less toxic, thermal
stable, but can decompose to boric acid, ammonia and hydrogen with water.
The bond length of the six B − N bonds is 0.1436nm. And it has a partial
delocalisation of the nitrogen lone pair electrons.
Applications for borazine and the derivatives are potential precursors to
boron nitride ceramics. It is also a starting material for other potential
ceramics, as boron nitride ceramics. Borazine is used as a precursor to grow
thin films on transition metals, to form structures like the nanomesh on
rhodium Rh(111) [15].
The borazine needs to be clean before it can be evaporated. Therefor it need
3 − 4 repetitions of a short cleaning cycle:

• The borazine is frozen with liquid nitrogen.

14
• It should melt slowly with low pressure pchamber ∼ 10−7 mbar. Con-
taminations melt and evaporate faster than the borazine itself, due to
the low pressure any dirt is transfered over the load-lock out of the
system.

• Before the borazine is melted completely, the valve between load-lock


and borazine chamber is closed, to avoid pumping the clean borazine.

After the cleaning cycles the borazine is ready to be evaporated on the


rhodium surface. Here, the frozen borazine is melted and then evaporated
into a small volume to get a saturated gas phase between the borazine cru-
cible and the needle valve leading to the preparation chamber. The borazine
valve should be closed again and the borazine cool back to T = 266K to avoid
decomposing. Meanwhile the rhodium is heated to Tdeposition = 1023K.
The borazine is deposited at a pressure of pdeposition = 3 · 10−7 mbar for
∆tdeposition = 2min. At the deposition rate of 10th of monolayers per sec-
onds the whole sample should be covered within the deposition time.
The hot rhodium surface has an initial energy of Einit ∼ 0.13eV to split
off the hydrogen of the borazine, which forms a new bond to a neighboring
boron nitride ring. Those boron nitride molecules form the regular, hexago-
nal nanomesh.

15
4 Measurements
Measurements on the nanomesh are complicated. Stable tip conditions are
hard to achieve. Changes in the topography and the pattern of the structure
of the acquired data are often seen. These changes are mainly random and
not reproducible. First measurements with low resolution on wide scan areas
show the hexagonal honeycomb structure as the theory predicted see Fig.4.
The observed periodicity of the structure is 3.2±0.1nm this is verified by the
line scans Fig.4b and 2D FFT Fig.4c. The 2D FFT confirms the hexagonal
structure visible in the 3 fold symmetry of the spots. Line scans show a
corrugation amplitude of 45±8pm, here again, we can confirm the theoretical
predictions, with a calculated corrugation amplitude of 50pm.

Figure 4: Figure (a) shows a zoom of the nanomesh of a larger scan area.
With wires ”bright” and holes ”darker”. Even in this small scan area insta-
bilities were found. Figure (b) shows a line profile of the topography (see a),
the maxima’s of the profile have a periodicity of the predicted 3.2 ± 0.1nm.
Also the corrugation amplitude of 45 ± 8pm is in agreement with the the-
ory Fig.:2b. Figure (c) is a 2D Fast Fourier Transformation (FFT) of the
topography. The position, distance and brightness of the spots denote the
direction of the wave vector k, wavelength of the of the k-vector and the
amplitude, with corresponds to the regularity of the nanomesh. The wave-
length of the k-vector is the inverse of the periodicity, it verifies the peri-
odicity measured by the line profile. The FFT spots shows a 3 fold sym-
metry, which confirms existence of a hexagonal structure. Image parameter:
Scan area 150x150nm2 , f1st = 153019Hz, Q1st = 33318, k1st = 25.1N/m,
∆f = −149Hz, A1st = 5nm, γ = 8.63f N m0.5 .

This measurement verifies the structural prediction, but shows a rather low
resolution at large frequency shifts of ∆f = −149Hz and large gamma fac-
tors γ = 8.63f N m0.5 .
The shown topography Fig.:4a is predicted, but not always seen. We mea-
sure three patterns and contrasts. First, the normal, predicted topography
with low resolution. Second, a topography with an inverted contrast but low
resolution. Third, a full inversion of the contrast of the topography, occuring

16
mainly with an improvement of the resolution. A detailed discussion on the
contrast inversion is given in chapter 4.2.

4.1 Mesh stability


The borazine precursor molecule is evaporated on the rhodium at Tdeposition =
1023K (see Section:3.4) for growing a regular nanomesh. This denotes a high
thermal stability of the nanomesh under UHV conditions. Due to a failed
deposition treatment of truxene molecules, the sample had to be cleaned.
Starting an annealing cycle at T = 1023K for t = 90min, no changes in the
coverage of the h-BN nanomesh could be observed, also longer annealing cy-
cles for t = 150min showed no changes. So annealing above the deposition
temperature and an increasing of the temperature in steps of ∆T = 25K
until a change in the coverage can be seen, was the logical step.

Temperature in K Annealing time in min


1023 90
1023 90
1023 150
1048 75
1048 90
1073 120
1098 90
1123 90

Table 1: Table of annealing temperatures and annealing times for cleaning


the nanomesh.

Finally, the nanomesh was annealed at T = 1123K for t = 90min. After


the cleaning cycle, small amounts of adsorbates could be registered, which
are not distinguishable, if they remain from the truxene deposition or from
remnant gases in the UHV.
Nevertheless, the measurements show still the regular pattern of the honey-
comb superstructure. No signs of damages, disordering or evaporation sites
of the h-BN nanomesh were visible. So, the thermal stability for a nanomesh
on a rhodium(111) substrate can be assumed to be even higher.
Beside the thermal stability the mesh shows also high stability at ambi-
ent conditions as well as a distinct longterm stability. Since the prepara-
tion of the nanomesh until the most recent measurements, the nanomesh
showed none significant change within on year, which can be clearly lead
back to changes in the superstructure. This is quite remarkable, because the
nanomesh is only a single layer with no further treatment for stabilizing the
structure.

17
4.2 Contrast inversion
The measured data is visualized by a color distribution, the global maxima
sets the upper end of the color range, while the global minima the lower.
This method allows to visualize an acquired data signal, like the frequency
shift, for better recognition. The contrast is then related as the change in the
color scale between two neighboring points. The corresponding change can
be positive, meaning a change from the lower color towards a higher color
from the defined range, or negative vice versa. If the change is large the
corresponding contrast makes these points more distinguishable from each
other. In an acquired data set, which is aligned regular picture like shape, we
recognise this set of data as a ”topographical” information, if we consider the
signal of the feedback loop of the z-tube piezo in an atomic force microscope,
the acquired data set represents the ”topographical height” information.
An inversion of the contrast relation between two neighboring points shows
a picture, which has the same shape and pattern, but reflects an inverted
”topographical” information. This inversion and its origins are showed and
investigated.
We measure not only the predicted contrast as mentioned before (see Fig.:4),
namely wire-hole contrast. Our measurements show over all, three clearly
distinguishable contrasts in the topography. The normal predicted contrast
and inverted contrasts, which we call the low resolution inversion and high
resolution inversion, shown below.

Figure 5: In the figures (a-c) we see from left to right the normal contrast
as expected (a). The inverted contrast with low resolution (b) and the high
resolution inversion (c).

18
Figure (a) Figure (b) Figure (c)
Scan frame in nm2 1502 2502 752
f1st in Hz 153019 153153 157502
Q1st 33318 32916 33544
k1st in N/m 25.1 25.1 27.3
∆f in Hz −149 −8.5 −10.7
A1st in nm 5 5 10
γ factor in f N m0.5 −8.63 −0.49 −1.86

Table 2: Table of image parameters of Figure 5.

In every image a hexagonal pattern related to the honeycomb structure is


shown. With a periodicity being within small errors 3.2 ± 0.2nm. But one of
the main difference is the corrugation amplitude, which is always enhanced
in the inversion by a factor of 2 to 3. The relative corrugation amplitude of
the low contrast inversion was measured to be 110 ± 15pm, while the high
contrast inversion shows a corrugation of 130 ± 20pm. Another difference
is seen in the high resolution inversion, in the line profile as well as in the
image. Here, it stands out that the z signal has a ”plateau” with a small
”dip” in the middle of an hole, which is less elevated. The other contrasts
show no indication of a plateau, neither in the normal contrast, nor in the
low resolution inversion.
In comparison with the DFT calculations of Laskowski et al. [27] the nanomesh
should show a flat area inside a hole. But in our case a really faint ”dip” is
recognizable in the middle of a hole.

4.2.1 Tip apex changes


A standard non contact cantilever has a conical tip of ∼ 20µm height. At
the top of the cone, we consider a half sphere with a radius of ∼ 10nm. The
front most atom of the tip, which has the closest distance to the surface, has
the highest contribution in the frequency shift at the lowest turning point
of a cantilever oscillation. A change in the conformation of the front most
atoms, towards a tip shape with more atoms having the same closest distance
to the surface, can change the frequency shift drastically.
It has been shown that tip apex changes in combination with polarity changes
can produce contrast inversions. This was shown in the studies of Enevoldson
et al. [34] about T iO2 (110) surfaces in 2008. Studies of our measurements
show often tip apex changes and corresponding height changes of the z piezo
tube. No additional offset or enhancement of the signal with respect to the
noise level could be measured.

19
4.2.2 Double tip
On the other hand a double tip might be induced during tip preparation or
scanning.
If two tips measure with a single atom at the same distance above the sur-
face at different lateral positions, then the acquired data is a superposition
of each topography measured by those tips. In addition, every topography
is a convolution of the surface structure with the tip apex measuring above.
Double tips occur most likely with a dominant tip and a side tip. The dif-
ference is the distance towards the surface, with the dominant tip having a
smaller tip surface distance than the side tip. The side tip influence is seen
at step edges, because the same step edge with the same shape is measured
twice.
Considering a double tip measuring above a corrugated surface at the same
tip surface distance for both tips. The regular surface pattern is superim-
posed with itself, the corrugation amplitude should change, due to inference
effects. But interference effects should affect the measured surface pattern
as well and even an inversion would be possible.

Figure 6: In this figure a scheme of a double tip is shown with the x condition
Eq.4.3. Both tips have the same tip-sample distance zc at the lower turning
point.

To prove the double tip theory and the assumption of a superposition of the
surface structure, a function for a hexagonal structure also used for HOPG
[35] was used. Taking into account the periodicity a = 3.2nm and the

20
corrugation amplitude ca = 0.05nm, we can write the function:

      
ca 2π 2π 4∗π
fnanomesh (x, y) = − 2 cos · x cos √ · y + cos √ ·y
4.5 a a 3 a 3
(4.1)

Figure 7: Figure (a) shows the nanomesh topography measured with a single
tip. Figure (b) shows the the nanomesh generated by the function 4.1. The
generated nanomesh agrees nicely with the measurement.

