Nearly-Free Electrons Model

Jacob Shapiro December 18, 2012

Abstract The following is my attempt to explain to myself what is going on in the discussion of the Nearly-Free Electron Model which can be found in Ashcroft & Mermins SSP book.

Introduction

We begin with some mathematical reformulation of the various functions in the system. The potential, V (r) has the periodicity of the lattice. That means that it will be equal with any arbitrary translation by an arbitrary Bravais lattice vector R. Explicitly, this means V (r + R) = V (r). Thus we may use a Fourier expansion of V (r), which will contain only reciprocal lattice veciKr tors: V (r) = , where K is a reciprocal lattice vector, and K V (K) e 1 (0) is V (K) = volume of unity cell unity cell d3 rV (r)eiKr . Observe that V the constant part of the potential (the part that does not depend on the location r) and so can be regarded as an arbitrary clibration of the energy scale. So for (0) = 0. In addition, since convenience we can pick our energy scale so that V 1 the potential is real, V (K) = volume of unity cell unity cell d3 rV (r)eiKr = (K). If, by suitable choice of the origin, the potential also has inversion V (K) = V (K). symmetry (V (r) = V (r)) then it is clear that V The wave-function will also undergo similar treatment, though, not precisely the sameit is not lattice-periodic. According to Blochs theorem, the wavefunction in a lattice-periodic potential (such as the one dened above) can be written as k (r) = eikr uk (r), where uk (r) are functions that also have the periodicity of the lattice, and k is any vector inside the rst Brillouin zone (in fact k+K (r) = k (r), where K is a recirpocal lattice vector). So we may expand uk (r) as a Fourier series, just like the potential: uk (r) = K u k (K) eiKr . Thus ikr ikr our Bloch-wave-function is: k (r) = e uk (r) = e Ku k (K) eiKr = i(k+K)r k (K) e . Ku Now our problem is not to nd k (r) and Ek , but rather, u k (K) and Ek , for any arbitrary vector k in the rst Brillouin zone and any arbitrary reciprocal lattice vector K.

Schroedingers Equation
2

Schroedingers equation is given by: 2 k (r) + V (r)k (r) = Ek k (r)

2m

Putting in our expansions into the equation we get:

2

2m

2
K

u k (K)ei(k+K)r +
K

(K )eiK r V
K

u k (K)ei(k+K)r = Ek
K

u k (K)ei(k+K)r

Applying the derivative and factoring by eikr :

2

2m

(k+K)2 u k (K)eiKr +
K K

(K )eiK r V
K

u k (K)eiKr = Ek
K

u k (K)eiKr

To take care of the product V (r) k (r) term we employ the following: (K ) eiK r V
K K

u k (K) eiKr =
K ,K

(K ) u V k (K) ei(K+K )r

Change of index: K := K + K , and summing over K instead of K , we get: (K K) u V k (K) eiK

K ,K r

Change the names of the indices, to the following: K K, K K and arrive at:

(K K ) u V k (K ) eiKr =
K,K K K

(K K ) u V k (K ) eiKr

Going back to the Schroedinger equation we get:

2

eiKr
K

2m

(k + K)2 Ek u k (K) +
K

(K K ) u V k (K )

=0

Since this equation is true for any arbitrary r, it follows that the expression inside the curly brackets has to be zero for any K: 2

2m

(k + K)2 Ek u k (K) +
K

(K K ) u V k (K ) = 0

(1)

Finally, we arrived at a set of equations (as many as there are Ks innitely many) for each value of k that shall determine the co-ecients u k (K). Note that we performed all the treatment where k was a xed parameter for the Schroedinger equation. The innitely many solutions for each value of k are labeled with the band-index n. Also note that since according to Blochs theorem the wave-function has periodicity of reciprocal lattice vectors, we only have to describe the solutions where k lies in the rst Brillouin zone.

Extreme CasesZero periodic potential

It can be shown (see Ashcroft & Mermin, Problem 8.1) using Blochs theorem that the relation between the energy in the lattice and crystal momentum obeys the following relation, for any aribtrary shape of periodic 1-D potential: cos(ka) = cos(a + ) |t|
2 2

where is related to the energy by E = 2m , whereas if the arbitrary potential barriers transmission co-ecient can be written as t = |t|ei , then |t| and assume those meanings then. This relationship assumes only one thing about the potential, namely, its periodicity. Any other information about the shape of the potential is embodied in the two parameters |t| and . At any rate, if we assume that the only eect of the potential is its periodicity, but otherwise, its not just weak, but completely zero, we get a special situation which is similar to the free-electron, but is also periodic. In such a case we assume that |t| = 1 and = 0 and so our equation simply becomes:

cos(ka) = cos(a) which has the solution: a = ka + 2n , n Z Which means that we get the following relationship between the energy and momentum:
2 2 n)2 = (k + Kn )2 2m a 2m This is quite an interesting result: we get the same free-electrion dispersion relation which we would expect, but there are additional indexed soltuions in periodicity of reciprocal lattice vectors. In other words, these are our bands (although, at the moment, there are no band gaps, youll notice that at the edges of the Brillouin zones as well as in its middle we have the bands meeting). 2

En (k ) =

(k +

When turning on the potential, we would have to then apply degenerate (or nearly degenerate) perturbation theory at these places where, for a given value of k , several (in 1D, only two) values of n give the same energy. The potential will lift the degeneracy at those places. Generalizing our result to 3 dimensions we would expect to arrive at:
2

