Professional Documents
Culture Documents
Abstract
As with any other microscopic technique, in atomic force microscopy (AFM), problems can arise. Some of
these happen due to improper use of the microscope by the operator, and some are due to particular
characteristics of the sample. Some occur depending on the type of instrument, or from probe damage.
Some of them are artifacts inherent in the technique. Knowledge of these issues is important for correct data
acquisition and interpretation, and in many cases, training in AFM is inadequate. In this chapter we show
examples of common artifacts in AFM and describe, where possible, how to overcome them. Other practical
issues important for best practice in AFM operation, such as noise reduction and data processing, are also
discussed.
1 Introduction
Nuno C. Santos and Filomena A. Carvalho (eds.), Atomic Force Microscopy: Methods and Protocols, Methods in Molecular Biology,
vol. 1886, https://doi.org/10.1007/978-1-4939-8894-5_1, © Springer Science+Business Media, LLC, part of Springer Nature 2019
3
4 Peter Eaton and Krystallenia Batziou
2 Types of Artifacts
2.1 Probe Artifacts Probe artifacts come about when the shape of the probe is nonideal.
In AFM, as in many other methods, the data output always depends
on the nature of the probe. The image obtained in any microscope
is a convolution of the probe with the sample. In the case of AFM,
the ideal situation is to have the probe with a tip diameter smaller
than the highest resolution required. The probe should also have a
high aspect ratio, such that the probe can reach to the bottom of
any depression or pits in the sample surface, with the minimum of
contact between the sides of the probe and sides of sample features.
However, this is not always the case. Probe-sample convolution and
the effect of different probe sizes are illustrated in Fig. 1.
Sample feature dilation by the probe is an effect that always
occurs in AFM imaging, and can be minimized, but not totally
Artifacts and Practical Issues in AFM 5
Fig. 1 Illustration of effects of probe dilation on the image formed in AFM. In each
case, the left images show the probe and sample feature, the right the image
formed. The first two images show the effect on a raised feature with a small
probe (a), and then with a large diameter tip (b). The lower two images show the
effects of imaging a depressed feature (a pit) with a small (c) and large diameter
tip (d). Raised features tend to be dilated by the probe, while maintaining
accurate height reproduction. On the other hand, depressed features will
become smaller, both in width and height
Fig. 2 Images showing the differences between imaging with a sharp (a) versus
a blunt (b) probe. Both images show the same sample (biaxially oriented
polypropylene film, BOPP), imaged with a sharp probe (a) and with a probe of
the same type that had been blunted by prolonged use (b)
Fig. 3 Examples of tip artifacts leading to repeating shapes in the height images. The samples were protein
nanoparticles (a and c) and a glass microscope slide coated with in proteins (b). All scale bars represent
500 nm
Fig. 4 Illustration of a double tip effect. Height image of DNA molecules, where
each molecule has an apparent “twin” caused by contamination of the probe.
The x–y scale of the image is 1.5 1.5 μm, and the image height is 2.0 nm
Fig. 5 Illustration of the effect of imaging holes in a sample surface with a blunt tip. (a) Height image of a
calibration artifact which features an array of square holes imaged with a sharp probe. When the same area is
scanned with a blunt probe (b), the holes in the surface appear artificially smaller
2.2 Scanner Artifacts In an AFM, the movement of the probe relative to the sample is
carried by a scanner, which translates a voltage produced by the
control electronics to a nano- or micro-metric movement. In nearly
all cases, these are based on piezoelectric crystals, which expand or
contract depending on the polarity of the voltage applied to them.
Since the coefficient of expansion of commonly used piezoelectric
crystals is commonly on the order of 0.01 nm per volt [15], it is
relatively simple with these devices to produce movements with the
nanometric precision required for an AFM.
An ideal AFM scanner would have a perfectly linear relation
between the voltage applied and the distance traveled in all three
axes, would not have any cross talk between movements in different
axes, would respond instantaneously to changes in voltage, and
would also cease movement immediately that the controlling volt-
age ceases changing.
In general, scanners do not produce such perfect movements,
but suffer from a number of imperfections, which give rise to
specific image artifacts which will be discussed in this section. Not
10 Peter Eaton and Krystallenia Batziou
all AFM scanners are the same, as many different designs are in use,
and some reduce or eliminate some of these artifacts.
