of 17
Spring constant calibration of atomic force microscope cantilevers of
arbitrary shape
John E. Sader, Julian A. Sanelli, Brian D. Adamson, Jason P. Monty, Xingzhan Wei et al.
Citation: Rev. Sci. Instrum. 83, 103705 (2012); doi: 10.1063/1.4757398
View online: http://dx.doi.org/10.1063/1.4757398
View Table of Contents: http://rsi.aip.org/resource/1/RSINAK/v83/i10
Published by the American Institute of Physics.
Related Articles
Compact metal probes: A solution for atomic force microscopy based tip-enhanced Raman spectroscopy
Rev. Sci. Instrum. 83, 123708 (2012)
Note: Radiofrequency scanning probe microscopy using vertically oriented cantilevers
Rev. Sci. Instrum. 83, 126103 (2012)
Switching spectroscopic measurement of surface potentials on ferroelectric surfaces via an open-loop Kelvin
probe force microscopy method
Appl. Phys. Lett. 101, 242906 (2012)
Enhanced quality factors and force sensitivity by attaching magnetic beads to cantilevers for atomic force
microscopy in liquid
J. Appl. Phys. 112, 114324 (2012)
Invited Review Article: High-speed flexure-guided nanopositioning: Mechanical design and control issues
Rev. Sci. Instrum. 83, 121101 (2012)
Additional information on Rev. Sci. Instrum.
Journal Homepage: http://rsi.aip.org
Journal Information: http://rsi.aip.org/about/about_the_journal
Top downloads: http://rsi.aip.org/features/most_downloaded
Information for Authors: http://rsi.aip.org/authors
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
REVIEW OF SCIENTIFIC INSTRUMENTS
83
, 103705 (2012)
Spring constant calibration of atomic force microscope
cantilevers of arbitrary shape
John E. Sader,
1,2,
a)
Julian A. Sanelli,
3
Brian D. Adamson,
3
Jason P. Monty,
4
Xingzhan Wei,
3,5
Simon A. Crawford,
6
James R. Friend,
7,8
Ivan Marusic,
4
Paul Mulvaney,
3,5
and Evan J. Bieske
3
1
Department of Mathematics and Statistics, The University of Melbourne, Victoria 3010, Australia
2
Kavli Nanoscience Institute and Department of Physics, California Institute of Technology, Pasadena,
California 91125, USA
3
School of Chemistry, The University of Melbourne, Victoria 3010, Australia
4
Department of Mechanical Engineering, The University of Melbourne, Victoria 3010, Australia
5
Bio21 Institute, The University of Melbourne, Victoria 3010, Australia
6
School of Botany, The University of Melbourne, Victoria 3010, Australia
7
Melbourne Centre for Nanofabrication, Clayton, Victoria 3800, Australia
8
MicroNanophysics Research Laboratory, RMIT University, Melbourne, Victoria 3001, Australia
(Received 15 June 2012; accepted 18 September 2012; published online 17 October 2012)
The spring constant of an atomic force microscope cantilever is often needed for quantitative measure-
ments. The calibration method of Sader
et al.
[Rev. Sci. Instrum.
70
, 3967 (1999)] for a rectangular
cantilever requires measurement of the resonant frequency and quality factor in fluid (typically air),
and knowledge of its plan view dimensions. This intrinsically uses the hydrodynamic function for a
cantilever of rectangular plan view geometry. Here, we present hydrodynamic functions for a series
of irregular and non-rectangular atomic force microscope cantilevers that are commonly used in prac-
tice. Cantilever geometries of arrow shape, small aspect ratio rectangular, quasi-rectangular, irregular
rectangular, non-ideal trapezoidal cross sections, and V-shape are all studied. This enables the spring
constants of all these cantilevers to be accurately and routinely determined through measurement of
their resonant frequency and quality factor in fluid (such as air). An approximate formulation of the
hydrodynamic function for microcantilevers of arbitrary geometry is also proposed. Implementation
of the method and its performance in the presence of uncertainties and non-idealities is discussed, to-
gether with conversion factors for the static and dynamic spring constants of these cantilevers. These
results are expected to be of particular value to the design and application of micro- and nanomechan-
ical systems in general.
© 2012 American Institute of Physics
.[
http://dx.doi.org/10.1063/1.4757398
]
I. INTRODUCTION
Knowledge of the stiffness of microcantilevers used in
the atomic force microscope (AFM) is essential for many ap-
plications of the instrument.
1
3
Over the past 20 years, many
techniques have been devised for the
in situ
measurement of
these spring constants. These methods allow the user to rou-
tinely and independently calibrate the spring constants of can-
tilevers during operation of the AFM. These calibration meth-
ods include dimensional approaches,
4
7
methods that probe
the static deflection of the cantilever induced by a calibrated
load,
8
12
and those that monitor the dynamic vibrational re-
sponse of the cantilever.
13
18
The performance of these tech-
niques has been widely explored and assessed, and the reader
is referred to Refs.
3
,
19
22
for detailed reviews.
The method of Sader
et al.
15
for rectangular cantilevers
makes use of the hydrodynamic load experienced by a can-
tilever as it oscillates in a fluid (such as air) – for clarity,
this approach shall henceforth be referred to as the “original
method” in this article. It was originally devised for rectangu-
lar cantilevers, for which the static normal spring constant
k
a)
Author to whom correspondence should be addressed. Electronic mail:
jsader@unimelb.edu.au.
is determined using the formula,
15
k
=
0
.
1906
ρb
2
LQ 
i
(
ω
R
)
ω
2
R
,
(1)
where
ρ
is the density of the fluid surrounding the cantilever,
b
and
L
are the cantilever width and length, respectively,
ω
R
and
Q
are the radial resonant frequency and quality factor in fluid
of the fundamental flexural mode, respectively, and

