4.
Surfaces and Interfaces
4.1 Introduction
Surface physics might seem to be unimportant because the number of atoms sitting
at a surface relative to those lying within the bulk is negligible for macroscopic bod-
ies. However, this judgment is incorrect. Surfaces provide the doorways through
which material inevitably passes when entering or leaving a solid. The state of
a material's surface determines both its strength and its resistance to chemical at-
tack. The manufacture of integrated circuits consists in the deposition of complex
sequences of surface layers to build up the desired pathways for electrons. Partly
for this last reason there is a continual rapid development of experimental meth-
ods for describing and characterizing surfaces. Characterization has in fact outrun
the immediate needs of technology, and it allows even surfaces without crystalline
order to be pictured in atomic detail.
4.2 Geometry of Interfaces
The simplest deviation from perfect crystalline behavior occurs when a crystal
comes to an end, abutting either another crystal or empty space. The first case
is a grain boundary, and the second one is a surface. As discussed by Wolf (1992),
these two types of interface can be described in a common way from the geometri-
cal point of view.
One can initially simplify the problem of interfaces by assuming that when a
perfect crystal is sliced along a plane, none of the remaining atoms moves from its
original location. This assumption is not generally true, but is often a reasonable
first approximation. The way in which it fails has to do with the fact that the crys-
tal deforms as it approaches a boundary. The deformation may take a mild form,
such as a slight increase or decrease of the volume per atom. It may involve the
formation of an entirely new crystal structure right at the interface, called surface
reconstruction. Or it may involve the formation of an entirely new phase; for ex-
ample, a thin liquid layer on an otherwise solid body. The region of deformation
relative to the perfect crystal is called the selvage.
Ignoring surface reconstructions and other deformations, a plane interface be-
tween two crystals may be described by ten variables, describing the ways that two
perfect planar interfaces can be brought together. Three of the variables describe
the precise microscopic three-dimensional positioning of one crystal with respect
to another, as shown in Figure 4.1(A): exactly how far the two crystals are from
one another, and how far each has been slid along the other in the two in-plane
77
78 Chapter 4. Surfaces and Interfaces
directions. The seven remaining variables have macroscopic significance. Each of
the two crystals has been sliced along a plane. The conventional way to describe
planes is with Miller indices, using three numbers, but of course to describe a plane
in three dimensions needs only two numbers, the angles along which its unit nor-
mal points. However, one needs to describe the precise point along the unit normal
where the cut is made. So specifying the two crystal planes takes a total of six
numbers. A tenth and final variable Θ is needed to specify the relative orientations
of the two crystals as they rotate about in the plane where they meet, shown in
Figure 4.1(B).
Figure 4.1. (A) Three microscopic variables describe the precise positioning of two ideal
interfaces with respect to one another. (B) Angle specifying the relative rotations of two
crystal interfaces.
For a free surface, only two variables need to be described; those giving the
plane along which it cuts the crystal.
4.2.1 Coherent and Commensurate Interfaces
When two crystal surfaces meet, it is valuable to know if they will mesh well with
one another. This question is particularly interesting if one is building a layered
structure consisting of alternating crystal types and wants to know if they will bind
together. If two crystal surfaces are placed together and the atoms are in perfect
registry with one another, then one has a coherent interface, and the process of
growing such an interface is called epitaxy. There is a related concept, that of a
commensurate interface which is slightly more general. Two interfaces are com-
mensurate if there is some larger two-dimensional lattice on which their atoms
coincide, as shown in Figure 4.2. The general condition for two lattices to be
commensurate is slightly complicated because one must include the possibility of
rotating them at some peculiar angle, as in Figure 4.2. Let a\, a-i and b\, &2 be
two sets of primitive vectors for the atoms on the two interfaces. Then the two are
commensurate if there is an infinite set of integers n\,ri2 and m\, m-i, together with
Geometry of Interfaces 79
an angle Θ such that
- , - / cos 6> sin ö \ , r
n\al+n2a2=\ . ,, „ \ [m\b\ + m2bτ2).Λ (4.1)
I — sin #cos 0 /
The set of points on which the two lattices agree is called the superlattice. One
can always define b\ and b2 so that the rotation angle in Eq. (4.1) is zero, although
it is included for generality.
