Hawkes P.W., Spence J.C.H. (Eds.) Science of Microscopy. V.1 and 2

Подождите немного. Документ загружается.

Chapter 3 Scanning Electron Microscopy 245

5 Ultrahigh Vacuum Scanning Electron Microscopy in

Surface Science

Ultrahigh vacuum (UHV) scanning electron microscopy is from the

point of view of the pressure inside the specimen chamber the opposite

of SEM at elevated pressure. Since the conventional SEMs typically

work at high vacuum with a pressure of about 10

−4

Pa inside the speci-

men chamber, the elevated pressure SEMs are operated at pressures

six to seven orders of magnitude higher and the UHV SEMs about

three to four orders of magnitude lower than 10

−4

Pa. UHV is required

for most surface science experiments for two principal reasons:

1. To obtain atomically clean surfaces for studies and to maintain such

clean surfaces in a contamination-free state for the duration of the

experiment.

2. To permit the use of a low-energy electron technique and, in addition

to that, ion-based and scanning probe techniques without undue

interference from gas phase scattering.

One crucial factor in determining how long it takes to build up a

certain surface concentration of adsorbed species is the incident molec-

ular ﬂ ux F of gas molecules on surfaces given by the Hertz–Knudsen

equation

Fp mkT

= /( )2π

(5.1)

where m is the molecular mass, k is the Boltzmann constant, and T is the

temperature (K). To obtain the minimum estimate of time it will take

for a clean surface to become covered with a complete monolayer of

adsorbate, a sticking probability S = 1 and a monolayer coverage typi-

cally in the order of 10

–10

molecules/m

are assumed. Then the

minimum time per monolayer simply equals

10 2

⋅ ()/πmkT p

, i.e.,

the lower the pressure the longer the time for coverage. At a pressure of

about 10

−4

Pa forming a monolayer takes about 1 s.

Clean surfaces and UHV are required to apply very surface-sensitive

methods such as Auger electron spectroscopy (AES) and spectromi-

croscopy, X-ray- and UV-induced photoelectron spectroscopies, and

LVSEM a nd VLVSEM. SEM in UHV can be done with customized

FESEMs equipped with an energy analyzer for AES, chemical state

analysis, or trace element detection, respectively. These UHV-SEMs

enable highly sensitive surface area observation using high-resolution

SE imaging and are very well suited for EBSD observation (see Section

7) of electron chan neling patterns without contamination as well as

low accelerating voltage EDX analysis (see Section 6). An attached

neutralizing ion gun allows analysis of insulating bulk materials and

thin ﬁ lms without electrical charge ups.

SEM in UHV can also be performed with commercial surface science

UHV chamber systems and a UHV SEM column attached to it. Such

chamber systems offer an in situ combination of several imaging

methods (e.g., LVSEM, SAM, AFM, and STM), methods for chemical

analysis such as AES, and diffraction methods such as low-energy

246 R. Reichelt

electron diffraction (LEED). Dual-chamber systems have a large

chamber for the imaging and analysis techniques mentioned above

and a second chamber for specimen preparation such as ion sputtering,

heating, and thin ﬁ lm growth.

SEM in UHV combined with other imaging, spectroscopic, and dif-

fraction methods enables the true surface structure and chemistry to

be studied. In particular, LVSEM and VLVSEM are most powerful in

UHV because of their high surface sensitivity, which is not noticeably

degraded by irradiation-induced contamination.

6 Microanalysis in Scanning Electron Microscopy

The generation of X-rays due to electron–specimen interactions was

discussed in Section 2.2.5. The characteristic X-rays emitted from the

specimen carry information about its local element composition, which

is utilized as a powerful microanalytical tool combining SEM with

EDX and WDX spectrometers. X-Ray microanalysis is by far the most

widely used method combined with SEM, which enables in various

modes qualitative and quantitative element analysis from a point or

area of interest as well as mapping of the distribution of various ele-

ments simultaneously with SE and BSE imaging. The size of the inter-

action volume emitting X-rays is signiﬁ cantly larger than the ones for

AE, SE, and BSE because of the weaker absorption of X-rays inside the

specimen (cf. Figure 3–14) and the secondary emission by X-ray ﬂ uo-

rescence outside the electron interaction volume. The secondary X-ray

emission volume is much larger than that for primary X-ray emission

since X-rays are more penetrating than electrons having the same

energy. For electron-excited X-ray spectrometry performed on thick

specimens in the SEM, the range R for X-ray excitation is given accord-

ing to Kanaya and Kayama (1972) by

R [µm] = (0.0276 A/ρZ

0.89

) · (E

1.67

− E

1.67

) (6.1)

