Scanning-probe microscopy has
been routinely employed as a surface
characterization technique for more
than 20 years. Tip deconvolution, the
longest-standing problem associated with particle image analysis in atomic
force microscopy (AFM), can be solved
by scanning a pre-characterized nanosphere
prior to imaging unknown particles.
INTRODUCTION
Knowledge of particle size, size
distribution, and other particle morphology
parameters on the nanometer
scale is becoming more important
with accelerating developments in the
nanotechnology branches of particle technology, toxicology, pharmaceutical,
semiconductor, composite, and coating
industries. For example, monitoring the
presence of particles, powder size, and
distribution data is an important aspect of
process control. Changes in these aspects
and/or the direction of such changes can
be quite indicative of events during the
manufacturing process, events which can
significantly impact the properties and
quality of the final product. Tracking such
changes can also indicate the point in the
manufacturing process at which there
is a problem. Particles and particle-size
distribution (PSD) significantly contribute
to the mechanical strength as well as
thermal and electrical properties of the
final material. Smart coatings with a great
variety of properties can be developed to
meet specific requirements of particular
applications. Mechanisms of agglomeration,
adhesion, and particle-particle
interaction in particulate materials need
to be understood to solve current challenges
in the pharmaceutical industry.
|
"Particles and particle-size
distribution significantly contribute
to the mechanical strength as well as
thermal and electrical properties of the
final material. " |
|
The most common set of morphology
parameters is listed in Table I. Parameters
involving measurements of the
third dimension such as height, surface
roughness, and volume are possible only
with atomic-force microscopy (AFM).
Table I. The Most Common Set of Morphology Parameters Available for Measurement by
Individual Particle Characterization Techniques
|
Morphology Parameters |
Optics
|
TEM/SEM
|
AFM
|
Size, Radii, Length, Width |
Yes
|
Yes
|
Yes
|
Aspect Ratio |
Yes
|
Yes
|
Yes
|
Height |
No
|
No
|
Yes
|
Perimeter |
Yes
|
Yes
|
Yes
|
Projected Area |
Yes
|
Yes
|
Yes
|
Surface Roughness |
No
|
No
|
Yes
|
Volume |
No
|
No
|
Yes
|
|
The ability to visualize and directly
measure dimensions of a few nanometers
is a necessity in nanotechnology. There
are few particle analysis techniques
capable of delivering morphological
information below 100 nm. Ensemble
methods, such as dynamic light scattering
and gravitational sedimentation
techniques, are starting to push the limits
of resolution at about the 40 nm to the
100 nm range, respectively.1
|
"In fact, AFM is a
non-intrusive technique with resolution
greater than or comparable to that of an
electron microscope." |
|
In certain cases, it is important to
ensure that PSD lies within certain
limits, making the precision of these
measurements more important than their
accuracy. Microscopy-based techniques,
where counted particles can be visually
examined, are usually used for absolute
measurement applications. Because the
resolution of the measuring technique
should be greater than the size of the
particles under investigation—in other
words, less than 1–100 nm—optical
microscopy is excluded (see Table II).
However, both AFM and scanning-electron
microscopy/transmission-electron
microscopy (SEM/TEM) have the
required resolution. In fact, AFM is a
non-intrusive technique with resolution
greater than or comparable to that of an
electron microscope. In addition, it is
easier to use and involves less sample
preparation (see Figure 1). Particles can
be measured at ambient condition, while
electron microscopy requires vacuum
chambers and conductivity of the specimen
(see Table II).
Table II. A Comparison of Resolution and the Most Common Artifacts Associated with
Individual Particle Characterization Techniques
|
|
Optics
|
TEM/SEM
|
AFM
|
Resolution |
~150 nm for green light
>1 µm for white light
|
a few A (TEM) a few nm (FEGSEM*,
secondary electrons)
~0.1 µm (SEM,
backscattered electrons)
|
sub-A (vertical)
|
Probe Size, Diameter |
|
1-2 nm
|
1-30 nm
|
Artifact Associated with a Probe |
Chromatic aberration, spherical aberration, astigmatism,
distortion
|
Spherical aberration, astigmatism, distortion
|
Tip convolution
|
Artifact Associated with a Sample |
|
Charging if sample is semi- or non-conductive
|
|
Environmental Requirements |
Ambient air
|
Vacuum, environment controlled
|
Ambient air liquid, UHV*
|
Artifact Associated with an Interaction |
Non-Intrusive
|
Distractive to sample surface
|
Non-instrusive
|
|
PARTICLE CHARACTERIZATION
|
The capabilities of atomic-force microscopy (AFM) in both qualitative and quantitative
particle characterization are described elsewhere.7 Table I shows the most common
morphology parameters required in particle characterization. The ability to measure the
width of the feature is critical because it contributes to all morphology parameters. If
the diameter of the probe is larger or comparable to the diameter of the particle, then
the broadening effect in measuring lateral dimension can be clearly seen (Figures Aa
and Ab). In turn, area, perimeter, surface roughness, and volume measurements all are
affected differently, while actual particle size is the same in both cases.
