The Surfaces of Bismuth: Structural and Electronic Properties

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 55

Progress in Surface Science 81 (2006) 191–245

www.elsevier.com/locate/progsurf

Review

The surfaces of bismuth: Structural


and electronic properties
Ph. Hofmann
Institute for Storage Ring Facilities and Interdisciplinary Nanoscience Center,
University of Aarhus, Ny Munkegade, 8000 Århus C, Denmark

Abstract

This paper reviews recent experimental and theoretical work on some low-index surfaces of the
group V semimetal bismuth. The main focus is on the geometric and electronic structure, including
the effect of the spin–orbit interaction, and on the electron–phonon coupling. The surface geometric
structures of Bi(1 1 1), Bi(1 1 0) and Bi(1 0 0) do not undergo any dramatic changes with respect to the
bulk structure but the electronic structure does: All three surfaces support metallic surface states
with larger Fermi surface elements and generally lower Fermi velocities than the bulk states. There-
fore the surfaces are considerably better metals than the bulk. This surface electronic structure can
only be understood when the spin–orbit coupling is taken into account, which leads to a strong split-
ting of the fact that the surface state bands due to the loss of symmetry at the surface. The surface
state bands contain only one electron per k point has a profound impact on effects such as screening
and the possible formation of charge density waves. The strength of the electron–phonon interaction
for the surface states of Bi is strongly energy dependent. At higher binding energies, where inter-
band scattering with bulk states can be important, strong electron–phonon coupling can be observed
(k > 0.7). Close to the Fermi level, where only scattering within the surface state bands is relevant,
the coupling is of intermediate strength with k  0.2–0.4, depending on the surface and the particular
state.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Bismuth; Single crystal surfaces; Density functional calculations; Angle resolved photoemission; Low
energy electron diffraction (LEED); Scanning tunneling microscopy; Scanning tunneling spectroscopies; Spin–
orbit interaction; Electron–phonon coupling; Charge density waves; Spin-dependent phenomena

E-mail address: philip@phys.au.dk

0079-6816/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.progsurf.2006.03.001
192 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
2. The bulk properties of Bi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
2.1. Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
2.1.1. Rhombohedral structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
2.1.2. Hexagonal structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
2.1.3. From rhombohedral to hexagonal indices . . . . . . . . . . . . . . . . . . . . . . . . 197
2.1.4. Pseudocubic structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
2.2. Electronic structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
2.3. Thin films, nano-wires and clusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
3. Surface geometric structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
3.1. Bi(1 1 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
3.2. Bi(1 0 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
3.3. Bi(1 1 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205
3.4. General trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4. Surface electronic structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4.1. Bi(1 1 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4.2. Bi(1 1 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
4.3. Bi(1 0 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
4.4. General trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
5. Spin–orbit coupling and splitting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
5.1. Splitting of the surface state bands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.2. Quasi-particle interference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.3. Possibility of a charge density wave on Bi(1 1 1) . . . . . . . . . . . . . . . . . . . . . . . . 223
6. Electron–phonon coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
6.2. Bi(1 0 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
6.3. Bi(1 1 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
6.4. Bi(1 1 0) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
6.5. General trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
7. Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

1. Introduction

One of the most fascinating aspects in the study of surfaces is that their properties can
be radically different from those of the corresponding bulk material. This has considerable
practical interest. One reason is, of course, that the surfaces dictates the chemical proper-
ties of a solid. In addition to this, the relative importance of the surface increases for smal-
ler structures. This is particularly relevant for nano-technology.
This review describes some of the most basic surface properties of the group V semi-
metal Bi. As it turns out, Bi is a striking example of the possible difference between surface
and bulk material properties. Most importantly, the surfaces are much better metals than
the bulk due to the existence of electronic surface states crossing the Fermi level. With
respect to the use of Bi structures in nano-technology, however, it is important to add that
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 193

the properties of bulk Bi are also radically altered by quantum size effects. This is already
observable in structures smaller than a few hundred nanometers or so [1].
The creation of a surface requires the breaking of atomic bonds. In most metals, cova-
lent bonding plays only a minor role. The effect of bond-breaking is small and the surface
properties are similar to those of the bulk, although localized electronic surface states may
be present. On semiconductors, creating the surface leaves so-called dangling bonds which
should give rise to half-filled and therefore metallic bands. However, it turns out that on
most semiconductor surfaces the atoms re-arrange their positions such that the dangling
bonds are removed and the surface is again a semiconductor and not a metal. Semimetals
such as Bi lie in between these two cases. On one hand, a semimetal is close to being a
semiconductor since directional bonding is important and the valence and conduction
bands are almost separated by a gap. On the other hand, a very small overlap between
both bands is found at some points of the Brillouin zone such that the material is formally
a metal. This delicate balance between being a metal and a semiconductor depends cru-
cially on the structural details [2], on external parameters such as the applied pressure
[3], and it can be expected to be severely disturbed at the surface. A semimetal surface
can be expected to turn either into a better metal or into a semiconductor. The former case
is more interesting because a good metallic surface on a semimetal (or, indeed, on a semi-
conductor) can be taken as a model for a nearly two-dimensional metal.
Metallic surface states have been found on Be [4–6] and a-Ga [7–9]. Both materials
could be called semimetals because directional bonding is important and the bulk density
of states has a minimum at the Fermi level, even though it is much less pronounced than in
a ‘‘traditional’’ group V semimetal such as As, Sb or, in the most extreme case, Bi. The
surfaces of Bi have been largely unexplored territory after the pioneering work of Jona
in 1967 [10]. In the recent years, however, there has been an increased interest, both from
the experimental and the theoretical side and considerable progress has been made. Still,
Bi remains rather challenging for both experiments and calculations.
A central result discussed in this review is that the surfaces of Bi are characterized by
metallic surface states and that they are therefore much better metals than the bulk. Indeed,
it is quite appropriate to view the Bi surfaces as quasi-two-dimensional metals. A simple
picture based on the creation of dangling bonds at the surface, however, is not sufficient
to understand the occurrence of metallic states because such states are also found for the
closed-packed (1 1 1) surface which can be created without breaking any covalent bonds.
Instead, even a qualitative understanding of the surface electronic structure requires the
inclusion of the spin–orbit interaction which leads to a splitting of the surface state bands
with respect to their spin direction. The spin–orbit interaction is very strong and many fea-
tures in the surface electronic structure can be understood by the tendency to achieve a
strong splitting of the bands on the one hand and the requirement for vanishing splitting
at certain high symmetry points in the surface Brillouin zone on the other hand. Thus the
Bi surfaces represent quasi-two-dimensional metals with unique spin properties. Indeed,
one of the most practical applications of Bi surface (or interface) states could lie in the con-
struction of spin-sources or spin-filters for the field of spintronics, a novel type of electron-
ics which is based on the spin degree of freedom in addition to, or instead of, the charge.
Spintronics devices would be capable of a much higher speed at very low power (because
of the reduced Ohmic losses) and could also find applications in quantum computing.
This review is structured as follows: Following this introduction, the bulk properties
of Bi are described briefly together with some findings for Bi thin films, nano-wires
194 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

and clusters. This is followed by a description of the most basic surface properties: structure
and electronic structure. In Section 5 the influence of spin–orbit coupling and splitting on
the surface band structure is described and the consequences of this in quasi-particle inter-
ference and the possible formation of charge density waves are discussed. Finally, the elec-
tron–phonon coupling for some surface states of Bi is described in Section 6 and the review
is closed by giving the main conclusions and an outlook in Section 7.

2. The bulk properties of Bi

2.1. Structure

Bismuth crystallizes with rhombohedral symmetry in a structure which is typical for the
group V semimetals (space group R 3m, Strukturbericht A7, arsenic structure). Each atom
has three equidistant nearest-neighbour atoms and three equidistant next-nearest neigh-
bours slightly further away. This results in puckered bilayers of atoms perpendicular to
the rhombohedral [1 1 1] direction in which each atom is covalently bonded to its three
nearest neighbours. The atoms’ next-nearest neighbours are in the adjacent bilayer and
the bonding within each bilayer is much stronger than the inter-bilayer bonding. This
explains why Bi crystals easily cleave along the (1 1 1) plane. The A7 structure has two
atoms per bulk unit cell, corresponding to the two atoms in the bilayers. Alternatively,
the structure can be described as hexagonal with six atoms per unit cell or as a pseudocubic
structure with one atom per unit cell.
The relation between these different unit cells is shown in Fig. 1. The rhombohedral unit cell
is indicated by green lines and the two different atoms in the unit cell are shown in blue and
red. The length ratio d1/d2 is 0.88 (instead of 1) and therefore the red atom is closer to the three
blue atoms below it than to the three atoms above it, forming the above-mentioned bilayers.
The trigonal (C3) axis of the rhombohedral structure is the c axis of the hexagonal structure.
The similarity to a cubic structure can be seen as follows. The angle between the vectors span-
ning the rhombohedral unit cell is a = 57.35 and the ratio d1/d2 = 0.88. For a = 60 and the
d1/d2 = 1, the rhombohedral structure would be simple cubic [10]. This is illustrated in the right
part of Fig. 1 which shows the rhombohedral unit cell in the pseudocubic lattice.
The basic symmetry elements of the bismuth structure are the trigonal axis, the binary
axis, the bisectrix axis, mirror planes and inversion [11]. In the rhombohedral system the
trigonal axis (C3) is lying in the middle of the three vectors spanning the lattice. In the
hexagonal system it is the c axis. The binary axis (C2) is perpendicular to C3. The bisectrix
axis (C1) is perpendicular to C3 and C2. This axis and the trigonal axis span the mirror plane
of the crystal structure. Finally, the crystal possesses inversion symmetry, a fact which will
be important for the discussion of spin–orbit splitting effects. Since the trigonal axis has
threefold symmetry, the binary and bisectrix axes and the mirror plane exist three times.
The different possibilities for describing the bulk structure can lead to considerable
confusion, in particular when discussing the properties of particular surfaces which have
to be specified by their crystallographic indices. There are four possibilities to describe
the surfaces: By rhombohedral indices, three or four hexagonal indices or pseudocubic
indices. The last choice is obviously not a good one, since the system is not truly cubic
and the notation therefore ambiguous. But even when confining the discussion to the
rhombohedral and hexagonal description, some non-trivial aspects appear. These are clar-
ified in the following treatment which gives the same results as in Refs. [10,12].
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 195

hexagonal + rhomb. pseudocubic + rhomb.


C3

α
d2
d1

C1 atom 1
C2
atom 2

Fig. 1. Bulk structure of Bi. Left: rhombohedral unit cell (dashed green lines) together with the hexagonal
unit cell (dashed pink lines). Not all the atoms are shown. Blue and red mark the two atoms in the rhombohedral
unit cell. The solid green and pink lines are the vectors spanning the rhombohedral and hexagonal lattice,
respectively. The three cartesian axes are: bisectrix (C1, y), binary (C2, x) and trigonal (C3, z). Right: Illustration of
the pseudocubic character of the structure together with the rhombohedral unit cell (after Ref. [10]). (For
interpretation of colours in this figure legend, the reader is referred to the web version of this article.)

2.1.1. Rhombohedral structure


The simplest description of the Bi structure is obtained when using a rhombohedral
Bravais lattice with two atoms per unit cell. The lattice is generated by three vectors
~
a1 ;~ a3 of equal magnitude arh [13]. These vectors are shown in Fig. 1 as solid green lines.
a2 ;~
The angle between any pair of the primitive vectors is a. The two basis atoms are chosen to
be at the origin of the co-ordinate system and at ð~ a1 þ ~
a2 þ ~a3 Þd 1 =ðd 1 þ d 2 Þ (see Fig. 1).
Thus the crystal structure is completely described by arh, d1/d2, a. For Bi these values
are arh = 4.7236 Å, d1/d2 = 0.88 and a = 57.35 at 4.2 K [14,15].
For a more convenient comparison with the hexagonal lattice system, we write the
rhombohedral lattice vectors using the characteristic parameters for the hexagonal system
a = 4.5332 Å and c = 11.7967 Å. We use cartesian coordinates such that the binary (C2)
axis is x, the bisectrix (C1) axis is y the trigonal (C3) axis is z. The vectors spanning the
rhombohedral unit cell are then
pffiffiffi !
1 3 1
a1 ¼  a; 
~ a; c ;
2 6 3
pffiffiffi !
1 3 1
a2 ¼
~ a;  a; c ; ð1Þ
2 6 3
pffiffiffi !
3 1
a3 ¼ 0;
~ a; c .
3 3
196 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

These vectors are shown as solid green lines in the left part of Fig. 1. The reciprocal lattice
is given by
pffiffiffi !
~ 1 3 1
b1 ¼ 2p  ;  ; ;
a 3a c
pffiffiffi !
~ 1 3 1
b2 ¼ 2p ;  ; ; ð2Þ
a 3a c
pffiffiffi !
~ 2 3 1
b3 ¼ 2p 0; ; .
3a c

As usual, the (m n o) surface in the rhombohedral notation is the surface plane which is
perpendicular to the reciprocal lattice vector m~
b1 þ n~
b2 þ o~
b3 [13].

2.1.2. Hexagonal structure


For the hexagonal structure the trigonal (C3) axis is the natural choice for the c axis. There
is, however, some ambiguity about how to assign the vectors in the close-packed planes. They
have to enclose an angle of 60 but they could be rotated by any angle with respect to the rhom-
bohedral vectors. A common choice is to define the hexagonal vectors such that they ‘‘connect
atoms’’, like the rhombohedral vectors do. This is achieved by the linear combinations:
ah1 ¼ ~
~ a1  ~
a3 ;
ah2 ¼ ~
~ a2  ~
a1 ; ð3Þ
ah3 ¼ ~
~ a1 þ ~
a2 þ ~
a3 .
With this the hexagonal unit cell is spanned by the following vectors
pffiffiffi !
1 3
ah1 ¼  a; 
~ a; 0 ;
2 2
ð4Þ
ah2 ¼ ða; 0; 0Þ;
~
ah3 ¼ ð0; 0; cÞ.
~

These vectors are shown as solid pink lines in the left part of Fig. 1. The reciprocal lattice
is given by

~ 4p
bh1 ¼ pffiffiffi ð0; 1; 0Þ;
3a
2p pffiffiffi
~
bh2 ¼ pffiffiffi ð 3; 1; 0Þ; ð5Þ
3a
~ 2p
bh3 ¼ ð0; 0; 1Þ.
c
Again, the (h k l) surface in the hexagonal notation is the surface plane which is perpendic-
ular to the reciprocal lattice vector h~
b1h þ k~
b2h þ l~
b3h . In specifying hexagonal surface ori-
entations, it is common to use four indices (h k i l) instead of three (h k l). In using this
redundant indexing, it is understood that i = (h + k). Here we will use four indices for
hexagonal surfaces, in particular to distinguish hexagonal and rhombohedral notations.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 197

2.1.3. From rhombohedral to hexagonal indices


Given the above definitions of the lattice vectors in real and reciprocal space, changing
the indices for a given surface from rhombohedral to hexagonal, and vice versa, is very
simple. For a given (m n o) surface in the rhombohedral notation, the hexagonal indices
(h k l) can be determined by the linear system of equations:
m~
b1 þ n~
b2 þ o~
b3 ¼ h~
b1h þ k~
b2h þ l~
b3h . ð6Þ
The result for the most important rhombohedral surfaces is given in Table 1.
The only remaining ambiguity is the above-mentioned possibility to rotate the hexago-
nal in-plane vectors with respect to the rhombohedral lattice. It is therefore useful to spec-
ify the surface orientation with respect to the trigonal axis and the mirror plane, both of
which are easily recognized in a Laue diffraction picture of the crystal. For the rhombo-
hedral (1 0 0) and (1 1 0) surfaces this is simple. Both directions lie in the mirror plane of
the crystal and enclose an angle of 71.59 and 56.35 with the trigonal axis, respectively.
This criterion is not sufficient to define the surface orientation but it can be used for ver-
ification purposes.

2.1.4. Pseudocubic structure


The similarity of the rhombohedral A7 structure to a simple cubic structure has been
discussed above and with special emphasis on the surfaces by Jona [10]. The pseudocubic
indices are therefore also given in Table 1. It is clear, however, that using the pseudocubic
notation is dangerous because it is not unambiguous. Already from Table 1 it is evident
that the rhombohedral (1 0 0) and (1 1 1) surfaces are both pseudocubic (1 1 1) surfaces.
But their detailed geometry and electronic structure is very different. We will therefore
make only very limited use of the pseudocubic notation. A more detailed description of
the pseudocubic indexing is given in Refs. [10,12].

2.2. Electronic structure

The classification of Bi as a semimetal and the many unique properties of this material
are closely related to its very special electronic structure [16]. Especially after the discovery
of the de Haas–van Alphen effect on Bi, much effort was devoted to the theoretical descrip-
tion of the electronic band structure close to the Fermi level [15,17–19]. Due to the small
energy scales involved, accurate ab initio calculations are difficult even though they have
been tried with some success [18]. If the objective of the calculation is a detailed description
of the Fermi surface at low computational cost then tight-binding calculations are the
method of choice. For Bi one such calculation was published by Liu and Allan in 1995
and the parameters in this calculation were later frequently used for the projection of

Table 1
Indexing of selected low-index surfaces on Bi
Rhombohedral Hexagonal Pseudocubic
(1 0 0) ð1 1 0 1Þ (1 1 1)
(1 1 0) ð1 0 1 2Þ (1 0 0)
(1 1 1) (0 0 0 1) (1 1 1)
ð1 0 
1Þ ð2 1 1 0Þ (1 1 0)
198 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

the bulk band structure on different Bi surfaces. The result of the calculation is shown in
Fig. 2, the bulk Brillouin zone is shown in Fig. 3.
An inspection of Fig. 2 immediately reveals the semimetallic character of Bi: The band
structure can very nearly be described by two filled s bands and three filled p bands,

holes
0 electrons spin-orbit
gap
binding energy (eV)

10

X Γ L U T Γ
Fig. 2. Bulk band structure of Bi from the tight-binding calculation of Liu and Allan (Ref. [15], green lines); first-
principles calculation by Gonze et al. (Ref. [18], red lines), only in the C–T direction. (For interpretation of
colours in this figure legend, the reader is referred to the web version of this article.)

