Venables J. Introduction to Surface and Thin Film Processes
Подождите немного. Документ загружается.
substantially developed in an evolutionary sense, year by year. By contrast, the scanned
probe techniques burst upon the scene in the early 1980s, in the revolutionary develop-
ment of first scanning tunneling microscopy (STM) (Binnig et al. 1982), followed in
quick succession by atomic force microscopy (AFM), scanning near-field optical
microscopy (SNOM), and related spectroscopies. The first two techniques as illustrated
in figure 3.2.
The central feature of STM operation is the tip, which has to be brought extremely
close to the sample in a controlled manner to effect tunneling, as shown schematically
in figure 3.2(a). In the simplest mode of operation, the z-motion is used to keep the
tunneling current I
T
constant. In addition, one has to be able to move the tip and
sample relative to each other, both as a shift to find out where you are, and as a scan in
x and y to produce the image. There are many different STM designs, but all success-
ful designs have been based on piezo-electric elements and have paid due regard for
design symmetry, which is necessary to minimize thermal drift.
A particularly appealing design is the ‘beetle’ STM developed by Besocke (1987) and
Frohn et al. (1989) in which the sample rests on three piezo-tube ‘legs’ and probes the
sample with the tip mounted on a piezo-tube ‘feeler’. This design is shown in figure
3.2(b) in the version developed by Voigtländer & Zinner (1993) and Voigtländer (1999)
for in situ deposition experiments. Coarse approach of the sample is effected by a
special holder, in which a rotational motion is translated via a shallow ramp into z-
motion; once the sample and tip can ‘feel’ each other, the feeler piezo takes over and
STM proper can begin. Coarse movement in x and y uses the leg piezo drives in stick-
slip motion; a fast jerk on the legs causes them to slip and the stage to move relative to
the leg and tip, but a slow movement translates the stage and legs together. By repeated
alternating stick and slip motions, stage translation can be made remarkably reprodu-
cible; the design will work either way up, though not on its side; it uses gravity. Either
the sample holder or the tip holder assembly can be readily withdrawn for sample prep-
aration.
The AFM also comes in many forms, and has the great advantage that it can be used
on insulators as well as conductors. A key element here was the development of sensi-
tive cantilever arms, whose deflection is typically monitored by a low powered He–Ne
laser reflected onto a position sensitive diode array detector, as shown in figure 3.2(c)
(Meyer & Amer 1988). These arms are usually made of lithographically etched silicon,
with silicon nitride (Si
3
N
4
) as the tip material. Such an arm will have a characteristic
resonant frequency, so that, in addition to steady (d.c.) measurements of tip displace-
ment, many a.c. and phase sensitive measurement schemes are possible. Figure 3.2(d)
shows a close up SEM view of such a Si
3
N
4
tip (Albrecht et al. 1990).
Of the many recent books on scanned probe microscopy, arguably the best to start
from are Chen (1993) and Wiesendanger (1994). There are several multi-author texts,
including Stroscio & Kaiser (1993) and many review and specialist articles. Indeed,
there are now a large number of techniques for studying surfaces on a microscopic
scale: a description of these techniques and their applications would take a very long
time. It is not possible to do justice to the full range of extraordinary possibilities
offered by these techniques here, but several examples are given throughout the book
which show how valuable they are in particular cases.
3.1 Classification of surface and microscopy techniques 67
68 3 Electron-based techniques
Figure 3.2. Scanned probe techniques: (a) principles of STM operation, indicating x, y and
z piezo-elements P
x
, P
y
and P
z
and contrast due to steps and electronic effects (after Binnig
et al. 1982); (b) the ‘beetle’ STM design of Besocke (1987) and Frohn et al. (1989), as used for
in situ deposition experiments by Voigtländer & Zinner (1993) and Voigtländer (1999);
P
x
P
y
P
z
scan
line
scan
line
Contrast due to
steps and electronic patches
Contrast due to
steps and electronic patches
z
I
T
I
Control
unit
Control
unit
(a)
(a)
(b)
(c)
3.1.4 Acronyms
Acronyms are defined at various places throughout this book and are summarized in
Appendix B. So far we have met and defined in the text: vacuum and electronics terms
UHV, RGA, QMS, TSP, SNR; surface and crystal growth terms ML, TLK, BCF, CVD,
MBE and others in section 2.5; diffraction techniques LEED, RHEED, THEED;
chemical analysis and ion scattering techniques AES, SIMS, ICISS; microscopy types
3.1 Classification of surface and microscopy techniques 69
Figure 3.2. (cont.) (c) a widely used AFM design, with a He–Ne laser to detect the deflection
of the cantilever on which the tip is mounted (after Meyer & Amer 1988); (d) close up SEM
view of a Si
3
N
4
AFM tip with a nominal tip radius ,30 nm, which has achieved atomic
resolution (after Albrecht et al. 1990; diagrams reproduced or redrawn with permission).
