We have indicated in the preceding section how energy bands arise from the existence of the translation group in a one-dimensional crystal. We will first generalize that description to three dimensions and then look in some detail at the nature of the energy bands themselves.
1. The Structure off the Bands
We have defined for an arbitrary structure the primitive lattice translations Ļ1, Ļ2, and Ļ3. The translation group of the crystal consists of sums of integral multiples of these three translations. For a general crystal structure these Ļi may be of different lengths and they need not be orthogonal. However, we have required that they not be coplanar.
We will again take periodic boundary conditions as we did in the one-dimensional case. We consider a system in the shape of a parallelepiped with edges N1Ļ1, N2Ļ2, and N3Ļ3, In a translation parallel to Ļ1, for example, we envisage the atoms transferred out of the boundary as reentering on the opposite face so that the crystal within our boundaries remains invariant under the translation. In addition, the periodic boundary conditions require that the slope and value of any wavefunction be the same on opposite faces of the system. The translation group contains N1N2N3 distinct translations, which may be written
with 0 ā¤ ni < Ni. In analogy with the one-dimensional case the irreducible representations are written
where k is now a vector and a dot product appears in the exponent. Each k corresponds to a different irreducible representation of the group. These kās may be written as linear combinations of primitive lattice wavenumbers, k1, k2, and k3.
where the ki are integers and
Substituting Eq. (2.1) into Eq. (2.2) and writing out k we may readily verify that these representations have the same multiplication table as the translation group and are therefore representations of the group. We have distinct representations for Ni consecutive integers ki. Thus there are as many irreducible representations given in Eq. (2.2) as there are symmetry operations in the translation group and this gives all of the irreducible representations of the translation group.
We again characterize each electron state by the wavenumber k corresponding to the irreducible representation according to which it transforms.
We again define Bloch functions uk = eāikā
rĻk. In terms of these the wavefunction is
and uk is seen to have the full translational symmetry of the lattice.
We may note that a representation corresponding to a wavenumber k is the same as a representation corresponding to
where the mi are integers. The addition of these extra terms does not modify the value of any of the representations of Eq. (2.2). Thus we have an arbitrariness in the choice of k which is to be associated with any given representation. It is desirable to prescribe a unique choice and this is customarily done by selecting for any representation the shortest wavenumber that will generate it. (This is consistent with our selection in the case of one dimension.) The resulting domain of k is the three-dimensional Brillouin zone, and the wavenumbers in the Brillouin zone are frequently called reduced wavenumbers.
Note that the set of wavenumbers m1k1 + m2k2 + m3k3 forms a lattice in wavenumber space. We will call this the wavenumber lattice and refer to the points making up the wavenumber lattice as lattice wavenumbers. In diffraction theory it is more customary to discuss the reciprocal lattice which is the corresponding lattice in reciprocal space. The reciprocal lattice vectors are (2Ļ)ā1 times the wavenumber lattice vectors. It is common terminology in solid state physics to call the wavenumber lattice the reciprocal lattice. However, in order to avoid the ambiguity with respect to the factor 2Ļ, we will use the term wavenumber lattice. The symbols k1, k2, and k3 are primitive lattice translations in the wavenumber lattice.
Before proceeding with the formal consequences of translational symmetry, it may be helpful to give a more pictorial description of the states and see the meaning of the multiple wavenumbers which could be associated with a given state. We are assuming a potential that has the full translational symmetry of the lattice. This is illustrated for a line through a series of atoms in Fig. 2.1a. Because of the periodicity this potential could be expanded in a Fourier series including only plane waves with wavenumber equal to a lattice wavenumber. (This follows in detail from the Fourier integral and will be demonstrated in Sec. 4.2 of this chapter.)
We have seen that the electronic eigenstates, illustrated in Fig. 2.1b, can be factored into a Bloch function, uk, and a plane wave, eikā
r, shown in Fig. 2.1c and 2.1d. The plane wave (as well as uk) satisfies periodic boundary conditions. Because uk has the full periodicity of the lattice, it could be expanded in a Fourier series including only lattice-wavenumber plane waves. It follows that the eigenstate can be expanded in a Fourier series containing plane waves of wavenumber k and wavenumbers differing from k by lattice wavenumbers; these are just the wavenumbers that generate the representation according to which Ļk transforms.
Consequently, if we could precisely measure the true momentum of an electron in the state Ļk, we could obtain a value corresponding to any one of these wavenumbers. The state Ļk is not an eigenstate of true momentum since by moving in a potential field the electron is continuously exchanging momentum with the lattice, but because of the periodicity it does not contain components of all momentum values, just those corresponding to the wavenumbers which generate the appropriate representation. Such a description is one way of giving physical content to the ambiguity in the wavenumber to be associated with a state.
Fig. 2.1 A schematic representation of electronic eigenstates in a crystal. (a) The potential plotted along a row of atoms. (b) A sample eigenstate; the state itself is complex but only the real part is shown. This state can be factored into Bloch function (c), which has the periodicity of the lattice, and (d) a plane wave, the real part of which is shown.
We have found that a given wavenumber k ...