T
U

X
L

L U

L L electron
pocket
T
hole
pocket
Fig. 3. Bulk Brillouin zone of Bi and a schematic sketch of the Fermi surface (not to scale). The C–T line
corresponds to the C3 axis and the [1 1 1] direction in real space.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 199

separated by a gap of several eV. These five bands can accommodate the 10 valence elec-
trons per unit cell, five from each Bi atom. However, this simple picture is not entirely cor-
rect: Close to the T and L points the p bands cross the Fermi level, creating hole pockets at
the T points and electron pockets at the L points. These pockets are very shallow; the
Fermi energy is 27.2 meV for the electrons and 10.8 meV for the holes. This in turn leads
to a very low carrier density of around 3 · 1017 cm3 and small effective masses of the car-
riers; for electrons along the trigonal axis m*  0.003 Æ me. Fig. 4 shows a calculation of the
bulk density of states near the Fermi energy using the tight-binding scheme of Liu and
Allen [15]. It illustrates the semimetallic character of Bi very well. In a narrow window
around the Fermi energy, the density of states drops dramatically by more than two orders
of magnitude.
The tight-binding calculation of Liu and Allen gives a good overall agreement with the
experimental measurements of the bulk Fermi surface (for references to the de Haas–van
Alphen measurements see Ref. [15]). This is not surprising because the tight-binding
parameters were optimized such that these Fermi surfaces are reproduced.
The dispersion of the bulk electronic states away from the Fermi energy was measured
using angle-resolved photoemission from Bi(1 1 1) by several authors with quite similar
results [20–26]. A reasonably good agreement with the calculated band structures (tight-
binding or ab initio) was found.
It is interesting to compare the electronic structure of rhombohedral Bi to that of the
very similar simple cubic structure. In the simple cubic case, the two atoms in the rhom-
bohedral unit cell become equivalent and the unit cell contains only one atom with five
valence electrons, i.e. an uneven number. Simple electron counting therefore predicts
the cubic structure to have a partially filled valence band and to be metallic. In the rhom-
bohedral unit cell the number of valence electrons is 10. This could lead to insulating
behaviour but a small band overlap results in the semimetallic situation. The important
point is that the difference between the two structures is small. Already from this simple
argument, one can expect structural changes at surfaces to have a major influence on
the electronic structure.
We conclude this section by addressing the effect of spin–orbit coupling on the bulk band
structure of Bi which has been discussed in detail by Gonze et al. [18]. Bi is a heavy element
with strong spin–orbit splitting in the atomic 6p levels (the atomic p3/2–p1/2 splitting in Bi is
DOS (arb. units)

3 2 1 0 -1 -2 -3
Binding energy (eV)

Fig. 4. Calculated bulk density of states close to the Fermi level using the tight-binding parameters of Liu and
Allen [15].
200 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

1.5 eV [27]). When the bulk electronic structure is calculated, the inclusion of spin–orbit
coupling has only a small effect on the two lowest bands. This is not surprising, since these
have predominantly s character and hence L = 0. The bands close to the Fermi level, how-
ever, are strongly affected: the strong spin–orbit interaction essentially accounts for the
existence of the hole Fermi surface at the T point [18]. Most importantly from a surface
point of view, the spin–orbit coupling does not lead to any lifting of the spin degeneracy,
i.e. there are still only six bands, each having two possible spin states per k point. This is
caused by the bulk inversion symmetry [28] and will be discussed in detail in connection
with the spin–orbit splitting on surfaces, where this symmetry is broken.
Another consequence of the spin–orbit interaction is the so-called projected spin–orbit
gap in the C–T direction, which corresponds to the normal emission direction of Bi(1 1 1).
This projected gap is formed between the p band which gives rise to the hole pocket at T
and the p band at higher binding energy (see Fig. 2). For the C–T direction the band result
of Gonze et al. is also shown in Fig. 2. Evidently, the first principles calculations predict
the projected band gap. The tight binding calculation also predict the lifting of the degen-
eracy at C but it does not give rise to a projected gap in the C–T direction.

2.3. Thin films, nano-wires and clusters

This section deals with the properties of small Bi structures. As it turns out, these can be
very different from those of the bulk. This may not be surprising because the surfaces gain
an increasing importance for small structures and we have seen how intimately the partic-
ular electronic structure of the rhombohedral phase is linked to its geometry. In addition
to this, quantum-size effects play an important role for small rhombohedral Bi systems. In
the following, we briefly review the properties of Bi thin films, nano-tubes and clusters.
Quantum mechanics dictates that the confinement of particles with an effective de
Broglie wavelength kB to structures with dimensions d smaller or comparable to kB should
lead to interference effects and quantized states. However, it was the seminal work of
Ogrin et al. in 1966 [1] which started a whole research field of quantum size effects in thin
films. Ogrin et al. argued that Bi films grown in the [1 1 1] direction should be ideal to
observe such effects for several reasons. First, the low Fermi energy and the small effective
mass of the carriers lead to a long de Broglie wavelength of around 120 Å. For such a large
value of kB, small irregularities in the film thickness are not expected to play a significant
role. Secondly, the mean free path of the carriers in Bi is long enough (1 mm at low tem-
peratures) not to destroy the interference phenomenon. Indeed, quantum size induced
oscillations of transport properties were observed as the thickness of the Bi layers is varied.
The typical distance between maxima in the transport properties is 400–500 Å.
Assuming a semimetallic band structure in thin films of Bi, similar to that of the bulk, a
semimetal-to-semiconductor transition has been predicted for films below a critical thick-
ness by Lutskii [29] and Sandomirskii [30]. This transition should occur when the energy
shift caused by the confinement is great enough to lift the energies of the lowest electron
states above those of the highest hole states (see Fig. 2). The critical thickness at which this
happens is expected to be between 230 Å and 320 Å. Considerable experimental effort has
been made in order to identify this transition. In the early work, it has been tried to iden-
tify a sudden change in the transport properties at the critical thickness (see discussion in
Ref. [31]). It was soon realized, however, that surface states could play an important role
on thin films [32,33] and that their presence could inhibit the discontinuity of the transport
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 201

properties at the transition thickness. Taking such effects into account, Hoffman and co-
workers have reported the observation of a semimetal-to-semiconductor transition at a
critical thickness of 320 Å [31]. Their interpretation was challenged by Chu who argues
that the same data can be interpreted by a model with no energy gap but different bound-
ary conditions for the wave functions [34]. To date, it appears that this long-standing issue
is still unresolved. Thin film studies have mostly been confined to films grown in the (1 1 1)
direction but other orientations may also be possible [35–37].
Recently, much attention was also directed towards the electronic structure and transport
properties of Bi nano-wires [38]. This is mostly due to the potential importance of such wires
or wire arrays in thermoelectric applications: For purposes of cooling or power generation,
the thermoelectric figure of merit ZT has to be maximized, but for conventional materials
the achievable values of ZT are thought to be reached. It has been proposed that the funda-
mental limits for optimization could be overcome by using nano-wires instead of bulk crys-
tals [39] and that Bi would be a promising material for such wires [40]. These ideas are
supported by recent experimental results [41]. However, the electronic structure of thin
nano-wires is disputed in a similar manner as for the thin films: For wires oriented in the
direction of the trigonal axis, Lin et al. have predicted a metal-to-semiconductor transition
for wire diameters below 55 nm [40]. This result could not be confirmed by the measurement
of Shubnikov–de Haas oscillations for wires of 30 nm diameter and the missing transition
was interpreted to be caused by electronic surface states [42].
Questions also arise regarding the electronic structure and superconducting properties
of small Bi clusters. Rhombohedral bulk Bi has not been found to be superconducting for
temperatures down to 50 mK [43]. However, Weitzel and Micklitz have reported the
observation of superconductivity in granular films of rhombohedral Bi clusters [44].
Depending on the cluster size, TC can reach values as high as several K. An initial expla-
nation was the occurrence of surface superconductivity on the clusters. This appeared to
be consistent with a number of experimental findings: a linear increase of TC for a smaller
cluster size L and a dependence of TC on the chemical nature of the embedding matrix.
Later, the view of a two-component cluster (metallic surface and semimetallic bulk) was
challenged by the same group [45]. The normal and superconducting properties show a
very similar dependence on the cluster size suggesting that dividing the system into two
independent components might be inappropriate.
In conclusion, there are two important aspects for non-bulk Bi. The first is the change
of the bulk electronic structure due to quantum confinement effects, as discussed above.
The second is the increasing importance of the surfaces for small structures because these
can, and do, have very different electronic properties than the bulk. These effects are of
course linked but it is important to bear in mind that the first sets in for length scales which
are considerably larger than in normal metals.

3. Surface geometric structure

In this chapter the geometric structures of three Bi surfaces are discussed: Bi(1 1 1),
Bi(1 0 0) and Bi(1 1 0). The first two are pseudocubic (1 1 1) surfaces whereas the last is a
pseudocubic (1 0 0) surface. For Bi(1 1 1) and Bi(1 1 0) both experimental and calculated
structural results are discussed, for Bi(1 0 0) only calculated structural parameters are
available. The geometric structures of the Bi surfaces have only recently been studied in
detail, a long time after the first work of Jona [10] who discussed the truncated bulk
202 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

structure of various Bi and Sb surfaces as well as the low energy electron diffraction
(LEED) patterns obtained from them.

3.1. Bi(1 1 1)

Bi(1 1 1) is by far the most important Bi surface for practical applications. It is the
natural cleavage plane of Bi crystals and it also turns out to be the preferred direction
of epitaxial growth, even though there are exceptions for very thin films [36].
The truncated bulk crystal structure of Bi(1 1 1) is shown in Fig. 5. For the figure, it is
assumed that the crystal is terminated with a Bi bilayer. By considering only the (1 · 1)
LEED diffraction pattern reported by Jona [10], a different termination would also be pos-
sible: the surface could be terminated with a long first interlayer spacing, i.e. with a bilayer
cut upon surface creation. This is utterly inconceivable because of the covalent bonds that
hold the bilayer together. The locations of these bonds are indicated in the figure. For the
formation of a split bilayer, three covalent bonds per surface unit cell have to be broken.
Creating a surface terminated with an intact bilayer, on the other hand, does not require
the breaking of any covalent bonds. This picture is confirmed by the overall mechanical
properties of Bi which is brittle and easily cleaves along the (1 1 1) plane, presumably
between two bilayers. The bilayer-type structure gives rise to alternating interlayer
distances.
The detailed structural parameters of this surface have been determined by Mönig et al.
using LEED and first principles calculations [46]. Given the (1 · 1) LEED pattern for this
surface, the only structural degrees of freedom are relaxations of the interlayer distances,

1st layer
2nd layer (a) Top view
3rd layer
4th layer

mirror plane(s)
4.54 Å

4.5

(b) Side view (parallel to mirror plane)

1.59 Å

2.35 Å

Fig. 5. Truncated-bulk structure of Bi(1 1 1). The dark solid lines indicate covalent bonds between the atoms
within the bilayers. (a) Top view of the first three atomic layers. Each layer consists of a two-dimensional trigonal
lattice and the lattice constants are given. The mirror planes of the structure are shown as dashed lines. (b) Side
view of the first four layers along a mirror plane. The alternating short and long interlayer spacings are evident.
After Ref. [46].
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 203

i.e. deviations from the truncated bulk values. These relaxations are quite small: at
T = 140 K the first interlayer spacing d12 is found to expand by Dd 12 =d b12 ¼ þ0:5 
1:1%, the second is found to expand by Dd 23 =d b23 ¼ þ1:9  0:8%, and for the third
Dd 34 =d b34 ¼ þ0:0  1:1%. Experimental data taken at higher temperatures suggest that
the first interlayer spacing is temperature-independent whereas there is a slight thermal
expansion of the second interlayer spacing.
These results agree well with intuition; small relaxations can be expected for a bilayer
structure with very weak, van der Waals type inter-bilayer bonds. Indeed, in the limit of a
vanishing interaction between the bilayers the creation of the surface would not have any
effect at all. It is even conceivable that the second interlayer spacing should relax more
strongly and be more affected by temperature.
Apart from the geometry changes, quantitative LEED also provides information about
the vibrations of the surface atoms because these vibrations lead to a damping of the dif-
fracted intensities via a Debye–Waller factor. The root mean-square vibrational ampli-
tudes of the atoms in the first two layers and the corresponding Debye temperatures
have also been determined. The resulting values are 71(+7/5) K and 77(+8/7) K for
the first and second layer, respectively. Thus, the surface Debye temperature is clearly
reduced with respect to the bulk value which is 120 K [47]. Reduced Debye temperatures
at surfaces are a common phenomenon [48] and the results by Mönig et al. are in good
agreement with early findings [49]. Note, however, that the absolute value of the Debye
temperature has to be taken with care: it can be temperature-dependent, both in the bulk
[47] and on the surface [46]. The reason for this is that the Debye model is only a very sim-
ple approximation.
The experimental values for the relaxations have been compared to the result of first
principles calculations, also in Ref. [46]. The calculations which include the spin–orbit
interaction give a small expansion of the d12 distance, Dd 12 =d b12 ¼ þ0:6%, and an expan-
sion of the second interlayer spacing, Dd 23 =d b23 ¼ þ6:2%. These data can be compared with
the experimental results linearly fitted and extrapolated to 0 K which are: Dd 12 =d b12 ¼
þð1:2  2:3Þ% and Dd 23 =d b23 ¼ þð2:6  1:7Þ%. For the Dd 12 =d b12 relaxation a very good
agreement is obtained with the computed value. For Dd 23 =d b23 the theoretical value clearly
lies outside the experimental error but the agreement can still be judged to be fair in view
of the remaining uncertainties. Indeed, it was shown that a variation of the theoretical
relaxations of a few percent around the optimum leads only to very small changes in
the total energy per surface atom, in the order to 10 meV. Such fine details can be influ-
enced by the choice of the approximation for the exchange–correlation potential, by the
finite thickness of the slab, by small inconsistencies in the k-point sampling between bulk
and film calculations, as well as by other factors.
A quite different picture of the surface structure of Bi(1 1 1) has been presented by Ast
and Höchst based on a comparison of the measured electronic structure and tight-binding
calculations [50]. Good agreement between the measured and calculated surface state dis-
persion was obtained, assuming a strong outward relaxation of the first interlayer spacing.
This relaxation was found to be Dd 12 =d b12 ¼ 71% where Dd12 (=1.13 Å) is the change in the
first to second interlayer distance and d b12 (=1.59 Å) is the truncated bulk value for this
distance. Such a big interlayer relaxation would be highly unusual. Moreover, the first
interlayer spacing, d12, would be larger than the inter-bilayer distance d23, leading to a
completely different type of bonding at the surface. The need to assume such a big relax-
ation probably stems from the very weak spin–orbit coupling parameter used in the same
204 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

calculation. It will be shown later that the spin–orbit interaction plays a major role for the
surfaces of Bi and that the coupling is strong.

3.2. Bi(1 0 0)

Bi(1 0 0) is, like Bi(1 1 1), a pseudocubic (1 1 1) surface. However, the actual structure of
Bi(1 1 1) and (1 0 0) is quite different. Fig. 6 shows the truncated bulk structure of Bi(1 0 0).
It can be viewed as built from quasi-hexagonal layers. In each layer an atom has two
neighbours at the same distance as on the Bi(1 1 1) surface (4.54 Å) and four next-nearest
neighbours at a slightly larger distance (4.72 Å). The three nearest neighbours of any Bi
atom in the bulk structure, however, do not lie in the same quasi-hexagonal layer but they
connect these layers. The connections to the nearest neighbours are indicated as ‘‘bonds’’
in Fig. 6. The side view shows that there are two different possible interlayer spacings. The
termination of the surface is not known, but it seems likely that the shorter interlayer spac-
ing prevails, not least because it requires only the breaking of one covalent bond per unit
cell instead of two. The broken bonds at the surface are also indicated as dashed lines in
Fig. 6. It can be seen that the structure is close to an ABCABC. . . stacking sequence of the
quasi-hexagonal layers, except that the fourth layer atoms are not quite in registry with
the first layer atoms. They are actually shifted by a distance of 0.57 Å. The symmetry
of the surface is very low. The only symmetry element is a mirror plane which is also indi-
cated in the figure. Consequently there are some important differences between Bi(1 0 0)
and Bi(1 1 1) even though both are pseudocubic (1 1 1) surfaces: In order to form
Bi(1 1 1) no covalent next-neighbour bonds have to be broken while one bond per unit cell
has to be broken to form Bi(1 0 0). Bi(1 1 1) also possesses a higher symmetry with a three-
fold axis and three mirror planes.

(a) Top view

4.7

1st layer
mirror plane
2nd layer

3rd layer

4th layer

(b) Side view (parallel to mirror plane)

1.74 Å
1.98 Å

Fig. 6. Truncated-bulk structure of Bi(1 0 0). The dark solid lines indicate covalent bonds between the atoms
within the bilayers. (a) Top view of the first four atomic layers. The single mirror plane of the structure is shown
as dashed line. (b) Side views of the first four layers parallel to the mirror plane. Dashed lines on the first layer
atoms indicate dangling bonds. From Ref. [51].
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 205

The LEED pattern for the surface has been found to be (1 · 1) [51] but no quantitative
LEED study has yet been performed. However, the energy-optimized geometric structure
has been calculated for this surface by first-principles [52]. The first interlayer distance d12
contracts by Dd 12 =d b12 ¼ 3:0%, and the second interlayer spacing expands by Dd 23 =d b23 ¼
þ14%.
There is a fundamental difference between Bi(1 1 1) and the other surfaces of Bi which
stems from the fact that Bi has two inequivalent atoms in the bulk unit cell. For Bi(1 1 1) a
shift of these atoms which does not break the translational symmetry parallel to the sur-
face corresponds to an interlayer relaxation. For Bi(1 0 0), Bi(1 1 0) and other surfaces, this
is not necessarily the case. It would, for example, be possible to shift the second layer
atoms of Bi(1 0 0) along the mirror line without breaking the translational symmetry
parallel to the surface, retaining a (1 · 1) LEED pattern. Such displacements have been
considered in the first-principles calculations but they have been found to be very small
(below 0.1 Å) [52].

3.3. Bi(1 1 0)

Bi(1 1 0) is a pseudocubic (1 0 0) surface. The truncated bulk structure is shown in Fig. 7.