Finally, apart from the image artifacts produced by piezoelectric
scanners, it is useful for the AFM operator to be aware that piezo-
electrics are quite sensitive materials. They are fragile, and easily
broken. They can also be depolarized by heat or exposure to water,
which will render them useless. Therefore AFM scanners should be
handled with care, and maintained in a dry, room temperature
environment at all times.
2.2.1 Nonlinearity The first major scanner-related artifact is nonlinearity in the x–y
plane. In general, the response of piezoelectrics to voltage is non-
linear. While there are a number of ways to ameliorate, or reduce
this effect, which will be discussed below, if uncorrected this can
introduce severe distortions in the image. Figure 6 illustrates the
effects of this nonlinearity on an image of an AFM calibration grid,
consisting of square, regularly spaced holes in a flat silicon surface.
The two images in Fig. 6 show the same area of the same sample,
scanned using the same instrument and the same AFM probe. The
height image in Fig. 6a was measured using the raw output from
the scanners to linear voltage ramps. It can be seen that in this case,
the response in the X-axis was extremely nonlinear, while linearity
in the y axis was somewhat better but not perfect. The image shown
in Fig. 6b was linearized using position sensors.
There are a number of ways to remedy or improve the linearity
in AFM scanners:
Fig. 6 Example of distortion in the x–y plane of an AFM image due to nonlinearity in the scanner response. The
two images show the same area of the sample. (a) Height image recorded with hardware linearization
disabled, and (b) height image recorded with hardware linearization (displacement sensors) enabled. Both
images are approximately 74 74 μm in x and y and 235 nm in z (height)
Artifacts and Practical Issues in AFM 11
Fig. 7 The causes and effects of creep. (a) Schematic illustration what happens when a piezoelectric
transducer has a voltage change applied to it—the movement of the piezoelectric tends to continue even
after the voltage stops changing. At the point marked “S,” the creep has stopped, and the system is stabilized,
and suitable to begin measuring an undistorted image. In (b), the result of this effect on atomic force
microscopy images is illustrated. The distortion at the start of the height image (the distortion is seen at the
top, in this figure) is due to creep. The image (b) is approximately 9.5 9.5 μm in x and y and 380 nm in
z (height).This effect should not occur if closed-loop operation is used
Artifacts and Practical Issues in AFM 13
2.2.3 Hysteresis (Edge Another effect caused by imperfections in piezo responses is hyster-
Overshoot) esis. Hysteresis in any material is the tendency to fail to return
perfectly to its original shape after extension or compression. In
the piezo actuator in an AFM, this means that in the x–y axis,
forward and back images may be slightly misaligned, although
this is not an important effect for most measurements. Hysteresis
in the z piezo is more important, since it can lead to inaccuracies in
some height measurements, and also in force-distance curve mea-
surements. Examples of the effects of this phenomenon are shown
in Fig. 8.
If the scanner is prone to this effect, it is actually hard to avoid.
The extent of its effects on the image can be lessened somewhat by
scanning more slowly, but will not disappear. The best way to avoid
it, if possible, is to use z sensor height data instead of raw height
data. This is only available if the instrument in use has a displace-
ment sensor on the z piezo. Use of the z sensor data will also
improve the accuracy of force-distance curves. In both cases, it’s
worth bearing in mind that if the z sensor has a considerable noise
level, using this data instead of the standard z height data may result
in better accuracy but lower precision.
Fig. 8 Effects of hysteresis in piezo actuators in AFM. (a) The image is from a height measurement over
square-profile posts that shows where errors can occur when passing over steep features. (b) An example of
an uncorrected deflection-distance curve (from which force-distance curves are calculated. Close to the
turnaround point at the right (arrowed), the forward and back traces do not match due to piezo hysteresis
14 Peter Eaton and Krystallenia Batziou
2.2.4 Scanner Calibration All AFM systems require calibration of the scanner movement in
order to deliver quantitative results. Commercial systems should be
delivered with an accurate calibration, and a certification indicating
the error in the calibration measurement. Nevertheless, piezoelec-
tric materials age over time, which means that, the exact number of
nanometers they move per applied volt changes. Typically, this
change is faster at the beginning of a scanner’s lifetime, and stabi-
lizes over time. It is highly recommended that the scanner is cali-
brated against a trusted calibration artifact 6 months after
installation, and every year thereafter. Calibration procedures have
been extensively described elsewhere [16].
