Professional Documents
Culture Documents
Marvin L. Cohen
University of California, Berkeley
Steven G. Louie
University of California, Berkeley
c Cambridge University Press 2016
1
Acknowledgement
The authors thank Mr. Meng Wu, together with Mr. Felipe H. da Jornada,
Mr. Fangzhou Zhao and Mr. Ting Cao, for their invaluable help in preparing
this problem solutions manual.
2
Sec. I
I.1. Crystal structure of MgB2 .
(a) The unit cell and Wigner-Seitz cell are displayed in Fig. 1.
Top view Side view
B1 B2 a2
Unit
cell
a3
Mg
B2 B1 Unit cell Mg Mg
Wigner-Seitz cell
B1 B2 a1 B B B
a1
y z
Wigner-Seitz
cell
x Mg
3
Top view Side view
b2
kz
b3 Reciprocal
Reciprocal
cell cell
Brilluoin
zone
b2
b1 ky Brilluoin zone
kx
(c) There are 24 point symmetry elements in MgB2 that leave the crystal structure invariant.
There are 12 in-plane symmetry elements: 6 rotation operations, including the identity oper-
ation: C6 , 2C6 , · · · , 6C6 = I. There are also 6 in-plane mirror symmetries that go along the B
atoms or between the B atoms, including, for instance, the mirror symmetry along the x = 0
and y = 0 planes, denoted by σx and σy , respectively. One can also compose any of these 12
in-plane symmetry elements with a mirror operation along the z = 0 plane, σz . This gives
the total of 24 symmetry elements. Note that this system has inversion symmetry, which is
equivalent to σz C2 .
(d) The irreducible part of the Brillouin zone is sketched below. It can be easily verified that, if
we apply all the 24 symmetry elements, we will recover the full Brillouin zone.
ky kz
Irr. BZ Irr BZ
kx
Figure 3: Irreducible BZ
I.2. The GaN crystal. This problem is similar to Problem I.1, except that a here is the length
of the in-plane lattice vectors. The details of the computation will be omitted.
(a) We will choose a different orientation for the unit cell vectors with respect to problem I.1. The
4
lattice vectors are given by √ !
1 3
a1 = a, − a, 0
2 2
√ !
1 3 (3)
a2 = a, a, 0
2 2
a3 = (0, 0, c) ,
in Cartesian coordinates. Note that, in real space, the Wigner-Seitz cell is a hexagon rotated
90◦ relative to that in problem I.1. √
The unit cell volume is Ω = |(a1 × a2 ) · a3 | = 23 a2 c, and the reciprocal lattice vectors are,
2π 1
b1 = 1, − √ , 0
a 3
2π 1
b2 = 1, √ ,0 (4)
a 3
2π
b3 = (0, 0, 1) ,
c
16π 3
in Cartesian coordinates. The Brillouin zone volume is ΩBZ = √
3a2 c
. The Brillouin zone is
sketched in Fig. 4.
b3
kz
b2
Brilluoin ky
zone
kx b2
Brilluoin zone
b1
Figure 4: BZ
(b) If we inspect the unit cell of GaN. which is enclosed by solid lines, we see that there are 2 N
atoms (1 spread throughout the corners, and 1 inside), and there are 2 Ga atoms inside (1
spread along the edges of the unit cell, and 1 completely inside the cell). So, there are 4 atoms
in the unit cell.
(c) A top view of the structure is displayed in Fig. 5.
There are 6 point symmetry elements that do not involve glide or screw axes: 3 mirror planes
along the atoms that form the hexagons, and 3 C3 rotation operations (including identity),
performed around the center of the hexagons.
(d) Ga has 3 valence electrons, and N has 5 valence electrons. So, one cell has 16 electrons, or
8 filled bands. There are no half-filled bands. so GaN is an insulator. So, it’s not surprising
that this material has a bandgap.
I.3. Born-Oppenheimer approximation for a molecule. Think about the simplest crude
model that captures the physics of this problem. Imagine that we have two rigid ions with mass M
5
and effective charge +Q each, separated by a distance x. Consider that you put a single effective
point charge of mass me and charge −2Q between the two ions, which corresponds to the effective
charge created by the bond. At rest, we know that x = a.
The electronic contribution to the energy can be written as a sum over kinetic energy and potential
energy, and we assume that the kinetic energy comes essentially from quantum confinement via
the uncertainty principle ∆p∆x ∼ ~,
~2
Our model has a free parameter, Q. But we have a constraint: E 0 (x = a) = 0 =⇒ Q2 = ame . So,
h i
~2 1 1
E(x) = me 2x2
− xa .
2
The electronic energy is the total energy at x = a, so Eel ∼ − 2m~e a2 .
The vibrational energy can be obtained if we realize that the total energy landscape as a function
of x can be approximated by a quadratic curve near x = a. This gives rise to a harmonic potential,
wherepthe vibrational modes are quantized, Evib = ~ω. But since we have a harmonic oscillator,
2
00
ω = k/M with k = E (x = a). Thus, Evib ∼ a2 √~M m .
e
Finally, the rotational energy can be calculated if we assume that the rotational motion is quan-
2
2~2
tized. We get that Erot = ~ `(`+1)
2I with I = M (a/2)2 . This gives Erot ∼ M a2
.
the ratio of the electronic to vibrational to rotational energy is given by Eel : Evib : Erot ≈
So, q
1: m me
M : M . The consequence of this finding is that, since M me , we can separate the three
e
degrees of freedom, and calculate the contribution to the vibrational and rotational degrees of
freedom as a perturbation to the electronic one. This validates the use of the Born-Oppenheimer
approximation, which treats the ions as fixed particles when solving for the electronic contribution.