As with the other surfaces, the covalent bonds have been drawn by solid lines and the dan-
gling bonds at the surface are indicated by dashed lines. The pseudosquare character of the
unit cell is evident: For a cubic Bi structure all the atoms in the first bilayer would have the

mirror plane
(a) Top view
m

1st layer

2nd layer

3rd layer

4th layer

4.75 Å

4.54 Å

(b) Side view (c) Side view


perpendicular to m parallel to m

Fig. 7. Truncated-bulk structure of Bi(1 1 0). The dark solid lines indicate covalent bonds between the atoms
within the bilayers. (a) Top view of the first two atomic layers. The single mirror plane of the structure is shown as
dashed line. (b, c) Side views of the first eight layers (four double layers) perpendicular and parallel to the mirror
plane, respectively. Dashed lines on the first layer atoms indicate dangling bonds.
206 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 8. LEED diffraction patterns taken from the Bi(1 1 0) surface. The mirror line is horizontal. The lack of any
other symmetry and the square-like reciprocal lattice are evident. The images have been colour-inverted and some
intensities are saturated for better presentation. From Ref. [53].

same height and the unit cell would be rotated by about 45 and contain only one atom.
The surface has the same low symmetry as Bi(1 0 0) with only one mirror line.
The low symmetry leads to a couple of difficulties with respect to LEED measurements
on Bi(1 1 0) (and all other surfaces possessing such a low symmetry). These difficulties are
illustrated in Fig. 8 which shows two measured LEED patterns taken at different kinetic
energies. The pseudosquare of the reciprocal lattice and the missing left/right symmetry is
clearly evident whereas the up/down symmetry is given by the mirror plane in the crystal.
For a standard structural determination by LEED it is necessary to align the sample
surface perpendicular to the incoming electron beam and this is usually achieved by
comparing the energy-dependent intensities of the symmetry-equivalent beams. Here this
procedure can only be applied for the up/down angle. In addition to this, most computer
codes for calculating LEED intensities make use of the surface symmetry and do not per-
mit having merely one mirror line.
Bi(1 1 0) has also been discussed by Jona who defined another unit cell and came to the
conclusion that the LEED pattern should not be exactly rectangular but that the recipro-
cal lattice vectors should include angles slightly different from 90. This, however, is not
correct.
Sun et al. have performed a quantitative LEED structural analysis of Bi(1 1 0) [53]. For
a sample temperature of 110 K, they have found the first four interlayer relaxations to be
Dd 12 =d b12 ¼ 13  23%, Dd 23 =d b23 ¼ 0:2  1:4%, Dd 34 =d b34 ¼ 105  19% and Dd 45 =d b45 ¼
þ4:3  1:5%. Note that the first and the third interlayer spacing is very small in the bulk
(0.21 Å), such that a seemingly dramatic relaxation of more than 100% is not very big in
absolute terms. First-principles calculations give relaxations only for the first two inter-
layer spacings. The values are Dd 12 =d b12 ¼ 141%, Dd 23 =d b23 ¼ 0:5% [53]. As for Bi(1 1 1),
the agreement between experiment and theory can be considered to be good, keeping in
mind that the absolute difference between a first interlayer relaxation of 13% and
141% is actually quite small. Moreover, in contrast to the case of Bi(1 1 1), the temper-
ature dependence of the structure has not been investigated. It is therefore not possible to
extrapolate the experimental relaxations to zero temperature before comparing them to
the calculations.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 207

As for Bi(1 1 1), the analysis of the LEED data also yielded surface Debye temperatures
for the first and the second layer, which were found to be 95 K and 116 K, respectively.
Again, the surface Debye temperature is lower than for the bulk but not as low as for
Bi(1 1 1).
Again displacements for the first and second layer atoms along the mirror plane have
been considered, both in the analysis of the experimental data and in the first-principles
calculations. In both cases, they have been found to be very small.

3.4. General trends

None of the three low-index surfaces studied so far shows a reconstruction. For Bi(1 1 1)
this is the expected result if one views the crystal as a weakly bonded stack of bilayers. In
this case, the surface should behave like the bulk, as it is the case for most layered mate-
rials. For Bi(1 1 0) and Bi(1 0 0), however, the absence of any reconstructions is unexpected
and in sharp contrast to the situation on most semiconductor surfaces. Explaining the lack
of the reconstruction with simple arguments is not possible but in the end of this review we
will argue that the strong spin–orbit splitting of the surface state bands could play a role.

4. Surface electronic structure

4.1. Bi(1 1 1)

Bi(1 1 1) is the only surface of Bi for which angle-resolved photoemission spectroscopy


(ARPES) studies have been performed by many authors. It is important to realize how
challenging such measurements are experimentally: The small energy scale and low effec-
tive masses call for high energy resolution and momentum resolution and the low Debye
temperature of Bi requires low temperatures in order to avoid final state broadening by
phonon assisted transitions. In fact, the first experiments suffer from the problem that
not all the relevant peaks have been resolved.
The bulk Brillouin zone (BZ) of Bi and the projection on the (1 1 1) surface are shown in
Fig. 9. Taking photoemission spectra at normal emission corresponds to sampling the C–T
direction. This is the only direction for which photoemission data for the determination of
the bulk electronic structure have been taken so far [20–26]. The bulk Fermi surface ele-
ments are projected on the surface Brillouin zone (SBZ) such that the electron pockets
can be found close to the M points and the hole pockets can be found close to the C point.
Nowhere else can there be bulk states very close to the Fermi level.
The first experiments were performed by Jezequel et al. [20]. Later, similar data were
published by the same group [21,22] whereas Tanaka et al. reported results from thin Bi
films [24,25]. The most important result from the early work was that Bi(1 1 1) supports
a surface state in the projected spin–orbit gap (see Fig. 2) at a binding energy of about
200 meV [20] but the energy resolution of around 250 meV was not good enough to pro-
vide further details.
Later experiments by Patthey et al. with much better energy resolution (40 meV) con-
firmed the existence of this surface state but also showed that the original peak observed
by Jezequel et al. actually contains several components [54]. Very recently, high-resolution
(40 meV) bulk band structure measurements in normal emission from Bi(1 1 1) have been
reported by Ast and Höchst [26]. These results also show the surface state in the spin–orbit
208 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

K
M

M
K

T
U

X
L

L U

L L electron
pocket
T
hole
pocket

Fig. 9. Bulk Brillouin zone of Bi and projection on the (1 1 1) surface. The elements of the bulk Fermi surface are
indicated but not to scale. The surface Brillouin zone has a threefold rotational axis and three mirror lines, shown
as pink dashed lines. (For interpretation of colours in this figure legend, the reader is referred to the web version
of this article.)

gap but at the same time the bulk bands in the C–T direction appear to cross the projected
gap. If this was the case, the surface state would not actually be situated in a projected gap
and should be called, more correctly, a surface resonance.
An important result from the work of Patthey et al. was that the photoemission spectra
taken in normal emission showed a high intensity at the Fermi level which was not be
expected for the semimetal Bi. This called for a detailed investigation of the band structure
close to the Fermi level, a task requiring even better resolution. Such a study was pub-
lished in 2000 by Hengsberger et al. [23]. The energy and angular resolution in that work
were 5 meV and 1, respectively, and the sample could be cooled down to 12 K. States
close to the M and C points were found and interpreted in terms of a modified bulk Fermi
surface of Bi. An electron pocket at M was found but it had a higher effective mass than
that expected for the bulk states. Near C the situation remained somewhat unclear with
several closely separated peaks in the spectra. However, a photoemission intensity plot
around C showed six pockets, grouped around C in a star-like fashion. An experimental
drawback in the work of Hengsberger et al. was the light source, a He lamp which works
at a fixed photon energy and does therefore not permit bulk and surface states to be dis-
tinguished by their different behaviour as the photon energy is changed.
The electronic structure around C was eventually clarified by Ast and Höchst, using
synchrotron radiation, an energy resolution of 25 meV and a very high momentum reso-
lution of 23 mÅ1 [55,56]. Their result for the dispersion of the states close to C and the
photoemission intensity at the Fermi level is reproduced in Fig. 10. Such photoemission
intensity plots can be viewed as a ‘‘picture’’ of the Fermi surface [57–60] even though this
simple interpretation has its limits [61–63]. There are two features in the Fermi surface; an
inner ring which is actually a hexagon (see also Ref. [64] and Fig. 24) and six narrow lobes
in the C–M directions. A detailed look at the dispersion of the states, shown in Fig. 10(b),
reveals that the inner hexagon encloses filled states and the lobes enclose empty states.
These Fermi surface elements have therefore been termed the electron pocket and hole
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 209

Fig. 10. (a) Photoemission intensity at the Fermi level of Bi(1 1 1). kx and ky are the parallel components of the
electron momentum along the C–M and C–K direction, respectively. (b) Band structure along the C–M direction.
From Ref. [55].

pocket, respectively [55]. Probing this electronic structure over a wider range of photon
energies has shown that both the electron and the hole pocket are surface states. From
more detailed measurements of the dispersion (for example in Ref. [65]) it is possible to
estimate the Fermi velocities of the states. For the hole pocket the velocity is in the order
of 1.3 eVÅ but the electron pocket is much steeper with a velocity of about 3.2 eVÅ. This
is quite a high velocity and mapping the band structure of the electron pocket is at the
limit of the capabilities of a modern spectrometer. Still, it is much smaller than typical bulk
velocities which can be up to 20 eVÅ [16].
The most important conclusion from these results was that the number of surface
charge carriers is much higher than the corresponding number of (projected) bulk carriers,
even if only the Fermi surface elements around C are taken into account. In other words,
the surface is a much better metal than the bulk. This certainly has a big influence on Bi
thin films and clusters and it confirms that surfaces states have to be considered when
investigating the semimetal-to-semiconductor transition for thin Bi films as discussed in
Section 2.
A first-principles calculation of the electronic structure of Bi(1 1 1) by Koroteev et al. is
shown in Fig. 19 together with the projected bulk band structure (red lines for the surface
state dispersion, yellow and brown lines for the bulk continuum) [65]. The agreement with
the experimental data of Ast and Höchst around the C can be judged to be good. Both the
electron pocket and the hole pocket are reproduced. The only differences are the maximum
binding energy of the electron pocket at C and the fact that the states forming the electron
and hole pocket are degenerate in the calculation but do not appear to be so in the exper-
iment. The first difference is probably due to a sample-misalignment in Ref. [55,64].
Indeed, for a carefully aligned sample the electron pocket is more clearly visible and
extends to higher binding energies (see Fig. 20). The question if the states are degenerate
at C or not is interesting in connection with the spin–orbit splitting discussed in Section 5,
but it is difficult to settle because the states dramatically loose their intensity as they move
210 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

into the projected bulk continuum near C (the projected bands together with the experi-
mental dispersion are also shown in Fig. 20). Very recently, an angle-resolved photoemis-
sion study on Sb(1 1 1) has shed some new light on this issue [66]. The electronic structure
of Sb(1 1 1) is quite similar to that of Bi(1 1 1) but the two surface state bands are clearly
seen to be degenerate at C as they lie outside the continuum of projected bulk bands.
A similar problem arises at the M point where the calculation predicts the two surface
state bands to be degenerate but the experimental situation is unclear. It is summarized in
Fig. 11 which shows results from Ast and Höchst [50] and from Hengsberger et al. [23].
Fig. 11(a) shows the complete dispersion of the states along C–M which agrees, on the
whole, well with the theoretical result in Fig. 19: The electron pocket crosses the Fermi
level very close to C, the hole pocket forms two crossings further away and even further
away the band which forms the electron pocket crosses the Fermi level again. However,
the absolute binding energy at M and the degeneracy of the two states in the calculation
does not seem to agree with the experiment (see Fig. 11(b) and (c)). But actually, this is
difficult to decide because, again, in the experiment the intensity of the peaks is lost as they
move into the bulk continuum, especially the peak at higher binding energy. We will come
back to this question when discussing the spin–orbit coupling.
Finally, it is puzzling that the Bi(1 1 1) surface is so different from the bulk in terms of its
electronic structure. There is, at first glance, not much reason to expect this: the geometric
structure is hardly changed upon the formation of the surface and, more importantly, it is
not necessary to create any dangling bonds to form the surface, as for the other surfaces.
The answer to this puzzle lies in loss of inversion symmetry and the effect of the spin–orbit
interaction and it will be discussed in Section 5.

Θ=

31º

28.5º
Intensity (arbitrary units)

26º
(a)
23.5º
_
M 22.3º
_
M
EF
21º
Energy (meV)

18.5º

-100
16º

13.5º

-200 exp. data, 12 K 11º


exp data, 300 K
theory (bulk)
fitted parabola 8.5º

0.4 0.6 0.8 1.0


k|| (Å-1) -300 -200 -100 EF
(c) (b) Energy (meV)

Fig. 11. Dispersion of the electronic states of Bi(1 1 1) close to the M point: (a) overview of the dispersion in the
C–M direction (from Ref. [64]). (b, c) Spectra near M and dispersion inferred from them, respectively (from Ref.
[23]).
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 211

4.2. Bi(1 1 0)

Before discussing Bi(1 0 0), the other pseudocubic (1 1 1) surface, we describe the elec-
tronic properties of Bi(1 1 0), which are considerably simpler and which were studied by
Agergaard et al. in 2001 by ARPES [67] and by Pascual et al. using first-principles calcu-
lations and scanning tunnelling microscopy (STM) [68].
The projection of the BZ in the (1 1 0) direction and the SBZ of Bi(1 1 0) are shown in
Fig. 12. As for the geometric structure, the only symmetry element is a mirror line. How-
ever, the band structure symmetry observed in a non-spin-resolved photoemission exper-
iment from the surface states is higher and quite different from that observed in the LEED
pattern of Fig. 8. In fact, the measured surface state dispersion appears to be symmetric
not only with respect to the mirror line along X 2 –C–X 02 but also with respect to the line
X 1 –C–X 1 . This is caused by a combination of mirror symmetry and time-reversal symme-
try. For a two-dimensional system, the latter leads to a symmetry in the dispersion of the
electronic states Eð~ k k ; "Þ ¼ Eð~
k k ; #Þ. This means that if one has a surface state at ~ k k with a
binding energy E and a spin ", then there must also be a state at ~ k k with the same energy
but spin #. As an example, take a surface state with ~ k k ¼ ðk x ; k y Þ where the x direction is
the mirror line. Time-reversal symmetry dictates that a state with the same energy is found
at ~k k ¼ ðk x ; k y Þ. Mirror symmetry leads to an equivalent state at ~ k k ¼ ðk x ; k y Þ and
another application of time-reversal symmetry to ~ k k ¼ ðk x ; k y Þ. In the end, the dispersion
looks as if there were two mirror lines instead of one, but only for a situation where the
spin is not measured. The role of the spin and its major impact on the surface state disper-
sion will be discussed in Section 5.
Fig. 13(a) shows the measured and calculated dispersion of the electronic states of
Bi(1 1 0). Several surface states are observed in the gaps of the projected band structure
around the Fermi energy and these surface states show several Fermi level crossings: There

X2 M2

X1
hole
pocket X1 Γ
T

X X'2
L

L M1

L L electron
pocket
T

Fig. 12. Bulk Brillouin zone of Bi and projection on the (1 1 0) surface. The elements of the bulk Fermi surface are
indicated but not to scale. The shaded plane is the bulk mirror plane and the dashed pink line its projection onto
the surface. (For interpretation of colours in this figure legend, the reader is referred to the web version of this
article.)
212 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 13. (a) Dispersion of the electronic states of Bi(1 1 0). The colour scale plot is the photoemission intensity, the
black dots are the projection of the bulk bands calculated using the tight-binding parameters of Ref. [15] and the
red dots are the result of a first-principles calculation. (b) Photoemission intensity at the Fermi energy. After Refs.
[67,68]. (For interpretation of colours in this figure legend, the reader is referred to the web version of this article.)

are hole pockets around the C and M points, a shallow electron pocket centred on the
M–X 1 line and a very small feature along C–X 1 . The shape of the surface Fermi contour
is more clearly seen in Fig. 13(b) which shows the photoemission intensity at the Fermi
level. The surface states forming the Fermi surface have Fermi level velocities in the order
of 1 eVÅ.
Fig. 13(a) also shows the result of the first-principles calculation of the surface elec-
tronic structure of Pascual et al. [68]. The agreement with the measured surface state dis-
persion is excellent.
In conclusion, the (1 1 0) surface of Bi is similar to the (1 1 1) surface in that it supports
metallic surface states. These surface states have much lower Fermi velocities than the bulk
carriers and the Fermi surface elements are much bigger than the projected bulk Fermi
surface. Both lead to a situation where the density of states at the Fermi level is consider-
ably higher at the surface than in the bulk.

4.3. Bi(1 0 0)

The last surface for which we discuss the electronic structure is Bi(1 0 0). As far as we
know, the only paper published on this surface is the work by Hofmann et al. [51]. All
the information shown in this section is taken from that paper. It turns out that
Bi(1 0 0) is quite similar to the two other surfaces in that it also supports metallic surface
states. But the character of these states is very different.
Fig. 14(a) shows the projection of the (1 0 0) SBZ from the bulk BZ. The picture is very
similar to Fig. 9 but the bulk BZ is rotated such that the bulk C–L direction corresponds to
the C point of the surface Brillouin zone. The SBZ is only a semi-regular hexagon and the
only symmetry element is a mirror line. As on Bi(1 1 1), there is a bulk Fermi surface ele-
ment close to the C point but here it is an electron pocket, not a hole pocket. The high
symmetry points of the SBZ have been called K and M in analogy to a truly hexagonal
case, but note that there are two in-equivalent M points, called M and M 0 .
Fig. 15 shows the measured dispersion of the electronic states along several directions in
the extended zone scheme. Equivalent directions have been marked with identical coloured
bars for easier comparison. Several surface states can be identified. There are three very
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 213

(a) K M'
M

M
M' K

X
T
L

Γ
T
electron L hole
pocket pocket

(b)

M1
Γ1 M2 Γ2

K1 K3
K2 K4
M'1 M'2 M'3

M3 Γ3

Fig. 14. (a) Bulk Brillouin zone of Bi and projection onto the (1 0 0) surface. The elements of the bulk Fermi
surface are indicated, but not to scale. The shaded plane is the bulk mirror plane and the dashed line its projection
onto the surface. (b) Surface Brillouin zone of Bi(1 0 0) in an extended zone scheme. The coloured lines in the
bottom part of the first surface Brillouin zone correspond to the lines along which the electronic dispersion was
measured and calculated (see Figs. 15 and 17). From Ref. [51].

steep bands close to C in the C–M direction, one along C–K, a band at higher binding
energy along C–K which shows a maximum in binding energy close to C and a minimum
close to K, and a state at the Fermi level exactly at M 0 .
The surface Fermi contour derived from the dispersion and the photoemission intensity
at the Fermi level shown in Fig. 16 contains three elements: An electron pocket around C,
a hole pocket centred on the C–M lines of the SBZ and the Fermi surface segment near M 0 .
As for Bi(1 1 1) the bands in the immediate vicinity of C are very steep. The Fermi veloc-
ities are 3.0 eVÅ, 4.9 eVÅ and 1.5 eVÅ, for the electron pocket along C3 –M 3 , the electron
pocket along C3 –K 3 and for the hole pocket along C3 –M 3 , respectively.
The result of first-principles calculations for the electronic structure of Bi(1 0 0) is shown
in Fig. 17. The comparison to the measured dispersion is much less favourable than for the
other two surfaces. However, in most parts of the SBZ a qualitative connection between
theory and experiment can be established despite the poor quantitative agreement. The
regions of high photoemission intensity around C1 also show a high density of surface
or near-surface states in the calculation, the three Fermi level crossings near the Brillouin
zone centre in the C1 –M 2 direction are visible, as well as the fact that there are only two
such crossings in the C1 –M 0 1 direction. The experiment reveals an unclear feature at the
Fermi level exactly at M 0 whereas the calculation shows clearly separated Fermi level
crossings of the surface state around M 0 . At first glance, this appears to be quite different
214 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 15. Measured dispersion of the electronic states along the several directions in the SBZ. The green hatched
area is the projection of the calculated tight-binding band structure after Ref. [15]. The coloured horizontal bars
can be used to identify equivalent directions in the extended zone scheme (see Fig. 14). From Ref. [51]. (For
interpretation of colours in this figure legend, the reader is referred to the web version of this article.)