where A is the atomic weight (g/mol), ρ is the density (g/cm

), Z is the

atomic number, E

is the incident electron energy (keV), and E

is the

X-ray energy (keV).

For reliable quantitative X-ray microanalytical studies, X-ray absorp-

tion, X-ray ﬂ uorescence, and the fraction of backscattered electrons, all

of which depend on the composition of the specimen, have to be taken

into account and are performed by the so-called ZAF correction. Z

stands for atomic number, which affects the penetration of incident

electrons into the material, A for absorption of X-rays in the specimen

on the path to the detector, and F for ﬂ uorescence caused by other X-

rays generated in the specimen.

Two different types of detectors are used to measure the emitted X-

ray intensity as a function of the energy or wavelength. In an EDX

system (Figure 3–64) the X-rays enter the solid-state semiconductor

detector and create electron hole pairs that cause a pulse of current to

ﬂ ow through the detector circuit. The number of pairs produced by

each X-ray photon is proportional to its energy (see Section 2.1.3.1). In

a WDX spectrometer (Figure 3–65) the X-rays fall on a bent crystal and

Chapter 3 Scanning Electron Microscopy 247

Figure 3–64. Scheme of Si(Li) X-ray diode coupled to a ﬁ eld effect transistor

(FET) with a resistive feed back loop (R

, C

). The shape of the output signal is

shown in the output voltage vs. time diagram. Typically, this principle is used

in energy-dispersive X-ray (EDX) detectors. (From Reimer, 1985; with kind

permission of Springer-Verlag GmbH, Heidelberg, Germany.)

Si (Li) diode

Preamplifier

Time

Output

FET

20 nm Au

X-ray

Bias voltage

–1000 V

3 – 5 mm

+–

Figure 3–65. Principle of a wavelength-dispersive X-ray (WDX) spectrometer.

Generated X-rays that hit the analyzing crystal are focused and due to Bragg

reﬂ ection directed to a slit in front of the proportional counter lying on a

Rowland circle with radius r

. The lattice planes of the crystal are bent to a

radius of 2r

. (From Reimer, 1985; with kind permission of Springer-Verlag

GmbH, Heidelberg, Germany.)

Analysing crystal with lattice planes of radius r

Specimen

Rowland circle

of radius r

Proportional

counter

Slit

π – 2Θ

= 2r

248 R. Reichelt

are reﬂ ected only if they satisfy Bragg’s law. The crystal bending is such

that it focuses X-rays of one speciﬁ c wavelength onto a proportional

counter and rotates to scan the wavelength detected. Some important

features of both types of X-ray spectrometer are listed in Table 3–9.

However, since instrumentation and analysis of data in X-ray micro-

analysis are usually considered a separate discipline, no further details

are discussed in this section [see, e.g., Heinrich (1982), Heinrich and

Newbury (1991), Reimer (1998), Goldstein et al. (1984, 2003), and

Newbury and Bright (2005)].

A selected application of the very powerful combination of SEM

imaging, X-ray micro analysis, and element mapping—the latter was

invented almost exactly 50 years ago by Cosslett and Duncumb (1956)—

is illustrated in Figure 3–66. The selected specimen is a Cr–Fe alloy

with an Si phase, which has a locally varying composition as clearly

indicated by the energy-dispersive spectra in Figure 3–66a and b

recorded at different locations (Figure 3–66c).

The area under each characteristic peak represents the amount of

X-ray counts, which is—after subtraction of the unspeciﬁ c background

below the peak and ZAF correction—a direct quantitative measure of

the number of atoms of the speciﬁ c element belonging to that peak.