(k + Kn )2 2m Using this information, we go back to our momentum-space Schroedinger equation, setting the potential to exactly zero: En (k) =
2

2m

(k + K)2 Ek u k (K) = 0

(2)

In this set of equations (for each k there will be as many equations as there are Kinnitely many), we are looking for u k (K) and Ek . This set of equations will have an innite number of solutions, which we shall conveniently index by K (this is our band index, n). 2 For a given value of k, and a given value of Kn , if 2m (k + K)2 is special (meaning no other value of K will give that same energy value), then only one of the co-ecients of u k (K) (namely, u k (Kn )) can be freely determined, while the rest of the co-ecients will be zero. Thus we arrive at a solution of the 2 form n,k (r) = u k (Kn ) ei(k+Kn )r and En (k) = 2m (k + Kn )2 , where all the co-ecients u k (K) = 0 for all K = Kn . If, however, there is a set of N such 2 2 2 2 2 2 values of K that obey 2m (k + K1 ) = 2m (k + K2 ) = ... = 2m (k + KN ) , then the coecients of the items in this set are again freely determined: k (r) = u k (K1 ) ei(k+K1 )r + u k (K2 ) ei(k+K2 )r + ... + u k (KN ) ei(k+KN )r ; all other co-ecients are zero: u k (K) = 0 K =K1 , K2 , ..., KN . As a concrete example, consider the one-dimensional lattice. In this setup, there are three special points in the Brillouin zone: k = 0, a . Any k value not equal to any of these three points will be result in a set of equations which have only one value of K equal to one particular energy. The wave-function will thus be only one plane-wave. However, for k = 0, a , there are, for one value of energy, two values of K that give rise to said energy value, and as a result the wave-function will be a linear combination of two plane-waves (one travelling left and the other right):

For more complicated lattices, like 2D, and 3D setups, there will be more than two values of K at the edge of the Brillouin zone resulting in the same energy. In the F.C.C., for example, at the W-point at the edge of the 1st Brillouin zone, there will be four dierent values of K which will result in the

same energy: . This is, thus, the starting point for our treatment of a weak periodic potential, which will have to employ nearly-degenerate-perturbation theory.

Weak Perturbing Potential

Now V (r) is weak but not zero, and we can analyze equation (1) to see what happens to it. We shall distinguish between clearly non-degenerate cases and nearly-degenerate cases.

4.1

Non-Degenerate Case

For any given value of k, there might be bands that are very near to each other, or bands that are very far apart. For example, for the 1D lattice, at k = 0, the lowest band (K = 0) is far apart from all other bands, or at k = 1 2 a , for any value of K, the bands are quite far apart. We would like to study how the latter type of bands (at those positions in k-space) are aected by the perturbing potential. Let us mark that one band with the index 1. For convenience well 2 dene q := 2m q2 . In the unperturbed case, we found that u k (K1 ) = 1 and u k (K) = 0 K = K1 . We thus assume that the potential is so weak that its k K = K1 , eect on the system is such that we could say that now, u k (K) V k is a typical Fourier-series co-ecient. In addition we assume that the where V k K = K1 . Equation energy levels are spaced such that | k+K E1 (k)| V (1) for the K1 band becomes: (
k+K1

+ E1 (k)) u k (K1 ) =
K

(K1 K ) u V k (K )

(3)

(0) = 0, the sum Observe that since we dened our energy scale such that V on the right hand side of the equation is over such K values such that K = K1 : (
k+K1

+ E1 (k)) u k (K1 ) =
K =K1

(K1 K ) u V k (K )

That same equation for any other band (K = K1 ) would be: ( + E1 (k))u k (K) =
K

k+K

(KK )u (KK1 )u V k (K ) = V k (K1 )+

K =K1

(KK )u V k (K )

Put dierently: u k (K) = (K K1 ) u V k (K1 ) + k+K + E1 (k) (K K ) u (K K1 ) u V k (K ) V k (K1 ) k )2 ) = +O((V k+K + E1 (k) k+K + E1 (k)

K =K1

If we put this result for the other bands coecients in equation (3) back:

k+K1

+ E1 (k))u k (K1 ) =
K =K1

k (K1 ) k )2 ) = (K1 K ) V (K K1 ) u V + O((V k+K + E1 (k)

=
K =K1

(K1 K )V (K1 K ) u V k (K1 ) k )3 ) + O((V k+K + E1 (k)

u k (K1 ) cancels out and we get: E1 (k)

k+K1

=
K =K1

(K1 K )|2 |V k )3 ) + O((V E1 (k) k+K

k )2 ), then obviously if we want to Since E1 (k) k+K1 is of the order of O((V solve for E1 (k) up to that order, we can substitute the E1 (k) in the denominator of the right hand side of the equation with k+K1 to arrive at the result of: E1 (k) =
k+K1 + K =K1 2 (K1 K )|2 |V k )3 ) +O((V (k+K1 )2 + = 2m k+K1 k+K

2m
2

K =K1

K1

(K1 K )|2 |V + K 2 + 2k(K1 K )

This result is important mainly in that it shows us that the lowest order corrections to these non-degenerate states is second order !

4.2

Nearly-Degenerate Case