(d)
TEM, REM, LEEM, PEEM, SEM, STEM, STM, AFM, SNOM. We also have two
techniques (EELS, CBED) indicated on figure 3.1, which have not yet been defined.
Now is a good time to check you know what these acronyms stand for, since we will be
adding more to the list as we start to study individual techniques in more detail. In the
following sections, we emphasize a common case, where electrons are both the probe
and the detected (response) particle.
3.2 Diffraction and quasi-elastic scattering techniques
3.2.1 LEED
The common electron-based diffraction techniques are LEED and RHEED. As with
all surface diffraction techniques, the analysis is based in terms of the surface recipro-
cal lattice. An important aspect of diffraction from 2D structures is that the compo-
nent of the wave vector k, parallel to the surface k
//
is conserved to within a surface
reciprocal lattice vector G
//
, whereas the perpendicular component k
⬜
is not. This leads
to the idea of reciprocal lattice rods; they express the fact that k
⬜
can have any value,
so that diffraction can take place at any angle of incidence. However, the intensity of
diffraction is typically not constant at all values of k
⬜
, but is modulated in ways which
reflect the partial 3D character of the diffraction (Lüth 1993/5 section 4.2, Woodruff
& Delchar 1986 section 2.3).
The equipment for both diffraction techniques is simple, involving a fluorescent
screen, with energy filtering in addition in the case of LEED, as indicated in figure 3.3,
to remove inelastically scattered electrons. There are three types of LEED apparatus
in regular use. The normal-view arrangement has the LEED gun and screen mounted
on a UHV flange, typically 8 inches (200mm) across, and the pattern is viewed past the
sample, which therefore has to be reasonably small, or it will obscure the view. Most
new systems are of the reverse-view type, where the gun has been miniaturized, and the
pattern is viewed through a transmission screen and a viewport. This enables larger
sample holders to be used, which helps for such operations as heating, cooling, strain-
ing, etc. The third, and potentially most powerful, technique is where, in addition to
viewing the screen, the LEED beams can be scanned over a fine detector using electro-
static deflectors and focusing, in order to examine the spot profiles, which can be sen-
sitive to surface steps and other forms of disorder at surfaces. This technique, which
has been perfected by the Hannover group (Henzler 1977, 1997, Scheithauer et al.
1986, Wollschläger 1995), is now known as SPA-LEED, emphasizing the capability for
spot profile analysis.
There are two aspects to electron diffraction techniques. The first, and simplest, is
that the positions of the spots give the symmetry and size of the unit mesh, i.e. the
surface unit cell. The common use of electron diffraction is primarily, often solely, in
this sense. The second effect is that the positions of atoms in the mesh is not determined
by this qualitative pattern (see discussion in section 1.4), but requires a quantitative
analysis of LEED intensities. Application of dynamic theory has so far ‘solved’ several
70 3 Electron-based techniques
hundred surface structures (Watson et al. 1996). This is impressive, but it pales besides
the number of bulk structures solved by X-rays, using (developments of) kinematic
theory.
Experimentally, the intensities are typically collected in the form of so-called I–V or
I(E) curves, where the size of the Ewald sphere is varied by varying the probe energy
(say from 20–150 eV), and the intensity data are obtained by ‘tracking’ along an indi-
vidual reciprocal lattice rod. Various computer controlled, and frame grabbing,
schemes have been developed to do this. A description of standard experimental and
theoretical methods is given by Clarke (1985); useful updates are reviews by Heinz
(1994, 1995). Lüth (1993/5, chapter 4) or Prutton (1994, chapter 3) are other starting
points for LEED, as they introduce the basic idea of multiple scattering in a relatively
short space.
The difference between LEED and X-ray structure analysis is that a kinematic
diffraction theory has limited usefulness, because the scattering is very strong, as
explored in problem 3.2. Averaging different I–V curves at constant momentum trans-
fer was once a promising method in the attempt to get around this problem, and some
successes were recorded, particularly in obtaining the distances between lattice planes
parallel to the surface, and surface vibration amplitudes (Webb & Lagally 1973,
Lagally 1975). However, dynamical theory is constantly being developed, e.g. via adop-
tion of the latest computational and approximation methods, which are closely related
to band theory and have similar constraints (Pendry 1994, 1997); LEED is still the
main method of surface structure analysis, now complemented by surface X-ray
diffraction using synchrotron radiation (Feidenhans’l 1989, Johnson 1991, Robinson
& Tweet 1992, Renaud 1998).