However, a simple visual inspection of the spectra shows, e.g., that the

location “Punkt1” (see Figure 3–66a) contains signiﬁ cantly more chro-

mium and less iron than location “Punkt2” (see the spectrum in Figure

3–66b). In addition, a strong silicon peak emerges in the spectrum of

location “Punkt2” not present in the spectrum of “Punkt1” (see the

spectrum in Figure 3–66a). The element distribution maps of four

important chemical elements in the specimen, namely iron, chromium,

silicon, and titanium, are shown in Figure 3–66d–g. Comparing the

information given by the four element distribution maps on the one

hand and the two spectra on the other hand immediately makes clear

why the titanium peak does not emerge in the spectra and the chro-

means of simple image processing procedures the SE micrograph

(Figure 3–66c) and the element distribution maps (Figure 3–66d–g) can

be superimposed in one image (Figure 3–66h) presenting information

for ﬁ ve individual images.

The most powerful tool in electron beam microanalysis is the ability

to depict the elemental compositional heterogeneity of matter with

micrometer to nanometer lateral resolution. Many developments have

occurred since the intervening years to advance this critical method.

We are now on the leading edge of extraordinary new X-ray mapping

performance: the emergence of the silicon drift detector (SDD) that

permits recording of X-ray spectrum images (XSI) in an energy-disper-

sive operating mode with output count rates of 600 kHz and even

higher (cf. Table 3–9). Further, computer-controlled SEM in connection

with image processing and EDX spectrometry enables the unattended

and automated determination of both the geometric parameters and

the chemical composition of thousands of individual particles down to

a size of 50–100 nm (see, e.g., Poelt et al., 2002). Consequently, correla-

tions between particle size, chemical composition, the number of

mium peak is dominant at location “Punkt1” but not at “Punkt2”. By

Chapter 3 Scanning Electron Microscopy 249

Table 3 –9. Characteristic features of energy-dispersive X-ray (EDX) and wavelength-dispersive X-ray (WDX) detectors.

Detector

Energy dispersive

Energy dispersive Energy dispersive Energy dispersive Si microcalorimeter Wavelength

Features Si(Li) X-ray detector

HPGe X-ray detector

drift (SD) X-ray X-ray detector

dispersive X-ray

detector

c,d,e

detector

Geometric collection <2% <2% <2% ª8 ¥ 10

−3

% <0.2%

efﬁ ciency

Quantum efﬁ ciency ª100% ª100% >90% 60–80% £30%

(within 1–2 keV)

Element detection Z ≥ 11 (Be window) Z ≥ 4 Z ≥ 5 Z ≥ 6

Z ≥ 4

Z ≥ 4 (windowless)

Energy resolution 150 eV (at 5.9 keV) 133 eV (at 10

cps) 150 eV (at 10

cps) 12 eV (at 5.9 keV) ª5 eV

115 eV (at 10

cps) 230 eV (at 6 ¥ 10

cps) 3.1 eV (at 1.49 keV)

Maximum counting rate 3 ¥ 10

cps 3 ¥ 10

cps ª6 ¥ 10

cps 8 ¥ 10

cps–10

cps 10

cps

Spectrum acquisition All energies All energies All energies All energies One wavelength

simultaneously simultaneously simultaneously simultaneously at a time

Probe current (nA) 0.1–20 0.1–10 0.1–10 0.1–10 1–100

Si(Li), Lithium drifted silicon; HPGe, high-purity germanium; cps: counts per second.

Reichelt (1995).

Strüder et al. (1998a).

Strüder et al. (1998b).

Lechner et al. (2001).

Newbury et al. (1999).

With BN/Moxtek window, thermal isolation, and Au-absorber.

