Sunteți pe pagina 1din 60

Chapter

6
TSEM: A Review of Scanning
Electron Microscopy in
Transmission Mode and Its
Applications
Tobias Klein , Egbert Buhr , and Carl Georg Frase

Contents

1. Introduction
I Technique
2. Common Electron Microscopy Techniques
2.1. Transmission Electron Microscopy
2.2. Scanning Electron Microscopy
2.3. Scanning Transmission Electron Microscopy
3. TSEM Signal Generation
3.1. Detection of Transmitted Electrons
3.2. Physical Background: Electron Scattering
and Diffraction
3.3. Contrast Mechanisms
3.4. Monte Carlo Simulation of TSEM Signals
4. TSEM Compared with Common Electron Microscopy
Techniques
4.1. Resolution
4.2. Contrast
4.3. Energy-Dispersive X-Ray Spectroscopy
4.4. Sample Preparation and Throughput
II Applications
5. Traceable Dimensional Measurements
of Nanostructures
5.1. Calibration of an SEM
5.2. Mask Metrology
5.3. Nanoparticle Size Measurement
6. Characterization of Different Material Classes

298
299
299
300
301
302
303
303
307
310
314
320
320
323
324
324
326
326
326
328
331
340

Physikalisch-Technische Bundesanstalt, Bundesallee 100, 38116 Braunschweig, Germany

Advances in Imaging and Electron Physics, Volume 171, ISSN 1076-5670, DOI: 10.1016/B978-0-12-394297-5.00006-4.
c 2012 Elsevier Inc. All rights reserved.
Copyright

297

298

Tobias Klein, Egbert Buhr, and Carl Georg Frase

6.1. Biological Samples


6.2. Polymers
6.3. Semiconductor
6.4. Material Science
7. Special Imaging Modes
7.1. TSEM in Liquids
7.2. Electron Energy-Loss Spectroscopy
7.3. Tomography
7.4. Visualization of Electric Fields
8. Conclusion
List of Abbreviations
References

340
341
341
343
344
344
345
345
346
346
347
348

1. INTRODUCTION
The history of electron microscopy begins with Hans Busch (1926, 1927),
who introduced electron optics, and continues with Ernst Ruska, who
was awarded the Nobel Prize for building the first transmission electron microscope (TEM) (Knoll and Ruska, 1932), which led to commercial
instruments as early as 1939 (Wolpers, 1991). In a TEM, the sample is
illuminated with a broad immobile beam. The principle of scanning a
focused electron beam across the sample was introduced by Manfred
Von Ardenne (1938a,b) and put into practice by Charles W. Oatley, whose
work led to the introduction of the first commercial scanning electron
microscope (SEM), the Cambridge Stereoscan, in 1965 (Oatley et al.,
1966). It featured detectors for secondary and backscattered electrons,
which represent the standard equipment of SEMs to this day. Parallel
development in Japan led to the launch of a second commercial SEM, the
JEOL JSM, only six months later (Fujita, 1986).
The JSM was additionally equipped with a transmission detector that
could be placed underneath the sample to detect primary electrons that
transmit through the specimen. Consequently, two employees of JEOL
were the first to discuss details of this mode, which today we call TSEM1
(Kimoto and Hashimoto, 1968).
Initially, there was little interest in transmission work with the SEM
because, at that time, TEMs had already been available for more than 25
years and had become capable of nanometer resolution. The transmission mode of an SEM did not seem to be a promising alternative since
it allowed resolutions of only about 10 nm (Kimoto and Hashimoto, 1968).
1

In the literature, diverse abbreviations for SEM in transmission mode are used: TSEM, STEM, STEM-in-SEM,
STEM/SEM, STEM-SEM, LVSTEM (for low-voltage STEM), and others. Throughout this text we use TSEM,
which was most likely introduced by Postek et al. (1989). In analogy to a STEM (a TEM used in scanning mode)
we think TSEM is instructive for an SEM operated in transmission mode. We do not recommend the usage
of STEM since this abbreviation is mainly used to denote dedicated high-voltage instruments. Unfortunately,
most SEM manufacturers use this delusive term. STEM-in-SEM is unambiguous since it explicitly names what
is done, but its length may be a little awkward.

TSEM: A Review of SEM in Transmission Mode

299

Today, things have changed. Due to improved electron optics


especially designed for low-energy electrons and the use of field emission
guns, resolutions down to 0.4 nm have been demonstrated using TSEM
(Van Ngo et al., 2007). At the same time, the demand for high-resolution
imaging is growing due to technological progress and ongoing miniaturization in semiconductor industry, nanotechnology, and material science.
In principle, TEM is capable of fulfilling this demand, whereas in practice there are obstacles. The purchase and operation of a TEM are costly
and analyzing samples takes a long time. If the highest resolution is not
needed, TSEM is an inexpensive and fast alternative that considerably
increases sample throughput and reduces the costs per sample.
TSEM is well suited for nanotechnological applications since the optimal sample thickness for TSEM ranges from some nanometers to some
hundred nanometers. In this range, TSEM yields high-contrast images
even for low-Z materials due to strong electron scattering. In addition, it
provides higher lateral resolution compared with common SEM imaging
modes, since it is not confined by the size of the interaction volume.
Consequently, there is rising interest in TSEM and today all SEM manufacturers offer the possibility of upgrading their SEMs with transmission
detectors. Although still on a small scale, the amount of publications of
work accomplished with the help of TSEM has been steadily increasing
since the turn of the millennium. Many of these publications do not explicitly deal with TSEM but use it as a given tool to accomplish the task at
hand.
This contribution summarizes the basics of the TSEM technique (Part I)
and presents an overview of the fields in which it has been applied to date
(Part II). Part I begins with a short introduction of the common electron
microscopy techniques. Subsequently, the configuration and the implementation of the transmission detector are introduced. Electron scattering
is briefly described, which is the basis of the different contrast mechanisms and may be simulated using Monte Carlo methods. In the following
section, we compare some aspects of TSEM with the common electron
microscopy techniques. The overview of TSEM applications in Part II
starts with traceable dimensional measurements before the characterization of various material classes is described and special imaging modes
are presented.

PART I: TECHNIQUE
2. COMMON ELECTRON MICROSCOPY TECHNIQUES
This section introduces the well-known and commonly used electron
microscopy techniques. They share many basic principles with TSEM as

300

Tobias Klein, Egbert Buhr, and Carl Georg Frase

discussed in Section 3 on TSEM signal generation. A comparison of some


aspects between the common techniques and TSEM follows in Section 4.

2.1. Transmission Electron Microscopy


The design of a transmission electron microscope is shown on the left side
of Figure 1. It basically resembles that of a transmission light microscope.
The electrons emitted by the source are accelerated (typically by 200 kV)
on their way to the sample. Condenser optics is used to achieve a spatially
uniform illumination on the electron-transparent specimen. After passing
through the sample, the electrons are collected and imaged on a projection
screen by means of electron optics. A scintillator converts the impinging
electrons to light pulses that may also be detected by a charge-coupled
device (CCD) for direct image recording using a computer.
The electrons are scattered, diffracted, and possibly absorbed by the
specimen, depending on different sample properties, such as mass, thickness, elemental composition, and crystallinity. The image is generated
by detecting electrons that are deflected within certain angular ranges,
which are determined by the aperture of the imaging lens. Dark-field (DF)
imaging uses only the deflected electrons for image formation, whereas in
bright-field (BF) mode the beam of undeflected electrons is registered. In
high-resolution TEM mainly phase contrast is exploited. The sample acts
as a phase object that distorts the wave front of the impinging electron
Detector
Source
Specimen
Aperture
Lens
Scanning
coils

Screen : Source

FIGURE 1 Schematic of the ray path in a TEM (left) and a STEM (right) demonstrating
the principle of reciprocity (Crewe and Wall, 1970). Image reprinted with kind permission
from Elsevier.

TSEM: A Review of SEM in Transmission Mode

301

wave. Under Scherzer defocus conditions, phase contrast is transformed


into visible amplitude contrast (Scherzer, 1949).

2.2. Scanning Electron Microscopy


In an SEM, the electron beam is focused on the sample and scanned in
a raster. As a consequence of multiple scattering of beam electrons in
the specimen the interaction volume evolves, which can exceed the size
of the electron probe considerably (Figure 2). The interaction of the primary electrons with the sample leads to a variety of signals that may
be exploited to gain information. Because the detected signal originates
from electrons that emerge from the surface, there is no restriction to thin
electron-transparent samples. Due to the interaction volume, electrons
may also be excited far away from the scan position, leading to background noise and a decrease in image quality. The intensity of the detected
signal is converted to a grey-scale value, which is attributed to the current
scan position. In this way, a grey-scale image evolves pixel by pixel.
A portion of the primary electrons is scattered back and escapes
the sample. The detection of these backscattered electrons (BSEs) leads
Objective lens

SE
detector
Electron
beam

X-ray

SE3
SE1
SE2

BSE

SE escape
depth

Interaction
volume

BSE
Region of BSE
production

FIGURE 2 Overview of the various signal types that may be exploited in SEM imaging.
The resolution is often limited by the size of the interaction volume, which forms inside
the specimen due to multiple electron scattering.

302

Tobias Klein, Egbert Buhr, and Carl Georg Frase

to intense material contrast because the backscatter efficiency strongly


depends on the atomic number of the element.
Imaging with secondary electrons (SEs), which exhibit energies below
50 eV by definition, is widely used. The number of SEs released depends
on the geometry of the sample and on the angle of incidence of the electron beam. In combination with a large depth of focus due to small cone
angles of the beam, the well-known quasi three-dimensional (3D) image
impression is obtained.
There are several excitation processes for SE. The most important one is
the excitation of plasmons (oscillations of the electron gas) by high-energy
electrons and their subsequent decay, which may lead to the generation
of SEs. SEs are generated in the whole interaction volume by primary and
backscattered electrons, but due to their low energies only those generated close to the surface are able to emit from the sample. SEs, which
are excited by primary electrons, the so-called SE1, are highly localized
(see Figure 2) and retain information about the sample area that is hit by
the beam spot. Because BSEs emerge from inside the interaction volume,
their travel range is not confined to the beam spot and the SE2 excited
by them originate from a relatively wide emission zone. If BSEs hit the
walls of the vacuum chamber or the electron column, SE3 are produced,
which contribute to the background noise, thereby decreasing contrast.
If the electron beam reaches an edge on the sample, SE emission is no
longer confined to the horizontal sample surface but may also take place
at the vertical edge. This so-called blooming effect may obscure the exact
position of the edge.
After impact ionization, the resulting vacancy in the shell may be filled
by an electron from a higher energy level, thereby possibly emitting an
X-ray photon specifically for the atomic element. These X-rays can be analyzed by energy-dispersive X-ray spectroscopy (EDX) to gain information
about the elemental composition of the sample.

2.3. Scanning Transmission Electron Microscopy


In a scanning transmission electron microscope (STEM), the scanning
principle is combined with high-voltage operation. The electrons transmitted through the sample are detected, depending on their angle of
deflection or their energy loss. In addition to BF and DF imaging the
high-angle annular dark-field (HAADF) method is often used because
it is highly element specific. Regarding energy-dependent detection, the
allocation and concentration of elements can be determined by electron
energy-loss spectroscopy because the distribution of energylosses of
primary electrons is characteristic for specific elements.
When Crewe and his coworkers put STEM into practice at the end of
the 1960s (Crewe and Wall, 1970; Crewe et al., 1970, 1968), the electron

TSEM: A Review of SEM in Transmission Mode

303

optics theory of TEM imaging was already well developed. Cowley (1969)
realized that the so-called principle of reciprocity (Pogany and Turner,
1968; Von Laue, 1935) allows the interpretation of diffraction phenomena in a STEM. The principle of reciprocity states that if the detector
acceptance angle of a STEM equals the angle of incidence of a TEM,
then exactly the same micrograph can be obtained with the two instruments (Humphreys, 1981). This is because the ray path in a STEM may
be regarded as being the same as in a TEM except for being reversed (see
Figure 1). Although the conditions usually do not match exactly, many
STEM imaging modes have equivalent and well-known TEM counterparts.
As described in the next section, TSEM and STEM share the same basic
concept. Thus the principle of reciprocity also holds for diffraction effects
in TSEM imaging.

3. TSEM SIGNAL GENERATION


3.1. Detection of Transmitted Electrons
3.1.1. Detector Configuration and Imaging Modes
TSEM is implemented into standard SEMs using their acceleration voltages usually not exceeding 30 kV. The electron beam focused by the
probe-forming lens is scanned across the sample by means of scanning
coils, and transmitted electrons (TEs) are registered by a transmission
detector underneath the sample.
The beam electrons interact with the sample in elastic and inelastic
scattering processes. If the sample is crystalline, diffraction from its lattice
planes must also be considered. Consequently, the energetic and angular
distribution of the beam electrons is changed, which may be exploited to
generate image contrast.
Since the energy loss of primary electrons is usually small compared
with their initial energy, mainly the angular distribution of the scattered
electrons is exploited in TSEM imaging by limiting the angular detection
range. Three ranges are differentiated (Figure 3), although not all of them
are necessarily implemented. In the BF mode only the central beam and
barely deflected electrons are detected, whereas the electrons scattered
to higher angles are blocked by a circular aperture. In the DF mode the
primary beam is blocked and only electrons scattered to medium angles
contribute to the image. This mode may be realized either by annular apertures or by separate detectors. If only electrons scattered to high angles
are detected, the resulting signal is usually not influenced by diffraction
effects which are mostly confined to small diffraction angles. This is why
HAADF detectors are sometimes implemented.

304

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Objective lens

Electron beam

Sample

HAADF

DF

BF

FIGURE 3 In transmission imaging, detector segments for bright-field (BF) and


dark-field imaging (DF) are usually provided, sometimes accompanied by segments for
the high-angle annular dark field (HAADF) mode.

100 nm

(a)

100 nm

(b)

FIGURE 4 Micrographs of silica particles in the hole of a lacy carbon film taken
(a) in bright-field and (b) in dark-field mode (Buhr et al., 2009). Images reprinted with
kind permission from IOP Publishing.

In BF imaging, sample areas that scatter electrons strongly have a dark


appearance (Figure 4a). The image contrast may be adjusted by varying
the acceptance angle of the BF detector, which determines the ratio of
electrons that are detected and omitted, respectively.

TSEM: A Review of SEM in Transmission Mode

305

The DF mode is characterized by high contrast since only scattered


electrons contribute to the image. The majority of electrons belong to the
unscattered beam, which is not detected. Therefore, DF imaging is sensitive to small variations in sample properties. The obtained micrographs
are often complementary to BF images (Figure 4b).
In dedicated STEMs the detector acceptance angle can be adjusted versatilely by means of a projection lens. Since there usually are no projection
lenses in TSEMs the adjustment is somewhat constrained. Main changes
are made by changing the geometry of the detector or using apertures of
different sizes. Smaller adjustments are possible by varying the distance
between the sample and the detector. The best choice of the acceptance
angle is governed by the sample and the desired type of information.