i
(
ω
R
)
is the imaginary part of the (dimensionless) hydrodynamic
function evaluated at the resonant frequency.
15
,
23
To imple-
ment this formula in practice, knowledge of the fluid density
and viscosity, cantilever width and length is required, and the
resonant frequency and quality factor must be measured. The
technique is independent of the thickness and material prop-
erties of the cantilever, which can be difficult to determine
in practice. The technique was extended to calibration of the
torsional spring constant of rectangular cantilevers in Ref.
24
.
Subsequently, the original method was generalized in
Ref.
25
to enable measurement of the spring constant of
any elastic body, including AFM cantilevers of arbitrary
geometry – this shall be referred to as the “general method.”
The general method relies on knowledge of the hydrodynamic
function
28
for a cantilever of arbitrary shape – a protocol for
its determination was also presented in Ref.
25
. With the
(dimensionless) hydrodynamic function for a specific type
0034-6748/2012/83(10)/103705/16/$30.00
© 2012 American Institute of Physics
83
, 103705-1
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-2
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
TR400(L)
BL-RC-150VB(L)
AC240TS
TR400(S)
ASYMFM
BL-RC-150VB(S)
AC160TS
TR800(L)
NCHR
AC240TM
FMR
TR800(S)
FIG. 1. SEM micrographs showing plan view geometries (shapes) of all can-
tilevers used in this study. Details of each cantilever, including dimensions,
are given in Table
I
.
of cantilever known (e.g., a particular V-shaped cantilever
model), the spring constant of any one of these cantilevers can
be determined from knowledge of its plan view dimensions,
and measurement of its resonant frequency and quality fac-
tor in fluid (typically air). The general method was tested and
validated for a series of rectangular and V-shaped cantilevers
in Refs.
22
and
25
, for which good agreement was found
with independent measurements of the spring constants. The
theoretical framework underpinning the general method also
explained its performance (and that of the original method)
under non-ideal but practical conditions, e.g., presence of an
imaging tip; see Ref.
25
.
Continuous advances in atomic force microscopy have
led to the development of a wide array of cantilever de-
signs, as evident from the micrographs presented in Fig.
1
.
Many of these cantilevers have designs that deviate strongly
from a rectangular geometry. Ability to calibrate the spring
constants of these cantilevers is thus of critical importance
to quantitative AFM measurements. The primary purpose of
this article is to report the hydrodynamic functions for a se-
ries of cantilevers that are commonly used in the AFM; see
Fig.
1
. This enables the normal spring constant of these
types of cantilevers to be routinely and accurately determined
through measurement of their resonant frequency and quality
factor in fluid (air). This is regardless of thickness variations,
compositional and size changes, provided the plan view ge-
ometry (shape) remains constant.
25
As we shall discuss, this
prerequisite is commonly satisfied in practice.
The cantilevers studied here possess significant non-
idealities in the form of arrow shaped ends, irregular rectan-
gular geometries, small aspect ratio rectangular geometries,
non-ideal trapezoidal cross sections of irregular shape and
V-shaped geometries. SEM micrographs of these cantilevers
are given in Fig.
1
. Importantly, the original method
15
im-
plicitly assumes that the cantilever plan view is rectangular
and that its aspect ratio (length/width) is large. This is be-
cause the fluid-structure theory
23
underpinning this method
is derived using this assumption. The method was found to
work well for ideally rectangular cantilevers with aspect ra-
tios (length/width) in the range 3.3
13.7, and rectangular
cantilevers of non-ideal geometry (ends slightly cleaved with
imaging tip) for aspect ratios 3.9–10;
15
the lower limit in as-
pect ratio for which the original method is valid was not de-
termined in Ref.
15
, and left as an open question. Some of the
cantilevers shown in Fig.
1
clearly do not satisfy these bounds
– they possess smaller aspect ratios, and have irregular and
non-rectangular shapes. It is important to emphasize that the
general method
25
is valid for any elastic body, and thus it au-
tomatically includes all non-ideal effects such as those due to
the imaging tip, arbitrary shape, and arbitrary aspect ratio. No
corrections to the reported hydrodynamic functions in this ar-
ticle, to account for any such non-idealities, are required or
warranted.
The AFM is frequently operated in two complementary
modes: (i) “static mode” where the static deflection of the
cantilever due to an applied force is monitored, and (ii) “dy-
namic mode” in which the cantilever is oscillated at or near
resonance.
1
3
,
19
22
Since the deflection functions of the can-
tilever for these two complementary modes are different, they
probe different spring constants. We remind the reader that the
spring constant of a cantilever at any point along its length is
defined as the second derivative of its potential energy with
respect to amplitude at that point.
26
The potential energy can
be written in terms of the elastic strain in the cantilever, and
therefore depends explicitly on the mode shape, i.e., the de-
flection function of the cantilever.
27
Consequently, the static
and dynamic spring constants of an AFM cantilever will dif-
fer, since the deflection functions in these two modes of oper-
ation are not identical. These complementary spring constants
are needed for quantitative analysis of static and dynamic
mode measurements. We therefore present numerical results
allowing for conversion between these two spring constants
for all cantilevers considered in Fig.
1
. This in turn allows the
general method to be used to determine both the static and dy-
namic normal spring constants. The dynamic spring constant
for only the fundamental flexural mode is considered in this
study.
Experimental protocols for determination of the hydro-
dynamic function
25
,
28
and implementation of the original and
general methods
22
are summarized. A discussion of the oper-
ation of the general method in the presence of random non-
idealities, such as uncertainty in cantilever dimensions and
cantilever clamping conditions, is also presented.
The article is organized as follows: We begin in Sec.
II
with a brief exposition of the theory underpinning the gen-
eral method and the experimental protocol for its implemen-
tation. Section
III
focuses on experimental determination of
the hydrodynamic function and implementation of the gen-
eral method. It is divided into several subsections that pro-
vide information on (a) cantilever dimensions, (b) spring con-
stant measurements, (c) measured hydrodynamic functions,
(d) a simplified approximate implementation of the general
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-3
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
method valid for any microcantilever, (e) conversion factors
for the static and dynamic spring constants, (f) effect of un-
certainty on the method, and (g) protocols for implementing
the method. Details pertinent to Secs.
II
and
III
can be found
in the Appendixes.
A. Summary
Readers primarily interested in the hydrodynamic func-
tions for the cantilevers in Fig.
1
are referred directly to
Table
III
and Eq.
(10)
, which are to be used with Eqs.
(7a)
and
(8)
. This completely specifies the general method for
each cantilever. Conversion factors for the static and dynamic
spring constants of these cantilevers are given in Table
IV
.
II. THEORETICAL AND EXPERIMENTAL
FRAMEWORK
We first summarize the theoretical framework of the gen-
eral method, which is applicable to any elastic body or de-
vice immersed in a viscous fluid. The device executes reso-
nant oscillations in the fluid. This theoretical framework is
then applied to the present case of interest: a cantilever of ar-
bitrary plan view shape undergoing resonant oscillations in
its fundamental flexural mode. The experimental protocol for
determination of the hydrodynamic function for a cantilever
of arbitrary shape is then given. For a detailed derivation of
this framework and a comprehensive discussion, the reader is
referred to Ref.
25
.
A. Arbitrary elastic device in fluid
The principal assumptions of the general method are:
(1) The body behaves as a linearly elastic solid;
(2) Energy dissipation due to vibration of the body occurs
in the fluid;
(3) The oscillation amplitude of the body is small, so that all
nonlinearities due to the body and fluid are negligible;
(4) The fluid flow generated by the oscillating body is in-
compressible.
These assumptions are commonly satisfied in practice,
from which the maximum energy stored in the oscillating
body at resonance directly follows:
E
stored
=
1
2
k
d
A
2
,
(2)
where
k
d
is the dynamic spring constant of the oscillation
mode, and
A
is the oscillation amplitude. The energy dissi-
pated in the fluid due to these resonant oscillations can be
quantified by the (dimensionless) quality factor,
Q
2
π
E
stored
E
diss
ω
=
ω
R
,
(3)
where
E
diss
is the energy dissipated per oscillation cycle, at
the resonant frequency
ω
R
.
Since the flow is linear, as discussed above, the energy
dissipated per oscillation cycle will depend on the square of
the oscillation amplitude,
A
. It therefore follows from Eqs.
(2)
and
(3)
that the dynamic spring constant (see definition in the
Introduction) is related to the quality factor by
k
d
=
(
1
2
π
2
E
diss
∂A
2
ω
=
ω
R
)
Q,
(4)
which is independent of the oscillation amplitude.
In accord with the above-mentioned assumptions, the en-
ergy dissipated per cycle
E
diss
must depend on (i) the square
of the device oscillation amplitude,
A
, (ii) the fluid density
ρ
and shear viscosity
μ
, (iii) the linear dimension (size) of
the device, denoted
L
0
, (iv) the relevant frequency of oscil-
lation, which from Eq.
(4)
is the resonant frequency of the
device immersed in fluid,
ω
R
, (v) the mode shape of the vi-
brating device, and (vi) the geometry of the device. Note that
the last two quantities are dimensionless. The relationship be-
tween the remaining quantities and
E
diss
can be rigorously
determined using dimensional analysis
29
– this gives two di-
mensionless groups. Use of Buckingham’s
π
theorem
29
then
yields the required result for the energy dissipated per cycle,
1
2
π
2
E
diss
∂A
2
ω
=
ω
R
=
ρL
3
0
ω
2
R

(
β
)
,
(5)
where

(
β
) is a dimensionless function that depends on the
dimensionless parameter,
β
ρL
2
0
ω
R
μ
,
(6)
which is often termed the inverse Stokes number or Womers-
ley number and is related to the Reynolds number, Re, defined
below. Substituting Eqs.
(5)
and
(6)
into Eq.
(4)
gives the re-
quired expression for the stiffness, as presented in Ref.
25
.
Note that

(
β
) also implicitly depends on the mode
shape and geometry of the body, as is evident from the above
discussion. Provided these dimensionless quantities do not
change,