The structural simulations were done by Matlab [36], by a small self made
program, see the appendix. These simulations are only mathematical func-
tions, which fit the structure. They don’t simulate the interaction forces or
take the shape of the tip into account. But Figs. 7 (a,b) are in agreement
with each other. Meaning, that the function is adaptable for the superposi-
tion of a second tip.
This function is then superimposed with itself, to simulate the influence of
the second tip, additional variables z1 , z2 where inserted to vary the influ-
ence of each tip. The range is z1,2 ∈ [0, 1], 0 for no influence, 1 maximal
influence.

fdoubletip (x1 , y1 , x2 , y2 , z1 , z2 ) =
      
ca 2π 2π 4∗π
−z1 · 2 cos · x1 cos √ · y1 + cos √ · y1
4.5 a a 3 a 3 (4.2)
      
ca 2π 2π 4∗π
−z2 · 2 cos · x2 cos √ · y2 + cos √ · y2
4.5 a a 3 a 3
x1,2 , y1,2 denote the positions of tip2 in relation to tip1 :
   
1 1
x2 = x1 + n ± · a~x y2 = y1 + m ± · a~y (4.3)
2 3

21
a~x,y denote the displacement vector in x and y direction with a corresponding
vector length of 3.2nm, the nanomesh periodicity. The condition in equation
4.3 is for a complete inversion of the measured topography. There are more
conditions in the measurements than the inversion condition, two other tip
conditions are shown in figure 8.

Figure 8: Figures 1-3 show for column (a) the nanomesh topography mea-
surements. Column (b) shows the the nanomesh generated by the function
4.1 with the special conditions for the second tip shown in column (c). The
generated nanomeshes are in agreement with the different measurements.

The measurements and calculations in Fig.8 give certain hints for double
tip conditions during scanning. The calculated ”topographies” match with
the measurements and explain the seen contrast. But the calculation could
not reproduce the high resolution inversion. The resolution of the calculated

22
Figure 1 Figure 2 Figure 3
Scan frame in nm2 502 502 752
f1st in Hz 154797 154797 157502
Q1st 33957 33957 33544
k1st in N/m 26.0 26.0 27.3
∆f in Hz −75 −55 −10.7
A1st in nm 6 6 10
γ factor in f N m0.5 −5.85 −4.29 −1.86

Table 3: Table of image parameters of Figure 8.

”topographies” is limited by the structural resolution of the function 4.1.


Another important feature of the double tip inversion is the position of the
holes, when switching contrast. Due to the superposition of two structural
functions the holes shift their position during the contrast change. A hint
for this feature, was given by a previously measured 3D spectroscopy of S.
Koch.
A 3D grid spectroscopy, for a small scan area, represents force distance spec-
troscopy curves at every grid point. Composing all spectroscopy curves to-
gether a 3 dimensional data cuboid is given, containing 3D frequency shift
information. A better visibility of the 3D data set, is provided by cross sec-
tion through isodistances of the force distance spectroscopy curves. Within
these cross sections frequency shift maps are included, allowing comparison
of different sites.
In this grid spectroscopy the thermal drift was not compensated and induced
tip changes, visible in the first third of the frequency shift maps Fig. 9.
Figs. 9a, c show the 3D spectroscopy cross sections at different tip sample
distances. The upper one has a larger tip sample distance than the lower. All
figures show measured or calculated frequency shifts with a negative contrast
pattern, because the frequency shift is negative proportional to the z-signal.
The upper figures show a frequency shift map with bright holes, for large
tip sample distances. At one hole, marked with a red dot, a calculation was
done by using the negative of equation 4.1. The calculated frequency map
was tilted to match the measurements. Again a bright hole was marked for
site comparison during the contrast switching.
The lower figures 9c, d show a contrast inversion in the frequency shift map.
The red dot of the measured frequency shift map has the same position, but
marks a site of a wire crossing. The figure 9d was calculated by the negative
of equation 4.2 with a double tip influence. The red dot marks the same spa-
tial position a above, but denotes a wire crossing site, too. Calculations are
in accordance with the measurements. A comparison inside every column, a
change of the hole sites during contrast change is visible.
With the upper formalism 4.2 and relation 4.3, the 3D grid spectroscopy

23
Figure 9: Figures (a, c) show layers of the frequency shift map of a 3D spec-
troscopy at different tip sample distances. Frequency shift map of Fig.:(a)
has a larger tip sample distance than the lower. Figures (b, d) show the
expected pattern calculated by the negative of Eg.:4.1. The 3D spectroscopy
showed an inversion of the wire-hole-contrast. A red dot marks the spot of a
hole in the frequency shift map of Fig.:(a). In the lower frequency shift map
Fig.:(c) the same red dot is on top of a wire. A simulation with the equation
4.2 with relation 4.3 showed the same behavior.

measurements as well as the other seen low resolution contrast could be


explained. Therefore, it can be concluded that the low resolution contrast
inversion (Fig.5b) is due to a double tip feature: superimposing two topogra-
phies, measured at different sites.

24
4.2.3 Tip geometry
The double tip induced pattern changes or even the contrast inversion brings
up the question of representativity and reproducibility. We cannot reproduce
artificially the patterns of figures 8, which show that, these happen by chance,
but they can be observed regularly. These patterns are seen with almost
every tip at some point, so that they should have a common feature.
In addition, a consistency with the mentioned theorem Eq.:4.2 is needed. The
properties for a ”good” AFM tip are firstly a small opening angle θ ∼ 15◦ , to
separate the interaction forces, to be more dependent on short range forces
and secondly the atomically sharp tip apex to resolve single atoms or small
defects. If a tip becomes blunt, meaning the tip apex is not atomically sharp
and can be assumed by a tip cone with a tip apex curvature, which is larger
than an atom, then the resolution will be limited by the curvature. A large
curvature results in a large force averaging and features smaller than the
curvature are not resolvable.
In our case, the tip cone with a low opening radius must be similar or smaller
than the periodicity of the nanomesh, otherwise it would not be possible
to resolve the wires or holes, meaning rconeopening = periodicity/2. But
the tip apex curvature should be large, to see the force averaging effect.
So a tip apex curvature, which is defined by a hole curvature is assumed.
With a periodicity of 3.2nm and a corrugation of 0.05nm, an arc, which
intersects with the left wire (−1.6, 0.05), the center (0, 0) and the right wire
(1.6, 0.05) can be defined. This results in an arc radius of rarc = 25.625nm
[47], which is also assumed for the tip apex curvature. An atom at the
tip apex could image the real topography, but would be accompanied by
an averaged background. A second dominant atom at the tip apex with
condition mentioned in equation 4.3 could lead to a stable superposition of
the topography.
The stability during the measurement, is then defined by a stable position of
the atoms to each other. If an atom change its position due to an influence
of an adsorbate or a step edge, the acquired pattern will change. A possible
way to distinguish a mono atomic tip or a double tip are step edges. A
double tip influences and broadens the measured path across an step edge.
In Scanning Probe Microscopy (SPM) step edges are imaged with the con-
volution of the tip geometry [14]. In the acquired data, step edges appear
broaden and smeared out. The transition distance between upper and lower
terraces can determine the tip apex geometry [37]. If the transition distance
is ”small”, then the tip apex has a small curvature and thus has small influ-
ence on the convolution between topography and the tip apex in the acquired
data. A single atomic tip resolves the actual topography the best.
In our modeled case, two atoms are assumed, each resolves the topography
correctly. The superposition of the topography will result in a broadening
of the transition distance of a step edge. The minimum transition distance

25
Figure 10: The Figure shows a schematic view of the assumed double tip.
The tip cone has an opening angle of ∼ 15◦ and at the lower cone end the
opening radius is modeled to ∼ 1.6nm, the apex curvature is r = 25nm.
Two atoms are supposed at the tip cone end, which measure the dominant
force contribution, with atom position relation 4.3 in respect to each other,
a superposition described by the figures 8 can be expected.

will be limited by the separation distance between those two atoms, the
tip cone opening radius rcone and the tip apex curvature. Assuming a con-
stant tip apex curvature as well as the tip cone opening radius, the tran-
sition distance has dominant influence by the measuring atoms, see Fig.10.
From the simple schematic point of view the transition distance for a single
atomic tip, which is placed in the center of the tip apex, should be in the
region of ∆strans ≤ 1.6. For a tip with two atoms, the separation in be-
tween adds up, but is maximal limited by the cone diameter plus two times
an atomic radius, because we are sensitive to the out of plane force com-
ponents. Meaning that the transition distance for two dominant atoms is
∼ 1.6nm ≤ ∆stransistion ≤∼ 3.2nm.
Figure 11 shows a pattern and contrast change of the topography. The seen
contrast will be discussed in the next section 4.2.4. The slow scan direction
was bottom up and the change in the image happened by chance. The
cross sections are perpendicular to the step edges to measure the acquired
transition distance, without artifacts. Both cross sections are parallel with
respect to each other, allowing a direct comparison of the data, due to the
same relation of slow and fast scan direction components. We suppose,
that the upper cross section shows the topography measured by a single
tip. We measure the transition distance of a step edge to ∆strans = 1.39 ±
0.07nm. This distance is an evidence, that the tip apex has broaden and
is decorated by an single dominant atom. The lower cross section shows a

26
Figure 11: The Figure 11a shows the topography, which shows a change
in the patter as well as in contrast. Two cross sections are taken across
step edges. The upper cross section shows the topography measured by
one tip. The lower cross section shows the topography measured, we as-
sume, by a double tip, see Fig. 11b. The step edge transitions distance
is measured for both cross sections. The transition is broadened for every
step of the lower cross section. Image parameters: Scan area 50x50nm2 ,
f1st = 154793Hz, Q1st = 33000, k1st = 26N/m, ∆f = −75Hz, A1st = 6nm,
γ = −5.84f N m0.5 .

transition distance, which is enhanced by almost one nanometer, ∆strans =


2.33 ± 0.20nm this holds with our simple assumption of an second dominant
atom at the tip apex with a broaden tip apex curvature, which lead to an
stable superposition of the acquired data. From the comparison of both
cross sections we can propose, that the topography in the upper part has
less spatial averaging of the forces, while the lower image part gives evidence
to a large spatial averaging, which reduces the corrugation amplitude of the
wire hole pattern. Even with a different acquired contrast pattern of the
topography, both cross sections seems to acquire the step height correctly.
The height difference between two terraces in both contrast is measured to
be ∆h = 230 ± 20pm, this is in correspondence to the surface relaxation of
rhodium in the (111)-direction, which was measured by low energy electron
diffraction (LEED) dbulk = 219pm [38].
The tip apex model we derive is capable to explain the measured pattern,
which leads to a low resolution acquisition of the topography. The stability
of the tip apex configuration we discussed in detail is consistent with the
data. We think the low resolution contrast inversion is a tip apex induced
feature. The High resolution contrast inversion, from this point of view, is

27
not explainable by the mentioned theory.