3.1.2. Implementation
The simplest way to enable transmission imaging without modifying
the SEM is to use a sample holder that can convert the transmitted
electrons to SEs which are subsequently registered using the standard
EverhartThornley SE detector. Crawford and Liley (1970) used a polished
aluminum block for conversion that was slanted and oriented towards the
SE detector. By moving the block it was possible to optimize image contrast and to choose for BF or DF imaging. Nemanic and Everhart (1973)
used a specimen stub for conversion. An aperture was mounted above
the stub to improve contrast. A specimen stub for DF imaging was introduced by McKinney and Hough (1976). The specimen was placed above
the gold-sputtered stub with a diametral line of carbon. While the scattered electrons impinging on the gold led to the generation of SEs that
were detected, the unscattered electrons were effectively absorbed by the
carbon.
The elaborate conversion sample holder of Oho et al. (1987b) consisted of three surfaces slanted at different angels. With this design the
BSEs produced when the transmitted electrons hit the uppermost surface could be also exploited in the conversion process. Vanderlinde (2002)
improved the achievable resolution of conversion imaging by introducing a graphite collimator to effectively absorb electrons scattered to high
angles.
The main disadvantage of the conversion approach is an inferior
signal-to-noise ratio (SNR) compared with dedicated transmission detectors, which are usually based either on solid-state detectors or a scintillator
coupled to a photomultiplier via a light tube. The latter approach was
used in the JEOL JSM, the first commercial SEM offering TSEM imaging (Kimoto and Hashimoto, 1968). Today scintillation detectors may be
found in SEMs from Hitachi and JEOL.

306

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Homemade transmission detectors using scintillators have also been


described: Swift et al. (1969) upgraded a Cambridge Stereoscan by coupling the scintillator for transmission measurements via a light pipe to
the photomultiplier of the EverhartThornley detector. Wells and Bremer (1970) used a similar approach with a turret that enabled them to
change between up to four detectors coupled to the Stereoscans photomultiplier, one of which was a transmission detector. In the arrangement
of Woolf et al. (1972) both the sample and the interchangeable aperture
could be moved independently by means of two separate xy-translation
stages. Krzyzanek et al. (2004) installed a Faraday cup at the central point
of a scintillator enabling DF transmission imaging simultaneous to beam
current measurements with the aim of thickness determination.
Although scintillation detectors may be faster and may exhibit better
SNRs, most commercial transmission detectors use solid-state detectors
in various configurations. They are smaller, more flexible, and thus easier to implement, especially if several detector areas for different imaging
modes are to be combined. Vendors of SEM accessories also distribute
transmission detectors that can be incorporated into most current SEMs.
One example is the transmission detector from K.E. Development that
was applied for nanoparticle characterization (see Section 5.3). It consists of five solid-state detectors (Figure 5). The four detectors on top
are used for DF imaging. The detector area for BF imaging is placed
underneath a small aperture. Since the geometry of the aperture cannot be

Specimen

Dark-field
detector

Bright-field
detector

FIGURE 5 Scheme of a typical solid-state transmission detector capable of bright-field


and dark-field imaging. Image courtesy of Carl Zeiss NTS GmbH.

TSEM: A Review of SEM in Transmission Mode

307

changed, the acceptance angle must be adjusted by varying the distance


between sample and detector. In principle, annular BSE detectors may
also be used as transmission detectors if they are mounted underneath the
sample.
To the authors knowledge the FEI transmission detector is the
only commercially available one that offers extra detector segments for
HAADF imaging (Pfaff et al., 2010; Volkenandt et al., 2010; Young et al.,
2008). The highest resolutions so far have been demonstrated using a
special SEM from Hitachi in which the sample is located inside the electron column (Tuysuz et al., 2008; Van Ngo et al., 2007). One disadvantage
is that this approach imposes similar restrictions on the sample size as
high-voltage instruments.

3.2. Physical Background: Electron Scattering and Diffraction


Scattering and diffraction are discussed briefly in this subsection since
they are the physical origin of contrast generation in the TSEM mode. The
books by Reimer (1998, 2008) cover many aspects of electron scattering
relevant to electron microscopy, whereas a thorough treatment is given by
Wang (1995).

3.2.1. Elastic Scattering


In solid state, elastic scattering is the most important interaction leading
to electron deflection. It is a result of Coulomb interaction of an electron with energy E with the potential of a core with atomic number Z.
In the Rutherford approximation the differential scattering cross section is
given by
1
e4 Z2
1
d
=

2
2
4
d
(4 0 ) 16E sin

.

(1)

Therein, denotes the scattering angle and e is the charge of an electron.


This formula is helpful in the interpretation of TSEM imaging, albeit
not absolutely exactly. For example, shell electrons are not considered that
screen the nucleus from the traversing electron, eliminating interaction
outside the neutral atom. They are accounted for in the Wentzel model by
an additional exponentially decreasing term (Wentzel, 1926).
For a thorough understanding, quantum mechanical and relativistic
effects such as spin-orbit coupling must also be considered. This leads
to Mott cross sections for which no analytical expression can be given
(Mott and Massey, 1933). Figure 6 shows the ratio of Mott and Rutherford
scattering cross sections for a few materials. As can be seen, Rutherford
scattering is a reasonable approximation for low atomic numbers and high

308

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Si
Z = 14

3.0

3.5
3.0

2.5

r ( )

1.5

2.0
1.5

1.0
180
150
120
at
90
te
rin
60
g
an
30
gl
e
0
()

1.0

0.5

Sc

0.1

10

180
150
120
at
90
te
rin
60
g
an
30
gl
e
()
0

0.5

Sc

50

eV)
y (k
erg

0.5

En

Au
Z = 79

1
0.1

0.5

10

50

eV)
y (k

rg

Ene

U
Z = 92

8.0

9.0
8.0

7.0

7.0

6.0
4.0

6.0
5.0

r ( )

5.0

4.0

3.0

3.0

2.0
180
150
Sc
120
at
90
te
rin
60
g
an
30
gl
e
()
0

1.0

1
0.1

0.5

rgy

Ene

V)
(ke

10

50

r ( )

2.5

2.0

r ( )

C
Z=6

180
150
Sc
120
at
90
te
rin
60
g
an
30
gl
e
()
0

2.0
1.0

1
0.1

0.5

rgy

Ene

10

50

V)
(ke

FIGURE 6 Ratio r( ) of differential Mott and Rutherford scattering cross sections as a


function of electron energy and scattering angle (Frase et al., 2009), calculated with the
Monte Carlo simulation program MOCASIM (Reimer et al., 1996). Image reprinted with
kind permission from IOP Publishing.

electron energies, whereas large discrepancies occur for heavy elements


and low energies.

3.2.2. Inelastic Scattering


Inelastic scattering occurs due to interactions between the incident electrons and the electrons of the specimen. Various interaction mechanisms
take place, all of which involve energy transfer (Reimer, 1998). A great
portion of the absorbed energy is eventually converted to heat, which is
the main reason for beam-induced sample damage.
The angular distribution of inelastic scattering (Figure 7) is restricted
mostly to very small angles (forward scattering), which are smaller than the
acceptance angle of typical BF detectors. Thus the number of electrons
registered by a BF detector is hardly influenced by inelastic scattering
processes. Consequently, variations of the signal detected in BF or DF
mode are predominantly due to elastically scattered electrons.

TSEM: A Review of SEM in Transmission Mode

309

104
1 d
2 d
aH
103

102

Ar

E = 11.7 eV

E = 0 eV

Elastic
Wentzel

10
2

1
104

103

102

rad

101

FIGURE 7 Angular dependence of differential scattering cross section for elastic and
inelastic scattering of electrons with an energy of 25 keV on an argon atom (normalized
by the square of the Bohr radius aH ) (Reimer, 2008). Image reprinted with kind
permission from Springer Science+Business Media.

3.2.3. Electron Diffraction


Bragg diffraction is evoked by regularly arranged lattice planes of crystalline materials. If electrons are reflected by two planes, positive interference occurs at certain angles that fulfill the Bragg condition. Although
Bragg diffraction is the most important contribution, further diffraction
effects may manifest themselves. For example, Fresnel fringes, which are
a result of diffraction at the edge of an aperture, have been shown using
TSEM (Broers, 1972).
Treacy and Gibson (1993, 1994) theoretically analyzed diffraction
effects by applying the concept of a coherence volume: Interference may
occur if a second atom is located inside the coherence volume around
the first one. Coherent intracolumn scattering vanishes above a certain
maximum scattering angle (Volkenandt et al., 2010)


p
max = 2 arcsin 0.61 /d ,

(2)

310

Tobias Klein, Egbert Buhr, and Carl Georg Frase

where denotes the electron wavelength and d the lattice spacing. This
knowledge may be used in Z-contrast imaging to avoid influences of
diffraction effects (see Section 3.3.2).

3.3. Contrast Mechanisms


3.3.1. Mass-Thickness Contrast
In Section 3.2, scattering on a single atom was discussed. In solid samples,
the probability of scattering events increases with growing sample thickness. The number of electrons that pass the sample without interaction can
be described by a simple exponential function as follows (Reimer, 2008):

n = n0 exp

t
3


(3)

with the number of incident electrons n0 , and the thickness of the sample t.
The mean free path 3 describes the average length of the path that an
electron travels between scattering events. It is inversely proportional to
the number of atoms per unit volume N, which can be expressed as N =
NA /ma with Avogadro constant NA , density , and atomic mass ma :
3=

1
ma
=
.
N
NA

(4)

The total scattering cross section of one atom, which depends on the
energy of the primary electron [see, e.g., Eq. (1) and Figure 6], is denoted
by .
In Figure 8, the mean free path 3 is shown for different materials and
electron energies. Usually, the sample is thicker than the mean free path
and thus multiple scattering occurs. Consequently, the electron beam is
broadened on its way through the sample. By varying the acceleration
voltage and thus the electron energy, the extent of beam broadening can
be changed to optimize the contrast of TSEM images.
For a constant electron energy, beam broadening grows with density
and thickness of the specimen (Figure 9). At the same time, more electrons are scattered to high angles and are registered by corresponding
DF detectors. Consequently, the BF signal decreases. For example, the
BF and DF images shown in Figure 4 reveal the divergent TSEM signal
intensities for silica spheres (high mass thickness), carbon film (low mass
thickness), and holes in the film. BF and DF imaging is usually dominated
by mass-thickness contrast, although it may be complemented by further
contrast mechanisms (e.g., by diffraction contrast in the case of crystalline
samples).

TSEM: A Review of SEM in Transmission Mode

311

15
C, Z = 6
AI, Z = 13
Cu, Z = 29

Mean free path (nm)

Ag, Z = 47
Au, Z = 79

10

10

15
20
25
Electron energy (keV)

30

35

FIGURE 8 Mean free path for varying electron energy and different materials, data
taken from Reimer (1998).

FIGURE 9 Illustration of beam broadening by a sample of varying thickness and


elemental composition: Only a few electrons are scattered out of the central beam by a
thin sample of light material (left). The number of scattered electrons grows if the
sample gets thicker (center) and if its density increases (right) (Goodhew et al., 2001).
Image reprinted with kind permission from Taylor & Francis.

312

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Initially, DF signal intensity grows if mass thickness is increased, but


this simple relationship does not hold constantly. At a certain thickness,
fewer electrons will be able to penetrate through the sample due to
absorption and backscattering, and thus the signal intensity decreases.
Even contrast inversion is possible if the thickness is increased further
(Grillon, 2006; Morikawa et al., 2006).

3.3.2. Z-Contrast
The deflection of electrons to large angles is dominated by elastic scattering, which depends strongly on the element as can be noted from
the Rutherford formula [Eq. (1)], which reveals that the scattering cross
section is proportional to the square of the atomic number: d/d Z2 .
This dependence is the origin of the term Z-contrast, which is also called
material contrast.
Consequently, by detecting only those electrons that are scattered to
high angles, information about the elemental composition of the specimen may be obtained. This is done with the help of an HAADF detector.
To avoid influences due to diffraction effects, its angular acceptance range
should be confined to angles larger than the maximum angle for coherent
intracolumn scattering [Eq. (2)], which is discussed in Section 3.2.3 (Volkenandt et al., 2010). For example, Figure 10 shows the minimum HAADF
detection angle for different primary beam energies.
Z-contrast is one of the most important imaging modes in STEM, but it
was only occasionally exploited in TSEM. Since TSEM detectors that offer
extra detector segments for HAADF are now commercially available, the
use of Z-contrast imaging is growing (see Section 6 for examples).

Detector inner radius (rad)

0.35
0.30
0.25
0.20
0.15
0.10
0.05

25

50 75 100 125 150 175 200


Primary beam energy (keV)

FIGURE 10 Minimum HAADF detection angle to avoid influences of diffraction effects


while studying a GaAs sample (Volkenandt et al., 2010). Image reprinted with kind
permission from Cambridge Journals.

TSEM: A Review of SEM in Transmission Mode

313

3.3.3. Diffraction Contrast


In STEM, electron diffraction is visualized with the aid of a projection
lens underneath the sample, which enables the detection of diffraction
patterns. The examination of diffraction effects may also be conducted
in TSEM, although projection lenses are usually not available. Two techniques have been described for this purpose.
In the so-called Grigson technique, scanning coils are placed underneath the sample (Figure 11). Two imaging modes may be realized. If
the incoming beam is kept fixed at a certain sample position and the
extra coils are used for scanning an angular range, the electron diffraction pattern is recorded that originates from the sample area hit by the
electron beam. Alternatively, the incoming electron beam may be scanned
across the sample while the coils below the sample are adjusted to record
a fixed diffraction order. In this way, different crystalline structures can
be resolved laterally. Using the Grigson technique, information is available concerning the type of lattice, its spacing, defects, and so on (Joy and
Maher, 1976).
The rocking beam technique was used only occasionally because its
application is more elaborate than that of the Grigson technique. It is the
electron-optical reciprocal of the Grigson technique without the need for
extra scanning coils (Van Essen et al., 1970). Instead of a variation of the
detection angle, the angle of incidence is varied. With an annular detector
usually used for DF imaging, the rocking beam technique adds the

Specimen

Scan coils

Detector
aperture

Diaphragm

Detector

FIGURE 11 In the Grigson technique, diffraction contrast may be depicted by scanning


coils positioned underneath the specimen (Reimer, 1998). Image reprinted with kind
permission from Springer Science+Business Media.

314

Tobias Klein, Egbert Buhr, and Carl Georg Frase

detection of Kikuchi patterns to the possibilities of the Grigson technique


(Woolf et al., 1972).

3.4. Monte Carlo Simulation of TSEM Signals


Accurate quantitative measurement results based on SEM or TSEM
require an understanding of the image formation process (Frase et al.,
2007; Postek and Vladar, 2011). An established method to gain insight
into SEM image generation is the simulation of electron sample interactions by means of Monte Carlo methods (Frase et al., 2009). In Monte
Carlo simulations, stochastic physical events are modeled using random
numbers. Whereas a single simulation is more or less meaningless, many
simulations of the same physical process lead to meaningful information.
Since the first implementations of Monte Carlo simulations in the 1960s
they have grown powerful due to the exponential increase in computing power, which allows the simulation of a large number of individual
events, thus reducing statistical noise to an adequate level. In addition,
progressively more sophisticated and realistic algorithms can be used to
numerically simulate the imaging process involving secondary or transmitted electrons. The basis of these simulations is a physical model of the
electron scattering process in solid-state (Section 3.4.1) and appropriate
models of electron detection (Section 3.4.2).
The interaction of primary electrons with the sample can be modeled on the basis of the fundamental scattering theory introduced above.
The detection of the transmitted electrons is easily modeled, taking
into account the geometry and sensitivity of the transmission detector.
Together this leads to robust and straightforward modeling of TSEM signals, which is a significant advantage of TSEM imaging compared with
the detection of SEs: The generation, emission, and detection of SEs with
energies below 50 eV is considerably influenced by many minor factors,
such as specimen charging, surface oxidation, carbon contamination, and
electromagnetic fields of SEM components (Frase et al., 2009). In practice,
many Monte Carlo program packages use adjustment factors to achieve
satisfactory agreement of simulated and measured SE signals (Frase et al.,
2009). In contrast, these effects barely affect primary and transmitted electrons with energies of some ten keVs, thus making the simulation of TSEM
imaging more reliable (Postek et al., 1993).