(
β
) will remain invariant. This point is examined
further in Sec.
III
.
B. Cantilevers of arbitrary shape
The hydrodynamic flow induced by the flexural oscilla-
tions of a thin cantilever is dominated by its plan view geom-
etry, with its thickness exerting a negligible effect;
23
,
30
,
31
see
Sect.
II C 1
. Since AFM cantilevers typically have both small
and large length scales in their plan view geometry, e.g., the
width
b
and length
L
of the cantilever, these can both affect
the flow. We therefore define, without loss of generality,
Re
ρb
2
ω
R
4
μ
=
(
b
2
L
0
)
2
β,
(7a)
(Re)
L
3
0
b
2
L

(
β
)
.
(7b)
Note that the linear dimension (size)
L
0
is a characteristic
length scale of the flow; see Sec.
II A
. This length scale has
been replaced by a combination of the length,
L
, and width,
b
, to accommodate details of the flow generated by a vibrat-
ing microcantilever. Specifically, the dominant hydrodynamic
length scale for the flow is often the smaller length scale, e.g.,
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-4
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
the width
b
of the cantilever.
23
As such, the flow varies slowly
along the cantilever length,
L
, and rapidly over its width,
b
.
23
It then follows that the hydrodynamic volume over which
energy dissipation occurs scales as
b
2
L
for viscous bound-
ary layers of comparable size to the cantilever width.
23
The
rescalinginEq.
(7)
thus ensures that the (dimensionless) hy-
drodynamic function,
(Re),
28
is an order one quantity in
such situations – this case is often encountered in practice,
23
and is demonstrated in Sec.
III C
. The Reynolds number Re
contains the width
b
only, and can thus be formally interpreted
as the squared ratio of the dominant hydrodynamic length
scale to the viscous penetration depth. The redefinitions in
Eq.
(7)
thus facilitate physical interpretation and proper nor-
malization of all dimensionless quantities.
Substituting Eq.
(7)
into Eq.
(5)
and subsequently into
Eq.
(4)
, gives the required formula connecting the dynamic
spring constant to the dissipative properties of the cantilever
at resonance,
k
d
=
ρb
2
L
(Re)
ω
2
R
Q.
(8)
Comparing Eq.
(8)
to Eq.
(1)
reveals that the general method,
which is rigorously applicable to a cantilever of arbitrary
shape, yields an equation of identical form to that for a rect-
angular cantilever. This establishes that the original method,
for rectangular cantilevers of high aspect ratio, can be directly
extended to cantilevers of arbitrary shape. All that is needed is
the hydrodynamic function,
(Re), for the cantilever geom-
etry and mode in question. Equations
(1)
and
(8)
show that
the hydrodynamic functions

