Sei sulla pagina 1di 35

Home Search Collections Journals About Contact us My IOPscience

Force calibration in lateral force microscopy: a review of the experimental methods

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

2010 J. Phys. D: Appl. Phys. 43 063001

(http://iopscience.iop.org/0022-3727/43/6/063001)

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

Download details:
IP Address: 124.247.231.3
The article was downloaded on 11/08/2010 at 12:03

Please note that terms and conditions apply.


IOP PUBLISHING JOURNAL OF PHYSICS D: APPLIED PHYSICS
J. Phys. D: Appl. Phys. 43 (2010) 063001 (34pp) doi:10.1088/0022-3727/43/6/063001

TOPICAL REVIEW

Force calibration in lateral force


microscopy: a review of the experimental
methods
Martin Munz
Analytical Science Division, National Physical Laboratory, Hampton Road, Teddington,
Middlesex TW11 0LW, UK
E-mail: martin.munz@npl.co.uk

Received 29 December 2008, in final form 6 November 2009


Published 29 January 2010
Online at stacks.iop.org/JPhysD/43/063001

Abstract
Lateral force microscopy (LFM) is a variation of atomic/scanning force microscopy
(AFM/SFM). It relies on the torsional deformation of the AFM cantilever that results from the
lateral forces acting between tip and sample surface. LFM allows imaging of heterogeneities
in materials, thin films or monolayers at high spatial resolution. Furthermore, LFM is
increasingly used to study the frictional properties of nanostructures and nanoparticulates. An
impediment for the quantification of lateral forces in AFM, however, is the lack of reliable and
established calibration methods. A widespread acceptance of LFM requires quantification
coupled with a solid understanding of the sources of uncertainty. This paper reviews the
available experimental calibration methods and identifies particularly promising approaches.
(Some figures in this article are in colour only in the electronic version)

1. Introduction developing technologies with requirements for nano-


mechanical characterization, such as micro- and nano-
Since its invention in 1987 [1], lateral force microscopy electromechanical systems (MEMS/NEMS) [7, 14, 15] or
(LFM), frequently also referred to as friction force microscopy biomedical engineering and biotechnology [16, 17]. For
(FFM), has found a steadily increasing number of applications instance, Bhushan et al [14] demonstrated the use of LFM
[2–4]. Among others, it has been employed for mapping spatial for the error analysis of arrays of micromirrors. Such
variations in the surface properties of heterogeneous polymers micromirrors are key components of digital micromirror
[5, 6], thin films [7] or surfaces patterned using lithography devices (DMDs), a type of MEMS devices with optical
techniques [8, 9]. functionalities. Related to the fabrication process of MEMS
The dependence of frictional forces on the chemical structures, Kim et al [18] employed LFM to evaluate the
species present can be utilized for differentiating chemical forces required for the collapse of photoresist patterns. Such
groups by measuring the force interactions with a chemically patterns tend to fail in the course of the rinsing required for the
functionalized tip [10–12]. Owing to this capability, LFM can development process.
be considered a form of chemical force microscopy (CFM). Hence, in addition to mere imaging of lateral forces,
Frequently, CFM is performed by measuring force–distance LFM can also be utilized for the nanomechanical testing of
curves (FDCs); however, the lateral forces measured upon thin films or small structures. Wright and Armstrong [16]
scanning with a chemically functionalized tip were shown to studied the removal of microbial cells from medically relevant
be closely related to the adhesion force values extracted from surfaces. Considering that in many cases microbial cells
FDCs [8, 10, 13]. are exposed to hydrodynamic shear fields, the application
Further major drivers for the advancement of nano- of lateral forces can be used to measure the critical shear
scale friction force measurements are emerging or rapidly strength of such cell–surface interfaces. Similarly, the

0022-3727/10/063001+34$30.00 1 © 2010 IOP Publishing Ltd Printed in the UK


J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

shear strength of nanoparticles adhering to a support surface parameters, and a major means of assessing the resulting data
has been studied [19]. As compared with macroscopic is to cross-check if they are consistent with the values derived
techniques averaging over a large number of nano-scale from several calibration routines [28].
entities, LFM allows for force measurements on single The AFM cantilever (CL) acts as the core element of
particles, a prerequisite for studies of the relationship between force sensing. The force interactions between the integrated
particle shape and the critical force required for dislodgement tip and the surface under investigation result in deformations
of a particle. Similarly, Ling et al [20] employed LFM to of the CL. Using a position-sensitive detector (PSD), these
investigate the friction between two single microspheres. deformations are converted into measurement signals. A
It is interesting to note that the forces measured in LFM bidirectional PSD allows for both normal and lateral forces to
can range from micronewton down to piconewton, depending be detected via corresponding deflections of the CL, i.e. via the
on the application. For the purpose of estimation of the CL bending (flexural deformation) and the CL twist (torsional
force ranges, it can be assumed that Amontons’ law [21] deformation), respectively. In the most common version of
is valid, i.e. that the friction forces scale linearly with the the PSD, a laser beam bounces off the back of the CL and
normal load. Comparatively high loads are applied when the angular deflection of the reflected beam is detected using a
performing nano-scale friction and tribology experiments on segmented photodiode. Essentially, quantitative measurement
hard coatings or materials [22, 23], whereas the loads need to of lateral forces by means of AFM requires two separate
be minimized when studying surfaces of biomaterials or soft calibrations [29], namely the calibration of the lateral PSD
matter [24, 25]. In the limit of measurements of the intrinsic response to convert the measured signal into the twist angle
forces of biomolecules, the forces can be around ∼10 pN, as of the CL, and the calibration of the torsional spring constant
marked by the unzipping of double-stranded DNA upon axial of the CL that converts the twist angle into a torque. This
loading [26, 27]. Thus, the vast spectrum of applications of is analogous to the calibration of normal forces, where both
LFM entails a wide range of forces, in particular if applications the related optical lever sensitivity and the CL normal spring
to biosystems are increasingly explored that involve forces in constant (stiffness) need to be calibrated [30].
the piconewton range. So far, for many applications of LFM However, the calibration of lateral forces is even more
the lateral forces are in the range of a few nanonewton to a few challenging than that of normal forces. Due to the lateral
CL spring constant (stiffness) being much larger than the
hundred nanonewton.
normal CL spring constant (stiffness), it is generally difficult to
In the case of friction measurements, the lateral forces
measure the lateral deflection sensitivity. For a stiff substrate,
scale linearly with the normal forces if Amontons’ law [21]
the contact stiffness measured in the normal direction is large
applies. With the values of the friction coefficient, µ, typically
compared with the stiffness of typical contact CLs. Hence
ranging from ∼0.01 to ∼1, the friction forces are in the range
the contact deformation can be considered negligible. This
of ∼1 nN to ∼100 nN if a normal load of 100 nN is applied. For
condition is utilized for the calibration of the normal PSD
materials that show friction coefficient values in the order of
sensitivity, where a FDC is measured on a stiff substrate and
0.01 [22, 24], a high force resolution is required to differentiate
the sensitivity identified with the slope of the linear compliance
such materials from each other. Moreover, the lateral force
region of the FDC [31]. However, an equivalent configuration
calibration method needs to allow for corresponding levels of for the lateral case is severely affected by the finite lateral
accuracy. Such requirements should be borne in mind when stiffness both of the tip and of the tip–sample contact. The
selecting methods for the calibration of lateral forces. The latter scales with the radius of the tip apex. Hence the lateral
force range, the force resolution and the level of accuracy need deformation related to the tip–sample contact can be neglected
to be in line with the application. for very large tips, such as colloidal probes. As shown by
It should be mentioned that the maximum force that can be Lantz et al [32], it certainly needs to be accounted for in the
applied via the AFM tip is limited by the mechanical strength case of sharp tips.
of the AFM tip. Since silicon is a brittle material that tends Over the last 15 years a variety of methods for the
to fracture, alternative tip materials are necessary if very high calibration of lateral forces have been demonstrated. Since the
forces need to be applied in tribological studies. For instance, year 2000 an increased level of activity can be observed. The
diamond tips can be a valuable option [23]. recent rise of the subject seems to indicate the widely accepted
Despite the early and continued demonstration of the relevance of LFM as well as the need for an established
use of LFM for a wide spectrum of applications, a lack calibration technique. Indeed a considerable variety of
of established calibration methods impeded its widespread methods for the calibration of lateral forces in AFM has been
acceptance. In particular, the reduced pace of development suggested. An overview of the existing approaches and their
of LFM manifests itself when comparing it with rapidly characteristics is given in table 1. Considering the way a
advancing techniques that rely on the analysis of FDCs, such torsional CL deformation is induced, five major groups of
as molecular pulling [12, 17]. calibration methods can be identified. Essentially, it can be
Another need for the development of accurate and reliable induced by direct application of a force acting at a position
calibration techniques is to understand and quantify the off the long axis of the CL [17, 29, 33–38], by loading the tip–
fundamental mechanisms governing sliding friction. The substrate contact in a lateral direction [28, 39–41], by loading
existing lack of understanding means that friction modelling a compliant structure of known stiffness [42–44], by scanning
frequently involves parameters such as surface energy or across a surface and utilizing the frictional forces [20, 45–50]
interfacial shear strength. Essentially, they are used as fitting or by exciting the torsional CL resonance [51, 52].

2
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Table 1. Overview of a number of approaches for the calibration of lateral forces measured in AFM. Aiming at an introduction, the
overview is not thought to be comprehensive. The quantities to be determined are the torsional CL spring constant and the sensitivity of the
PSD. Rather than the torsional CL spring constant, some methods yield an overall lateral calibration constant that depends on the sensitivity
of the PSD.
Group of
calibration Major Main source
methods Approach advantage of uncertainty Sections
Direct application of a A force is applied directly to No friction model Length of the lever arm;
force off-axis to the long the beam or via the equator involved the tip height needs to be 4.2, 4.4,
axis of the CL beam of a colloidal probe, an offset measured separately and 5.1
(static loading) between the CL shear centre it may differ from the
and the point of the force effective tip height
results in a torque

Compliance (static loading) Measuring lateral forces while No friction model Lateral deformation of
loading the tip–substrate con- involved, applying the tip (unless a colloidal 4.3
tact in the lateral direction the lateral forces probe is used) or of the
(static friction) or running a via the adhesive joint (if a
FDC on a sloped surface tip–substrate colloidal probe is used)
contact

Suspended platform Applying a lateral force to a Applying the Stiffness of the


(scanning method) platform suspended by arms lateral forces via suspension springs; 5.3, 5.4
of known stiffness, operation the tip–substrate friction law if the
either in the static or in the contact method relies on sliding
dynamic friction regime friction

Wedge (scanning method) Scanning across two sloped Applying the Analysis relies on a
surfaces and analysing the lateral forces via friction model, in 6.1
friction loop the tip–sample particular on Amontons’
contact law

Torsional resonance Analysis of the fundamental Non-contact Accurate description of


(non-contact method) torsional resonance of the CL technique, i.e. no the interaction with the 7
risk of tip damage surrounding medium or
attached masses

It should be emphasized that table 1 is thought to give an the wedge calibration methods [20, 45–47], and the parallel
introduction to the major groups of calibration methods. In this scan methods where the scan movement is parallel to the
sense it is not intended to be comprehensive. A more detailed CL long axis [48–50]. In particular, the wedge method
table of the particular methods is provided in section 7. is frequently used as an experimental reference technique
One obvious way of classification of the available methods [36, 40, 42].
is the involvement of a sliding motion, i.e. whether or not the Considering that these methods are based on empirical
calibration method relies on a scanning motion between tip and relationships between friction force and normal load, Asay et al
sample [40]. Among the scanning calibration techniques are [40] classified them as indirect. Thus, the lateral forces causing

3
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

the CL twist are not directly known or measured. For instance, the CL are introduced and discussed. In section 8, an overview
techniques [20, 40, 45, 46, 48, 49] based on Amontons’ law and a discussion of the various methods are given. A table is
[21] involving the friction coefficient, µ, assume that its value shown that provides a schematized overview of the methods
is constant over time; however, in the case of sharp AFM tips it and their key characteristics. Furthermore, it indicates the
is likely to change in the course of the experiment. In particular, particular quantity provided by each method and the reported
the tip can be affected by wear, and an increase in tip radius uncertainties. For the purpose of clarity, such details are not
entails increases in contact area as well as in total adhesion typically provided in the main text. Rather, the main text
force. aims at an outline of the basic laws involved in the various
Another point of criticism is that the friction law can be a approaches. Finally, in section 9 some concluding remarks
matter of the tip shape and properties. A friction force scaling are made and an outlook on potential future developments is
in a linear manner with the load, as given by Amontons’ law, given.
is known to hold for multi-asperity contacts only, whereas
contacts with a single-asperity are more likely to follow a 2. Basic definitions and relationships
power law, as given by the Johnson–Kendall–Roberts (JKR)
or the Derjaguin–Muller–Toporov (DMT) model [53–56]. 2.1. The calibration constants involved in a lateral force
A classification of the calibration method, in terms measurement
of direct or indirect, has been applied by various authors,
however, with slight differences in meaning. In line with The quantitative analysis of LFM data requires knowledge of
the elucidation given above, a calibration method can be the factor linking the measured signal either with the force
referred to as direct if the force applied to the CL is known or with the deformation. In general, lateral force calibration
from basic principles or experiments, e.g. by measurement of involves two steps [29]:
the deformation of a calibrated reference spring. However, (i) the calibration of the lateral PSD response to convert the
other authors [40, 42] also used the term ‘direct’ for one-step measured signal to the twist angle of the CL and
techniques, i.e. calibration methods that yield a lateral force (ii) the calibration of the torsional spring constant of the CL
calibration factor in a single step, without the requirement that converts the twist angle to a torque.
of a separate measurement for the determination of the PSD
sensitivity [40, 41]. In these terms, the wedge method is a The related calibration factors are the torsional sensitivity
direct technique, since it delivers an overall lateral calibration of the PSD, Sφ , and the torsional spring constant, kφ . Assuming
constant and does not require separate measurements for the small deflections of the CL, the relationships between the
determination of torsional CL spring constant and lateral PSD acting forces, the deformations of the CL and the resulting
sensitivity. In contrast, Asay et al [40] classified the wedge measurement signals are of linear form. Hence, the output
voltage Vφ of the lateral direction PSD (position-sensitive
method as indirect since it involves a friction law, and Li
detector) can be written as
et al [42] classified it as a semi-direct method, considering the
facts that it is a single-step technique but that some geometric
Vφ = Sφ φ, (2.1)
variables need to be evaluated separately.
On the down side, as a result of its single-step nature where φ is the angle of torsional deformation (twist) of the
the wedge method does not yield a separate value of the CL and Sφ is the torsional angle sensitivity of the PSD (in
torsional spring constant. The overall torsional conversion V rad−1 ). It depends on the geometry of the laser beam path
factor value resulting from the wedge method also depends on and frequently also on the total signal on the PSD. Hence, Sφ
the particulars of the PSD system and its settings. Furthermore, needs to be calibrated every time the position of the laser beam
it requires knowledge of the normal CL spring constant. In this or of the PSD changes.
sense, it is not genuinely a one-step technique. Varenberg’s The linear relationship between the torsional CL
version of the wedge method relies on the measurement of deformation and the acting torque, T , involves the torsional
FDCs and their evaluation in terms of the pull-off force [46]. CL spring constant, kφ :
Similarly, the initial approach of Ogletree et al [45] requires
normal force calibration since it relies on the analysis of friction T = kφ φ. (2.2)
forces for a range of loads.
This paper is organized as follows: the basic definitions Correspondingly, kφ is measured in N m rad−1 . As a measure
of the calibration constants related to LFM as well as for the amount of deformation resulting from a certain torsion
some relationships between these constants are given in moment or force, the spring constant is a stiffness by nature.
section 2. In section 3, an overview of the available methods In principle, both the terms spring constant and stiffness can
for the calibration of the lateral sensitivity of the PSD is be used interchangeably. For the purpose of clarity and to
given. An outline of the configurations employing a force align with the standardized vocabulary for surface chemical
balance between the tip and a rigid structure is presented in analysis [57], however, it is advisable to refer to the CL stiffness
section 4, and configurations involving compliant structures as spring constant.
are considered in section 5. Methods relying on a scanning Furthermore, it is worth noting that as a quantity
movement between tip and substrate are described in section 6. characterizing the CL mechanical properties, kφ is more
In section 7, the methods related to the torsional resonances of fundamental than Sφ . By definition, it is independent of

4
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Table 2. Equations that define the calibration constants related to normal and torsional deformations. The vertical displacement and the
twist angle are denoted by z and φ, respectively, and the related normal force and torque are denoted by Fz and T , respectively. The PSD
sensitivities, Sz and Sφ , are measures for the signal changes upon deformation of the CL. The CL spring constants, kz and kφ , are measures
for the forces related to the deformations. The conversion factors, αz and αφ , are measures for the changes in the measurement signal upon
torque acting on the CL.
PSD sensitivity CL spring constant Overall conversion factor
Torsional (in terms of φ) Vφ = Sφ φ (D1) T = kφ φ (D2) T = αφ Vφ (D3)
Lateral (in terms of x) Vφ = Sx xφ (D4) Fx = kx xφ (D5) Fx = αx Vφ (D6)

Normal Vz = Sz z (D7) Fz = kz z (D8) Fz = αz Vz (D9)

Figure 1. (a) Schematic representation of the core sensing elements of an AFM instrument, consisting of a CL with integrated tip, laser
beam and segmented photodiode. In the case of a lateral force acting at the apex of the tip, the length of the torsional arm is given by
H = R + h + t/2, where R is the radius of curvature of the tip apex, h is the tip height and t is the thickness of the CL beam. (b) Schematic
representation of the calibration triangle, illustrating the relationships between the applied forces, the resulting deformations and the
measured signals. Such relationships can be applied to both the flexural and the torsional CL deformations.

the particulars of the read-out system for the CL deflection. tip (figure 1(a)). If the calibration of the system has been
Knowledge of the overall conversion factor α (table 2) is performed under such a condition, then the values of all three
sufficient to convert the readout values into force units. calibration constants (table 2) are effective ones. In particular,
However, α changes with every repositioning of the laser this also applies to the torsional spring constant, kφ , since the
focus or of the segmented photodiode. Thus, quantitative amount of deflection at the sensing position of the laser beam
comparison of various CLs or comparison with alternative is different from the amount of deflection at the position of
calibration techniques requires knowledge of the torsional the tip. However, a differentiation between the true and the
spring constant, kφ . Notably, the overall conversion factor, effective values of the torsional spring constant is unnecessary
α, is representative for one-step calibration techniques that do if it has been determined using an alternative method that is
not involve a separate measurement for the determination of not affected by such an offset.
the PSD sensitivity (section 1). An overview of the defining equations is given in table 2,
Furthermore, it should be noted that the position of the and a schematic representation of the mutual relationships of
laser spot on the CL can differ from the position of the the calibration constants is shown in figure 1. For the purpose