3.4.1. Electron Diffusion in Solid State


In Monte Carlo simulation of electron diffusion in solid state, individual
electron trajectories caused by scattering in the sample are determined.
An electron that enters the sample at (x0 , y0 , z0 ) with an energy E0 travels
a distance s1 before it is scattered for the first time at (x1 , y1 , z1 ) (Figure 12).
The scattering event leads to a change in the direction of travel specified

TSEM: A Review of SEM in Transmission Mode

315

x0, y0, z0
s1
x1, y1, z1

s2
x2, y2, z2

xn, yn, zn

FIGURE 12 Scheme of an electron trajectory iteratively calculated as the result of a


series of scattering events.

by the scattering angle 1 and the azimuthal angle 1 . Subsequently, the


electron travels the distance s2 in this direction before it is scattered for the
second time at (x2 , y2 , z2 ), leading to angles 2 and 2 , and so forth.
The iterative simulation of scattering events stops when the electron
leaves the sample. If its direction of travel has been reversed and it
emerges from the sample surface in backward direction, it becomes a BSE.
If thin samples are examined, most electrons cross the sample and emerge
from its bottom as transmitted electrons moving in the direction specified by = n and = n . It is also possible for the electron to lose all its
energy along its path, leading to absorption by the sample.
Starting with the distance s1 , all parameters are determined randomly.
The probability p(s) that an electron has not yet been scattered at distance
s depends on the mean free path 3 (Reimer, 1998)

s
.
p(s) = exp
3


(5)

The probability that a scattering event occurs is given by


Rs

 
s
0 p(s)ds
P(s) = R
= 1 exp
,
3
0 p(s)ds

(6)

which can be simulated by a uniformly distributed random number R


(0, 1]. Depending on R, the traveled distance is determined by the inverse

316

Tobias Klein, Egbert Buhr, and Carl Georg Frase

function
s(R) = 3 ln(1 R)

(7)

or by s(R) = 3 ln(R), which is equivalent because both R and (1 R) are


uniformly distributed between 0 and 1.
The scattering angle of elastic scattering events is usually determined
by differential Mott scattering cross sections (Mott and Massey, 1933).
Because no analytical expression can be given and their numerical calculation is elaborate and time-consuming, they must be calculated beforehand.
Czyzewski et al. (1990) compiled a dataset used by Browning et al. (1994)
to deduce an approximation formula. As an alternative, Salvat and Mayol
(1993) published FORTRAN algorithms that enable everyone to compute
datasets that are adapted to the simulation task at hand and to the desired
accuracy. Values between precalculated data points are determined by
interpolation.
Because the azimuthal deflection of the electron is radially symmetric, the azimuthal angle is uniformly distributed and may be easily
determined by
= 2R

(8)

with a second, independent, random number R as defined above.


Inelastic scattering is characterized by small scattering angles (see
Section 3.2.2) and energy losses usually not exceeding 50 eV, which is
small compared with the initial energy of primary electrons. The small
deflections may often be neglected and inelastic scattering events do
not necessarily have to be simulated individually. Instead, a continuous
energy loss of the primary electrons may be assumed along their path
through the sample. This so-called continuous slowing down approximation (CSDA) was introduced by Bethe (1930). The mean energy loss dE per
path element ds is called stopping power S. For nonrelativistic energies the
simplified Bethe formula reads



dE NA e4 Z
1.166E


S= =
ln
(9)
ds
J
8 02 AE
with the atomic weight A. The stopping power depends on the ionization
energies of the shell electrons, which are generalized by a mean ionization
energy J. Its value (in eV) can be approximated as (Berger and Seltzer,
1964)
J = 11.5Z

for

Z 12

J = 9.76Z + 58.5Z0.19

for

Z 13.

(10)

TSEM: A Review of SEM in Transmission Mode

317

This approach fails if the electron energy is too small and the logarithmic term in Eq. (9) becomes negative, describing an increase in the
electron energy, which is physically impossible. Based on a semi-empirical
approach and physical considerations, Joy and Luo (1989) introduced a
mean ionization energy J0 that depends on the energy of the electron and
replaces J in Eq. (9):
J0 =

J
1+

kJ
E

(11)

The variable k is close to but always less than unity. Because the stopping
power is not strongly sensitive to k, the small differences between its values for various elements may be neglected and a constant value of 0.8576
may be chosen to simplify the argument of the logarithm in Eq. (9), which
then reads (Joy, 1991)


1.166E
NA e 4 Z
S=
ln
+1 .
J
8 02 AE

(12)

3.4.2. Detection of Transmitted Electrons


After a transmitted electron emerges from the sample at (x, y, z) with
energy E, its direction is specified by the angles and (Figure 12).
The normalized signal intensity registered by a detector can be written
in general form as
P
h(i , i , xi , yi , zi ) s(Ei )
I= i
.
(13)
N0 s(E0 )
Therein, N0 is the number of primary electrons impinging on the sample, s(E) denotes the energy sensitivity of the detector, and the acceptance function h( , , x, y, z) describes whether or not an electron hits the
detector (h = 1 or h = 0, respectively).
At high magnifications, the detector dimension is orders of magnitude
larger than the field of view; hence the exact emerging position (x, y, z) and
even the different scan positions of the electron beam may be neglected.
Since transmission detectors are usually radially symmetric, the dependence on the azimuthal angle also vanishes. A transmitted electron is
registered if the angle lies between the minimum and maximum detection angles of the respective detector segment: min max . This can
be expressed as
h( , , x, y, z) h( ) = H( min ) H(max )

(14)

318

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Sensitivity (1/pA)

0.5
0.4
0.3
0.2
0.1
0

14

16

18

20
22
24
26
Electron energy (kV)

28

30

FIGURE 13 Electrons with energies below a threshold of 13 kV are omitted by the


solid-state detector whose sensitivity increases linearly above this value (Buhr et al.,
2009). Image reprinted with kind permission from IOP Publishing.

using the Heaviside function H. If the assumptions made above are not
validfor example, due to azimuthal subdivided detectors or due to low
magnificationscorresponding geometric considerations lead to other
expressions for h( , , x, y, z).
The sensitivity s(E) depends on the type of detector. In the simplest
case of sufficiently fast electron counting, it is unity. Another example is
presented in Figure 13, which shows the linear sensitivity of the detector
used for nanoparticle characterization in Section 5.3.
To summarize, two parameters that can be easily determined are usually sufficient to characterize a transmission detector, its acceptance function h( ), and its sensitivity s(E). The acceptance function is determined on
the basis of the detector geometry, whereas the energy-dependent sensitivity can be measured by varying acceleration voltage and probe current
in the absence of a sample.

3.4.3. Two Simulation Programs: MONSEL and MCSEM


Various Monte Carlo program packages are available for the simulation
of SEM image formation, which differ in the implementation of electron
specimen interaction and in their intended use (Frase et al., 2009). These
packages are usually developed to simulate SE and BSE emission, but
some may also be used to simulate transmitted electrons. Two examples

TSEM: A Review of SEM in Transmission Mode

319

are presented by Young et al. (2008) and by Demers et al. (2011). In the
following, two simulation programsMONSEL and MCSEMare introduced, which were used in the analysis of linewidths and nanoparticles,
respectively, as described in Section 5.
The Monte Carlo simulation program MONSEL was implemented at
the National Institute of Standards and Technology (NIST) by Lowney and
Marx (1994) on the basis of an older Monte Carlo code of Myklebust et al.
(1976). A predecessor of MONSEL approximated Mott cross sections by a
screened Rutherford potential multiplied by the factor (1 + Z/300) (Postek
et al., 1993), while later Brownings formula (1994) is used. Inelastic
scattering is modeled by the CSDA.
MONSEL was specifically designed for the application discussed
in Section 5.2: the examination of X-ray masks using transmitted and
backscattered electrons. For example, the implementation of the homebuilt transmission detector was optimized using simulation results
(Postek et al., 1993). In the initial version, MONSEL-I, the sample geometry was restricted to one line on top of a substrate consisting of up to
three layers, and the excitation of quasi-free valence electrons was treated
as the only reason for SE emission using Mollers cross section (1932).
Subsequently, the code has evolved to be more flexible and to incorporate new implications, such as the proximity effects of neighboring lines
(Lowney, 1995b). With the second version, MONSEL-II (Lowney, 1995a),
the treatment of SEs was improved using Koteras (1990) formalism for
the generation of plasmons and their decay into SEs. MONSEL-III introduced a new specimen structure, a 2 2 array of finite lines, and allowed
two-dimensional (2D) plots (Lowney, 1996). Recently, MONSEL has been
rewritten in Java and merged with NISTMonte (Ritchie, 2005) to simulate arbitrary 3D specimens such as transistors (Postek and Vladar, 2011;
Villarrubia et al., 2007).
The Monte Carlo Simulation for Electron Microscopy (MCSEM) was
developed at the Physikalisch-Technische Bundesanstalt (PTB), Germanys national metrology institute (Gnieser et al., 2008; Johnsen et al.,
2010). It is a general-purpose simulation program that provides thorough
and realistic modeling for SEM and electron beam lithography up to acceleration voltages of 50 kV. It has been implemented in C++ (Stroustrup,
2003) using object-oriented techniques. In the meantime, a second version
has been written in Matlab (Matlab, 2009). A key feature of MCSEM is
its modular design leading to great versatility. The software may be easily adapted to new simulation tasks by enhancing existing modules or
by integrating new ones. Third-party code may also be readily adapted.
Figure 14 shows the standard modules for input and output, which interact with the core module that simulates diffusion of primary electrons in
the specimen using precalculated Mott cross sections (Salvat and Mayol,

320

Tobias Klein, Egbert Buhr, and Carl Georg Frase

3D Electron
beam
3D Specimen
model
Charge
model

Core:
Electron diffusion
in solid state
Scattering:
Mott cross section
Bethe approximation

Detector model:
SE / BSE / TE yield
Electron trajectories
Exit angles and energies

Output:
Image processing

FIGURE 14 Overview of the program modules of MCSEM grouped into input (left),
core (center), and output (right).

1993) as well as SE generation and emission based on a semi-empirical,


parametric model (Joy, 1987; Seiler, 1983). The CSDA in the modification
of Joy and Luo (1989) is used to describe the energy loss of the electrons.
Aside from point sources and Gaussian beams with different diameters, parallel and conical illumination can also be modeled at any focal
plane. Arbitrary specimen structures with free elemental composition
may be simulated in two or three dimensions. The structures are composed of possibly overlapping, basic predefined geometric bodies that can
consist of different elements or compounds.
Whereas charging effects are supposed to have little effect on highenergy primary electrons transmitting through thin samples, they may be
crucial for SE imaging. The transport of SEs in electromagnetic fields is
implemented into MCSEM, and specimen charging also can be modeled
on the basis of dielectric properties of the material.
Different detectors may be modeled at the same time in MCSEM,
based on the known exit angle and energy of all individual secondary,
backscattered, and transmitted electrons. The simulated signals at adjacent points may be composed as linescans producing signal profiles or as
whole synthetic grey-scale SEM or TSEM images. Furthermore, electron
trajectories in the specimen may be visualized in three dimensions.
Figure 15 shows a number of electron trajectories in a nanosphere.

4. TSEM COMPARED WITH COMMON ELECTRON


MICROSCOPY TECHNIQUES
4.1. Resolution
In this section, the main effects that limit the lateral resolution of TEM,
STEM and SEM are discussed and compared with TSEM.

TSEM: A Review of SEM in Transmission Mode

321

FIGURE 15 Hundred individual electron trajectories of beam electrons transmitting


through a latex sphere having a diameter of 100 nm.

4.1.1. Chromatic Aberration


While the energy spectrum of the escaping electrons is quite narrow for
thin samples examined in a TEM, energy losses of the electrons due to
inelastic scattering on their way through the specimen are more diverse
for thicker samples. Due to the resulting broader energy spectrum, the
impact of chromatic aberration of the objective lens behind the sample
grows, which leads to a variation in focal length for electrons exhibiting
different energies. The loss of resolution may be seen in Figure 16a, which
shows indium crystals on a thin Formvar film covered with a latex sphere
1.1 m in diameter (see Figure 16d). In comparison, Figure 16b shows
the same sample imaged with STEM. Because no objective lens is used in
STEM (and TSEM), the achievable resolution is not limited by chromatic
aberration and depends mainly on the size of the beam spot.

4.1.2. Beam Broadening


In STEM and TSEM, the resolution of thick samples is influenced by the
top-bottom effect (Gentsch et al., 1974): Because the beam size widens by
multiple scattering on the way through the sample, the resolution of features at the bottom of the sample is inferior to that at the top. As shown
in Figure 16c, a sphere placed underneath the indium crystals leads to
reduced intensities but the crystals can be depicted with high resolution. In contrast, if the sphere is on top of the crystals it leads to beam
broadening, which results in blurred edges and reduced resolution (see
Figure 16b).

322

Tobias Klein, Egbert Buhr, and Carl Georg Frase

100 kV, TEM, Bottom

100 kV, STEM, Bottom

(a)

(b)

Scanning
electron probe
Formvar film
Evaporated
In layer

Polystyrene
sphere

100 kV, STEM, Top


(c)

(d)

(e)

FIGURE 16 Demonstration of chromatic aberration (discussed in Section 4.1.1) and the


top-bottom effect (Section 4.1.2). The TEM image (a) of a layer of indium crystals
deposited on a Formvar film and covered by a latex sphere (d) is blurred due to
chromatic aberration of the objective lens. The STEM image of the same sample (b)
shows mediocre image quality due to beam spreading. Sharp STEM images (c) may be
taken if the sphere is located underneath the indium crystals (e). The arrows indicate the
same elongated feature. Results first published by Gentsch et al. (1974). Images reprinted
with kind permission from Springer Science+Business Media (Reimer, 2008).

Concerning resolution, TSEM is somewhat inferior to STEM due to


the reduced acceleration voltage and technical constraints. By going from
some hundred kilovolts to below 30 kV, the impact of lens aberrations
increases and thus the beam diameter grows. The extended focal length

TSEM: A Review of SEM in Transmission Mode

323

further deteriorates the size of the electron probe. Therefore, the distance
between the final lens and the sample should be kept as small as practical. Consequently, in a STEM, the sample is placed inside the electron
column for best resolution. A further advantage of this assembly is the
rigid interconnection between the sample and the electron-optical system.
In contrast, sample stage and electron optics move more or less independently in the case of TSEM. Hence, the resulting vibrations sometimes
ultimately limit the achievable resolution in practice, whereas the electron
optics is in principle capable of higher resolutions.
In more common SEM imaging modes, such vibrations play only a
minor role because usually the resolution is limited by the size of the interaction volume. BSEs and SEs are generated inside the interaction volume
and may be registered by corresponding detectors if their energy is high
enough to exit the sample. As discussed in Section 2.2, the emission zone
of SEs and BSEs is larger than the beam spot, thus leading to a deterioration of resolution. One possible way to improve the resolution in SE
and BSE imaging is to restrict the size of the interaction volume by reducing the acceleration voltage. However, this also leads to a larger beam
spot because the influence of the aberrations of the electronoptical system grows. Compared with those common SEM imaging modes, TSEM
exhibits a higher resolution because it is not limited by the interaction volume (Golla et al., 1994). Thus, maximum acceleration voltage may be used
for best resolution.