Fig. 16. Normalized photoemission intensity at the Fermi level (see text). Black corresponds to high intensity.
The solid line is the SBZ boundary. From Ref. [51].

but a small upward shift of the surface state energy at M 0 in the calculation would lead to
Fermi level crossings much closer to M 0 and in better agreement with the experiment.
The most important question which arises when comparing Bi(1 0 0) to the other two
surfaces, is why the calculations do not reproduce the measured dispersion as well as they
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 215

Fig. 17. First-principles calculation of the Bi(1 0 0) band structure. States with a large weight near the surface are
marked as filled red circles. Projected bulk bands are shown as a green hatched area. A tentative connection
between the surface states is marked by a red line. The coloured horizontal bars can be used for a direct
comparison with the equivalent directions in Fig. 15. From Ref. [51]. (For interpretation of colours in this figure
legend, the reader is referred to the web version of this article.)

Fig. 18. Charge density contours of surface states on the Bi(1 0 0) surface: (a) is a state from the middle of C1 –M 2
almost at the Fermi level, (b) is the highest occupied state at K 2 , and (c) is a state from the middle of the highest
occupied band in C1 –K 2 direction. The position of the surface plane (s) and of four lower layers are indicated by
the dashed lines. The contour lines are plotted on a logarithmic scale. From Ref. [51].

do on the other surfaces. The main cause for the limited accuracy of the calculation for
Bi(1 0 0) can be understood by inspecting the charge density contours of different surfaces
states shown in Fig. 18. Some states near the Fermi level penetrate very deeply into the
216 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

bulk (Fig. 18(a) and (b)) while others (Fig. 18(c)) are localized in the first five layers. This
penetration depth is remarkably high, compared to Bi(1 1 1) or Bi(1 1 0), where the surface
state is mostly localized in the first bilayer of the surface [69]. Also, the weight of some
Bi(1 0 0) surface states in the vacuum is rather low. In some cases the state even has its larg-
est weight in the fifth layer from the surface, even though it is clearly in a large projected
bulk band gap. This limits the possible accuracy of the theory which is restricted to a film
of finite thickness because the states from both sides of the film couple strongly.

4.4. General trends

All the three Bi surfaces described here drastically differ from the bulk in at least two
important respects. First, the surface Fermi contours are much larger than the (projected)
size of the bulk Fermi contour. This means that the surface carrier density is also much
higher than the corresponding bulk value. In the case of Bi(1 1 1), this has been quantified
for the surface states near the C point where the total surface carrier density is in the order
of 3 · 1013 cm2 [55]. This is obviously significant for a semimetal with a bulk carrier den-
sity of only 3 · 1017 cm3. The second important difference to the bulk electronic structure
is that the Fermi velocity of the surface states is generally lower than that of the bulk
states, also leading to an increase of the density of states at the Fermi level.
The metallic character of the surface is of high importance when considering Bi in small
systems, like clusters or thin films. Two important considerations apply: First, the bulk
electronic structure changes drastically when reducing the dimensions of Bi to values lower
than a few hundred nanometres, due to quantum-size effects. Second, for small structures
the metallic surface contribution becomes increasingly important. A good example for the
interplay of these two effects is the long sought semimetal to semiconductor transition for
thin Bi films described in Section 2.3. Given the importance of the surface states, it is dif-
ficult to find the transition because the electronic surface states dominate the density of
states at the Fermi level even for relatively thick films.
This simple distinction between surface and bulk properties holds for Bi(1 1 1) and
Bi(1 1 0) where the surface states are localized in the first layers [69]. Indeed, the strong sur-
face localization is the very reason why the first-principles calculations agree so well with
the experiment for these two systems. For Bi(1 0 0) the situation is different. There are also
metallic surface states but they penetrate deeply into the bulk. So for a thin film grown in
the (1 0 0) orientation, it is no longer possible to distinguish between surface effects and the
modification of the bulk electronic structure due to the finite size of the film.
In fact, the strong penetration of the surface states on Bi(1 0 0) could be important for
understanding some results concerning the electronic structure of Bi clusters and nano-
tubes. Take, for example, the superconductivity in granular films of Bi clusters reported
by Weitzel and Micklitz [44] (Section 2.3). While the initial explanation was based on
metallic cluster surfaces and a semimetallic bulk, this simple picture was not found to
agree with other experimental data [45]. Deeply penetrating surface states could well be
the cause for this apparent contradiction. States similar to those in Fig. 18(a) and (b) could
exist on the (1 0 0) surface of the clusters. Strictly speaking, the states are surface states and
therefore influenced by, e.g. the surroundings of the crystal. At the same time, they would
be delocalized over a cluster which is a only few nanometres in size, in the order of the film
thickness used to produce Fig. 18. Similar surface states could inhibit a predicted [40] semi-
metal-to-semiconductor transitions in Bi nano-wires [42].
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 217

There are still a number of open questions about the surface electronic structure of Bi.
One concerns the general picture of the electronic structure. It is conceivable that Bi(1 0 0)
and Bi(1 1 0) support partially filled surface states because of the formation of dangling
bonds upon surface formation, combined with the absence of a surface reconstruction
which could remove these dangling bonds. But for Bi(1 1 1) the situation is different: no
dangling bonds are formed and the surface creation can even be viewed as the mere sep-
aration of two bilayers which are not strongly interacting. Still the surface electronic struc-
ture is metallic and very different from the bulk electronic structure. Ast and Höchst have
explained this by strong surface relaxations but both the LEED experiment and the first-
principles calculations show that such relaxations are not there. Finally, a characteristic
feature of the Bi surfaces is the very steep dispersion of surface states around some high
symmetry points (C on Bi(1 1 1) and Bi(1 0 0) and M 0 on Bi(1 0 0)). Where does this come
from? All these questions will be answered in the next section which describes the influence
of the spin–orbit interaction on the surface band structure of Bi.

5. Spin–orbit coupling and splitting

Spin–orbit coupling (SOC) is a relativistic effect which is important for the electronic
structure of heavy atoms and their solids. An electron which moves in an electric field ~E
with a velocity ~
v experiences a magnetic field in its rest frame, which is given by
~ v~
B ¼ cð~ EÞ=c2 ; ð7Þ
where c is the factor (1  v2/c2)1/2. This magnetic field couples to the electron’s spin, giv-
ing rise to an energy term of the form:
h
 h

H SOC ¼ p ~
ð~ r¼
EÞ~ ðrV  ~ r;
pÞ~ ð8Þ
4m2 c2 4m2 c2
where V is the potential and ~ r is the Pauli spin operator. In atomic physics, a special form
of this correction for a spherical potential is frequently added to the (non-relativistic)
Schrödinger equation. It leads to a splitting of the states proportional to the product of
spin and orbital angular momentum operator, hence the name ‘spin–orbit interaction’.
For heavy atoms such as Au or Bi, the splitting can be quite large. For instance, the split-
ting of the 6p states, i.e. the energy difference between the 6p1/2 and 6p3/2 is 0.47 and 1.5 eV
in Au and Bi, respectively.
The SOC also causes the lifting of degeneracies in the bulk electronic structure of solids,
for example in Bi (see discussion in Section 2.2 and Refs. [18,28]). However, for solids with
inversion symmetry the interaction does not lead to a lifting of the spin degeneracy of the
bands, i.e. every band still contains two possible spin directions. The reason for this is
the combination of time-reversal symmetry and inversion symmetry. The former leads
to the so-called Kramers degeneracy between a Bloch wave w~k ð~ r; sÞ and its complex con-
jugate w~k ð~
r; sÞ. To arrive at the complex conjugate both the wavevector and the spin have
to be reversed. This means that there is a degeneracy in the band structure between states
with energy Eð~ k; "Þ and Eð~k; #Þ. Inversion symmetry, on the other hand, requires that
Eð~k; "Þ ¼ Eð~ k; "Þ. From the combination of both it is evident that for every state with
Eð~k; "Þ there is also a state with Eð~ k; #Þ.
Even for materials with inversion symmetry in the bulk, this symmetry is lost at the sur-
face and a ~ k k -dependent splitting of the bands can be expected. In the most simple picture
218 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

the surface state can be viewed as a two-dimensional free electron gas with a potential gra-
dient perpendicular to the surface. The electrons moving in the surface plane experience a
magnetic field which also lies in the surface plane and perpendicular to the direction of
propagation, i.e. perpendicular to ~ k k . In this model, the electron wave functions and the
size of the splitting can be found analytically [70]. For k = 0 the splitting vanishes, consis-
tent with the symmetry requirement E(0, ") = E(0, #). Away from ~ k k ¼ 0, two eigenstates
for each ~k k can be found which differ in energy and spin direction. The spin of both states
lies in the surface plane and perpendicular to ~ k k , i.e. it is either parallel or antiparallel to
the magnetic field the electron experiences. The energy splitting is proportional to the mag-
nitude of ~ k k . This situation at surfaces strongly resembles that in semiconductor hetero-
structures where the so-called Rashba effect leads to a similar splitting [71].
Spin–orbit splitting for surface states was found first for the free-electron like surface
state on Au(1 1 1) by LaShell et al. [72]. As expected from the simple model outlined above,
the surface state dispersion is modified by a spin-dependent linear term proportional to the
magnitude of ~ k k . The spin–orbit splitting of the Au(1 1 1) surface state was later studied in
considerable detail, both theoretically [70,73–75] and experimentally [73,74,76]. In fact, it
was even possible to demonstrate the spin-splitting directly using spin-resolved photoemis-
sion [77]. Other cases of spin–orbit splitting in surface state bands have also been reported,
for example on the Li-covered surfaces of W and Mo and for H on W(1 1 0), also studied
by spin-resolved photoemission [78].
At this point it is worth noting that the simple free electron model for the spin–orbit
splitting of surface states has some severe limitations. Firstly, it does not correctly predict
the size of the splitting which is still mainly related to atomic properties. This can easily be
seen from the case of the noble metal surfaces; the splitting of the Au(1 1 1) free electron
surface state can be routinely observed by state-of-the-art angle-resolved photoemission
whereas the predicted splitting for the same surface state on Ag(1 1 1) is far too small to
be observed at present [73]. The simple picture of electrons with a certain spin moving
in the surface plane is also problematic because, strictly speaking, the spin is not a good
quantum number for heavy atoms, only the total angular momentum.
The most interesting application for spin-split states lies arguably in the emerging field
of spintronics, a new type of electronics which is based on the electron’s spin degree of
freedom in addition to, or instead of, its charge. A key-ingredient of spintronics are
sources and filters of spin-polarized currents and to this date, the most promising
approach to build these is based on the Rashba-splitting in semiconductor heterostruc-
tures [79,80]. Surface or interface states with sizeable splitting could be an alternative
but most examples studied so far suffer from two restrictions: The splitting is not very
big and, more importantly, the split states contribute only very little to the total density
of states (at the Fermi level) so the splitting will not show up in transport properties or
in resonant tunnelling devices such as the one proposed by Koga et al. [80]. The surfaces
of Bi have the potential to solve both issues; the spin–orbit splitting in the atomic 6p levels
is three times stronger than in Au and the density of states close to the Fermi level is dom-
inated by metallic surface states for most surfaces, as discussed in Section 4.

5.1. Splitting of the surface state bands

The possibility of a strong spin–orbit splitting of the Bi surface state bands was first dis-
cussed by Agergaard et al. in connection with the electronic structure of Bi(1 1 0) [67]. The
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 219

problem, however, was that no clear indication of split states were found in the non-spin-
resolved data: No occupied surface states were present at the high-symmetry points were
the splitting would have to be zero and away from these points it was not possible to easily
identify pairs of bands split by the SOC.
The proof of a strong surface state splitting was provided by combining experimental
data from angle-resolved photoemission with first-principles calculations [65]. These calcu-
lations can be performed with and without the inclusion of the spin–orbit interaction and
they can also reveal the dispersion of the unoccupied surface states above the Fermi level.
Fig. 19 shows the electronic structure of Bi(1 1 1) calculated with and without SOC
together with the projected bulk band structure for the (1 1 1) surface from Koroteev
et al. [65]. The latter was calculated using the tight-binding model of Ref. [15], the former
from first-principles. In the case without SOC, a parabolic surface state is located around
C in the non-relativistic energy gap. This surface state band gives rise to a hexagonal elec-
tron-like Fermi surface element. When the SOC term is included in the calculation it
results in a spin-splitting of the surface state in all directions and leaves it degenerate only
at C and at M. Very close to C the relativistic surface state bands are degenerate with bulk
states and show less clear surface character.
The lift of the spin degeneracy leads to radical change of the surface Fermi surface:
First, the radius of the electron pocket Fermi surface is smaller by 30% compared to
the non-relativistic calculation; second, in the C–M symmetry direction the hole pockets
are formed (see also Fig. 10). Another remarkable feature of the Bi(1 1 1) surface electronic
structure is the very strong anisotropy of the spin–orbit splitting: it is 0.2 eV in the C–M
direction and even more along C–K. Finally, the strong spin–orbit splitting gives a qual-
itative explanation for the fact that Bi(1 1 1) is metallic despite of the lack of dangling
bonds. The SOC-caused splitting leads to a very steep dispersion of the bands near C
which, in turn, leads to Fermi level crossings of the bands.

Fig. 19. Surface states of Bi(1 1 1) calculated without (black) and with (red) spin–orbit splitting included. The
shaded areas show the projection of the bulk bands obtained without (violet) and with (yellow) SOC and their
superposition (brown). After Ref. [65]. (For interpretation of colours in this figure legend, the reader is referred to
the web version of this article.)
220 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Exactly at M the situation is somewhat unclear. A degeneracy of the bands as shown in


Fig. 19 is expected because M is a high-symmetry point on the SBZ boundary which lies
half-way between two C points, e.g. in contrast to K. This means that there are two rele-
vant symmetry elements. Time-reversal symmetry assures that EðM; "Þ ¼ EðM; #Þ.
Translational symmetry requires that EðM; "Þ ¼ EðM; "Þ because M and M are sepa-
rated by a reciprocal lattice vector. The experimental situation shown in Fig. 11(b) and
(c), however, does not appear to confirm this. One reason for this difficulty could be that
the surface state bands lie in the bulk continuum close to M and are therefore actually sur-
face resonances. Spin-resolved photoemission experiments from the states observed
around M could help to clarify their character.
The ultimate proof of the strong spin–orbit splitting lies in the comparison with the
experimental findings. As seen in Fig. 19, the surface state dispersion is radically different
for the cases with and without SOC. A comparison to the experimental dispersion can
therefore immediately reveal if the spin–orbit splitting picture is correct. Such a compar-
ison is shown in Fig. 20, also from Ref. [65]. The figure shows the calculated and measured
surface state dispersion for three different surfaces in the vicinity of high symmetry points.
In all cases the agreement between the calculated and measured dispersion is very good. In
addition to this figure, the calculated surface state dispersions for Bi(1 1 1), Bi(1 1 0) and
Bi(1 0 0) shown in Section 4 all include the spin–orbit coupling. Their good agreement with
the experiment clearly proves the validity of the strong splitting.
The only case where the agreement between the calculated and measured surface state
dispersion is less good is that of Bi(1 0 0), for reasons explained in Section 4. In this con-
text, it is also worth mentioning that the calculated dispersion of Bi(1 0 0) in the vicinity of
the M 1 point is slightly different in Figs. 20 and 17. The reason is that the surface geometry
has been allowed to relax to the lowest total energy in the calculation of Ref. [51] whereas
it has been kept in the truncated bulk geometry in Ref. [65]. This difference is only of
minor importance. In fact, both calculations show the very steep dispersion of the bands
close to the symmetry point which is characteristic for the strong SOC.

Fig. 20. Calculated and measured electronic structure in the vicinity of two high symmetry points on three
surfaces of Bi. (a) C on Bi(1 1 1), (b) C on Bi(1 1 0), and (c) M 1 on Bi(1 0 0). The small black dots are the projected
bulk band structure calculated using the tight-binding model of Liu and Allen [15]. The red filled circles are the
calculated surface state energies, thin red line is a guide to the eye. After Ref. [65]. (For interpretation of colours
in this figure legend, the reader is referred to the web version of this article.)
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 221

5.2. Quasi-particle interference

From the foregoing discussion it is evident that the surfaces of Bi are not only quasi-
two-dimensional metals but that they also have rather interesting properties with respect
to their spin. This has opened the possibility to use them as a test case to study the influ-
ence of the spin on the so-called quasi-particle interference observed by scanning tunnel-
ling microscopy (STM), a tool of increasing importance in the fields of electronic structure
and nano-technology.
The term quasi-particle interference is used to describe periodic modulations in the local
density of states on surfaces which are caused by the interference of electrons, or quasi-
particles, in the vicinity of defects. These modulations are closely related to the Friedel
oscillations [81] which appear in the Lindhard picture of screening in metals [82]. One pos-
sible approach to investigate the quasi-particle interference is the observation of oscilla-
tory patterns in the local density of states by STM [83]. STM conductance images taken
at a certain bias voltage are analysed by Fourier transformation (FT) to give detailed
information about the electronic structure of the surface in k space [84–87]. Analysing
quasi-particle interference patterns has found a wide range of applications such as probing
the electronic structure of nano-scale objects [88], contacting of molecular wires [89], as
well as measuring the superconducting energy gap function and possible local ordering
on high-temperature superconductors [90,87,91].
In a simple picture, a quasi-particle interference pattern near EF arises because a quasi-
particle (or an electron) with wavevector ~ k F encounters a defect such as an impurity or a
step edge and is reflected into a state with wavevector ~ k F . The interference of incoming
and reflected waves gives rise to a modulation in the local density of states with a period-
icity of 2~k F , i.e. with the vector connecting the two states. In many cases the FT of
conductance images taken with small bias voltages can therefore be interpreted as a direct
image of the Fermi contour, magnified by a factor of two [84–86]. On high-temperature
superconductors, similar interference patterns are observed but the intensities in the FT
maps are strongly influenced by the density of the states causing the scattering events [87].
An aspect which has been almost completely neglected in the analysis of quasi-particle
interference patterns is the fact that the quasi-particles can also have a spin. The notable
exception is the case of Au(1 1 1) which has been studied in great detail by Petersen and
Hedegård [70]. As described above, the free-electron like surface state on Au(1 1 1) is split
by the SOC. This results in a small change of the Fermi surface. Instead of a simple circle,
the Fermi surface consists of two concentric circles with Fermi wavevectors different by
0.025 Å1 [73]. One could argue that this small splitting should be detectable by analysing
the quasi-particle interference in a Fourier-transformation of a sufficiently large STM
image but a careful experiment revealed only one circle with a radius corresponding to
the average of the two concentric circles [92]. This was explained by the fact that quasi-
particles with opposite spin cannot give rise to observable interference patterns [70].
Indeed, keeping in mind the simple picture for quasi-particle interference outlined above,
one can argue as follows: The characteristic features in quasi-particle interference near the
Fermi level are created by the scattering of states at ~ k F into ~ k F . In the case of strong
spin–orbit splitting, these two states will have opposite spins because of time-reversal
symmetry and the wave functions will therefore be orthogonal. Hence, all of the interfer-
ence processes building up the Fermi surface topology in the FT maps should be forbidden
[70].
222 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