28
4.2.4 Height dependent contrast inversion
The hole wire contrast seen in Fig.5c is characterised by high resolution,
reproducibility and stability. Still, it is not clear whether it is a real structural
change, a surface feature or if it is a tip induced measurement artefact.
The investigation was difficult, because a second 3D grid spectroscopy was
no good solution, due to the lack of thermal drift compensation and tip
instabilities. Normal frequency modulated DFM topography measurements
seemed to be the best solution, due to the adequate acquiring time, thus
low thermal drift and high fidelity of the tip conditions. But a topography
measurement on a clean surface at a fixed frequency shift set point gives
no indication on small hole-wire site changes, during a contrast inversion.
Therefore, a position marker was needed, for a clear evaluation of hole-wire
sites. A simple solution for position markers are random adsorbates aligned
in a recognisable way.

Figure 12: Topography maps at the exact same position are shown, visible
by three random adsorbates at the corners. The frequency shift set point was
increased slowly from left to right dfa,b,c = {−10.5Hz, −12.5Hz, −15.5Hz}.
The measurements shows a contrast inversion for a lower frequency set point.
Line profiles at the same position, recognisable by the markers, show the
corrugation amplitude behavior and the maxima/minima position for the
nanomesh. Figure (b) seems like an ”intermediate” state, here the wires and
the middle of a hole is lower and the rimes of the nanomesh are highlighted
even the corrugation amplitude decreases for the ”intermediate” state. An
important fact of these topography measurements is the not changing posi-
tion of holes, during the high resolution contrast inversion.

Figures 12(a-c) show topography maps of the same position with increased
frequency shift set point. From left to right a clear contrast inversion of
the nanomesh is visible. Figure 12 b seems to show an ”intermediate” state

29
with enhanced rims and low wires and hole centers. The markers fix the
position of the line profile for a comparison. First visible difference of these
line profiles is the corrugation amplitude. The ”intermediate” state has a
reduced corrugation amplitude by a factor of ∼ 2 in comparison to normal
and inverted contrast. Secondly, the profile structure changed in these three
measurements. The normal contrast shows a ”sinusoidal” profile curve, while
the ”intermediate” state as well as the inverted state shows additional peaks.
All line profiles were recorded a the same position judged by the three mark-
ers. This allows clear comparison of hole-wire positions. In the line profile
of Fig. 12a a maxima was marked at a fixed position. In figures 12(b-c) the
same position marker denotes a minima of the measured topography. This
shows clearly, that holes and wires does not change their position during
a high resolution contrast inversion. In conclusion, the high resolution in-
version is a real surface property and is not a measurement artefact due to
multiple tips. Any superposition of multiple nanomeshes should change the
position of holes and wires.
If the high resolution contrast inversion is height dependent, a 2D line spec-
troscopy should show a height dependent variation in the force fields. Figure
12 denote a rather intense dependence ranging within δf = 5Hz between
normal contrast and inverted contrast. Recalling the nanomesh theory, fig-
ure 2c shows a shift in energy levels by virtue of a non uniform electronic
distribution [15]. This enforces the expectation of spatial variation of forces
above the surface.

4.2.5 2D spectroscopy
The z-spectroscopy correlates one point of the surface (x, y) with a large
number of measuring points in the surface normal z. Spectroscopy gives
insight in the forces above a certain point, with a high vertical resolution.
Here, 2D spectroscopy was used.
Theory predicted vertical as well as lateral force deviation within the range of
about a few nN or in the frequency shift ∆f of some Hertz. These deviations
are supposed to be measurable and extend to several nanometer above the
surface into the vacuum [15].
The 2D spectroscopy was performed with the second flexural resonance f2nd ,
this enhances the short range force sensitivity, due to the measurement of
the force gradient Eq.:2.17 [20]. The second flexural amplitude was set to
A2nd = 400pm to have a enhanced sensitivity and to ensure an appropriate
signal to noise ratio. In the first spectroscopy measurements only the second
flexural resonance was excited. First we wanted to acquire a set of data by
monomodal measurement, to compare this with bimodal spectroscopy data,
for additional features and cross talk relations.
The 2D spectroscopy was recorded at the path of the cross section seen in
Fig. 13f. The acquired spectroscopy data was evaluated and further com-

30
Figure 13: (a) frequency shift of the second flexural resonance f2nd in de-
pendence of z. The average frequency shift of the line was subtracted 13b.
The force is calculated by the Sader algorithm 13c and the same spec-
troscopy curve is plotted after post processing. A force average of each
line is then subtracted 13d. The acquired excitation was converted to the
excitation per cycle 13e. Figure 13f shows the cross section of the 2D spec-
troscopy before the measurements. Spectroscopy parameters: Line distance
4.372nm, sweep distance ∆z = 5.0nm, VCP D = 0.304mV , f2nd = 967594Hz,
Q2nd = 13099, k2nd = 1701.0N/m, A2nd = 400pm. Image parameter: Scan
area 10nmx10nm, f2nd = 967594Hz, Q2nd = 13099, k2nd = 1701.0N/m,
∆f = −230Hz, A2nd = 400pm, γ = −0.29f N m0.5 .

31
puted by a self written MathWorks Matlab program [36] (see. appendix
B). The data is correctly aligned by a force gradient correction of the long
range tail of the spectroscopy curves. A manual drift correction was not
applied, because the gradient correction error is indistinguishable from the
cross section error. The fully processed data was smoothed vertically and
horizontally by an Svitzky-Golay filtering just before plotting. This reduces
the measurement and post processing noise in the data.
One frequency shift curve after post processing is shown beside figure 13a.
A long range spectroscopy curve of a sweep distance of ∆z = 20nm was
added to each spectroscopy curve to aligned each towards zero frequency
shift, meaning the cantilever is at its free oscillation with no perturbation.
The post processed frequency shift data should match the derivative of the
Lennard Jones potential Fig.1. The spectroscopy curves end before the
equilibrium bond distance is reached, to avoid tip instabilities or tip apex
changes.
The subtraction of the average frequency shift at distance z in each line
enhances the visibility of small deviations. The resulting figure denotes the
change in frequency shift for a ”quasi” constant height measurement. Figure
13b indicates clearly long range frequency shift deviations above wire sites,
which extend more than ∼ 3nm above the surface. The periodicity of the
features coincident with the nanomesh periodicity. The deviations from the
mean value increase drastically, if the tip sample distance becomes small,
∆ztip ≤ 1nm and the actual surface topography becomes visible and domi-
nant in the acquired data set.
In addition figure 13b shows some feature at the end of every spectroscopy
curve bélow z-distances of z ≤ 80pm. The frequency shift deviation from
the average line value switch their signs abruptly, within a small distance
∆z ∼ 50 ± 20pm. This feature will be discusses later in detail 4.2.6.
Figure 13c shows corresponding to the frequency shift the calculated force.
The calculation of the force includes the 20nm long range tail. This was
added to the spectroscopy to apply the Saders algorithm [39]. A single force
curve at the same position as the frequency shift curve was picked for clear
visibility. This force distance curve reaches a peak of F2nd = −73.02nN .
The force distance curve should be related to the Lennard Jones potential,
Fig.:1, which shows in comparison to the frequency shift curve a smoother
decline, due to the dominance of the integral tail.
After subtraction of the average force at distance z in a line, the force de-
viations at a ”quasi” constant height become enhanced. The force difference
of the wire to hole sites at a fixed distance is clearly visible. The force de-
viations show again the periodicity of the nanomesh. The variance in the
force deviation rises presumably from the averaging effect of the integration
and the topography irregularity. In comparison with figure 13b no change
in the force deviation is visible close to in the region below z ≤ 80pm. This
feature is presumably hidden, due to the dominance in the frequency shift

32
integration.
Simultaneously to the frequency shift measurements the excitation voltage
for the piezo actuation was recorded. This data set was computed to the ex-
citation per a single oscillation cycle by the conversion formula of Giessible,
see [18]. The excitation per cycle is related to the dissipation and the energy
loss of the cantilever-sample system. On the other hand, a tip apex deforma-
tion would be visible in the actuation energy. The excitation data coincident
with the recorded frequency shift. A larger actuation energy is needed on
top of wire sites. From the excitation data, we have no visible evidence for
a tip induced inversion of the frequency shift in the sweep distances below
z ≤ 80pm. Presumably, it can not be seen in the excitation per cycle, due
to the point, that the interaction force acting on the cantilever during one
oscillation, changes the frequency shift and the amplitude. The amplitude
changes in correspondence to the force, while the frequency shift is related
to the force gradient 2.17 [20]. The force does not show this small variations
close to the surface, due to that the excitation force exciting the cantilever
to a constant amplitude does not show this variations, too.
The detailed analysis of the inversion in the frequency shift given below 4.2.6,
but from the excitation energy per cycle it becomes unlikely that this inver-
sion lasts from tip induced effects, like double tips or any chemical bondings
towards the substrate, because of the relative small changes in the actuation
energy of ∆E = 0.13 ± 0.01eV .

4.2.6 Detailed spectroscopy analysis


The observed inversion Fig. 13b was neglected in the first place, because
of uncertainty in data post processing. The data seem to show an smooth-
ing error or some artifacts from the applied gradient correction. The post
processing code was checked and possible errors were tested. It can be con-
cluded that the inversion is not introduced by the data post processing of
the Matlab program [36].
In the frequency shift plot Fig. 14, iso lines with the same frequency shift
have been plotted at the frequencies:

∆fisolines = {−230, −200, −170, −140, −110, −80, −50, −20}

The plotted contour lines in the frequency shift data Fig.14a show for ∆f2nd =
−20Hz at distance of z = 2.15nm only small deviations, which seem to be
related to the theoretical expected topography. Above this contour line the
frequency shift show a smooth homogeneous decrease to ∆f2nd = 0Hz. Be-
low the contour line the frequency shift deviations show the theoretical to-
pography more pronounced. For ”tip-sample” distances between z = 0.25nm
and z = 0.5nm the contour lines show the largest corrugationamplitude in
their deviations, the maxima and minima of the deviations have the predicted
nanomesh periodicity and show a surface corrugation within the predicted

33
Figure 14: Contour lines of constant frequency shifts were added to the
acquired frequency shift data of the spectroscopy, Fig.: 14a. The vertical
white lines denote the positions of the distance - frequency shift curves.
With the left blue curve for a hole and the right red curve for a wire cross
section. Figure 14b is a zoom of the acquired frequency shift map with
contour lines. The contour line for ∆f2nd = −230Hz is marked, because
it shows a inverted curvature, in spite of the other contour lines. The two
cross sections are plotted together on a 5nm scale, the interval between 0
and 0.5nm is magnified to show a crossing in the frequency shift curves at
the sweep distance of z ∼ 80pm.