5
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

of distinction from the other equations, the defining equations (iii) Furthermore, inserting equations (2.4) and (D1) into
are referred to by the numbers (D1)–(D9). equation (D5) yields an expression for the lateral force, Fx ,
These definitions establish a loop that can be represented in terms of the voltage Vφ :
by a triangle (figure 1(b)). Once two calibration constants are kx H kx
known, the third one can be calculated. Combination of the Fx = Vφ = Vφ . (2.9)
S Sx
equations given in table 2 yields the following relationships
between the calibration constants: (iv) The combination of equation (2.4) with the expressions for
the torsional and the lateral spring constants (equations (D2)
kz
αz = , (2.3a) and (D5)) as well as with equation (2.6) results in the following
Sz relationship:
k φ = H 2 kx . (2.10)

αφ = . (2.3b) Using this equation in combination with equation (2.3b), the

lateral CL spring constant, kx , can be written in terms of the
A torque, T , is applied to the CL by the lateral force, Fx , acting torsional conversion factor, αφ :
via a lever arm of length H , between the CL neutral axis and
αφ Sφ
the contact point. The torque causes a CL twist, and for small kx = . (2.11)
values of the twist angle, φ, the related displacement at the H2
tip–sample interface, xφ , can be approximated by (v) The lateral conversion factor, αx , is related to the torsional
conversion factor, αφ , by the relationship:
xφ = H φ. (2.4) αφ
αx = . (2.12)
In the case of a symmetric tip position and a symmetric CL H
beam, it acts as a torsional arm of length H , which is given by Equations (2.8), (2.10) and (2.12) provide direct links between
the sum of the tip radius, R, the tip height, h, and half of the the relationships based on the torsional angle, φ, and the
CL thickness, t, i.e. H = R + h + t/2 (figure 1(a)). relationships based on the lateral displacement, xφ .
In addition to the torsional CL spring constant, kφ , also a (vi) For the convenient comparison of the normal and the
lateral CL spring constant, kx , can be defined: lateral conversion factors, both the lateral and the normal PSD
sensitivities can be defined in terms of the angular deflections
Fx = kx xφ . (2.5) at the position of the tip. Denoting these angles with φ and
η, respectively, the PSD sensitivity, Ŝz , related to the angular
Further to these basic equations, a couple of related equations deflection, η, resulting from the vertical tip displacement,
and comments shall be given: z, is defined by
Vz = Ŝz η. (2.13)
(i) In vector representation, the torque, T , acting on the CL is
given by the vector product of the distance vector, q , and the This alternative PSD sensitivity, Ŝz , is measured in V rad−1 .
lateral force, Flat : For a rectangular CL that is uniform along its entire length
T = q × Flat . (2.6) and that has the tip at a distance L from the clamped end, the
relationship between z and η is given by [58]
Hereby, q is the distance vector between the tip apex and the
z = 23 Lη, (2.14)
CL long axis. Assuming a perfectly symmetric CL, the tip apex
is situated exactly beneath the CL long axis, and the CL shear Thus, the normal force, Fz , can be written as
centre is situated on its long axis. In this case, the amount
of torque can be written as T = |q × Flat | = qFx . As the Fz = kz z = 23 kz Lη ≡ k̂z η, (2.15)
force acting perpendicularly to the tip, the lateral force, Flat ,
with the alternative spring constant, k̂z , in N rad−1 . Similarly,
is horizontal if q is vertical. Then it can be identified with Fx ,
an alternative formulation for the normal PSD sensitivity can
the force acting along the x-axis. As discussed in section 4.5,
be given by
this does not hold in the case of a positional offset between tip
and CL shear centre. Vz 3 1 Vz 31
Sz = = ≡ Ŝz . (2.16)
For the case of a perfectly symmetric CL, q is vertical, and z 2 L η 2L
equation (2.6) can be written in terms of the total tip height, H :
(vii) Assuming validity of Amontons’ law for the given tip–
sample system, the lateral force signal, Vφ , and the CL twist
T = H Fx . (2.7)
angle, φ, scale in a linear manner with the applied load,
(ii) The total tip height, H , represents the lever arm along which Fz [59]. Thus, when writing Vφ = mFz and φ = m̃Fz
the lateral force, Fx , generates a torque and causes a CL twist. with the constant factors m and m̃, the defining equation for
the torsional PSD sensitivity, equation (D1), transforms into a
By comparing equations (D1) and (D4) and inserting equation
particularly simple expression:
(2.4), an alternative equation for the lateral PSD sensitivity can
be derived: Vφ m
Sφ = = . (2.17)
Sx = Sφ /H. (2.8) φ m̃

6
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

2.2. The lateral and the normal spring constants of a


rectangular CL
For a homogeneous CL of rectangular shape, the normal as
well as the lateral spring constant can be calculated from its
dimensions and the moduli of its material. The lateral spring
constant, kx , is given by

Gwt 3
kx = , (2.18)
3LH 2
where G is the shear modulus of the CL material. w, t and L are
the CL width, thickness and the distance from the fixed end of
the CL to the position near its free end where the tip is situated.
For an isotropic and homogeneous material of Poisson’s ratio
ν, the shear modulus, G, and Young’s modulus, E, are coupled Figure 2. Plot of the stiffness ratio kx /kz versus the CL length, L.
via the equation G = E/ (2 (1 + ν)) [60]. Using equation (2.20), the ratio was calculated for various values of
the tip height, H .
Similarly to kx , the normal CL spring constant, kz , can be
calculated from
Ewt 3 in terms of Young’s modulus, E, by employing the equation
kz = . (2.19) G = E/(2(1 + ν)) [60]; however, this approach is valid only
4L3
for CLs made of isotropic materials. Indeed, in the case of
Herewith, E denotes Young’s modulus of the CL material.
single crystal Si CLs this approach was found to result in
Analytical equations are more difficult to derive for the
large uncertainties, as compared with the case of CLs made
case of V-shaped CLs. Neumeister and Ducker [61] used
of Si3 N4 [51].
approximations such as a triangular plate or two prismatic
Due to the scaling of kx and kz with t 3 (equations (2.19)
beams.
and (2.20)), any method involving thickness measurements
Importantly, the pronounced dependence of kz and kx
is inherently error-sensitive [65]. For applications beyond
on the CL thickness, t, means that any uncertainty in the
the quick estimation of the CL spring constant, mitigation
measurement of t is bound to entail a large uncertainty in
strategies should be applied, such as the replacement of the
the CL spring constants. Similarly, spatial variations in
CL thickness with a parameter related to its effective thickness
the CL thickness can lead to a large difference between the
[66]. Considering that the fundamental flexural resonance
calculated and the real CL spring constants. For instance, frequency is directly related to the CL spring constant, it is
spatial variations in the CL thickness or in its shear modulus evident that it can be utilized as a measure for its thickness
can result from imperfections related to irregularities in the [51, 66]. From the Euler–Bernoulli equations for a rectangular
processing steps involved in the CL fabrication [62]. and homogeneous beam, an equation for the CL thickness can
A combination of equations (2.18) and (2.19) yields the be derived [41, 67, 68]:
ratio of the lateral to the normal spring constant:

 2 ωn ρ
kx 4G L t = 2 12 . (2.21)
= . (2.20) Kn E
kz 3E H
Herewith, ωn is the nth flexural resonance frequency, Kn is
The ratio kx /kz is independent of the CL dimensions w and the related wavenumber and ρ is the mass density. For the
t, since both spring constants scale with wt 3 . With G/E in fundamental mode (n = 1), the related value of the wave
the order of 1 and L/H  1, the value of the ratio kx /kz is number is given by K1 L0 = 1.875, with the CL length, L0 .
1. As shown in figure 2, the value of the ratio increases It should be noted that the equations given above
with the length L. Furthermore, it decreases with increasing assume a rectangular CL geometry and perfectly homogeneous
values of the tip height, H . The scaling of kx /kz with H −2 materials properties. However, if the CL geometry becomes
(equation (2.20)) accounts for the fact that the tip acts as a more complex and variations in its local materials properties
moment arm. For a higher tip a certain lateral force causes a and its coatings are significant, then the simple beam
larger torque and hence a larger torsional deformation of the mechanics approach leads to large uncertainties [64, 69].
CL. Thus the lateral CL stiffness, kx , decreases with increasing The particulars of the CL geometry are also of relevance
values of H . Since kz is independent of H , the ratio kx /kz for the amount of CL deformation parallel to its surface
shows the same scaling as kx . plane. Throughout the existing calibration methods it is
Notably, equation (2.20) can be used to calculate the lateral assumed that the lateral force acting on the tip leads purely
CL spring constant, kx , if the normal spring constant, kz , is to a torsional CL deformation. In principle, however, it
known [51, 63, 64]. However, propagation of the uncertainties can also lead to an in-plane deformation. This deformation
in kz as well as in the dimensions, H and L, and in the elastic mode can be neglected only within certain ranges of the CL
properties, E and G, can lead to a considerable uncertainty in geometry parameters, such as its width, thickness or tip height.
kx . Moreover, the shear modulus, G, is frequently expressed Considering the spring constants for torsional and for in-plane

7
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

deformation, the torsional deformation is dominating if the (i) Tilting a reflective surface. The lateral sensitivity of the
following condition is satisfied: PSD can be measured by mimicking the torsional deformation
of the CL and recording the change in the LFM signal. For
kφ instance, the angle between PSD and reflecting surface can be
λ=  1, (2.22)
kin altered by using a mirrored substrate and by tilting it by means
of a stepper motor [75]. In a similar approach, the AFM head
Herewith, kin denotes the spring constant for in-plane can be tilted [35]. By recording the associated changes in the
deformation of the CL. lateral force signal, Vφ , and using equation (D1), the PSD
For rectangular CLs with L  w, kin is given by [70] sensitivity, Sφ , can be calculated.
Provided a CL holder with an integrated lever and a hinge,
1  w 3 such measurements can also be achieved with a test CL as
kin = Et . (2.23)
4 L the reflective surface. The CL holder developed by Xie et al
[72] converts a vertical displacement into a rotation about the
Notably, this expression refers to a lateral loading
horizontal axis through the hinge. If this axis is in line with
perpendicular to the CL long axis. In combination with the
the CL long axis, then a rotation of the CL is achieved that
expression for the lateral spring constant (equation (2.18)), the simulates a torsional deformation of the CL.
ratio λ is given by A highly linear relationship between the PSD response
 2 and the tilt angle was observed by Rutland’s group [17, 35].
4G tL The laser spot position on the PSD did not show a significant
λ= . (2.24)
3E Hw effect on the PSD sensitivity [17]. On the other hand, Cannara
et al [36] found a parabolic dependence of the lateral sensitivity
Clearly, the condition λ  1 (equation (2.22)) is satisfied if on the lateral laser spot position. Such findings are reminiscent
t  H and L  w, since the ratio G/E is in the order of the fact that the PSD response depends on the particulars
of 1. This simple consideration shows that for CLs with of the instrumental design. Thus, it should be borne in mind
L  w, equation (2.22) does not necessarily hold and that the that the instrument to be used for LFM measurements should
assumption of negligible in-plane deformation is invalid [70]. be thoroughly tested [17]. As pointed out by Asay et al
For instance, with H = 10 µm, t = 2 µm, L = 400 µm, [40], in many cases the assumption of a negligible optical
and w = 40 µm, the ratio λ has the value of 2.13, if the crosstalk does not hold, thus leading to a poor reproducibility
shear modulus is replaced with G = E/(2(1 + ν)) [60] and of measurements performed in different laboratories.
if ν = 0.25. For a CL of length L = 200 µm, however, the
value of λ is 0.53, i.e. < 1. (ii) Geometrical optics method. As an alternative to tilting
Furthermore, it is worth mentioning that also the the AFM head or a mirrored substrate, the PSD response can
torsional resonance can be utilized for lateral force calibration be measured as a function of its displacement, i.e. by moving
(section 7). Similar to the flexural resonances of the it along its two major axes [73]. The PSD sensitivity can be
CL, torsional resonance frequencies also depend on the calculated by means of a suitable model for changes in the beam
CL spring constant, mass distribution and dimensions. A path as the CL bends and twists [28]. Using such an approach
comprehensive review of the CL dynamic properties and the based on geometrical optics, Liu et al [73] converted the PSD
related dynamic AFM operation modes has been given by Song displacement into an equivalent lateral tip displacement. The
and Bhushan [71]. method requires knowledge of the tip height and of the lateral
CL spring constant, kx . The equation derived by Liu et al [73]
relates the lateral force, Fx , to the friction loop width, W . It
3. Calibration of the lateral sensitivity of the PSD involves the sensitivities of the PSD for vertical and horizontal
movements [28, 73].
3.1. Available methods As reported by Cain et al [28], the sensitivity for horizontal
PSD movements can depend on the level of the bending signal,
Various methods for the measurement of the lateral deflection Vz , i.e. on the setpoint value of the topography feedback.
sensitivity have been suggested. In the simplest case the PSD In such a case, the scaling of raw friction data, Vφ , is a
is assumed to be rotationally symmetric, i.e. to be equally function of the load applied during the friction measurement.
sensitive for normal and for torsional angular deflections. In A similar dependence was also found for friction versus load
terms of the PSD sensitivities defined in table 2, this approach data measured on a polished SrTiO3 (3 0 5) surface.
can be expressed as Sx = Sz . However, this is not likely In many cases, the measurements need to be undertaken
to be the case. Possible reasons for different values of the with the sample surface covered by a liquid, thus requiring
normal and the lateral PSD sensitivities are asymmetries in the the usage of a fluid cell. In particular, this is more the rule
shape of the reflected laser spot [41] and different electronic than the exception when it comes to experiments involving
amplification factors for the normal and the lateral PSD signals. biological or biochemical specimens. With a refractive index,
Hence, dedicated methods for the calibration of the lateral PSD n, of the liquid different from the one of air (n ≈ 1), it
sensitivity are required. An overview of the available methods affects the path of the laser beam. When comparing the
is given in table 3. results for air and water, Pettersson et al [17] observed a

8
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Table 3. Overview of calibration techniques for the PSD sensitivity.

Group of
calibration Major Major sources
methods Approach advantages of uncertainty References
Tilting a reflective surface The laser beam is Non-contact technique Different conditions of [29, 35, 72]
reflected off a mirrored (i.e. no risk of tip the reflective surface
substrate or a reflective damage) (distance from the
CL and the PSD output mirror, shape of the
measured as a function reflected laser beam
of the substrate tilt angle cross-section); different
neutral axis

Geometrical optics Geometrical analysis of Non-contact technique Non-linearities in the [28, 73, 74]
the optical beam path (i.e. no risk of tip detector system
and measuring the signal damage) (including mechanical
changes upon backlash in the PSD
displacement of the PSD positioner)

Lateral FDC Pressing a colloidal A standard calibration Length of the torsional [36]
probe (test probe) grating can be used; no moment arm, both for
against a vertical and risk of damage to the tip the test and the target
rigid step, recording the of the target CL since a probe
changes in the LFM test probe of similar
signal as a function of mechanical and optical
the lateral displacement properties is used
(lateral FDC)

Scanning across a vertical Scanning an integrated Requires no particular Assumes that the global [59]
step tip across the vertical experimental measures, tip shape applies to the
step of a calibration only a standard tip apex; relies on the
grating calibration grating and a readout of the peak
sharp tip voltage in the lateral
force signal

significant change in the sensitivity, Sφ . Hence, the PSD substrate, i.e. by a purely normal loading, an offset in the
calibration needs to be performed under the same conditions tip position results in a torque and a corresponding torsional
as the real measurement. More recently, Tocha et al [74] CL deformation. However, for typical CLs the vertical tip
presented a geometrical analysis accounting for the different displacement related to the torsional deformation of the CL,
refractive indices of air and the liquid contained in the fluid zφ = z − znorm , is comparatively small, owing to a large
cell. They derived a correction factor that relates the PSD ratio of torsional and normal CL spring constants, kφ /kz  1.
response for measurements in air and in the liquid to be used. Bogdanovic et al [35] reported that the relative amount of
This correction factor is given by the ratio of the refractive zφ was less than 10%. Nevertheless, the occurrence of such
indices of the two media [74]. mechanical crosstalk (section 4.5) underscores the necessity
Alternatively to the approach taken by Liu et al [73], of placing the colloidal probe as close as possible to the long
torsional CL deflections can also be generated via an offset axis of the CL.
in the tip position. In the case of an AFM measurement,
such a positional offset with respect to the CL long axis (iii) Lateral FDC method. Approaches relying on geomet-
causes an apparent frictional force even without any sliding rical optics or a mirrored substrate circumvent the usage of
movement. In the course of the calibration of the PSD response an AFM CL. However, under certain conditions the calibra-
for CL deflection by going into contact with a hard and stiff tion can also be achieved using an AFM probe. Given a