3.2 Diffraction and quasi-elastic scattering techniques 71
Figure 3.3. LEED apparatus types, illustrating schematically the configurations of normal and
reverse view LEED, and spot profile analysis. The 15 kV is applied to a fluorescent screen,
which for reverse view must be transparent (after Bauer 1975, redrawn with permission).
electron gun
specimen
amplifier
Normal view
E/e
E/e
–E/e
–E/e
+5 kV
Spot profile
analysis
Reverse view
3.2.2 RHEED and THEED
The basis of RHEED is very similar in LEED, but the language used is rather different,
being similar to that used for THEED. The discussion can again be separated into
geometry and intensities. The glancing angle geometry of RHEED means that the
reciprocal lattice rods are closely parallel to the Ewald sphere near the origin, as shown
in figure 3.4. This means that the low angle region often consists of streaks, rather than
spots. This part of the pattern corresponds to the zero order Laue zone (ZOLZ) in
THEED patterns. The higher angle parts of a RHEED pattern are then equivalent to
higher order (HOLZ) rings in THEED patterns.
The apparatus for RHEED can consist of a simple 5–20 kV electrostatically focused
gun, for instance to monitor the surface crystallography in an MBE experiment, where
the glancing angle geometry has many practical advantages over LEED, especially in
the ease of access around the sample. Or it can utilize an electron gun approaching elec-
tron microscope quality, operating at higher voltages, and produce finely focused
diffraction spots over a wide angular range.
Several workers have perfected this technique especially in Japan (Ino 1977, 1988,
Sakamoto 1988, Ichikawa & Doi 1988), and two examples of Si(111) 737 in different
azimuths are seen in figure 3.5. Ino’s group in particular have also developed a curved
screen, centered on the sample, which could be viewed in two directions at right angles.
In the normal view, from the right of figure 3.4, we see spots distributed on a series of
arcs as shown for the Si(111)冑33冑3R30° Ag structure in figure 3.6(a); the same screen,
viewed via a mirror in the perpendicular direction along the reciprocal lattice rods is
seen to be an undistorted view of the reciprocal lattice (as in LEED) in figure 3.6(b).
We can note that the (111)737 and (111)冑3 patterns are strikingly different in a qual-
itative sense.
As in LEED, the question of intensities is much more detailed, involving multiple
scattering and inelastic processes, and there are many discussions/assertions in the lit-
erature about whether streaks or spots constitute evidence for good (i.e. well prepared,
flat) surfaces. Some general remarks are made in the next section.
72 3 Electron-based techniques
Figure 3.4. RHEED geometry in (a) reciprocal space and (b) real space.
k9
k
C
O
g
(b)
Reciprocal space
Real space
Electron gun
Sample
Screen
(a)
3.2 Diffraction and quasi-elastic scattering techniques 73
Figure 3.5. RHEED patterns (20 kV) of Si(111) with the 737 reconstruction along: (a) [1
¯
21
¯
]
and (b) [011
¯
] incidence. Note the reciprocal lattice unit cell O9ACB, and the six superlattice
spots in each direction between these fundamental spots (from Ino 1977, reproduced with
permission).
3.2.3 Elastic, quasi-elastic and inelastic scattering
Models of LEED and RHEED concentrate on elastic scattering, where the energy of
the outgoing electron is the same as that of the incoming electron. But experimentally
we cannot discriminate in energy very well in a typical diffraction apparatus. LEED
grids/screens are able to remove plasma loss electrons (⬃10–20 eV loss), but the inten-
sities measured include phonon scattering (⬃25 meV losses and gains). This is thermal
diffuse scattering, and is accounted for in the models using a Debye–Waller factor, as
in standard X-ray treatments. At higher temperature, the intensities in the Bragg peaks
fall off exponentially, as
I/I
0
5exp 2 (K
2
冓u
2
冔/2), with 冓u
2
冔53"
2
T/(mk
u
d
2
), (3.1)
74 3 Electron-based techniques
Figure 3.6. RHEED patterns (20 kV) of Si(111) in the ‘冑3’ structure associated with the ML
phase of Ag deposited at 500°C: (a) in the normal view, (b) in the perpendicular view, showing
reciprocal lattice unit cell O9ACB, and the two spots between these spots (from Gotoh & Ino
1978, reproduced with permission).
u
d
being the Debye temperature, and K the scattering vector. This means that intensity
measurements as a function of temperature measure 冓u
2
冔, and several such studies have
been done with LEED. An interesting feature of such experiments is that the value of
冓u
2
冔 decreases towards the bulk value as the incident energy is increased, reflecting the
increased sampling depth of the electrons (Lagally 1975, Woodruff & Delchar
1986,1994 chapter 2.7).