34
limits. For ”tip-sample” distances, below z ≤ 0.25nm the amplitude of the
deviations in the frequency shift decreases and show even an inversion of the
theoretical topography for a contour line of ∆f2nd = −230Hz.
In figure 14b a detailed view of the closest region with a contour lines are
given. Here, the contour line of ∆f2nd = −230Hz is marked with an arrow.
This contour line shows an inversion of the theoretical predicted topography
in the frequency shift. Meaning, if the tip-sample distance is controlled by
a set point below ∆f2nd = −230Hz, the measured z signal or topography is
inverted.
For a detailed insight, two cross section were extracted from the frequency
shift data. Fig.14c. The blue line is a frequency shift distance curve at a
hole site and the red at a wire site. In the upper plot, the full curves are
shown. The long range tail of those curves match each other quite well,
down to a sweep distance of z = 2.5nm. The red curve, for the wire site,
has a less negative frequency shift in comparison to the hole site, blue curve.
The difference between both curves become largest for ”tip-sample” distances
of z = 0.16nm. For smaller distances the difference in the frequency shift
becomes smaller and the curves show an intersection at z ∼ 0.08nm. For
distances below this intersection the wire sites has a more negative frequency
shift than the hole site.
The actual distance at which the crossing occurs is not accurate, because
the measured frequency shift is strongly tip shape dependent. But the data
show no indication of tip induced features, as it could be estimated by the
acquired date in Fig. 13.
To distinguish a height dependent contrast inversion clearly from other ef-
fects, we can think of probable boundary conditions for such an effect. In
FM-DFM mode with small amplitudes the frequency shift is sensitive to the
force gradient. The tip apex is sensitive to a assumed surface potential with
a dominance of the Lennard-Jones potential. Any physical property, which
shifts the turning point or the slope of these potentials, would lead to a
change of the measured frequency shift on the surface at a certain tip sam-
ple distance. If we assume, that wires and holes have independent properties
and can be described by linear independent potentials with different slopes,
we could explain a height dependent contrast inversion by a transition of the
potentials at a certain frequency shift level above the surface. This could re-
sult in a distant dependent enhancement of holes compared to wires. In the
2D line spectroscopy Fig.13 and Fig.14 such a distant dependent inversion
of holes is visible.
For the difference in the frequency shift curves Fig.: 14c, a explanation can
be the non uniform electronic properties of the surface [15]. But the com-
bination of long and short range forces and the dependence of the tip apex
sets a sophisticated task. But, a procedure for measuring and post precess-
ing calculations for separation of the influences can be performed.
The tip shape or tip apex curvature can be in situ estimated by measuring

35
the tip sample distance at a fixed frequency shift for ramping the sample
voltage [40]. This method allows the evaluation, of the tip apex curvature
and the dependence of the conical part. The method should be applied be-
fore and after a spectroscopy, if the measured voltage dependence does not
show any difference, it can be assumed, that the tip apex does not change. In
comparison to a frequency shift versus distant curve the advantage of ramp-
ing the sample voltage at a larger distance is to avoid touching the surface.
For the acquired frequency shift data a analytical fit of the electrostatic forces
between the tip and surface can be extracted [41]. This fit takes the tip shape
and the tip apex into account. This values for an initial fit values can be
assumed by the bias dependence of the tip position at fix frequency shift
[40]. After the subtraction of the long range electrostatic part, the frequency
shift data should contain mainly contributions of the van der Waals force
and the chemical as well as the electrostatic short range force. The chemical
interaction can be neglected for tip sample separations larger than z ≥ 1nm.
For distances below the tip radius the curve can be fitted by the spherical tip
model Eq.2.5 [17] and the Hamaker constant can be extracted from the data.
After subtracting the van der Waals contribution of the frequency shift data
the chemical short range forces remain. For the limit, that the A1st,2nd  z
the data can be fitted by an equation Guggisberg [17] introduced. In the
measurements of Fig.:13 this equation does not apply, because we have an
amplitude A2nd , which is comparable to the decay length λ of the chemical
forces.
Even though, that the chemical forces cannot be fitted correctly, the contri-
bution of the chemical force should be dominant after post processing. This
separation method, can be a powerful tool to give a cross check of the fre-
quency shift variations of the 2D Spectroscopy curves of the nanomesh data
Fig.14.

4.2.7 2D bimodal spectroscopy


To get a detailed insight of the short range interactions a bimodal 2D spec-
troscopy was performed exciting a normal and torsional oscillation. The
second flexural resonance frequency was used for tip sample distance control
with an amplitude of A2nd = 600pm. Simultaneously, the torsional reso-
nance was excited and measured with an amplitude of AT R = 80pm. This
spectroscopy had two purposes, first, the analysis of the an expected lateral
force variation of the nanomesh [15]. Second, the influence of the torsional
excitation on the flexural 2D spectroscopy.
Figures 15 with the marked cross sections denote the position of the 2D
spectroscopy. The data was measured simultaneously by bimodal DFM.
The height control was warranted by the second frequency shift ∆f2nd , in
addition the torsional frequency shift was acquired with a high sensitivity.

36
Figure 15: (a) topography measurement controlled by the second amplitude.
(b) the simultaneously measured frequency shift of the torsional resonance.
Graph (c) denotes a line profile of the position of the 2D spectroscopy.
Image parameters figure (a): Scan area 20x20nm2 , f2nd = 967617Hz,
Q2nd = 13065, k2nd = 1701.0N/m, ∆f = −46Hz, A2nd = 600pm, γ =
−0.11f N m0.5 , fT R = 1.486589M Hz, QT R = 123253, kT R =∼ 2000N/m,
AT R = 80pm

In the recorded frequency shift the wires are visible with a fast decreasing
frequency shift on top of them (at the positions ∼ 1.5nm, ∼ 4nm, ∼ 6.75nm
and ∼ 9nm). This spectroscopy data is influenced by uncompensated ther-
mal drift in the range of ∆zthermal = 35±7pm calculated by the residual data
mismatch after post processing. Compared to the spectroscopy Fig.:13 the
bimodal spectroscopy Fig.:16 suffered from a bad tip shape and presumably
a large tip apex curvature, visible in the strong averaging of the frequency
shift ∆f2nd .
Spatial ∆f2nd variations in the range of δf = 5Hz are visible after subtrac-
tion of the mean value in each line, Fig.16b. For large tip sample distances
a spatial variation of the ∆f2nd signal is visible. On top of the wires, more
negative frequencies are recorded compared to the line average as seen before
Fig.:13. Below a tip sample separation z ≤ 0.2nm the wires and holes are
clearly distinguishable.
The calculated forces Fig.16c show a deviation, which is in correspondence
to the corrugation of the nanomesh below. More detailed information are

37
Figure 16: Figures (a-d) show the recorded frequency shift of the second
resonance (a), second frequency shift less the line average (b). From the
frequency shift the distant dependent force was calculated by the Sader
formalism (c), while the line average is subtracted in figure (d). The dis-
tance dependent excitation is shown in (e). Figure (f) shows the recorded
frequency shift of the torsional resonance. Spectroscopy parameters: Line
distance 9.203nm, sweep distance ∆z = 1.1nm, bias CP D = 0.589mV ,
f2nd = 967617Hz, Q2nd = 13065, k2nd = 1701.0N/m, A2nd = 600pm,
fT R = 1.486589M Hz, QT R = 123253, kT R =∼ 2000N/m, AT R = 80pm

obtained, if the line average of the force is subtracted Fig. 16d. On the left
side of the data plot, holes show a quite extended field with values above
the line average, marked in blue colors. In comparison to a wire site, the
values are slightly below the line average, except the case close to the sur-
face, here forces have a larger deviation from the mean value. On the right

38
side in comparison to the left side the interplay between more attractive
regimes and less attractive regimes seem to change. One possibility could be
the uncompensated thermal x − y − z drift. The Saders formalism is highly
distance dependent. If the distance between tip and sample is varied during
the spectroscopy measurement, it would show an influence in the force cal-
culation. For large tip sample distances the wires are more pronounced. The
force calculation with the subtracted line average shows this behaviour more
clear. This mentioned feature was also visible in the force calculations of the
spectroscopy above, Fig.13b. It seems that this is a systematical error, due
to z drift or large lateral x − y drift, which changes the position to be only
on top of wires.
The excitation signal shows almost no features except close to wires. In
comparison to the spectroscopy of Fig.:13e the excitation energy range has
shifted to larger values.
The last figure 16f shows the recorded frequency shift signal of the first tor-
sional resonance. Theory has predicted an additional lateral binding force
component close to the surface [15]. This bimodal spectroscopy measurement
shows the predicted property for small tip sample distances. The frequency
shift pattern corresponds to the hole wire sites, seen above. Again a small
gradient in the frequency shift is visible, from less negative shifts on the left
to more negative on the right.

39
Figure 17: A small intersection of the torsional frequency shift of Fig.16f is
shown with a projection of the surface on the x − z-plane. Two wires with
a negative frequency shift and three holes with a positive. Far away, small
frequency shift deviations are measured, which are in accordance with the
wire-hole positions. Close to the surface, the lateral variation, due to the
corrugated surface, become more visible. A doubling of the peaks or holes
evolves at close distances.

The torsional frequency shift is shown figure 17 with a projection of the


surface on the x − z plane. The sweep distance of the data was stretched,
due to the density of data points. The long range part of the torsional fre-
quency shift was cut, because of no additional information. The z distance
is arbitrary. For large tip sample distances, the wires have an extended tail
in the torsional frequency shift and are showing a more negative shift. The
periodicity of the extended features are in correspondence with the measured
topography. The average torsional frequency shift value is for large tip sam-
ple separations at fT R = −2Hz. Close to the sample the faint variations
become enhanced. First the signal to noise ratio becomes better and the
wires and holes are clearly visible. At very small tip sample distances the
wires as well as the holes show a doubling of the depressions or peaks. The
holes separate towards two peaks and in the center of a hole the torsional
frequency shift is almost zero. The wires show a doubling of the depressions,
too. But, in comparison with the holes, the frequency shift is almost zero or

40
even slightly positive for positions directly above the wires.
This torsional frequency shift pattern close to the surface, which relates to
the lateral force variation, becomes comprehensible, if it is assumed, that the
lateral force variation is proportional to the second derivative of the topogra-
phy. Lateral forces arise, if the surface becomes rough or periodic corrugated
[15]. The lateral forces become large, if the lateral gradient of the topogra-
phy is large, see Fig.:17 at wires. The measured torsional frequency shift at
extreme small amplitudes is related to the derivative of the torsional force
Eq. 2.20 [21].
If we consider a geometry consisting out of step functions like the Fermi func-
tion at non zero temperature, we can model a cross section of a corrugated
topography, see Fig.18. The shape of the lateral force can be approximated
with the derivative of the modeled topography in x-direction. The mea-
sured torsional frequency shift, is then related to the second derivative in
x-direction of the topography. The lateral force gradient should become
zero, if the gradient of the topography is zero.
If those Fermi functions are in a small interval, the lateral force gradient
would start overlapping. At a sufficient small interval, the response would
express itself in a doubling of the peak with only one dip or vice versa.
The depth of a dip between two peaks should be dependent to the distance
between the Fermi functions. The double dip feature at wires depend, pre-
sumably, on the finite width of the wires itself and the convolution of the tip
shape.
Figure 18a was modeled in accordance to the periodicity and corrugation of
the nanomesh for a line cross section through holes. The modeled nanomesh
was derived in x-direction to visualize the shape of the lateral force, Fig.18b.
The lateral force is not simulated and does not take real forces into account.
Therefore, it is arbitrary in units. The second derivative of the topography
denotes the change in the force and should correspond to the lateral force
gradient 18c, which is measured by the torsional frequency shift Eq.2.20
[21]. Figure 18c shows the doubling of the structure, as it was seen in in
the measured frequency shift Fig. 17. Although, the second derivative of
the topography does neglect any forces and considers only the topographical
shape of the surface.
In a detailed comparison between figures 18c and 17, differences can be seen.
First, hole peaks are much more broaden and have a less pronounced depres-
sion in the real spectroscopy. Second, in the spectroscopy, wire depressions
have a strongly pronounced peak, which has a small peak width. The peak
height reaches the zero frequency shift level and rises also above the zero
level into positive frequency shifts. Third, the slope between wire depression
and hole peaks in the spectroscopy is not as steep as the change in the mod-
eled approach. Forth, due to the fact, that the model has no real physical
property the peaks have a arbitrary scale and cannot be compared to the
values of the torsional frequency shift.