6
I.4. Hartree-Fock approximation. Let us first derive some useful identities related to Slater
determinants. Let’s define the antisymmetrization operator A as
1 X
 = (−1)P P̂ , (9)
N! P
where N is the number of particles and occupied orbitals, P is a permutation, and P̂ is the
operator associated to the permutation P . If the overall permutation contains an odd number
of pairwise permutation, (−1)P = −1, otherwise the permutation is even and (−1)P = 1. The
identity permutation is even. For instance, consider the following permutations and their signs:
P = (1, 2, 3, 4, . . . , N ) → (−1)P = 1
P = (2, 1, 3, 4, . . . , N ) → (−1)P = −1 (10)
P
P = (2, 3, 1, 4, . . . , N ) → (−1) = 1
where φ is a single-particle orbitals, the subscript i in φi (j) denotes the orbital index, and the
argument j denotes the space-time coordinate. We take a spinless wavefunction for the sake of
simplicity, although the proof here can be trivially extended to the case with spin. We define the
permutation operator P̂ acting on Ψ by permuting the orbitals indices,
In addition, Â commutesh with anyi operator that preserves particle indistinguishability, such as
the crystal Hamiltonian, Â, Hxtal = 0. In particular, we can show that:
fˆ(i1 , . . . , iM ) = 0
X
Â, (14)
i1 ,...,iM
7
We will now particularize to the case where i = 1 and j = 2, Since ĝ(1, 2) doesn’t depend on any
coordinate larger than 2 and the orbitals are orthogonal to each other, we can separate the orbitals
into two sets, one with indices 1 or 2, where ĝ(1, 2) will act upon, and another set with indices
larger than 2. There can be no cross-terms between orbitals because they are orthogonal. This
way, we get:
X
hΨHF |ĝ(1, 2)|ΨHF i = (−1)P hφ1 (1) . . . φN (N )|ĝ(1, 2)|φP1 (1) . . . φPN (N )i
P
X
= (−1)P hφ1 (1)φ2 (2)|ĝ(1, 2)|φP1 (1)φP2 (2)i
P
X
= (−1)P hφ1 (1)φ2 (2)|ĝ(1, 2)|φP1 (1)φP2 (2)i (16)
P
× hφ3 (3)|φP3 (3)i · · · hφN (N )|φPN (N )i
X
= (−1)P hφ1 (1)φ2 (2)|ĝ(1, 2)|φP1 (1)φP2 (2)i δ3 P3 · · · δN PN
P
Now, there are only two permutations such that Pi = i for i > 2, namely:
P = (1, 2, 3, 4, . . . , N ) → (−1)P = 1
(17)
P = (2, 1, 3, 4, . . . , N ) → (−1)P = −1
So,
hΨHF |ĝ(1, 2)|ΨHF i = hφ1 (1)φ2 (2)|ĝ(1, 2)|φ1 (1)φ2 (2)i
− hφ1 (1)φ2 (2)|ĝ(1, 2)|φ2 (1)φ1 (2)i (18)
.
= hφ1 φ2 |ĝ|φ1 φ2 i − hφ1 φ2 |ĝ|φ2 φ1 i,
where we assume that the coordinate indices (e.g., (1) and (2)) follow the same order in the bras,
kets, and in the operators. A generalization for other orbital indices give relation 3, as desired.
Relations 1 and 2 are easier cases that can be proved in the same fashion, but in those cases only
the identity permutation yields a non-zero contribution to the matrix element.
Now, we can write the crystal Hamiltonian as
N
X N
X
Hxtal = ĥ(i) + ĝ(i, j), (19)
i i6=j
In order to find the Hartree-Fock equations, we wish to minimize the expectation value of the total
energy with respect to the single-particle orbitals that make up the Hartree-Fock wavefunction,
with the constraint that the orbitals are orthonormal, which can be introduced via a Lagrange
multiplier,
L = hΨHF |Hxtal |ΨHF i − λ (hΨHF |ΨHF i − 1) (21)
8
We set δL = 0 and obtain,
N N
(N )
X 1X X
δ hφi |ĥ|φi i + [hφi φj |v̂|φi φj i − hφi φj |v̂|φj φi i] −δ hφi |φi i =0 (22)
i
2 ij
i
Note that we don’t have to impose orthogonality constraint because, for small perturbations in
the orbitals, the change in the orthogonality will be quadratic in |δφi i. Then,
N N N
X X 1X
hδφi |ĥ|φi i − λ hδφi |φi i + [hδφi φj |v̂|φi φj i − hδφi φj |v̂|φj φi i]
i i
2 ij
(23)
N
1 X
+ [hφi δφj |v̂|φi φj i − hφi δφj |v̂|φi φj i] + h.c. = 0
2 ij
Note that the double sums can be combined into a single sum, since v̂(1, 2) = v̂(2, 1). We seek
critical solutions for any orbitals. Because the linear independence of the orbitals, we only need
to solve the linear system for the bras. For a given orbital with index i, we obtain
N
X
ĥ|φi i + [hφj (2)|v̂|φi (1)φj (2)i − hφj (2)|v̂|φj (1)φi (2)i] − λ|φi i = 0. (24)
j
For the second part of the question, we assume that we promote one electron from an occupied
orbital a to an unoccupied orbital b, i.e., we assume that single-particle orbitals remain frozen, and
9
that orbital a, which was previously occupied, becomes unoccupied, and that orbital b, which as
previously unoccupied, is now occupied. We need to assume that single-particle orbitals are frozen:
if we solve the HF equations self-consistently, we always arrive at the ground-state orbitals). The
excitation energy ∆E is the energy difference from the ground-state energy, E 0 , to the excited
state energy, E a→b .
The energy difference is composed of two parts: ∆E 1 , which depends on ĥ(1), and ∆E 2 , which
depends on ĝ(1, 2). It is clear that ∆E 1 = hφb |ĥ|φb i − hφa |ĥ|φa i. The change ∆E 2 is a little bit
harder to compute as the whole Fock operator changes as an electron gets promoted,
N
1X
∆E 2 = [hφb φj |v̂|φb φj i − hφb φj |v̂|φj φb i]
2 j
(29)
N
1X
− [hφa φj |v̂|φa φj i − hφa φj |v̂|φj φa i]
2 j
Note that, within the HF approximation, the excitation energy is not just the difference between the
eigenvalues, but it also contains a correction term, which can be understood as an approximation
to the exciton binding energy.