9
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

well-defined contact geometry and a rigid structure, a lateral quadratic in the CL tilt angle measured on the flat area [59].
displacement, x, can be assumed to be fully converted into Provided knowledge of the torsional CL spring constant, kφ ,
a torsional deflection of the CL beam. The approach chosen the quadratic equation delivers a value for the CL twist angle,
by Cannara et al [36] takes benefit of the sphere-like shape of and combination with equation (D1) yields the torsional PSD
the colloidal probe. By approaching it to a vertical step, they sensitivity.
recorded lateral force–distance curves (lateral FDCs). Accord- In their experimental approach, Choi et al [59] used
ing to equation (D4), the required sensitivity, Sx , is given by rectangular CLs and calculated the kφ values via the linear-
the slope of the linear compliance regime of such lateral FDCs elastic continuum-mechanical model, i.e. employing equations
(see sections 4.4 and 5.1). (2.18) and (2.19), with the CL thickness measured indirectly
In greater experimental detail, a colloidal glass sphere via the torsional resonance frequency of the CL.
of ∼70 µm diameter was attached to a Si3 N4 test CL of It should be noted that this method assumes a conically
40 µm width. A lateral force–displacement measurement was symmetric global tip shape only, as expressed in terms of a tip
achieved by approaching a vertical step on the surface of opening angle, ψ. In reality, however, the local tip shape can
a freshly cleaved GaAs wafer [36]. Owing to its cleavage significantly deviate from the shape of the global tip (tip shank).
behaviour, a plane at a perfect 90◦ angle to the (1 0 0) surface Typically, the local tip shape is accounted for by describing
can be generated, the (1̄ 1 0) plane of the GaAs crystal. the tip apex in terms of a sphere of radius R. The approach
In order to allow for calibration of standard CLs with an involving exclusively the global tip shape is more likely to
integrated tip or of colloidal probes with a small diameter, deliver useful results in the case of very sharp tips, i.e. for very
they suggested a procedure utilizing the colloidal probe as a small values of R.
test probe. Under the assumption of an unchanged condition of
the PSD, the sensitivity can be calculated for a CL of arbitrary (v) Reverse calculation. Considering the mutual relationship
length provided that the sensitivity was measured for the test between the PSD sensitivity, the CL spring constant and the
probe. However, the width and reflectivity of the target CL overall conversion factor as given in equations (2.3a) and
should not deviate significantly from those of the test probe. (2.3b), respectively, the PSD sensitivity can be calculated if
A major advantage of the test probe method is that the risk of the two other calibration constants are known. For instance,
damage to the tip of the target CL is eliminated. when using the torsional Sader method (section 7) to determine
It should be mentioned that in addition to the torsional kφ and a direct method (section 4.2) that generates a torque
response, also lateral in-plane bending may occur. Since and delivers the overall conversion factor, αφ , then the PSD
a lateral movement of the CL surface has no effect on the sensitivity, Sφ , can be calculated. Such an approach was
angle of incidence at the position of the laser spot, the laser demonstrated by Pettersson et al [17].
deflection technique is not sensitive for the in-plane bending
of the CL. Owing to the fact that the in-plane bending is 3.2. Optical crosstalk
negligible for typical CLs with w  t, however, the torsional
CL deformation dominates and it is justified to rely on the laser A major complication of LFM measurements can result from
deflection technique. Nevertheless, this assumption should be a rotational misalignment of the PSD [76, 77]. Such a
validated for the particular CL under consideration [70]. Hence misalignment causes crosstalk between the PSD signals related
both the in-plane bending of the test and the target CL should to the normal and the torsional deflections of the CL. It
be negligible. Under the assumption that the calibration is leads to systematic errors in lateral force as well as in height
independent of the laser spot position if it is situated between measurements (see figure 3). In particular, such crosstalk can
the tip and the free end of the CL, Cannara et al [36] provided be dominating in the case of nano-scale structures featuring
an equation to correct for the in-plane bending of the target significant variations in frictional properties [77, 78], i.e. the
probe. superposition of significant variations in the lateral force signal
Furthermore, calibration of the target CL requires onto an almost constant normal force signal.
accounting for differences in the torsional moment arm length It should be pointed out that the crosstalk related to
as well as in the total signal on the PSD. a misalignment of the PSD (optical crosstalk) needs to
be differentiated from mechanical crosstalk, resulting from
(iv) Scanning across a vertical step. In the search for a a positional offset of the CL shear centre (section 4.5).
simple calibration technique without any special experimental Furthermore, it should be borne in mind that crosstalk can
requirements, Choi et al [59] suggested a method involving also result from signal mixing effects in the electronic parts of
only a standard CL and a grating featuring vertical steps. the AFM instrument [79].
Similar to the approach of Cannara et al [36], the method Crosstalk between the lateral and the normal responses is
is based on the changes in lateral force upon contact with a likely to occur, and for many AFM instruments it is difficult
vertical wall. However, the method of Choi et al [59] employs to eliminate. For instance, if the PSD is not oriented strictly
a sharp tip rather than a colloidal probe and considers the lateral perpendicular to the incident laser beam, the projection of the
forces at two distinct points of contact, namely a point on the beam onto the PSD surface is distorted [76, 79]. In this way,
flat area and the point at the upper edge of the step. At the the spatial distribution of the light intensity on the PSD surface
latter, the lateral force signal is assumed to peak. does not exactly reflect the deformation state of the CL.
An expression for the CL twist angle can be derived from Experiments by Cannara et al [36] have shown a decrease
the balance of the torsional moments. This expression is in the lateral deflection sensitivity, Sφ , with increasing offset

10
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 3. Schematic representation of the optical crosstalk effect. (a) The ideal case of a perfectly symmetric and centred laser spot. The
segmented photodiode comprises four segments, I–IV. (b) The laser spot is shifted along the x-axis, thus indicating a torsional deformation
of the CL. (c) Result of a misalignment of the segmented photodiode. The misalignment of the PSD is given by the angle γPSD . In the
coordinate system associated with the segmented photodiode, (l, n), the vector representing the lateral shift of the laser spot has a non-zero
component in the vertical direction.

of the PSD position. Approximately, the observed dependence detection system consisting of CL and the optical beam path
was of a quadratic form. This finding was attributed to the to enable a simulation of the LFM system and to perform
Gaussian distribution of the laser spot. That is, the PSD parametric studies. The model involves a finite element
output may also be affected by the intensity distribution of the analysis (FEA) of the CL deformation and matrix algebra to
reflected laser spot, depending on its extent as compared with describe the path of the reflected laser beam. The access to
the area of the PSD. In order to reduce such non-linearities, it parametric simulations allows for analysis of the effects of the
is advisable to avoid large offsets between the laser spot and system variables. In particular, the effect of misalignments
the centre of the PSD. in the optical beam path on the measurement signals can be
Prunici et al [79] analysed the amount of optical crosstalk studied in a systematic manner. Hence, crosstalk effects can
in terms of a misalignment factor deduced from the changes be identified more easily and the consideration of crosstalk
in friction loop with sample height. They observed strong coefficients allows for quantification.
changes in their misalignment factor with changes in the
position of the reflection point on the CL. In particular, for
narrow CLs strong variations were found, due to the amount 4. Methods relying on a force balance upon contact
of laser light not reflected by the CL. with a rigid structure
In the case of a scanning-by-probe (SBP) scheme, optical
4.1. Normal loading upon contact with a sloped substrate
crosstalk can occur even in the case of a free CL, i.e. when it
is in its undeformed condition [80]. In contrast to a scanning- A torsional deformation of the CL can be generated in various
by-sample (SBS) scheme, the scanning and the measurement ways. In addition to the typical case of a frictional force
functions are not separated. If the axes of the scanning system resulting from sliding motion, a lateral force can also be applied
are not properly aligned with the axes of the PSD, the reflected to the tip via the contact with a sloped surface, i.e. without
laser spot does not move along the PSD axes even if the scanner application of a scanning motion. Asay et al [40] utilized
motion is purely along one of its inherent axes. Varenberg such a static force balance to determine the lateral force at
et al [80] devised an electronic circuit to compensate for the any applied load. Two major advantages of such non-scanning
linear crosstalk characteristic observed when approaching the calibration techniques are that no frictional law is involved
CL to the sample surface. (the validity of which may depend on the particulars of the
Such an electronic compensation scheme allows for tip–sample interaction) and that the risk of tip degradation by
real-time correction of the PSD signals. However, with wear is minimized.
most commercial AFM systems these signals are not readily The characteristic quantities can be extracted from the
available and such a scheme can be difficult to implement by changes in lateral force upon approach of a ridged calibration
the AFM user. Yet corrections can also be achieved through grid featuring facets of the slope angles 0 and +θ.
an analysis scheme that quantifies the amount of crosstalk and Torsional deformation results from any force acting
can be used to adjust the experimental setup [79] or through a perpendicular to a moment arm. As can be seen from figure 4,
data processing scheme that can be applied retrospectively to for an applied load, FL , and a slope angle, θ, the friction force,
the measurement data. Ff , is
A data processing scheme based on an affine transforma-
Ff = FL sin θ. (4.1)
tion has been presented by Onal et al [81]. It delivers a trans-
formation matrix that is applied to the vector of the measured Assuming a perfectly symmetric CL and a zero tilt angle of the
voltages Vz and Vφ . The components of this matrix result from tip, i.e. φ = 0, the torque can be calculated from the total tip
measurements where the AFM tip is glued to a flat substrate height, H , and the horizontal force, Fx (equation (2.7)). The
and the changes in Vz and Vφ are recorded as a function of the latter is given by the sum of the horizontal components of FN ,
displacement. FA and Ff (see the horizontal dashed line in figure 4):
A similar approach has been chosen by Michal et al [82],
although with a different aim. They modelled the complete Fx = (FN − FA ) sinθ − Ff cos θ. (4.2)

11
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 5. Schematic signal–distance curves for the case of a FDC


measured on a sloped surface. The changes in the normal signal,
Figure 4. Vector analysis of the forces acting on the AFM tip when
Vz , with the vertical displacement are shown in the upper plot. The
loading a slope of angle θ . The point XG denotes the intersection
corresponding changes in the lateral force signal, Vφ , are shown in
with the CL long axis.
the lower plot. In the regime of the FDC contact branch, Vφ scales
with Vz , and its sign is given by the sign of the slope angle, θ .
Using this expression for Fx and equation (4.1), Sx can be
calculated from the defining equation, equation (D4): grid featuring trenches with sloped edges and by following
Vφ Vφ the procedure given by Varenberg et al [46], Asay et al [40]
Sx = = kx compared their results from their non-scanning technique with
xφ Fx
Vφ the results from the wedge method (section 6.1). Reportedly,
= kx . (4.3) they found reasonable agreement both for the lateral PSD
(FN − FA ) sin θ − FL sin θ cos θ
sensitivity, Sφ , and for the friction coefficient, µ, measured
This equation holds for the ideal case of a perfectly on the horizontal surface.
symmetric CL. Potential experimental complications are a slip-like
Importantly, any torsional deformation of the CL involves movement of the tip while recording a FDC and a change
a rotation of the tip axis and a corresponding slip in the tip– in contact point between tip and facet surface. A stick–slip
substrate contact. This means that the force analysis needs transition should be recognizable from a step-like non-linearity
to consider the condition of kinetic rather than static friction. in the FDC. Since nano-scale slipping would imply slight
Thus, a no-slip condition leads to erroneous results. In a more variations in contact point and moment arm, it is advisable
recent publication, Asay et al clarified this point and provided to check the FDCs for the presence of stick–slip effects.
a revised analysis including kinetic friction [83]. This revised
analysis is meant to replace the one given in [40]. It should be
4.2. Normal loading with the contact point off the CL long
pointed out that the models given in these publications do not
axis
include adhesion forces.
A positional offset of the AFM tip causes an asymmetry A force applied at a lever arm distance, a, from the CL
with respect to the curve measured on the flat area (facet 2), i.e. symmetry axis causes a CL with a horizontal lever attached to
in a plot of the voltage Vφ versus
 the
 z-position
  the voltage it to deflect flexurally as well as torsionally (figure 6(a)). A CL
differences are not identical, Vφ1  = Vφ3  (figure 5). with a large offset between its centre axis and the tip position
Since the resulting lateral signal shifts in the same manner for may lend itself to a lateral force calibration experiment;
positive and for negative slopes, the optical crosstalk can be however, a general calibration method should be applicable
described in terms of an offset voltage, Vφ0 , in the lateral PSD to any CL.
signal. Hence, the voltage representing the degree of optical
 Provided a CL similar to the target CL (i.e. the CL to be
 
crosstalk,
 Vφ0 , is given by the arithmetic mean of Vφ1 and used for quantitative LFM measurements), it can be facilitated
Vφ3  [40]. for lateral force calibration by attaching a horizontal lever to
The data given by Asay et al [40] indicate that both the upper side of the CL beam. Hence a normal force can
the moment arm angle, ε, and the substrate mounting angle, be applied to the free end of the horizontal CL, resulting in a
θ0 , have a minor effect on the lateral calibration factor, Sφ . torque [29, 34]. A schematic representation of the approach
Furthermore, by using a commercially available calibration is shown in figure 6. The schematic (figure 6(a)) depicts the

12
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 6. (a) Schematic of the calibration method involving a horizontal bar attached to the CL beam. Using a piezoelectric transducer, it is
lifted at a distance, a, from the CL long axis and a torque is applied to the CL. The distance, a, defines the torsional arm length. (b) Plot of the
lateral signal, Vφ , versus the vertical displacement, z. Lifting the right side of the CL leads to voltage changes, Vφ , of the opposite sign.

situation of a lever arm of length a > w/2, where w denotes a fibre to the CL, with its axis perpendicular to the CL long
the width of the CL. axis. In contrast to a crossbar used by Feiler et al [29] and
Using equations (D7) and (D8), the applied force Fz can Reitsma [34], the fibre was mounted asymmetrically to cause
be written in terms of the related voltage change in the PSD an imbalance in the gravitational pull. The resulting CL twist
output, Vz : was quantified by approaching the assembly to a flat substrate
kz and measuring the distance required to level the fibre.
Fz = Vz , (4.4)
Sz In general, a torque can be generated by any loading point
where Sz is the normal force sensitivity. The torque resulting off the CL long axis. In contrast to the case a > w/2 where an
from the force Fz applied at a distance a from the CL long axis external loading point is generated by attachment of a crossbar
is given by or by a special CL design, a loading point with a < w/2
T = Fz a. (4.5) is provided by a tip or probe at a position off the CL long
axis. Using a colloidal particle as a probe, such a configuration
In combination with equations (D3) and (D9), this equation has been employed by Quintanilla and Goddard [38]. As the
delivers an expression for the ratio of the two voltages read out main sources of uncertainty, they identified the particle offset
from the PSD: distance and the alignment of the CL with the plane given by
Vφ αz
= a. (4.6) the laser beam path. Although such a positional offset of the
Vz αφ
probe can be beneficial for purpose of calibration, the coupling
Similar to the approach of Feiler et al [29], Reitsma [34] between normal and lateral forces impairs the orthogonality of
applied a lever technique by using a CL with a crossbar attached the related signals necessary for a facile interpretation of LFM
to it, i.e. a bar across both the left and the right edge of the CL. measurements (section 4.5).
This prototype is referred to as ‘hammerhead’ CL [34]. The
Si cross beam was glued to the free end and at right angles to 4.3. Lateral loading of a horizontal surface
the long axis of the CL.
A torque can be applied via a tungsten sphere glued to a When bringing the AFM tip in contact with a stiff sample
ramp chip attached to a piezoelectric actuator. When lifting surface and loading the contact in the lateral direction, the
the ramp chip, the sphere presses against the left or the right contact region is sheared and the CL shows a torsional
end of the cross beam. The distances, aL and aR , of the deformation. In the sticking regime, i.e. before the onset of
contact points between sphere and hammerhead from the CL sliding motion, these two deformations add up to the lateral
central axis were ∼73 µm [34]. When plotting their data displacement applied to the tip–sample contact. This condition
according to equation (4.6), clearly Reitsma [34] observed a of static friction corresponds to the initial slope of a friction
linear relationship, with the curve passing through the origin. loop. In the limit of a very stiff tip–sample contact, the
The latter finding indicates a proper alignment between the lateral displacement is completely converted into a torsional
CL and the detection system. Furthermore, the distance of the CL deformation. Hence such a measurement can be exploited
sphere–lever contact from the CL long axis was identified as a to determine the lateral deflection sensitivity [39].
main source of uncertainty [34]. Considering that in the case of sharp tips also a significant
Reitsma [34] found that the precision was increased by a tip deformation can occur [32], it is recommendable to make
factor of 5 as compared with the single-point method of Feiler such a measurement using a large colloidal probe and a stiff
et al [29]. Micro-fabricated CLs of a hammerhead-like shape substrate. That is, a configuration should be chosen that
are now commercially available [84], and in the future similar ensures that the torsional CL deformation dominates. Since
CLs could be dedicated to lateral force calibration. the lateral stiffness of the tip–sample contact increases with the
A torsional deformation of the CL can also be generated tip radius, R, the condition of a stiff tip–sample contact is met
by an asymmetric mass distribution. Toikka et al [33] attached with such a probe as well, i.e. the total lateral deformation is

13
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

With increasing sphere size, the ratio xc /xφ is 1 and


H 2 kx,tot approaches kφ .
The experimental results given by Cain et al [39], however,
departed from the asymptotic behaviour of H σ expected for
the limit of high loads. In fact, the product H σ was found to
decrease. Since it is unlikely that the lateral CL spring constant
or the lateral contact stiffness decreases with increasing load,
other effects need to be considered, such as the finite stiffness
of the adhesive used to fix the colloidal probe. For larger
values of the sphere diameter, H 2 kx,tot was found to approach
kφ more rapidly with increasing load [39]. The torsional spring
Figure 7. Schematic representation of the compliance method. constant, kφ , was calculated using finite element analysis
When operating in the sticking regime, a lateral displacement of the (FEA) [39].
sample, x, causes a lateral deformation of the tip–substrate
contact, xc , and a lateral deformation of the CL, xφ . The approach of analysing the static friction regime for
the purpose of lateral force calibration was referred to as
static friction method [36, 39]. Alternatively, it can be called
dominated by the torsional deformation of the CL. In addition
lateral compliance method, reflecting the fact that it relies
to the calibration of friction forces, the technique allows for
on an analysis of the deformations resulting from a lateral
the extraction of both the lateral contact stiffness, kxc , and the
displacement.
contact shear strength, τ , from the friction trace.
For a quantitative analysis of the initial slope of the friction
loop, the deflections of two springs need to be considered. 4.4. Lateral loading of a vertical surface
With the springs arranged in series, one represents the lateral From the torsional arm length, H , related to the tip, the full
CL spring constant, kx , and the other represents the lateral length, L0 , of the CL, the distance, L, of the tip position from
contact stiffness, kxc [85]. As indicated in figure 7, the lateral the CL base, and the torsional spring constant, kφ , the lateral
scanner displacement, x, can be decomposed into the lateral spring constant can be calculated:
deformation of the contact, xc , and the lateral movement at
the interface leading to twisting of the CL, xφ : kφ L0
kx = . (4.13)
H2 L
x = xφ + xc . (4.7)
Once kx is known, the lateral conversion factor, αx , can be
The related forces are given by calculated:
Fx = kx,tot x = kx xφ = kxc xc . (4.8) Fx xφ 1
αx = = kx = kx . (4.14)
Vφ Vφ Sx
The PSD signal, Vφ , resulting from a twist angle φ of the
CL is given by equation (D1). Accordingly, the slope, σ , of Equation (4.14) results from the defining relationships given
the stick region of the friction loop is given by in table 2, namely equations (D4), (D5) and (D6).
Vφ φ If the target probe is also the test probe, the lateral in-plane
σ = = Sφ . (4.9) bending can be ignored. The uncertainty analysis performed
x x
by Cannara et al [36] shows that ignoring the in-plane bending
Combined with equation (2.4), this relationship can be written can lead to a significant increase in the error margin. Other
as follows: major sources of uncertainty are the tip height, Htarget , of the
xφ
H σ = Sφ . (4.10) target CL, and the CL width, w (due to αtarget ∝ w4 ). Another
x assumption is that the spatial distribution of reflected laser
Using equation (4.8), the ratio xφ /x can be expressed in intensity on the PSD is very similar for both CLs.
terms of the related spring constants: Comparison of the test probe technique with the wedge
kx,tot method delivered similar results, but a slightly lower
H σ = Sφ . (4.11) uncertainty in the case of the wedge method where the
kx
calculated uncertainty in the lateral conversion factor, αx , was
Provided a constant sensitivity Sφ and a constant lateral CL 4.2% [36].
spring constant, kx , H σ varies with the total stiffness, kx,tot = As a major methodical advantage of the test probe
(1/kx + 1/kxc )−1 . As pointed out by Cain et al [39], in the technique, it was pointed out that the tip of the target CL is
limit kx  kxc , kx,tot ≈ kx and the left side of equation (4.11) not in mechanical contact [36]. In particular, this approach
is equal to the scaling factor Sφ . can be valuable for functionalized or coated tips.
Using equations (2.10), (4.7) and equation (4.8),
equation (4.11) can be rewritten as follows: 4.5. Mechanical crosstalk
 
xc Inspection of many commercially available AFM CLs reveals
kφ = H kx,tot 1 +
2
. (4.12)
xφ an offset in the tip position with respect to the CL symmetry

14
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

axis. The schematic representation shown in figure 5 indicates


that such an offset results in a dependence of the lateral force on
the normal force. This effect can be employed for lateral force
calibration; however, it should be borne in mind that any LFM
measurement can be affected by such crosstalk between the
normal and the lateral forces. Notably, this purely mechanical
crosstalk effect can occur on top of the optical crosstalk effect
resulting from a misalignment of the PSD (section 3).