41
Figure 18: (a) a topography cross section was modeled with Fermi function
at non zero temperature, with the periodicity and the corrugation of the
nanomesh. (b) is the derivative of the topography in x-direction and denotes
the expected lateral forces. (c) shows the second derivative in x-direction,
it should show the expected lateral force gradient, which should assumed to
be in accordance with the measured torsional frequency shift.

These differences are mainly due to the fact that lateral forces can corre-
spond to the surface topography, but there are more reasons influencing the
lateral force, like electrical properties. On the other hand, the measurements
depend on the tip apex shape.
Beside the differences, the agreement between the measured lateral force
and the modeled is good. It can explain the shape and the doubling of the
peaks or depressions. The superpositions of Fermi function at non zero tem-
perature was a good choice for modeling the topography. Superimposing
sinusoidal functions would not show the features for the second derivative.

42
The outlook would be, to link the modeled topography with real forces and
compare these with further spectroscopy data.
We can confirm lateral force variations close to the surface by the torsional
measurements [15], which should provide additional binding forces. Third, a
distant dependent enhancement of holes is seen in the spectroscopy Fig. 14.
Further bimodal spectroscopy measurements are needed to affirm these mea-
surements, as well as bias spectroscopy for a comparison of the acquired data
to the electronic landscape of the nanomesh.

4.3 Domain orientation


Beside the contrast inversions, irregularities in the nanomesh pattern could
be observed. At bigger scan areas of 150x150nm2 distortions, tilting and
bending of the h-BN nanomesh were found. A clear nanomesh symmetry
with the 3 fold relation of the holes can be observed in every domain. On a
more detailed investigation, these regions with distinct symmetry are tilted
towards each other. It was found, that four different symmetry regions, with
a difference in the tilt angle of 15◦ , are existing, which will be called domains.

Both images 19(a, b) show large range topographical measurements of the


nanomesh, in which one finds domains with a 3 fold symmetry. The dif-
ference between those measurements are the scan position and the time of
measument. Figure 19(a) was measured 3 days after preparation and figure
19(b) 6 months later.
The topography signal in figure 19a was classified, in a detailed investigation,

43
Figure 19: figure (a) shows a large scan area with 131x131nm2 in size. This
measurement was done 3 days after the nanomesh preparation. The to-
pography signal showed clearly domains with different orientations. Right,
every domain was classified and colored, domains with the same orienta-
tion have the same color. Beneath, histograms of the four different do-
mains with respect to their size are shown. figure (b) is a measurement
with 150x150nm2 in size, done 6 months later. Due to a probable multitip
influence the nanomesh is not shown properly, but this defect enhances the
visibility of the difference in the domain orientation. Again, the domains
were colored with respect to their orientation.

on its domain orientation Fig. 19. The right overlay picture Fig. 19 was cre-
ated afterwards to enhance the visibility of the single domain areas. As well
as allowing a domain size measurement. These were used in the histogram,
with respect to their orientation. This was done for both figures 19a,b.
After a full classification almost no area is left without a domain affiliation.
In figure 19b, were still some unclassified spots left, due to distortions and tip
artifacts, which does not allow a proper orientation analysis. By considering
only the coloring, it is not possible to define a preferred domain orientation.
The position and orientation seems to be random, even at step edges the
domains are randomly aligned.
Histograms of figure (a) indicate an average domain size between 200nm2 −

44
Figure (a) Figure (b)
Scan frame in nm2 2502 1502
f1st in Hz 153153 158664
Q1st 32916 6323
k1st in N/m 25.1 28.0
∆f in Hz −8.5 −283
A1st in nm 5 4.9
γ factor in f N m0.5 −0.48 −17.1

Table 4: Table of image parameters of Figure 19.

400nm2 for every orientation. Some satellites at very large values, could
stem from aggregation of smaller domains. From the histogram, there is
no evidence of continuous domain sizes, due to the lack of counts in the
histogram, because the overall count of domains is too small for significant
statistic.
From these measurements, it can be assumed that the h-BN nanomesh is
growing on the rhodium(111) substrate in an island growth mode with many
nucleation points. We consider two possibilities for the domain creation.
First, the rhodium substrate was grown epitactical on sapphire (Al2 O3 ) for
some hundreds of nanometer with a distinguished orientation of (111) with
respect to the Miller notation. By the surface treatment of the prepara-
tion cycles large terraces and a plane surface could be achieved. DFM in-
vestigations at room temperature of the clean rhodium surface could not
resolve rhodium atoms, therefore the substrate surface orientation remains
unknown, whether it has grains with a tilt angle with respect to other grains,
or if it is a perfect epitactical layer with a large symmetry. If we consider a
grain formation with tilt angle, the borazine would nucleate on such a grain
and adept to the orientation. Such a nucleation center would start growing
the nanomesh in the energetic preferred orientation calculated by DFT by
Laskowski [27]. If two domains of the nanomesh would intersect the domain
start to form a domain wall region, were an orientation shift will occur.
Due to the energetic favoured orientation with respect to the substrate, two
different domains would not formate towards an aggregated domain with
only one orientation. This can be deduced by calculations of the rigidity
of the sp2 -nanomesh layer itself. If we take the second measurement into
account, which was done 6 months later, after many annealing cycles with
temperatures up to Tmax = 1123K, we do not observe recognisable changes
in the domain sizes and preferred orientation Fig. 19b. In the frame of do-
main growth on substrate grains with different tilt angle, it can be assumed,
that there is no change in the grain orientation of the rhodium substrate.
Therefore, no domain reorganization of the nanomesh was observed. One
possibility could be the lateral rigidity of the nanomesh and the π-bonding

45
towards the substrate, the nanomesh may prevent the substrate atoms of
reorganization.
Additional, four different orientations with a tilt angle of 15◦ were measured.
If the substrate was grown in grains, the grain orientation in principal can
have every orientation. The relation between the measured nanomesh do-
main orientation and the grain orientation is not clear, why four distinct
orientations are stable with respect to each other.
Second, if the rhodium was grown epitactical with a perfect crystal orienta-
tion at the surface, the borazine nucleation centers grow towards different
domain orientations on the substrate. This would query the DFT calcula-
tion [27], which does not predict other possible superstructure alignment on
the rhodium. If a perfect substrate orientation is considered, long annealing
cycles at high temperatures should provide enough kinetic energy to enhance
the preferred alignment of the domains. In the histograms of figure 19b, no
clear indication of a preferred orientation is recognisable.
If the substrate is grown perfect epitactical the question of the orientation
stabilization process remains. Considering a different domain orientation on
a perfect surface the boron nitride rings are less bounded on the substrate,
due to stress and the lack of the lock-in energy of the π orbitals 2.3 [15].
From this, a lower surface binding is assumed, which should be indicated by
different heights, different spacings between the sp2 layers and the rhodium
substrate, which is not observed.
A LEED analysis of our sample would solve the task of the crystal composi-
tion, whether it has a epitactical or a polycrystalline structure. Nevertheless,
it does not explain the stabilization mechanism of the four domain orienta-
tions.
From the measurements we assume that the nanomesh grows by an island
growth with different nucleation centers and is independent of the step edge
influence. Also, it is assumed, that the post annealing process does not
change the domain sizes and domain orientations towards a preferred orien-
tation and triggers no higher ordering.
Recently, the island growth of the h-BN nanomesh was observed in-situ [44].
Here, STM measurements of the nanomesh growth on Rh(111) surface were
made. Different domain orientations were observed in addition to domain
wall forming. The domain size seems to be related to the deposition rate
and the deposition temperature of the surface. For slow deposition rates and
large temperatures islands become bigger in size.

4.4 Atomic contrast measured by torsional resonance


Atomic resolution on the nanomesh with DFM were demonstrated earlier
in my master project work [48]. The resolution was achieved by standard
DFM, meaning first resonance for height control and small scan areas. Here,
I want to present atomic resolution of the nanomesh recorded by bimodal

46
dynamic force microscopy with the second flexural resonance at amplitudes
of A2nd = 600pm for height control and the first torsional resonance for ultra
sensitive probing of the atomic corrugation by changes of the lateral short
range forces. The torsional amplitude could be set to AT R = 20pm, which
provides the drastically increased resolution.

Figure 20: Figure (a) shows the topography signal controlled by the sec-
ond flexural resonance. This shows the nanomesh in lower resolution. In
figure (b) the torsional signal clearly indicates the ”atomic” corrugation of
the nanomesh with high resolution. figure (c) shows a 2D FFT of figure
(b). The bright spots in the corners denote the periodicity of the ”atomic”
corrugation and the one in the center due to random features in the mea-
surement. Additional, the FFT has no information in the y direction which
is predicted by the torsional oscillation mode. Image parameters figure (a):
Scan area 18x18nm2 , f2nd = 967613Hz, Q2nd = 12236, k2nd = 1881N/m,
∆f = −160Hz, A2nd = 600pm, γ = −0.40f N m0.5 . figure (b): Scan
area 18x18nm2 , fT R = 1.486602M Hz, QT R = 124429, kT R ∼ 2200N/m,
AT R = 20pm.