2.3 Noise Vibrations in the AFM system inevitably give rise to noise in the
data produced. This is common in all high-resolution microscopes.
2.3.1 Acoustic
In AFM, compared to electronic microscopes, the system can be
and Mechanical Noise
more troublesome for two reasons. Firstly, the AFM is often a small
system, so it’s more prone to mechanical vibration. Secondly, in
AFM, the probe physically touches the sample, forming a mechani-
cal loop, which transmits all vibrations in the system into the probe-
sample interface. Typically, AFM installations include some form of
vibration isolation, which reduces these effects to an acceptable
level. Acoustic isolation might be included as well. Nevertheless,
these measures only reduce the level of noise. Figure 9 shows image
recorded in an AFM system located in a fully working vibration, and
acoustic enclosure. In Fig. 9, parts A and B are images recorded in
quiet conditions. Figure 9 (parts c and d) are similar images
recorded with people in the room talking and performing various
noisy activities. These produce significant transient noise in the
image.
In Fig. 9, the noise is clearly visible in the lower right hand error
image (Fig. 9d), and somewhat harder to perceive in the
corresponding height image, at the top right (Fig. 9c). But height
images do contain marked vibrations. These are illustrated in the
height profiles in Fig. 10, which were extracted from the images a
and c from Fig. 9.
In order to avoid the problems shown in Fig. 9, it is important,
firstly, to ensure the AFM instrument is shielded toward both
acoustic and vibrational influences. There are many commercial
vibration solutions available, both active and passive. Passive solu-
tions generally work by connecting the AFM instrument to a large
mass on the end of a soft spring. In this way, high frequency
vibrations are absorbed by the mass-spring system. Active systems,
on the other hand, use accelerometers to measure vibrations com-
ing into the system, and actively compensate for these using actua-
tors. Either kind of system can work extremely well to reduce (but
not to eliminate) vibrations entering the AFM system. Thus it’s
important to install the AFM in a low-noise environment, and avoid
the kind of transient impulses illustrated in Fig. 9. Acoustic
Artifacts and Practical Issues in AFM 15
Fig. 9 Examples of transient noise effects in AFM images. All images show the same area of an HOPG sample.
(a, b) show height and amplitude images, respectively, of the samples taken under quiet conditions, i.e., with
the AFM enclosed in a vibration isolation cabinet, and no acoustic or vibration sources in the room. (c, d) show
the equivalent images of the same area with talking and noisy activities occurring in the room. The lines
marked P1 and P2 refer to profiles through these images shown in Fig. 10
Fig. 10 Height profiles indicating interference in height profiles from acoustic and vibrational noise. The graphs
show two line profiles (P1, P2) from Fig. 9a, c. In both graphs, the blue solid line shows the height profile
measured during noisy conditions (from Fig. 9c), while the red dashed line shows the same line recorded
under quiet conditions (from Fig. 9a)