The quasi-particle interference on Bi(1 1 0) was studied by Pascual et al. [68]. To lay the
ground for understanding their results, Fig. 21 shows a sketch of the surface Fermi surface
of Bi(1 1 0) with an indication of the spin directions (see also Fig. 13). As in the free elec-
tron model, the spin is taken to lie in the plane of the surface and perpendicular to the
propagation direction. The spin directions can be worked out by inspecting the band
dispersion in Fig. 13. For example, the hole pocket states near C and M are formed by
the lower branch of the spin–orbit split bands. Therefore, the relative orientation between
the two-dimensional momentum ~ k and the spin has to be the same for both states. Hence,
the indicated direction of the spin has to rotate anti-clockwise and clockwise around the C
and M contours, respectively. For the electron pocket along M–X 1 the spin polarization
merely ‘‘wiggles’’ around the preferred direction because of the need to take into account
the symmetry-requirement imposed by the crossing of the SBZ boundary.
Fig. 22(a) and (b) shows an STM conductance image of Bi(1 1 0) taken at V = +40 mV
together with its Fourier transformation [68]. The bias voltage is sufficiently small to allow
a direct comparison of the observed interference features with the Fermi surface. Fig. 22(c)
shows a schematic drawing of the FT map (coloured features) together with the expected
interference pattern for a spin-independent situation (grey dashed lines). It is evident that
the latter is totally absent from the FT map, in agreement with the simple picture of quasi-
particle interference outlined above. Instead, a number of other structures (A, B, C) are
observed in addition to the so-called lattice spots which arise from the lattice periodicity.
Finally, Fig. 22(d) illustrates the origin of these features which are interpreted as spin-con-
serving scattering events.
These results confirm the analysis by Petersen and Hedegård [70] that the spin-forbid-
den scattering events do not contribute to the quasi-particle interference pattern. They also
illustrate that the actually observed interference pattern results from spin-conserving scat-
tering events. The pattern has to be interpreted in a very different way compared to a
situation in which the spin is not an issue. It cannot be viewed as an image of the two-
dimensional Fermi contour but together with this Fermi contour it can provide informa-
tion about the spin orientation of the states. These statements are also true for the case of
Au(1 1 1) where the single circle observed in FT transformed images [92] does not represent
the actual Fermi contour but a spin-conserving average. But the situation on Bi(1 1 0) is

X1 M

X2
Γ

Fig. 21. Sketch of the surface Brillouin zone and the Fermi surface of Bi(1 1 0) with an indication of the
approximate spin direction. After Ref. [68].
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 223

Fig. 22. (a) Differential conductance map (V = 40 mV; I = 1 nA) and (b) its FT map. (c) Schematic drawing of
the FT map. The dashed grey lines are the modulations which would be expected for quasi-particle interference
from a Fermi surface in Fig. 21(b) if the spin was not important. The features A, B and C are the actually
observed non-lattice-periodic structures. The yellow markers represent the lattice spots. (d) Illustration of the
spin-conserving scattering events causing features A, B and C. The dashed scattering vectors have to be translated
to the origin and scaled down by a factor of 2 to yield the features in (c). From Ref. [68]. (For interpretation of
colours in this figure legend, the reader is referred to the web version of this article.)

more drastic: due to the strong splitting the observed features do not even resemble the
Fermi contour anymore. Indeed, this effect can be taken as a direct proof of the surface
states’ spin–orbit splitting, supplementing and reinforcing the information from photo-
emission and calculations.
Finally, these findings suggest that the most appropriate way of interpreting STM
images is that they show the electronic response of the surface to small perturbations,
i.e. that they are a direct image of the screening of defects. This interpretation will be used
in the next section when the possibility of a charge density wave on Bi(1 1 1) is discussed.

5.3. Possibility of a charge density wave on Bi(1 1 1)

Reducing the dimensions of a metal from three to two or one leads to an increased ten-
dency for instabilities such as charge density waves (CDW) [93]. In fact, there have been sev-
eral cases of CDWs reported on metal surfaces over the recent years [94,95], and even typical
semiconductor surface reconstructions which turn the surface from a metal into an insulator
can be viewed as CDWs [96]. It can be tempting to argue that the surfaces of Bi are partic-
ularly good candidates for CDWs because they are very nearly two-dimensional metals. In
fact, Ast and Höchst have reported a spectroscopic indication of a charge density wave for-
mation on Bi(1 1 1)[64]. Using high-resolution angle-resolved photoemission, they have
224 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

found that the Fermi surface contour of the electron pocket state around C has a hexagonal
shape and therefore fulfils the condition of good ‘‘nesting’’ (for details see Section 4.1). They
have also found that the leading edge of the spectra taken at the Fermi surface undergoes a
discontinuous shift of about 5 meV as the sample temperature is lowered from 75 K to 50 K.
This was interpreted as the opening of a gap, caused by the formation of a CDW. However,
it was also pointed out that the mere existence of a leading-edge gap would not necessarily
imply a CDW ground state. Other techniques would be needed to confirm this.
The most remarkable fact about such a CDW on Bi(1 1 1) is that it should not be per-
mitted because of the role of the spin. A key ingredient in the formation of a CDW is a
singularity in the Lindhard susceptibility, which leads to the unstable situation. Such sin-
gularities can be created by a so-called ‘‘nesting’’ of the Fermi contour, i.e. by having
many states at the Fermi contour separated by the same wavevector 2~ k F . In fact, the phys-
ics of CDW formation is very similar to the screening response needed for quasi-particle
interference and which is discussed in the preceding section. The only difference is that the
response of the system is confined in the case of quasi-particle interference whereas it is
unconfined and created by a vanishingly small perturbation in the case of a CDW. In
the particular case of Bi(1 1 1), the states on opposite sides of the hexagonal electron
pocket have opposite spins. Therefore the pairing needed for a CDW formation is forbid-
den but the possibility of a spin-density wave would still exist [65,93].
Kim and co-workers have tried to address two questions; is there a CDW on Bi(1 1 1)
and if not, can the leading edge gap in the data from Ast and Höchst be reproduced [97]?
To this effect, they have used three different experimental techniques: STM, transmission
electron microscopy and angle-resolved photoemission from two different set-ups. The
main conclusion from their work is that there is no CDW on Bi(1 1 1) and that the opening
of a leading edge gap in the photoemission spectra cannot be confirmed.
The strongest and most direct proof against a CDW is arguably the failure to observe it
in STM. Fig. 23(a) shows an STM constant current image taken at a temperature of 5 K,
a tunnelling voltage of 13 mV and a current of 1.1 nA. The image clearly shows the atomic
corrugation corresponding to the hexagonal lattice of Bi(1 1 1), but no sign of a charge
density wave. In order to illustrate this, Fig. 23(b) shows a simple model of the
Bi(1 1 1) atomic structure with the superimposed modulation of the wavevector ~ qCDW ,
i.e. the wave vector needed to connect the two sides of the hexagonal electron pocket
shown in Figs. 10 and 24. The STM image does not resolve such a CDW modulation even
for tunnelling into states close to the Fermi level. This can be taken as solid evidence
against the presence of a CDW on Bi(1 1 1), since CDWs usually show up very clearly
in constant current STM images [98–101]. It is also interesting to note that earlier low-
temperature STM investigations of Bi(1 1 1) did not report the observation of a CDW
superstructure either, although the precise temperature in these experiments is not known
[102,103]. Further evidence comes from detailed STM conductance measurements and
their analysis, also shown in Fig. 23. The observed quasi-particle interference pattern
can be interpreted along exactly the same lines as for Bi(1 1 0), confirming the spin-split
character of the bands.
The analysis of diffraction patterns from transmission electron microscopy for Bi(1 1 1)
samples of different thickness and at different temperatures lead to the same conclusion,
that a superstructure due to a CDW could not be confirmed.
Finally, Kim et al. re-investigated the issue of a possible gap opening at low tempera-
tures and how this would be reflected in the photoemission spectra. Data have been taken
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 225

Fig. 23. (a) Typical constant current image of Bi(1 1 1) showing the atomic corrugation of the surface
(Vs = 13 mV, I = 1.1 nA, 19 · 17 nm2), (b) Model of a CDW modulation superimposed to the Bi(1 1 1) atomic
periodicity, in a similar surface area as (a). (d) Differential conductance image (Vs = 19 mV, Vac = 6 mV
rms, I = 1.1 nA, 21 · 19 nm2) and (c) its two-dimensional Fourier transformation. (e) Schematic representation of
the Fourier transformed image. The dashed grey lines are the modulations which would be expected for a spin-
independent quasi-particle interference from the Fermi surface in Fig. 10(a). Feature A is the only non-lattice-
periodic structure as seen in (c). The yellow markers represent the lattice-periodic structure. (f) Illustration of the
Fermi surface with the approximate spin direction indicated. The transition giving rise to the ‘A’ features is
indicated. The dark arrow illustrates one of the many possible small-j~ kj transitions which do not violate the spin
conservation. From Ref. [97]. (For interpretation of colours in this figure legend, the reader is referred to the web
version of this article.)

on two different set-ups: In Aarhus using synchrotron radiation, moderate energy resolu-
tion (better than 26 meV) and high angular resolution (±0.2) as well as in Dresden using a
laboratory UV source, with high energy resolution (13.5 meV), moderate angular resolu-
tion (±0.7) and the ability to take a large range of emission angles simultaneously. All the
results are shown in Fig. 24.
Fig. 24(a) shows the photoemission intensity at the Fermi level for a grid of ~ k points
close to C, revealing the hexagonal shape of the electron pocket. Fig. 24(b) shows a
momentum distribution curve (MDC) taken at the Fermi level along C–K together with
a fit using a Lorentzian line and a linear background. These results are in good agreement
with the data from Ast and Höchst [64] even though the MDC linewidth of the electron
pocket at the Fermi level is slightly smaller in the work of Kim et al. Fig. 24(c) shows a
series of energy distribution curves (EDCs) taken exactly at the Fermi level crossing of
226 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 24. ARPES data taken in Aarhus: (a) Large image: Photoemission intensity at the Fermi level (Fermi surface
map) around normal emission, revealing the hexagonal shape of the electron pocket (hm = 15 eV and T = 30 K).
Smaller image: Large scale Fermi surface map to determine sample orientation and work function (hm = 25 eV and
T = 30 K). (b) MDC taken at the Fermi level for the crossing of the electron pocket in the C–K direction together
with a Lorentzian fit (hm = 15 eV and T = 30 K). (c) EDCs taken at the Fermi level crossing in the C–K direction as
a function of temperature. (d) ARPES data taken in Dresden: Temperature-dependent dispersion of the surface
states along K–C–K (hm = 21.2 eV). The band crossing the Fermi level is the electron pocket. The band at a binding
energy of about 100 meV is the state which forms the hole pockets in the C–M direction. From Ref. [97].

the electron pocket in the C–K direction, as a function of temperature. The EDCs consist
of one clear peak and a shoulder. The peak at a binding energy of about 80 meV stems
from the surface state band forming the hole pocket. The shoulder close to the Fermi level
corresponds to the electron pocket. It does not show up as a clear peak because its max-
imum lies at the Fermi level. As a consequence of this, the Fermi edge appears to be much
broader than expected from the energy resolution, a fact which is well-known from other
studies [104,105]. Most importantly, there is no sign of a leading-edge gap opening at the
lowest temperatures. Fig. 24(d) shows the temperature-dependent dataset taken in
Dresden. Also in these data, no gap-opening at the Fermi level crossing can be observed.
The information in the Dresden data goes beyond the temperature-dependent EDCs from
Fig. 24(c) because the entire dispersion at each temperature is measured. If there was a
CDW one should not only measure a gap opening but one could also expect to observe
a change of the dispersion close to the Fermi level (for a recent example see [106]).

6. Electron–phonon coupling

6.1. Introduction

The electron–phonon coupling (epc) is an important many-body interaction in metals


which has been studied in great detail between 1960 and 1980. Most of the results from this
period are summarized in the comprehensive book by Grimvall [107]. In the most simple
picture, the epc changes the dispersion and the lifetime of the electronic states in a material.
This situation is illustrated in Fig. 25. Very close to the Fermi energy EF, within a typical
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 227

energy (arb. units)


Σ''

Σ'
0
hωD 0
energy
E/EF
1

hωD
E(k)
ε(k)

1.00
k/kF

Fig. 25. Re-normalization of the electronic dispersion close to the Fermi energy (schematic). The dashed line is
the bare dispersion in the absence of the electron–phonon interaction and the solid line is the re-normalized
dispersion. Inset: Real and imaginary part of the complex self-energy for the electron–phonon coupling, R 0 and
R00 . After Ref. [108].

phonon energy ⁄xD, the dispersion is re-normalized such that it is flatter at the EF. Conse-
quently, the effective mass of the electrons at EF and the density of states is increased [107].
The increase of the effective mass is described by the electron–phonon mass enhancement
parameter k such that m* = m0(1 + k) where m* and m0 are the effective masses with and
without epc, respectively. The effect of the epc on the dispersion and lifetime of the states
can be expressed by the complex self-energy R, where the real part R 0 re-normalizes the dis-
persion and the states acquire a finite lifetime through the imaginary part R00 . R 0 and R00 are
related by a Kramers–Kronig transformation. R 0 is small except for energies very close to
EF. R00 is changing rapidly close to EF and is constant at higher energies. R 0 vanishes exactly
at EF such that the Fermi surface is not affected by the interaction. At T = 0, R00 is also zero
at EF and the lifetime of the quasi-particles exactly at EF becomes infinite (see Fig. 25).
On the scale of the electronic state dispersion the re-normalization shown in Fig. 25 is a
rather small effect. However, the epc is still extremely important because it affects the elec-
trons near EF, i.e. the states which are relevant for many of the metal’s properties. The
consequences of the epc are found in the transport properties, in the electronic heat capac-
ity and, most spectacularly, in BCS-type superconductivity.
In recent years, the epc has seen a renewed interest. This has several reasons but the most
important one is probably the discovery of high-temperature superconductivity and the
experimental progress in angle-resolved photoemission (ARPES) which was triggered by
this. ARPES is a particularly valuable technique to learn something about many body inter-
actions like the epc: given a sufficiently high resolution, detailed information about the real
and imaginary parts of the self-energy can be obtained in a energy and k-resolved fashion.
This means that ARPES permits a spectroscopic approach to the interactions, unlike, e.g.
the measurement of transport properties which implies averaging over the whole Fermi sur-
face. A lucky coincidence is also that quasi-two-dimensional systems permit a simple inter-
pretation of ARPES measurements in terms of the initial state spectral function (see below)
228 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

and the high-temperature superconductors happen to be such systems. In fact, ARPES has
already lead to many important results in this field [109]. Based on ARPES data it has even
been suggested that the epc could play an important role in the mechanism of high-temper-
ature superconductivity [110,111], but this point is subject of heavy debate (see for example
Ref. [112]). Along with these studies, ARPES has also been used to investigate the many-
body effects in electronic surface states on clean and adsorbate-covered metals surfaces
and in particular the electron–phonon coupling [113,9,114–120]. Surfaces states are also
quasi-two-dimensional and thus well suited for ARPES investigations.
Investigating the epc on Bi surfaces has been motivated by several factors. Firstly, the
simple quasi-two-dimensional metallic character of the low-index surfaces sets the stage
for an apparently simple test system which permits studying the spectroscopic signature
of the epc in a two-dimensional metal. Secondly, the research on superconductivity in
granular systems of rhombohedral Bi clusters by Weitzel and Micklitz [44,45] suggests that
the surfaces could play an important role in the superconductivity. Obviously this raises
questions about the strength of the electron–phonon coupling on the surfaces and if per-
haps the surfaces could be superconducting by themselves. Finally, it is well known that
the electron–phonon coupling parameter depends strongly on the Bi structure. For amor-
phous Bi, k is very large (k = 2.46) [121,107]. For rhombohedral single crystals only a
rough estimate based on the calculated density of states at EF [15] and the low-temperature
heat capacity [122] is possible. In any event, the resulting k is small (k = 0.13(0.13 + 0.2))
[123]. It seems therefore worthwhile to study the epc on the Bi surfaces which have very
different properties than the bulk.
Before we discuss the particular results for Bi surfaces, it is appropriate to give a brief
overview on how information about the epc can be obtained by ARPES. We summarize
only the most essential facts without trying to be concise. Several detailed treatments on
ARPES [124–128] and on the relation between ARPES and epc [129–133] can be found
elsewhere. We assume that ARPES from a quasi-two-dimensional system measures the ini-
tial state spectral function A, convoluted with the appropriate functions for the energy
and angular resolution. For A we assume

p1 jR00 ðx; T Þj


Aðx; ~
k; T Þ ¼ 2 2
; ð9Þ
hx  ð~
½ kÞ  R0 ðx; T Þ þ R00 ðx; T Þ

where hx is the energy, ~


k is the two-dimensional crystal momentum vector, T is the tem-
perature and ð~kÞ is the single-particle dispersion. R 0 and R00 are the real and imaginary
part of the self-energy. The self-energy depends on the energy but here it is assumed
not to depend on ~ k.
If we further assume that the energy and angular resolutions of the spectrometer are
sufficiently high to be neglected, the measured photoemission intensity, as a function of
emission direction or energy or temperature, merely represents a cut through the spectral
function given in (9).
The most important measurement modes in ARPES are energy distribution curves
(EDC) where the photoemission intensity is measured as a function of kinetic energy at
a fixed emission angle, and momentum distribution curves (MDC) where the kinetic
energy is kept constant and the emission angle (or ~ k) is scanned.
One could be tempted to describe an EDC as a cut through the spectral function at a
constant ~k but this is not quite correct because it is the emission angle which is held
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 229

constant and not ~ k. Under certain conditions, this problem can be neglected
[134,135,117,133], e.g. near normal emission or near a ~ k point for which the state in ques-
tion shows an extremum in its dispersion. Even if we assume that an EDC is taken at con-
stant ~
k, such a cut through (9) is fairly complicated. A simple equation is obtained when we
further assume that that R 0 (x, T) = 0 and that R00 (x, T) does not depend on x. Then we get
p1 jR00 ðT Þj
Aðx; ~
k; T Þ ¼ 2 2
; ð10Þ
½hx  ð~
kÞ þ R00 ðT Þ

which is a Lorentzian with the maximum at ð~ kÞ and a full-width at half-maximum


(FWHM) of 2jR00 (T)j. This approximation is relevant in the case of the electron–phonon
coupling, as long as the binding energy of the peak is not too small (see Fig. 25). Another
difficulty one has to keep in mind is the cut-off by the Fermi function at higher tempera-
tures. So even if the peak binding energy is much higher than typical phonon energies, it
might be necessary to fit the data including a Fermi function.
The situation is simpler in case of MDCs because they are readily represented by (9).
The maximum of an MDC is reached when hx  ð~ kÞ  R0 ðx; T Þ ¼ 0. Based on this, the
renormalized dispersion is defined as the self-consistent solution of