The measured topography signal, shown in figure 20a is acquired by the sec-
ond flexural resonance. The achieved resolution is rudimental, but denotes
the theoretical nanomesh hole wire pattern. The set point for the height
controlling frequency shift was ∆f2nd = −160Hz. A larger frequency shift
set point would drastically increase the probability of a tip apex change or
snap in contact instabilities, which would not allow stable measurement con-

47
ditions. No atomic resolution would be possible with a flexural oscillation
with this cantilever. A cantilever change or another cantilever preparation
would be recommendable.
In the torsional oscillation measurements of the atomic corrugation by the
lateral force variation Fig. 20b is imaged. Although we have a small signal
to noise ratio we still achieve high resolution. One of the main goals of the
torsional resonance is the extremely high Q-value of QT R = 124429, which
allows to measure the frequency shift with high fidelity. While the Q-value
provides a high signal to noise ratio the relatively high spring constant of
kT R ∼ 2200N/m reduces the over all noise level. These conditions in link
with bimodal dynamic force microscopy are a powerful tool for surface anal-
ysis.
Figure 20b shows the frequency shift of the torsional signal simultaneously
measured as the topography. At wire positions the corrugation of the BN-
rings between the atomic sites and the center of a ring is visible. This denotes
the atomic sites of the h-BN nanomesh. In a second approach, atomic res-
olution can be achieved by the torsional resonance. The elevations have a
hexagonal alignment and have relative distance of ∆a = 220 ± 20pm mea-
sured by 2D FFT Fig. 2c. The torsional signal implies not only the molecule
corrugation, additionally an enhancement of the hole rims is measured. This
is in correspondence with the theoretical predicted lateral force variation of
the nanomesh, see Fig.2c. The molecule corrugation is clear visible on the
wires and faintly visible in holes, this difference arises due to the poor ∆f2nd
signal. Nevertheless, it is still possible to distinguish between them in the
acquired signal.
In figure 20c the 2D FFT of the torsional signal is shown. The spots on
the corners denote the molecular elevations, while the bright spots at the
center have information of all other features including the enhancement of
the rims. Even though we recognise a hexagonal pattern of the molecule
corrugation, which corresponds to a 3-fold symmetry, only four bright spots
at the corners of the FFT can be seen. The angle between the two right
spots is almost 60◦ , but the angle of the right between the left upper spots
is about 120◦ . For a 3-fold symmetry 2 spots seem to be missing in the FFT
of the torsional frequency shift.
The torsional resonance oscillates in x direction. This allows a detection of
forces with a x-component, y-components are not detectable, due to the os-
cillation mode. This is nicely demonstrated in the FFT. A lack information
is seen in the pure y-direction. This is in perfect agreement with the theory
of cantilever dynamics.
The torsional frequency shift signal demonstrates the ultra high sensitivity of
the torsional resonance mode used in bimodal dynamic force microscopy. For
further enhanced resolution measurements on the nanomesh bimodal DFM
in combination with the torsional resonance is a unremitting tool.

48
5 Conclusion
The h-BN nanomesh is a amazing monolayer. A surface characterization
opened up a lot of questions, like a greek hydra it raised two more questions, if
one was solved. Nevertheless, we distinct three different contrast observable
on the surface. A normal predicted one, a low resolution inversion, which
lasts from a double tip influence Fig. 8 and a high resolution inversion, which
seems to be a property of the surface. We have seen this high resolution
contrast in scans and 2D spectroscopy Fig. 13. The spectroscopy showed
changes in the force distance curves Fig. 14, but the origin of this could not
be explained appropriately. Torsional measurements revealed the predicted
lateral force variations Fig. 17 [15] close to the surface. The structure of
the torsional frequency shift was modeled by a simple approach, which takes
only the topographical structure into account and no forces Fig. 18. This
approach is nevertheless quite accurate for such simple assumptions.
In addition structural domains have been observed. These domains have
been classified and it was found that there are four different orientations
with a 3 fold symmetry. No preferred domain orientation could be observed
and no change in the domain distribution after annealing could be observed
Fig. 19.
Atomic resolution at a large scan area was achieved by bimodal DFM. The
amplitude was controlled by the second flexural oscillation with amplitudes
of A2nd = 600pm and the atomic corrugation was measured by the ultra
sensitive second torsional resonance with an amplitude of AT R = 20pm,
Fig.20.
After all, tasks and open questions, I still have the confidence, that this
surface opens a lot of possibilities for room temperature single molecule
spectroscopy, due to the high reactivity molecules can be immobilized. If the
growth mode can be controlled and a preferred orientation is achievable, a
magnetic cobalt clusters could be aligned in arrays, controlling the magnetic
spins of those clusters could lead to a storage of information with a huge
density. Organic transistor arrays, organic solar cells, even vertical quantum
dots would be possible. The high environmental and thermal stability leave
an open range of applications in the nano mechanics. This opens a interesting
future for the h-BN nanomesh.

49
6 Acknowledgment
I thank Prof. Ernst Meyer for giving me the possibility to do my masterthesis
in his group. Additionally, I am thankful to have access to the UHV system
with such an amazing microscope. Without this system most of the measure-
ments weren’t possible. I really thank Sascha, Shigeki, Marcin and Thilo for
the never ending brain support. Every idea was created or improved by fruit-
ful discussions with them. Thank you, Sascha for the the critics, which lead
to further improvement of the ideas. Thank you, Shigeki for the amazing
knowledge and technical support for the spectroscopy measurements. Thank
you, Thilo for so many initial ideas and so many discussions.
I’d like to thank Ernst Meyer and Thilo to open up me the possibility of
presenting my work at conferences, this was a great opportunity for me and
helped me at my decision of starting a PhD study in physics.

50
save(strcat(Folder_path,Spec_folder,backsl,Base_name,Longbase_name,underl,'001'))

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% manual drift correction
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear hold_df1_F
clear hold_Exc1_F
clear hold_df2_F

drift= (0*100/(Nr_file));
for xi = 1:Nr_file
drift_corr(xi) = ceil(-xi*(drift/Nr_file)+drift);
for zi = 1:drift_corr(xi)
hold_df1_F(xi,zi)=df1_F(xi,zi);
hold_Exc1_F(xi,zi)=Exc1_F(xi,zi);
hold_df2_F(xi,zi)=df2_F(xi,zi);
end
for zi = 1:(nr_z-drift_corr(xi)-1)
zi = (nr_z)-zi+1;
df1_F(xi,(zi))= df1_F(xi,(zi-drift_corr(xi)));
Exc1_F(xi,(zi))=Exc1_F(xi,(zi-drift_corr(xi)));
df2_F(xi,(zi))= df2_F(xi,(zi-drift_corr(xi)));
for i = 1:drift_corr(xi)
df1_F(xi,i) = hold_df1_F(xi,i);
Exc1_F(xi,i) = hold_Exc1_F(xi,i);
df2_F(xi,i) = hold_df2_F(xi,i);
end
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% calculate shift in Longdata
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average_df1_long = mean(long_df1_F(1:75));
df1_correction_long = 0-average_df1_long;
long_df1_F = long_df1_F + df1_correction_long;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Shift all the Data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for xi = 1:Nr_file
for zi = 1:nr_z
df1_F(xi,zi) = df1_F(xi,zi)+(df1_correction_long);
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% calculate the gradient in Longdata
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
grad_nr=150;
Gradient_long_df1 = zeros(1,grad_nr);
Gradient_long_x = zeros(1,grad_nr);
Gradient_long_y = zeros(1,grad_nr);
for zi = nr_z-grad_nr:nr_z
Gradient_long_x(zi) = long_Z_F(zi);
Gradient_long_y(zi) = long_df1_F(zi);
end
ans = [Gradient_long_x' ones(length(Gradient_long_x),1)]\Gradient_long_y';
Gradient_long_df1 = ans(1);

Max_Gradient_long = max(Gradient_long_df1);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% calculate the gradient
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
gradient_nr=400;
Gradient_x = zeros(1,gradient_nr);
Gradient_y = zeros(1,gradient_nr);
for xi = 1:Nr_file
Gradient_df2=zeros(Nr_file);
for zi = 1:gradient_nr
Gradient_x(zi) = Z_F(zi);
Gradient_y(zi) = df1_F(xi,zi);
end
ans = [Gradient_x' ones(length(Gradient_x),1)]\Gradient_y';
Gradient_df2(xi) = ans(1);
end
Max_Gradient = max(max(Gradient_df2));

shift_z = zeros(Nr_file,gradient_nr);
for xi = 1:Nr_file
while(Gradient_df2(xi) < Max_Gradient)
for zi = 1:gradient_nr
zi = zi+shift_z(xi);
Gradient_x(zi) = Z_F(zi);
Gradient_y(zi) = df1_F(xi,zi);
end
ans = [Gradient_x' ones(length(Gradient_x),1)]\Gradient_y';
Gradient_df2(xi,:) = ans(1);
shift_z(xi) = shift_z(xi) +1;
end
end

for xi = 1:Nr_file
for zi = (gradient_nr+1):(nr_z - shift_z(xi))
df1_F(xi,zi)= df1_F(xi,(zi+shift_z(xi)));
Exc1_F(xi,zi)=Exc1_F(xi,(zi+shift_z(xi)));
df2_F(xi,zi)= df2_F(xi,(zi+shift_z(xi)));
end
for zi = (nr_z - shift_z(xi))+1 : nr_z
df1_F(xi,zi) = df1_F(xi,(nr_z-shift_z(xi)));
Exc1_F(xi,zi) = Exc1_F(xi,(nr_z-shift_z(xi)));
df2_F(xi,zi) = df2_F(xi,(nr_z-shift_z(xi)));
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Add Long data to the normal set of data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear integ_A
clear integ_B
clear z_step
clear root_amplitude
clear F_F
clear root_distanceall_Z_F;

all_A1_F = zeros(1,(2*nr_z-(grad_nr+1)));
all_df1_F = zeros((Nr_file),(2*nr_z-(grad_nr+1)));

F_F = zeros(Nr_file,(2*nr_z-(grad_nr+1)));
for xi = 1:Nr_file
for zi = 1:(2*nr_z-grad_nr)
if zi <=nr_z-grad_nr
all_Z_F(zi)=long_Z_F(zi);
all_A1_F(zi)=long_A1_F(zi);
all_df1_F(:,zi)=long_df1_F(zi);
else
all_Z_F(zi)=Z_F(zi-(nr_z-grad_nr));
all_A1_F(zi)=A1_F(zi-(nr_z-grad_nr));
all_df1_F(xi,zi)=df1_F(xi,zi-(nr_z-grad_nr));
end
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Calculating the Force
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for zi = 2:(2*nr_z-grad_nr)
integ_A = zeros(Nr_file,1);
integ_B = zeros(Nr_file,1);
for zzi = 1:(zi-1)
z_step = all_Z_F(zzi)-all_Z_F(zzi+1);
root_distance = sqrt(abs(all_Z_F(zzi)-all_Z_F(zi))) ;
root_amplitude = sqrt(abs(all_A1_F(1,zzi)-all_A1_F(1,zi)));
integ_A =integ_A+(1+root_amplitude*z_step/(8*sqrt(pi)*root_distance))*all_df1_F(:,zzi);
integ_B =integ_B+ 1/(sqrt(2)*root_distance)*(all_df1_F(:,zzi+1) - all_df1_F(:,zzi));
end
F_F(:,zi) = integ_A+integ_B * root_amplitude^3;
end
all_F_F = 2 * k_1st * F_F / f_1st;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% CUT OFF the Long data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear F_F
cutoff = zeros(Nr_file,nr_z-grad_nr);

for xi = 1:Nr_file
for zi = 1:(2*nr_z-grad_nr)
if zi <= nr_z-grad_nr
cutoff(xi,zi)=all_F_F(xi,zi);
else
F_F(xi,zi-(nr_z-grad_nr))=all_F_F(xi,zi);
end
end
end
size(F_F);
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Add Long data to the normal set of data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear integ_pot_A
clear integ_pot_B
clear integ_pot_C
clear z_step
clear root_amplitude
clear U_F
clear root_distanceall_Z_F;

all_A1_F = zeros(1,(2*nr_z-(grad_nr+1)));
all_df1_F = zeros((Nr_file),(2*nr_z-(grad_nr+1)));

