Part II Theory
2.6 Lattice Dynamics
andΦ0is the static potential energy of the crystal. Because the force on any atom must vanish in the equilibrium configuration,
Φαlκ =0. (2.47)
As an approximation that we will revisit in the next section, we neglect terms of order three and higher in the displacements in Eq. 2.44. The Hamiltonian can be written as
Hn =X
lκ
p2lκ 2mκ
+Φ0+ 1 2
X
αlκ
X
α0l0κ0
Φαα0lκl0κ0uαlκuα0l0κ0. (2.48)
This is the harmonic approximation of lattice dynamics.
It is possbile to rewrite the Hamiltonian in matrix form
Hn=X
lκ
p2lκ
2mκ +Φ0+ 1 2
X
lκ
X
l0κ0
uTlκ Φlκl0κ0 ul0κ0, (2.49)
where Φlκl0κ0 is a 3×3 force-constant matrix defined for each atom pairlκ and l0κ0 ((l, κ)≠(l0, κ0)):
Φlκl0κ0 =[Φαα0lκl0κ0] . (2.50)
This is the famous Born-von Kármán model. [8,26] If(l, κ)=(l0, κ0),Φαα0lκlκ is a
“self-force constant”, determined by the requirement of no overall translation of the crystal
Φlκlκ = − X
(l0,κ0)≠(l,κ)
Φlκl0κ0. (2.51)
Because equal and opposite forces act between each atom of a pair, the matrix
Φlκl0κ0 must be real and symetric.
Φlκl0κ0 =
a b c b d e c e f
(2.52)
To satisfy the translational symmetry, the force constant matrices must also have the following property
Φlκl0κ0=Φ0κ(l0−l)κ0 =Φ(l−l0)κ0κ0. (2.53)
It should be noted that the interatomic force constants, such asΦαα0lκl0κ0, are in fact not constant but functions of temperature and lattice parameters.
2.6.2 Equations of Motion
Using the the harmonic approximation and the potential above, the equations of motion for nuclei are
mκulκ(t)= − X
l0,κ0
Φlκl0κ0ul0κ0(t). (2.54)
There are 3× R ×Ncell equations of motion for a finite crystal containingNcell unit cells, each containingRatoms. [8]
It is easily calculated that for any crystal of reasonable size (on the order of 1023 atoms) the atoms on the surface account for only a very small fraction of all atoms. Typical atom in that crystal has many, many atoms in any direction extending beyond the distance that interatomic forces can reach. As a result, it is convenient to apply the approximation that displacements separated by a certain number of cells are equal. Imposing these periodic boundary conditions on the crystal, the solutions to the equation of motions can be written in the form of
plane waves of wavevectork, angular frequencyωkj, and “polarization”eκj(k)
ulκkj(t) =
s 2~
N mκωkj eκj(k)ei(k·rl−ωkjt), (2.55)
= ~ v u
t2n(εkj, T )+1 N mκεkj
eκj(k)ei(k·rl−ωkjt), (2.56)
where we take the real part to obtain physical displacements. The phase factor, eik·rl, provides all the long-range spatial modulation, while the dependence onκ, a short-range basis vector index, is placed in the complex constanteκj(k). It is convenient for this constant and the exponential to have modulus unity. j is the
“branch index” discussed below and there are 3Rdifferent such branches from symmetry.
2.6.3 The Eigenvalue Problem of the Phonon Modes
The polarization vectors,eκj(k), contain all information on the excursion of each atom κ in the unit cell for the phonon mode k, j, including the displacement direction of the atom and its relative phase with respect to the other atoms. These
“polarization” vectors and their associated angular frequenciesωkj(normal modes) can be calculated by diagonalizing the “dynamical matrix” D(k). The detailed descriptions can be found elsewhere. [8,23–25]
The dynamical matrix can be obtained by substituting Eq. 2.56into2.54. It has the dimensions 3N×3N and is constructed from 3×3 submatrices Dκκ0(k)
D(k)=
D11(k) . . . D1N(k) ... . .. ... DN1(k) · · · DNN(k)
, (2.57)
where each sub-matrix Dκκ0(k) is the Fourier transform of the force-constant
matrixΦlκl0κ0
Dκκ0(k)= 1
√mκmκ0 X
l0
Φ0κl0κ0eik·(rl0−r0). (2.58)
To simplify, the equations of motions (Eq. 2.54) with the plane wave solutions (Eq. 2.56) can be represented as an eigenvalue problem:
D(k)ej(k)=ω2kjej(k), (2.59)
where
ej(k)=
ex1j(k) ey1j(k) ez1j(k) ex2j(k)
... ezNj(k)
(2.60)
is the eigenvector(s).
It can be shown that the dynamical matrix D(k)is hermitian for any value of k. As a result it is thus fully diagonalizable and the eigenvaluesω2kj are real. The eigenvectors and eigenvalues of the dynamical matrix evaluated at a particular wavevector k correspond to the 3R eigenmodes of vibration of the crystal for that wavevector. It should be noted that the angular frequency (ωkj) dependence on the wavevector (k), called a dispersion relation, can be quite complicated. The speed of propagation of phonons in solids (speed of sound) is given by the group velocity as the slope of this relation
vsound =vg= ∂ωkj
∂k , (2.61)
instead of the phase velocity
vp= ωkj
k . (2.62)
2.6.4 The Phonon Density of States
With the phonon vibrations calculated above, it is straightforward to calculate the phonon density of states (DOS) of a crystal. This is done by diagonalizing the dynamic matrixD(k)at a large number ofkpoints in the first Brillouin zone, and then binning them into the DOS histogram.
Similarly, a phonon partial DOSgd(ε), which gives the spectral distribution of motions by one atom speciesdin the unit cell, can be obtained through weighting the vibrational contribution
gd(ε)=X
k|
X
ακj
δdκ|eακj(k)|2g(ε). (2.63)
Because the eigenvalues of the dynamical matrix are normalized for eachkas
X
ακj
|eακj(k)|2=1, (2.64)
the total DOS is the sum of the partial DOSs of all atoms in the unit cell,
g(ε)=X
d
gd(ε). (2.65)
This seems simple but some software for phonon calculations does not implement it correctly. As a result, the partial DOSs do not add up to the total DOS, and the total DOS may also have an incorrect weighting on the different atom species.
2.6.5 Group Theory
With group theory, it is possible to find eigenvectors and eigenvalues without diagonalizing the dynamical matrix. [8] Maradudin and Vosko’s review showed how group theory can be used to understand the normal modes of crystal vi- brations and classify them. [27] In their work, the degeneracies of the different normal modes were addressed rigorously, and their association with the symmetry elements of the crystal were used to label them. In the same issue of the journal, Warren discussed the point group of the bond, [28] which refers to the real space symmetries of the interatomic interactions, i.e., how the force constants transform under the point group operations at a central atom.