(i) Considering the effect of an offset in the tip position. In


contrast to a PSD misalignment (section 3.2), an off-centred
tip makes the lateral force responses on sloped surfaces non-
symmetric about the horizontal segment. As shown in figure 8,
such a positional offset with respect to the CL long axis can Figure 8. Consideration of the case of an offset, xt , in the tip
be accounted for by considering the angle, ε, of the moment position. It results in a moment arm, q , that shows a non-zero angle,
ε, relative to the vertical. The point XG denotes the intersection with
arm of length q. The moment arm is defined by the vector, q , the CL long axis. The torque due to the slope (θ < 0) of the
between the tip end and the CL centre axis, and its angle, ε, substrate counteracts the torque due to the tip offset. Here, it is
with respect to the line perpendicular to the CL beam. Forces assumed that the former outweighs the latter, i.e. the total torque is
perpendicular to the moment arm q cause CL twisting and the counter-clockwise.
corresponding torque, T , is given by
T =q×F
   
−H tan ε −Ff cos θ + (FN − FA ) sin θ
= 0 × 0 
−H Ff sin θ + (FN − FA ) cos θ
 
0
Ff (cos θ + sin θ tan ε) + (FN − FA )
=H
,
 (4.15)
× (cos θ tan ε − sin θ )
0
where θ is the slope angle (figure 8). Only the y-component
of T is non-zero. In the case of θ = 0, it is given by
H (Ff + (FN − FA ) tan ε).
The force causing the CL to twist can be calculated from
the projection of the lateral force Flat onto the axis othogonal
to the moment arm q (figure 8). Here, this force is denoted
oq oq
as Flat . In the general case of a non-zero slope, Flat depends
on both the angle ε related to the offset in the tip position and
the slope angle, θ . In the case of a negative facet (slope angle
oq
θ < 0, see figure 5), Flat is given by
oq
Flat = Ff · cos (|θ | − ε) − (FN − FA ) · sin (|θ | − ε) .
(4.16)
oq
depends on the difference
It should be noted that the sign of Flat
angle |θ | − ε. This reflects the fact that the configuration
Figure 9. Similar to figure 4, however, for the case of an offset
of figure 8 involves two contributions of opposite sign. The
between the symmetry centre and the shear centre of the CL.
momentum due to the negative slope tends to push the tip to Considering the plane of the tip cross-section, the two centres are
the right and twist the CL beam counter-clockwise, whereas represented by the points XG and XC , respectively. The offset is
the momentum due to the tip offset tends to twist the CL beam accounted for via an offset angle, γm . Furthermore, the case of
clockwise. The situation depicted in figure 8 assumes that the lateral loading is considered and the force FT is the traction
generated by the scanner.
contribution due to the slope dominates.

(ii) Considering the effect of an offset in the position of the shear a consequence, a comprehensive characterization of the CL
centre. In general, any deviation of the real CL geometry torsional properties includes a crosstalk constant, i.e. a measure
from the ideal one as well as any heterogeneity in its material for the change in lateral force upon a change in the normal load
properties can lead to an offset between the nominal CL applied to the tip.
symmetry axis and the CL shear centre. The shear centre can As depicted in figure 9, such an offset between the
be defined as the point at which any resulting concentrated geometrical centre, XG , of the local cross-section of the CL
force will not induce a torsional twist on the CL [42]. As on the projection plane and the CL shear centre, XC , can be

15
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

accounted for via the misfit angle, γm (figure 9). It contributes also its augmentation can be of interest. Recently, a new CL
to the initial tilt angle, φ0 , in the undeformed configuration, i.e. design has been devised where the tip is deliberately offset
φ0 = θ − γm , where θ denotes the slope angle of the surface from the CL long axis [87]. When operating in intermittent
feature. contact mode, the resulting torsional oscillations enrich the CL
The coordinate systems involved are the friction force frequency spectrum and can help to reconstruct the tip–sample
system, x  , z , associated with the tip–sample contact, interactions. As compared with the higher harmonics of the
the mechanical sensing system, (x, z), associated with the fundamental CL resonance, the torsional resonance peaks were
undeformed CL and the optical sensing system, (n, l), found to show a good signal-to-noise ratio (SNR).
associated with the PSD (figure 9). The relationships between
the forces and the deformation quantities can be derived by 5. Methods relying on a force balance upon contact
describing
  the forces acting in the tip–substrate contact in the with a compliant structure
x , z -system and the CL deformations in the (x, z)-system.
Application of the orthogonal transformation matrix yields an 5.1. The case of a vertical reference beam
expression for the friction force, f (κ) , in terms of the normal
and the lateral output voltages [42]: As discussed in section 3.1, the lateral PSD sensitivity can
  be determined by pressing a colloidal probe against a vertical
f (κ) = sgn (κ) αxx Vφ(κ) + αxz Vz(κ) . (4.17) and rigid wall. If this wall is replaced with a compliant beam
of known stiffness and vertical alignment, also the torsional
Herewith, αz denotes the normal conversion factor, αxx denotes CL spring constant can be measured. This approach was
the 2nd rank lateral conversion factor, and αxz denotes the demonstrated by Ecke et al [37].
crosstalk conversion factor. The prefactor sgn(κ) is equal to As depicted in figure 10(a), a beam-shaped reference
+1 or –1 for scans in the forward (trace) or in the backward lever was pushed against the equator of the probe. Such
(retrace) direction, respectively. The factors αxx and αxz of an experiment yields a lateral FDC (figure 10(b)), and the
equation (4.17) are related to intrinsic structural quantities of lateral sensitivity, Sxref , of the setup is given by the slope of
the CL, namely the misfit angle, γ m , and the reduced parameter the compliant region of the FDC. Repeating this experiment
ξ = R/H : using a rigid edge rather than a reference lever yields a lateral
force curve with a steeper compliant region. The respective
1
αxx = αx , (4.18) sensitivity can be denoted as Sx0 .
(1 − ξ ) cos γm + ξ cos θ Similar to the calibration technique for normal CL spring
(1 − ξ ) sin γm + ξ sin θ constant [76, 88], the lateral CL spring constant can be
αxz = αz . (4.19) calculated from the condition of force balance between the
(1 − ξ ) cos γm + ξ cos θ
two springs:
In the case of θ = γm = 0, the crosstalk lateral conversion Fx = kx xφ = knref x ref . (5.1)
factor is zero, αxz = 0, and the 2nd rank lateral conversion
factor is equal to αx , i.e. αxx = αx . That is, no mechanical The lateral displacements xφ and x ref sum up to the total
crosstalk between the normal and the lateral force occurs and lateral displacement generated by the scanner:
αxx reduces to the lateral conversion factor of equation (D6),
respectively. x = xφ + xref . (5.2)

(iii) Eliminating the mechanical crosstalk effect by novel design Considering equation (2.10), the lateral CL spring constant,
concepts. Considering the susceptibility of conventional CLs kx , for the case of loading at the position of the probe equator
for mechanical crosstalk, novel concepts of micromechanical is related to the lateral CL spring constant, kx , for the standard
probes have been suggested. Essentially, such studies aim case of loading at the lowest point of the probe:
at a complete decoupling of the deformations resulting from
normal and lateral forces. kx H 2
= , (5.3)
Fukuzawa et al [86] devised such a probe by combining a kx H2
double CL with a torsion beam. The torsion beam is horizontal
and supports the double CL beam. The amount of its torsional where H  = D/2 + t/2 with the diameter, D, of the colloidal
deformation is a measure for the normal force acting on the tip. probe.
The lateral force, however, is measured via the lateral bending Taking into account that in the case of a rigid counter-
of the double CL beam. Owing to its stiffness against torsional body the total displacement is accommodated by the CL
deformation, it is deflected parallel to a lateral force. Yet a deformation, the lateral sensitivity, Sxref , of the measurement
standard PSD setup can be used, i.e. a laser beam deflection involving both CLs can be written in the form
    
system. By providing a groove-like structure etched into the Vφ xφ + xref −1 1 xref −1
probe surface, the lateral movement of the groove-like structure Sxref = = = + ,
x Vφ Sx0 Vφ
causes spatial variations in the laser beam spot bounced off the
(5.4)
CL, thus resulting in variations in the lateral force signal.
In contrast to a scheme aiming at elimination of the where Sx0 is the lateral sensitivity in the case of a rigid vertical
mechanical crosstalk effect, for special measurement modi edge. Combining this with the equation representing the force

16
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 10. Schematic representation of the vertical beam method. (a) Sketch of the configuration of colloidal probe and vertical beam of
stiffness knref . (b) The related lateral signal–displacement curves both for the case of a beam and of a rigid structure. (c) Sketch of a similar
approach, applying the lateral force via the contact between the tip apex and a beam, i.e. operating in the sticking regime of a lateral contact.
If the reference beam is made of a piezoresistive material, its deformation is converted into a voltage.

equilibrium (equation (5.1)), the ratio kx /knref can be expressed 10.36 ± 0.27 µN V−1 [41]. The effect of torsional loading
in terms of the sensitivities: on the output voltage, Vp , was found to be negligible. Thus,
it was concluded that points on the side edges as well as on
kx Sx0 − Sxref
= . (5.5) the tip edge of the piezoresistive CL can be used as loading
knref Sxref locations [41].
For the calibration of AFM CLs, the piezoresistive force
By accounting for the shorter torsional arm length in the case
sensor was either mounted vertically or horizontally on the
of the calibration measurement, equation (5.3) leads to an
AFM stage along its longitudinal axis. Then the tip of the CL
expression for the lateral CL spring constant, kx :
to be calibrated was brought into contact with the top edge of
 2 the piezoresistive force sensor. The respective measurement
Sx0 D/2 + t/2
kx = knref . (5.6) configurations are referred to as top-loading (figure 10(c)) and
Sxref − Sx0 D + t/2
side-loading (not shown), respectively. In both configurations,
 the AFM tip was moved laterally after the loading location
The lateral
0  sensitivity, Sx , results from S x /S 0
x = S φ /H /
Sφ /H = H /H : had been determined by scanning in contact mode. Upon
lateral movement, the voltage outputs of both the piezoresistive
D/2 + t/2 sensor, Vp , and the lateral PSD signal, Vφ , were recorded.
Sx = Sx0 . (5.7) For the top-loading configuration, force equilibrium
D + t/2
between the lateral force exerted by the CL, Fx , as given by
Xie et al [41] employed a commercially available equation (D5) and the restoring force exerted by the reference
piezoresistive CL as a portable microforce calibration standard. beam, knref x ref , yields the following equation:
As a reference lever with an integrated piezoresistive
sensor, this calibration standard allows for a voltage output kx xφ = Fx = knref x ref = Cp Vp . (5.8)
proportional to its deflection.
The normal spring constant, knref , of the reference lever was As resulting from equations (4.7) and (4.8), the total lateral
calibrated using Cleveland’s mass loading method [89]. The stiffness, kx,tot , of the mechanical system consisting of CL and
resulting value of knref was given as 18.21 ± 0.47 N m−1 [41]. tip–sensor contact is given by
For the calibration of the piezoresistive force sensor sensitivity  −1
[41], the piezoresistive CL was mounted horizontally on a 1 1
kx,tot = + . (5.9)
movable platform. It was loaded normal to its longitudinal kx kxc
axis via a glass microsphere attached to a nanopositioning
stage. The voltage output, Vp , of the Wheatstone bridge Herewith, kx and kxc are the lateral stiffness of the AFM CL
connected to the piezoresistive CL was recorded as a function and the tip–substrate contact, respectively.
of the microsphere displacement. The resulting value of the Using equation (5.8), the lateral conversion factor, αx ,
piezoresistive force sensor sensitivity, Cp = dFz /dVp , was can be expressed in terms of the voltage output, Vp , of the

17
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

piezoresistive force sensor:


Fx Vp
αx = = Cp . (5.10)
Vφ Vφ
For the side-loading configuration, however, the loading
direction is perpendicular to the reference beam, at a certain
distance from its clamping end. In order to eliminate the effects
of finite lateral contact stiffness, Xie et al [41] applied the
lateral force via the tip shank rather than via the tip apex.
Furthermore, from the condition that the same moment and
voltage are generated for both configurations, a correction
formula for the conversion factor measured in the side-loading
configuration was deduced.
By considering a CL of rectangular shape and the related
beam mechanics, the lateral force can be calculated using Figure 11. Schematic representation of the fibre bending method.
equations (2.7), (D2), (2.10) and (2.18): (a) Top view of the configuration involving fibre and CL. The upper
end of the fibre is fixed to a substrate. The distance xhf denotes the
T kφ φ H 2 kx φ
Fx = = = lateral displacement of the fibre at the position of the tip–fibre
H H H contact. (b) Cross-section of the fibre and its contact with the tip
Gwt 3 1 Gwt 3 shank. The vector q denotes the torsional arm between the tip–fibre
=H φ = φ. (5.11) contact and the CL shear point (assumed to be situated on the CL
3LH 2 3 LH long axis).
Using this expression as well as equations (D6), (D4) and (2.4),
the lateral conversion factor, αx , can be expressed in terms
of the lateral PSD sensitivity, Sx , and the lateral CL spring with the fibre bending stiffness
constant, kx : 4
3π Ehf rhf
Fx Fx 1 Fx 1 Gwt 3 kx khf = . (5.14)
αx = = = = = . 4 L3hf
Vφ Sx xφ Sx H φ Sx 3LH 2 Sx
(5.12) Herewith, Ehf , rhf and Lhf are Young’s modulus, the radius
and the free length of the fibre, respectively [69]. The fibre
In a simplified approach, the PSD can be assumed to be
bending stiffness, khf , scales with 1/L3hf . When plotting the
rotationally symmetric, i.e. to be equally sensitive for normal
PSD output versus the restoring force, Fhf , the lateral force
and torsional deflections. Then, Sx can be replaced with Sz .
conversion factor results from the slope of the curve.
However, as mentioned in section 3, the PSD response is likely
In the experimental realization of Liu et al [69], a glass
to be asymmetric, e.g. due to asymmetries in the laser spot
fibre was used. The loading was achieved by scanning across
shape [41].
the fibre. The loading was assumed quasi-static and the lateral
Xie et al [41] gave an overall error of 12.4% in the lateral
force conversion factor derived from the balance of restoring
conversion factor, α. Notably, it was found to depend largely
forces.
on the uncertainty in Cp . For the case of a piezoresistive
The conversion factor value was found to increase slightly
force sensor calibrated using an absolute force standard, they
expect an error of <6%. The uncertainty of their method was with the free length of the glass fibre, Lhf . Notably, for smaller
compared with that of a beam mechanics method where the values of Lhf , the assumption of small deflection (xhf  Lhf )
thickness was determined via the CL fundamental resonance may not hold. Furthermore, for increasing deflection, xhf ,
frequency. With the uncertainties δLeff = 2%, δw = 2%, the contact point between tip and glass fibre is likely to move a
δh = 10%, δt = 10% and δSz = 10%, they arrived at an small distance along the fibre axis, thus resulting in an effective
overall error of 33% in α0 . increase in Lhf , i.e. in a slightly reduced spring constant. That
is, for a larger free length, Lhf , a more accurate lateral force
conversion factor should be obtained. With uncertainties of 3%
5.2. The case of a horizontal reference beam
in both rhf and Lhf , and of 5% in xhf , Liu et al [69] calculated
A horizontal fibre of known dimensions and modulus can be an overall uncertainty of the technique of 26%. However, the
utilized to apply a lateral force to the tip. If laid down on a effective free length of the fibre may differ from the value of
flat substrate and attached to it at one of its two ends, the free Lhf extracted from the images, if the glue used to attach the
end of the fibre bends laterally when it encounters the scanning fibre to the substrate is not rigid but accommodates a fraction
AFM tip (figure 11(a)). of the deformation. In these terms, effectively the value of Lhf
For small deflections of the horizontal fibre, xhf  is likely to be slightly larger, resulting in a lower value of the
Lhf , the restoring force exerted by the in-plane fibre, Fhf , is restoring force, Fhf .
proportional to the displacement at the position of the tip, xhf : Another point of concern is the offset between the tip apex
4
and the contact point with the fibre. Even if the lower side of
3π Ehf rhf the fibre is in genuine contact with the substrate, a fibre of
Fhf = xhf , (5.13)
4 L3hf micrometre-size radius touches the tip shank rather than the