U_F = zeros(Nr_file,(2*nr_z-(grad_nr+1)));
for xi = 1:Nr_file
for zi = 1:(2*nr_z-grad_nr)
if zi <=nr_z-grad_nr
all_Z_F(zi)=long_Z_F(zi);
all_A1_F(zi)=long_A1_F(zi);
all_df1_F(:,zi)=long_df1_F(zi);
else
all_Z_F(zi)=Z_F(zi-(nr_z-grad_nr));
all_A1_F(zi)=A1_F(zi-(nr_z-grad_nr));
all_df1_F(xi,zi)=df1_F(xi,zi-(nr_z-grad_nr));
end
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Calculating the Potential
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for zi = 2:(2*nr_z-grad_nr)
integ_pot_A = zeros(Nr_file,1);
integ_pot_B = zeros(Nr_file,1);
integ_pot_C = zeros(Nr_file,1);
for zzi = 1:(zi-1)
z_step = all_Z_F(zzi)-all_Z_F(zzi+1);
distance = abs(all_Z_F(zzi)-all_Z_F(zi));
root_distance = sqrt(distance);
root_amplitude = sqrt(abs(all_A1_F(1,zzi)-all_A1_F(1,zi)));
integ_pot_A = integ_pot_A + (root_amplitude/4)*(root_distance/sqrt(pi))*all_df1_F(:,zzi);
integ_pot_B = integ_pot_B + (root_amplitude^3/(sqrt(2)*root_distance))*all_df1_F(:,zzi);
integ_pot_C = integ_pot_C + distance*all_df1_F(:,zzi);
end
U_F(:,zi) = integ_pot_A + integ_pot_B + integ_pot_C;
end
all_U_F = 2 * k_1st * U_F / f_1st;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% CUT OFF the Long data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear U_F
cutoff = zeros(Nr_file,nr_z-grad_nr);

for xi = 1:Nr_file
for zi = 1:(2*nr_z-grad_nr)
if zi <= nr_z-grad_nr
cutoff(xi,zi)=all_U_F(xi,zi);
else
U_F(xi,zi-(nr_z-grad_nr))=all_U_F(xi,zi);
end
end
end
size(U_F);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% ddf_1st and dF_F
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
df1_average_F = zeros(1,nr_z);
ddf1_F = zeros(Nr_file,nr_z);
for xi = 1:Nr_file
for zi = 1:nr_z
df1_average_F(zi)=mean(df1_F(:,zi));
ddf1_F(xi,zi)=df1_F(xi,zi)-df1_average_F(zi);
end
end

dF_average_F = zeros(1,nr_z);
dF_F = zeros(Nr_file-init_nr,nr_z);
for zi = 1:(nr_z)
dF_average_F(zi)=mean(F_F(:,zi));
for xi = 1:Nr_file
dF_F(xi,zi)=F_F(xi,zi)-dF_average_F(zi);
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Get amplitude
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
A_1st= mean(mean(A1_F(xi,1:100)));

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% calculate the average excitation
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for xi = 1:Nr_file
Average_Exc1_F(xi) = mean(Exc1_F(xi,1:50));
end

for xi = 1:Nr_file
Exc1_F(xi,:)= Exc1_F(xi,:) ./ Average_Exc1_F(xi) - 1 ;
end

Exc1_F_inst = pi* k_1st * A_1st^2 / Q_1st /eV*1000;


Exc1_F = Exc1_F * Exc1_F_inst;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Filter df signal in Z direction
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for xi = (1+1):Nr_file
df1_F(xi,:) = sgolayfilt(df1_F(xi,:),2,5);
Exc1_F(xi,:) = sgolayfilt(Exc1_F(xi,:),2,5);
ddf1_F(xi,:) = sgolayfilt(ddf1_F(xi,:),2,5);
df2_F(xi,:) = sgolayfilt(df2_F(xi,:),2,5);
%F_F(xi,:) = sgolayfilt(F_F(xi,:),2,5);
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Filter df signal in X direction
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
for zi=2:nr_z
df1_F(:,zi) = sgolayfilt(df1_F(:,zi),2,5);
Exc1_F(:,zi) = sgolayfilt(Exc1_F(:,zi),2,5);
ddf1_F(:,zi) = sgolayfilt(ddf1_F(:,zi),2,5);
df2_F(:,zi) = sgolayfilt(df2_F(:,zi),2,5);
%F_F(:,zi) = sgolayfilt(F_F(:,zi),2,5);
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Make axis information
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
X = 1:Nr_file;
Z = 0:(nr_z-1);
Z_F = Z_F * 1.00e+9;
X_max = max(Z_F);
X_min = min(Z_F);
z_nm = Z * (X_max - X_min)/(nr_z-1);
x_nm=9.203*(X)/(Nr_file);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Plot Range and Contour Spacings
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear figure
scrsz=get(0,'ScreenSize');
figure('Position',[1 1 scrsz(3) scrsz(4)],'Name',fname, 'NumberTitle','off');
colormap (flipud(jet(96)))

clear clevs_df1
clear temp_df1
temp_df1 = rot90(df1_F,1);
%C_max_df1= 55;
%C_min_df1= -130;
C_max_df1=max(max(temp_df1));
C_min_df1=min(min(temp_df1));
% for i=1:clev_nr
% clevs_df1(i) = C_min_df1+(i*(C_max_df1-C_min_df1))/clev_nr;
% end

clear clevs_df2
clear temp_df2
temp_df2 = rot90(df2_F,1);
C_max_df2= 2;
C_min_df2= -6;
% C_max_df2=max(max(temp_df2));
% C_min_df2=min(min(temp_df2));
% for i=1:clev_nr
% clevs_df2(i) = C_min_df1+(i*(C_max_df2-C_min_df2))/clev_nr;
% end

clear clevs_Exc1
clear temp_Exc1
temp_Exc1 = rot90(Exc1_F,1);
%C_max_Exc1= 5;
%C_min_Exc1= -5;
C_max_Exc1= max(max(temp_Exc1));
C_min_Exc1= min(min(temp_Exc1));
% C_min_Exc1= 0;
%for i=1:clev_nr
%clevs_Exc1(i) = C_min_Exc1+(i*(C_max_Exc1-C_min_Exc1))/clev_nr;
%end

clear clevs_ddf1
clear temp_ddf1
temp_ddf1 = rot90(ddf1_F,1);
C_max_ddf1= 3;
C_min_ddf1= -6;
% C_max_ddf1=max(max(temp_ddf1));
% C_min_ddf1= min(min(temp_ddf1));
%for i=1:clev_nr
%clevs_ddf1(i) = C_min_ddf1+(i*(C_max_ddf1-C_min_ddf1))/clev_nr;
%end

clear clevs_dF_F
clear temp_dF_F
temp_dF = rot90(dF_F,1);
C_max_dF= 1.75;
C_min_dF= -1.75;
% C_max_dF=max(max(temp_dF));
% C_min_dF=min(min(temp_dF));
% for i=1:clev_nr
% clevs_dF(i) = C_min_dF+(i*(C_max_dF-C_min_dF))/clev_nr;
% end

clear clevs_F_F
clear temp_F_F
temp_F_F = rot90(F_F,1);
% C_max_F_F= 1;
% C_min_F_F= -1;
C_max_F_F=max(max(temp_F_F));
C_min_F_F=min(min(temp_F_F))
%for i=1:clev_nr
%clevs_F_F(i) = C_min_F_F+(i*(C_max_F_F-C_min_F_F))/clev_nr;
%end

clear clevs_U_F
clear temp_U_F
temp_U_F = rot90(U_F,1);
%C_max_U_F= 55;
%C_min_U_F= -130;
C_max_U_F=max(max(temp_U_F));
C_min_U_F=min(min(temp_U_F));
%for i=1:clev_nr
%clevs_U_F(i) = C_min_U_F+(i*(C_max_U_F-C_min_U_F))/clev_nr;
%end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Plotting XZ plane
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

subplot(7,1,1)

% [c,h] = contour3(x_nm, z_nm, temp_df1,clevs_df1,'k');set(h,'LineWidth',2); hold 'on';


surf(x_nm, z_nm, temp_df1,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_df1,C_max_df1]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('\Delta f_{2nd}(z) (Hz)','fontsize', 22)
axis equal tight

subplot(7,1,2)

% [c,h] = contour3(x_nm, z_nm, temp_ddf1,clevs_ddf1,'k');set(h,'LineWidth',2);hold 'on';


surf(x_nm, z_nm, temp_ddf1,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_ddf1,C_max_ddf1]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('\Delta f_{2nd}(z) - f_{average}(z) (Hz)','fontsize', 22)
axis equal tight

subplot(7,1,3)

% [c,h] = contour3(x_nm, z_nm, temp_df2,clevs_df2,'k');set(h,'LineWidth',2); hold 'on';


surf(x_nm, z_nm, temp_df2,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_df2,C_max_df2]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('\Delta f_{TR}(z) (Hz)','fontsize', 22)
axis equal tight

subplot(7,1,4)

%[c,h] = contour3(x_nm, z_nm, temp_F_F,clevs_F_F,'k');set(h,'LineWidth',2);hold 'on';


surf(x_nm, z_nm, temp_F_F,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_F_F,C_max_F_F]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('Force_{2nd}(z) in (nN)','fontsize', 22)
axis equal tight

subplot(7,1,5)

% [c,h] = contour3(x_nm, z_nm, temp_dF,clevs_dF,'k');set(h,'LineWidth',2);hold 'on';


surf(x_nm, z_nm, temp_dF,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_dF,C_max_dF]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('Force_{2nd}(z) - Force_{average}(z) in (nN)','fontsize', 22)
axis equal tight

subplot(7,1,6)

%[c,h] = contour3(x_nm, z_nm, temp_Exc1,clevs_Exc1,'k');set(h,'LineWidth',2);hold 'on';


surf(x_nm, z_nm, temp_Exc1,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_Exc1,C_max_Exc1]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('Excitation_{1st}(mV)','fontsize', 10)
axis equal tight

subplot(7,1,7)

%[c,h] = contour3(x_nm, z_nm, temp_U_F,clevs_U_F,'k');set(h,'LineWidth',2);hold 'on';


surf(x_nm, z_nm, temp_U_F,'EdgeColor','none','FaceLighting','phong'),caxis manual,caxis([C_min_U_F,C_max_U_F]), view(0,+90)
colorbar
xlabel('X(nm)','fontsize', 18);
ylabel('Z(nm)','fontsize', 18);
set(gca,'XminorTick','on','YminorTick','on','fontsize', 18);
title('Potential(z)','fontsize', 10)
axis equal tight

Plane='XZ';
save(strcat(Folder_path,Spec_folder,backsl,Base_name,underl))
Outputname=strcat(Base_name,underl,Plane);
saveas(gcf,strcat(Folder_path,Spec_folder,backsl,Outputname),'tif')
%close(gcf)
References
[1] F. J. Giessibl Atomic resolution of the Silicon(111)-(7x7) Surface by
Atomic Force Microscopy, Science (1995), 267, 68

[2] D. Fotiadis, Y. Liang, S. Filipek, D.A. Sperstein, A. Engel and K.