The particle sample is the same in both Figure Aa and Ab. The examined sample
consists of 102 nm polystyrene spheres (Duke Scientific , Fremont, California) deposited
substrate.
Figure B illustrates the way in which three different types of tips used for the simulation
can alter topographical AFM data of the spheres.8 Shown are three simulated images and
the corresponding height profiles for four latex spheres with radii 56, 68, 91, and 102
nm, respectively. The image size in parts a, c, and e is always 3,000 nm × 500 nm. Three
different types of tips have been used for the simulation: (a) ideal tip (assumed as delta
function), (c) conical tip with 6 nm apex radius and 50° total opening angle, and (e) tip
with 50 nm apex radius and 50° total opening angle. The corresponding height profiles
across the centers of the spheres are displayed in parts b, d, and f, respectively. The influence of the tip geometry on the measured topography of the latex spheres (especially
its apparent width) is clearly visible. This is a phenomenon of tip convolution.
To circumvent this type of problem, the authors use half of the measured height of the
particles as a true measure for the sphere radius. Note that the height is not affected by
the size of the probe.
For an exact determination of the sphere radii, one must consider elastic deformation
under the influence of attractive adhesive forces, compressibility due to the interaction
with the styli, and the possibility of particle conformation on the surface. The amount of
elastic deformation was calculated by U. Schwartz.8 The calculated deformation gives
the range of 0.5–0.9 nm. The largest deformation of 0.9 nm was estimated for a 220 nm
latex sphere and is less than 0.4%. Scanning-electron microscopy images of polystyrene
spheres of 100 nm and colloidal gold of 28 nm do not reveal any evidence that particles
are being deformed on the substrate. The compressibility comparison on colloidal gold
spheres and the tobacco mosaic virus has been reposted by J. Vasenka.9 The study suggests
that colloidal gold particles are incompressible and can be independently characterized
through electron microscopy. This characteristic makes colloidal gold particles a good
candidate for tip geometry characterization. Some particles may collapse into a pancake-like
shape or retain their spherical shape depending on rigidity of the particle and substrate
surface free energy.10 However, the authors are not aware of any reported data on latex or
colloidal gold spheres suggesting the possibility of conformation on the surface.
The narrowness of the size distribution curve for the nominal size value is critical
to AFM particle characterization analysis. There are several manufacturers producing
spherical particles with a nominal size below 100 nm. Usually spherical shape and size
is independently measured with transmission-electron microscopy or photon correlation
spectroscopy (PCS). The standard deviation varies between 4.5–7.2 nm, which is 5–16%,
with a mean diameter of 100 nm and 50 nm, respectively. The specification for particles
with a 20 nm diameter is 21 nm ± 1.5 nm, calibrated with PCS (Duke Scientific, Inc.).
In the case of colloidal gold particles, the reported coefficient of variation for a mean
diameter of 4.79 nm is 11.1%, for 20.51 nm–5.4%, and for 30.7 nm–3.9% (Ted Pella,
Inc., Redding, California).
In addition, it has been suggested that fullerenes may be used as uniform spheres for
tip characterization.4 The outer diameter of carbon-60 is 10.18A.
|
Probe artifacts in the case of optical
and electron microscopy, such as aberration,
astigmatism, and distortion, are well
known, studied, and mostly compensated
for in the commercially available microscope.
Note that theoretical resolution
in TEM cannot be obtained, primarily
due to the spherical aberration of the
lens.2 A tip artifact or a tip dilation, the
phenomenon specific to AFM, manifests
itself in a broadening of the lateral
dimensions of the surface topography.