i
(
ω
) and
(Re) are related by a
constant factor for rectangular cantilevers of high aspect ratio.
Since the static and dynamic spring constants differ
by a constant multiplicative factor for a cantilever of fixed
plan view geometry, Eq.
(8)
is equally applicable to the
static spring constant under the appropriate renormalization.
The renormalization factors for all cantilever geometries in
Fig.
1
are given in Sec.
III E
.
Throughout we only consider the fundamental flexural
mode of vibration, even though the general method is rigor-
ously applicable to any mode. Note that the material prop-
erties of the cantilever do not enter into the derivation of
Eq.
(8)
, and thus the original and general methods are applica-
ble to cantilevers composed of any elastic material. The appli-
cability of these methods to devices whose thickness and/or
material properties vary along their length is discussed in
Sec.
II C
.
C. Properties of the general method
In this section, we present a discussion of several features
of the general method that are pertinent to its implementation.
1. Effect of finite cantilever thickness
For thin cantilevers executing flexural oscillations, the
hydrodynamic function,
(Re), depends only on the plan
view geometry of the cantilever and its mode shape, which
are both dimensionless quantities. Cantilever thickness plays
a relatively minor role in the hydrodynamic load (and en-
ergy dissipation) experienced by a cantilever undergoing flex-
ural oscillations, even for quite thick devices.
30
,
31
This is
because the load is dominated by contributions from the hy-
drodynamic pressure rather than the shear stress.
30
,
31
As such,
the cantilever plan view dimension to thickness ratio, e.g.,
the width-to-thickness ratio, does not exert a significant ef-
fect on the hydrodynamic function, and can be ignored.
30
,
31
This property is used in determination of the hydrodynamic
function in Sec.
II D
.
2. Effect of non-uniform cantilever thickness
and material properties
Spatial variations in thickness and/or material properties
also exert a weak effect on the general method and only enter
via their effect on the cantilever mode shape. This is because
the right hand side of Eq.
(8)
depends on the energy dissi-
pated in the fluid, not the cantilever thickness or material; see
above. The energy dissipated in the fluid is a weighted aver-
age of the mode shape over the cantilever plan view geometry.
Since the fundamental mode shape is a simple monotonically
increasing function of distance from the clamp, spatial varia-
tions in thickness and/or material have a weak effect on this
mode shape and hence the general method. This explains re-
cent theoretical findings demonstrating the robustness of the
original method, with respect to thickness variations along the
cantilever axis, in all but the extreme cases of very strong vari-
ations in thickness
32
– in these extreme cases, the mode shape
was significantly altered. The same property holds true for the
general method.
For the same reason, the presence of an imaging tip mass
also has a very minor effect on this mode shape, even in the
high tip mass to cantilever mass limit.
25
,
32
Nonetheless, if the
imaging tip is comparable in size to the dominant hydrody-
namic length scale of the flow, its presence will enhance the
true energy dissipation and thus increase the hydrodynamic
function.
24
,
25
While this can lead to an underestimate of the
spring constant obtained using the original method,
25
the gen-
eral method intrinsically accounts for any such extra energy
dissipation. This is because the hydrodynamic function is de-
termined in the presence of the imaging tip; see Sec.
II D
.
As such, the general method is rigorous and accurate in such
non-ideal cases.
3. Effect of non-uniform widths and trapezoidal
cross sections
Some of the cantilevers in this study possess strongly
non-ideal geometries, with varying widths along the can-
tilever axis and highly scalloped trapezoidal cross sections;
see SEM micrographs of the NCHR and FMR devices in
Fig.
2
. Since the pressure load on the plan view area of a
cantilever dominates its hydrodynamic load (see Sec.
II C 1
above), and the original and general methods probe the net
energy dissipated in the fluid, these geometric properties are
inconsequential to the performance of these methods. The
maximum width of the trapezoidal cross section should thus
be employed in both methods; this explains the finding of
Ref.
18
. The average of the maximum width is used to
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-5
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
AC160TS
BL-RC-150VB(L)
NCHR
NCHR
FMR
TR800(S)
FIG. 2. SEM micrographs showing perspective images for a selection of can-
tilevers used in this study. All cantilevers used are shown in Fig.
1
. Details of
each cantilever studied, including dimensions, are given in Table
I
.
estimate the net energy dissipated by corrugated devices:
NCHR and FMR in Fig.
1
. The inner width is approximately
one third the outer width, see Fig.
2
, and is not relevant to the
net energy dissipated and hence implementation of the origi-
nal and general methods.
D. Determination of the hydrodynamic function
We now summarize the experimental protocol
25
to deter-
mine the hydrodynamic function,
(Re), for a cantilever of
arbitrary shape – this function is needed to close Eq.
(8)
.
Importantly, the hydrodynamic function is a dimension-
less quantity that remains unchanged as the size and/or com-
position of the cantilever are varied. It is formally the scaled
energy dissipation in the fluid and thus depends only on the
mode shape and plan view geometry. It is also independent
of cantilever thickness for the reasons discussed in Sec.
II C
.
The hydrodynamic function can therefore be determined for
all cantilevers of the same plan view geometry by studying
a single “test” cantilever immersed in a fluid. Gas is used in
these measurements, since it allows for easy modification of
its transport properties, produces sharp resonance peaks and
thus enables rigorous extraction and measurement of the qual-
ity factor.
25
While theoretical calculations and simulations can be
used to determine the hydrodynamic function, performing
measurements on a test cantilever (i) automatically accounts
for the true geometry of the cantilever device, and (ii) intrin-
sically includes all complexities such as hydrodynamic cou-
pling between the cantilever and the supporting chip. It also
accounts for all non-ideal structures and other effects due
to the manufacturing process that may be difficult to quan-
tify and thus theoretically model in an accurate fashion, e.g.,
shape of imaging tip.
25
To proceed, we rearrange Eq.
(8)
to give an expression
for the hydrodynamic function in terms of the properties of
the cantilever and the gas
(Re)
=
k
d
ρb
2
2
R
Q
.
(9)
Our goal is to evaluate the hydrodynamic function over a
range of Reynolds numbers, Re, for a single test cantilever.
Varying the gas pressure facilitates systematic variation of
the Reynolds number, Re, because the gas density and de-
vice quality factor are both strongly dependent on gas pres-
sure; the resonant frequency is relatively insensitive to pres-
sure, whereas gas viscosity is invariant. The hydrodynamic
function is measured by placing the test cantilever in a pres-
sure chamber, systematically sweeping the gas pressure and
recording the gas density, resonant frequency, and quality fac-
tor. These results are then substituted into Eqs.
(7a)
and
(9)
.
This can be performed for a number of different gases to en-
sure consistency between measurements; this is implemented
in Sec.
III C
.
The dynamic spring constant of each test device is also
needed to complete the determination of the hydrodynamic
function,
(Re); see Eq.
(9)
. Since the general method re-
quires the hydrodynamic function for its implementation, an
alternate calibration method must be employed for this mea-
surement on the test device – the approach used is detailed in
Sec.
III B
and Appendix
A
.
We emphasize that once the hydrodynamic function,
(Re), for a particular test cantilever is determined, imple-
mentation of the general method for all cantilevers of that
same type is independent of the above specified gas pressure
and spring constant measurements. This then allows for non-
invasive and accurate spring constant measurements of AFM
cantilevers in practice using Eq.
(8)
.
III. RESULTS AND DISCUSSION
Hydrodynamic functions for all AFM cantilevers are re-
ported in this section, together with fit functions to facili-
tate their use in practice. The dimensions and properties of
these cantilevers are listed. The apparatus developed for the
gas pressure measurements is detailed in Appendix
B
. Perfor-
mance of the general method in the presence of non-idealities
is also discussed, together with finite element calculations al-
lowing for conversion between the static and dynamic spring
constants of all cantilevers. We remind the reader that only
the fundamental flexural modes are considered.
A. AFM cantilevers and dimensions
The cantilevers used in this study are from Asylum Re-
search, Nanoworld, and Olympus (Japan); see Fig.
1
. These
cantilevers are commonly used in practice. They possess sig-
nificant non-idealities, as outlined above. To highlight their
geometric features, perspective SEM images of some of these
devices are given in Fig.
2
– other devices possess similar
non-idealities to those evident in Fig.
2
. No cantilever in this
study exhibits an ideal rectangular plan view; such devices
were studied previously.
15
20
,
33
,
34
Plan view dimensions of all
cantilevers were measured from SEM images using ImageJ
and a S003 carbon grating replica (2160 lines/mm) as refer-
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-6
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
TABLE I. Measured plan view dimensions (in micrometer) of all can-
tilevers, as obtained from SEM micrographs. Definitions of the listed dimen-
sions are illustrated in Fig.
3
.
Cantilever
bb
C
dLL
C
L
TIP
AC160TS
51.0
0
...
151
94.4
151
AC240TM
30.2
0
...
227
196
227
AC240TS
29.6
0
...
229
199
229
BL-RC150VB(L)
29.8