-3π/a -2π/a -π/a π/a 2π/a 3π/a -3π/a -2π/a -π/a π/a 2π/a 3π/a -3π/a -2π/a -π/a π/a 2π/a 3π/a
(a) (b) (c)
Figure 6: Simple bandstructure diagrams for a one dimensional periodic solid in the limit of V (r) →
0 expressed in the extended zone (a), repeated zone (b), and reduced zone (c) scheme.
(b)
Ω ΩΩBZ Ω
X Z
1= d3 k = = =N (31)
k
(2π)3 (2π) 3 Ωc
10
(a) Semiconductors:
• Diamond: Si, Ge, Sn(grey tin);
• Rhombohedral: B, P(red phosphorus);
• Hexagonal chains: Se, Te.
(b) Semimetals:
• Graphite/hexagonal: C;
• Rhombohedral: As, Sb, Bi.
(c) Metals in fcc:
• Main-group metals: Al, Ca, Sr, Pb;
• Transition metals: Ni, Cu, Rh, Pd, Ag, Ir, Pt, Au.
(d) Metals in bcc:
• Main-group metals: Li, Na, K, Rb, Sn(white tin), Cs, Ba;
• Transition metals: V, Cr, Fe, Nb, Mo, Ta, W.
(e) Metals in hcp:
• Main-group metals: Be, Mg, Tl;
• Transition metals: Sc, Ti, Co, Zn, Y, Zr, Tc, Ru, Cd, Hf, Re, Os.
I.7. Point group. Let ai be the lattice translation vectors and bi the reciprocal lattice vectors.
If R(θ) is a symmetry operation, then
X (k) (k)
Rak = mi ai , mi ∈Z (32)
i
we have,
1
B · [RA] = (matrix of
integers ) (34)
2π
Note that,
1
Tr[ B · R · A] = integer (35)
2π
Using Tr[XY Z] = Tr[Y ZX], we get,
1 A·B
Tr[ B · R · A] = Tr[R ] = Tr[R] (36)
2π 2π
Expand R(θ) as,
cos θ sin θ 0
R(θ) = − sin θ cos θ 0 (37)
0 0 1
we get,
Tr[R] = 2 cos θ + 1 = integer (38)
or, θ = n(60◦ , 90◦ , 120◦ , 180◦ , 360◦ ).
11
I.8. Lattice sums. Suppose our crystal has N = N1 N2 N3 unit cells with ai the lattice vectors
and bi the reciprocal lattice vectors.
Rn = n1 a1 + n2 a2 + n3 a3
(39)
q = q1 b1 + q2 b2 + q3 b3
where qi = mi /Ni , 0 ≤ ni < Ni and mi , ni ∈ Z. The condition on the qi comes from the imposition
of periodic boundary conditions for a finite-size crystal.
(a) Then:
3 Nj −1
X X Y X
eiq·Rn = ei2π(q1 n1 +q2 n2 +q3 n3 ) = ei2πqj nj (40)
n n1 n2 n3 j=1 nj =0
We can think of the terms of the above sum as representing αj evenly spaced points on the
complex unit circle, each point with multiplicity βj , where αj βj = Nj . It’s easy to see that
these complex points will sum to zero unless qj is an integer, in which case the summand is
always 1. Furthermore, the product will vanish unless the above condition is true for all j,
which is equivalent to saying that q is a reciprocal lattice vector. Therefore, if q is a reciprocal
lattice vector, G, then,
3 Nj −1
X Y X X
eiG·Rn = 1 = N1 N2 N3 = N =⇒ eiq·Rn = N δq,G (41)
n j=1 nj =0 n
This expression can be reduced to the expression in (a) by noting that in (a) qj = mj /Nj .
Therefore, in this part, the sums will be zero unless the nj /Nj are all integers. Since 0 ≤ nj <
Nj , the only possibility is if n1 = n2 = n3 = 0, in which case we have,
X
eiq·Rn = N1 N2 N3 δRn ,0 = N δRn ,0 (44)
q∈BZ
12
(b) Use
3/2
Ω 2mEF
N (EF ) = 2 =n (48)
3π ~2
to get,
3N
D(EF ) = (49)
2EF
(c) The average energy per electron, referred to the Fermi level (chemical potential, really) is,
Z ∞
1
U/N = d ( − µ)D()f () (50)
N 0
where h i−1
f () = e(−µ)/kB T + 1 (51)
For fixed N , we get, Z ∞
∂f ()
Cel = d ( − µ)D() (52)
0 ∂T
For low T , ∂f /∂T is peaked sharply at µ and µ is constant: µ ≈ EF . Then if the density of
states is smooth near EF or if T is small enough, we have,
( − EF )2 e(−EF )/kB T
Z ∞
1 2 2
Cel ≈ D(EF ) d 2 = π D(EF )kB T (53)
kB T 2 3
0 e (−E F )/k B T +1
or,
1 T
Cel = π 2 N kB , kB TF = EF (54)
2 TF
I.10. Heat capacity from electrons in a semimetal. We assume T is very low, and then,
3/2 3/2
Ω 2me Ω 2mh
3/2
# of electrons = 2 (µ − Ee ) = # of holes = 2 (Eh − µ)3/2
3π ~2 3π ~2 (55)
me Ee + mh Eh
=⇒ µ =
me + mh
Also inter-band transition is negligible compared to intra-band excitation, which means the density
of states is given by,
3/2 3/2
Ω 2me Ω 2mh
D() = ( − Ee )1/2 Θ( − Ee ) + (Eh − )3/2 Θ(Eh − ) (56)
2π 2 ~2 2π 2 ~2
And therefore the electronic contribution to the heat capacity of a semimetal at very low T is,
π2 2 2T
1 ΩkB
Cel (T ) = kB T D(µ) = (me + mh )(2µ)1/2 (Eh − Ee )1/2 (57)
3 3 ~3
1 1 1
where µ = mh + me .