16 Peter Eaton and Krystallenia Batziou
2.3.2 Electronic Noise It is possible for the AFM to pick up electronic noise from the
surrounding environment (e.g., radio-frequency interference), or
from an internal fault in the AFM to give rise to electronic noise in
the image. External noise sources can be diagnosed by switching off
possibly interfering devices. Fluorescent lights are a common cul-
prit in this regard, either malfunctioning, or not. Internal problems
in the AFM may be harder to deal with, but it’s important that the
user check for loose wires if such a problem is suspected. Very
commonly such problems are simple to diagnose, since they are
either completely constant, or switch on and off suddenly, unlike
other sources of noise, and have fixed frequencies. Commonly, this
will be the frequency of the mains AC source (i.e., either 50 or
60 Hz), which can be measured in the AFM image (Fig. 11).
2.3.3 Bit Noise “Bit noise” is an effect due to the limited bit resolution in the
analogue to digital converters (ADCs) which are a vital part of the
AFM. Usually this is only seen in z height images, when looking at
extremely flat samples, or measuring very small images. It comes
about because the AFM electronics usually record the height data
as 16-bit numbers. With 16 bits, the number of discrete height
values that can be registered is 216, i.e., 65,536. While this is a very
large number, and is sufficient for most cases, since the AFM is
extremely sensitive in z, it can be a limitation. If we take the case of
a very large scanner, with 10 μm height range, we can determine
that the smallest value that can be distinguished is 10,000/
65,536, i.e., 0.15 nm. This is a pretty small measurement, espe-
cially in a range of 10 μm; however, most AFMs will have a noise
floor several times smaller than this value. Thus for high-resolu-
tion imaging, bit noise can be a significant factor in image quality.
Fortunately, it is a limitation which is generally easy to overcome.
One solution is to use a smaller scanner. AFM instruments are
often purchased with multiple scanners for imaging samples with
difference size features. This does work, but it’s an expensive
solution to a simple problem with a simple solution. With most
systems, it’s possible to artificially limit the range of the scanner
used when imaging under high-resolution/low-noise conditions.
This has the effect of greatly reducing the bit noise. For example,
in the case above, the same scanner might be used with the
effective range of only 1 μm. Thus, the sampling would occur at
Artifacts and Practical Issues in AFM 17
Fig. 11 Example of electronic noise in an AFM image. In this case, the noise that
can be seen at the bottom of the image was highly reproducible. It always
occurred in the lower part of the image, irrespective of the sample scanned, or
the conditions used. The cause is likely a loose connection in the scanner.
Thanks to Gleb Turlakov for permission to use the image
2.4 Image Image processing in AFM is very important. In fact, the directly
Processing Issues digital nature of AFM data makes image processing inherently easy.
Standard topographical AFM images are in fact a matrix of height
values. This makes the extraction of quantitative date from the
image relatively easy, although some care always needs to be taken.
Considering AFM height images, the data is, in general, stored
as 16 bit values. Therefore, there are 65,536 possible height values.
In other words, if we attempted to represent all this data on
computer monitor as a grayscale image, we would need to show
more than 65,000 shades of gray. There are currently no computer
monitors capable of this bit-depth. Even if they were, it would not
be possible for us to perceive this level of detail, since our eyes are
capable of distinguishing between 50 and 100 shades of gray
[17]. The same goes for printed images. An example of this is
shown in the Fig. 12. In Fig. 12a is shown a height image, with
the height illustrated using a grayscale, the most common form of
display used for AFM images. On Fig. 12b is a 3D-rendered and
light-shaded version of the exact same image. It is possible to see
18 Peter Eaton and Krystallenia Batziou
a 388 nm
b
350
300
250
200 0.39 µm
0.00 µm
150
100 y:
2.4 µm
µm 2.4
50 x:
500 nm
0
Fig. 12 Illustration of the difficulty of visualizing details in color-coded AFM height images. The two images
derive from the same data set. The conventional way to show AFM height data is in (a). This “color mapping”
representation is useful because the colors can be directly related to height values. However, many height
details are hidden since only a few shades of gray can be represented. The light-shaded rendering in (b)
shows many of these hidden details
many details in this shaded image which are “hidden” in the gray-
scale height image.
It is important that the user be aware that it is not always
possible to see all the details available in an AFM image at first
sight. Thus it is imperative that any modifications made to the data
via processing steps, such as histogram adjustment, shading, level-
ing, and filtering, be made non-destructively. It is critically impor-
tant that the original data file as saved by the instrument software is
not altered in anyway, as this can destroy data forever.
In this section of the chapter, some issues involving data pro-
cessing are discussed, with particular attention paid to processing
steps which could result in artifacts or image degradation.
2.4.1 Leveling Leveling is the most important step in AFM image processing,
which is applied to nearly all images [18]. Depending on the
image, it can be either extremely simple or quite difficult to get
right. One of the most common artifacts associated with leveling is
illustrated in Fig. 13. In this case, the effect is introduced by using a
polynomial leveling routine, which attempts to level the back-
ground, without taking account of features that do not form part
of the background. In Fig. 13a, an image of nanoparticles
distributed on a flat surface has been leveled with a horizontal
second-order polynomial fit. This results in the dark streaks or
shadows around the particles. If an adequate leveling routine is
used [18, 19], the image on the right can result.