Eð~
kÞ ¼ ð~
kÞ þ R0 ðEð~
kÞ; T Þ. ð11Þ

Eq. (9) takes on a particularly simple form in the case of a linear dispersion. We consider
only one direction in ~
k space and write (k) = vk such that the origin of the co-ordinates is
at the Fermi level crossing. Then it is easy to show that (9) is a Lorentzian line in k for a
given x with the maximum at
hx  R0 ðxÞÞ
k max ¼ ð1=vÞð ð12Þ

and
FWHM ¼ 2jR00 ðxÞ=vj. ð13Þ
The practical consequences of this can be summarized as follows: for states which are far
away from the Fermi energy (with respect to typical phonon energies) R00 , the inverse life-
time, can be determined directly from the EDC linewidth, provided that the EDC can be
treated as if it was measured at constant ~ k. If it cannot, a simple correction may be applied
[114,117,123]. For states close to EF such a simple interpretation of the EDC lineshape is
not possible and MDCs are to be preferred. Assuming a (locally) linear dispersion of the
state, the MDC lineshape is always Lorentzian but to extract R00 , the bare dispersion of the
state has to be known.
After having briefly sketched two approaches to obtain information about R00 and
thereby about the epc, we give some basic equations defining R00 . If the only relevant
many-body interaction is the epc, it is given by
Z xm
00
jR ðx; T Þj ¼ ph dx0 a2 F ðx0 Þ½1  f ðx  x0 Þ þ 2nðx0 Þ þ f ðx þ x0 Þ; ð14Þ
0

where f and n are the Fermi and Bose distribution functions, respectively and xm is the
maximum phonon frequency. The basic function describing the epc (14) is the so-called
Eliashberg coupling function a2F(x) which describes the transition probability to a state
~kf from a state ~ki involving a phonon of energy ⁄x. The expression a2F(x) suggests that
230 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

the Eliashberg function is a product of a (squared) matrix element and the phonon density
of states. Indeed, in the quasi-elastic approximation a2F(x) can be written as [120,129]:
X ~q;m 2
a2 F ~ki ðxÞ ¼ jgi;f j dðx  x~q;m Þdð~ki  ~kf Þ; ð15Þ
~
q;m;f

which, if we assume a constant matrix element jgi, fj, is simply a sum over the different pos-
sibilities to fill a hole state with an electron using a phonon to provide energy and momen-
tum (the small phonon energy is neglected in the d function for the energy conservation).
In general, the electron–phonon mass enhancement parameter can be evaluated from
the Eliashberg coupling function by
Z xm
k¼2 dx0 a2 F ðx0 Þ=x0 . ð16Þ
0

If the aim of the experimental study is the determination of k, R00 has to be determined as
outlined above and then fitted to (14). In order to do this, simple models for a2F(x) have
to be used. Of particular importance is the three-dimensional Debye model in which
 2
2 x
a F ðxÞ ¼ k ; x < xD ;
xD ð17Þ
a2 F ðxÞ ¼ 0; x > xD ;
where xD is the Debye energy, but other models have been used as well [129].
A particularly useful equation is the high-temperature limit of (14) which is
R00 ¼ pkk B T ; ð18Þ
and therefore does not depend on the characteristics of the phonon model anymore.
Finally, the electron–phonon coupling may not be the only process which limits the life-
time of a state. The two most important other decay mechanisms are the electron–electron
interaction (Auger decay) and the electron–impurity interaction in which an electron is
scattered elastically into a hole state, receiving momentum from the defect. In most cases,
these two can be assumed to have a negligible temperature dependence such that their
effect can be described as a constant off-set to the self energy (14) and (18). The energy
dependence of the electron–electron interaction is more complicated but usually more
gradual than the rapid change of the electron–phonon self-energy close to EF [113,115].

6.2. Bi(1 0 0)

The epc on Bi(1 0 0) was studied by Gayone et al. by studying the EDC linewidth of a
surface state as a function of energy and temperature [123]. Keeping in mind the results
from the previous section, this approach is only possible for binding energies much larger
than a typical phonon energy. For Bi this is not a problem because the maximum phonon
energy is very small, only 13.8 meV [136]. The surface state used for the study was the state
in the C–K 2 direction which has two local extrema in the dispersion, a maximum at a bind-
ing energy of 330 meV and a minimum at 70 meV (see Fig. 15). For the states in between
these extrema, the EDC linewidth cannot be interpreted directly in terms of R00 but it has
been corrected accordingly to permit this interpretation.
By performing temperature dependent measurements of the linewidth on many points
of this surface state band, the strength of the epc was extracted for many different binding
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 231

34 K
-1 -1
K|| = 0.34 Å K|| =0.78 Å
55 K

74 K

94 K
114 K
Photoemission Intensity (arb. units)

134 K
35 K
154 K
54 K
74 K 174 K
94 K 194 K
114 K
134 K 213 K
154 K 233 K
174 K
253 K
194 K
214 K 274 K

234 K 293 K
254 K
274 K
294 K
0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.6 0.5 0.4 0.3 0.2 0.1 0.0
Binding Energy (eV) Binding Energy (eV)

Fig. 26. Temperature dependence of the surface state on Bi(1 0 0) for different binding energies. The solid lines are
the result of a peak fitting procedure. From Ref. [123].

energies. Fig. 26 shows the temperature-dependent EDCs for the two extrema in the dis-
persion. From this plot of the raw data it is already evident that the state at the higher
binding energy broadens more strongly than the low binding energy state as the temper-
ature is raised.
In order to obtain a quantitative description, the peaks in the EDCs have been fitted
using Lorentzian lines and the temperature dependent linewidth for several states is plot-
ted in Fig. 27. For each binding energy the temperature dependent linewidth has been fit-
ted using (14) plus a temperature independent offset. The Eliashberg coupling function has
been described using a Debye model with a Debye temperature of 50 K. The fit functions
are close to being straight lines, due to the low Debye temperature and the fact that (14)
can be approximated well by (18).
The final result of the analysis is the electron–phonon mass enhancement parameter k
as a function of binding energy as shown in Fig. 28. The energy dependence of k is very
strong; it changes from 0.72 to 0.20 in an energy range of less than 300 meV. From this
it is evident that k determined in this way cannot be interpreted as the mass enhancement
parameter at EF. A spectroscopic interpretation as a parameter measuring the epc strength
at a certain binding energy and ~ k is more appropriate.
The strong energy dependence of k was essentially explained by the transition from a
three-dimensional system at high binding energies (bulk plus surface) to a merely two-
dimensional system close to EF. The Eliashberg function (15) is a sum over the different
possibilities to fill a hole state with an electron using a phonon to provide energy and
momentum. If we neglect the small phonon energy in the process and assume the matrix
element for the scattering process to be only weakly energy dependent, the change of k
232 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

-1
BE=330 meV K ||=0.34 Å
λ=0.72 ± 0.05
-1
0.20 BE=280 meV K ||=0.42 Å Γ0 = 0.090 ± 0.005 eV
-1
BE=180 meV K ||=0.54 Å
-1
BE=70 meV K ||=0.78 Å

0.18

0.16
λ =0.45 ± 0.04
Γ0 = 0.080 ± 0.005 eV
λ =0.25 ± 0.03
Γ (eV)

Γ0 = 0.080 ± 0.002 eV
0.14

0.12

0.10
λ=0.2 ± 0.02
Γ0 = 0.065 ± 0.003 eV

0.08

0 100 200 300 400


Temperature (K)

Fig. 27. Temperature-dependent linewidth of the surface state on Bi(1 0 0) for four different binding energies and
fit to the data within the Debye model. From Ref. [123].

can be understood from simple phase space arguments; at high binding energies there are
many bulk states available in which a hole can scatter with the help of a phonon. Close
to the Fermi energy the density of bulk states is very small. In order to illustrate
the argument, Fig. 28 also shows a plot of the calculated bulk density of states
(DOS), scaled in an arbitrary way. For high binding energies the change in the bulk
DOS mimics the change in k, as expected from the simple argument above. For small
binding energies the bulk DOS essentially vanishes but k stays finite. The scattering pro-
cesses leading to epc coupling in this energy range are therefore most likely to involve
other surface states.
Apart from the technical point that there can be strong energy dependent changes in k,
these results support the view of the Bi surfaces as nearly two-dimensional metals. An
analysis based on EDCs, however, cannot give any information about the coupling
strength at EF, which would be desirable in order to learn more about the properties of
these two-dimensional metals. In the next two sections, an analysis of k at EF is described
for Bi(1 1 1) and Bi(1 1 0).
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 233

0.8

0.6

Bulk density of states (arb. units)


λ

0.4

0.2

0.0
0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00
Binding Energy (eV)

Fig. 28. Electron–phonon coupling parameter k as a function of the binding energy. The dashed line between the
data points connects the points as they lie on the C–K 2 line (see Fig. 15). Also shown is the bulk density of states
of Bi as a function of binding energy as calculated using the tight-binding parameters from Liu and Allen [15].
After Ref. [123].

6.3. Bi(1 1 1)

The epc close to EF for the hole pocket on Bi(1 1 1) has been studied by two groups. Ast
and Höchst have analysed the MDC linewidth as a function of binding energy near EF [56]
and Gayone et al. have analysed the temperature dependence of the MDC linewidth of the
same state, also near the Fermi level crossing [108].
Fig. 29 shows the result from the work of Ast and Höchst [56]. R00 was obtained from a
fit to MDCs taken near EF and R 0 from a Kramers–Kronig transformation of R00 . How-
ever, a technical mistake was made in the MDC analysis by using an incorrect expression
for the bare dispersion instead of (11). It is not clear how much this error affects the
results, but due to the small R 0 , it is probably not too important.
The imaginary part of the self-energy has been fitted to (14) plus a constant offset using
the Debye model (17) to describe the Eliashberg function. The dashed and solid lines in
Fig. 29 are found for very different parameters: Using the bulk Debye energy of
⁄xD = 10 meV, k is found to be 0.6 ± 0.05 (solid line) but when a lower Debye energy
of 5 meV is assumed, as appropriate for the surface [49,46], k is found to be 2.3 ± 0.2.
The constant off-set needed to fit the actual data, is also very different. Assuming that only
electron-defect scattering is important, almost the entire linewidth at EF (22 meV) stems
from defect scattering for the solid line, whereas the defect scattering is negligible for
the dashed line. Note that a strong correlation between the Debye energy and k in the
fit is not unexpected. Already from an inspection of (14) and (17) it is evident that the
234 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 29. Electron–phonon coupling for the hole state on Bi(1 1 1): (a) Imaginary part of the self-energy obtained
from fits to MDCs near EF. (b) Real part of the self-energy calculated from the imaginary part. The solid and
dashed lines are fits to the Debye model of the epc at T = 50 K. See text for details. From Ref. [56].

latter two parameters are highly correlated in the fit [137,138]. To make matters worse, the
experimental sample temperature also plays an unusually important role for the fit because
the experiment has been carried out at a temperature similar to the surface Debye
temperature.
The results of the work from Gayone et al. are shown in Fig. 30. The temperature
dependence of MDCs at a binding energy of 25 meV is shown in Fig. 30(a) together with
a fit using four Lorentzian lines (only shown for the lowest temperature). From the tem-
perature-dependent linewidth and the local group velocity of the state one can determine
R00 which is shown in Fig. 30(b). The group velocity at a binding energy of 25 meV can be
obtained from a fit to the dispersion of the state in a dataset such as that shown in
Fig. 20(a). Note that this binding energy is sufficiently large such that R 0 and thereby
the re-normalization of the bands will be very small (below 2 meV). Fig. 30(b) also shows
a fit using (18). This results in a value of k = 0.40(5). A clear advantage in this experiment
is the use of temperature dependent data at higher temperatures. Because of this the data
can be analysed using (18), the high-temperature limit of (14), and no assumptions about
the surface Debye temperature have to be made.
The values of k obtained by Ast and Höchst (0.6 or 2.3) and by Gayone et al. (0.4) are
quite different, especially if one takes the larger value of 2.3 in the first study for compar-
ison. Indeed, there is a good case for comparing to this value because it has been obtained
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 235

T=298 K

Photoemission intensity (arb. units)


8
70
T=248 K
65
6 λ = 0.4
T=199 K
60

Σ'' (meV)
T=148 K
4 55

T=98 K
50
2
e e h h
T=48 K 45

0 40
0.0 0.1 0.2 0.3 0.4 50 100 150 200 250 300
(a) Γ k|| (Å-1) (b) Temperature (K)

Fig. 30. Electron–phonon coupling for the hole state on Bi(1 1 1): (a) Temperature-dependent MDCs at a binding
energy of 25 meV in the C–M direction. On the bottom part of the figure the fit with four Lorentzian lines to the
spectrum taken at the lowest temperature is shown. The solid Lorentzian line represents the branch of the hole
pocket for which k is determined. The branches belonging to the electron and hole pockets are labelled by e’s and
h’s, respectively. (b) Imaginary part of the self-energy extracted from the temperature-dependent MDCs with a
linear fit according to (18). From Ref. [108].

for a Debye temperature which is more realistic for the surface layers, according to the
LEED results [49,46]. This apparent contradiction was eventually solved by Kirkegaard
et al. [133]. These authors have shown that the finite spectrometer energy resolution needs
to be taken into account in the type of analysis performed by Ast and Höchst. A simple
estimate shows that if this is done, the final value of k is quite close to that obtained by
Gayone et al. [133].

6.4. Bi(1 1 0)

Kirkegaard et al. have studied the epc for two different surface states of Bi(1 1 0), the hole
pockets near C and M (see Fig. 13) [133]. R00 was extracted from MDCs, both as a function
of temperature and energy. These large two-dimensional datasets have been fitted to (14)
using both the Debye and the Einstein model for the Eliashberg function. An example of
the data and the fit is shown in Fig. 31. Note that R00 shows the characteristic and expected
increase with energy and temperature. Because of the large dataset, the problem of deter-
mining both k and the Debye (or Einstein) energy can be solved in a satisfying way. The
resulting values of k actually turn out to be the same in the Debye and Einstein models,
a fact which is not surprising because much of the data was taken at elevated temperatures
where the precise nature of the phonon spectrum becomes unimportant. The k values
obtained from this analysis are 0.19(3) and 0.27(2), near the C and M points, respectively.
It is interesting to remark that Kirkegaard et al. have also studied the effect of a finite
spectrometer resolution on the results for this system. This is certainly an important issue,
in particular the energy resolution which has been lower than the phonon bandwidth of Bi
in all the experiments described above. It turns out that the type of analysis used for
Bi(1 1 0) gives reliable values of k even if the resolution broadening is ignored. The resulting
Debye or Einstein frequencies (xD and xE) and the constant defect scattering contribution
to R00 , however, can be seriously overestimated. Indeed, the xD and xE obtained from
236 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

Fig. 31. R00 for the hole pocket near M on Bi(1 1 0) in the M–X 2 direction, fitted to the Debye model. (a)
Experimental data, (b) fit to data, (c) deviation between the data and the model and (d) plot of the reduced v2 and
k.

fitting data for Bi(1 1 0) have been unrealistically high, far above the highest phonon fre-
quencies in Bi or the values determined using LEED [53]. Even after correcting for the
energy resolution induced increase, the values remain high; an issue which is not yet solved.

6.5. General trends

Summarizing the results from the different surfaces, a consistent picture of the epc on
the surfaces of Bi emerges. Bi(1 0 0) shows strong energy dependence of k in the binding
energy region where the bulk DOS changes rapidly. This behaviour can also be expected
for the other surfaces of Bi and it fits well into the picture of the Bi surfaces as quasi-two-
dimensional metals on a semimetallic bulk.
The argument that the epc strength is dominated by the available phase space for the
scattering also agrees with the general picture of epc for Bi. The epc for bulk rhombohe-
dral Bi at the Fermi level should be very weak due to the semimetallic nature. To the best
of our knowledge, there is no experimental value for k in bulk Bi. Gayone et al. have esti-
mated k by calculating the density of states and thereby the electronic heat capacity from
the tight-binding parameters given by Liu and Allen [15] and comparing this to the mea-
sured heat capacity [122]. The result is k = 0.13(0.13 + 0.2). Even though the uncertainty
is big, the value is small. However, the semimetallic character is specific to the rhombohe-
dral structure, i.e. the DOS at EF is much higher for other Bi phases. Indeed, in the liquid
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 237

the pseudogap at the Fermi energy is less pronounced [139] and the same probably also
holds for the amorphous state. The epc is very strong for amorphous Bi (k = 2.46) [140]
but there can be other factors contributing to this change besides the change in the density
of states.
The coupling strength at or near the Fermi energy is found to be of the same order of
magnitude for all surfaces investigated here. For Bi(1 0 0) the smallest binding energy is
70 meV, not very close to EF but k appears to level off at a value around 0.2 (see
Fig. 28). For Bi(1 1 1) and for Bi(1 1 0) the values of k near EF are also in the range between
0.4 and 0.2. There is a strong indication of a certain ~ k-dependence in the coupling strength.
Fig. 28 shows that different values of k are obtained for states with the same binding ener-
gies but different ~ ks and also for Bi(1 1 0) the two states investigated show different k values
at the Fermi energy. A certain ~ k-dependence is not surprising given the complex physics
involved. On the other hand, it seems that the assumption of a locally ~ k-independent R
is justified, because the fitting of MDCs with simple Lorentzian lines generally works very
well.
Finally, it is interesting to turn back to the possible influence of the Bi surface states on
the superconductivity of the granular cluster films observed by Weitzel and Micklitz
[44,45] and to the intriguing but unanswered question if the surfaces of Bi as such could
be superconducting. If we compare the measured k values at EF to those of three-
dimensional materials, they are clearly smaller than those for the strong coupling super-
conductors such as Pb or a-Hg which have k values around 1. On the other hand, they
are similar to the values found for other superconducting metals such as Zn or Cd
(k  0.4) and clearly higher than for the non-superconducting noble metals which have
k  0.15 [107].
Genuine surface-state mediated superconductivity would be a fascinating new finding.
Superconductivity involving bulk-like states but localized at surfaces or in thin films is a
well-known phenomenon [141] and it is two-dimensional in the sense that its extension
in the third dimension is much smaller than the coherence length of the Cooper pairs.
It has even been shown that the transition temperature of thin films can be changed via
quantum size effects which affect the density of states at the Fermi level [142]. The idea
of superconductivity through the coupling between electronic surface states has been dis-
cussed for at least two promising cases: Be(0 0 0 1) [114] and hydrogen on W(1 1 0) [118] but
it has not been confirmed so far. An advantage of Bi surfaces is that the electrons near the
Fermi level are extremely localized at the surface. On the other hand, however, the phonon
energies are low, much lower than in the two cases mentioned above, and if surface-state
superconductivity exists, one would expect to find it at quite low temperatures only. An
experimental test is therefore challenging. There are several possibilities. One could try
to observe the opening of a gap using ARPES or scanning tunnelling spectroscopy. This
puts very high requirements on sample cooling and resolution [143]. Alternatively, one
could try surface-sensitive, temperature dependent resistivity measurements using an
ultra-high vacuum version of a micro-four-point probe [144,145].
It is interesting to note that there is a link between the epc discussed here and the spin–
orbit splitting of the surfaces states. The spin splitting is very likely to reduce the available
phase space for epc in a similar way as it reduces the possibilities for quasi-particle inter-
ference such that only spin-conserving scattering events are possible. This means that the
actual epc strength on the surfaces of Bi could be higher than the measured k values
suggest.
238 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