13
E
U0
x
0 b a a+b
(a) The potential is sketched in Fig. 7: Let’s define the region 1 inside the barrier, and the region
2 outside the barrier. Define p
2M (U0 − E)
Q=
√ ~ (58)
2mE
K=
~
Then,
ψ1 (a) = eika ψ1 (0)
ψ10 (a) = eika ψ10 (0)
(59)
ψ1 (b) = ψ2 (b)
ψ10 (b) = ψ2 (b)
(b) We write the wave functions as
ψ1 (x) = AeQx + Be−Qx
(60)
ψ2 (x) = CeiKx + De−iKx
Imposing the boundary conditions from item (a), we get a linear system,
e−Qb −e−iKb
eQb −eiKb
QeQb −Qe−Qb −iKeiKb iKe−iKb
=0 (61)
e−ika −e−iKa
ika
e −eiKa
Qeika −Qeika −iKeiKa iKe−iKa
The solution is
Q2 − K 2
cos ka = cos K(a − b) cosh Qb + sin K(a − b) sinh Qb (62)
2KQ
2
(c) We take b → 0 and U0 → ∞, but with U0 b → ma
~
. Clearly, Qb → 0. Performing a Taylor
expansion, we obtain,
Q2 − K 2
cos ka ≈ cos Ka + sin ka Qb
2KQ
Q2 b (63)
≈ cos Ka + sin Ka
2K
W0
= cos Ka + sin Ka
Ka
14
(d) The extended zone scheme is displayed in Fig. 8, and the reduced zone scheme is displayed in
Fig. 9.
15
where w` is the number of atoms of type ` in the unit cell, and τ` is the basis vector. The
structure factor can be understood as a Fourier transform of a bunch of delta functions at
the atomic coordinates of atom ` in the crystal structure. The structure factor is quite useful
because it allows one to write the crystal potential due to all atomic centers in reciprocal space
as a simple produce of the Coulomb potential in reciprocal and the structure factor,
X
V (q + G) = S` (G)V` (G), (65)
`
where V` (G) is the potential of a single atom of type ` in reciprocal space. The advantage of
expressing the total potential due to the atoms in this form is that it explicitly separates all
quantities that depend on the geometry of the system, encoded in the structure factor, from
the quantities that depend of the charge and pseudopotential used for a given type of atom,
independent of its position on the Bravais lattice.
(b) Si has a diamond (FCC) structure, with a basis located at a8 (1, 1, 1) and− a8 (1, 1, 1). So, the
a
structure factor is S(G) = cos G · 8 (1, 1, 1) = cos π4 (G1 + G2 + G3 ) , where a reciprocal
P
lattice vector G are written in the basis of the reciprocal lattice vectors, i.e., G = i Gi bi ,
where
2π
b1 = (ŷ + ẑ − x̂)
a
2π
b2 = (ẑ + x̂ − ŷ) (66)
a
2π
b3 = (x̂ + ŷ − ẑ)
a
Also, note that
In order for the pseudopotential V (G) for a given G-vector to be important, we need to have a
non-zero structure factor, i.e., cos( π4 (G1 +G2 +G3 )) 6= 0. We can search for other combinations
of G1 , G2 and G3 that yield a non-zero structure factor and compute the corresponding
structure factor. At the end of the day, we can compile a table such as Table 1.
By inspection, we see that G2 /(2π/a)2 = 3, 8, 11, 16, 19, 24, 27, 32, 35, · · · yield a non-zero
crystal pseudopotential.
(a) Because the atomic orbits are spherical symmetric, only s orbital contributes to the interaction.
Assume the hopping integrals:
Z
dV ψ ∗ (r)Hψ(r) = −α (68)
Z
dV ψ ∗ (r − ρ)Hψ(r) = −τ (69)
where ρ is the vector pointing to the nearest neighbor. So the energy for a state with wavevector
k is:
e−ik·ρm
X
hk|H|ki = −α − τ (70)
m
16
G1 G2 G3 G2 / (2π/a)2 G1 + G2 + G3 S(G) = 0? Relevant?
000 0 0 No Yes
100 3 1 No Yes
111 3 3 No Yes
110 4 2 Yes No
211 8 4 No Yes
210 11 3 No Yes
221 11 5 No Yes
200 12 2 Yes No
222 12 6 Yes No
220 16 4 No Yes
322 19 7 No Yes
311 19 5 No Yes
321 20 6 Yes No
332 24 8 No Yes
310 24 4 No Yes
320 27 5 No Yes
300 27 3 No Yes
331 27 7 No Yes
333 27 9 No Yes
422 32 8 No Yes
432 35 9 No Yes
421 35 7 No Yes
330 36 6 Yes No
433 36 10 Yes No
411 36 6 Yes No
17
for simple cubic lattice
Ek = −α − 4τ (cos(kx a/2) cos(ky a/2) + cos(kx a/2) cos(kz a/2) + cos(ky a/2) cos(kz a/2)) (73)
.
(c) See Fig. 10.
-1
-2
-3
-3 -2 -1 0 1 2 3
18
Figure 11: Orbitals on CuO plane
(a) There are three atoms/cell and three basis sets. We will use the following convention for the
unit cell and basis set associated with each atom, including the direction of the orbitals: We
assign an index to each orbital as in Fig. 11:
i. O1 px
ii. O2 py
iii. Cu dx2 −y2
Since we are only including nearest-neighbor hopping, we have that H11 = H22 ≡ O , and
H33 = Cu . The hopping matrix elements between different oxygen atoms are zero, so H12 =
H21 = 0. If we denote by Vdp the (negative) matrix element between a d and a p orbital aligned
such that the positive lobes of the orbitals overlap, we obtain:
H13 = Vdp −1 + eikx a
2
kx a
2
|H13 | = 2Vdp sin
2
(77)
H23 = Vdp 1 − eiky a
2
ky a
2
|H23 | = 2Vdp sin
2
1 = O
O + Cu 1
q
2,3 = ± (Cu − O )2 + 4λ2 (79)
2 2
2 kx a 2 ky a
2 2
λ ≡ 4Vdp sin + sin .
2 2
There are three bands and we can fit 2 electrons per band, so band theory predicts a metal.
(b) We plot the band structure in Fig. 12, assuming that the energy is set to zero at O and that
Cu = 3 eV and Vdp = −1 eV. Note the half occupied band at the Fermi energy (at 0).