The image in Fig. 13a does show correctly the positions of the
nanoparticles, but misrepresents the shape of the background.
However, there is a more serious implication of using inadequate
Artifacts and Practical Issues in AFM 19
Fig. 13 Example of artifacts introduced into an image by a poor choice of leveling parameters. (a) A height
image of nanoparticles which has been leveled by a horizontal second-order polynomial fitting routine. As a
consequence, the particles are accompanied by dark streaks, since the routine fits the particles as if they were
a part of the background. (b) A correctly leveled version of the same image
2.4.2 Filtering Most AFM processing software allows the user to filter the images,
for a variety of reasons. One very common reason why users could
use filters on their images is to remove noise present in the data.
While this may be useful for display purposes, it’s important that
the user be aware that all filtering alters the data. Thus, any analysis
to be performed on the AFM data should be made using un-filtered
data only. As explained above, in many cases, subtleties in AFM
images are hard to discern by eye. Thus, it can be that an image
looks similar after filtering (albeit with lower noise), but on close
examination, the shape of features in the image may be considerably
altered by the filtering process. An example of the kind of alteration
that can occur upon filtering is illustrated in Fig. 15.
In Fig. 15, it’s possible to observe that a low-pass filter enables
smoothing, and reduction of noise, but can also change features
20 Peter Eaton and Krystallenia Batziou
100 100
vertical
nm
nm
101 nm 124 nm
50 50
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
150 150
100
100
horizontal
nm
118 nm
50
nm
127 nm
50
0
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 14 Examples of vertical and horizontal cross-section profiles over the same nanoparticle from two
differently processed images coming from the same AFM height image (Fig. 13). On the left profiles from the
image shown in Fig. 13a; on the right are shown images from Fig. 13b. It can be seen that the way leveling is
performed can affect the measured height of features
2.5 Other Issues There are many sample preparation issues that can affect the results
obtained in AFM imaging. Proper sample preparation is vital for
2.5.1 Sample Adhesion
AFM measurements, and typically problems in sample preparation
cannot be overcome when imaging. Thus this first step in an AFM
experiment can be considered one of the most important. One issue
that can affect AFM imaging is cleanliness of the sample. AFM is a
surface characterization technique. Thus, a very thin (nanometric)
contamination layer could completely eliminate the sample features
which the user desires to scan. Dirty or contaminated samples can
also lead to probe contamination, which can also prevent imaging
as described above. Another issue is that the features to be imaged
should not move about under the forces from the probe. Most
commonly if this happens, horizontal streaks will appear in the
images, and the sample features will not be imaged at all. On
some occasions, however, the sample features will move only
slightly, leaving tracks across the image. This is illustrated with the
height image of nanoparticles in Fig. 16a.
Artifacts and Practical Issues in AFM 21
Fig. 15 Example of the effects of filtering. The original height image (a) and a cross-section profile (c) through
several nanoparticles. A low-pass (smoothing) filter was applied, giving the result on (b). Shape and height
changes can be seen on (b and d, respectively) indicating that this data is not suitable for further analysis
Fig. 16 Height images illustrating the effects of marginal sample-substrate adhesion. The images show a
sample of nanoparticles on a surface to which they adhere poorly. (a) The first image obtained which shows
“drag marks,” caused by the samples being nudged down the image (the slow scan direction was down). This
results in image artifacts and underestimation of the number of particles present. (b) The same area imaged a
second time with higher PID gains, and lower force applied, which does not show dragging of the
nanoparticles
22 Peter Eaton and Krystallenia Batziou
2.5.2 Laser Interference This kind of artifact is caused by reflections of the laser originating
from the sample surface. As illustrated in Fig. 17a, the laser light
that follows this path will tend to interfere with the laser that is
properly reflected from the back of the cantilever. In principle, if the
optical setup of the AFM was perfect, there would of course be very
little signal reflecting from the sample reaching the photodetector.