18
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

very end of the tip (figure 11(b)). Hence, the torsional arm is given in figure 12(b). The suspension system is represented
length differs from the one acting in the case of an apex– by the reference spring of stiffness ks and the restoring force
surface contact. If the torsional arm length is shorter than generated by the suspension system is denoted by Fs .
the full tip height, H , then the same lateral force generates By integrating a number of such MEMS devices of various
a smaller torque, and the CL appears stiffer for torsional suspension stiffness values into a single chip, a wide range of
deformations. This effect can make a major contribution to CL stiffness values can be covered.
the total uncertainty budget, in addition to the uncertainties MEMS devices involving a suspended platform were also
considered by Liu et al [69]. Provided accurate knowledge of demonstrated for the calibration of normal forces in AFM [91].
the glass fibre diameter as well as the tip shape, however, it In this case, the platform is displaced in the normal direction.
should be possible to account for this effect. With the aim of providing traceability to SI units,
Furthermore, sliding motion of the fibre across the Cumpson et al [43] developed a procedure for the calibration
substrate surface may involve a frictional force contribution of the suspension system, involving electrical current
to the total lateral force acting on the tip. The assumed force measurement, laser Doppler vibrometry and white light
balance neglects such a frictional contribution. interferometry. Owing to such a set of complementary
Owing to its cylindrical geometry, a perfect fibre is measurements, this approach does not depend on knowledge
isotropic for all radial directions. Thus the same bending of the elastic properties or dimensions of the suspension arms.
stiffness applies to both lateral and normal loading conditions. Hence, the procedure is able to cope with deviations of the
Once the bending stiffness of the fibre is known, it can be fabricated structures from the layout, typically resulting from
utilized to calibrate the CL for normal as well as for lateral variations in the etching steps involved in the fabrication
loading. Such a measurement configuration can be achieved process of such MEMS structures. Typical etch effects are
with a fibre that is clamped at one end and that is not suspended edge biases and non-orthogonal sidewall angles, and they can
by a substrate, i.e. if the other end of the fibre is free to move result in significant changes in the mechanical properties, such
in all radial directions. By loading the free end of the fibre as the fundamental resonance frequency [62].
at various distances from the clamped end, Morel et al [90] Similarly to the lateral electrical nanobalance (LEN)
determined the CL stiffness and the PSD sensitivity. Using a developed by Cumpson et al [43], Ando et al [44] demonstrated
calibrated weight, the fibre bending stiffness was measured by a calibration device based on MEMS technology, dubbed
loading the fibre at various distances from the clamped end. microlateral force sensor (MLFS). In contrast to the LEN
device, the MLFS employs a feedback circuit to keep the
5.3. The case of a mechanically suspended platform lateral position of the platform constant. Upon application
of a voltage to the comb drive, the suspension system deforms
Provided a calibrated reference spring, the lateral CL spring elastically and the platform moves in the x-direction. The
constant can be calculated in a straightforward manner. Force electrodes of a detector attached to the platform are used to
equilibrium between the spring representing the lateral CL measure a tunnelling current. Depending on the magnitude of
spring constant, kx , and the reference spring of stiffness, ks , the tunnelling current, the voltage applied to the comb drive is
yields the following equation: adjusted to keep the tunnelling current constant.
xs When the AFM tip is in mechanical contact with the
kx = ks . (5.15) suspended platform, the latter is laterally displaced by a
xφ
distance xs , as a function of the lateral force exerted by the
Herewith, xs is the lateral displacement of a platform CL. The position of the platform is restored by means of a
suspended by the reference spring. The restoring force exerted feedback circuit. It applies a voltage to the comb drive that
by this spring is denoted as Fs . A total displacement x scales with the tunnelling current measured.
generated by an actuator causes a lateral displacement xs The voltage, Vd , applied to the comb drive scales with
of the platform, depending on the stiffness of its suspension the displacement xs :
arms, ks .
With the aim of providing a user-friendly calibration Vd = gcd xs . (5.16)
standard, purpose-made micro-electromechanical systems
(MEMS) were prototyped and demonstrated [43, 44]. The Herewith, the gain factor of the feedback connected with the
core element of the demonstrated MEMS devices is a platform comb drive is denoted with gcd .
suspended by four arms (figure 12(a)). Similar to an AFM The lateral force applied by the tip, Fx , can be measured
CL, the stiffness of the suspension system can be adjusted by recording the changes in the driving voltage, Vd . The
via the dimensions of the arms. The suspension arms of the driving force, Fd , generated when applying the voltage Vd to
platform should have a large stiffness for deformation in the the comb drive is given by the dimensions of the interdigitating
normal direction but a much lower stiffness for deformation in fingers and the permittivity of the medium between the comb
the direction perpendicular to their long axis. When scanning electrodes:
Ncd Tcd 2
across the platform surface, the AFM tip applies a lateral Fd = 2ε0 V , (5.17)
Dcd d
force and displaces the platform laterally. The demonstrated
suspension systems [43, 44] consist of two pairs of parallel where ε0 is the permittivity of air (8.85 × 10−12 F m−1 ), Dcd
arms. A schematic representation of such a calibration device is the width of the gap between the comb fingers, Tcd is the

19
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 12. (a) Top view of a MEMS-based device, consisting of a suspended platform and a comb drive. A driving voltage, Vd , can be
applied to the comb drive to apply a lateral force. (b) Schematic with springs representing the stiffness of the suspension system. The
restoring force exerted by the platform suspension arms is denoted as Fs . The force FT is the traction generated by the scanner while moving
the device from right to left.

thickness of the comb fingers and Ncd is the number of fingers In their design of the LEN device, Cumpson et al [43]
constituting the comb drive. included a slit-shaped groove, situated in the centre of the
By considering the dimensions of the suspension system, platform. When scanning across such a groove, the lateral
Ando et al [44] employed finite element analysis (FEA) to force acting between tip and platform increases, resulting both
calculate its lateral stiffness, ks . The resulting value of ks in a significant change in the CL torsional deformation and in
was 26 ± 3 N m−1 , with the main error source given by the a lateral displacement of the platform.
error in the thickness of the leaf springs. The relationship In the case of such slit-shaped grooves the global tip shape
between the driving voltage, Vd , and the displacement of the comes into play when the tip touches the groove edges with its
sensing table, xs , was measured using a confocal laser- shank rather than with its apex, thus leading to an ill-defined
scanning microscope. The displacement xs was found to tip–sample contact and large forces acting on the groove edges.
increase linearly with Vd1.8 and a sensitivity of 1.1 nN mV−1 To ensure stable and continuous contact between tip apex and
was calculated for a bias voltage of Vd = 90 V. platform, more shallow grooves seem to be advisable. For
When scanning on the platform surface a lateral force acts, instance, a trench-shaped groove could be employed, similar
depending on the surface properties of both tip and platform.
to the elementary unit of the grating structure (figure 15(a))
In the case of a low friction coefficient, µtp , of the tip–
used in Varenberg’s version of the wedge method [46, 47].
platform contact, the signals resulting from the torsional CL
deformation and the platform displacement can be as small as
the noise level. Hence the calibration measurement will suffer 5.4. The case of a magnetically suspended platform
from a large uncertainty. To ensure a sufficient SNR, it would
By employing a compliant load cell, Li et al [42] demonstrated
be necessary to select a useful LEN device depending on the
a direct method for the calibration of the lateral conversion
spring constant of the CL as well as on the largely unknown
factor of an AFM CL-tip system. In a similar manner to the
friction coefficient µtp . Hence to make the selection solely
dependent on the CL spring constant, it is helpful to provide a MEMS devices it relies on a suspended platform; however,
suitable topographical structure leading to significant changes the suspension is achieved via diamagnetic forces rather than
in the measured signals. via suspension arms. When providing an inhomogeneous
Using focused ion beam (FIB) etching, Ando et al [44] magnetic field exhibiting a minimum, a diamagnetic object
created an array of asperities on the platform surface. When is levitated, forced to the point of lowest magnetic field, and
scanning across such asperities, variations in the lateral forces gets trapped at the position of its minimum. The gradient
occur. The array pitch size was 250 nm. The normal spring of the diamagnetic levitation force represents a spring that
constant, kz , of the CL used for the demonstration of the can be used as the reference spring, ks , of the calibration
technique was ∼2 N m−1 . It should be noted that the AFM measurement. Accordingly, the suspension system was
measurement was performed under vacuum conditions (3 × dubbed diamagnetic lateral force calibrator (D-LFC) [42].
10−5 Pa) to ensure a stable tunnelling current. Li et al [42] relied on a static friction measurement, i.e.
Rather than providing an array of topographic structures, the lateral force is kept below the force required for the stick–
also a single but more extended structure can be utilized. slip transition (section 4.3). The total lateral displacement,

20
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 13. Schematic representations of (a) the array of four


permanent magnets used to create a magnetic spring, and (b) a PG
sheet levitated and trapped in the region of high magnetic field
gradients. When operating in the sticking regime of the
tip–substrate contact, a small lateral displacement, x, of the CL Figure 14. Schematic representation of a friction loop and the
causes a lateral deformation, xt , of the CL, the tip and the contact related parameters. The loop can be described in terms of its
as well as a lateral displacement of the levitated PG sheet, xs . If half-width, W , and the value of its centre line, .
the lateral stiffness of the levitation system is very small, the
component xt is negligible. (figure 13(a)). They utilized the highly anisotropic magnetic
behaviour of pyrolytic graphite (PG). PG has a layered
structure and is highly diamagnetic. Its magnetic susceptibility
x, generated by the scanner results in lateral displacements in the direction perpendicular to the basal plane is several times
of the suspended platform, xs , and of the tip apex, xt , higher than that in the direction parallel to the plane [42, 92].
respectively: Provided a vertical magnetic field, a PG sheet can be suspended
x = xs + xt . (5.18) with its basal plane aligned horizontally. Lateral stability can
In turn, the tip displacement, xt , can be split into a fraction be achieved by providing a lateral modulation of the magnetic
due to the compliance of the tip–sample contact, xc , and field, e.g. by arranging a number of permanent magnets
a fraction due to the torsional deformation of the CL, xφ with alternating magnetization. A system of four permanent
(section 4.3). Owing to the low lateral stiffness, ks , of the magnets allows for stability along both lateral directions, with
magnetic suspension system, however, the component xt is the minimum of the magnetic field at its centre (figure 13(b)).
negligible, xt  xs , and the lateral force acting on the tip For the lateral spring constant of their diamagnetic levitation
can be written as system, Li et al [42] gave a value of 10 pN nm−1 .
Key parameters for the diamagnetic levitation effect are
f  = ks xs ≈ ks x. (5.19) the magnetic flux density, B, and its gradient, ∇B, the
susceptibility of the diamagnetic material, χ , as well as its
When writing equation (4.17) in a difference form, f = mass density, ρ. The magnetic force, Fm , acting on a particle
αxx Vφ + αxz Vz , the inverse slope, ∂f/∂Vx , of the resulting of volume, V , and susceptibility, χ , is given by
plot of Vφ versus f ≈ ks xs represents the lateral conversion
factor, αxx . The measurement can be made for various values 1
of Vz by changing the setpoint value of the force feedback. Fm = χ V ∇B 2 . (5.20)
2µ0
Similarly to the determination of α xx , the crosstalk lateral
conversion factor, αxz , is given by the inverse slope, αxz = For diamagnetic materials, the susceptibility is negative, i.e.
∂f/∂Vz . Alternatively to a two-step procedure, the lateral χ < 0. Balance between the diamagnetic and the gravitational
conversion factors can be determined in a single step by a 3D force yields the following condition [93]:
fit procedure for the full dataset f (Vx , Vz ). ρ
It is worth underscoring that apart from the normal B∇z B = µ0 g , (5.21)
χ
conversion factor, αz , and the lateral conversion factor, α xx , the
method provides a conversion factor, αxz , related to crosstalk. where g denotes the gravitational acceleration. That is,
As elucidated in section 3.2 and in section 4.5, respectively, materials with a large ρ/χ -ratio require a large value of the
crosstalk between normal and lateral forces can be induced product B∇z B.
optically or mechanically.
Interestingly, the availability of both lateral conversion
6. Methods relying on a scanning motion
factors allows for the measurement of the absolute value of the
friction force. Thus friction force measurements do not need 6.1. The wedge method
to rely on the friction loop width, W . Although this approach
is well-established, it is not suitable for studies of directional When plotting the lateral force versus the lateral position,
effects, i.e. in the case that the absolute value of the friction x, for both the forward (trace) and the backward (retrace)
force measured in the forward direction differs from the one scan directions, a loop results (figure 14). It is generally
measured in the backward direction [42]. referred to as a friction loop. Essentially, it is defined by
The levitation system of Li et al [42] consists of a set two parameters, namely its width, W , and its centre line,
of permanent magnets and the levitated (suspended) platform . Both these parameters are related to the torsional moment

21
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

acting on the tip. In the case of a positive slope, the forward of sliding friction is not necessarily identical with the pull-off
and the backward scans are related to uphill and downhill force extracted from a normal FDC. This issue was pointed out
motions, respectively. The uphill motion causes an additional by Butt et al [95]. They identified the approach of determining
contribution to the torsional moment acting on the CL. This FA via FDCs [46] as a major source of uncertainty.
effect has been referred to as a ratchet mechanism [94]. With Indeed, Varenberg et al [46] reported a load dependence of
the local slope angle θ, the related contribution scales with both the calibration coefficient, αφ , and the friction coefficient,
tan θ. µ. In contrast to the approach taken by Varenberg et al [46],
By denoting the related torsional moments as T u and T d , the original wedge method of Ogletree et al [45] relied on
respectively, the friction loop parameters W and  can be measuring friction loops for a range of loads and evaluating
written as follows: their load dependence.
Wang and Zhao [96] attributed the load dependence to the
2W = T u − T d , (6.1a)
finite contact stiffness. They expressed the overall conversion
2 = T u + T d . (6.1b) factor as the product of a factor given by the CL properties and a
contact factor. The latter is given by the ratio of the CL stiffness
For a perfectly symmetric CL with a tip exactly positioned on
to the stiffness of the tip–substrate contact. Similarly to Wang
its long axis, the lateral forces acting during the uphill and the
and Zhao, Butt et al [95] employed a mathematical separation
downhill motions differ in direction only but are identical in
to address the load dependence of the wedge method. With the
amount, and the difference in related torques is constant. It
aim to separate the influence of FA on the lateral deflection,
should be noted that potentially the frictional forces depend
they developed an ansatz that is linear in FL and folds the FA -
on the direction of scanning even in the case of a perfectly
dependence into the offset. The offset constants are denoted
flat surface; however, this case seems to be rare and is widely
as IW and I . Hence the key equations given by Varenberg
neglected.
Thus, when evaluating lateral force data, the friction et al [46] can be written in the following form:
loop width, W , is typically considered rather than the lateral
W00 (M, θ ) = SW FL + IW (6.6a)
force signal, Vφ , since the lateral force contribution related to
topography is eliminated as well as any offset in Vφ .
and
In terms of the measured PSD signal, the moments acting
on the tip can be written as follows: 00 (M, θ ) = S FL + I , (6.6b)

T (κ̃) = αφ Vφ(κ̃) (6.2) where


R (1 + cos θ ) + t/2 µ
with κ̃ = u and κ̃ = d for an uphill and a downhill scan, SW = · , (6.6c)
β cos θ − µ2 sin2 θ
2
respectively.
R (1 + cos θ ) + t/2 µ cos θ
By expressing the moments in terms of the applied load, IW = · FA , (6.6d)
FL , and the adhesion force, FA , the following equation for the β cos θ − µ2 sin2 θ
2

friction loop width, W , can be derived [46]: R (1 + cos θ) + t/2 µ2 sin θ cos θ + sin θ cos θ
S = ·
β cos2 θ − µ2 sin2 θ
W FL + FA cos θ
=µ . (6.3) R sin θ
H cos2 θ − µ2 sin2 θ − , (6.6e)
β
Similarly, for the centre line value, , of the friction loop the R (1 + cos θ) + t/2 µ2 sin θ
following equation holds: I = · FA . (6.6f)
β cos2 θ − µ2 sin2 θ
 
 sin θ FL cos θ + µ2 (FA + FL cos θ ) The values of the slope SW result from linear plots of W00 values
= . (6.4)
H cos2 θ − µ2 sin2 θ measured for various loads, FL . Then the friction coefficient,
µ, can be calculated using equation (6.6c), and the torsional
From the ratio of equation (6.3) to equation (6.4), an equation calibration factor, αφ , can be calculated using equation (6.6e).
can be derived that is quadratic in µ: Notably, both these equations do not depend on FA .
 In addition to the effective adhesion force, FA , Butt et al
sin θ (FA + FL cos θ ) µ2 − (FL + FA cos θ ) µ
W [95] identified a group of error sources related to the CL
+ FL sin θ cos θ = 0. (6.5) deflection and its detection. The effects to be considered are:
The solutions of equation (6.5) depend on the quantities  and
W , resulting from the lateral force measurement, as well as on (i) a non-zero lateral CL deflection even if no torque is
the forces FL and FA . In particular, this means that calculation applied,
of µ requires calibration of the normal CL spring constant, (ii) a non-constant vertical CL deflection due to poor feedback
kz . Thus, strictly speaking the wedge method is not truly a performance or due to pronounced topography features,
single-step method, although it provides an overall torsional and
conversion factor. Furthermore, it should be mentioned that (iii) changes in the background deflection signal with CL
the effective adhesion force, FA , acting under the conditions height due to optical interference effects.

22
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Figure 15. (a) Sketch of the calibration grating TGF11 and (b) depiction of its application to a colloidal probe of radius R. Owing to its
comparatively extended slopes, the TGF11 calibration grid is suitable for calibrating colloidal probes with a tip diameter up to
2R ≈ 5.6 µm. However, for particle diameters larger than 2R ≈ 3.9 µm the colloidal probe cannot reach the planar bottom of the grating.