Figure 3–66. X-Ray microanalysis of a Cr–Fe-alloy with a Si phase. The EDX spectra (a and b) were

recorded with the Röntec XFlah3001 from locations “Punkt1” and “Punkt2” marked in the SE micro-

graph of the specimen (c). The positions of the characteristic X-ray energies for the various elements

emerging in the spectra are indicated by thin colored lines, which are labeled with the chemical

symbol of the corresponding chemical element. The elements iron, chromium, and vanadium occur

with one K

peak each in the energy range from 4.95 to 6.40 keV and with one less intense L

peak

each in the energy range from 0.51 to 0.71 keV. Elemental distribution maps of Fe (d), Cr (e), Si (f), and

Ti (g) were recorded using the K

lines. (h) Mixed micrograph obtained by superimposition of the SE

image and the maps of the distribution of Fe, Cr, Si, and Ti within the ﬁ eld of view. Experimental

conditions: SEM, LEO 438VP. For recording spectra: E

= 20 keV; count rate, ≈ 3 × 10

cps; acquisition

time, 300 s. For recording maps: E

= 25 keV; count rate ≈ 1.5 × 10

cps; acquisition time 600 s. (EDX

spectra, SE micro graph, and elemental distribution maps were kindly provided by Röntec GmbH,

Berlin, Germany.) (See color plate.)

x 1E3 Pulses/eV

4.0

3.0

2.0

1.0

0.0

- keV -

810

x 1E3 Pulses/eV

3.0

2.0

1.0

0.0

246

- keV -

810

Chapter 3 Scanning Electron Microscopy 251

Figure 3–66. Continued

252 R. Reichelt

different compounds, and their contribution to the overall concentra-

tion can be established. Problems may arise in connection with speci-

men preparation, optimization of the image contrast, the sometimes

nonhomogeneous composition of particles, shadowing of the X-rays in

case of large particles, and the lack of a rigorous ZAF correction pro-

cedure for particles of arbitrary shape.

Another increasingly important application is multilayer analysis,

i.e., the nondestructive measurement of the thickness and composition

of thin ﬁ lms both unsupported and on substrates, which can be per-

formed with high accuracy down to a thickness of 2 nm by a combina-

tion SEM and EDX (see, e.g., Hu and Pan, 2001; Rickerby and Thiot,

1994; Rickerby et al., 1998; Smith et al., 1995). For this purpose the ratio

between the X-ray intensity of the ﬁ lm and the intensity of the same

element of a bulk standard is used.

Low-voltage scanning electron microscopy (E

< 5 keV) offers both

high spatial resolution and a signiﬁ cantly reduced X-ray generation

depth. This enables the composition of thin layers on substrates to be

determined without the need to use a dedicated thin ﬁ lm analysis

program. The analysis of small phases down to a size of 50 nm is also

possible (see, e.g., Poelt, 2000; Wurster, 1997). However, the disadvan-

tages are that the X-ray intensity produced is small because of low ﬂ uo-

rescence yield ω (cf. Figure 3–22) and not all elements can be analyzed.

In principle elements of higher atomic number can be identiﬁ ed using

L- and M-shell X-rays, but these are much more complicated than the

rather simple K-shell X-ray emissions. Although the enhanced

surface sensitivity may be useful sometimes, in many cases it is more

likely to be a problem. If thin ﬁ lm coating is required this ﬁ lm

may be a signiﬁ cant fraction of the electron range, and without it the

specimen may charge up. Working at E

= E

, where incoming

and emitted electrons are balanced, is usually not sufﬁ cient. Further-

more, having the electron beam stationary on the specimen for a long

time to integrate the weak signals is exactly the way to produce con-

tamination, which can stop the electrons before they reach the real

specimen surface. Then X-ray microanalysis does not relate to the

specimen at all.

7 Crystal Structure Analysis by Electron Backscatter

Diffraction

In crystalline specimens electrons are diffracted at lattice planes

according to Bragg’s law given as

2d sin ϑ = nλ (7.1)

where d is the lattice-plane spacing, ϑ is Bragg’s angle, and λ is the

electron wavelength. Bragg’s law requires that the incident and emer-

gent angles should be equal ϑ. As previously mentioned, the backscat-

tering coefﬁ cient sensitively depends on the tilt of the incident electron

beam relative to the lattice and Bragg position. Changing the tilt of the

Chapter 3 Scanning Electron Microscopy 253

incident beam relative to the lattice, e.g., by rocking the electron beam

or tilting the specimen, affects the backscattering coefﬁ cient, which

results in an electron channeling pattern (ECP). To obtain from an ECP

information about the crystal structure, e.g., the crystal orientation and

the lattice plane spacings, a so-called panorama diagram recorded by

successively tilting the specimen over a large angular range is required

(Joy et al., 1982; van Essen et al., 1971; Reimer, 1998). However, establish-

ing a panorama diagram is a somewhat difﬁ cult task.