7. Conclusion and outlook

The most important result of the work reviewed here is that the surfaces of Bi are rather
different from the semimetallic bulk: The (1 1 1), (1 0 0) and (1 1 0) surfaces support metallic
surface states which turn them into a considerably better metal than the bulk. However,
these metals have some quite unusual properties because of the strong spin–orbit interac-
tion in Bi. In bulk Bi the spin–orbit interaction is important for the (mainly) 6p-derived
valence band, but because of inversion symmetry, the spin-degeneracy in the band struc-
ture is not lifted. On the surface, the inversion symmetry is broken and the surface band
structure and Fermi surface are strongly influenced by the spin–orbit splitting. Finally, the
electron–phonon coupling for the surface states near the Fermi level was found to be of
intermediate strength, markedly different from the very weak coupling expected for the
bulk.
For both the geometric structure and the electronic structure, a good agreement
between the experimental data and first-principles calculations was found. Indeed, the ini-
tial interpretation of the electronic surface band structure in terms of spin–orbit splitting
had been entirely based on the comparison between experiment and calculations. Later,
additional direct evidence was found in the characteristic quasi-particle interference on
Bi surfaces. The only case for which the agreement between the calculated and measured
surface electronic structure is less favourable is Bi(1 0 0). Here the calculated electronic
structure is less reliable because of the deep penetration of the surface states.
These findings have a number of interesting implications for Bi nano-structures which
have recently attracted much interest [38]. It is well known that the properties of bulk Bi
are strongly changed by quantum confinement effects as the dimensions of a structure
become smaller than a few hundred nanometres [1]. However, the results summarized in
this review show that it is not sufficient to consider the changed bulk properties when dis-
cussing nano-scale objects of Bi. For instance, a central difficulty for unambiguously iden-
tifying the predicted semimetal-to-semiconductor transition for Bi films thinner than
30 nm or so [29–31,34] was long thought to be linked to the existence of surface states,
a fact confirmed by the results outlined here. Indeed, a simple estimate of the surface car-
rier densities for the electron and hole pockets of Bi(1 1 1) lead to values in the order of
1 · 1013 cm2, similar to that assumed in thin film experiments [33,31,55]. On the other
hand, it can be misleading to view the Bi nano-structures as being composed from a bulk
which is modified by quantum size effects and a metallic surface because the surface states
can actually penetrate rather deeply, as on Bi(1 0 0). This could be a source for the diffi-
culty in establishing whether electronic surface states are responsible for the superconduc-
tivity found in systems of Bi nano-clusters [44,45].
The results on the electron–phonon coupling in Bi fit well into the picture of a metallic
surface on a semimetallic bulk. For higher binding energies, were phonon-mediated scat-
tering between a surface state and bulk states is possible, the coupling is rather strong. On
the other hand, only an intermediate coupling strength has been found close to the Fermi
level where the density of bulk states is vanishingly small. Still, the coupling for the surface
states at the Fermi level is much stronger than that expected for a bulk semimetal, suggest-
ing that the most of the scattering is happening between surface state bands.
At the same time, our understanding of the electron–phonon coupling on the Bi
surfaces remains somewhat rudimentary and the interpretation of the data is confined
to simple phase space arguments. On the experimental side a detailed determination of
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 239

the Eliashberg function as reported for Be by Shi et al. [132] would require an energy res-
olution which is much better than the typical phonon energies in Bi, i.e. a resolution in the
meV or leV range. This is at the limit of what is possible now. In addition, virtually noth-
ing is known about the surface phonons on Bi, as opposed to the electronic structure.
Indeed, the discussion here has been confined to surface Debye temperatures determined
by LEED. While this gives a rough idea of the vibrational properties, it is insufficient for
any detailed analysis. On the theory side, the situation is equally difficult, mainly because
Bi is a heavy atom, having many electrons, strong spin–orbit interaction and a compli-
cated bulk electronic structure. Unlike the case of light metals such as Be, where very
detailed calculations of the many-body interactions are possible [146], this is not feasible
at present for Bi.
Apart from the metallic surface states, an important property of the Bi surfaces is the
strong spin–orbit splitting. There are many unexplored phenomena related to this, such as
the electronic response, quasi-particle interference, electron–phonon and electron–electron
coupling and the possibility of two-dimensional superconductivity [147]. So far, most of
the dynamical aspects of strong spin–orbit coupling remain unexplored territory.
The spin–orbit splitting is probably also the most important property of Bi surfaces and
interfaces with respect to applications in spintronics. Because of the considerable size of
the splitting and the high density of states, Bi interfaces could be promising candidates
for the construction of spin sources or filters instead of, or more probably in addition
to, semiconductor heterostructures [79,80].
Finally, the strong spin–orbit splitting may be linked to the remarkable fact that neither
Bi(1 1 0) nor Bi(1 0 0) reconstruct, despite the dangling bonds on these surfaces. Many
reconstructions of semiconductor surfaces can be viewed as charge density waves [96],
or at least as a mechanism of energy-lowering by removing the states at the Fermi level.
The discussion of a possible charge density wave on Bi(1 1 1) has shown that the standard
reason for a charge density wave formation, the 2~ k F singularity in the Lindhard suscepti-
bility, is absent for Bi surfaces due to the role of the spin. This may contribute to the fact
that Bi surfaces have a weaker tendency for reconstruction than most semiconductor
surfaces.

Acknowledgements

I would like to thank all my many co-workers in this project. In particular, I would like
to mention the students and postdocs who have been working on Bi surfaces in my group,
as well as my collaborators, who did first-principles calculations, LEED data analysis,
STM, photoelectron diffraction and transmission electron microscopy.
I thank Christian Søndergaard, Casper Kirkegaard and Justin Wells for a critical read-
ing of the entire paper. Finally, I thank the Danish Research Council, The Danish Foreign
Ministry, the Carlsberg Foundation and the European Union for financial support.

References

[1] Y.F. Ogrin, V.N. Lutskii, M.I. Elinson, Observation of quantum size effects in thin bismuth films, JETP
Letters 3 (1966) 71–73.
[2] A.B. Shick, J.B. Ketterson, D.L. Novikov, A.J. Freeman, Electronic structure, phase stability, and
semimetal-semiconductor transtitions in Bi, Physical Review B 60 (1999) 15484–15487.
240 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

[3] A.A. Giardini, G.A. Samara, The compressibility of bismuth and its upper transition pressure, Journal of
Physics and Chemistry of Solids 26 (1965) 1523, and references therein.
[4] R.A. Bartynski, E. Jensen, T. Gustafsson, E.W. Plummer, Angle-resolved photoemission investigation of
the electronic structure of Be: surface states, Physical Review B 32 (1985) 1921–1926.
[5] E.V. Chulkov, V.M. Silkin, E.N. Shirykalov, Surface electronic structure of Be(0 0 0 1) and Mg(0 0 0 1),
Surface Science 188 (1987) 287.
[6] Ph. Hofmann, R. Stumpf, V.M. Silkin, E.V. Chulkov, E.W. Plummer, The electronic structure of
Beð1 0 
1 0Þ, Surface Science 355 (1996) L278.
[7] M. Bernasconi, G.L. Chiarotti, E. Tosatti, The (0 0 1) surface of a-Ga(0 0 1) is covered with Ga III, Physical
Review Letters 70 (1993) 3295–3298.
[8] M. Bernasconi, G.L. Chiarotti, E. Tosatti, Theory of the structural and electronic properties of a-Ga(0 0 1)
and (0 1 0) surfaces, Physical Review B 52 (1995) 9999–10013.
[9] Ph. Hofmann, Y.Q. Cai, C. Grütter, J.H. Bilgram, Electron–lattice interactions on a-Ga(0 1 0), Physical
Review Letters 81 (1998) 1670–1673.
[10] F. Jona, Low-energy electron diffraction study of surfaces of antimony and bismuth, Surface Science 8
(1967) 57–76.
[11] M.H. Cohen, Energy bands in the bismuth structure. I. A nonellipsodial model for electrons in Bi, Physical
Review 121 (1961) 387–395.
[12] C. Søndergaard, An investigation of surface electronic structure and electron–phonon interaction by angle-
resolved photoemission spectroscopy, Ph.D. thesis, University of Aarhus, 2001.
[13] N.E. Ashcroft, N.D. Mermin, Solid State Physics, international edition., Saunders College, Philadelphia,
1976.
[14] P. Cucka, C.S. Barrett, The crystal structure of Bi and of solid solutions of Pb, Sn, Sb and Te in Bi, Acta
Crystallographica 15 (1962) 865–872.
[15] Y. Liu, R.E. Allen, Electronic structure of the semimetals Bi and Sb, Physical Review B 52 (1995) 1566.
[16] V.S. Édel’man, Electrons in bismuth, Advances in Physics 25 (1976) 555–613.
[17] S. Golin, Band structure of bismuth: pseudopotential approach, Physical Review 166 (1968) 643–651.
[18] X. Gonze, J.-P. Michenaud, J.-P. Vigneron, First-principles study of As, Sb, and Bi electronic properties,
Physical Review B 41 (1990) 11827–11836.
[19] J.H. Xu, E.G. Wang, C.S. Ting, W.P. Su, Tight-binding theory of the electronic structures for
rhombohedral semimetals, Physical Review B 48 (1993) 17271–17279.
[20] G. Jezequel, Y. Petroff, R. Pinchaux, F. Yndurain, Electronic structure of the Bi(1 1 1) surface, Physical
Review B 33 (1986) 4352–4355.
[21] G. Jezequel, J. Thomas, I. Pollini, Experimental band structure of semimetal bismuth, Physical Review B 56
(1997) 6620–6626.
[22] J. Thomas, G. Jezequel, I. Pollini, Photoemission study of the bulk and surface electronic structure of
Bi(1 1 1), Journal of Physics: Condensed Matter 11 (1999) 9571–9580.
[23] M. Hengsberger, P. Segovia, M. Garnier, D. Purdie, Y. Baer, Photoemission study of the carrier bands in
Bi(1 1 1), The European Physical Journal B 17 (2000) 603–608.
[24] A. Tanaka, M. Hatano, K. Takahashi, H. Sasaki, S. Suzuki, S. Sato, Growth and angle-resolved
photoemission studies of bismuth epitaxial films, Surface Science 433–435 (1999) 647–651.
[25] A. Tanaka, M. Hatano, K. Takahashi, H. Sasaki, S. Suzuki, S. Sato, Bulk and surface electronic structures
of the semimetal Bi studied by angle-resolved photoemission spectroscopy, Physical Review B 59 (1999)
1786–1791.
[26] C.R. Ast, H. Höchst, High-resolution photoemission mapping of the three-dimensional band structure of
Bi(1 1 1), Physical Review B 70 (2004) 245122.
[27] C.E. Moore, Atomic Energy Levels, National Bureau of Standards, Washington, DC, 1949.
[28] J.M. Ziman, Principles of the Theory of Solids, second ed., Cambridge University Press, 1972.
[29] V.N. Lutskii, Features of optical absorption of metallic films in region where metal turns into a dielectric,
Soviet Physics JETP Letters 2 (1965) 245.
[30] V.B. Sandomirskii, Quantum size effect in a semimetal film, Soviet Physics JETP 25 (1967) 101.
[31] C.A. Hoffman, J.R. Meyer, F.J. Bartoli, A.D. Venere, X.J. Xi, C.L. Hou, H.C. Wang, Semimetal-to-
semiconductor transition in bismuth thin films, Physical Review B 48 (1993) 11431–11434.
[32] Y.F. Komnik, E.N. Bukhshtab, Y.V. Nikitin, V.V. Andrievskii, Some features of the temperature
dependence of the resistance of thin bismuth films, Zhurnal Eksperimentalnoi I Teoreticheskoi Fiziki USSR
(2) (1971) 669–687.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 241

[33] Y.F. Komnik, V.V. Andrievskii, Kinetic properties of thin bismuth films, Fizika Nizkikh Temperatur
(Ukrainian SSR) 1 (1) (1975) 104–119.
[34] H.T. Chu, Comment on ‘‘Semimetal-to-semiconductor transition in bismuth thin films’’, Physical Review B
51 (1995) 5532–5534.
[35] C. Koitzsch, M. Bovet, F. Clerc, D. Naumovic, L. Schlapbach, P. Aebi, Growth of thin Bi films on W(1 1 0),
Surface Science 527 (1–3) (2003) 51–56.
[36] T. Nagao, J.T. Sadowski, M. Saito, S. Yaginuma, Y. Fujikawa, T. Kogure, T. Ohno, Y. Hasegawa, S.
Hasegawa, T. Sakurai, Nanofilm allotrope and phase transformation of ultrathin Bi film on Si(1 1 1)-7 · 7,
Physical Review Letters 93 (10) (2004) 105501.
[37] S.A. Scott, M.V. Kral, S.A. Brown, A crystallographic orientation transition and early stage growth
characteristics of thin Bi films on hopg, Surface Science 587 (3) (2005) 175–184.
[38] M.S. Dresselhaus, Y.M. Lin, O. Rabin, A. Jorio, A.G. Souza Filho, M.A. Pimenta, R. Saito, G.
Samsonidze, G. Dresselhaus, Nanowires and nanotubes, Materials Science and Engineering: C 23 (1–2)
(2003) 129–140.
[39] L.D. Hicks, M.S. Dresselhaus, Thermoelectric figure of merit of a one-dimensional conductor, Physical
Review B 47 (1993) 16631.
[40] Y.-M. Lin, X. Sun, M.S. Dresselhaus, Theoretical investigation of thermoelectric transport properties in
cylindrical Bi nanowires, Physical Review B 62 (2000) 4610.
[41] J.P. Heremans, C.M. Thrush, D.T. Morelli, W. Ming-Cheng, Thermoelectric power of bismuth
nanocomposites, Physical Review Letters 88 (2002) 216801.
[42] T.E. Huber, A. Nikolaeva, D. Gitsu, L. Konopko, C.A. Foss Jr., M.J. Graf, Confinement effects and the
surface-induced charge carriers in Bi nanowires, Applied Physics Letters 84 (2004) 1326.
[43] J.-P. Issi, Low temperature transport properties of the group V semimetals, Australian Journal of Physics
32 (1979) 585–628.
[44] B. Weitzel, H. Micklitz, Superconductivity in granular systems built from well-defined rhombohedral Bi
clusters: evidence for Bi-surface superconductivity, Physical Review Letters 66 (1991) 385–388.
[45] C. Vossloh, M. Holdenried, H. Micklitz, Influence of cluster size on the normal- and superconducting-state
properties of granular Bi films, Physical Review B 58 (1998) 12422–12426.
[46] H. Mönig, J. Sun, M. Koroteev, G. Bihlmayer, J. Wells, E.V. Chulkov, K. Pohl, Ph. Hofmann, The
structure of the (1 1 1) surface of bismuth: LEED analysis and first principles calculations, Physical Review
B 72 (2005) 085410.
[47] K.G. Ramanathan, T.M. Srinivasan, Specific heat of bismuth at liquid helium temperatures, Physical
Review 99 (1955) 442.
[48] C. Walfried, D.N. McIlroy, J. Zhang, P.A. Dowben, G.A. Katrich, E.W. Plummer, Determination of the
surface Debye temperature of Mo(1 1 2) using valence band photoemission, Surface Science 363 (1996) 296–
302.
[49] R.M. Goodman, G.A. Somorjai, Low-energy electron diffraction studies of surface melting and freezing of
lead, bismuth and tin single-crystal surfaces, Journal of Chemical Physics 52 (1970) 6325.
[50] C.R. Ast, H. Höchst, Electronic structure of a bismuth bilayer, Physical Review B 67 (2003) 113102.
[51] Ph. Hofmann, J.E. Gayone, G. Bihlmayer, Y.M. Koroteev, E.V. Chulkov, The electronic structure and
Fermi surface of Bi(1 0 0), Physical Review B 71 (2005) 195413.
[52] Y.M. Koroteev, G. Bihlmayer, E.V. Chulkov, private communication.
[53] J. Sun, A. Mikkelsen, Y.M. Koroteev, G. Bihlmayer, E.V. Chulkov, M.F. Jensen, D.L. Adams, K. Pohl,
Ph. Hofmann, Physical Review B, submitted for publication.
[54] F. Patthey, W.-D. Schneider, H. Micklitz, Photoemission study of the Bi(1 1 1) surface, Physical Review B
49 (1994) 11293–11296.
[55] C.R. Ast, H. Höchst, Fermi surface of Bi(1 1 1) measured by photoemission spectroscopy, Physical Review
Letters 87 (2001) 177602.
[56] C.R. Ast, H. Höchst, Two-dimensional band structure and self-energy of Bi(1 1 1) near the C point, Physical
Review B 66 (2002) 125103.
[57] A. Santoni, L.J. Terminello, F.J. Himpsel, T. Takahashi, Mapping of the Fermi surface of graphite with a
display-type photoelectron spectrometer, Applied Physics A 52 (1991) 299–301.
[58] P. Aebi, J. Osterwalder, R. Fasel, D. Naumović, L. Schlapbach, Fermi surface mapping with
photoelectrons at UV energies, Surface Science 307–309 (1994) 917–921.
242 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

[59] P. Aebi, J. Osterwalder, P. Schwaller, L. Schlapbach, M. Shimoda, T. Mochiku, K. Kadowaki, Complete