Interestingly, the vertical resolution of
AFM is not affected by the finite size of
the probe. In mathematics, this problem
is known as tip deconvolution. Imaging
very sharp vertical surfaces (those with
high relief) is influenced by the sharpness
of the tip. Only a tip with sufficient
sharpness and aspect ratio can properly
image a given vertical or horizontal
profile. Some profiles can be steeper or
sharper than any tip can be expected to
image without artifact. False images are
generated that reflect the convolution of
the tip geometry and the geometry of
the object being imaged, rather than the
object surface. Mathematical methods
of tip deconvolution can be employed
for image restoration.3–6 The effectiveness
of these methods depends on the
specific characteristics of the sample and
of the probe tip. A known tip implies a
known sample; if this is the case, then
the morphological subtraction3,4 or
envelope method5 can be used. If the
sample is unknown, then the method of
blind reconstruction can be utilized.3,6
It has been shown that the combination
of erosion with blind reconstruction can
produce the most optimum deconvolution
results on a known sample.4 The
major obstacle in AFM, a tip artifact
associated with the finite size of the
probe, can be compensated for by using
a deconvolution method before performing
analytical measurements. (Note that
all AFM images shown in this paper
are obtained on Nano-Rp™: Nanoflat
substrates are used for the sample
preparation; and AFM image analysis
is performed in NanoRule +™ particle
analysis software.)
See the sidebar for details on particle
characterization.
TIP RECONSTRUCTION ON A PARTICLE SAMPLE
It is possible to reconstruct tip
geometry on both known and unknown
samples. This article will focus on known
samples. To refer to a sample as “known”
implies that the size and geometry
of particles used as tip characterizers
can be independently verified. From a
mathematical perspective, it is important
to be aware of errors associated with a
known sample. Naturally, it is desirable
to have a minimum of such errors. The
errors in the estimate of characterizer
dimensions propagate to comparable
errors in the reconstructed tip shape.
The characterizer should be stable at
the nanoscale with dimensional errors
that are small compared to the tip size.
Commercially available probes have tip
diameters at about 20–30 nm. However,
there are some special probes with tip
diameters as small as 1 nm.
|
"The uncertainty associated with the
pitch size is comparable to the size of
the probe." |
|
Patterned silicon wafers are routinely
used for dimensional calibration of AFM. National Institute of Standards
and Technology traceable gratings are
available from VLSI Standards, Inc., of
San Jose, California. The grid consists
of an array of alternating bars and spaces
with uniform pitch in both X and Y directions.
The uncertainty associated with the
pitch size is comparable to the size of
the probe. The reported value is on the
order of 20 nm for a 2.99 μm pitch. Note
that the size of the square impression is
about 1.5 µm × 1.5 μm. The calibration
grating consisting of an array of sharp
tips, with strict symmetry on tip sides,
small cone angle (less than 20°), and
small curvature radius of the tips (less
than 10 nm) is a very good candidate for
determination of styli shape parameters11 (NT-MDT, Moscow). However, the lack
of specifications for the parameters
mentioned and the complicated shape
of the spikes makes the problem more
complicated. Figure 2 shows an AFM
image of the pillars in comparison to
the SEM image. Both AFM and SEM
data agree that the spikes appear to have
variations in shape and size. The use of
single spherical particles has been proposed
for tip characterization for several
reasons.
The high degree of symmetry, manufacturability,
consistent uncertanties on
the scale of 1–3 nm, noncompressibility,
negligible elastic deformation, and conformation
on the substrate make colloidal
gold and polystyrene particles favorable
candidates for tip geometry checkers.3,4,9,12 Figure 3 shows one example
of a three-dimensional view of a tip
reconstructed from an AFM image of a
102 ± 3 nm polystyrene sphere.
The reconstructed tip appears to
be asymmetrical in shape with a tip
radius of ~38 nm, a left angle of 65°,
and a right angle of 50° (Figure 4). The
reconstructed tip obtained on scanning
28 nm colloidal gold spheres (Ted Pella,
Inc., Redding, California) is shown on
Figure 5. The tip has a radius of ~2 nm,
a left angle of 58°, and a right angle of
46°. The reconstructed tip is outer bound
on the part of the tip which contacts the
particle during the scan. This happens
due to the fact that dilation and erosion
operators are not strict inverses of one
another for an arbitrary tip.4 The closing
operation tends to smooth the edge from
the outside. The tip’s outer bound estimate
is the more important one because
it leads to a lower bound estimate of the
sample surface.3
SURFACE RECONSTRUCTION
Once the geometry of the tip is estimated,
it is possible to reconstruct the
surface topography. Figure 6 shows AFM
images of the particles before and after
reconstruction. Ideally, if a tip is perfectly
reconstructed, the reconstructed
sample surface will be an outer bound
that contains an actual sample. This outer
bound will be equal to the true sample
surface at those points where the tip
actually touched, and it will be larger at
points where the tip was unable to touch.
Notice that the true particle surface is a
cylinder with a hemisphere cap. That is,
it contains vertical sidewalls where the
tip was not able to touch. This is one
of the reasons that these reconstructed
samples seem dilated (compare Figures
Be, computer-simulated tip dilation, and
Figure 6, tip dilation on measured and
reconstructed surfaces).