(p̂ + ~k)2
uk (x) + V (x)uk (x) = Euk (x) (80)
2m
where p̂ = −i~∇.
I.17. Dirac equation and spin-orbit interaction. The spin-orbit contribution to the one-
electron Hamiltonian starting with the Dirac equation is derived by Schiff (pp. 332-3) and by
Bjorken-Drell (Rel. Q. M. pp48). The Dirac equation:
! !
0 σi 1 0
α · p + βm + V = H, αi = , β= (81)
σi 0 0 −1
!
ψ
Write Ψ = , HΨ = E 0 Ψ, E 0 = E + m. Assume E and V are small compared to m.
φ
σ · pφ = (E 0 − m − V )ψ = (E − V )ψ
(82)
σ · pψ = (E 0 + m − V )φ = (E + 2m − V )φ
20
which leads to,
φ = (2m + E − V )−1 σ · pψ (83)
and therefore,
−1
1 E−V
(E − V )ψ = σ · p φ = σ·p 1+ σ·p ψ (84)
2m 2m
−1
E−V E−V
Approximately, 1 + 2m ≈ 1− 2m . Note that (σ · A)(σ · B) = A · B + iσ · (A × B) and
pV = (pV ) − i∇V ,
" #
E−V p2 i 1
Eψ = 1− +V ψ− 2
∇V · pψ + σ · (∇V × p)ψ
2m 2m 4m 4m2
" # (85)
p2 p4 i~ ~
= − 2 2
+V − 2 2
∇V · p + σ · (∇V × p) ψ
2m 8m c 4m c 4m2 c2
And the last term on the right side of Eqn. (85) is the spin-orbit Hamiltonian,
~
HSO = σ · (∇V × p) (86)
4m2 c2
p2
h i
In the following discussion, we assume the symmetry properties hold already for the terms 2m + V (r) .
Under translational invariance: x → x + n1 a1 + n2 a2 + n3 a3 , we have
21
Define
Z Z Z
0 0
A= dr ψk∗ (r)ψk0 (r) = dr e−ik·r u∗k (r)eik ·r uk0 (r) = dre−i(k−k )·r u∗k (r)uk0 (r) (91)
So, two Bloch functions are orthogonal unless (k − k0 ) is a reciprocal lattice vector.
1 ∂ε(k)
≡ (95)
m∗ µν ∂kµ ∂kν
The expression for mm∗ µν is referred to as the f -sum rule. Assume that ε(k0 ) is s-fold degenerate
and V (−x) = V (x). Since the crystal has inversion symmetry, the states are eigenfunctions of
parity. For degenerate states, they may be chosen such that they are also eigenstates of parity.
These eigenstates of the parity are the total wavefunctions ψnk = eik·r un (k), not the states un (k)
themselves,
∂ε(k0 )
∼ hu(k)|(p̂ + k)|u(k)i · δkµ = hψk |p̂|ψk i · δkµ = 0 (96)
∂kµ
since p̂ has odd parity and ψk∗ ψk has even parity. (p̂ could also include spin-orbit terms, which
also have odd parity.) For no first order effects for a crystal with inversion symmetry.
22
The second order shifts are eigenvalues of,
0 H0
X Hµl lν
Mµν = = Aµν
αβ kα kβ , µ, ν = 1, 2, ..., s (97)
l
εµ − ε l
If s = 2,
s
M11 + M22 (M11 − M22 )2
q
E (2) = + 2 =B k k ±
+ M12 αβ α β Cαβγδ kα kβ kγ kδ (98)
2 4
except for special values of Cαβγδ , E (2) cannot be expanded about k = 0. Thus an effective mass
cannot be defined at k = 0, although for other values of k it is possible. The calculations are
tedious and need not have been performed in detail.
I.21. Delta function potential. Assume the delta function potential V (x) = −αδ(x), The
Schrödinger equation:
~2 d2 ψ
− − αδ(x)ψ = Eψ (99)
2m dx2
The wavefunction is assumed to be,
(
Beκx , x≤0
ψ(x) = (100)
Be−κx , x≥0
√
−2mE
where κ = ~ . By solving the equation, we have
mα
κ= (101)
~2
mα2
E=− 2 (102)
2~
For bound state, the wavefunction must decay to zero at infinity, so κ must be real. As a result,
the energy E must be negative to have the bound state.
For two one-dimensional delta function separated by a.
The solution must either have odd parity or even parity. For even parity solution:
κx x ≤ −a/2
Ae ,
ψ(x) = Ae−κx , x ≥ a/2 (104)
B(eκx + e−κx ), −a/2 < x < a/2
Solving the Schrödinger equation by applying boundary conditions at the discontinuity, we can
get the equations:
κb = 1 + e−κa (105)
~2
where b = mα , there will always be one and only one solution.
23
For the odd parity solution, the wavefunction is assumed to be
Aeκx , x ≤ −a/2
ψ(x) = −Ae −κx , x ≥ a/2 (106)
B(eκx − e−κx ), −a/2 < x < a/2
Solving the Schrödinger equation by applying boundary conditions at the discontinuity, we can
get the equations:
κb = 1 − e−κa (107)
The condition to have a solution is b < a, which is ~2 < mαa. As a result, there will be only one
bound state if ~2 > mαa, but two bound states if ~2 < mαa.
By solving the Schrödinger equation for triple delta function potentials with distance a from each
other, we can get the equation for even parity:
1 + κb
e2κa = (108)
(κb − 1)2
Solving the equation, When 2a < 3b, there is only one solution, and when 2a > 3b, there are two
solutions. For odd parity:
1
e2κa = (109)
(1 − κb)
When 2a < b, there is no solution, and when 2a > b, there is one solution. So there is one solution
when 2a < b, two solutions when b < 2a < 3b, three solutions when 2a > 3b.
When it goes to many delta functions, the bound states form a band.
I.22. High symmetry points. The electronic energies in the free electron gas model are given
by,
~2
E(k) = (k + G)2 (110)
2m
where k is in the first Brillouin Zone (BZ), and G is any reciprocal lattice vector. Clearly, there
will be more than one branch of E(k), with different branches corresponding to different G’s.