4.2. Contrast
The size of the interaction volume not only affects the resolution but also
governs the achievable contrast. It limits the size of objects that may be
depicted using SEs (Goldstein et al., 2003). If the object is smaller than
the interaction volume, the electron beam transmits through it and enters
the substrate. The smaller the object, the more scattering and SE emission take place in the substrate while the number of SEs emitted from
the object decreases. This leads to a high background noise level, a poor
SNR, and deteriorating contrast. At some point, scattering in the substrate
dominates and the contrast is no longer sufficient to observe the object of
interest.
Emission of SE2 from the substrate is reduced if a thin film of
light material is used. In relation, more well-localized SE1 are detected.
Although the noise is reduced, this does not help to raise the low overall
signal since the number of SEs emitted from the object due to inelastic scattering remains small. In comparison, many more elastic scattering events
occur, which mostly lead to deflections of the primary electrons sufficient

324

Tobias Klein, Egbert Buhr, and Carl Georg Frase

to distinguish them from unscattered ones. Consequently, small objects


can still be imaged with good contrast in TSEM mode.
The contrast of TSEM images can be improved further by reducing
the acceleration voltage as discussed in Section 3.3.1. This is particularly
important for low-Z materials such as biological samples (Section 6.1),
as well as for samples composed of materials exhibiting only small differences in densitylike polymer blends (Section 6.2). Being limited to a
maximum acceleration voltage of 30 kV, TSEM is able to show more details
of such samples than dedicated high-voltage instruments. Most current
SEMs can be operated at acceleration voltages as low as 1 kV, or even less,
with still quite good resolution (Pawley, 1992).

4.3. Energy-Dispersive X-Ray Spectroscopy


If the electron beam transmits small objects of interest, EDX spectra also
deteriorate because X-rays are mainly generated in the substrate. Because
X-rays are able to escape not only close to the surface but basically from
the whole interaction volume, the resolution and contrast of EDX are usually even worse than in SE imaging. Therefore, EDX analysis benefits
extraordinarily from thin supporting films and, thus, from reduced interaction volumes common in TSEM (Kotula, 2009; Laskin and Cowin, 2001;
Maggiore and Rubin, 1973; Vanderlinde and Chernoff, 2005), leading to
higher resolutions approaching the size of the electron beam for very thin
samples.
Since there is no more need to confine the interaction volume, large
acceleration voltages become possible, leading to additional high-energy
peaks in the X-ray spectrum. These peaks generally facilitate the identification of elements and without them, the determination of some elements
would not be possible at all. However, minimizing the interaction volume
also leads to a decrease in count rate, which needs to be compensated by
longer integration times. Also, in conjunction with a transmission detector, attention must be paid to peaks originating from the detector (Habicht
et al., 2001).

4.4. Sample Preparation and Throughput


Due to the restriction to electron-transparent specimens, the preparation
of bulk samples for TSEM is as sophisticated and time-consuming as in the
case of TEM. Fortunately, there is a rich knowledge of TEM sample preparation techniques available that may also be exploited for TSEM (Ayache
et al., 2010). For example, TEM grids are also conveniently used.

TSEM: A Review of SEM in Transmission Mode

325

In the case of focused ion beam (FIB) milling, TSEM is a valuable tool to speed up sample preparation (Kendrick et al., 2008).
Instead of transferring the milled sample to a TEM, quality control and
basic analysis can be accomplished parallel to milling if the SEM is
equipped with both FIB capabilities and a transmission detector (see also
Section 6.3).
A serious constraint of dedicated high-voltage instruments is the
reduced space that is available for the sample. Consequently, only one
sample may be analyzed at a time and the size of the sample is confined.
In contrast, arbitrary sample sizes and shapes can be examined by TSEM
due to the large specimen chamber of an SEM. Furthermore, the specimen
chamber can accommodate many samples simultaneously, enabling their
batch analysis without the need to break and reestablish the vacuum. For
example, up to 12 samples can be mounted on the turret-type multisample holder shown in Figure 17. They may be exchanged simply by rotating
the sample stage.
This leads to a considerable speed-up of batch sample analysis that is
also facilitated by further advantages over high-voltage techniques that
TSEM shares with SEM. In addition to versatility and relative ease of use,
there is the advantage of better and faster orientation due to the possibility
of very low magnifications. The acquisition and maintenance costs of an
SEM with a transmission detector are significantly lower than those of
TEM or STEM. Thus, TSEM cuts down on time and cost per sample.

FIGURE 17 Multisample holder used in Zeiss SEMs for TSEM examination.

326

Tobias Klein, Egbert Buhr, and Carl Georg Frase

PART II: APPLICATIONS


5. TRACEABLE DIMENSIONAL MEASUREMENTS
OF NANOSTRUCTURES
Todays modern lifefor example, electronic devices incorporating
integrated circuitsis based largely on progress in manufacturing structures with dimensions at the nanometer level. Because the functionality
of these structures depends critically on the structure size, accurate and
reliable measurement technologies for verification and quality assurance
are required. SEM is often used for this purpose since it offers lateral
resolutions in the nanometer range.
A quantitative indication of the quality and reliability of measurement
results is the measurement uncertainty, which characterizes the dispersion
of the measured results (Joint Committee for Guides in Metrology, 2008b).
Typically, the measurement uncertainty is stated for a 95% coverage probability, meaning that the true value lies in the stated interval with a probability of 95%. The Guide to the Expression of Uncertainty in Measurement
(GUM) establishes the general rules for evaluating and expressing uncertainties (Joint Committee for Guides in Metrology, 2008a). In general, the
measurement uncertainty consists of numerous contributions from the
measuring instrument, the sample under test, and from the measuring
procedure.
To enable comparability of measurement resultsfor instance, between different methods and instrumentsthe scale of the dimensional
measurement needs to be traceable to the definition of the SI unit meter.
This is ensured by relating the measuring result to a reference through an
unbroken chain of calibrations (each of which contributes to the measurement uncertainty). Thus for traceable measurements, a calibration of the
measuring instrument must be performed. One possibility is the use of
an adequate reference standard as described in Section 5.1. In the remainder of this section, two applications of TSEM for traceable dimensional
measurements are discussed: linewidth measurements of masks used in
X-ray lithography (Section 5.2) and, in more detail, size measurements of
nanoparticles (Section 5.3).

5.1. Calibration of an SEM


In SEM measurements, pixelated images are obtained and the relation to
the SI unit meter must be established by a calibration procedure. The pixel
size depends on the SEM parameters set during image acquisition, such as
magnification and scan speed. SEM manufacturers typically calibrate their
instruments for a couple of parameter settings and interpolate between

TSEM: A Review of SEM in Transmission Mode

327

them. This is not sufficient for highly accurate measurements because


the conversion factors determined in this way are typically exact only
within some percent, traceability is not ensured, and the uncertainty of
the calibration is not stated by the manufacturer.
A simple approach to calibrate an SEM is the use of a calibrated
gratingthat is, an artifact containing features with known dimensions.
For example, in the following, the calibration of the SEM that was used
for nanoparticle sizing (Section 5.3) is discussed. A grating from Advanced
Surface Microscopy was used (Chernoff et al., 2008), which consists of a 2D
pattern of aluminum bumps on silicon (Figure 18). The mean grating pitch
of about 144 nm is calibrated with an uncertainty of less than 10 pm using
traceable ultraviolet laser diffraction (Buhr et al., 2007).
The calibration of the SEM revealed two interesting points that need
to be considered for highly accurate measurements using an SEM. First,
the pixel sizes along the fast scanning direction (x-axis) and slow scanning direction (y-axis) differ by some tenths of a percent. This may be
attributed to the repositioning of the beam in the y direction after linescans in the x direction. The so-called leading edge distortion is the second
effect to be considered. The scan speed and thus also the pixel size of the
first 200 pixels in the x direction differ from those of the remaining image

FIGURE 18 Micrograph of the 2D grating used for calibration, taken in SE mode.

328

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Pixel size (nm)

4.5

4.4

4.3

4.2

4.1

200

400
600
x-coordinate in pixel

800

1000

FIGURE 19 Due to the leading edge distortion, the pixel size of the first 200 pixels
deviates from that of the remaining image.

(Figure 19). Therefore, these pixels are omitted during both calibration and
measurement.
An alternative way to ensure traceability of SEM measurements is
applied for mask metrology (see Section 5.2). In this approach, a sample
scanning technique that includes the measurement of the sample position
by interferometry is used; thus, there is no need for the calibration of pixel
sizes using an artifact. On the other hand, this approach is feasible only
for capturing one-dimensional (1D) signal profiles. Making full 2D images
would take too long because moving the stage is significantly slower than
scanning the electron beam.

5.2. Mask Metrology


The production of integrated circuits (ICs) relies on a lithographic process
to pattern functional layers on top of silicon wafers. Optical lithography is still able to fulfill the demands of todays integrated circuit
production due to constant technological improvements. Nevertheless,
a number of alternative production techniques have been and are still
discussed to overcome the physical limits of optical lithography. Among
others, the use of X-rays instead of light has been proposed (Cerrina,
1992; Peckerar and Maldonado, 1993). This approach used a mask consisting of a thin X-raytransparent silicon membrane with opaque gold
structures on top. The use of electron beams and so-called SCALPEL

TSEM: A Review of SEM in Transmission Mode

329

masks (for SCattering with Angular Limitation Projection Electron beam


Lithography) also has been proposed as another alternative to optical
lithography (Harriott, 1997; Waskiewicz, 1997). These masks consist of a
150-nm thin electron-transparent membrane of silicon nitride, structured
with bilayers of chromium and tungsten.
Farrow et al. (1997) stated that for the analysis of SCALPEL masks,
the TSEM approach is particularly suitable due to exploitation of the
same interaction process that is used during exposure of the wafer. They
concluded that with the aid of simulations the achievable measurement
uncertainty was not limited by the instrumentation but by the measurement object due to disturbing line edge roughness of the SCALPEL test
mask.
Because X-ray masks are also essentially electron transparent, TSEM
was also proposed as an ideal tool for traceable measurements of the
linewidth of the gold structures (Postek et al., 1989). In the following
decade of research, Postek and coworkers accomplished fundamental
work on traceable TSEM measurements as well as on its simulation.
Since their work remains relevant for TSEM measurements to this day,
it is outlined in this section regardless of the fact that X-ray lithography
has ultimately not been implemented on a large scale due to difficulties
associated with the production and handling of the fragile masks.
In 1989, Postek et al. laid out the basic concepts of X-ray mask metrology using TSEM and presented first experimental results, demonstrating
nanometer precision (Postek et al., 1989). The authors stressed the advantages of using transmitted electrons instead of SEs. The width of line
structures can be determined with higher accuracy because no blooming
of the edge occurs. Furthermore, linewidth measurements using TSEM
are sensitive to the base of the line, which is relevant to lithography. Further advantages of TSEM include the inspection of X-ray masks for defects
such as particles and voids (Postek et al., 1991). Buried voids cannot be
seen using SEs and the influence of low-Z particles may be exaggerated by
SE imaging due to topographic contrast mechanisms. In contrast, TSEM
imaging is a good approximation of the X-ray illumination of the wafer.
Voids manifest themselves as contrast differences and the detection of particles is restricted to those that block electrons and are thus prone to also
block X-rays.
In a comprehensive contribution to the subject, Postek et al. (1993) gave
a detailed description of the instrumentation, the measuring approach
based on simulations, and the main factors influencing measuring accuracy. They used a standard SEM retrofitted with a transmission detector
and an interferometer stage. Instead of scanning the electron beam, the
specimen stage was scanned while the electron probe was fixed. From the
signal profile (Figure 20a), the edge of the feature of interest, usually a

330

Tobias Klein, Egbert Buhr, and Carl Georg Frase

(a)

(b)

FIGURE 20 Measured (a) and simulated (b) transmission signal profiles across a gold
line with a linewidth of 500 nm showed good agreement (Lowney, 1995b). Note the small
notch on the sidewall. Increasing transmission is plotted downward to resemble the
actual shape of the line. Images reprinted by permission of John Wiley & Sons, Inc.

gold absorber line, could be deduced by comparison with simulations.


The findings of the simulation were also used for an improved detector design: Because the silicon membrane resulted in a small amount of
unscattered electrons, a semiconductor transmission detector with a large
acceptance angle was used to maximize the detected TSEM signal. The
detector was insensitive for electrons with energies below 35 keV; thus
SEs were omitted.
Uncertainty contributions of the instrument as well as those from
the X-ray mask were analyzed, revealing some advantages of the TSEM
method compared with SE imaging. One is its robustness to misalignment.
Due to the large detector area, axial alignment and spacing between detector and sample were not critical. Also, carbon contamination was no
problem under typical measurement conditions because it did not noticeably affect the high-energy primary electrons. However, the edges of the
studied gold absorber lines were slanted by a few degrees and TSEM was
sensitive to the side wall angle. Changing the angle resulted in significant
changes of the signal profile. Thus it was important that the surface of
the mask was perpendicular to the electron beam and that the slope angle
was accurately determined. Unfortunately, line edge roughness randomly
affected the effective slope angle. The determination of the linewidths
relied on Monte Carlo simulations using a predecessor of MONSEL (see
Section 3.4.3) that used the slope angle as an input parameter; thus the
achievable accuracy was limited by those mask imperfections. For example, the combined uncertainty for the measurement of a 250-nm line was
estimated to be 10 nm.
The simulations revealed the presence of a small notch in the linescan
(Figure 20b). However, in practice, the SNR was often not sufficient to

TSEM: A Review of SEM in Transmission Mode

331

reproducibly reveal the notch in the experimental data. After improving


the instrument by using a field emission gun and reducing the distance
of the sample, the notch could finally be routinely resolved as shown in
Figure 20a (Lowney, 1995b). Subsequently, the notch served as a beneficial
point to determine the linewidth with reduced uncertainties by automatic
algorithms.

5.3. Nanoparticle Size Measurement


Nanoparticles, whose size is confined to the nanometer range in all
three dimensions, exhibit new properties different from the bulk material
(Daniel and Astruc, 2004). Therefore, nanoparticles are increasingly
exploited not only in science and technology but also for improved consumer goods. Prominent examples of the latter are titanium dioxide
particles in sunscreen and silver particles in clothing.
Since the desired functionality of nanoparticles depends on their size,
reliable size measurements are required. They may be accomplished by
ensemble techniques, such as dynamic light scattering (DLS) and smallangle X-ray scattering (SAXS), that probe a large amount of particles at
the same time and provide their mean diameter. But ensemble techniques
share disadvantages in case of complex size distributions or when morphological examinations are needed (Gleber et al., 2010; Rasteiro et al.,
2008). Consequently, as a direct imaging method, electron microscopy is
often used to study nanoparticles. However, it struggles with poor statistics since probing many particles is usually time-consuming, especially
using TEM. SEM measurements in SE mode suffer from a blooming of
the SE signal at the particle boundary, thus hampering reliable size measurements. In contrast, TSEM is perfectly suited for this measurement task
due to high resolution (see Section 4.1) combined with simple and wellunderstood signal generation that is reliably simulated on the foundation
of fundamental scattering theory (see Section 3.4).
TSEM has been used in a number of studies examining nanoparticles. Habicht et al. (2001, 2004, 2006) studied microtubules of different
shapes as well as metal nanoparticles whose synthesis uses the tubules
as templates. Probst et al. (2007) examined tin-palladium particles as small
as 5 nm incorporated into carbon nanotubes. They compared the resolution for varying instrument parameters and detection modes. Maximum
resolution was demonstrated by Tuysuz et al. (2008), who analyzed mesoporous particles by means of SE and transmission imaging. Barkay et al.
(2009) were able to estimate the 3D shape of nanoparticles by combining the 2D projection image with thickness information gained from the
transmitted intensities. Krzyzanek and Reichelt (2009) determined the
height of latex nanospheres by comparing annular DF signals with Monte

332

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Carlo simulations. Laskin et al. (2006) showed how to overcome the problem of poor statistics by analyzing large numbers of particles, thanks
to automation. They used computer-controlled SEM to study different
aerosol particles. The micrographs were composed of a mixture of BSE
and DF TSEM signals, allowing the automated detection of a wide size
range of nanoparticles made from various materials.
Whereas the contributions discussed above present a profound basis,
none of the instruments used was traceably calibrated and no attempt
was made for highly accurate measurements. Aiming for traceable
size measurements of nanoparticles with small uncertainties, we developed a dedicated TSEM measuring procedure (Buhr et al., 2009; Klein
et al., 2011). We calibrated the instrument as described in Section 5.1
and used Monte Carlo simulations (Section 3.4) for quantitative analysis of the experimental TSEM signals. In the remaining parts of this
section (Sections 5.3.1 to 5.3.4) our approach is presented, which has
been developed within the framework of a European joint research project
(Implementing Metrology in the European Research Area, 2008). Exemplary measurement results of gold particles are shown compared with
TEM measurements.