10.1
...
93.1
83.8
93.1
BL-RC150VB(S)
29.9
10.1
...
51.7
42.7
51.7
ASYMFM
31.0
a
0
...
241
207
241
FMR
30.7
b
0
...
242
223
235
NCHR
38.3
c
0
...
136
107
128
TR400(S)
15.6
...
110
104
...
100
TR400(L)
29.5
...
164
198
...
194
TR800(S)
15.4
...
109
103
...
99.0
TR800(L)
30.4
...
170
206
...
202
a
Width tapers from clamp to end-tip in the range 32.1–30.6
μ
m.
b
Width non-uniform along length, and varies between 28.7 and 32.1
μ
m.
c
Width non-uniform along length, and varies between 36.6 and 39.5
μ
m.
Average width listed in these cases, and used in analysis.
ence; these dimensions are listed in Table
I
and Fig.
3
.The
estimated uncertainty in any given dimension measurement is
less than 1%.
1. Olympus cantilevers
The devices from Olympus, namely, AC160TS,
AC240TM, AC240TS, and the BL-RC150VB, TR400,
TR800 series, all have smooth and uniform edges. The
arrow-shaped cantilevers (AC160TS, AC240TS, AC240TM)
are composed of silicon with reflective aluminum coatings;
AC240TM has an additional platinum coating. The imaging
tips for AC160TS, AC240TS, and AC240TM are positioned
at the very end of the cantilever, i.e., the tip coincides with
the maximum extension of the plan view; see Fig.
2
.Both
biolevers (BL-RC150VB series) are composed of silicon
nitride with a reflective gold coating; their imaging tips also
coincide with the end of the cantilever and are formed by a
depression in the silicon nitride; see Fig.
2
. The V-shaped
cantilevers (TR400, TR800 series) are also made of silicon
nitride with a reflective gold coating. However, their imaging
tips are set back from the cantilever end. The TR400 and
TR800 cantilevers possess identical plan view geometries,
but have thicknesses of 400 and 800 nm, respectively, as
specified by the manufacturer. Consequently, they present
ideal candidates for demonstrating the invariance of the
measured hydrodynamic function to thickness variations
between devices.
2. Nanoworld cantilevers
The two Nanoworld devices (FMR, NCHR) exhibit sig-
nificantly different geometric features to the other quasi-
rectangular cantilevers of Asylum Research and Olympus.
These cantilevers are composed of silicon with aluminum re-
flective coatings. However, their widths are non-uniform and
vary significantly along their lengths; see Fig.
1
. Perspective
images of these devices (in Fig.
2
) reveal that they possess
L
TIP
L
C
L
b
b
C
L
b
d
b
L
TIP
(a)
(b)
FIG. 3. Schematic diagrams of all cantilevers in Fig.
1
, illustrating dimen-
sions listed in Table
I
. (a) Quasi-rectangular cantilevers; (b) V-shaped can-
tilevers. Position of imaging tip shown as a square dot.
quasi-trapezoidal cross sections with pronounced scalloping
of their sloped edges along the cantilever length. Their imag-
ing tips are set back from the cantilever end. Given their non-
uniform geometries, these devices allow the robustness of the
original and general methods to be assessed in the presence of
significant non-idealities.
3. Asylum Research cantilever
The cantilever from Asylum Research, ASYMFM, is of
identical geometry to the Olympus AC240TS and AC240TM
devices. It is also composed of silicon but has a CoCr mag-
netic coating. The imaging tip coincides with the end of the
cantilever. It possesses a slightly tapered plan view with its
width slightly narrowing from the clamp to the cantilever end;
see Table
I
.
B. Spring constants of test cantilevers
The dynamic spring constant of each test cantilevers was
measured noninvasively by monitoring its Brownian motion
under ambient conditions using a laser Doppler vibrometer
(LDV). This approach eliminates additional uncertainties in-
herent in the standard AFM thermal method
3
,
14
,
16
,
19
,
20
that
arise from calibration of the AFM photodiode displacement
sensitivity. These uncertainties originate from a number of
factors and include required corrections for laser position and
finite spot size, non-ideal contact and friction between the
cantilever tip and sample, compliance of the sample, and con-
version factors relating the cantilever angle-to-displacement
under static and dynamic loads;
16
,
20
,
35
,
36
these vary with the
cantilever used. Since the spring constant is inversely propor-
tional to the displacement squared in this method, the addi-
tional uncertainty introduced by these effects is doubled in
all AFM thermal method measurements. Elimination of these
additional uncertainties is thus highly desirable in the present
study and is achieved by using a LDV; see Appendix
A
for
details. Instruments calibrated to the SI standard (e.g., see
Ref.
37
) provide an alternate approach for measuring the
spring constants of the test cantilevers; these were not imple-
mented in this study.
We emphasize that these LDV measurements are needed
only for the test cantilevers to determine their hydrodynamic
functions. The general method is specified independently of
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-7
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
TABLE II. Measured dynamic spring constants
k
d
of the test cantilevers,
at their imaging tip positions, using a laser Doppler vibrometer (uncertainty
based on a 95% confidence interval). Resonant frequencies
f
R
and quality
factors
Q
in air (1 atm) also shown.
Cantilever
k
d
(N/m)
f
R
(kHz)
Q
AC160TS
57.3
±
1.9
370
646
AC240TM
1.65
±
0.065
65.9
162
AC240TS
2.90
±
0.13
83.0
213
BL-RC150VB(L)
0.00683
±
0.00017
12.4
12.5
ASYMFM
2.13
±
0.055
69.4
187
FMR
2.19
±
0.085
70.6
163
NCHR
33.9
±
1.5
299
480
TR400(S)
0.0971
±
0.0052
34.6
39.7
TR400(L)
0.0293
±
0.0027
11.8
21.5
TR800(L)
0.194
±
0.0062
22.9
57.1
the LDV instrument once these measurements have been per-
formed. The determined hydrodynamic functions then allow
for accurate noninvasive calibration of these cantilever types
using the general method, Eq.
(8)
, regardless of their plan
view dimensions, composition, and thickness variations; see
Secs.
I
and
II
.
Several measurement points along the cantilever were in-
terrogated to ensure robust measurements. Details of these
measurements are given in Appendix
A
. The measured dy-
namic spring constants at the imaging tip position of each
cantilever are listed in Table
II
; uncertainties due to fitting
the spring constant data at all measurement points are listed
in Table
II
and discussed in Appendix
A
. These are typically
between
±
2% and
±
5% based on a 95% confidence inter-
val. Since the LDV does not possess additional significant un-
certainties in velocity, due to its inherent calibration relative
to the speed of light, these observed fit uncertainties specify
the total uncertainties in the measured spring constants; see
Appendix
A
. The measured resonant frequency and quality
factor in air (1 atm) for each cantilever are also given, and
possess smaller uncertainty: between
±
0.0006% and
±
0.02%
for the resonant frequency, and
±
0.3% and
±
1% for the qual-
ity factor.
53
The spring constants of two test cantilevers, BL-
RC150VB(S) and TR400(L), could not be measured; also see
Appendix
A
.
C. Hydrodynamic functions
The hydrodynamic function for each test cantilever was
determined by measuring its resonant frequency and quality
factor as a function of gas pressure; see Sec.
II D
. These re-
sults were then combined with the measured plan view di-
mensions and spring constant (Secs.
III A
and
III B
), and
substituted into Eqs.
(7a)
and
(9)
to give the required hydro-
dynamic function. The experimental procedure and apparatus
developed for the gas pressure measurements are detailed in
Appendix
B
.
To ensure accurate data collection, independent measure-
ments were performed using two different gases: dry nitro-
gen (N
2
) and carbon dioxide (CO
2
). One reason for using
CO
2
is that it possesses a kinematic viscosity (shear viscos-
0.1
0.3
1
3
10
0.2
0.3
0.5
1
2
3
5
Re
Λ(
Re
)
AC240TS
FIG. 4. Measured hydrodynamic function of the AC240TS device using N
2
(open circles) and CO
2
(filled circles). Theoretical result (solid line) calcu-
lated using Eq. (20) of Ref.
23
.
ity/density) a factor of two lower than that of air, at 1 atm.
54
This in turn yields a Reynolds number, Re, a factor of two
higher than the result in air; see Eq.
(7a)
. Any subsequent
measurement on another cantilever in air, with a resonant fre-
quency twice that of the test cantilevers, will thus be within
the characterized range of Re. Importantly, the use of CO
2
eliminates the need to use gas pressures larger than 1 atm to
increase the upper limit of Re.
25
1. Quasi-rectangular cantilevers
We first assess the robustness of the characterization pro-
tocol underpinning the general method. This is illustrated for
cantilevers with quasi-rectangular plan views. The hydrody-
namic function of the AC240TS device is given in Fig.
4
.In-
dependent measurements in N
2
and CO
2
overlap precisely,
illustrating the accuracy of the measurements – we remind
the reader that the kinematic viscosities of these gases differ
by a factor of two. Also shown is the theoretical result for
the hydrodynamic function, as calculated using Eq. (20) of
Ref.
23
. Note the excellent agreement between all three data
sets – there are no adjustable parameters in this comparison.
Since the theoretical result for the hydrodynamic function is
used implicitly in the original method (for rectangular can-
tilevers), the level of agreement in Fig.
4
demonstrates the
validity of the original method for this type of device. The
original method and the general method coincide in this case
because their hydrodynamic functions overlap.
Figure
5
compares the measured hydrodynamic func-
tions for the AC240TS, AC240TM, ASYMFM, and FMR de-
vices, which have similar plan view geometries. Note that
(i) these devices have different stiffnesses, resonant frequen-
cies, and quality factors, see Table
II
, and (ii) measurements
on each device were performed in both N
2
and CO
2
.The
measured hydrodynamic functions for all these devices again
overlap with each other and with the theoretical prediction of
Ref.
23
, see Fig.
5
. The FMR device possesses a scalloped
trapezoidal cross section, in contrast to the uniform cross sec-
tions of the other devices; see Fig.
2
. The average of the
maximum width of such cantilevers should be used in the