Depending on the crystal symmetry of the points of interest in the BZ, each branch may be
multiply degenerate (i.e., several G’s might map different parts of the extended E(k) back to the
same branch in the BZ.)
24
(c) fcc: Γ(000) to X( 2π 2π
a (001)). Here k = ξ a (001) where 0 ≤ ξ < 1, and,
2π
G = n1 b 1 + n2 b 2 + n3 b 3 = (−n1 + n2 + n3 , n1 − n2 + n3 , n1 + n2 − n3 ) (115)
a
2~2 π 2 2
Ek = [ξ + 2ξ(n1 + n2 − n3 ) + (−n1 + n2 + n3 )2 + (n1 − n2 + n3 )2 + (n1 + n2 − (116)
n3 )2 ]
ma2
(d) hcp: Γ(000) to A( πc (001)). Here we assume an ideal hcp structure, so c/a =
p
8/3. So,
q
k = ξ πc (001) = ξ πa 3
8 (001) where 0 ≤ ξ < 1.
I.23. Deep core level. The width of the core-level band is proportional to the overlap integral,
Z
W = dr φ(r)[V (r) − Va (r − a)]φ(r − a) (119)
− r 2
e ac aB
where φ(r) = (πa3c )1/2
, ac = Z = 0.09 × 10−8 cm. Also V (r) ∼ potential in solid, Va (r) = − Zer
for r → 0.
Approximate V (r) − Va (r − a) ≈ Vc (r) near r = 0; and V (r) − Va (r − a) near r = a. These are
the only two places the wavefunction have appreciable overlap.
!
Ze2 e2
Z
W ≈ drφ(r) − φ(r − a) = Z 2 e−a/ac ≈ Z 2 e−60 Ryd (120)
r aB
I.24. Buckyballs.
(a) In fcc C60 , clusters are condensed by rather weak van der Waals force. As discussed in [Phys.
Rev. Lett. 66, 2637(1991)] and [Phys. Rev. B 46, 1754(1992)] , fcc C60 is a semiconductor
with direct energy gap at the Brillouin-zone boundary (X point). In a primitive cell, there is
one C60 molecule, and each carbon atom has 4 valence electrons. In this way, there will be
4 × 60 = 240 valence electrons, and 120 filled electron bands. The band structure is plotted in
Fig. 13
(b) 0 partially filled bands.
(c) The alkali metals can form interstitial alloys, such as K3 C60 , with solid C60 . The crystal
structure is shown in Fig. 14, In one primitive cell, there are 3 K atoms and 1 C60 molecule,
so there will be 3 ∗ 1 + 60 ∗ 4 = 243 valence electrons. According to [Phys. Rev. B 46,
1766(1992)], there is little difference between the band structure of fcc C60 and that of fcc
K3 C60 (Fig. 15), except for the position of Fermi level. In this way, the effect of 3 potassium
atoms is to increase the Fermi level so as to fit in 3 more valence electrons. There will be 120
filled bands.
25
Figure 13: The band structure of fcc C60 . The top of the valence band was chosen as the zero of
the energy. [Phys. Rev. B 46, 1754(1992)]
26
Figure 15: The band structure of fcc K3 C60 . [Phys. Rev. B 46,1766(1992)]
(a) The states in the frequency interval (ω, ω + δω) are those contained in the region between the
surfaces ω(q) = ω and ω(q) = ω + δω. The distance between the surfaces, ∆l, is given by:
|∇q ω| · ∆l = δω. So,
dS
Z Z
{Volume in q-space} = dS · ∆l = δω (121)
S S |∇q ω|
and
Ω
D(ω)dω = {volume in q-space} (122)
(2π)3
So, in three dimensions,
Ω dS
Z
D(ω) = (123)
(2π)3 S |∇q ω|
(b) D(ω) near critical point can be expanded as,
ω ≈ ωc + αξ 2 + βη 2 (124)
where ξ = q1 − q1c and η = q2 − q2c . Also note that |∇q ω| = 2(α2 ξ 2 + β 2 η 2 )1/2 , and dl =
1 2 2 2 2 1/2 dξ = 1 (α2 ξ 2 + β 2 η 2 )1/2 dη.
|βη| (α ξ + β η ) |αξ|
(i) Minimum (α, β > 0): for ω < ωm , there is no contour, D(ω) = 0; for ω > ωm ,
Z ω−ωm √
A d( α) A 1
D(ω) = 2 √ = √ (125)
(2π)2 0 αβ(ω − ωm − αξ 2 )1/2 4π αβ
27
A √1 ; for ω > ωm , D(ω) = 0.
(ii) Maximum (α, β < 0): for ω < ωm , D(ω) = 4π αβ
Hence, D(ω) ∼ Θ(ωm − ω) + analytic terms.
(iii) Saddle point (α > 0, β < 0): contours are hyperbolas, both above and below ωs . Intro-
duce a cutoff R: αξ 2 , |β|η 2 < R2 . For ω < ωs ,
Z R √
A d( αξ) A 1 h p i
D(ω) = 2 √ = √ ln(R + R 2 + ω − ω) − ln(ω − ω)
s s
(2π)2 0 −αβ(ωs − ω + αξ 2 )1/2 2π 2 −αβ
A ln(ωs − ω)
=− 2 √ + const. for |ω − ωs | R
2π −αβ
(126)
For ω > ωs :
Z R √
A d( −βη) A 1
D(ω) = 2 √ 2 1/2
= 2√ ln(ω − ωs ) (127)
2π 0 −αβ(ω − ωs − βη ) 2π −αβ
Hence, D(ω) ∼ ln |ω − ωs | + analytic terms.
(b) These two waves differ as they are propagating not due to collision but due to quantum effect
caused by the Pauli exclusion principle. These waves would propagate even for uncharged
particles which do not react as slowly as electrons.
(c)
!1/6
~ π4
vs = n1/3 = 2.07n1/3 for electron (134)
m 3
and,
vs = 3.75 × 10−4 n1/3 for He3
= 1.13 × 10−3 n1/3 for neutron
−2 1/3
= 1.00 × 10 n for muons (135)
= 2.07n1/3 for positrons
= 1.13 × 10−3 n1/3 for protons
(d) kF is same for all particles because it is independent of mass of the particle.