5.3.1. Sample Preparation and Image Acquisition


Nanoparticles are often distributed in suspension. Sample preparation
should ensure that a representative fraction of the particles is present on
the substrate and it should facilitate their analysis. Therefore, a homogeneous distribution of individual particles across a TEM grid without
drying artifacts is desirable. To achieve this, a droplet of the suspension is deposited on TEM grids and after some time excess suspension
is removed using clean room tissue (Klein et al., 2011).
For the measurement, a Zeiss Leo Supra 35 VP is used, which is
equipped with the solid-state transmission detector from K.E. Development (see Section 3.1.2). Micrographs are taken in BF mode with an
acceptance half-angle of about 16 mrad, using an accelerating voltage of
30 kV. Figure 21 shows some examples. Thanks to automatic acquisition,
a series of TSEM images can be taken at a speed of more than one image
per minute in a predefined area of interest on the sample. Thus, the shortcoming of poor statistics often associated with electron microscopy can be
overcome.

5.3.2. Analysis of TSEM Images of Nanoparticles


In order to take advantage of the automatic image acquisition and to avoid
the risk of systematic deviations between different operators, an automatic
image analysis routine has been developed. The size is deduced from the

TSEM: A Review of SEM in Transmission Mode

100 nm

333

100 nm

(a)

(b)

200 nm

(c)

FIGURE 21 TSEM images of three gold nanoparticle samples with nominal diameters of
10 nm (a), 30 nm (b), and 60 nm (c).

projected area, which is determined by simply counting the pixels that


belong to the particle. The critical task is to distinguish these pixels from
the ones belonging to the background or to other particles. This is often
accomplished using global thresholding techniques (Sezgin and Sankur,
2004). These techniques calculate a threshold based on the distribution of
grey-scale values of the image, regardless of the depicted objects, which
leads to varying results depending on the chosen algorithm (Sadowski
et al., 2007).
For accurate measurements, a threshold determination based on the
physical effects of the image formation process is necessary. In our
approach, the threshold has been determined by Monte Carlo simulations using the program package MCSEM (see Section 3.4). Assuming
homogeneous spheres, the output of the simulation is a signal profile of
a scan across the center of the particle. For example, Figure 22 shows the
simulated signal profile across a latex sphere.
The threshold signal Sthres at the boundary of the sphere can be easily
deduced and converted to the corresponding grey-scale value in a real
TSEM image as follows:
gthres =

Sthres S0
(g1 g0 ) + g0 .
S1 S0

(15)

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Simulation

S1

g1

Experiment

gthres

Simulated signal

Sthres

200

0.7
150

100

0.5

Particle
boundary

S0
90

60

30

0
30
x-coordinate (nm)

60

Measured grey value

334

g0
90

FIGURE 22 The simulated signal profile across a latex sphere agrees well with the
measured data (Klein et al., 2011). Image reprinted with kind permission from IOP
Publishing.

The variables S0 and S1 denote the signal level in the center of the particle
and in the background outside it, respectively (see Figure 22). They can
be related to the grey-scale values g0 and g1 . The mean grey-scale value
of a few pixels at the particles center of mass yields g0 , whereas g1 is
calculated as the median grey-scale value of the background pixels of the
region of intrest (ROI) around the particle. This approach is insensitive
regarding changes in brightness and contrast as long as the image is not
under- or oversaturated.
As shown in Figure 23, the threshold signal at the particle boundary depends on both the material and the particle diameter. This is due
to different scattering properties of various materials and to increasing
interaction paths for growing particle diameters, respectively (Klein et al.,
2011). The interdependence can be taken into account by an iterative procedure of the image analysis routine that determines threshold and size
individually for every single particle. Global thresholding (Prewitt and
Mendelsohn, 1965) is used to obtain an initial guess of the particle size.
With this guess, an improved estimation of the threshold can be determined, leading to an improved estimation of the size, and so on. After a
couple of iterations both threshold and size remain stable.

TSEM: A Review of SEM in Transmission Mode

335

0.9
Latex
Silica
Gold

Threshold signal level

0.8

0.7

0.6

0.5

0.4

30

60
90
Diameter (nm)

120

150

FIGURE 23 The threshold signal at the particle boundary depends on both its size and
material (Klein et al., 2011). Image reprinted with kind permission from IOP Publishing.

The final threshold deduced from the iteration is used to separate the
particle from the background. Beforehand the ROI containing the particle is interpolated to obtain subpixel accuracy. The particle size is then
determined as the diameter of a sphere exhibiting the same projected area.
During image analysis, real particles must be distinguished from spurious objects, such as agglomerates or drying artifacts. For this purpose,
three intuitive geometric parameters have to be set: minimum and maximum particle size and minimum circularity. To simplify the selection, the
software presents all objects sorted by their size or by their circularity.

5.3.3. Uncertainty Budget


The uncertainty of the determined mean particle size consists of numerous contributions (Table 1), which are listed in this section in the order
of their importance. The largest part of the uncertainty budget is related
to image analysis. The choice of minimum and maximum particle size
and minimum circularity is often ambiguous with an interval of potentially appropriate values. The uncertainty is estimated from the impact of
varying the parameters within the limits of these intervals.
The second most important contribution to the overall uncertainty
originates from the Monte Carlo simulations on which the determination

336

Tobias Klein, Egbert Buhr, and Carl Georg Frase

TABLE 1 Overview over uncertainty contributions


Effect

Contribution

Image analysis
Simulation
Digitalization
Statistics
Determination of grey-scale values
Calibration of pixel size
Pixel noise
Sample preparation

Significant
Significant
Minor
Minor
Minor
Minor
Negligible
Unknown

of the threshold level used for image analysis relies. Beside general issues
related to Monte Carlo simulations, the main reason for this uncertainty
is the lack of knowledge about the diameter of the electron beam. Reasonable estimates range from 3 nm to 8 nm, and again the uncertainty is
estimated from this interval, whereas a value of 5 nm is assumed for image
evaluation.
Digitalization errors occur because round objects are projected onto
square pixels. Interpolating the image before size determination reduces
these effects.
Because the electron microscopic examination of many similar objects
is a tedious and time-consuming task, usually only a small number of
objects are analyzed and thus electron microscopy is often associated with
poor statistics. On the contrary, the uncertainty attributed to statistics
plays only a minor role if reference samples with narrow size distributions
are measured and if thousands of particles are evaluated in reasonable
times thanks to automatic image acquisition and analysis.
As stated in Section 5.3.2, the grey-scale values g1 and g0 in the background and in the particle center, respectively, must be determined in
order to convert the simulated signal at the particle boundary to a greyscale value required for thresholding the image. The determination may
be influenced by an inhomogeneous carbon background, drying residue,
and so on, resulting in a corresponding uncertainty.
Based on the calibration described in Section 5.1, the pixel size can be
determined quite accurately, thus contributing only to a minor extent to
the overall uncertainty. Pixel noise may lead to erroneous inclusion or
exclusion of noisy pixels at the particle boundary, and thus to faulty particle sizes. Compared with the other uncertainty sources, its effect can be
neglected.

TSEM: A Review of SEM in Transmission Mode

337

If the mean particle size of a particle ensemble present in a suspension


is to be determined, sample preparation should ensure that a representative subsample is present on the substrate. While there are no
obvious objections to the preparation technique used, its effect on the
size distribution, if any, has not yet been quantified. Thus the stated
uncertainties do not include any preparation effects.
To determine the overall uncertainty, all stated uncertainty contributions must be summarized quadratically, resulting in expanded uncertainties (coverage factor of k = 2, 95% probability coverage) associated with
the mean particle size of about one to a few nanometers. The stated uncertainty should not be confused with the statistical uncertainty component,
which is sometimes mentioned exclusively. The statistical component is
only one part of the overall uncertainty and, in our case, it is a rather
small one.

5.3.4. Measurement of Gold Particles and Comparison


with TEM Results
Three gold nanoparticle standards have been chosen as test samples,
which have been extensively studied by TEM (Kaiser and Watters, 2007)
at NIST. These samplesnamed RM8011, RM8012, and RM8013have
nominal diameters of 10 nm, 30 nm, and 60 nm, respectively, and allow
a comparison of our TSEM measurement results with results gained by
TEM.
Table 2 presents a summary of the TSEM results. The expanded
uncertainty of the mean diameter is as low as 1.2 nm for RM8011 and
RM8012. For the 60-nm particles (RM8013) the uncertainty almost doubles
to 2.3 nm. This is mainly due to a higher uncertainty contribution from
the image analysis that could possibly be reduced if more particles were
measured.
For comparison, the results of traceable TEM measurements are also
given in Table 2. The mean diameters determined by TEM are slightly
smaller than those measured by TSEM, but they are quite close and they
perfectly agree within the scope of the stated uncertainties. The differences
are only 0.2 nm for the 10-nm gold particles, 0.3 nm for the 30-nm particles and 1.2 nm for the 60-nm gold particles. The report of investigation
(Kaiser and Watters, 2007) states that for TEM [...] reference values were
calculated from the ampoule means and the uncertainty level is based on
a prediction interval approach (Neter et al., 1996), where the combined
uncertainty is calculated
as the standard deviation of the ampoule means
p
multiplied by 1 + 1/N (N is the number of ampoules analyzed) [. . .].
Hence, the uncertainty values stated for the TEM results contain only

338

Tobias Klein, Egbert Buhr, and Carl Georg Frase

TABLE 2 Results of TSEM measurements of three different gold particle samples and
comparison with NIST TEM data
TSEM
Name

Nominal
diameter (nm)
Mean particle
size (nm)
Expanded
uncertainty of the
mean size (nm)
Statistical standard
deviation of mean
size (nm)
Spread of size
distribution in nm
(standard
deviation)
Median size (nm)
Mode size (nm )
Number of analyzed
particles

TEM

RM8011 RM8012 RM8013

RM8011 RM8012 RM8013

10

30

60

10

30

60

9.1

27.9

57.2

8.9

27.6

56.0

1.2

1.2

2.3

0.1*

2.1*

0.5*

0.02

0.08

0.23

0.02

0.07

0.09

0.8

2.2

4.2

1.1

4.3

5.0

9.1
9.4
2318

27.7
28.1
747

56.6
55.8
325

8.8
8.8
5098

26.9
26.7
4364

55.4
55.8
3030

Solely based on statistics.

the statistical contribution, whereas the uncertainty analysis for TSEM


measurements also includes important systematic influences as discussed
in Section 5.3.3. Yet in the case of RM8012, the uncertainty associated with
TSEM is smaller than the one ascribed to TEM measurements.
The size distributions of the samples are quite similar, resembling a
Gaussian distribution with the addition of a small fraction of larger particles (Figure 24). Small but consistent differences can be seen by comparing
the size distributions determined by TSEM and TEM. Both size distributions have similar widths but differ in their absolute position on the size
axis. The shift ranges from 0.3 nm to 1.3 nm, with TSEM yielding slightly
larger size values than TEM.
Whereas the different portions of large particles measured by TSEM
and TEM may be an effect of subsampling or statistics, the small shift
between the size distributions measured by TEM and TSEM seems to
be systematic. Which distribution is closer to reality cannot be decided
based on present knowledge. However, the differences are small compared with the stated uncertainties and the results of both measurements
fit well within the scope of those uncertainties.

TSEM: A Review of SEM in Transmission Mode

Relative frequency

0.2

TEM
TSEM

RM8011

0.15

0.1

0.05

12
8
10
Particle size (nm)
(a)

14

16

TEM
TSEM

0.3 RM8012
Relative frequency

339

0.25
0.2
0.15
0.1
0.05
0

0.25

10

20
30
40
Particle size (nm)
(b)

50

60

TEM
TSEM

RM8013

Relative frequency

0.2
0.15
0.1
0.05
0
10

20

30

40
50
60
70
Particle size (nm)
(c)

80

90

FIGURE 24 Size distributions of three gold nanoparticle samples with nominal sizes of
10 nm (a), 30 nm (b), and 60 nm (c) as measured with TEM and TSEM.

340

Tobias Klein, Egbert Buhr, and Carl Georg Frase

6. CHARACTERIZATION OF DIFFERENT MATERIAL CLASSES


6.1. Biological Samples
From the very beginning, transmission detectors in SEMs were used to
examine biological specimens. Already in the first publication regarding
TSEM, Kimoto and Hashimoto (1968) showed a micrograph of a cotyledon
cell of a soybean seed to demonstrate the possibilities of the TSEM technique. Nemanic and Everhart (1973) used a conversion stub for the study
of biological samples demonstrating the resolution of an 8-nm membrane
in a mitochondrion. Swift et al. (1969) modified a Cambridge Stereoscan
for transmission electron detection and examined the internal structure
of keratin fibers (Swift, 1972a,b). Sample preparation was similar to TEM
preparationnamely, embedding in polymer, sectioning, and staining.
Oho et al. (1987a) used a commercial SEM capable of transmission imaging and steadily improved it (e.g., using an adjustable detector aperture).
Studying biological samples, a detailed comparison between TEM images
and micrographs taken with the improved TSEM instrument revealed
similar image qualities.
One of the main benefits of TSEM for biological applications is the
good contrast of light elements without the need for metal staining
(Takaoka and Hasegawa, 2006; Takaoka et al., 2004) due to larger scattering cross sections for light elements at low electron energy. However,
the use of low-energy electrons is a trade-off between higher resolution at
high energies but lower contrasts and higher contrasts at low energies but
reduced resolution. Takaoka and Hasegawa (2006) also discuss the influence of the support film in the imaging of biological samples. The granular
image of a carbon film overlaps the original image, thus degrading contrasts. Therefore, support-free imaging using, for example, microgrids is
desirable for background-free images with high contrast.
The relatively large specimen chamber compared with TEM allows
the integration of other instruments. Stemmer et al. (1991) fitted a scanning tunneling microscope (STM) in the specimen chamber of an SEM.
The STM was tilted at 45 to allow for simultaneous TSEM and STM
examination of biological structures on electron-transparent films.
One step along the route to 3D reconstructions of the mammalian brain
has been taken using TSEM (Liu and Yorston, 2010). Such reconstructions rely on a large number of ultramicrotome cross sections that must be
imaged at high resolution. Using TEM to obtain these micrographs takes
a tremendous amount of time. With the development of a new, powerful
software the task of acquiring, storing, and combining many images can
be automated. After setting up a measuring task, the instrument can work
unattended for days, generating combined images of whole tissues.
Hondow et al. (2011) established TSEM as a practical tool for
nanotoxicology. They studied the in vitro uptake of nanoparticles and
carbon nanotubes in cells. For reliable results it is necessary to distinguish

TSEM: A Review of SEM in Transmission Mode

341

nanomaterial inside the cell from material that is solely deposited on top
of the cell and also from preparation artifacts originating from microtome slicing. The combination of SE imaging using an in-lens detector
and transmission imaging proved to be an efficient method for this
differentiation.