Palczewski Atomic-force-microscopy: Rhodopsin dimers in native disc
membranes, Nature (2003), 421, 127-(2)

[3] H. Seelert, A. Poetsch, N.A. Dencher, A. Engel, H. Stahlberg and D.J.


Müller Structural biology: Proton-powered turbine of a plant motor, Na-
ture (2000), 405, 418-(2)

[4] Y. Sugimoto, P. Pou, M. Abe, P. Jelinek, R. Pérez, S. Morita and O.


Custance Chemical identification of individual surface atoms by atomic
force microscopy, Nature (2007), 446, 64-(3)

[5] D. Rugar, R. Budakian, H. J. Mamin and B. W. Chui Single spin de-


tection by magnetic resonance force microscopy, Nature (2004), 430,
329-(3)

[6] J. Martin, N. Akerman, G. Ulbricht, T. Lohmann, J. H. Smet, K. von


Klitzing and A. Yacoby Observation of electronhole puddles in graphene
by using a scanning single electron transistor, Nature Physics (2008),
4, 144 - 148

[7] K.V. Emtsev, A. Bostwick, K. Horn, J. Jobst, G.L. Kellogg, L. Ley,


J.L. McChesney, T. Ohta, S.A. Reshanov, J. Röhrl, E. Rotenberg, A.K.
Schmid, D. Waldmann, H.B. Weber and T. Seyller Towards wafer-size
graphene layers by atmospheric pressure graphitization of silicon carbide,
Nature Materials (2009), 8, 203 - 207

[8] K.S. Kim, Y. Zhao, H. Jang, S.Y. Lee, J.M. Kim, K.S. Kim, J.H. Ahn, P.
Kim, J.-Y. Choi and B.H. Hong Large-scale pattern growth of graphene
films for stretchable transparent electrodes, Nature (2008), 447, 706-(4)

[9] G. Binning, C.F. Quate, Ch. Gerber, Atomic Force Microscope, PRL.
(1986), 56, 930-933

[10] G. Binning, h. Rohrer, Ch. Gerber, E. Weibel Surface Studies by Scan-


ning Tunneling Microscopy, PRL. (1982), 49, 57-61

[11] Christoph Bruder, Van der Waals- und Casimir-Kräfte, Phys. in unserer
Zeit (1997), 28, 149-154

[12] D.C. Prieve and W.B. Russel Simplified predictions of Hamaker con-
stants from Lifshitz theory, Journal of Colloid and Interface Science,
(1987), Vol. 125, No.1

51
[13] R. Smoluchowski Anisotropy of the electronic work function of metals,
Physical Review (1941), 60, 661-674
[14] Ernst Meyer, Hans Josef Hug, Roland Bennewitz Scanning Probe Mi-
croscopy: The Lab on a Tip, Springer, Berlin (2003), 1. Auflage
[15] Thomas Greber Chapter:Graphene and Boron Nitride Single Layers,
Taylor and Francis Books, (2009) arXiv: 0904.1520v1
[16] Lukas Howald, Raster-Kraftmikroskopie an Silizium und Ionenkristallen
im Ultrahochvakuum, Dissertation, Universität Basel (1994)
[17] M.Guggisberg, M. Bammerlin, Ch. Loppacher, O. Pfeiffer, A. Abdurixit,
V. Barwich, R. Benewitz, A. Baratoff, E. Meyer, H.-J. Güntherodt Sep-
aration of interactions by noncontact force microscopy, PRB (2000),
61, 11151-11155
[18] Franz J. Giessibl, Advances in atomic force microscopy, Rev. Mod. Phys.
(2003), 75-3, 949-978,
[19] Demtröder, Experimentalphysik 3: Atome, Moleküle und Festkörper,
Springer, Berlin (2005), 3. Auflage
[20] S. Kawai, T. Glatzel, S. Koch, B. Such, A. Baratoff and Ernst Meyer
Systematic achievement of Improved Atomic-Scale Contrast via Bimodal
Dynamic Force Microscopy, PRL. (2009), 103, 220801 (4)
[21] S. Kawai, T. Glatzel, S. Koch, B. Such, A. Baratoff and Ernst
Meyer Ultrasensitive detection of leteral atomic-scale interactions on
graphite(0001) via bimodal dynamic force measurements, PRB. (2010),
81, 085420 (7)
[22] A. Nagashima, N Tejima, Y. Gamou, T.Kawai, C. Oshima, Electronic
Structure of Monolayer Hexagonal Boron Nitride Physisorbed on Metal
Surfaces, PRL (1995), 75, 3918(4)
[23] Martina Corso, Willi Auwärter, Matthias Muntwiler, Anna Tamai,
Thomas Greber, Jürg Osterwalder Boron Nitride Nanomesh, Science
(2004), 303, 217-220
[24] Simon Berner, Martina Corso, Roland Widmer, Oliver Groening,
Robert Laskowski, Peter Blaha, Karlheinz Schwarz, Andrii Goriachko,
Herbert Over, Stefan Gsell, Matthias Schreck, Hermann Sachdev,
Thomas Greber, Jürg Osterwalder, Boron Nitride Nanomesh: Function-
ality from a Corrugated Monolayer, Angew. Chem. (2007), 46, 5115-
5119
[25] J. Hölzl, F.K. Schulte and H. Wagner Work Function of Metals, Springer
(1979), Volumen 85 of Springer Tracts in Modern Physics.

52
[26] G. B. Grad, P. Blaha, K. Schwarz, W. Auwarter and T. Greber Density
functional theory investigation of the geometric and spintronic structure
of h-bn/ni(111) in view of photoemission and stm experiments, Phys.
Rev. B (2003), 68, 085404(8)

[27] Robert Laskowski, Peter Blaha, Thomas Gallauner, Karlheinz Schwarz,


Single-Layer Model of the Hexagonal Boron Nitride Nanomesh on the
Rh(111) Surface, PRL (2007), 98, 106802(4)

[28] A. Stock and E. Pohland Chem. Ber. (1926), 59B, 2215(8)

[29] S. Belaidi, P. Girard, G. Leveque, Electrostatic forces acting on the tip


in atomic force microscopy: Modelization and comparison with analytic
expressions, J. Appl. Phys. (1997), 81-3, 1023-1030

[30] Andrii Goriachko, Yunbin He, Marcus Knapp, Herbert Over Self-
Assembly of a Hexagonal Boron Nitride Nanomesh on Ru(0001) Lang-
muir (2007), 23, 2928-2931

[31] W. Auwaerter, T.J. Kreutz, T. Greber and J. Osterwalder XPD and


STM investigation of hexagonal boron nitride on Ni(111) Surface Sci-
ence, (1999) 429, 229-236

[32] , Th. Glatzel, M.Ch. Lux-Steiner, E. Strasbourg, A. Boag, Y. Rosen-


waks I.4 Principles of Kelvin Probe Force Microscopy, Springer, Berlin
(2007), 113-134

[33] G.H. Enevoldsen, A.S. Foster, M.C. Christensen, J.V. Lauritsen, F.


Besenbacher Noncontact atomic force microscopy studies of vacancies
and hydroxyls of T iO2 (110): Experiments and atomistic simulations
Phys. Rev. B (2007) 76, 205415(14)

[34] G.H. Enevoldsen, T. Glatzel, M.C. Christensen, J.V. Lauritsen, F. Be-


senbacher Atomic Scale Kelvin Probe Force Microscopy Studies of the
Surface Potential Variations on the T iO2 (110) Surface PRL, (2008)
100, 236104 (4)

[35] P. Steiner, R. Roth, E. Gnecco, A. Baratoff, S. Maier, T. Glatzel and


Ernst Meyer Two-dimensional simulations of superlubricity on NaCl and
highly oriented pyrolytic graphite PRB, (2009) 79, 045414 (9)

[36] The MathWorks, Matlab R2008a, (2008)

[37] T. Glatzel, L. Zimmerli, S. Koch, B. Such, S. Kawai and Ernst Meyer


Determination of effective tip geometries in Kelvin probe force mi-
croscopy on thin insulating films on metals Nanotechnology, (2009) 20,
264016 (7)

53
[38] Van Hove, M. A. Hermann, K. Watson Tables for 4.1.
Bonzel SpringerMaterials - The Landolt-Börnstein Database
(http://www.springermaterials.com), (06.06.2010) DOI:
10.1007/10783464-4

[39] J. E. Sader, S. P. Jarvis Accurate formulas dor interaction force and


energy in frequency modulated force spectroscopy APL, (2004) Vol. 84,
No. 10

[40] L. Olsson, N. Lin, V. Yakimov and R. Erlandsson A method for insitu


characterization of tip shape in ac-mode atomic force microscopy using
electrostatic interaction APL, (1998) Vol. 84, No. 8

[41] S. Hudlet, M. Saint Jean, C. Guthmann and J. Berger Evaulation of the


capacitive force between an atomic force microscopy tip and a metallic
surface Eur. Phys. J. B, (1998) 2, 5-10

[42] S. Hudlet, M.Saint Jean, B. Roulet, J.Berger, C. Guthmann Electro-


static forces between metallic tip and semiconductor surfaces J. Appl.
Phys. (1995), 77-7, 3308-3314

[43] T. R. Albrecht, P. Grutter, H. K. Horne, D. Rugar Frequency modu-


lation detection using high-Q cantilevers for enhanced force microscope
sensitivity J. Appl. Phys. (1991), 69-2, 668-673

[44] G. Dong, E.B. Fourre, F. C. Tabak and J.W.M. Frenken How Boron
Nitride forms a regular Nanomesh on Rh(111) PRL (2010), 104, 096102

[45] Th. Glatzel, S. Sadewasser, M.Ch. Lux-Steiner Amplitude or frequency


modulation-detection in Kelvin probe force microscopy Appl. Surf. Sci.
(2003), 210, 84-89

[46] Lars A. Zimmerli, Assemblies of Organic Molecules on Insulating Sur-


face Investigated by nc-AFM, Dissertation, Universität Basel (2007)

[47] Arndt Brünner, www.arndt-bruenner.de/mathe/scripts/kreis3p.htm,


Homepage, (6.März.2010), 28.Mai.2010

[48] Markus Langer, nc-AFM Analyse des h-BN Nanomesh, Masterprojekt


Work, Universität Basel (2009)

54

Potrebbero piacerti anche