Fermi surface mapping of Bi2Sr2CaCu2O8+d(0 0 1): coexistence of short range antiferromagnetic correla-
tions and metallicity in the same phase, Physical Review Letters 72 (1994) 2757–2760.
[60] J. Avila, C. Casado, M.C. Asensio, J.L. Perez, M.C. Munõz, F. Soria, Bulk Fermi surface determination by
tuning the photoelectron kinetic energy, Journal of Vacuum Science and Technology A 13 (1995) 1501–
1505.
[61] T. Straub, R. Claessen, P. Steiner, S. Hüfner, V. Eyert, K. Friemelt, E. Bucher, Many-body definition of a
Fermi surface: application to angle-resolved photoemission, Physical Review B 55 (1997) 13473–13478.
[62] L. Kipp, K. Roßnagel, C. Solterbeck, T. Strasser, W. Schattke, M. Skibowski, How to determine Fermi
vectors by angle-resolved photoemission, Physical Review Letters 83 (1999) 5551–5554.
[63] S.V. Borisenko, A.A. Kordyuk, S. Legner, C. Dürr, M. Knupfer, M.S. Golden, J. Fink, K. Nenkov, D.
Eckert, G. Yang, S. Abell, H. Berger, L. Forro, B. Liang, A. Maljuk, C.T. Lin, B. Keimer, Estimation of
matrix-element effects and determination of the Fermi surface in Bi2Sr2CaCu2O8+d systems using angle-
scanned photoemission spectroscopy, Physical Review B 64 (2001) 094513.
[64] C.R. Ast, H. Höchst, Indication of charge-density wave formation on Bi(1 1 1), Physical Review Letters 90
(2003) 016403.
[65] Y.M. Koroteev, G. Bihlmayer, J.E. Gayone, E.V. Chulkov, S. Blügel, P.M. Echenique, Ph. Hofmann,
Strong spin–orbit splitting on Bi surfaces, Physical Review Letters 93 (2004) 046403.
[66] K. Sugawara, T. Sato, S. Souma, T. Takahashi, M. Arai, T. Sasaki, Fermi surface and anisotropic spin–
orbit coupling of sb(1 1 1) studied by angle-resolved photoemission spectroscopy, Physical Review Letters
96 (4) (2006) 046411-4.
[67] S. Agergaard, C. Søndergaard, H. Li, M.B. Nielsen, S.V. Hoffmann, Z. Li, Ph. Hofmann, The effect of
reduced dimensionality on a semimetal: the electronic structure of the Bi(1 1 0) surface, New Journal of
Physics 3 (2001) 15.1–15.10.
[68] J.I. Pascual, G. Bihlmayer, Y.M. Koroteev, H.-P. Rust, G. Ceballos, M. Hansmann, K. Horn, E.V.
Chulkov, S. Blügel, P.M. Echenique, Ph. Hofmann, Role of the spin in quasiparticle interference, Physical
Review Letters 93 (2004) 196802.
[69] Y.M. Koroteev, G. Bihlmayer, E.V. Chulkov, to be published.
[70] L. Petersen, P. Hedegård, A simple tight-binding model of spin–orbit splitting of sp-derived surface states,
Surface Science 459 (2000) 49–56.
[71] E.I. Rashba, Properties of semiconductors with an extremum loop. I. Cyclotron and combinational
resonance in a magnetic field perpendicular to the plane of the loop, Soviet Physics – Solid State 2 (1960)
1109.
[72] S. LaShell, B.A. McDougall, E. Jensen, Spin splitting of an Au(1 1 1) surface state band observed with angle
resolved photoelectron spectroscopy, Physical Review Letters 77 (1996) 3419–3422.
[73] G. Nicolay, F. Reinert, S. Hüfner, P. Blaha, Spin–orbit splitting of the l-gap surface state on Au(1 1 1) and
Ag(1 1 1), Physical Review B 65 (2001) 033407.
[74] A. Mugarza, A. Mascaraque, V. Repain, S. Rousset, K.N. Altmann, F.J. Himpsel, Y.M. Koroteev,
E.V. Chulkov, F.J.G. de Abajo, J.E. Ortega, Lateral quantum wells at vicinal au(1 1 1) studied with angle-
resolved photoemission, Physical Review B 66 (24) (2002) 245419.
[75] J. Henk, A. Ernst, P. Bruno, Spin polarization of the l-gap surface states on Au(1 1 1), Physical Review B 68
(2003) 165416.
[76] F. Reinert, Spin–orbit interaction in the photoemission spectra of noble metal surfaces, Journal of Physics:
Condensed Matter 15 (2003) S693.
[77] M. Hoesch, M. Muntwiler, V.N. Petrov, M. Hengsberger, L. Patthey, M. Shi, M. Falub, T. Greber, J.
Osterwalder, Spin structure of the Shockley surface state on Au(1 1 1), Physical Review B 69 (2004) 241401-
1.
[78] M. Hochstrasser, J.G. Tobin, E. Rotenberg, S.D. Kevan, Spin-resolved photoemission of surface states of
W(1 1 0)-(1 · 1)H, Physical Review Letters 89 (2002) 216802.
[79] S. Datta, B. Das, Electronic analog of the electro-optic modulator, Applies Physics Letters 56 (1990) 665.
[80] T. Koga, J. Nitta, H. Takayanagi, S. Datta, Spin-filter device based on the Rashba effect using a
nonmagnetic resonant tunneling diode, Physical Review Letters 88 (2002) 126601.
[81] J. Friedel, Metallic alloys, Nuovo Cimento 7 (1958) 287.
[82] J. Lindhard, On the properties of a gas of charged particles, Det Kongelige Danske Videnskabernes Selskab
Matematisk-fysiske Meddelelser 28 (8) (1954) 1.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 243

[83] M.F. Crommie, C.P. Lutz, M. Eigler, Imaging standing waves in a two-dimensional electron gas, Nature
363 (1993) 524.
[84] P.T. Sprunger, L. Petersen, E.W. Plummer, E. Lægsgaard, F. Besenbacher, Giant Friedel oscillations on the
beryllium(0 0 0 1) surface, Science 275 (1997) 1764–1767.
[85] Ph. Hofmann, B.G. Briner, M. Doering, H.-P. Rust, E.W. Plummer, A.M. Bradshaw, Anisotropic two-
dimensional Friedel oscillations, Physical Review Letters 79 (1997) 265–268.
[86] L. Petersen, P.T. Sprunger, Ph. Hofmann, E. Lægsgaard, B.G. Briner, M. Doering, H.-P. Rust, A.M.
Bradshaw, F. Besenbacher, E.W. Plummer, Direct imaging of the two-dimensional Fermi contour: Fourier-
transform STM, Physical Review B 57 (1998) R6858–R6861.
[87] K. McElroy, R.W. Simmonds, J.E. Hoffman, D.-H. Lee, J. Orenstein, H. Eisaki, J.C. Davis, Relating
atomic-scale electronic phenomena to wave-like quasiparticle states in superconducting Bi2Sr2CaCu2O8+d,
Nature 422 (2003) 592.
[88] L. Diekhöner, M.A. Schneider, A.N. Baranov, V.S. Stepanyuk, P. Bruno, K. Kern, Surface states on cobald
nanoislands on Cu(1 1 1), Physical Review Letters 90 (2003) 236801.
[89] F. Moresco, L. Gross, M. Alemani, K.-H. Rieder, H. Tang, A. Gourdon, C. Joachim, Probing the different
stages in contacting a single molecular wire, Physical Review Letters 91 (2003) 036601.
[90] J.E. Hoffman, K. McElroy, D.-H. Lee, K.M. Lang, H. Eisaki, S. Uchida, J.C. Davis, Imaging quasiparticle
interferences in Bi2Sr2CaCu2O8+d, Science 297 (2002) 1148.
[91] M. Vershinin, S. Misra, S. Ono, Y. Abe, Y. Ando, A. Yazdani, Local ordering in the pseudogap state of the
high-tC superconductor Bi2Sr2CaCu2O8+d, Science 303 (2004) 1995.
[92] L. Petersen, L. Bürgi, H. Brune, F. Besenbacher, K. Kern, Comment on ‘Observation of two-dimensional
Fermi contour of a reconstructed Au(1 1 1) surface using Fourier transform scanning tunneling microscopy’
D. Fujita, K. Amemiya, T. Yakabe, H. Nejoh, T. Sato, M. Iwatsuki [Surf. Sci. 423 (1999) 160], Surface
Science 443 (1999) 154.
[93] G. Grüner, Density Waves in SolidsFrontiers in Physics, vol. 89, Perseus Publishing, Cambridge,
Massachusetts, 1994.
[94] E. Tosatti, Surface states, surface metal–insulator transition, and surface insulator–metal transitions, in: E.
Bertel, M. Donath (Eds.), Electronic Surface and Interface States on Metallic Systems, World Scientific,
Singapore, 1995.
[95] T. Aruga, Charge-density waves on metal surfaces, Journal of Physics: Condensed Matter 14 (2002) 8393.
[96] E. Tosatti, Electronic superstructures of semiconductor surfaces and of layered transition metal
compounds, Festkörperprobleme XV (1975) 113–147.
[97] T.K. Kim, J. Wells, C. Kirkegaard, Z. Li, S.V. Hoffmann, J.E. Gayone, I. Fernandez-Torrente, P. Haberle,
J.I. Pascual, K.T. Moore, A.J. Schwartz, H. He, J.C.H. Spence, K.H. Downing, S. Lazar, F.D. Tichelaar,
S.V. Borisenko, M. Knupfer, Ph. Hofmann, Evidence against a charge density wave on Bi(1 1 1), Physical
Review B 72 (8) (2005) 085440.
[98] R.V. Coleman, B. Drake, P.K. Hansma, G. Slough, Charge-density waves observed with a tunneling
microscope, Physical Review Letters 55 (1985) 394.
[99] W. Han, R.A. Pappas, E.R. Hunt, R.F. Frindt, Atomically resolved charge-density waves in 1T-TaS2,
Physical Review B 48 (1993) 8466.
[100] J.M. Carpinelli, H.H. Weitering, E.W. Plummer, R. Stumpf, Direct observation of a surface charge density
wave, Nature 381 (1996) 398–400.
[101] I. Ekvall, J.-J. Kim, H. Olin, Atomic and electronic structures of the two different layers in 4HB-TaS2 at
4.2 k, Physical Review B 55 (1997) 6758.
[102] V.S. Édel’man, STM observation of twin microlayers on cleaved bismuth surfaces, Physics Letters A 210
(1996) 105–109.
[103] A.V. Ofitserov, V.S. Édel’man, Sts study of spectrum of surface electronic states in bismuth, Physica B 329–
333 (2003) 1094.
[104] J. Kröger, T. Greber, T.J. Kreutz, J. Osterwalder, The photoemission Fermi edge as a sample thermometer?
Journal of Electron Spectroscopy and Related Phenomena 113 (2001) 241–251.
[105] A.A. Kordyuk, S.V. Borisenko, M. Knupfer, J. Fink, Measuring the gap in angle-resolved photoemission
experiments from cuprates, Physical Review B 67 (2003) 64504.
[106] S. Hatta, H. Okuyama, M. Nishijima, T. Aruga, Temperature dependence of the charge-density-wave
energy gap of In/Cu(1 1 0), Physical Review B 71 (2005) 041401(R).
[107] G. Grimvall, The Electron–Phonon Interaction in Metals, North-Holland, 1981.
244 Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245

[108] J.E. Gayone, C. Kirkegaard, J.W. Wells, S.V. Hoffmann, Z. Li, Ph. Hofmann, Determining the electron–
phonon mass enhancement parameter k on metal surfaces, Applied Physics A 80 (2005) 943–949.
[109] A. Damascelli, Z. Hussain, Z.-X. Shen, Angle-resolved photoemission studies of the cuprate supercon-
ductors, Reviews of Modern Physics 75 (2003) 473.
[110] A. Lanzara, P.V. Bogdanov, X.J. Zhou, S.A. Kellar, D.L. Feng, E.D. Lu, T. Yoshida, H. Eisaki,
A. Fujimori, K. Kishio, J.-I. Shimoyama, T. Noda, S. Uchida, Z. Hussain, Z.-X. Shen, Evidence
for ubiquitous strong electron-phonon coupling in high-temperature superconductors, Nature 412 (2001)
510–514.
[111] Z.-X. Shen, A. Lanzara, S. Ishihara, N. Nagaosa, Role of the electron–phonon interaction in the strongly
correlated cuprate superconductors, Philosophical Magazine B 82 (13) (2002) 1349–1368.
[112] P.W. Anderson, P.A. Lee, M. Randeria, T.M. Rice, N. Trivedi, F.C. Zhang, The physics behind high-
temperature superconducting cuprates: the ‘plain vanilla’ version of rvb, Journal of Physics: Condensed
Matter 16 (2004) R755.
[113] B.A. McDougall, T. Balasubramanian, E. Jensen, Phonon contribution to quasiparticle lifetimes in Cu
measured by angle-resolved photoemission, Physical Review B 51 (1995) R13891.
[114] T. Balasubramanian, E. Jensen, X.L. Wu, S.L. Hulbert, Large value of the electron-phonon coupling
parameter (k = 1.15) and the possibility of surface superconductivity at the Be(0 0 0 1) surface, Physical
Review B 57 (1998) R6866–R6869.
[115] T. Valla, A.V. Fedorov, P.D. Johnson, S.L. Hulbert, Many-body effects in angle-resolved photoemission:
quasiparticle energy and lifetime of a Mo(1 1 0) surface state, Physical Review Letters 83 (1999) 2085–2088.
[116] M. Hengsberger, D. Purdie, P. Segovia, M. Garnier, Y. Baer, Photoemission study of a strongly coupled
electron–phonon system, Physical Review Letters 83 (1999) 592–595.
[117] S. LaShell, E. Jensen, T. Balasubramanian, Nonquasiparticle structure in the photoemission spectra from
the Be(0 0 0 1) surface and determination of the electron self energy, Physical Review B 61 (2000) 2371–2374.
[118] E. Rotenberg, J. Schaefer, S.D. Kevan, Coupling between adsorbate vibrations and an electronic surface
state, Physical Review Letters 84 (2000) 2925–2928.
[119] E. Rotenberg, S.D. Kevan, Electron–phonon coupling in W(1 1 1)-(1 · 1)H, Journal of Electron
Spectroscopy and Related Phenomena 126 (2002) 125.
[120] A. Eiguren, B. Hellsing, F. Reinert, G. Nicolay, E.V. Chulkov, V.M. Silkin, S. Hüfner, P.M. Echenique,
Role of bulk and surface phonons in the decay of metal surface states, Physical Review Letters 88 (2002)
066805.
[121] T.T. Chen, J.D. Leslie, H.J.T. Smith, Electron tunneling study of amorphous Pb–Bi superconducting alloys,
Physica 55 (1971) 439.
[122] H.K. Collan, M. Krusius, G.R. Pickett, Specific heat of antimony and bismuth between 0.03 and 0.8 k,
Physical Review B 1 (1970) 2888.
[123] J.E. Gayone, S.V. Hoffmann, Z. Li, Ph. Hofmann, Strong energy dependence of the electron–phonon
coupling on Bi(1 0 0), Physical Review Letters 91 (2003) 127601.
[124] S. Hüfner, Photoelectron Spectroscopy, third ed., Springer, Berlin, 2003.
[125] S.D. Kevan (Ed.), Angle-Resolved Photoemission, Studies in Surface Chemistry and Catalysis, vol. 74,
Elsevier, Amsterdam, 1992.
[126] W. Eberhardt, E.W. Plummer, Angle-resolved photoemission determination of the band structure and
multielectron excitations in Ni, Physical Review B 21 (1980) 3245–3255.
[127] F.J. Himpsel, Angle-resolved measurements of the photoemission of electrons in the study of solids,
Advances in Physics 32 (1983) 1–51.
[128] R. Matzdorf, Investigation of line shapes and line intensities by high-resolution UV-photoemission
spectroscopy – some case studies on noble-metal surfaces, Surface Science Reports 30 (1998) 153–206.
[129] B. Hellsing, A. Eiguren, E.V. Chulkov, Electron–phonon coupling at metal surfaces, Journal of Physics:
Condensed Matter 14 (2002) 5959–5977.
[130] E.W. Plummer, J. Shi, S.-J. Tang, E. Rotenberg, S.D. Kevan, Enhanced electron–phonon coupling at metal
surfaces, Surface Science Reports 74 (2003) 251.
[131] P.M. Echenique, R. Berndt, E.V. Chulkov, T. Fauster, A. Goldmann, U. Höfer, Decay of electronic
excitations at metal surfaces, Surface Science Reports 52 (2004) 219.
[132] J. Shi, S.-J. Tang, B. Wu, P.T. Sprunger, W.L. Yang, V. Brouet, X.J. Zhou, Z. Hussain, Z.-X. Shen, Z.
Zhang, E.W. Plummer, Direct extraction of the Eliashberg function for electron-phonon coupling: a case
study of Beð1 0 1 0Þ, Physical Review Letters 92 (2004) 186401.
Ph. Hofmann / Progress in Surface Science 81 (2006) 191–245 245

[133] C. Kirkegaard, T.K. Kim, Ph. Hofmann, Self-energy determination and electron-phonon coupling on
Bi(1 1 0), New Journal of Physics 7 (2005) 99.
[134] N.V. Smith, P. Thiry, Y. Petroff, Photoemission linewidth and quasiparticle lifetimes, Physical Review B 47
(1993) 15476–15481.
[135] E.D. Hansen, T. Miller, T.-C. Chiang, Observation of photoemission line widths narrower than the inverse
lifetime, Physical Review Letters 80 (1998) 1766–1769.
[136] J.L. Yarnell, J.L. Warren, R.G. Wenzel, S.H. Koenig, Phonon dispersion curves in bismuth, IBM Journal
of Research and Development 8 (1964) 234.
[137] C. Søndergaard, C. Schultz, S. Agergaard, H. Li, S.V. Hoffmann, Z. Li, Ph. Hofmann, C. Grütter, J.H.
Bilgram, Interplay between the electronic structure and phase transition on a-Ga(0 1 0), Physical Review B
67 (2003) 165422.
[138] T.K. Kim, T.S. Sorensen, E. Wolfring, H. Li, E.V. Chulkov, Ph. Hofmann, Electron–phonon coupling on
the Mg(0 0 0 1) surface, Physical Review B 72 (7) (2005) 075422.
[139] A. Kakizaki, M. Niwano, H. Yamakawa, K. Soda, S. Suzuki, H. Sugawara, H. Kato, T. Miyahara, T. Ishii,
A UPS study of liquid and solid bismuth using synchrotron radiation, Journal of Physics F – Metal Physics
18 (1988) 2617.
[140] T.T. Chen, J.T. Chen, J.D. Leslie, H.J.T. Smith, Phonon spectrum of superconducting amorphous Bi and
Ga by electron tunneling, Physical Review Letters 223 (1969) 526.
[141] P.G. De Gennes, Boundary effects in superconductors, Reviews of Modern Physics 36 (1964) 225.
[142] Y. Guo, Y.-F. Zhang, X.-Y. Bao, T.-Z. Han, Z. Tang, L.-X. Zhang, W.-G. Zhu, E.G. Wang, Q. Niu, Z.Q.
Qiu, J.-F. Jia, Z.-X. Zhao, Q.-K. Xue, Superconductivity modulated by quantum size effects, Science 306
(5703) (2004) 1915–1917.
[143] F. Reinert, G. Nicolay, B. Eltner, D. Ehm, S. Schmidt, S. Hüfner, U. Probst, E. Bucher, Observation of a
BCS spectral function in a conventional superconductor by photoelectron spectroscopy, Physical Review
Letters 85 (2000) 3930–3933.
[144] P. Bøggild, T.M. Hansen, O. Kuh, F. Grey, T. Junno, L. Montelius, Scanning nanoscale multiprobes for
conductivity measurements, Review of Scientific Instruments 71 (2000) 2781.
[145] S. Hasegawa, I. Shiraki, T. Tanikawa, C. Petersen, T. Hansen, P. Boggild, F. Grey, Direct measurement of
surface-state conductance by microscopic four-point probe method, Journal of Physics: Condensed Matter
14 (2002) 8379–8392.
[146] I.Y. Sklyadneva, E.V. Chulkov, W.-D. Schöne, V.M. Silkin, R. Keyling, P.M. Echenique, Role of electron–
phonon interactions versus electron–electron interactions in the broadening mechanism of the electron and
hole linewidths in bulk Be, Physical Review B 71 (17) (2005) 174302.
[147] S.K. Yip, Two-dimensional superconductivity with strong spin–orbit interaction, Physical Review B 65
(2002) 144508.

You might also like