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-8
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
0.1
0.3
1
3
10
0.2
0.3
0.5
1
2
3
5
Re
Λ(
Re
)
AC240TS, AC240TM
FMR, ASYMFM
FIG. 5. Measured hydrodynamic function of the AC240TS (open circles),
AC240TM (filled circles), ASYMFM (squares) and FMR (diamonds) de-
vices using both N
2
and CO
2
. Theoretical result (solid line) calculated using
Eq. (20) of Ref.
23
.
original and general method; see Sec.
II C 3
. As discussed in
Sec.
II C
, this non-uniformity in cross section is inconsequen-
tial to both methods – the measured hydrodynamic functions
for all devices overlap over the entire Reynolds number range
studied. This establishes that the original and general methods
coincide for all these devices and demonstrates the robustness
of these methods.
2. Non-rectangular cantilevers
Next, we demonstrate the validity of the general method
for non-rectangular geometries by comparing the measured
hydrodynamic functions for the TR400(L) and TR800(L) de-
vices. These cantilevers possess identical plan view geome-
tries, but differ in thickness by a nominal factor of two. Their
measured resonant frequencies also differ by a factor of ap-
proximately two (11.8 and 22.9 kHz) while their dynamic
stiffnesses are vastly different, in agreement with theoretical
considerations; see Table
II
. Since the hydrodynamic func-
tion is only sensitive to the plan view geometry, and this is
identical for these two devices, the independently measured
hydrodynamic functions for these two devices are expected
to coincide. This is borne out in Fig.
6
, where overlap of the
(independent) measurements on these two different devices is
observed. This fundamental property is used in Sec.
III C 3
to
derive a single fit function for the hydrodynamic function of
devices with identical plan view geometries.
The AC160TS and BL-RC150BV(L) cantilevers also
have strongly non-ideal geometries. The AC160TS device
does not satisfy the fundamental requirement for use of the
original method: a rectangular plan view of large aspect ra-
tio (length/width); it has an arrow shaped geometry. While
the BL-RC150BV(L) device has a rectangular geometry, its
aspect ratio is small (length/width
3), with a significant
imaging tip. The performance of the original and general
methods for these cantilevers, in comparison to higher aspect
ratio quasi-rectangular cantilevers, is discussed in Sec.
3
of
Appendix
A
. Overlap in the measured hydrodynamic func-
tions using N
2
and CO
2
, as observed in Figs.
4
6
, was also
0.01
0.03
0.1
0.3
1
3
0.5
1
2
3
5
10
30
Re
TR400(L), TR
8
00(L)
Λ(
Re
)
FIG. 6. Measured hydrodynamic functions of the TR400(L) (open circles)
and TR800(L) (filled circles) devices using both N
2
and CO
2
. These devices
have identical plan view geometries but their thicknesses differ by a nominal
factor of two.
found for the AC160TS and BL-RC150BV(L) devices (data
not shown).
The results in Figs.
4
6
and measurements for the
AC160TS and BL-RC150BV(L) devices, serve to demon-
strate the robustness of the experimental protocol and valid-
ity of the general method for both quasi-rectangular and non-
rectangular plan view geometries of varying composition.
3. Formulas for hydrodynamic functions
To facilitate their use in practice, analytical formulas for
all hydrodynamic functions were obtained by fitting the mea-
sured data to the following functional form:
(Re)
=
a
0
Re
a
1
+
a
2
log
10
Re
,
(10)
where
a
0
,
a
1
, and
a
2
are constant coefficients, specific to each
type of cantilever. This functional form was chosen because
the hydrodynamic function is approximately linear on a log-
log scale and is a monotonically decreasing function of Re;
see Figs.
4
6
. Fitting the data to a second-order polynomial
on a log-log scale yields the general form in Eq.
(10)
.The
resulting fit coefficients
a
0
,
a
1
, and
a
2
for all test devices are
presented in Table
III
.
Three sets of devices have identical plan view geome-
tries: (1) AC240TS, AC240TM, ASYMFM; (2) TR400(L),
TABLE III. Coeffi
cients
a
0
,
a
1
,and
a
2
in functional form Eq.
(10)
for mea-
sured hydrodynamic functions
(Re) of test cantilevers.
Cantilever
a
0
a
1
a
2
AC160TS
0.7779
0.7230
0.0251
AC240TM, AC240TS, ASYMFM
0.8170
0.7055
0.0423
BL-RC150BV(L)
1.0025
0.7649
0.0361
BL-RC150BV(S)
...
0.7613
0.0374
FMR
0.8758
0.6834
0.0357
NCHR
0.9369
0.7053
0.0438
TR400(S), TR800(S)
1.5346
0.6793
0.0265
TR400(L), TR800(L)
1.2017
0.6718
0.0383
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-9
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
0.1
0.2
0.5
1
2
5
10
0.2
0.5
1
2
5
Re
Λ(
Re
)
AC240TS, AC240TM
ASYMFM
FIG. 7. Combined data of measured hydrodynamic functions for AC240TS,
AC240TM, and ASYMFM devices (red dots), and resulting fit of this data
to Eq.
(10)
(solid line). Fit coefficients are given in Table
III
. These devices
have identical plan view geometries.
TR800(L); (3) TR400(S), TR800(S). The hydrodynamic
functions within each set must therefore coincide. Single fit
functions for each set were determined in the following man-
ner. Set (1): The fit function was evaluated by combining
independent data from all cantilevers in this set, and fitting
the result to Eq.
(10)
. The combined data and fit function
for Set
(1)
is given in Fig.
7
; similar fits were obtained for
other sets and devices (not shown). Set (2): Since the mea-
sured spring constant of TR400(L) possesses greater uncer-
tainty than TR800(L) [see Table
II
and Appendix
A
], its mea-
sured spring constant was not used but chosen such that its
hydrodynamic function overlapped precisely with that of the
TR800(L) device over the entire Reynolds number regime.
This yielded a spring constant 5% lower than that reported
in Table
II
, which is within its measured uncertainty. These
combined data were then fit to Eq.
(10)
. Set (3): The spring
constant of the TR800(S) device could not be measured; see
Appendix
A
. Rather than discarding this data, its spring con-
stant was chosen such that the hydrodynamic function for the
TR800(S) coincided with that of the TR400(S) device over
the entire Reynolds number range studied – again excellent
overlap was observed. The resulting data were fit to Eq.
(10)
.
While single devices could have been used to determine the
hydrodynamic function for each of these sets, the chosen re-
dundancy of devices facilitates accurate evaluation of the hy-
drodynamic functions.
a. Evaluation of unknown coefficients
Note that the coeffi-
cient
a
0
for the BL-RC150VB(S) test device is not spec-
ified, because its spring constant could not be measured
and no other device studied possesses an identical geome-
try. Nonetheless, the results in Table
III
can in the future be
used to generate the hydrodynamic function,
(Re), for this
device model by performing independent measurements on
other test cantilevers, of identical geometry. The dimensions
and materials need not be the same, but the plan view geom-
etry (shape) must remain unchanged. All that is required is
knowledge of the plan view dimensions of the additional de-
vice, measurement of its resonant frequency and quality fac-
tor in gas (at a single known pressure) and its stiffness. From
these measurements, the Reynolds number, Re, of the addi-
tional device follows and the value of the hydrodynamic func-
tion at this Reynolds number is immediately determined using
Eq.
(9)
. This value can then be substituted into the fit formula
for
(Re), in Eq.
(10)
and Table
III
, from which the coeffi-
cient
a
0
is uniquely specified. Using this procedure, the hy-
drodynamic function
(Re) for the BL-RC150VB(S) device
can be evaluated.
b. Refinement of hydrodynamic functions
The procedure out-
lined immediately above allows the accuracy of the pre-
sented hydrodynamic functions to be enhanced. This for
example could involve calibrating additional test cantilevers
using methodologies referenced to the SI standard.
37
These
additional measurements would lead to a refinement in the
coefficient
a
0
.
c. Uncertainty analysis
All fit functions specified in Table
III
possess uncertainties of less than
±
1% (based on a 95% con-
fidence interval), relative to measured data for the hydrody-
namic function. Note that the data are well represented by a
simple power-law as specified by
a
0
,
a
1
(see above discussion
regarding linearity on a log-log scale) – the higher-order cor-
rection due to
a
2
exerts a minor influence. Measurements of
the dynamic spring constants introduce additional uncertainty
of approximately
±
2% to
±
5% (also based on a 95% confi-
dence interval) into the coefficient
a
0
only; the coefficients
a
1
and
a
2
are unaffected by the spring constant. It is noteworthy
that individual fits to the hydrodynamic functions for devices
with identical plan views differ by only a few percent, e.g., in-
dividual fits to data for AC240TM, AC240TS, and ASYMFM.
This is in line with the measured spring constant uncer-
tainty and the use of an empirical fit function to represent
the measured data; see above. Combining the data from de-
vices with identical plan view geometries yields a more accu-
rate estimate of the hydrodynamic function, and is reported in
Table
III
.
d. Scaling behavior
The scaling chosen in Eq.
(7)
was mo-
tivated by the expectation that the dominant hydrodynamic
length scale is given by the minimum plan view dimension
of the cantilever. This is strongly supported by the values of
a
0
in Table
III
, which are all of order unity, i.e., the hydrody-
namic functions
(Re) are of order unity when the Reynolds
number is one, as required; see discussion in Sec.
II B
.
Note that the hydrodynamic functions for the V-shaped
cantilevers, TR400 and TR800, are larger in magnitude to
those for other devices. This is expected because V-shaped
cantilevers possess two skewed rectangular arms, enhancing
the net hydrodynamic load and hence energy dissipation.
It is interesting that the hydrodynamic functions of all
devices exhibit a power-law dependence on Reynolds num-
ber of approximately Re
0
.
7
, i.e.,
a
1
∼−
0
.
7; see Table
III
.
This can be understood by considering the following asymp-
totic limits. In the high Reynolds number limit (Re