6.2. Polymers
Besides biological specimens, TSEM is especially suited for the characterization of low-Z polymers due to increasing scattering cross sections
for low-energy electrons. This behavior enables high-contrast imaging
and enhances the capacity to differentiate between materials of similar
atomic composition. Consequently, the technique is successfully applied
to the study of polymers and rubber blends without the need for chemical
staining (Cudby, 1998; Guise et al., 2011). Lednicky et al. (2000) used electron energies of 5 keV to distinguish between individual components of
polymer blends differing in density by as little as 0.04 g/cm3 . Due to
the low-energy electrons, the preparation of sufficiently thin specimens
caused some difficulties.
TSEM is also beneficial for analytical X-ray mapping of polymer samples. Beam damage and charging effects are avoided as far as possible
(Brown and Westwood, 2003). Williams et al. (2005) used TSEM in an environmental SEM to provide both compositional and structural details of
thin films of semiconducting polymeric materials used in electronic applications. Environmental TSEM imaging in the liquid state has been applied
to observe surfactant layers absorbed on the surface of latex particles
(Faucheu et al., 2009).
In contrast to TEM, the large range of magnifications offered by TSEM
enables the study of both large-scale phenomena such as crack propagation using low magnifications and nanoscale morphology of polymer
composites using high magnifications (Guise et al., 2011). Together with
its flexibility in sample handling and the use of sample carousel systems,
TSEM is a practical and affordable alternative to the TEM analysis of
polymers.

6.3. Semiconductor
The small feature sizes in semiconductor devices require high-resolution
imaging for the inspection of the manufactured structures. A common
procedure uses FIB preparation of thin cross sections in a FIB/SEM followed by TEM imaging (Giannuzzi and Stevie, 1999). This approach
yields high-resolution cross-sectional images, but the effort to transfer
the sample to the TEM is time-consuming. Often the ultimate resolution
offered by TEM is not necessary, and in those cases, TSEM is a promising
and valuable alternative.

342

Tobias Klein, Egbert Buhr, and Carl Georg Frase

In this field of application, TSEM can use its capabilities to full extent
while being easier to use compared with TEM, resulting in an increased
throughput and a reduced cost per sample. TSEM offers an imaging
resolution that comes closer to TEM than ordinary SEM (Moore, 2003;
Vanderlinde, 2002). Instrument parameters, such as acceleration voltage
and detector acceptance angle, may be optimized to achieve high material
contrast. For example, Young et al. (2008) showed that the contrast at the
interface between two materials can be enhanced if the HAADF detector
is composed of several segments allowing differentiation between various
azimuthal angular ranges.
TSEM can easily be incorporated into an FIB/SEM, allowing in situ
investigations and making tedious sample transfer redundant. With the
help of a flipstage (Young et al., 2004) or of a special probe tip holder
(Kendrick et al., 2008), the orientation of the sample can be adapted for
either FIB milling or TSEM observation. Due to its pronounced thickness
dependence, TSEM can be used to monitor and control FIB processing
during sample preparation (Golla-Schindler, 2008). Furthermore, TSEM
enables direct thickness measurements based on appropriate model simulations and can be used to ensure thickness uniformity (Young et al.,
2008).
TSEM has been successfully applied to imaging, inspection, and failure analysis of ICs (Coyne, 2002; Gignac and Wells, 2011; Nakagawa
et al., 2002; Tracy, 2002); Figure 25 shows an example. Coyne et al. (2005)
observed structural changes, such as defects in the crystal lattice and
thermal-mechanical damage, which are induced by the micromachining
of wafer-grade silicon. Furthermore, high-resolution elemental analysis is
possible if TSEM is combined with EDX spectroscopy (Iannello and Tsung,
2005).

(a)

(b)

FIGURE 25 (a) The electron-transparent sample is ready for lift-off after FIB
preparation. (b) The BF TSEM micrograph of a faulty semiconductor structure clearly
reveals the defect as a dark spot (Gnauck, 2005). Images reprinted by permission of John
Wiley & Sons, Inc.

TSEM: A Review of SEM in Transmission Mode

343

6.4. Material Science


In addition to monitoring FIB thinning, the previously mentioned thickness and material dependence of TSEM may also be used for spatially
resolved sample thickness determination or material concentration measurements. Since TSEM signals areeven for thin samples of a few
nanometersthe result of multiple electron scattering, quantitative measurements require appropriate model calculations or calibration procedures using samples with known properties.
Extensive work on this subject has been performed by Merli and
co-workers in the first decade of this century. They used the TSEM technique to measure arsenic dopant profiles in silicon (Merli et al., 2002),
thereby demonstrating the verification of two monolayers of AlAs in
a GaAs matrix (Figure 26). They also analyzed TSEM contrast and lateral resolution in dependence on sample thickness (Morandi and Merli,
2007) and generalized their method for the study of biological samples
(Morandi et al., 2007).
Pfaff et al. (2010) used HAADF detection at electron energies below
30 keV to study electron scattering in amorphous carbon and carbonbased materials. Their experimental and theoretical findings quantitatively revealed the relationships between TSEM signal intensity, sample
thickness, and system parameters, such as electron energy and detector
acceptance angle. In addition to thin film samples, they also used wedgetype specimens to vary sample thickness in a wide range up to a few
hundred nanometers. The authors conclude that the method is capable
of determining specimen thickness with a precision of 10%, provided that
40 nm

250

GaAs

100 nm
GaAs

GaAs

Pixel value

20

200

t
10

150

200

40 nm 20 8 3 1 1ml
Al As 10 5 2 2 ml
5
3
2
2 ml
1

400

600

800

1000

nm

FIGURE 26 The specimen shown as an inset was analyzed using a conversion-type


transmission detector. It consists of thin layers of AlAs (40 nm to 1 monolayer)
sandwiched between slices of 100-nm GaAs. Layers as thin as two monolayers can be
verified by averaging over 200 scan lines (Merli et al., 2002). Image reprinted with
permission from American Institute of Physics. Copyright 2002.

344

Tobias Klein, Egbert Buhr, and Carl Georg Frase

the composition of the specimens is known. In another study (Volkenandt


et al., 2010), HAADF-TSEM was applied to quantify sample thickness and
indium concentration in InGaAs quantum wells. Here, the authors state
that thickness measurements can be performed with an accuracy of 5 nm
and that composition changes of 5% can be detected.
Acevedo-Reyes et al. (2008) studied the application of TSEM to characterize precipitates in microalloyed steels, concluding that TSEM is a useful
technique because it allows the analysis of large populations of precipitate
particles due to the high contrast of the micrographs. The application of
TSEM is restricted to cases where knowledge of the particles chemistry
is not required. The authors prefer HAADF-TEM over TSEM to resolve
the chemical composition of very small precipitates with a size of a few
nanometers because these precipitates are difficult to analyze by EDX.
TSEM has been used to investigate the morphology of thin films of aluminum and its alloys (Shimizu et al., 2004). The authors report that TSEM
enables the study of fine film features and near-surface metal regions
with resolutions similar to TEM. In contrast to SE imaging, coating of
non-conducting material is not required to avoid charging problems.
TSEM has also been demonstrated successfully in mineralogical applications as a quick and easy method for imaging submicrometer-sized crystals in rock samples or for characterizing fine-scale intergrowths (Lee and
Smith, 2006; Smith et al., 2006). Russias et al. (2008) used TSEM to study
calcium silicate hydrates, which are the main components of cement. They
compared TSEM with high-energy TEM for this application, reporting
that TSEM causes less beam damage and may be regarded as a low-dose
technique. However, TSEM could not provide information on the hydrate
crystal structure; for this purpose, TEM is the appropriate means.

7. SPECIAL IMAGING MODES


7.1. TSEM in Liquids
Transmission detectors may also be used in so-called environmental SEMs
(ESEM), which enable the investigation of samples in their liquid environment. This combination has been called wet STEM by Bogner et al. (2005).
The liquid layer must be thin enough to enable the transmission of electrons, which can be achieved by controlling the environmental conditions
in the ESEM chamber. The option of detecting the transmitted electrons
is a significant extension of the ESEM technique since the entire liquid
volume can be imagedthat is, objects underneath the surface of the
liquid are accessible and can be studied. This technique was used to investigate suspensions in their wet environment (Bogner et al., 2007, 2005). It is
also ideally suited to the study of dynamic processes at the nanoscale, such
as colloidal crystal formation (Moh et al., 2010) or condensation processes

TSEM: A Review of SEM in Transmission Mode

345

(Barkay, 2010). Stokes and Baken (2007) imaged vacuum-sensitive soft


matter nanomaterials in their native state.

7.2. Electron Energy-Loss Spectroscopy


Electron energy-loss spectroscopy (EELS) is a standard method for elemental analysis in TEM or STEM, featuring a very good spectral resolution
of less than 1 eV. In SEM, however, the standard method for elemental analysis is EDX spectroscopy. Its energy resolution of about 100 eV
is clearly inferior. Therefore, an EELS attachment for standard SEM in
transmission mode is a promising alternative to EDX. Luo, Kursheed, and
coworkers developed a miniaturized EELS attachment that fits into a standard SEM and works in transmission as well as in BSE mode (Khursheed
et al., 2003; Luo and Khursheed, 2006, 2008). The instrument is secondorder corrected for spherical aberration and has a spectral resolution of
about 4 eV. As a demonstration, the K-edge spectrum of an amorphous
carbon film was recorded.

7.3. Tomography
Electron tomography is applied in both life science (Koning and Koster,
2009) and material science (Kubel et al., 2005) to obtain high-resolution 3D
object information. Typically, TEMs are used for electron tomography due
to the ultimate resolution achievable, but also because the damage to biological objects is less severe at such high energies compared with electron
energies in the range of some ten kV. In material science, where crystalline
objects are often studied, STEM using HAADF detection is applied to
avoid artifacts due to electron diffraction effects and to exploit the strong
Z-contrast mechanism of HAADF detection.
High-voltage STEM images of low-Z materials have a weak object contrast. The application of low-energy electrons as used in SEM improves
the imaging of low-Z materials and thus may supersede the application
of staining procedures. Recently electron tomography using transmission detection in a SEM has been successfully applied to study proteins
(Furusho et al., 2009). Sample damage effects due to low-voltage electron
beam irradiation could be effectively reduced by cooling the sample down
to about 100 K.
Jornsanoh et al. (2011) introduced a new sample holder mounted to an
eucentric tilting stage for tomography in an ESEM. They highlight the usefulness of a tomography technique which is intermediate between X-ray
and STEM tomography with regard to both resolution and sample size.
They demonstrated the applicability of their device to study samples that
are difficult to image in STEM: filler particles incorporated into a polymer
(Figure 27).

346

Tobias Klein, Egbert Buhr, and Carl Georg Frase

1 m
1 m

1 m

FIGURE 27 Three-dimensional view of the filler particles inside a polymer sample


obtained by TSEM tomography (Jornsanoh et al., 2011). Image reprinted with kind
permission from Elsevier.

7.4. Visualization of Electric Fields


Sharp tips and protrusions such as carbon nanotubes or sharp emitter
tips can generate strong local electric fields. TSEM can be used for an
in situ visualization of these fields. Fujita et al. (2007a,b, 2008) placed
tungsten tips in a TSEM above the BF detector. They observed dark shadows around the apex of the biased tip because the electric field caused
a deflection of the primary electron beam, resulting in a loss of BF signal. The size and form of the shadow were recorded as a function of the
tip voltage. This technique is, for instance, useful to visualize the local
field enhancement behavior of ultrasharp tungsten tips used as electron
emitters.

8. CONCLUSION
The studies presented in this review demonstrate that SEM in transmission mode is a valuable technique that bridges the gap between SEM and
TEM. It combines the versatility of a SEM with the advantageous imaging
modalities using TEs. TSEM can replace TEM or STEM if atomic resolution
is not required, and it offers new imaging possibilities and applications
due to the usage of low-energy electrons.
Although the TSEM technique has been described and applied since
the early days of SEM, it was rather a side issue and gained increasing interest especially in the past decade when high-resolution SEMs
and reliable target preparation techniques using FIB became increasingly

TSEM: A Review of SEM in Transmission Mode

347

available. Besides the lower cost, the ease of handling and the high sample throughput of TSEM compared with TEM have encouraged its further
dissemination. Today TSEM does not intend to rival TEM and STEM in
terms of achieving highest resolution, but it is a good choice to reduce
workload by taking over less demanding tasks.
Since scattering cross sections of low-energy electrons are large, which
is advantageous especially for low-Z materials, TSEM yields high-contrast
images of, for example, polymers or biological samples. For instance, it is
possible to distinguish between different components in polymer blends
or to study biological samples without metal staining. Moreover, using
TSEM in an environmental SEM enables the investigation of suspensions in their wet environment and allows the study of dynamic colloid
processes at the nanoscale.
TSEM has strong potential particularly for the investigation of
nanoparticles since they are nanoscaled by nature and hence do not
require elaborate preparation techniques. It is possible to carry out accurate and traceable measurements of nanoparticle size and shape. The
accuracy of these measurements benefits from the fact that TSEM images
can be reliably modeled and simulated in acceptable time using modern
computers, thus enabling quantitative comparisons between experiment
and theory. In addition, due to the strong correlation between TSEM signal and sample thickness, TSEM is able to quantitatively measure film
thickness with lateral resolutions at the nanoscale.
The broad range of applications that are already visible today demonstrate that TSEM is a versatile technique that presumably will be used
with increased frequency and might become a standard methodfor
example, in material science and biology. Further developments in SEM
technology, such as continuing improvement of lateral resolution down
to the subnanometer range, will shorten the distance to TEM regarding
resolution and might make TSEM a serious competitor of high-voltage
instruments. Hence, the affirmative answer to the rhetoric question Is
STEM possible in a SEM? given by Joy and Maher (1976) stating that
TSEM will be a versatile instrument for many applications is still valid.
Many of the applications they had in mind have been demonstrated,
and further applications of TSEM will most probably become possible in
future.

LIST OF ABBREVIATIONS
1D
2D
3D
BF
BSE

One-dimensional
Two-dimensional
Three-dimensional
Bright-field
Backscattered electron

348

Tobias Klein, Egbert Buhr, and Carl Georg Frase

CCD
CSDA
DF
DLS
EDX
EELS
ESEM
FIB
GUM
HAADF
IC
MCSEM
MONSEL
NIST
PE
PTB
ROI
SAXS
SCALPEL
SE
SE1
SE2
SE3
SEM
SNR
STEM
STM
TE
TEM
TSEM
Z (low-Z)

Charge-coupled device
Continuous slowing-down approximation
Dark-field
Dynamic light scattering
Energy-dispersive X-ray spectroscopy
Electron energy-loss spectroscopy
Environmental Scanning Electron Microscope (Microscopy)
Focused ion beam (milling)
Guide to the expression of uncertainty in measurement
High-angle annular dark-field
Integrated circuit
Monte Carlo Simulation of Electron Microscopy
MONte carlo Simulation of secondary ELectrons
National Institute of Standards and Technology
Primary electron
Physikalisch-Technische Bundesanstalt
Region of interest
Small-angle X-ray scattering
SCattering with Angular Limitation Projection Electron
beam Lithography
Secondary electron
Secondary electron (excited by PE)
Secondary electron (excited by BSE in the sample)
Secondary electron (excited by BSE in the vacuum chamber)
Scanning electron microscope (microscopy)
Signal-to-noise ratio
Scanning transmission electron microscope (microscopy)
Scanning tunneling microscope (microscopy)
Transmitted electron
Transmission electron microscope (microscopy)
Transmission scanning electron microscope (microscopy),
SEM in transmission mode
(Material having a low) atomic number

REFERENCES
Acevedo-Reyes, D., Perez, M., Verdu, C., Bogner, A., & Epicier, T. (2008). Characterization of
precipitates size distribution: Validation of low-voltage STEM. Journal of Microscopy, 232,
112122.
Ayache, J., Beaunier, L., Boumendil, J., Ehret, G., & Laub, D. (2010). Sample Preparation
Handbook for Transmission Electron Microscopy: Techniques. Springer, Heidelberg.