The various contributions to the torsional deflection signal can Although the method is suitable for colloidal probes, for
be summarized in the following equation: particle diameters larger than 2R ≈ 3.9 µm they cannot reach
  the planar bottom of the grating [95].
Vφ = Vφ0 (T ) + VφI VzSP + VφII Vzerror + VφIII (noise) , (6.7)
6.2. Methods involving the normal spring constant
where Vφ0 denotes the contribution by the torsional moment,

T , and where VφI VzSP denotes the coupling to the static When scanning parallel to the CL long axis rather than

component of Vz . The term VφII Vzerror accounts for the perpendicular to it, the frictional forces acting between tip and
coupling to the dynamic component of Vz , including any sample result in flexural CL deformations. In the ideal case of
change in lateral deflection associated with a limited response a perfectly flat and smooth surface, the flexural deformation
of the topography feedback. This term is likely to scale with can be exclusively attributed to the frictional forces. If the
the scan speed. Finally, the term VφIII (noise) represents the normal spring constant, kz , and the related PSD sensitivity, Sz ,
contribution due to the noise floor associated with optical are known, the frictional force can be quantified. Hence, by
interference effects. first measuring the frictional forces in a parallel scan and then
In a similar manner, the various contributions to the recording the lateral force signal in the usual perpendicular
friction loop offset, 0 , and to the friction loop half-width, W0 , scan, the lateral force measurement can be calibrated [48, 76].
can be summed up. Only the terms 00 and W00 of order zero This approach relies on the assumption that the frictional forces
are connected to the torsion moment, T . Since the first and the measured in parallel scan mode are equal to the ones measured
third terms are independent of the scan direction, they do not in perpendicular scan mode. Furthermore, it should be noted
contribute to W0 but to 0 . However, their contribution to 0 that such modes with the frictional forces acting along the CL
can be evaluated by scanning on a flat segment of the calibration long axis were also referred to as axial [97] or longitudinal [50].
grid and subtracting the related offset from the offset measured In the general case of a real surface, the flexural CL
on a sloped facet. deformation includes a contribution from surface height
Another point of concern is the effect of optical or variations, and this contribution needs to be separated from the
mechanical crosstalk on the wedge method. The resulting frictional one. The separation can be achieved by accounting
values of the calibration factors may vary significantly with for the fact that the sign of the frictional forces changes
the degree of crosstalk, and parameters such as the friction with the scan direction, whereas the height variations due
coefficient could show a non-linear impact on the signals. to the surface topography are independent of it. With the
Reportedly, the friction loops measured on two flat surfaces topography feedback being engaged, the changes in total
of different height can show a mutual offset, reflecting a flexural deformation with scan direction will translate into a
coupling between the topography and the lateral force signal difference in the height profiles measured in forward and in
[80]. At present, however, there are no dedicated experimental backward scan direction. That is, when scanning parallel to the
techniques available for the quantification of crosstalk effects. CL long axis, the height profiles of the forward and backward
A valuable approach to such quantification could be a rigorous scans will differ due to the frictional forces. As shown by Ruan
modelling of the LFM detection system. Michal et al [82] and Bhushan [48], the frictional force can be calculated from
suggested parametric studies based on a system model. the height difference between the two scans, the normal CL
A depiction of the calibration grating used by Varenberg spring constant, the CL length and the tip height.
et al [46] is shown in figure 15(a). The grooves have a Another assumption is that the friction force is identical
trench-like structure. The slope angle of ∼54.7◦ is given by for the forward and the backward scans [48, 76]. However,
the Si crystal lattice. When scanning across such grooves, the normal load is slightly different for the two scans, since the
good feedback performance is required to ensure a reasonably normal deflection signal, Vz , is kept constant by the topography
constant load. Furthermore, in the case of colloidal probes, feedback loop but affected by both normal and lateral forces.
the limited width of the grooves can be an issue (figure 15(b)). This issue was addressed by Wang et al [49]. Starting from

23
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

an analysis of the forces acting parallel and perpendicular to constant. However, it should be borne in mind that large
the CL, they developed a modified set of equations. For large uncertainties in kz will propagate into the uncertainty of kx
values of the ratio of tip height and CL length, h/L, the values (section 2.2).
of the friction coefficient and of the overall conversion factor Provided validity of the macroscopic friction law, a
were found to differ significantly from the ones calculated by linear relationship between frictional and normal forces can
the method of Ruan and Bhushan [48, 76]. be expressed in terms of the friction coefficient, µ. Its
In a more recent publication, Karhu et al [50] experimental determination involves variation of the normal
demonstrated another modification of the parallel scanning force and measurement of the lateral force signal. For instance,
method. By considering particular friction laws, they in the approach demonstrated by Bilas et al [63], normal force
quantitatively extracted friction data from topography images variation along a scan line was generated by exploiting the bow
measured in longitudinal scan mode. Also, they provided a set of a tube scanner. In their scanning-by-probe (SBP) system, the
of equations for the case of colloidal probes. laser beam was not moved along with the CL. In combination
Similarly, the contribution of frictional forces to the with the action of the topography feedback, the corresponding
normal CL deflection observed over the measurement of motion of the laser spot resulted in normal force variations.
normal force–displacement curves can be evaluated. This Finally, it is worth mentioning that a scanning motion can
approach was demonstrated by Stiernstedt et al [97]. Due also be employed to load a compliant structure. This is the
to the non-zero tilt angle between CL and the horizontal, upon case for the methods of Liu et al [69], Cumpson et al [43] and
contact with the tip a vertical displacement induces a sliding Ando et al [44]. Such methods are discussed in section 5.
motion between tip and surface. The resulting friction force
in the longitudinal direction causes an additional contribution
to the normal deflection of the CL. Since the direction of the
7. Methods relying on torsional resonances of the CL
friction force changes upon reversal of the vertical motion, i.e.
In addition to the common measurement of lateral forces in
from approach to retraction, the friction force component can
LFM, i.e. the lateral forces induced by the scanning motion
be extracted from the difference in slopes of the two constant
of the tip across the sample surface, more recently developed
compliance branches of the force–distance curve [97]. Such a
imaging techniques involve the torsional resonance of the CL.
contribution related to friction is hardly detectable for standard
In the torsional resonance mode a special configuration of
AFM tips; however, it is more pronounced for larger probes,
piezoelectric actuators is used to drive the CL into torsional
such as colloidal probes with a radius of ∼10 µm or more [97].
resonance [71, 98]. The related amplitude and phase signals
In addition to avoiding calibrations of the torsional spring
are monitored and their changes upon tip–sample interaction
constant and of the torsional PSD sensitivity, the approach
were shown to enable distinct materials contrasts [98].
delivers friction force data as a continuous function of the
applied load. In contrast, in calibration methods based on In another mode relying on the CL torsional resonance,
the measurement of friction loops (section 6.1) the load is a hammerhead-shaped CL is operated in intermittent contact
a parameter, i.e. a separate measurement for each value of mode [87]. The tip of the asymmetric CL is placed off
the load is required. On the other hand, the method does the CL long axis by a distance larger than the CL half-
not allow for (not even partly) quantification of typical LFM width. As a result of this asymmetry, a torsional moment
measurements, i.e. of the separate signal, Vφ , related to is generated (section 4.5), and vibrations in the fundamental
torsional CL deformations that occur upon scanning in the torsional resonance can be excited. While the amplitude
lateral direction. Due to this limitation, here it is not listed as of the CL flexural deformation is used for operation of the
a lateral force calibration method. topography feedback, the amplitude and phase signals related
It should be noted that these methods assume that to the torsional resonance provide additional information on
the hysteresis in the normal force signal can be fully the sample properties, such as the elastic modulus [87, 99].
attributed to the frictional contribution. However, piezoelectric Due to the development of such techniques relying on the
scanners typically show non-zero hysteresis due to non- CL torsional resonance, it seems to be of increasing relevance.
linear behaviour. Although most AFM microscopes feature However, it also can be exploited for the calibration of the CL
a built-in linearization, it is not always fully efficient. Since torsional spring constant. In many aspects, there is a direct
the parameters characterizing the non-linearity are likely analogy with the use of the fundamental flexural resonance
to change over time, in general numerical linearization for the calibration of normal forces. Commonly used methods
schemes (software linearization) are less effective than for the calibration of the normal CL spring constant are the
active ones that involve a real-time measurement of the added mass method of Cleveland et al [89] and the unloaded
actual displacements generated by the piezoelectric scanner resonance technique of Sader et al [100]. The former relies
(hardware linearization). on the change in resonance frequency with known masses
Another route for lateral force calibration is provided by attached to the free end of the CL. The latter relies on the
equation (2.20). In addition to Young’s modulus, E, the shear resonance frequency and quality factor of the fundamental
modulus, G, the effective CL length, L, and the tip height, H , flexural resonance as well as on the plan view dimensions of
its employment requires knowledge of the normal CL spring the CL. In a similar manner, both these methods can be applied
constant, kz . to determine the torsional CL spring constant [51].
Hence, in an indirect way, this approach can take benefit of The fundamental torsional resonance frequency, ωφ ,
the methods available for the calibration of the normal spring depends on the torsional spring constant, kφ , and on the mass

24
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

moment of inertia, J : constant. It involves the imaginary part of the hydrodynamic


function for torsional vibrations, it (ωt ) [51]:

ωφ2 = . (7.1)
J kφ = 0.1592 · ρw 4 L0 Qt ωt2 it (ωt ) , (7.6)
In addition to the mass moment of inertia, Ji , inherent to the where ωt and Qt are the radial frequency and quality factor of
CL, the moment J also comprises the contribution, Ja , due to the fundamental torsional resonance, respectively, as measured
a mass attached to the CL: with the CL immersed in a fluid of density ρ. The
hydrodynamic function  t (ω) has been derived for the case
J = Ji + Ja . (7.2) of a rectangular CL. It depends only on the Reynolds number
Re = ρw 2 ω/ (4η), where η is the fluid viscosity. It is
For a sphere-shaped mass, Ms , attached to the free end of the
independent of the CL thickness and density [51]. Tabulated
CL, the total added mass moment of inertia, Ja , is given by [51]
values of the function it (ωt ) can be found in [102].
Good agreement between experimental and theoretical
Ja = 75 Ms rs2 . (7.3)
data has been achieved for the fundamental oscillation modes;
Herewith, Ms and rs denote the mass and the radius of the however, significant deviations have been reported for higher
attached sphere. Rearrangement of equation (7.3) shows that harmonic modes [103]. When operating in viscous liquids, the
a plot of Ja versus ωφ−2 results in a linear curve, the slope motion of the liquid can strongly affect both the CL frequencies
of which is given by the torsional spring constant, kφ . That and the damping of the higher harmonic modes. Under such
is, the calibration procedure requires a series of torsional conditions, the motion of the liquid around the oscillating CL
resonance frequency measurements, using different spherical is of a three-dimensional nature and calculations based on
masses attached to the free end of the CL. the three-dimensional Navier–Stokes equation were found to
Notably, the mass should be placed at the free end of the deliver more appropriate results [103].
CL and aligned with the CL long axis. Furthermore, for measurements under ambient conditions
Off-end loading can be accounted for by consideration the effects of temperature or relative humidity on the
of equations (2.10) and (2.18). That is, the torsional spring hydrodynamic function, it (ωt ), need to be known.
constant for the case of loading at a distance L = L0 − L Considering that the quality factor of the CL resonance shows
some dependence on the system temperature or on the humidity
from the CL free end is given by
level, it can be expected that the value of the hydrodynamic
kφ L0 function changes slightly with these parameters [17].
= , (7.4) Another source of uncertainty can be differences between
kφ0 L
assumed and actual CL shape. The model assumes a perfectly
where L0 is the full CL length and kφ0 the related torsional rectangular CL. Most types of commercially available CLs,
spring constant. however, exhibit some variations along their long axis, such as
Off-axis loading can be accounted for via the mass a protruding edge near the free end of the CL. It is conceivable
moment of inertia, Ja , due to the added mass. With the that the hydrodynamic behaviour of the CL is affected by such
distance, e, between the sphere centre of mass and the CL deviations from the perfect CL shape [17].
long axis (eccentricity), Ja can be expressed as follows [51]: Various techniques for the deliberate excitation of
  torsional resonances can be employed. In the torsional
7 e2 resonance mode a pair of piezoelectric actuators is used
Ja = + Ms rs2 . (7.5)
5 rs2 [71, 98]. They are integrated into the CL holder in a way that
the left- and the right-hand sides of the CL can be displaced
Hence, the mass moment of inertia of a mass applied off-axis separately, in the direction normal to the CL plane. When they
is larger than that of an identical mass placed on the long axis are operated 180◦ out of phase, a torsional motion is generated.
of the CL. Consideration of equations (7.1) and (7.2) shows Since the usage of piezoelectric actuators involves a
that the increase in Ja will cause ωφ to decrease. potential hazard of electrical shorting when operating in
Moreover, off-axis positioning of the mass can also result liquids, Mullin and Hobbs [104] devised a different setup.
in a coupling of the flexural and torsional modes of vibration. They replaced the piezoelectric actuators with slabs of a
A similar asymmetry in mass distribution occurs for the case of magnetostrictive material. In combination with a modulated
a positional offset of the tip from the CL long axis (section 4.5). magnetic field, one slab expands whilst the other one
By imaging CLs upon vibration in their resonance modes, it contracts. The modulated magnetic field can be generated by
has been shown that asymmetries in the CL geometry lead to superimposing the static magnetic field of a permanent magnet
significant mode coupling [101]. with the alternating magnetic field of a solenoid connected to
The underlying theory of the Sader method considers the an ac voltage source [104].
response of a CL surrounded by a fluid of a certain density Another way of exciting torsional resonances has been
and viscosity. Since air can be considered a fluid, the method demonstrated by Jeon et al [52]. Using a triangular-shaped
can be used to describe experiments performed under ambient CL, they induced a torque by exploiting the Lorentz force. Via
conditions. Calculation of the effect of fluid dissipation on the an electrically conductive coating on the backside of the CL, an
CL response yields an expression for the torsional CL spring electric current was conducted across the two legs of the CL.

25
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

In combination with an in-plane static magnetic field aligned employ a lever arm different from the AFM tip (as is the case for
perpendicular to the CL long axis, a Lorentz force is induced the methods (4)–(9) of table 4). Then, the moment arm length
that acts upwards on one side of the CL and downwards on consisting of the AFM tip needs to be evaluated separately. As
the other side of the CL. The magnitude of the resulting torque can be seen from equation (2.10), the lateral spring constant,
can be calculated from the magnetic field strength, the electric kx , is equal to the torsional spring constant, kφ , divided by
current and the area enclosed by the two legs of the CL [52]. the square of the moment arm length, i.e. by the square of the
distance between the contact point and the CL neutral axis.
8. Overview and discussion A determination of the moment arm length related to the
tip can be achieved by measuring the tip height using an SEM
Considering the large number of methods related to force and by assuming that the CL shear centre is identical with
calibration in LFM, it is advisable to organize the comparison its long axis. Provided a uniform and perfectly symmetric
of the methods in the form of a table. As can be seen from CL, such an approach can be sufficiently accurate. However,
table 4, the applied criteria are: in practice it is likely to cause additional uncertainties since
heterogeneities and asymmetries of the CL beam can result
– the resulting calibration constant, i.e. if the method yields in an offset of the shear centre from the symmetry centre
separate values of the PSD sensitivity and of the torsional (section 4.5). Similarly, additional uncertainties can arise from
spring constant or if it yields the overall conversion factor. a positional offset of the tip from the CL long axis. Then,
Typically, measurement of an overall (combined) torsional accurate calculation of the torsional CL deformation is difficult
calibration factor means that neither kφ nor the sensitivity to achieve since it requires intimate knowledge of the CL shape
of the optical detection system was calibrated separately. and materials properties.
The related pattern of indicators is N N Y (No–No–Yes).
Calibration methods that apply a lateral force via the tip
In contrast, a pattern Y Y N (Yes–Yes–No) means that
shank (methods (15), (16) of table 4) are somewhat in between
separate calibrations are achieved; then the overall factor
the methods that use the tip apex and the lever methods.
could be calculated but is not primarily measured
Considering that in such a configuration the moment arm length
– the involvement of a model such as the friction law, i.e. a
is shorter than in the case of an LFM measurement, a correction
method is classified as direct if the applied lateral force is
for the difference in moment arm length needs to be included.
known from basic principles or direct measurements
Again, this requires detailed knowledge of the tip shape and
– the suitability of the method for colloidal probes, i.e. if
in these terms, methods based on the tip shank are similar to
it is applicable to probes incorporating micrometre-size
lever methods, even though they rely on the integrated tip.
beads rather than a standard AFM tip
Being independent of the tip–substrate interaction, lever
– the risk of tip damage, i.e. if the method requires
methods can be used as a reference to cross-check results
mechanical contact with the tip and if this contact is likely
from other methods and to estimate the uncertainties resulting
to cause some damage
from the tip–substrate interaction or from mechanical crosstalk
– the degree of user-friendliness, i.e. the level of skills
effects. Furthermore, the risk of damage to the tip occurring
required
in the course of the calibration procedure is reduced. In this
– the methodological requirements or complications
sense, it is similar to the methods that do not involve the tip but
inherent to the method
benefit from force interactions with the surrounding medium
– the uncertainties as reported by the authors and the
or the CL holder, such as the methods that analyse the torsional
identified main source of uncertainty.
CL resonances (methods (28)–(30) of table 4).
For the evaluation of the risk of tip damage as well as For the calibration methods that involve a sliding motion,
of the degree of user friendliness, a scoring range 0–2 was the friction force is usually calculated on the basis of
applied. For instance, a score of 0 indicates a low risk of tip Amontons’ law (methods (20)–(26) of table 4). As a linear
damage, and it was attributed to the methods where the tip is law it allows for comparatively simple equations. Although
not involved in the calibration procedure. A scoring value of many experimental studies show that it applies to a wide range
1 was applied if a static contact with the tip is involved and a of tip–substrate systems, in the case of truly microscopic tips,
scoring value of 2 if a scanning motion is involved. typically a friction law based on contact mechanics applies,
For the purpose of comparison, also two theoretical such as the JKR or the DMT model (see [56] and the references
methods are listed, namely a method based on analytical given therein). Such models show a non-linear dependence
continuum mechanics [61] and a method involving FEA [64]. on the load applied and result in analytic expressions that
With an overview of the individual characteristics of the are more difficult to deal with. Nevertheless, before applying
various methods given in table 4, the following discussion Amontons’ law for the evaluation of lateral forces, one should
looks at some overarching characteristics. verify its validity for the particular tip–substrate system. In
As already indicated in table 1, many calibration methods the case that a contact model applies, the equations of the
apply the force immediately via the tip apex. The key calibration method relying on a friction law need to be derived
advantage of this approach is that it gets as close as possible by including the equations of that contact model. To assess
to a typical LFM measurement. Otherwise, additional sources if a particular tip is likely to show macroscopic friction
of uncertainty may result from differences in the mechanical behaviour (i.e. if it obeys Amontons’ law) or microscopic
configuration. This is the case for the lever methods that friction behaviour (i.e. if it follows a contact mechanics law