Another way of obtaining information about the crystal structure of

the specimen is the use of electron diffraction effects associated with

the scattered electrons. As described in Section 2.2, the beam electrons

are scattered elastically and inelastically due to the electron–specimen

interaction, thus scattered beam electrons travel within the excitation

volume in all directions (cf. Figure 3–13). This scattering process can

be considered as a small electron source inside the crystalline speci-

men emitting electrons in all directions as shown in Figure 3–67. These

electrons may be diffracted at sets of parallel lattice planes according

Electron beam

Crystal surface

Lattice planes

Cone 1

Cone 2

2ϑ

Figure 3–67. Scheme of the formation of one pair of cones from diffraction of

scattered electrons at one set of parallel lattice planes.

254 R. Reichelt

to Bragg’s law [(Eq. (7.1)] as ﬁ rst observed by Venables and Harland

(1973). Figure 3–67 illustrates that the electrons emitted from one point

and diffracted at lattice planes will form pairs of cones centered with

respect to the normal vector of these lattice planes. The opening angle

of the cones is 180° −2ϑ and the angle between them is 2ϑ. Since the

Bragg angle is in the order of 1° for 15–30 keV electrons and lattice-

plane spacings of 0.2–0.3 nm, the intersections of the cones with a ﬂ at

observation screen positioned at a distance much large than d (usually

2–4 cm from the point of beam–specimen intersection) and tangential

to the propagation sphere of the scattered electrons are almost straight

and parallel pairs of lines. The are called Kikuchi lines. The angle

between parallel pairs of Kikuchi lines is 2ϑ. The whole EBSD pattern

with the Kikuchi lines reveals local information about the crystalline

structure within the individual excitation volume. A detailed treat-

ment of EBSD, which is beyond the scope of this section, is given by

Wilkinson and Hirsch (1997). Nevertheless, it is worth mentioning that

the relationship between EBSD and the previously introduced ECP is

characterized by the reciprocity of their ray diagrams (Reimer, 1985).

The recording of the EBSD pattern, which originally was performed

by exposure of a photographic ﬁ lm, is now done by position-

sensitive detectors such as a phosphorous screen coupled to a TV

camera or a scintillation window with a CCD camera attached. To

allow the diffracted electrons to escape from the specimen its surface

is usually tilted approximately 60° or 70° toward the screen. The EBSD

patterns are digitally acquired by computer, which also controls the

positioning of the beam. In case of automated crystal orientation

mapping, the computer scans the electron beam stepwise across the

specimen and controls the dynamic focusing of the beam (cf. Figure

3–9e and Section 2.1.5.2). Since special software algorithms for the

automated detection and indexing of Kikuchi lines in EBSD pattern

were introduced by Krieger-Lassen et al. (1992), Adams et al. (1993)

developed a new scanning technique called “orientation imaging micro-

scopy” (OIM). In OIM, which is also called automated crystal orienta-

tion mapping, the EBSD pattern from each individual point at the

specimen surface radiated by the electron beam is recorded and ana-

lyzed. The current state of the art of OIM is reviewed by Schwarzer

(1997) and Schwartz et al. (2000).

In the past decade EBSD has become a powerful tool for crystallo-

graphic analysis such as the determination of the orientation of indi-

vidual crystallites of polycrystalline materials in the SEM, phase

identiﬁ cation, and characterization of grain boundaries, which is illus-

trated by the following examples. Figure 3–68 shows two different

EBSD patterns from an as-cast niobium sample recorded with high

(Figure 3–68a) and low resolution (Figure 3–68b), respectively (Zaef-

ferer, 2004), at a stationary beam position. Figure 3–68c shows the

colored pairs of Kikuchi lines generated by automatic indexing and

overlayed to the EBSD monitored in Figure 3–68a.

Figure 3–69a shows an SE micrograph of polycrystalline austenite

with the related color-coded orientation map of polycrystalline austen-

ite monitored in Figure 3–69b (Zaefferer et al., 2004). The spatial resolu-