TSEM: A Review of SEM in Transmission Mode

349

Barkay, Z. (2010). Dynamic study of nanodroplet nucleation and growth on self-supported


nanothick liquid films. Langmuir, 26, 957985.
Barkay, Z., Rivkin, I., & Margalit, R. (2009). Three-dimensional characterization of drugencapsulating particles using STEM detector in FEG-SEM. Micron, 40, 480485.
Berger, M., & Seltzer, S. (1964). Tables of energy losses and ranges of electrons and positrons.
In Studies in penetration of charged particles in matter (No. 39 in Nuclear Science Series).
National Academy of Sciences, Washingtion, DC.
Bethe, H. (1930). Zur Theorie des Durchgangs schneller Korpuskularstrahlen durch Materie.
Annalen der Physik, 5, 325400.
Bogner, A., Jouneau, P. H., Thollet, G., Basset, D., & Gauthier, C. (2007). A history of scanning
electron microscopy developments: Towards wet-STEM imaging. Micron, 38, 390401.
Bogner, A., Thollet, G., Basset, D., Jouneau, P. H., & Gauthier, C. (2005). Wet STEM: A new
development in environmental SEM for imaging nano-objects included in a liquid phase.
Ultramicroscopy, 104, 290301.
Broers, A. N. (1972). Observation of Fresnel fringes in the conventional scanning electron
microscope. Applied Physics Letters, 21, 499501.
Brown, G., & Westwood, A. (2003). Characterization of polymers and catalysts using scanning transmission electron microscopy (STEM) in a field emission SEM. Microscopy and
Microanalysis, 9, 10201021.
Browning, R., Li, T. Z., Chui, B., Ye, J., Pease, R. F. W., & Joy, D. C. (1994). Empirical forms
for the electron/atom elastic scattering cross sections from 0.1 to 30 keV. Journal of Applied
Physics, 76, 20162022.
Buhr, E., Michaelis, W., Diener, A., & Mirande, W. (2007). Multi-wavelength VIS/UV optical
diffractometer for high-accuracy calibration of nano-scale pitch standards. Measurement
Science and Technology, 18, 667674.
Buhr, E., Senftleben, N., Klein, T., Bergmann, D., Gnieser, D., Frase, C. G., & Bosse, H. (2009).
Characterization of nanoparticles by scanning electron microscopy in transmission mode.
Measurement Science and Technology, 20, 084025.
Busch, H. (1926). Berechnung der Bahn von Kathodenstrahlen im axialsymmetrischen
elektromagnetischen Felde. Annalen der Physik, 386, 974993.
Busch, H. (1927). Uber die Wirkungsweise der Konzentrierungsspule bei der Braunschen
Rohre. Electrical Engineering (Archiv fur Elektrotechnik), 18, 583594.
Cerrina, F. (1992). Recent advances in X-ray lithography. Japanese Journal of Applied Physics,
31, 41784184.
Chernoff, D. A., Buhr, E., Burkhead, D. L., & Diener, A. (2008). Picometer-scale accuracy in
pitch metrology by optical diffraction and atomic force microscopy. Proceedings of SPIE,
6922, 69223J.
Cowley, J. M. (1969). Image contrast in a transmission scanning electron microscope. Applied
Physics Letters, 15, 5859.
Coyne, E. (2002). A working method for adapting the (SEM) scanning electron microscope to
produce (STEM) scanning transmission electron microscope images. In ISTFA 2002: 28th
international symposium for testing and failure analysis (pp. 9399). Phoenix, Arizona.
Coyne, E., Magee, J. P., Mannion, P., OConnor, G. M., & Glynn, T. J. (2005). STEM (scanning
transmission electron microscopy) analysis of femtosecond laser pulse induced damage
to bulk silicon. Applied Physics A: Materials Science and Processing, 81, 371378.
Crawford, B. J., & Liley, C. R. W. (1970). A simple transmission stage using the standard collection system in the scanning electron microscope. Journal of Physics E: Scientific
Instruments, 3, 461462.
resolution. Journal of Molecular
Crewe, A. V., & Wall, J. (1970). A scanning microscope with 5 A
Biology, 48, 375393.
Crewe, A. V., Wall, J., & Langmore, J. (1970). Visibility of single atoms. Science, 168, 1338
1340.

350

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Crewe, A. V., Wall, J., & Welter, L. M. (1968). A high-resolution scanning transmission electron
microscope. Journal of Applied Physics, 39, 58615868.
Cudby, P. (1998). Characterization of vulcanized blends by microscopy. In A. J. Tinkler & K. P.
Jones (Eds.), Blends of natural rubber: Novel techniques for blending with speciality polymers
(pp. 2139). Chapman & Hall, London.
Czyzewski, Z., MacCallum, D., Romig, A., & Joy, D. C. (1990). Calculations of Mott scattering
cross section. Journal of Applied Physics, 68, 30663072.
Daniel, M. C., & Astruc, D. (2004). Gold nanoparticles: Assembly, supramolecular chemistry, quantum-size-related properties, and applications toward biology, catalysis, and
nanotechnology. Chemical Reviews, 104, 293346.
Demers H., Poirier-Demers N., Couture A. R., Joly D., Guilmain M., De Jonge N., & Drouin
D. (2011). Three-dimensional electron microscopy simulation with the CASINO Monte
Carlo software. Scanning. 33, 135146.
Farrow, R. C., Postek, M. T., Keery, W. J., Jones, S. N., Lowney, J. R., Blakey, M., . . . Hopkins,
L. C. (1997). Application of transmission electron detection to SCALPEL mask metrology.
Journal of Vacuum Science and Technology B: Microelectronics and Nanometer Structures, 15,
21672172.
Faucheu, J., Chazeau, L., Gauthier, C., Cavaille, J. Y., Goikoetxea, M., Minari, R., & Asua, J. M.
(2009). Latex imaging by environmental STEM: Application to the study of the surfactant
outcome in hybrid alkyd/acrylate systems. Langmuir, 25, 1025110258.
Frase, C. G., Buhr, E., & Dirscherl, K. (2007). CD characterization of nanostructures in SEM
metrology. Measurement Science and Technology, 18, 510520.
Frase, C. G., Gnieser, D., & Bosse, H. (2009). Model-based SEM for dimensional metrology
tasks in semiconductor and mask industry. Journal of Physics D: Applied Physics, 42, 183001.
Fujita, H. (1986). History of electron microscopes. In Commemoration of the 11th international
congress on electron microscopy, Kyoto, Japan.
Fujita, J., Ikeda, Y., Okada, S., Higashi, K., Nakasawa, S., Ishida, M., & Matsui, S. (2007a). In
situ visualization of local electric field in an ultrasharp tungsten emitter under a low voltage scanning transmission electron microscope. Journal of Vacuum Science and Technology
B: Microelectronics and Nanometer Structures, 25, 26242627.
Fujita, J., Ikeda, Y., Okada, S., Higashi, K., Nakazawa, S., Ishida, M., & Matsui, S. (2007b).
In-situ visualization of local field enhancement in an ultra sharp tungsten emitter under a
low voltage scanning transmission electron microscope. Japanese Journal of Applied Physics,
46, L498L501.
Fujita, J., Ikeda, Y., & Suzuki, I. (2008). Multilevel visualization of local electric field at
probe apex using scanning electron microscopy. Journal of Vacuum Science and Technology
B: Microelectronics and Nanometer Structures, 26, 20692072.
Furusho, H., Mishima, Y., Kameta, N., Yamane, M., Masuda, M., Asakawa, M., . . . Shimizu, T.
(2009). Lipid nanotube encapsulating method in low-energy scanning transmission electron microscopy analyses. Japanese Journal of Applied Physics, 48, 097001.
Gentsch, P., Gilde, H., & Reimer, L. (1974). Measurement of the top bottom effect in scanning
transmission electron microscopy of thick amorphous specimens. Journal of Microscopy,
100, 8192.
Giannuzzi, L. A., & Stevie, F. A. (1999). A review of focused ion beam milling techniques for
TEM specimen preparation. Micron, 30, 197204.
Gignac, L. M., & Wells, O. C. (2011). Characterization of semiconductor nanostructures by
scanning electron microscopy. In R. Haight, F. M. Ross, & J. B. Hannon (Eds.), Handbook of instrumentation and techniques for semiconductor nanostructure characterization. World
Scientific Publishing, Singapore.
Gleber, G., Cibik, L., Haas, S., Hoell, A., Muller, P., & Krumrey, M. (2010). Traceable size determination of PMMA nanoparticles based on small angle X-ray scattering (SAXS). Journal of
Physics: Conference Series, 247, 012027.

TSEM: A Review of SEM in Transmission Mode

351

Gnauck, P. (2005). Failure analysis and defect review using extended accuracy of the
CrossBeam technology. Vakuum in Forschung und Praxis, 17(S1), 58.
Gnieser, D., Frase, C. G., Bosse, H., & Tutsch, R. (2008). MCSEMa modular Monte Carlo
simulation program for various applications in SEM metrology and SEM photogrammetry. In M. Luysberg, K. Tillmann, & K. Weirich (Eds.), EMC 2008: 14th European microscopy
congress, Aachen, Germany (pp. 549550). Springer, Heidelberg.
Goldstein, J., Newbury, D. E., Joy, D. C., Echlin, P., Lyman, C. E., & Lifshin, E. (2003). Scanning
Electron Microscopy and X-Ray Microanalysis. Springer, New York.
Golla, U., Schindler, B., & Reimer, L. (1994). Contrast in the transmission mode of a lowvoltage scanning electron microscope. Journal of Microscopy, 173, 219225.
Golla-Schindler, U. (2008). Quantitative in situ thickness determination of FIB TEM lamella
by using STEM in a SEM. In M. Luysberg, K. Tillmann, & K. Weirich (Eds.), EMC 2008:
14th European microscopy congress, Aachen, Germany (pp. 667668). Springer, Heidelberg.
Goodhew, P. J., Humphreys, F. J., & Beanland, R. (2001). Electron Microscopy and Analysis.
Taylor & Francis, London.
Grillon, F. (2006). Low voltage contrast with an SEM transmission electron detector.
Microchimica Acta, 155, 157161.
Guise, O., Strom, C., & Preschilla, N. (2011). STEM-in-SEM method for morphology analysis
of polymer systems. Polymer, 52, 12781285.
Habicht, W., Behrens, S., Boukis, N., & Dinjus, E. (2001). Scanning transmission type imaging and analysis (EDX) of protein supported metallic nanoparticles. G.I.T. Imaging and
Microscopy, 1/2001, 4244.
Habicht, W., Behrens, S., Wu, J., Unger, E., & Dinjus, E. (2004). Characterization of metal
decorated protein templates by scanning electron/scanning force microscopy and microanalysis. Surface and Interface Analysis, 36, 720723.
Habicht, W., Behrens, S., Unger, E., & Dinjus, E. (2006). Cylindrical and ring-shaped tubulin
assemblies as metallization templates explored by FE-SEM/EDX and SFM. Surface and
Interface Analysis, 38, 194197.
Harriott, L. (1997). Scattering with angular limitation projection electron beam lithography
for suboptical lithography. Journal of Vacuum Science and Technology B: Microelectronics and
Nanometer Structures, 15, 21302135.
Hondow, N., Harrington, J., Brydson, R., Doak, S., Singh, N., Manshian, B., & Brown, A.
(2011). STEM mode in the SEM: A practical tool for nanotoxicology. Nanotoxicology, 5,
215227.
Humphreys, C. J. (1981). Fundamental concepts of STEM imaging. Ultramicroscopy, 7, 712.
Iannello, M. A., & Tsung, L. (2005). STEM role in failure analysis. Microelectronics and
Reliability, 45, 15261531.
Implementing Metrology in the European Research Area. (2008). T3.J1.1: Traceable characterisation of nanoparticles. Funded by the European Commission.
Joint Committee for Guides in Metrology. (2008a). Evaluation of measurement dataguide
to the expression of uncertainty in measurement (GUM).
Joint Committee for Guides in Metrology. (2008b). International vocabulary of metrology
basic and general concepts and associated terms (VIM).
Johnsen, K. P., Frase, C. G., Bosse, H., & Gnieser, D. (2010). SEM image modeling using the
modular Monte Carlo model MCSEM. Proceedings of SPIE, 7638, 76381O.
Jornsanoh, P., Thollet, G., Ferreira, J., Masenelli-Varlot, K., Gauthier, C., & Bogner, A.
(2011). Electron tomography combining ESEM and STEM: A new 3D imaging technique.
Ultramicroscopy, 111, 12471254.
Joy, D. C. (1987). A model for calculating secondary and backscattered electron yields. Journal
of Microscopy, 147, 5164.
Joy, D. C. (1991). An introduction to Monte Carlo simulations. Scanning Microscopy, 5, 329
337.

352

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Joy, D. C., & Luo, S. (1989). An empirical stopping power relationship for low-energy
electrons. Scanning, 11, 176180.
Joy, D. C., & Maher, D. M. (1976). Is STEM possible in a SEM? In Proceedings of the Workshop
on scanning electron microscopy (pp. 361368). Chicago, IL.
Kaiser, D. L., & Watters, R. L. (2007). Report of Investigation. Reference Material 8011.
Gold nanoparticles, nominal 10 nm diameter. National Institute of Standards and Technology, Gaithersburg, MD. Retrived from https://www-s.nist.gov/srmors/reports/8011
.pdf.
Kendrick, A. B., Moore, T. M., Zaykova-Feldman, L., Amador, G., & Hammer, M. (2008).
Cassette-based in-situ TEM sample inspection in the dual-beam FIB. Journal of Physics:
Conference Series, 126, 012082.
Khursheed, A., Karuppiah, N., Osterberg, M., & Thong, J. T. L. (2003). Add-on transmission attachments for the scanning electron microscope. Review of Scientific Instruments, 74,
134140.
Kimoto, S., & Hashimoto, H. (1968). On the contrast and resolution of the scanning electron microscope. In Proceedings of the workshop on scanning electron microscopy (pp. 6378).
Chicago, IL.
Klein, T., Buhr, E., Johnsen, K. P., & Frase, C. G. (2011). Traceable measurement of nanoparticle size using a scanning electron microscope in transmission mode (TSEM). Measurement
Science and Technology, 22, 094002.
Knoll, M., & Ruska, E. (1932). Das Elektronenmikroskop. Zeitschrift fur Physik A: Hadrons and
Nuclei, 78, 318339.
Koning, R., & Koster, A. (2009). Cryo-electron tomography in biology and medicine. Annals
of AnatomyAnatomischer Anzeiger, 191, 427445.
Kotera, M., Ijichi, R., Fujiwara, T., Suga, H., & Wittry, D. (1990). A simulation of electron
scattering in metals. Japanese Journal of Applied Physics, 29, 22772282.
Kotula, P. G. (2009). STEM in SEM for medium-resolution X-ray microanalysis. Microscopy
and Microanalysis, 15, 474475.
Krzyzanek, V., Nusse, H., & Reichelt, R. (2004). Quantitative microscopy with a highresolution FESEM in the transmission mode. In Proceedings of the 13th European microscopy
congress, volume 1: Instrumentation and methodology.
Krzyzanek, V., & Reichelt, R. (2009). Quantitative scanning electron microscopy in the
transmission mode: a development for nanoanalytics. In Microscopy conference 2009
(pp. 157158). Graz, Austria.
Kubel, C., Voigt, A., Schoenmakers, R., Otten, M., Su, D., Lee, T., . . . Bradley, J. (2005). Recent
advances in electron tomography: TEM and HAADF-STEM tomography for materials
science and semiconductor applications. Microscopy and Microanalysis, 11, 378400.
Laskin, A., & Cowin, J. P. (2001). Automated single-particle SEM/EDX analysis of submicrometer particles down to 0.1 m. Analytical Chemistry, 73, 10231029.
Laskin, A., Cowin, J. P., & Iedema, M. J. (2006). Analysis of individual environmental particles
using modern methods of electron microscopy and X-ray microanalysis. Journal of Electron
Spectroscopy and Related Phenomena, 150, 260274.
Lee, M. R., & Smith, C. L. (2006). Scanning transmission electron microscopy using a SEM:
Applications to mineralogy and petrology. Mineralogical Magazine, 70, 579590.
Lednicky, F., Coufalova, E., Hromadkova, J., Delong, A., & Kolarik, V. (2000). Low-voltage
TEM imaging of polymer blends. Polymer, 41, 49094914.
Liu, X., & Yorston, J. B. (2010). Automated extreme field of view low voltage multi-mode
STEM imaging of biological ultramicrotome cross sections with Atlas. Carl Zeiss NTS.
Application note.
Lowney, J. R. (1995a). MONSEL-II: Monte Carlo simulation of SEM signals for linewidth
metrology. Microbeam Analysis, 4, 131136.
Lowney, J. R. (1995b). Use of Monte Carlo modeling for interpreting scanning electron
microscope linewidth measurements. Scanning, 17, 281286.