26
J. Phys. D: Appl. Phys. 43 (2010) 063001
Table 4. Overview of the available methods for the calibration of lateral forces in LFM.
PSD Over-all Suitable for Risk of tip User Methodological Reported uncertainties;
Name of method, sens. sens. colloidal damage friendliness requirements or main source of Compared with
reference Approach S φ kφ α φ Direct probes 0 (lo)–2 (hi) 0 (lo)–2 (hi) complications uncertainty method
(1) Optical geometry, Beam elasticity Y Y N N Y 0 1 Requires knowledge of the δkφ = 10%; None;
Liu et al [73], theory + geometrical analysis optical beam path, non-linearities in the [28]: comparison with
section 3.1 of the optical beam path, including the PSD dimensions, detector system and its the methods of [39]
measuring signal changes distance from CL to PSD, and associated mechanical and [45]
upon displacement of the PSD probe moment arm positioning
(2) Optical geometry Geometrical analysis of the Y N N N Y 0 1 Requires knowledge of the Not available; Optical geometry
method for liquid cells, optical beam path, accounting optical beam path and of the non-linearities in the method of [73]
Tocha et al [74], for differences in refractive refractive index of the liquid detector system and its
section 3.1 indices between air and liquid associated mechanical
positioning
(3) CL rotation using a Rotation of a test CL about its Y N N Y Y 0 1 CL holder with integrated δSφ = 5.6%; None
CL holder with long axis via the vertical lever and hinge that converts a alignment of the CL with
integrated hinge, displacement of a special CL vertical displacement into a the hinge, distance between
Xie et al [72], holder with integrated lever rotation the barrier and the hinge
section 3.1 and hinge
(4) Pivot and mirror Pushing against a protuberance Y Y N Y N 2 1 Demonstrated for a tipless δkφ = 15%; Dimensional
tilting, offset from the CL long axis, target CL only, requires off-axis distance a,
Bogdanovic et al [35], PSD sens. by changing the well-defined tilting of the different conditions of the
27

section 3.1 angle between AFM-head and AFM head mirrored substrate
a mirrored substrate
(5) Pivot and torsional Torsional Sader + pivot as Y Y N Y Y 2 1 Requires measurement of the δkφ = 16.8%; Torsional
Sader, of [35], reverse calculation torsional resonance frequency off-axis distance a Sader + lever method,
Pettersson et al [17], yields the PSD sens. also torsional added
section 3.1 mass method [51],
mirror tilting
(6) Horizontal lever, Generating a torque via a lever Y N Y Y Y 0 0 Requires attachment of a stiff δαφ = 10%; Dimensional via
Feiler et al [29], attached to the target CL lever to the CL, assumes that lever arm length analytical continuum
section 4.2 the CL of interest as well as mechanics [61], tilting
the PSD settings are equivalent a mirrored substrate
to the ones used for calibration
(7) Horizontal lever, Generating a torque via the N Y N Y Y 1 0 Requires attachment of a stiff Not available; Dimensional via
Toikka et al [33], gravitational pull of a lever lever to the CL, significant distance needed to level the analytical continuum
section 4.2 attached in an asymmetric deflection for soft CLs only lever mechanics [61]
manner to the CL
(8) Hammer-head CL, Lever method by lifting the N N Y Y Y 0 0 Requires attachment of a δαφ = 0.8%; None
Reitsma [34], outer ends of a cross-beam cross-beam to the CL, the lever arm length
section 4.2 attached to the CL additional mass reduces its
resonance frequency

Topical Review
(9) Off-axis colloidal Torque via a colloidal particle Y Y N Y Y 0 0 Requires a test probe with a δSx = 11.9%, Lateral compliance
probe, Quintanilla mounted off-axis + normal colloidal particle δkφ = 16.6%; particle method of [39], kφ /kz
et al [38], section 4.2 added mass method [89] for kz offset distance ratio
J. Phys. D: Appl. Phys. 43 (2010) 063001
Table 4. Continued.
PSD Over-all Suitable for Risk of tip User Methodological Reported uncertainties;
Name of method, sens. sens. colloidal damage friendliness requirements or main source of Compared with
reference Approach S φ kφ α φ Direct probes 0 (lo)–2 (hi) 0 (lo)–2 (hi) complications uncertainty method
(10) Normal FDCs on Vector analysis of forces N N Y Y Y 1 1 Grid commercially available δαx = 15.2%; Wedge method
wedge sample, acting when measuring FDCs slope dependence of (dev. ∼13%)
Asay et al [40, 83], on slopes adhesion force
section 4.1
(11) Lateral Calculation of kφ using Y Y N Y Y 1 1 Assumption of stiff Not available; [28] provides
compliance method, FEA + static friction probe–substrate contact (lateral contact stiffness) comparison with the
Cain et al [28, 39], measurement and relying on requires colloidal probe methods of [45]
section 3.1, section 4.3 high contact stiffness by using and [73]
colloidal probes
target
(12) Lateral FDC using Torsional Sader + lateral FDC Y Y N Y Y 0 1 Requires a test probe of δSx = 0.5%, Wedge method
target
a test probe, against vertical wall similar width and reflectivity, δkx = 6.8%, wedge
δαx = 4.2%
Cannara et al [36], assumes same laser intensity target
δαx = 7.5%;
section 3.1, section 4.4 distribution for test CL and for length of target CL
target CL torsional moment arm
(13) Vertical lever, Pushing a colloidal probe Y Y N Y Y 1 1 Requires a colloidal probe and Not available; Dimensional,
Ecke et al [37], laterally against a rigid vertical a test lever of known normal piezo response horizontal lever [29]
section 5.1 surface and a vertical lever stiffness
28

(14) Vertical lever with Bending a reference CL via N N Y Y Y 1 1 Requires a calibrated reference δαx = 12.4%; Modified dimensional
integrated lateral loading of the contact lever with integrated piezoresistive force sensor (t via f1 ),
piezo-resistive sensor, between its edge and the tip, piezoresistive sensor and sensitivity, vertical offset δαxdim = 33%
Xie et al [41], its bending is measured using related electronics, contact between contact point and
section 5.1 an integrated between tip and the edge-sided tip apex
piezoresistive sensor face of the reference lever
(15) Bending a fibre Loading a glass fibre at various Y Y N Y Y 1 1 Contact point between tip and δkx = 9%, δSx = 3%; Friction force
clamped at its one end, distances from its clamped end fibre ill-defined tip–fibre contact point measurement on
Morel et al [90], defining the torsional arm various materials
section 5.2 length
(16) Bending a fibre Scanning across the free end N N Y Y Y 2 1 Severe risk of damage to the δαφ = 26%; Dimensional
supported by a of a glass fibre clamped at the tip, contact point between tip tip–fibre contact point δαcdim = 56%
substrate, Liu other end and supported by a and fibre ill-defined, neglects defining the torsional arm
et al [69], section 5.2 substrate frictional forces length
fibre–substrate
(17) Lateral electrical Scanning on a suspended N N Y Y Y 2 1 Design, fabrication, and δαx = 7%; None
nano-balance (LEN), platform of a MEMS device calibration of the MEMS stiffness of the suspension
Cumpson et al [43], devices are time-consuming system
section 5.3
(18) Micro-lateral Scanning on a suspended N Y N Y Y 2 1 Design, fabrication, and Not available; Plausibility test using

Topical Review
force sensor (MLFS), platform of a MEMS device calibration of the MEMS stiffness of the suspension an expression for
Ando et al [44], with feedback via comb drive devices are time-consuming system tangential force
section 5.3 from [45]
J. Phys. D: Appl. Phys. 43 (2010) 063001
Table 4. Continued.
PSD Over-all Suitable for Risk of tip User Methodological Reported uncertainties;
Name of method, sens. sens. colloidal damage friendliness requirements or main source of Compared with
reference Approach S φ kφ α φ Direct probes 0 (lo)–2 (hi) 0 (lo)–2 (hi) complications uncertainty method
(19) Diamagnetic Lateral loading (static friction) N N Y Y Y 1 1 The magnetic levitation δαxx = 0.5%, Wedge
lateral force calibrator of a platform suspended by system can be too large for δαxz = 2.4%; method + Sader for kz
(D-LFC), Li et al [42], magnetic forces many AFM systems, n.a. for lateral displacement of PG
section 5.4 CLs with magnetic coating sheet
 
α
(20) Wedge method by Scanning across well-defined N N Y N N 2 1 The sharp edges between two δ αφz = 12%; Dimensional; [28]:
Carpick (facets of a slopes of a crystal surface facets can lead to tip damage, piezo response comparison with the
crystal surface), (SrTiO3 ) and analysing the restriction to a CL with a sharp methods of [39]
Ogletree et al [45], friction loops, the tip, assumes Amontons’ law, and [73].
section 6.1 measurements need to be considers the adhesion as a
made for a range of loads force offset
(21) Wedge method by Scanning across well-defined N N Y N Y 2 1 Grid commercially available, δαx = 11% for integrated None
Varenberg slopes of a grid (TGF11) and scanning steep slopes is probe, δαx = 8% for
(micro-fabricated analysing the friction loops, demanding for the colloidal probe;
trenches), requires measurement of the topography feedback, adhesion force via pull-off
Varenberg et al [46], pull-off force but no variation calibrated factor was found to measurement results in load
section 6.1 in load be load dependent, n.a. to dependence of calibration
large colloidal probes factor
29

(22) Wedge method by Improved analysis method, by N N Y N Y 2 1 Grid commercially available, δαφ = 4%; None
Varenberg, case of replacing the measurement of scanning steep slopes is ill-defined contact (contact
colloidal probes, the pull-off force with the demanding for the area, load) due to large
Ling et al [20], measurement of friction loops topography feedback, n.a. to slope angles
section 6.1 for a range of loads large colloidal probes
(23) Wedge method by Scanning across well-defined N N Y N Y 2 1 Fabrication of the slopes using δαx = 5%; Calculation of the CL
Varenberg, using shallow slopes and analysing FIB, prone to optical non-zero global sample tilt spring constant
shallow slopes, the friction loops interference on the sloped (analytical & via
Tocha et al [40], areas FEA) + opt. geometry
section 6.1 method of [73]
(24) Parallel scanning, Employing the calibrated N N Y N Y 2 2 Requires calibration of the Not available; Macro-scale friction
Ruan et al [48], normal forces by scanning in normal force signal; assumes piezo response, tip measurement
section 6.2 direction parallel to the CL negligible hysteresis of the asymmetry, normal spring
long axis (longitudinal scanner; assumes a symmetric constant
scanning) tip
(25) Parallel scanning Employing the calibrated N N Y N Y 2 2 Requires calibration of the Not available; Parallel scanning
with modified force normal forces by scanning in normal force signal; assumes piezo response, tip method of Ruan
analysis, direction parallel to the CL negligible hysteresis of the asymmetry, normal spring et al [48]
Wang et al [49], long axis (longitudinal scanner; assumes a symmetric constant; CL inclination
section 6.2 scanning) and analysing the tip angle

Topical Review
forces relative to the long axis
of the CL beam
J. Phys. D: Appl. Phys. 43 (2010) 063001
Table 4. Continued.
PSD Over-all Suitable for Risk of tip User Methodological Reported uncertainties;
Name of method, sens. sens. colloidal damage friendliness requirements or main source of Compared with
reference Approach S φ kφ α φ Direct probes 0 (lo)–2 (hi) 0 (lo)–2 (hi) complications uncertainty method
(26) Parallel scanning Employing the calibrated N N Y N Y 2 2 Requires calibration of the δαx > 12%; Wedge method
and friction force normal forces by scanning in normal force signal; assumes piezo response, tip wedge
δαx > 10%
scenarios, direction parallel to the CL negligible hysteresis of the asymmetry, normal spring
Karhu et al [50], long axis (longitudinal scanner; assumes a symmetric constant
section 6.2 scanning) and applying tip
various friction force models
(27) Piezotube bow, Employing the calibrated Y Y N N Y 2 1 Generates variation in normal Not available; None
Bilas et al [63], normal forces by calculation of force along a scan line via bow piezo response
section 6.2 lateral forces via analytical of a tube scanner
relationship between normal
and lateral CL spring constant
(28) Torsional added Attaching small masses to the N Y N Y Y 0 2 Requires accurate attachment δkφ = 25%; Torsional Sader,
mass, CL and analysing the changes of small spheres to the CL and on-axis positioning of the normal added mass
Green et al [51], in its fundamental torsional the measurement of its sphere on the CL method
section 7 resonance frequency fundamental torsional [89] + equation for
resonance frequency kφ /kz -ratio
30

(29) Torsional Sader, Exploiting the fundamental N Y N Y Y 0 2 Requires measurement of the Not available; Torsional added mass
Green et al [51], torsional resonance torsional resonance peak and value of the hydrodynamic method [51], normal
section 7 of the plan view dimensions of function added mass
the CL method [89] + equation
for kφ /kz -ratio
(30) Torsional Applying a static magnetic N Y N Y Y 0 0 Requires a static magnetic Not available; None
deflection via Lorentz field and driving an electric field and an electrically magnetic field strength and
force, current across the two legs of conductive pattern that allows area of the current loop if
Jeon et al [52], the CL the flow of electrical current the absolute value of kφ is
section 7 across the two legs of a calculated, kz if the
triangular CL rel. value is calculated
(T1) Analytical Analytical equations derived N Y N N Y 0 2 Knowledge of the CL δkφ = 30–50%; Finite element
continuum mechanics, from continuum mechanics dimensions and of the elastic CL thickness analysis (FEA)
Neumeister and properties of the CL material
Ducker [61],
section 2.2
(T2) FEA and Computational modelling N Y N N Y 0 1 Knowledge of the CL δkφ = 30–50%;
analytical expression using finite element analysis dimensions and of the elastic CL thickness
for kx /kz , (FEA) properties of the CL material,
Hazel and definition of a suitable mesh
Tsukruk [64],

Topical Review
section 2.2
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

of a single asperity), it can be helpful to evaluate the shape of lateral forces. Although non-zero mechanical crosstalk is
its foremost end, its apex. Importantly, its shape and its friction likely to occur for most CLs, its quantification seems to be
characteristics can change in the course of an experiment [105]. widely neglected. However, a thorough calibration method for
In particular, the wedge method (entries (20)–(23) in lateral forces should provide a comprehensive quantification,
table 4) is frequently used as a reference when a novel method including crosstalk coefficients. They could support the user
is evaluated. Provided a suitable substrate including sloped in his efforts to reduce misalignments in the optical detection
features, the experimental realization of the wedge method is system. Also, such information can be used to select CLs
comparatively easy. In addition to the choice of the friction law suitable for quantitative lateral force measurements. Finally,
used to derive the related equations, a practical point of concern by elimination of crosstalk effects, the discrepancies of data
can be the steepness of the slopes. A sluggish topography obtained with different experimental setups should be reduced,
feedback may lead to significant variations in the load if the i.e. the measurement data become less dependent on the
slope angle is large, thus reducing the accuracy of the method. particulars of the instrument.
Furthermore, it should be mentioned that the suspended Interestingly, all calibration methods that apply the lateral
platform methods (17) and (18) of table 4 do not rely on a force via the tip of the CL can be affected by mechanical
friction law, although they employ a sliding motion of the crosstalk. By coming close to the configuration of an LFM
tip. Provided a proper alignment of the scanning motion measurement, also its less favourable properties are included.
with the calibrated suspension system, methods that involve In contrast, the direct methods that generate a torque via a
a suspended platform allow for quantification of the friction lever other than the tip and the torsional method could also
forces through its lateral displacement. In the approach of be applied to a tip-less CL. That is, they are not affected by
Ando et al (method (18) of table 4), the lateral displacement is a positional offset of the tip. Deviations in results from a
kept constant by a feedback loop that applies a variable voltage lever method and from a method that involves the tip may
to a comb drive. In contrast to the methods (17) and (18), Li indicate the presence of a non-zero mechanical crosstalk effect.
et al operated their diamagnetic levitation system (method (19) Thus, the lever methods and the torsional resonance methods
of table 4) in the static friction regime, i.e. they measured the can be considered complementary methods, in that they can
platform displacement upon lateral loading of the tip–sample be employed to verify the results from other methods. In
contact. particular, comparison with results from methods that involve
Calibration methods involving static lateral loading the tip can indicate non-zero contributions from mechanical
(methods (11), (14) and (19) of table 4) of the tip–substrate crosstalk.
contact can suffer from micro-slip that occurs before the actual The diamagnetic levitation method (method (19) of
stick–slip transition. Furthermore, for tip–substrate systems table 4) delivers a calibration factor related to crosstalk.
that show very low friction forces or a stick–slip transition Although the experimental setup involves strong magnetic
for very low force values, the force range over which static fields and may not be compatible with all commercially
friction occurs is very small and it can be difficult to measure available microscopes, it can be interesting for basic
significant changes in the lateral signal before the stick–slip experiments on lateral force calibration since it provides a
transition occurs. compliant suspension system. As a method involving a
Elegant and comparatively facile methods of lateral force suspended platform, it is conceptually similar to the MEMS
calibration are given by the approaches that employ the devices of methods (17) and (18), and a similar analysis might
torsional resonance (methods (28)–(30) of table 4), similar be achieved using such devices.
to the corresponding methods for normal force calibration.
With the Sader method (method (29) of table 4), there is
almost no risk of damage to the tip or to the CL beam. 9. Concluding remarks and outlook
Implementation of the method requires a setup for capturing
the thermomechanical response of the CL up to the megahertz Eventually, for the establishment of one or several methods
range. To ensure a sufficient SNR, also a setup for the for the quantification of lateral forces in AFM they will
excitation of torsional resonances can be beneficial. Although need to meet a number of requirements. Assuming that the
such spectra can be recorded with most commercially available evolution of lateral force calibration methods will take a similar
AFM instruments of the latest generation, setups for the development than the evolution of normal force calibration
excitation of torsional resonances do not seem to be widely methods, it can be worth analysing the trends in normal force
available yet. However, the recent rise in interest in calibration.
measurement techniques based on torsional resonances is After an initial phase of exploration of a large variety of
likely to spark further instrumental developments on the part experimental and theoretical approaches for the calibration
of the manufacturers. Concerning the hydrodynamic function, of normal forces, efforts need to be made to ensure that
it seems important to map the parameter space that allows for the most promising calibration methods are widely accepted.
negligible deviations from the tabulated values. That is, they should be accepted worldwide across labs
In addition to the determination of the torsional spring operated by industry, academia or governments. In addition
constant and the sensitivity of the optical detection system, to general applicability, ease of use and a high level of
an ideal calibration method should also deliver quantitative accuracy, ultimately this also means traceability to the Système
information on the amount of crosstalk between normal and International d’Unités, i.e. to the SI system of units.