In a typical RHEED setup there is no energy filtering, other than that caused by the
fact that higher energy electrons produce more light from phosphor screens. Yet the
geometry is such that plasmons, especially surface plasmons, will be produced very
efficiently. Because plasmon excitation produces only a small angular deflection, the
diffraction pattern is not unduly degraded. A few groups have studied energy-filtered
RHEED (Marten & Meyer-Ehmsen 1988, Ichimiya et al. 1997, Weierstall et al. 1999),
but it is difficult to construct filters which work over a large angular range.
LEED (especially SPA-LEED) and RHEED, and the corresponding microscopies
(LEEM and REM) have been shown to be very sensitive to the presence of surface steps
and other types of defects, including domain structures. Some of these effects are due to
the extra diffraction spots associated with particular domains; some are due to exploit-
ing the difference between in-phase and out-of-phase scattering between terraces separ-
ated by steps; some again depend on the small static distortions and rotations produced
by surface steps, and the increase in diffuse scattering (Yagi 1988, 1989, 1993, Henzler
1977, 1997, Bauer 1994, Wollschläger 1995). SPA-RHEED has also been demonstrated
(Müller & Henzler 1995).
The basic reason for the surface sensitivity of LEED is the short inelastic mean free
path for the excitation of plasmons (and other forms of electron–electron collision);
this means that information from deeper in the crystal is effectively filtered out. One of
the few calculations which is straightforward (see problem 3.2) is the pseudo-kinematic
case, where one has single scattering and exponential attenuation. This calculation
shows that the attenuation causes only a few layers at the surface to be sampled, which
give rise to modulated reciprocal lattice rods, the width of the modulations being
inversely proportional to the imfp.
In the full dynamical LEED calculations, the attenuation effect is included by an
imaginary potential, V
0i
. This is similar to the high energy case of RHEED and
THEED, but the language is a little different. In TEM, imaginary potentials (V
0i
and
V
gi
) are used to describe contrast in images caused by inelastic scattering; but these are
dependent on the aperture size used, and are typically due to the scattering of phonons
and defects. In contrast to plasmons, these scattering events cause a wide angular
spread, and very little energy loss. Calculating RHEED intensities is a suitable combi-
nation of layer slicing, as in LEED, and high energy forward scattering as in THEED;
reviews of these methods have been given by Peng et al. (1996) and Maksym (1977,
1999).
There are new electron diffraction techniques emerging, such as DLEED (Diffuse-
LEED) (Heinz 1995) and electron holography (Saldin 1997), and continuous develop-
ment of related theoretical methods (Pendry 1997). The above (outline) discussion
has concentrated on the effect of inelastic processes on the interpretation of elastic
3.2 Diffraction and quasi-elastic scattering techniques 75
scattering processes. In the next two sections we are concerned with the understanding
and use of the inelastic processes in their own right.
3.3 Inelastic scattering techniques: chemical and electronic state
information
3.3.1 Electron spectroscopic techniques
If we bombard a sample with electrons or photons, electrons will be emitted which have
an energy spectrum, shown schematically in figure 3.7 for the case of electron bom-
bardment. The most well-known historical example is the photoelectric effect, and the
modern version in UHV is called photoemission (Cardona & Ley 1978, Bonzel &
Kleint 1995). Electron emission is commonly used; for example, secondary electrons
are the signal normally used to form an image in the scanning electron microscope
(SEM), and AES uses Auger electrons to determine surface chemical composition. Ion
emission is also known, but is less widely used.
The problem of measuring the energy spectrum is non-trivial, and is discussed in
many books (Bauer 1975, Ibach 1977, Walls 1990, Briggs & Seah 1990, Rivière 1990,
Smith 1994); introductions are given by Prutton (1994, chapter 2) and Woodruff &
Delchar (1986/1994, section 3.1). The field also supports specialist publications such
as the Journal of Electron Spectroscopy, and Surface and Interface Analysis. There are
various possible geometries for the analyzers and the measurements can be performed
in an angle-integrated or angle-resolved (AR) mode. Thus we have a profusion of acro-
nyms, e.g. UPS, ultra-violet photoelectron spectroscopy; ARUPS, the angular resolved
version of the same technique, which is used to study band structure and surface states;
XPS, X-ray photoelectron spectroscopy, also known as ESCA, electron spectroscopy
for chemical analysis, so named by Siegbahn et al. (1967). The massive body of work
by this Swedish group resulted in the Nobel Prize being awarded to Kai Siegbahn in
76 3 Electron-based techniques
Figure 3.7. Electron energy spectrum, showing secondary, Auger, energy loss and
backscattered electrons.
E
0
Secondary
electrons
Backscattered
electrons
Auger
electrons
Energy loss
electrons
Intensity E N(E)
Electron energy E
.