AFM RESOLUTION ON PARTICLES
The conventional definition of resolution
in the field of light and electron
microscopy is based on the Lord Rayleigh
criterion: when the maximum
intensity of an Airy disc coincides with
the first minimum of the second, then
two points can be distinguished.2 This sets the resolution limit as d/2, as shown
in Figure 7. The Rayleigh criterion can
be applied to three-dimentional images
of AFM. Here, the intensity peaks are
analogous to height line profiles. Let us
assume “r” is tip radii, “R” is sphere radii,
and “h” is the height difference between
the top of the sphere and the lowest
point in a line profile drawn through two
spheres. H is the distance that actually
can be measured on the height profile.
These line profile measurements should
be performed on deconvoluted surfaces
to compensate for tip artifact.
The easiest case, that in which particles
are far apart, is shown in Figure 8a. The
particles are clearly resolved, and the
probe goes all the way down between
two particles. Criterion “h” measures
the height difference between the top
of the sphere and the lowest point in the
line profile drawn between two adjacent
spheres. In this case, h = 2R (Figure 8b).
In Figure 8c, the spheres touch each
other and it is assumed that there is no
interaction. The height line profile is
shown in Figure 8d. The height of the
sphere remains the same, while the ability
of the probe to go between the particles
will be limited by the finite diameter of
the probe. If we assume the ideal case,
in which the probe is represented by
delta-function, then, h = R. If the finite
size of the probe and the upper-lower
bounds of the tip reconstruction are taken
into account, then separation, t, between
two spheres will be smaller. This can be
shown, if the size of the probe is equal
to the diameter of the sphere, h = 0.26R.
If the diameter of the sphere is one-half
of the diameter of the probe, then h =
0.39R. If r = 0.2R—the most realistic
case for 50 nm radius spheres—then h
=0.54R = 27 nm. This criteria could be
considered the spatial resolving limit for
AFM. To summarize: h = R-((2Rr+r2)1/2
–r); h = 0.26-corresponds to the r = R;
h = 0.54R-resolution criteria, r = 0.2R;
h = R, probe-delta functions, r = 0.
CONCLUSION
In this work, tip deconvolution and
image restoration are done on known,
pre-characterized manufacturable
spheres from a commercial vendor,
rather than an arbitrary unknown surface
or mathematically modeled structure.
This is a practical, innovative solution
to the long-standing problem of tip
deconvolution.
REFERENCES
1. A. Jillavenkatesa, S. Dapkunas, and L. Lum, Particle
Size Characterization (Gaithersburg, MD: NIST, 2001).
2. P. Goodhew, J. Humphreys, and R. Beanland,
Electron Microscopy and Analysis (New York, NY:
Taylor & Francis, 2001).
3. J.S. Villarrubia, Surf. Science, 321 (1994), pp. 287–300.
4. J.S. Villarrubia, J. Vac. Sci. Technol. B, 14 (2) (1996),
pp. 1518–1521.
5. D. Keller, Surf. Science, 294 (1993), pp. 409–419.
6. J.S. Villarrubia, Journal of NIST, 102 (1997), pp.
425–454.
7. P. West and N. Starostina, “Quantitative and
Qualitative Nanopowder Nanoparticle Characterization
with AFM,” Proceedings of Advances in Powder
Metallurgy & Particulate Materials (Princeton, NJ:
Metal Powder Industries Federation, 2006)
8. C. Ritter et al., Langmuir, 18 (2002), pp. 7798–7803.
9. J. Vasenka et al., Biophysical Journal, 65 (1993), pp.
992–997.
10. T. Dukette and M. Mackay, Nano Lett., 5 (9) (2005),
pp. 1704–1709.
11. V. Bykov, A. Gologanov, and V. Shevyakov, Applied
Physics A, 66 (1998), pp. 499–502.
12. M. Van Cleef et al., Journal of Microscopy, Pt. 1,
181 (1996), pp. 2–9.
P.E. West and N.V. Starostina are with Pacific
Nanotechnology Inc., 17984 Sky Park Circle, Suite
J, Irvine, California 92614. K.S. Olson and M.L.
Mecartney are with the Department of Chemical
Engineering and Material Science at the University
of California–Irvine. Ch. Wong is with the Claremont
Graduate University School of Mathematical Sciences,
Claremont, California. N. Starostina can be
reached at (949) 253-8813; fax (949) 253-8816; e-mail
nstarostina@pacificnanotech.com
|