28
I.27. Vibrational modes in graphene.
(a) Because there are 2 atoms in a unit cell, and there are 2 degrees of freedom in the plane which
corresponds to one longitudinal and one transverse mode, there are 2 acoustic branches and 2
optical branches.
(b) In Debye model, ω = vk, where v is the velocity of sound.
A 1 Aω
D(ω) = k = (137)
2π dω/dk 2πv 2
The internal energy (the factor 2 is included because of there are 2 branches,
ω2
Z ωD
A~
Z
U =2 dωD(ω) < n(ω) > ~ω = dω (138)
πv 2 0 e~ω/kB T − 1
The capacity,
dU A kB3 T2 Z xD
x3
Cv = = dx (139)
dT πv 2 ~2 0 (ex − 1)2
(c) The minimum number of singularities of type n is CnN , where N = 2 and n = 0, 1, 2. So, we
have
(i) n = 0, C02 = 1. There is 1 Van Hove singularity of type 0 (minimum)
(ii) n = 1, C12 = 2. There are 2 Van Hove singularities of type 1 (saddle points)
(iii) n = 2, C22 = 1. There is 1 Van Hove singularity of type 0 (maximum)
So, there is a minimum of 4 Van Hove singularities in total.
To sketch the density of states around an M0 singularity, lets assume that we have an optical
branch with dispersion ω(q) ∼ ω0 + αq 2 , for small q. We now that the density of states is zero
for ω < ω0 . For ω > ω0 , we evaluate,
A 1 2πqA 1
Z
D(ω) = dq = ∼ const (140)
(2π)2 |∇q ω| (2π)2 2αq
So, the M0 Van Hove singularity in this case looks like a step function. A M0 Van Hove
singularity in the DOS for an optical branch of this system is sketched in Fig. 16.
(a) Near absolute zero temperature, the energy of each phonon mode is Eλq = 12 ~ωλq . We use
the virial theorem
D E and write the average potential energy for each phonon mode as D hPEi E=
1 2 2 11 2
2 mωλq |ξλq | = 2 2 ~ωλq . So, the mean-square displacement of the phonon mode is |ξλq | =
~
2mωλq .
The displacement in real space per atom within the Debye approximation is,
Z ωD
D E 1 X 3 ~ Dλ (ω) 9 ~
R2 = ξλq = dω = (141)
N λq N 2m 0 ω 4 mωD
29
Figure 16: Sketch of a M0 Van Hove singularity.
30
(d) For Aluminum, we have: TD = 428 K, vs = 3100 m/s, ρ = 2400 kg/m3 , so ~ωD ∼ 37 meV. If
we evaluate the entropy at T = 300 K (xD ∼ 1.43), we obtain
Al
RRMS ∼ 0.1 Å
Al (148)
EZP /N ∼ 42 meV
S/N ∼ 0.1 meV/K
(a) Let’s label the atoms with masses M and m with indices 1 and 2, respectively. We want to
compute the dynamical matrix Djj 0 (Rl ), where j is an atom index. The force F`,j acing on
each atom j is
F`,1 = α (R`,2 − R`,1 ) − β (R`,1 − R`−1,2 )
(149)
F`,2 = β (R`+1,1 − R`,2 ) − α (R`,2 − R`,1 )
The dynamical matrix reads,
α+β
D11 = δ``0
M
1
D12 =√ (−αδ``0 − βδ`+1,` )
Mm
(150)
1
D21 =√ (−αδ``0 − βδ`+1,` )
Mm
α+β
D22 = δ``0
m
With Fourier transform, we have,
To be specific,
α+β
D11 (q) =
M
1 iqa/2
D12 (q) = − √ αe + βe−iqa/2
Mm (152)
∗
D21 (q) = D12 (q)
α+β
D22 =
m
We diagonalize the dynamical matrix as,
D − ω 2 I = 0 (153)
4αβ 3qb
Define µ−1 = M −1 + m−1 and C = Mm sin2 2 . Then,
s
2
1 α+β α+β
ω2 = ± − 4C (154)
2 µ µ
31
ω ω
0 q 0 D(ω)
π/a
We now change the dummy indices such that ` − `0 = L and perform the sum over L and `.
The sum over ` yields a constraint that q0 = −q and a factor of N . We obtain
1 X −iq·τj iq·(τj 0 −RL )
U= e e Djj 0 (L)ξqj ξ−qj 0
2 Lqjj 0
(157)
1 X †
= Djj 0 (q)ξqj ξqj 0
2 Lqjj 0
†
where we used the fact that ξq = ξ−q .
We now define, s
Mj X −iq·(R` −τj )
p`j = e pqj (158)
N q
32
We get rid of the sum over j 0 by diagonalizing the dynamical matrix. Let’s denote the eigen-
values and eigenvectors of D(q) for each q by ω and C, respectively. Then, we can transform
from the atomic basis j to an eigenvector basis λ and define,
X
ξqλ = Cjλ (q)ξqj
j
X (161)
pqλ = Cjλ (q)pqj
j
Then,
1 Xh † †
i
H= pqλ pqλ + ω 2 ξqλ ξqλ (162)
2 qλ
as desired.
N (ω)(~ω)2 e~ω/kB T
Z ωD Z ωD
∂ N (ω)~ω
Cv = dω = dω (165)
∂T 0 e ~ω/kBT − 1 0 (e~ω/kB T − 1)2 kB T 2
dk L
Take Debye dispersion ω = vs k, and 1D DOS is given by N (k) = L/π, N (ω) = N (k) dω = πvs .
R ωD L
Note that, 0 N (ω)dω = N ⇒ N = πvs ~ kB TD ,
Let x = ~ω/kB T , we have
2T x2 e x
Z xD
L kB
Cv dx
πvs ~ 0 (ex − 1)2
x2 e x x2 e x
Z xD Z ∞ Z ∞
2x
= N kB (T /TD ) x 2
dx → N kB (T /TD ) x 2
dx = N kB (T /TD ) x
dx
0 (e − 1) 0 (e − 1) 0 e −1
(166)
in the limit of T → 0.