In particular, this interference is reduced because the cantilever is
usually at an angle compared to the sample surface (in most systems
this angle is around 10–12 ). This difference in angle can be
enough to direct laser light from the sample away from the photo-
detector. Moreover, most of the laser light should be tightly
focussed on the cantilever, with very little light “spilled” over the
edges of the lever. However in the real world, there are several
reasons why this can in fact occur. Firstly, many lasers in AFM
Fig. 17 (a) The principle behind laser interference in AFM. The single incoming laser beam can reflect from the
sample as well as the cantilever, creating interference at the photodetector. (b) The effect of such interference
on imaging. Vertical “waves” appear in the height image while scanning micron-sized areas. Typically the
wavelength of the artifacts is the same as that used in the laser. The sample is a silicon surface with deposited
nanoparticles. The x–y scale is 15 μm and the z scale is approximately 10 nm
Artifacts and Practical Issues in AFM 23
systems are not as well focussed as y should be. Secondly, the user
might accidentally set the focus position of the laser on the edge of
the probe, which will allow some light to spill on to the sample. In
addition, many AFM cantilevers are actually semitransparent to the
laser light, allowing a small amount of it to pass straight through.
This particular problem is reduced if the cantilevers are coated.
Many levers are available coated with a gold film on the backside
for increased reflectivity.
The effect of this issue is illustrated in Fig. 17b. Typically
vertical waves appear in the image, with a wavelength dependent
on the laser used. The height scale of the image in Fig. 17b is about
10 nm, so it can be seen that this effect is quite small. Thus it does
not significantly affect large-scale images. However, for images with
low z scales, it can be troublesome. The simplest solution is to
realign the laser, ensuring as little overspill as possible. It is worth
noting that this particular effect also affects force-distance curves.
These are measured during a vertical movement of the probe
relative to the surface. Considering the diagram in Fig. 17a, we
can see that vertical movement is also likely to result in changes in
interferences. In fact, this effect is seen more often in force curves
than in images. An illustration of this is shown in Fig. 18.
Just as with the effect on imaging, the best way to avoid this
issue is to carefully realign the laser, or avoid using a reflective
substrate. This artifact can make further analysis of the force-
curve more complicated, but many force-curve analysis software
2.5.3 Feedback Incorrectly set feedback gain parameters can cause considerable
Parameters problems in AFM images. These parameters are sometimes
known as PID, or proportional, integral, derivative values. In real-
ity, tuning of the PID parameters is a basic requirement of AFM
operation, so an experienced AFM user should be accustomed to
adjusting and refining these numbers. However, many factors can
alter the specific values required, including sample topography,
scanning speed, scan size, etc. Therefore it is rather important
that the user is used to recognizing when these parameters need
adjusting. Figure 19 shows different images of the same region of a
Fig. 19 Examples of the effect of using incorrect feedback gains. Panels (a) and (b) are height and error signal
images, respectively, of a sample obtained using adequate feedback gains. Panels (c) and (d) show
corresponding images of the same area, obtained with the same settings, except for the feedback gains
(PID values), which were set too low. It is noticeable that the topography is not followed properly by the probe
in this case. All images are 4 4 μm in x and y
Artifacts and Practical Issues in AFM 25
2.5.5 Sample Drift Sample drift refers to uncontrolled movement of the sample while it
is being scanned. As the term “drift” implies, this tends to be a slow
movement. For large images, this effect often causes very few
problems. However, slow movements tend to have serious implica-
tions when scanning at high magnifications—i.e., small scale
images. Sample drift is an important problem in all kinds of high-
resolution microscopy. For example, in electron microscopy, a lim-
iting factor when obtaining high-resolution imaging tends to be
26 Peter Eaton and Krystallenia Batziou
Fig. 20 Example of uncontrolled z oscillations. The effect is seen around the base of the bacteria in these
images, probably due to incomplete sample drying. The effect is seen most easily in amplitude or phase
images (amplitude image, b), but also occurs in height images (a). The images are 10 10 μm in x and y, and
the height image is approximately 1.3 μm in z
Fig. 21 Illustration of the effects of sample drift. (a) This height image was
obtained with a sample (a monolayer on a gold film) that was not at thermal
equilibrium. The features are stretched, and imaging unstable. (b) The height
image on the right was obtained under exactly the same conditions, after the
sample had cooled to reach thermal equilibrium
3 Conclusions
Acknowledgments
References
1. Michael Hollas J (2004) Modern spectroscopy. 13. Chen Y, Cai JY, Liu ML et al (2004) Research
In: Modern spectroscopy. Wiley, Hoboken, NJ, on double-probe, double- and triple-tip effects
p 41–71 during atomic force microscopy scanning.