Figure 4.2. Two square lattices whose lattice constants differ by VS/2 are commensurate
but incoherent, because not all the atoms lie on top of one another.
4.2.2 Stacking Period and Interplanar Spacing
Another interesting geometrical quantity is the stacking period, which is defined by
counting the number of lattice planes one encounters, while heading directly away
from the interface, before hitting a new lattice plane that has all its atoms sitting
directly above those on the interface, along the normal. There are some fairly
general results for crystals of cubic symmetry. If the Miller index of a surface is
(ijk), then the stacking period in a cubic crystal is
P = ö(i2 + j 2 + k2), (4-2)
where δ equals 1 or 2, depending on the particular lattice being considered and
whether /, j , and k are odd or even, as shown in Figure 4.3.
Probably more important is the interplanar spacing
d = ea/yji2 + j 2 + k2, (4.3)
where e equals 1/2 or 1 depending upon circumstances. In fee crystals, for ex-
ample, e is 1 if i, j , and k are all odd, but otherwise it equals 1/2. Because the
80 Chapter 4. Surfaces and Interfaces
fee ( 100) surface, P = 2 fee ( 111 ) surface, P = 3
Figure 4.3. The stacking periods of the (100) and (111) planes for the fee lattice. In the
first case, δ of Eq. (4.2) is 2 and e of Eq. (4.3) is 1/2; in the second case, δ = 1 and e = 1.
product of the interplanar spacing and the area per atom on the interface must
be constant, interfaces with large interplanar spacings have the highest density of
points on the surface. This situation is energetically favorable if two crystals are
brought together, because it gives them opportunities to bond that resemble those
of the unbroken crystal. For this reason, interfaces with small Miller indices, (110),
(100), and (112), occur frequently in practical situations.
Yet another concept, not so easy to define precisely, is the number of nearest-
neighbor bonds one must snap in order to build a given interface. This idea is
connected approximately with surface energy, although the energy needed to form
a surface cannot usually be encompassed completely by so simple a calculation as
the snapping of nearest-neighbor bonds. The process of snapping bonds is rela-
tively easy to visualize in simple cases, such as the (100) surface of an fee lattice,
shown in Figure 4.4.
There is a large vocabulary that has developed to describe the geometries of
Figure 4.4. Diagram showing the bonds broken in cutting along the (001) plane in an
fee crystal. For planes at oblique angles, determination of the analogous quantity is more
difficult, particularly because it may be advantageous for the surface of separation to rise
and fall slightly on the atomic scale to cut the smallest number of bonds.
Geometry of Interfaces 81
interfaces. If a single crystal is sliced along some plane, and the top half turned
upside-down and replaced, the result is called a twin boundary, shown in Figure
4.5. The grain boundary in this case is a mirror plane; the high degree of symmetry
makes such boundaries energetically favorable and quite common. If instead one
rotates the top half of the crystal through some angle around a normal to the inter-
face, one produces a twist boundary. If a single crystal has a wedge cut out of it and
is then glued back together, the result is a tilt boundary; if the wedge is small, the
result may be called a low-angle grain boundary. Finally, if the two halves of the
crystal are brought back together without any tilting or rotation, but out of registry,
the interface is a stacking fault.
1 nm
Ni· Al O
0
(110) X C?.0»#o 0
o no / ^
/Tuo
Λ
(001) r 1 (001)
(A) (B)
Figure 4.5. (A) (111) twin boundary in NiAl visualized with atomic resolution in a high-
resolution electron micrograph. (B) Sketch clarifying locations of atoms. The mirror sym-
metry between right- and left-hand sides virtually eliminates defects along the interface.
[Source: Nadarzinski and Ernst (1996), p. 651.]
4.2.3 Other Topics in Surface Structure
Because surfaces are open to the environment, they easily form thin layers of new
structures, generically referred to as adsorption. If the attraction of the foreign
material is weak, it is called physisorption. De Gennes (1985) reviews the phe-
nomenon of wetting in which a thin fluid layer covers a solid. Gomer (1975) re-
views chemisorption, where a chemical species forms a strong chemically bonded
layer. All these topics are discussed in more detail by Zangwill (1988).