1), the
hydrodynamic function is expected to scale as Re
1
/
2
due to
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-10
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
the presence of thin viscous boundary layers in the vicinity of
the cantilever surface. In the opposite (creeping flow) limit of
low Reynolds number (Re

1), a scaling behavior of Re
1
must be exhibited. Thus, the observation in measurements of
a power-law dependence intermediate to these two limiting
cases is not unexpected, since the Reynolds number is of or-
der unity for all devices studied. This property can be used to
devise an approximate method that circumvents the need for
pressure measurements. This is discussed in Sec.
III D
.
D. Simplified approximate implementation
of the general method
The general method relies on knowledge of the hydro-
dynamic function,
(Re), which can be measured for a sin-
gle “test” device using the gas pressure protocol discussed in
Sec.
II D
; these measurements are reported in Sec.
III C
.Im-
portantly, the preceding discussion demonstrates that the hy-
drodynamic function of a microscale device (whose Reynolds
number is of order unity) is well approximated by
(Re)
a
Re
0
.
7
,
(11)
where the coefficient
a
depends only on the plan view geom-
etry of the device model in question.
Since only one coefficient is unknown in Eq.
(11)
,asin-
gle measurement on a test device is required to determine its
value. For example, by measuring the spring constant, reso-
nant frequency, and quality factor of a test device in air, the
value of
a
then directly follows from Eqs.
(7a)
,
(9)
, and
(11)
:
a
=
k
d
Re
0
.
7
ρb
2
2
R
Q
test device
,
(12)
where all parameters on the right hand side of Eq.
(12)
are
determined from this single test measurement.
Equations
(11)
and
(12)
then uniquely specify the hy-
drodynamic function for arbitrary Reynolds numbers. While
this approach introduces a systematic error into the general
method (since Eq.
(11)
is approximate), this error is expected
to be small since the power-law dependence of the hydro-
dynamic function is bounded between
0.5 and
1; see
Sec.
III C
. This error is of course minimized if the Reynolds
number of the test device is comparable to (other) devices of
the same geometry to be calibrated.
If the test cantilever has identical plan view dimensions to
any subsequent (uncalibrated) device, Eqs.
(9)
and
(12)
sim-
plify, yielding
k
d
=
k
d,
test
Q
Q
test
(
f
R
f
R,
test
)
2
α
,
(13)
where
α
=
0.7 (see above), the subscript “test” refers to the
(known) test cantilever parameters, and all other parameters
are for the uncalibrated device. Equation
(13)
thus enables
the spring constant
k
d
for an uncalibrated device to be eas-
ily determined from measurement of its resonant frequency
and quality factor alone. The actual dimensions of the test
and uncalibrated devices are not required; they simply need
to be identical. Note that small deviations in the plan view
dimensions between the test and uncalibrated devices have a
minimal effect, for reasons discussed in Sec.
III F
.
TABLE IV. Conversion factors relating the dynamic
k
d
and static
k
s
spring
constants of all devices in this study. These are evaluated at the imaging tip
positions; see Table
I
. Finite element (FE) analysis is used for all calcula-
tions based on geometries as measured from SEM micrographs. The FE mesh
is systematically refined to ensure convergence of 99.9%. Poisson’s ratio of
0.25 is used in all calculations.
Cantilever
k
d
/
k
s
AC160TS
1.101
AC240TM, AC240TS, ASYMFM
1.043
BL-RC150VB(L)
1.035
BL-RC150VB(S)
1.042
FMR
1.029
NCHR
1.036
TR400(S), TR800(S)
1.054
TR400(L), TR800(L)
1.072
The above approximate implementation of the general
method facilitates the calibration of microscale devices in sit-
uations where equipment for the required pressure measure-
ments is not available.
E. Conversion factors for static and dynamic
spring constants
As discussed in Sec.
I
, either the static or dynamic spring
constant is needed for quantitative measurements, depending
on the mode of operation. Importantly, the general method
can be applied to measure both the static and dynamic spring
constants. In results for the hydrodynamic function presented
in Sec.
III C
, the dynamic spring constant of the fundamental
flexural mode was used. To calibrate the static spring con-
stant associated with a static force applied at the imaging tip
position, conversion factors between these two spring con-
stants are required. These were calculated using finite element
analysis,
38
and are given in Table
IV
. Devices with identical
plan view geometries gave identical results to within dimen-
sional uncertainty; the average of these results is reported.
Note that the conversion factors in Table
IV
are dimension-
less and thus independent of the cantilever dimensions.
F. Effect of dimensional uncertainty
on the general method
Next, we study the effect of uncertainty in the plan view
dimensions on the spring constant determined by the general
method. Naive inspection of Eq.
(8)
appears to suggest that
the resulting uncertainty in the spring constant scales with the
cube of the plan view dimensions. However, the width of the
cantilever,
b
, is embedded in the hydrodynamic function via
the Reynolds number; see Eq.
(7a)
. Since the hydrodynamic
function scales as Re
a
1
, to leading order, it follows that the
spring constant scales as
b
2(1
+
a
1
)
L
in the general method;
note that
a
1
∼−
0
.
7 as discussed above. As such, the mea-
sured spring constant exhibits a weak dependence on the can-
tilever width with sub-linear scaling; the dependence on can-
tilever length is linear. This leads to an overall uncertainty in
the measured spring constant that scales with the
3/2 power
of the plan view dimensions. Consequently, knowing the plan
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-11
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
view dimensions only approximately, imposes a weak penalty
on the overall uncertainty of the method.
Measurement of the hydrodynamic function
(Re), for
a single test cantilever, also exhibits this robustness to di-
mensional uncertainty. The uncertainty in the measured hy-
drodynamic function again scales with the
3/2 power of the
test device plan view dimensions. This is evident by compar-
ing Eqs.
(7a)
and
(9)
, which are used in the measurement of
(Re), to the power-law dependence, Re
a
1
, of the hydrody-
namic function.
The general method has been implemented so that the
dynamic spring constant is determined at the imaging tip po-
sition of all test cantilevers. If the cantilever under consid-
eration has a different tip position, relative to the cantilever
length, the spring constant will need to be adjusted; the imag-
ing tip positions of the test cantilevers are specified in Table
I
.
This adjustment can be achieved using the following proper-
ties: (i) the normal spring constant varies approximately with
the cube of the distance along the cantilever length,
39
41
and
(ii) this spring constant is insensitive to tip position variations
parallel to the clamp.
39
G. Implementation of the original and general
methods
We now summarize some practical issues relevant to im-
plementation of the original and general methods:
(1) Measurement of the thermal noise spectrum facilitates
determination of the resonant frequency and quality fac-
tor, by eliminating any spurious effects due to the fre-
quency response of the piezoactuator.
42
,
43
Thus, while
active excitation of the cantilever can be used to mea-
sure these fundamental quantities, and may be desirable
for very stiff cantilevers,
15
interrogation of the thermal
noise poses fewer issues and is simple to implement.
(2) Hydrodynamic functions for the original and general
methods were determined for cantilevers well away from
any surface, i.e., gas surrounding the cantilever was as-
sumed to be unbounded. This implicit assumption must
therefore be satisfied in all implementations of these
methods, since proximity to a surface can reduce the
measured quality factor due to squeeze film damping
– this would lead to an (artificial) underestimate of
the spring constant. Since the dominant hydrodynamic
length scale for many cantilevers is given by their width,
the thermal noise spectra of cantilevers should be mea-
sured at least several widths away from any solid sur-
face. Exploration of these effects and measurement pro-
tocols for implementation of the original and general
methods have been reported in Refs.
22
,
44
, and
45
.
(3) The resonance behavior of a cantilever is independent
of the measurement position along its axis, provided a
measurement is not taken at a zero of the slope of its de-
flection function (for which no signal may result). This
is always satisfied for the fundamental flexural mode of a
cantilever, away from the clamp. Consequently, the orig-
inal and general methods are insensitive to position of
the measurement laser on the cantilever plan view and
the laser spot size. These quantities are not required and
can be specified arbitrarily provided the signal-to-noise
is sufficient.
(4) The original and general methods yield the spring con-
stants directly and do not require knowledge of the ab-
solute deflection of the cantilever. Consequently, the
equipartition theorem can be used to calibrate the dis-
placement sensitivity of the optical lever deflection sys-
tem commonly used in the AFM, from the measured
spring constant. This approach was proposed and sys-
tematically studied in Ref.
46
for a series of rectangu-
lar cantilevers – the same method is applicable to non-
rectangular cantilevers.
22
(5) Some cantilevers possess a significant “over-hang” at the
clamped end, due to the manufacturing process, e.g., see
TR400(S) and TR800(S) devices in Fig.
1
. Since the
general method only requires the mode shape of the vi-
brating structure to be identical to the test cantilever,
such non-idealities do not pose an issue. The displace-
ment near the base of the cantilever is always small, and
hence its contribution to the net energy dissipation is
small. The general method is therefore expected to be
robust to such non-ideal variations.
(6) An “under-hang” would shorten the overall length of
the vibrating structure and could affect the net hydro-
dynamic load and energy dissipation. To understand its
effect, consider a long rectangular cantilever of high as-
pect ratio (length/width). The presence of an under-hang
would reduce the overall length of the cantilever, while
not changing its geometry. Since the original method
scales linearly with the cantilever length, this would have
a weak effect for under-hangs of relatively small length
in comparison to the cantilever length, if the original
length were used; use of the shortened length would
eliminate this effect. For non-rectangular cantilevers,
their geometry may also be affected in addition to a
change in cantilever length. Nonetheless, this effect is
still expected to be weak unless the under-hang were a
significant fraction of the cantilever length.
(7) The original and general methods are most easily imple-
mented in air, at 1 atm. The density and viscosity of air
are thus required. These are weakly dependent on atmo-
spheric variations in temperature and pressure, and are
insensitive to humidity. Even so, for more precise mea-
surements the temperature and pressure should be mea-
sured and SI data used to determine the true density and
viscosity.
IV. CONCLUSIONS
Ability to characterize the static and dynamic mechan-
ical properties of AFM cantilevers and in general, micro-
and nanomechanical devices, is critical to many applications.
Manufacturing and measurement specifications often lead to
device geometries that are complex and non-ideal. Dynamic
methods provide a versatile tool for extracting the mechani-
cal properties of such devices. In this study, previous work on
ideal rectangular cantilevers has been extended and applied
to actual devices currently employed. Specifically, we have
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions
103705-12
Sader
et al.
Rev. Sci. Instrum.
83
, 103705 (2012)
presented hydrodynamic functions for a series of non-ideal
and non-rectangular cantilevers that are commonly used in
practice. These functions allow the spring constants of these
cantilevers to be easily and routinely determined from mea-
surements of the resonant frequency and quality factor in fluid
(such as air). Performance of the original method
15
in the
presence of non-idealities was also examined. For all quasi-
rectangular cantilevers of high aspect ratio (length/width), the
original and general methods agreed closely. This demon-
strated the robustness of both methods to the presence of
non-idealities. For highly non-rectangular cantilevers, robust-
ness of the experimental protocol and general method was
also confirmed and deviations between the original and gen-
eral methods were discussed. Simple and accurate formu-
las for the hydrodynamic functions of all cantilevers were
presented to facilitate implementation in practice. A simpli-
fied approximate implementation of the general method was
proposed, facilitating implementation of the general method
when the specified gas pressure measurements are not avail-
able. Finally, conversion factors relating the dynamic and
static spring constants of all cantilevers studied were pre-
sented, together with a discussion of practicalities for imple-
menting the original and general methods.
ACKNOWLEDGMENTS
The authors would like to thank the Melbourne Cen-
tre for Nanofabrication for access to the MSA-400 Micro
System Analyzer, and Toby Ban, Jerome Eichenberger, and
Mario Pineda from Polytec Headquarters, Irvine, CA, for
use of the MSA-500 Micro System Analyzer. This research
was supported by the Australian Research Council Grants
Scheme. An iPhone application implementing the general
method for the cantilevers used in this study is available from:
http://www.ampc.ms.unimelb.edu.au/afm/
.
APPENDIX A: SPRING CONSTANT MEASUREMENTS
The dynamic spring constants of the fundamental flexural
modes of all test cantilevers were measured using a LDV;
47
49
MSA-400 and MSA-500 Micro System Analyzers, Polytec
(Waldbronn, Germany). The LDV provides an independently
calibrated measurement of velocity using a laser system that
incorporates a modified Mach-Zehnder interferometer.
50
The
spot size of the incident LDV laser is nominally 1
μ
m, al-
lowing for precise placement and subsequent measurement of
cantilever velocity at any position on its plan view – all can-
tilever plan view dimensions are an order of magnitude larger
than the optical spot size.
The spring constants were determined by monitoring the
Brownian fluctuations of each cantilever. The equipartition
theorem provides a unique connection between the mean
squared velocity of the device,