31
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

To address such requirements, traceably calibrated force particulars of the molecular system and on the chemical
cells have been developed that can generate forces in the nN functionalization of the tip, additional effects need to be
range. New approaches need to be taken since the traditional accounted for, such as the deformation of tether molecules
way of generating forces via deadweights would lead to high used as a linker [117].
uncertainties [106, 107]. For realizing and disseminating small Although thoroughly studied systems are available and
forces in a way that is traceable with the established framework the related rupture force values seem to be widely accepted,
of SI units, National Measurement Institutes (NMIs) developed additional validation of such molecular force standards by
alternative methods, such as the electrostatic force balance alternative measurement techniques is recommendable. A
(EFB) [106, 108–110]. For a simple parallel-plate capacitor, promising technique is optical trapping since it allows for
the force between the capacitor plates is given by the voltage ultra-low spring constants, ranging from ∼0.1 to 100 fN nm−1 ,
between the plates and the capacitance gradient in the direction
and low loading rates [118]. Albeit the force regime of this
of the force. Since displacement, voltage and capacitance can
emerging measurement technique is somewhat complementary
all be measured traceably and with a high accuracy, EFBs seem
to the one of typical AFM CLs, bond rupture force values
to be a viable approach.
measured with optical traps (laser tweezers) can be used to
The related electrical units Volt (V) and Farad (F) are
validate AFM-based force measurements. Using carefully
derived units in the SI; however, they can be linked to quantum
selected molecular systems and appropriate experimental
invariants via the Josephson and the quantum Hall effect,
respectively [106]. Owing to these links to quantum effects, configurations, normal as well as lateral force measurements
they can be measured over large dynamic ranges with small may benefit from molecular force standards. An even more
variations in relative uncertainty. direct comparison of forces may be achieved by combining
In turn, an EFB can be used to calibrate secondary AFM systems with optical traps [119].
artefacts, e.g. piezoresistive CLs [111, 112]. In an AFM Finally, improvements in the quantification of lateral (and
instrument, the calibrated piezoresistive CL can be loaded normal) forces may also be achieved through new approaches
like a specimen and the resulting electric readout be used in the probe design. Novel CL designs could allow for
to calibrate the mounted CL [112]. In this way, an indirect elimination of some of the shortfalls of the widely used
comparison with the EFB is achieved. rectangular or triangular CLs. Some recent developments
Other approaches to generating and measuring ultra- may indicate the way forward. As mentioned in section 4.5,
small forces are conceivable. A MEMS-based approach was Fukuzawa et al [86] suggested a CL design that allows for
demonstrated by Koch et al [113]. They generated forces in the full decoupling of the deflections resulting from normal and
piconewton range by exposing a magnetic bead to the field of from lateral forces, thus avoiding mechanical crosstalk. In
an electromagnet. The bead was attached to a microfabricated particular, the availability of two separate springs for normal
spring, and its deflection was measured as a function of the and lateral forces means that the respective spring constants are
electric current through the coils of the electromagnet. Such a independent of each other, although the fabrication process
MEMS system could be used for the calibration of AFM CLs, may impose some limitations. For instance, such a design
similar to the approach of Cumpson et al [43]. It would need approach allows for the realization of a ratio of lateral to normal
to be shown if this approach is SI traceable. stiffness that is <1, i.e. for the application of high loads while a
A practical approach to verifying the force readout of high sensitivity for small lateral forces is maintained. A similar
an AFM system is the measurement of the intrinsic forces configuration is provided by the approach of Dienwiebel
of well-studied molecular systems. A typical example is et al [120]. However, their design involves separate spring
the B–S transition of DNA [27, 114, 115]. It is known to systems for the two lateral directions and can be interfaced
occur around ∼65 pN for most types of DNA. Similarly, the with fibre-optic interferometers. Also Beyder and Sachs [121]
rupture force of binding sites of complementary biomolecules
demonstrated dual-axis probes, although with a focus on high-
can be used, although the dependence on the loading rate
frequency operation in liquids.
needs to be accounted for [12, 27, 116]. In the case of the
biotin–streptavidin system, a typical example of a biological In conclusion, the ideal calibration method allows for high
system that shows specific ligand–receptor interactions, the accuracy, is traceable to the SI units, and provides a measure for
bond rupture force was found to range between ∼100 and the amount of crosstalk. To ensure its widespread application,
400 pN [116]. Alternatively, such well-known intrinsic forces it should be compatible with most AFM instruments, easy to
of widely available molecular systems can be used as a use, applicable to wide ranges of CL shapes and of lateral
force standard and employed for the calibration of CL spring forces. Although a considerable number of approaches for
constants. the calibration of lateral forces has been demonstrated so far,
However, to establish such calibration methods, robust further work is required to provide the necessary levels of
protocols are required that allow for reproducibility of the accuracy, to ensure compatibility with general metrological
measurements and sufficient accuracy of the data. For instance, requirements, to establish them in the AFM community as well
it needs to be ensured that force contributions due to non- as adjoining user groups, and to cover the full range of forces,
specific binding are negligible or can be separated. Similarly, from the piconewton scale of single molecule experiments up
general acceptance of such an approach requires establishment to the micronewton scale of tribological testing of hard surfaces
of a suitable data analysis technique. Depending on the or coatings.

32
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

Acknowledgments [34] Reitsma M G 2007 Rev. Sci. Instrum. 78 106102


[35] Bogdanovic G, Meurk A and Rutland M W 2000 Colloids
The author would like to thank C A Clifford, D Roy and Surf. B 19 397
[36] Cannara R J, Eglin M and Carpick R W 2006 Rev. Sci.
I S Gilmore for helpful comments. Also, the author is grateful Instrum. 77 053701
for valuable comments by the referees. This work is supported [37] Ecke S, Raiteri R, Bonaccurso E, Reiner C, Deiseroth H-J and
by the National Measurement System of the UK Department Butt H-J 2001 Rev. Sci. Instrum. 72 4164
for Business, Innovation and Skills through the Chemical and [38] Quintanilla M A S and Goddard D T 2008 Rev. Sci. Instrum.
Biological Metrology Programme. This paper is reproduced 79 023701
[39] Cain R G, Biggs S and Page N W 2000 J. Colloid Interface
with the permission of Her Majesty’s Stationery Office (Crown Sci. 227 55
copyright, 2010). [40] Asay D B and Kim S H 2006 Rev. Sci. Instrum. 77 043903
[41] Xie H, Vitard J, Haliyo S, Régnier S and Boukallel M 2008
Rev. Sci. Instrum. 79 033708
References [42] Li Q, Kim K-S and Rydberg A 2006 Rev. Sci. Instrum.
77 065105
[1] Mate C M, McClelland G M, Erlandsson R and Chiang S 1987 [43] Cumpson P J, Hedley J and Clifford C A 2005 J. Vac. Sci.
Phys. Rev. Lett. 59 1942 Technol. B 23 1992
[2] Sheiko S S 2000 Adv. Polym. Sci. 151 61 [44] Ando Y and Shiraishi N 2007 Rev. Sci. Instrum. 78 033701
[3] Perry S S 2004 MRS Bull. 29 478 [45] Ogletree D F, Carpick R W and Salmeron M 1996 Rev. Sci.
[4] Perry S S and Tysoe W T 2005 Tribol. Lett. 19 151 Instrum. 67 3298
[5] Munz M, Schulz E and Sturm H 2002 Surf. Interface Anal. [46] Varenberg M, Etsion I and Halperin G 2003 Rev. Sci. Instrum.
33 100 74 3362
[6] Raczkowska J, Montenegro R, Budkowski A, Landfester K,
[47] Tocha E, Schonherr H and Vancso G J 2006 Langmuir 22 2340
Bernasik A, Rysz J and Czuba P 2007 Langmuir 23 7235
[48] Ruan J A and Bhushan B 1994 J. Tribol.—Trans. ASME
[7] Bhushan B, Kasai T, Kulik G, Barbieri L and Hoffmann P
116 378
2005 Ultramicroscopy 105 176
[49] Wang Y-L, Zhao X-Z and Zhou F-Q 2007 Rev. Sci. Instrum.
[8] Papastavrou G and Akari S 1999 Nanotechnology 10 453
78 036107
[9] Sun S, Chong K S L and Leggett G J 2005 Nanotechnology
[50] Karhu E, Gooyers M and Hutter J L 2009 Langmuir
16 1798
25 6203–13
[10] Vezenov D V, Noy A, Rozsnyai L R and Lieber C M 1997
[51] Green C P, Lioe H, Cleveland J P, Proksch R, Mulvaney P and
J. Am. Chem. Soc. 119 2006
Sader J E 2004 Rev. Sci. Instrum. 75 1988
[11] Smith D A, Connell S D, Robinson C and Kirkham J 2003
Anal. Chim. Acta 479 39 [52] Jeon S, Braiman Y and Thundat T 2004 Appl. Phys. Lett.
[12] Butt H J, Cappella B and Kappl M 2005 Surf. Sci. Rep. 59 1 84 1795
[13] Frisbie C D, Rozsnyai L F, Noy A, Wrighton M S and [53] Meyer E, Lüthi R, Howald L, Bammerlin M, Guggisberg M
Lieber C M 1994 Science 265 2071 and Güntherodt H J 1996 J. Vac. Sci. Technol. B 14 1285
[14] Liu H and Bhushan B 2004 Ultramicroscopy 100 391 [54] Putman C and Kaneko R 1996 Thin Solid Films 273 317
[15] Williams J A and Le H R 2006 J. Phys. D: Appl. Phys. 39 R201 [55] Schwarz U D, Zwörner O, Köster P and Wiesendanger R 1997
[16] Wright C J and Armstrong I 2006 Surf. Interface Anal. 38 1419 Phys. Rev. B 56 6987
[17] Pettersson T, Nordgren N, Rutland M W and Feiler A 2007 [56] Munz M, Cappella B, Sturm H, Geuss M and Schulz E 2003
Rev. Sci. Instrum. 78 093702 Adv. Polym. Sci. 164 87
[18] Kim S-K, Jung M-H, Kim H-W, Woo S-G and Lee H 2005 [57] Draft of the standard ‘Surface chemical
Nanotechnology 16 2227 analysis—Vocabulary: II. Terms for the scanned probe
[19] Munz M, Cox D C and Cumpson P J 2008 Phys. Status Solidi microscopies’, ISO TC 201/SC 1 – ISO/DIS 18115-2,
a 205 1424 version 29 August 2008
[20] Ling X, Butt H-J and Kappl M 2007 Langmuir 23 8392 [58] Marti O 1999 AFM instrumentation and tips Handbook of
[21] Bowden F P and Tabor D 1964 The Friction and Lubrication Micro/Nano Tribology 2nd edn, B Bhushan (Boca Raton,
of Solids (Oxford: Clarendon) FL: CRC Press) chapter 2
[22] Bhushan B 1999 Micro/nanotribology and [59] Choi D, Hwang W and Yoon E 2007 J. Microsc. 228 190
micro/nanomechanics of magnetic storage devices [60] Bergmann-Schaefer 1974 Lehrbuch der
Handbook of Micro/Nano Tribology 2nd edn, ed B Bhushan Experimentalphysik. Bd. 1: Mechanik, Akustik, Wärme. 9th
(Boca Raton, FL: CRC Press) chapter 14 edn (Berlin: Walter de Gruyter) chapter 5
[23] Degiampietro K and Colaço R 2007 Wear 263 1579 [61] Neumeister J M and Ducker W A 1994 Rev. Sci. Instrum.
[24] Luzinov I, Julthongpiput D, Gorbunov V and Tsukruk V V 65 2527
2001 Tribol. Int. 34 327 [62] Gupta R K 2000 J. Microelectromech. Syst. 9 380
[25] Watson J A, Brown C L, Myhra S and Watson G S 2006 [63] Bilas P, Romana L, Kraus B, Bercion Y and Mansot J L 2004
Nanotechnology 17 2581 Rev. Sci. Instrum. 75 415
[26] Bockelmann U, Essevaz-Roulet B and Heslot F 1997 Phys. [64] Hazel J L and Tsukruk V V 1998 J. Tribol. 120 814
Rev. Lett. 79 4489 [65] Clifford C A and Seah M P 2005 Nanotechnology 16 1666
[27] Alessandrini A and Facci P 2005 Meas. Sci. Technol. 16 R65 [66] Hazel J L and Tsukruk V V 1999 Thin Solid Films 339 249
[28] Cain R G, Reitsma M G, Biggs S and Page N W 2001 Rev. Sci. [67] Sarid D 1991 Scanning Force Microscopy—With Applications
Instrum. 72 3304 to Electric, Magnetic, and Atomic Forces (New York:
[29] Feiler A, Attard P and Larson I 2000 Rev. Sci. Instrum. 71 2746 Oxford University Press) chapter 1
[30] Clifford C A and Seah M P 2005 Appl. Surf. Sci. 252 1915 [68] Mendels D-A, Lowe M, Cuenat A, Cain M G, Vallejo E,
[31] Cappella B and Dietler G 1999 Surf. Sci. Rep. 34 1 Ellis D and Mendels F 2006 J. Micromech. Microeng.
[32] Lantz M A, O’Shea S J, Hoole A C F and Welland M E 1997 16 1720
Appl. Phys. Lett. 70 970 [69] Liu W, Bonin K and Guthold M 2007 Rev. Sci. Instrum.
[33] Toikka G, Hayes R A and Ralston J 1997 J. Adhes. Sci. 78 063707
Technol. 11 1479 [70] Sader J E and Green C P 2004 Rev. Sci. Instrum. 75 878

33
J. Phys. D: Appl. Phys. 43 (2010) 063001 Topical Review

[71] Song Y and Bhushan B 2008 J. Phys.: Condens. Matter [96] Wang F and Zhao X 2007 Rev. Sci. Instrum. 78 043701
20 225012 [97] Stiernstedt J, Rutland M W and Attard P 2005 Rev. Sci.
[72] Xie H, Vitard J, Haliyo S and Régnier S 2008 Rev. Sci. Instrum. 76 083710
Instrum. 79 096101 [98] Huang L and Su C 2004 Ultramicroscopy 100 277
[73] Liu E, Blanpain B and Celis J P 1996 Wear 192 141 [99] Sahin O and Erina N 2008 Nanotechnology 19 445717
[74] Tocha E, Song J, Schonherr H and Vancso G J 2007 Langmuir [100] Sader J E, Chon J W M and Mulvaney P 1999 Rev. Sci.
23 7078 Instrum. (Boca Raton, FL: CRC Press) 3967
[75] Meurk A, Larson I and Bergstrom L 1998 Mater. Res. Soc. [101] Reinstaedtler M, Rabe U, Scherer V, Turner J A and
Symp. Proc. 522 427 Arnold W 2003 Surf. Sci. 532–5 1152
[76] Bhushan B 1999 Introduction—measurement techniques and [102] Van Eysden C A and Sader J E 2007 J. Appl. Phys.
applications Handbook of Micro/Nanotribology 2nd edn 101 044908
ed B Bhushan (Boca Raton, FL: CRC Press) chapter 1 [103] Maali A, Hurth C, Boisgard R, Jai C, Cohen-Bouhacina T
[77] Such M W, Kramer D E and Hersam M C 2004 and Aimé J-P 2005 J. Appl. Phys. 97 074907
Ultramicroscopy 99 189 [104] Mullin N and Hobbs J 2008 Appl. Phys. Lett. 92 053103
[78] Piner R and Ruoff R S 2002 Rev. Sci. Instrum. 73 3392 [105] Carpick R W, Agraı̈t N, Ogletree D F and Salmeron M 1996
[79] Prunici P and Hess P 2008 Ultramicroscopy 108 642 Langmuir 12 3334
[80] Varenberg M, Etsion I and Halperin G 2003 Rev. Sci. Instrum. [106] Pratt J R, Kramar J A, Newell D B and Smith D T 2005
74 3569 Meas. Sci. Technol. 16 2129
[81] Onal C D, Sumer B and Sitti M 2008 Rev. Sci. Instrum. [107] Jabbour Z J and Yaniv S L 2001 J. Res. Natl Inst. Stand.
79 103706 Technol. 106 25
[82] Michal G, Lu C and Tieu A K 2008 Nanotechnology [108] Leach R, Chetwynd D, Blunt L, Haycocks J, Harris P,
19 455707 Jackson K, Oldfield S and Reilly S 2006 Meas. Sci. Technol.
[83] Asay D B, Hsiao E and Kim S H 2009 Rev. Sci. Instrum. 17 467
80 066101 [109] Ku Y S 2004 Meas. Sci. Technol. 15 2108
[84] Pittenger B 2008 HarmoniX Application Note AN112 Rev. AO [110] Choi I-M, Kim M-S, Woo S-Y and Kim S H 2004 Meas. Sci.
Veeco Instruments Inc. Technol. 15 237
[85] Carpick R W, Ogletree D F and Salmeron M 1997 Appl. Phys. [111] Gates R S and Pratt J R 2006 Meas. Sci. Technol. 17 2852
Lett. 70 1548 [112] Langlois E D, Shaw G A, Kramar J A, Pratt J R and
[86] Fukuzawa K, Terada S, Shikida M, Amakawa H, Zhang H and Hurley D C 2007 Rev. Sci. Instrum. 78 093705
Mitsuya Y 2007 J. Appl. Phys. 101 034308 [113] Koch S J, Thayer G E, Corwin A D and de Boer M P 2006
[87] Sahin O, Magonov S, Su C, Quate C F and Solgaard O 2007 Appl. Phys. Lett. 89 173901
Nature Nanotechnol. 2 507 [114] Cluzel P, Lebrun A, Heller C, Lavery R, Viovy J-L,
[88] Gibson C T, Watson G S and Myhra S 1996 Nanotechnology Chatenay D and Caron F 1996 Science 271 792
7 259 [115] Smith S B, Cui Y and Bustamante C 1996 Science 271 795
[89] Cleveland J P, Manne S, Bocek D and Hansma P K 1993 Rev. [116] Lo Y S, Zhu Y-J and Beebe T P 2001 Langmuir 17 3741
Sci. Instrum. 64 403 [117] Ray C, Brown J R and Akhremitchev B B 2007 Langmuir
[90] Morel N, Tordjeman P and Ramonda M 2005 J. Phys. D: Appl. 23 6076
Phys. 38 895 [118] Ota T, Sugiura T and Kawata S 2005 Appl. Phys. Lett.
[91] Cumpson P J and Hedley J 2003 Nanotechnology 14 1279 87 043901
[92] Simon M D, Heflinger L O and Geim A K 2001 Am. J. Phys. [119] Huisstede J H G, Subramaniam V and Bennink M L 2007
69 702 Microscopy Res. Techn. 70 26
[93] Dunne P A, Hilton J and Coey J M D 2007 J. Magn. Magn. [120] Dienwiebel M, de Kuyper E, Crama L, Frenken J W M,
Mater. 316 273 Heimberg J A, Spaanderman D-J, Glastra van Loon D,
[94] Bhushan B and Ruan J A 1994 J. Tribol.—Trans. ASME Zijlstra T and van der Drift E 2005 Rev. Sci. Instrum.
116 389 76 043704
[95] Supporting Information of [20] [121] Beyder A and Sachs F 2006 Ultramicroscopy 106 838

34

Potrebbero piacerti anche