TSEM: A Review of SEM in Transmission Mode

353

Lowney, J. R. (1996). Monte Carlo simulation of scanning electron microscope signals for
lithographic metrology. Scanning, 18, 301306.
Lowney, J. R., & Marx, E. (1994). Users Manual for the Program MONSEL-1: Monte Carlo
Simulation of SEM Signals for Linewidth Metrology. National Institute of Standards and
Technology, Gaithersburg, MD.
Luo, T., & Khursheed, A. (2006). Second-order aberration corrected electron energy loss spectroscopy attachment for scanning electron microscopes. Review of Scientific Instruments, 77,
043103.
Luo, T., & Khursheed, A. (2008). Elemental identification using transmitted and backscattered electrons in an SEM. Physics Procedia, 1, 155160.
Maggiore, C. J., & Rubin, I. B. (1973). Optimization of an SEM X-ray spectrometer system for
the identification and characterization of ultramicroscopic particles. In Proceedings of the
workshop on scanning electron microscopy (pp. 129136). Chicago, IL.
Matlab. (2009). Release 2009a. The MathWorks Inc., Natick, MA.
McKinney, W. R., & Hough, P. V. C. (1976). Simple SEM to dark field STEM conversion stub.
Technical report, Brookhaven National Laboratory, Upton, NY.
Merli, P. G., Corticelli, F., & Morandi, V. (2002). Images of dopant profiles in low-energy
scanning transmission electron microscopy. Applied Physics Letters, 81, 45354537.
Moh, K., Werner, U., Koch, M., & Veith, M. (2010). Silver nanoparticles with controlled dispersity and their assembly into superstructures. Advanced Engineering Materials, 12, 368373.
Moller, C. (1932). Zur Theorie des Durchgangs schneller Elektronen durch Materie. Annalen
der Physik, 406, 531585.
Moore, M. V. (2003). Using in situ scanning transmission electron microscopy to increase
resolution in a focused ion beam scanning electron microscope. Scanning, 25, 159160.
Morandi, V., & Merli, P. G. (2007). Contrast and resolution versus specimen thickness in low
energy scanning transmission electron microscopy. Journal of Applied Physics, 101, 114917.
Morandi, V., Merli, P. G., & Quaglino, D. (2007). Scanning electron microscopy of thinned
specimens: from multilayers to biological samples. Applied Physics Letters, 90, 163113.
Morikawa, A., Kamiya, C., Watanabe, S., Nakagawa, M., & Ishitani, T. (2006). Low-voltage
dark-field STEM imaging with optimum detection angle. Microscopy and Microanalysis,
12, 13681369.
Mott, N. F., & Massey, H. S. W. (1933). The Theory of Atomic Collisions. Claredon Press, Oxford,
UK.
Myklebust, R., Newbury, D., & Yakowitz, H. (1976). NBS Monte Carlo electron trajectory
calculation program. In Use of monte carlo calculations in electron probe microanalysis and
scanning electron microscopy (pp. 105128). Proceedings of a workshop held at the National
Bureau of Standards, Gaithersburg, MD.
Nakagawa, M., Dunne, R., Koike, H., Sato, M., Perez-Camacho, J. J., & Kennedy, B. J. (2002).
Low voltage FE-STEM for characterization of state-of-the-art silicon SRAM. Journal of
Electron Microscopy, 51, 5357.
Nemanic, M. K., & Everhart, T. E. (1973). The SEM to STEM conversion stub. In Proceedings
of the scanning electron microscopy symposium (pp. 2532). Chicago, IL.
Neter, J., Wasserman, W., Kutner, M., & Li, W. (1996). Applied Linear Statistical Models.
McGraw-Hill, New York.
Oatley, C. W., Nixon, W. C., & Pease, R. F. W. (1966). Scanning electron microscopy. Advances
in Electronics and Electron Physics, 21, 181247.
Oho, E., Baba, M., Baba, N., Muranaka, Y., Sasaki, T., Adachi, K., . . . Kanaya, K. (1987b).
The conversion of a field-emission scanning electron microscope to a high-resolution,
high-performance scanning transmission electron microscope, while maintaining original
functions. Journal of Electron Microscopy Technique, 6, 1530.
Oho, E., Sasaki, T., Adachi, K., Muranaka, Y., & Kanaya, K. (1987a). An inexpensive and
highly efficient device for observing a transmitted electron image in SEM. Journal of
Electron Microscopy Technique, 5, 5158.

354

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Pawley, J. (1992). LVSEM for high resolution topographic and density contrast imaging.
Advances in Electronics and Electron Physics, 83, 203274.
Peckerar, M. C., & Maldonado, J. R. (1993). X-ray lithographyAn overview. Proceedings of
the IEEE, 81, 12491274.
Pfaff, M., Muller, E., Klein, M. F. G., Colsmann, A., Lemmer, U., Krzyzanek, V., . . .
Gerthsen, D. (2010). Low-energy electron scattering in carbon-based materials analyzed
by scanning transmission electron microscopy and its application to sample thickness
determination. Journal of Microscopy, 234, 3139.
Pogany, A. P., & Turner, P. S. (1968). Reciprocity in electron diffraction and microscopy.
Acta Crystallographica Section A: Crystal Physics, Diffraction, Theoretical and General
Crystallography, 24, 103109.
Postek, M. T., Larrabee, R. D., & Keery, W. J. (1989). A new approach to accurate X-ray mask
measurements in a scanning electron microscope. IEEE Transactions on Electron Devices,
36, 24522457.
Postek, M. T., Larrabee, R. D., Keery, W. J., & Marx, E. (1991). Application of transmission
electron detection to x-ray mask calibrations and inspection. Proceedings of SPIE, 1464,
3547.
Postek, M. T., Lowney, J. R., Vladar, A. E., Keery, W. J., Marx, E., & Larrabee, R. D. (1993).
X-ray lithography mask metrology: Use of transmitted electrons in an SEM for linewidth
measurement. Journal of Research of the National Institute of Standards and Technology, 98,
415445.
Postek, M. T., & Vladar, A. (2011). Modeling for accurate dimensional scanning electron
microscope metrology: Then and now. Scanning, 33, 111125.
Prewitt, J. M., & Mendelsohn, M. L. (1965). The analysis of cell images. Annals of the New
York Academy of Sciences, 128, 10351053.
Probst, C., Gauvin, R., & Drew, R. A. L. (2007). Imaging of carbon nanotubes with
tin-palladium particles using STEM detector in a FE-SEM. Micron, 38, 402408.
Rasteiro, M. G., Lemos, C. C., & Vasquez, A. (2008). Nanoparticle characterization by PCS:
The analysis of bimodal distributions. Particulate Science and Technology, 26, 413437.
Reimer L., Kassens M., & Wiese L. (1996). Monte Carlo simulation program with a free configuration of specimen and detector geometries. Mikrochimica Acta, Supplement 13, 485492.
Reimer, L. (1998). Scanning Electron Microscopy: Physics of Image Formation and Microanalysis.
Springer, Heidelberg.
Reimer, L. (2008). Transmission Electron Microscopy: Physics of Image Formation. Springer,
Heidelberg.
Ritchie, N. (2005). A new Monte Carlo application for complex sample geometries. Surface
and Interface Analysis, 37, 10061011.
Russias, J., Frizon, F., Cau-Dit-Coumes, C., Malchere, A., Douillard, T., & JoussotDubien, C. (2008). Incorporation of aluminum into CSH structures: From synthesis to
nanostructural characterization. Journal of the American Ceramic Society, 91, 23372342.
Sadowski, T. E., Broadbridge, C. C., & Daponte, J. (2007). Comparison of common segmentation techniques applied to transmission electron microscopy images. Materials Research
Society Symposium Proceedings, 982, 2530.
Salvat, F., & Mayol, R. (1993). Elastic scattering of electrons and positrons by atoms.
Schrodinger and Dirac partial wave analysis. Computer Physics Communications, 74,
358374.
Scherzer, O. (1949). The theoretical resolution limit of the electron microscope. Journal of
Applied Physics, 20, 2029.
Seiler, H. (1983). Secondary electron emission in the scanning electron microscope. Journal of
Applied Physics, 54, R1R18.
Sezgin, M., & Sankur, B. (2004). Survey over image thresholding techniques and quantitative
performance evaluation. Journal of Electronic Imaging, 13, 146168.

TSEM: A Review of SEM in Transmission Mode

355

Shimizu, K., Fujitani, H., Habazaki, H., Skeldon, P., & Thompson, G. E. (2004). Examination
of surface films on aluminium and its alloys by low-voltage scanning and scanning
transmission electron microscopy. Corrosion Science, 46, 25492561.
Smith, C. L., Lee, M. R., & MacKenzie, M. (2006). New opportunities for nanomineralogy
using FIB, STEM/EDX and TEM. Microscopy and Analysis, 20, 1720.
Stemmer, A., Reichelt, R., Wyss, R., & Engel, A. (1991). Biological structures imaged in a
hybrid scanning transmission electron microscope and scanning tunneling microscope.
Ultramicroscopy, 35, 255264.
Stokes, D. J., & Baken, E. (2007). Electron microscopy of soft nanomaterials. Imaging and
Microscopy, 9, 1820.
Stroustrup, B. (2003). The C++ Programming Language. John Wiley & Sons, Hoboken, NJ.
Swift, J. A. (1972a). A further technique for the routine examination of keratin-fibre sections
by transmission scanning electron microscopy. Journal of the Textile Institute, 63, 129133.
Swift, J. A. (1972b). The routine examination of keratin-fibre internal structure by
transmission scanning electron microscopy. Journal of the Textile Institute, 63, 6472.
Swift, J. A., Brown, A. C., & Saxton, C. A. (1969). Scanning transmission electron microscopy
with the Cambridge Stereoscan Mk II. Journal of Physics E: Scientific Instruments, 2,
744746.
Takaoka, A., & Hasegawa, T. (2006). Observations of unstained biological specimens using
a low-energy, high-resolution STEM. Journal of Electron Microscopy, 55, 157163.
Takaoka, A., Hasegawa, T., Kuwae, A., & Mori, H. (2004). Low energy and high resolution
STEM for unstained or weak stained biological specimens. In Proceedings of the 13th
European microscopy congress (pp. 349350). Antwerp, Belgium.
Tracy, B. (2002). Materials analysis and process monitoring in MegaFabs. In ISTFA 2002: 28th
international symposium for testing and failure analysis (pp. 6975). Phoenix, Arizona.
Treacy, M. M. J., & Gibson, J. M. (1993). Coherence and multiple scattering in Z-contrast
images. Ultramicroscopy, 52, 3153.
Treacy, M. M. J., & Gibson, J. M. (1994). Erratum to Coherence and multiple scattering in
Z-contrast images. Ultramicroscopy, 54, 93.
Tuysuz, H., Lehmann, C. W., Bongard, H., Tesche, B., Schmidt, R., & Schuth, F. (2008). Direct
imaging of surface topology and pore system of ordered mesoporous silica (MCM-41,
SBA-15, and KIT-6) and nanocast metal oxides by high resolution scanning electron
microscopy. Journal of the American Chemical Society, 130, 1151011517.
Van Essen, C. G., Schulson, E. M., & Donaghay, R. H. (1970). Electron channelling patterns
from small (10 m) selected areas in the scanning electron microscope. Nature, 225,
847848.
Van Ngo, V., Hernandez, M., Roth, B., & Joy, D. C. (2007). STEM imaging of lattice fringes
and beyond in a UHR in-lens field-emission SEM. Microscopy Today, 15, 1216.
Vanderlinde, W. E. (2002). STEM (scanning transmission electron microscopy) in a SEM
(scanning electron microscope) for failure analysis and metrology. In ISTFA 2002:
international symposium for testing and failure analysis (pp. 7786). Phoenix, Arizona.
Vanderlinde, W. E., & Chernoff, D. (2005). X-ray nanoanalysis in the SEM. In ISTFA 2005:
international symposium for testing and failure analysis (pp. 370379). San Jose, California.
Villarrubia, J., Ritchie, N., & Lowney, J. (2007). Monte Carlo modeling of secondary electron
imaging in three dimensions. Proceedings of SPIE, 6518, 65180K.
Volkenandt, T., Muller, E., Hu, D. Z., Schaadt, D. M., & Gerthsen, D. (2010). Quantification
of sample thickness and In-concentration of InGaAs quantum wells by transmission measurements in a scanning electron microscope. Microscopy and Microanalysis, 16, 604613.
Von Ardenne, M. (1938a). Das Elektronen-Rastermikroskop. Praktische Ausfuhrung.
Zeitschrift fur technische Physik, 19, 404416.
Von Ardenne, M. (1938b). Das Elektronen-Rastermikroskop. Theoretische Grundlagen.
Zeitschrift fur Physik A: Hadrons and Nuclei, 109, 553572.

356

Tobias Klein, Egbert Buhr, and Carl Georg Frase

Von Laue, M. (1935). Die Fluoreszenzrontgenstrahlung von Einkristallen. Annalen der Physik,
415, 705746.
Wang, Z. (1995). Elastic and Inelastic Scattering in Electron Diffraction and Imaging. Plenum
Press, New York.
Waskiewicz, W. K. (1997). SCALPEL proof-of-concept system: Preliminary lithography
results. Proceedings of SPIE, 3048, 255.
Wells, O. C., & Bremer, C. G. (1970). Collector turret for scanning electron microscope. Review
of Scientific Instruments, 41, 10341037.
Wentzel, G. (1926). Zwei Bemerkungen uber die Zerstreuung korpuskularer Strahlen als
Beugungserscheinung. Zeitschrift fur Physik A: Hadrons and Nuclei, 40, 590593.
Williams, S. J., Morrison, D. E., Thiel, B. L., & Donald, A. M. (2005). Imaging of semiconducting polymer blend systems using environmental scanning electron microscopy and
environmental scanning transmission electron microscopy. Scanning, 27, 190198.
Wolpers, C. (1991). Electron microscopy in Berlin 19281945. Advances in Electronics and
Electron Physics, 81, 211229.
Woolf, R. J., Joy, D. C., & Tansley, D. W. (1972). A transmission stage for the scanning electron
microscope. Journal of Physics E: Scientific Instruments, 5, 230233.
Young, R. J., Buxbaum, A., Peterson, B., & Schampers, R. (2008). Applications of in-situ sample preparation and modeling of SEM-STEM imaging. In ISTFA 2008: 34th international
symposium for testing and failure analysis (pp. 320327). Portland, Oregon.
Young, R. J., Bernas, M. P., Moore, M. V., Wang, Y. C., Jordan, J. P., Schampers, R., & Van
Hees, I. (2004). In-situ sample preparation and high-resolution SEM-STEM analysis.
In ISTFA 2004: 30th international symposium for testing and failure analysis (pp. 331337).
Oregon, Massachusetts.

S-ar putea să vă placă și