2. Goldstein JI, Newbury DE, Michael JR et al Scanning 26:155–161
(2017) Image defects. In: Scanning electron 14. Gruber A, Gspann J, Hoffmann H (1999)
microscopy and X-ray microanalysis. Springer, Nanostructures produced by cluster beam
New York, pp 133–146 lithography. Appl Phys A Mater Sci Process
3. Spence JCH, Spence RPPJCH, DeWitt BS 68:197–201
(2003) High-resolution Electron Microscopy, 15. Eaton P, West P (2010) Chapter 2: instrumen-
3rd edn. Oxford University Press, New York, tal aspects of AFM. In: Atomic force micros-
pp 15–47 copy. Oxford University Press, Oxford, pp
4. Nie H-Y, Walzak MJ, Mcintyre NSCN (2002) 9–48
Use of biaxially oriented polypropylene film for 16. Eaton P, West P (2010) Appendix B: scanner
evaluating and cleaning contaminated atomic calibration and certification procedures. In:
force microscopy probe tips: an application to Atomic force microscopy. Oxford University
blind reconstruction. Rev Sci Instrum Press, Oxford, pp 192–197
73:3831–3836 17. Russ JC (2006) Human vision. In: The image
5. Eaton P. SPM. Standards and reference sam- processing handbook, 5th edn. CRC Press,
ples. In: AFMHelp.com. http://afmhelp.com/ Boca Raton, p 83–134
index.php?option¼com_content& 18. Klapetek P (2012) 4.4.2 Data levelling and
view¼article&id¼48&Itemid¼54. Accessed background extraction. In: Quantitative data
3 May 2018 processing in scanning probe microscopy:
6. Ramirez-Aguilar KA, Rowlen KL (1998) Tip SPM applications. William Andrew, Norwich,
characterization from AFM images of nano- NY, p 64–67
metric spherical particles. Langmuir 19. Eaton P, West P (2010) Atomic force micros-
14:2562–2566 copy. Oxford University Press, Oxford
7. Bykov V, Gologanov A, Shevyakov V (1998) 20. Eaton P, West P (2010) Processing AFM
Test structure for SPM tip shape deconvolu- images. In: Atomic force microscopy, 1st edn.
tion. Appl Phys A Mater Sci Process Oxford University Press, Oxford, p 104–109
66:499–502 21. Eaton P, West P (2010) Substrates for AFM.
8. Ho H, West P (1996) Optimizing AC-mode In: Atomic force microscopy. Oxford Univer-
atomic force microscope imaging. Scanning sity Press, Oxford, pp 87–88
18:339–343 22. Chada N, Sigdel KP, Gari RRS et al (2015)
9. Nie HY, McIntyre NS (2001) A simple and Glass is a viable substrate for precision force
effective method of evaluating atomic force microscopy of membrane proteins. Sci Rep
microscopy tip performance. Langmuir 5:12550
17:432–436 23. Wagner P (1998) Immobilization strategies for
10. Sirghi L, Kylián O, Gilliland D et al (2006) biological scanning probe microscopy. FEBS
Cleaning and hydrophilization of atomic force Lett 430:112–115
microscopy silicon probes. J Phys Chem B 24. Haugstad G (2012) Chapter 5: Probing mate-
110:25975–25981. https://doi.org/10. rial properties I: phase imaging. In: Atomic
1021/JP063327G force microscopy: understanding basic modes
11. Lo YS, Huefner ND, Chan WS et al (1999) and advanced applications. Wiley, Hoboken,
Organic and inorganic contamination on com- NJ, p 187–257
mercial Afm cantilevers. Langmuir 25. Sang X, LeBeau JM (2014) Revolving scanning
15:6522–6526 transmission electron microscopy: correcting
12. Nie H-Y, McIntyre NS (2007) Unstable ampli- sample drift distortion without prior knowl-
tude and noisy image induced by tip contami- edge. Ultramicroscopy 138:28–35
nation in dynamic force mode atomic force
microscopy. Rev Sci Instrum 78:23701