v
2

, and its dynamic spring
constant,
k
d
, at the measurement position
48
k
d
=
ω
2
R
k
B
T

v
2

,
(A1)
where
k
B
is Boltzmann’s constant and
T
is absolute temper-
ature. Note that
k
d
and

v
2

in Eq.
(A1)
are specified at the
same measurement position on the cantilever.
1. Spring constant at the imaging tip position
LDV measurement sensitivity is intrinsically dependent
upon strong reflection of the incident laser – sensitivity is de-
graded on highly curved surfaces. Curved or slanted surfaces
are thus not easily interrogated, and direct measurement at the
tip position was therefore not always possible. Spring con-
stants at the tip positions were always determined by mea-
suring the dynamic spring constant at a series of defined
positions along the cantilever length – absolute distance cal-
ibration of these points was not required. Interpolation be-
tween these data point values and extrapolation allowed for
the dynamic spring constant at the tip position to be acquired.
This was achieved by fitting the measured spring constants to
the function
k
d
=
k
tip
d
(1
αx
)
β
,
(A2)
where
k
tip
d
is the required spring constant at the imaging tip
position,
k
d
is the spring constant at the LDV measurement
position,
x
is the (uncalibrated) measurement position along
the cantilever relative to the imaging tip position, and
α
and
β
are constants. Choice of this functional form is driven by the
corresponding result for the static spring constant, for which
β
=
3.
39
,
40
,
51
The dynamic spring constant possesses a
slightly weaker dependence on position, which motivates the
use of an adjustable constant power-law,
β
.
To demonstrate the validity of this approach, a compar-
ison is made to calculations from Euler-Bernoulli theory for
a beam with a uniform cross section along its major axis; see
Fig.
8
. Note that the data point closest to the cantilever end is
a significant distance away (10% of the cantilever length). An
absolute length scale along the cantilever axis is not required
and is automatically determined by fitting the data to Eq.
(A1)
0
2
4
6
8
10
1
1.2
1.4
1.6
1.8
x
k
d
k
d
tip
=
2
=
3
FIG. 8. Theoretical simulation of fitting procedure, Eq.
(A2)
, to extract the
dynamic spring constant at the imaging tip position. Solid circles are data
calculated using Euler-Bernoulli beam theory. Lines are fits to this data using
Eq.
(A2)
. Each unit on the horizontal axis is 0.02
L
,where
L
is cantilever
length; point furthest back from end-tip is at a distance of 0.2
L
. Imaging tip
coincides with the end-tip here. (Upper fit curve [blue])
β
=
2; (Lower fit
curve [red])
β
=
3.
Downloaded 20 Dec 2012 to 131.215.71.79. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/about/rights_and_permissions