(a) There are 3 atoms/cell and we are assuming this is a 2D universe. So, there are 6 phonon
modes. The number of acoustic modes is the number of dimensions, and one mode will always
be longitudinal. Since this is a 2D problem, there is only one transverse mode. So, we have 2
acoustic modes (1 LA + 1 TA), and 4 optical modes (2 LO + 2 TO).
(b) For each acoustic phonon branch λ, we require that
Z ωD
N= Dλ (ω) (167)
0
33
Following the same procedure done in the book for 3D systems, we obtain, for a 2D system,
A 2πq A ω
Dλ (ω) = 2
= , (168)
(2π) vs 2π vs2
where ωD is the Debye frequency, i.e., the maximum frequency a phonon can have, vs is the
sound velocity, and A = N a2 . Solving the integral for ωD , we obtain
2vs √
ωD = π. (169)
a
(c) We want the T-dependence of CvLA (T ). From the previous item, we know the expression for
DLA (ω). The total energy of the system due to phonons is simply
N a2 ~ ω2
Z ωD Z ωD
U= dω D(ω)n(ω)~ω = dω . (170)
0 2πvs2 0 e~ω/kB T − 1
~ω ~ωD
Let x = kB T and xD = kB T . We can write
3 Z xD 3
N a2 ~ kB T x2 . N a2 ~ kB T
U= dx x = I(T ). (171)
2πvs2 ~ 0 e −1 2πvs2 ~
The low-T limit of the integral is I(T → 0) = 2ζ(3) + O (e−xD ), where ζ is the Riemann zeta
function. So,
3N a2 ζ(3) 3 2
LA ∂U
Cv (T → 0) = = k T ∼ T2 (172)
∂T T →0 πvs2 ~2 B
for each TA phonon branch.
I.32. Zero-point energy correction. The general expression for the zero-point energy per atom
is,
1 X1 1 XX 1
EZP M = ~ωλ (q) = ~ωλ (q) (173)
N λ 2 N λ BZ 2
(a) For a three dimensional Debye model (ωλ (q) = cs q) with one atom per cell and all acoustic
modes degenerate, we get,
3 X1 3~cs V
Z
EZP M = ~cs q → qd3 q
N BZ 2 2N (2π)3 D.S.
Z qD (174)
3~cs V 3 3 V 3
= 4π q dq = q ~ωD
2N (2π)3 0 16π 2 N D
3 = 6π 2 N , so that,
From the definition of qD we get qD V
9
EZP M = ~ωD (175)
8
(b) For a three-dimensional Einstein model (ωλ (q) = ωE ) with one atom per cell and all three
modes degenerate, we get,
3 X 3
EZP M = ~ωE 1 = ~ωE (176)
2N BZ
2
34
(a) Let γ = mω 2 ,
( )
p2 γ p2 mω eF 2 e2 F 2
H= + x2 + eF x = ~ω + (x + ) −
2m 2 2~mω 2~ mω 2 2mω 2
r r 2
p mω eF p mω eF 1 eF
= ~ω √ +i (x + ) √ −i (x + ) + − √ /ω
2~mω 2~ mω 2 2~mω 2~ mω 2 2 2mω
1 λ2
= ~ω(A† A + ) −
2 ω
(177)
(b)
1 λ h i
A = a − i√ =⇒ A, A† = [a, a† ] = 1 (178)
~ω
(c)
1 p
ẋ = [x, H] =
i~ m (179)
1
ṗ = [p, H] = −(γx + eF )
i~
define x0 = x + eF
γ ,
γ 0 p(0)
ẍ0 = − x =⇒ x0 (t) = x0 (0) cos ωt + sin ωt (180)
m mω
or,
eF p(0) eF
x(t) = (x(0) + ) cos ωt + sin ωt − (181)
γ mω γ
The effect of eF x term is simply to displace the oscillator to a new equilibrium point: − eF
γ .
35
Because hn|x|ni = 0
C 1
xn = 0
(hn|x|n0 ihn0 |x3 |ni + hn|x3 |n0 ihn0 |x|ni) (188)
~ω n − n
√ √
Usinghn|a|n0 i = n0 δn,n0 −1 , hn|a† |n0 i = n0 + 1δn,n0 +1
(1/2) √ √
1 2~
0
hn|x|n i = n0 δn,n0 −1 + n0 + 1δn,n0 +1 (189)
2 mω
√
hn|a† a† a|n0 i = n0 n0 + 1δn,n0 +1 (190)
√
hn|a† aa|n0 i = (n0 − 1) n0 δn,n0 −1 (191)
This gives
(3/2) h
3 0 3
2~ √ √ i
hn|x |n i = n nδn,n0 +1 + (n + 1) n + 1δn,n0 −1 (192)
8 mω
2C 1 2~ (1/2) 3
(3/2) h
2~
i
xn = (n + 1)2 − n2 (193)
~ω 2 mω 8 mω
3 ~
= C 2 3 (2n + 1) (194)
2 m ω
As a result, the mean position of the particle depends on the number of phonons excited. That is,
the mean position depends on the temperature. This is thermal expansion.
Sec. II
II.1. Wannier functions.
Use planewaves as an approximation for the band states, and use a cubic cell with dimension
a,
1
ψnk (r) = √ eik·r (196)
V
and therefore,
1
e−ik·Ri eik·r
X
Wn (r − Rl ) = √
N V k∈BZ
s s
V V 1 sin( πa (x − Rlx )) sin( πa (y − Rly )) sin( πa (z − Rlz ))
Z
= f racdk(2π)3 eik·(r−Rl ) =
N N π3 x − Rlx y − Rly z − Rlz
(197)
36
(b)
1 X ik1 ·Ri ∗ 1 X −ik2 ·Rj
Z
hWn (Ri )|Ĥ|Wm (Rj )i = dr √ e ψnk1 (r)Ĥ √ e ψmk2 (r)
N k1 N k2
1 X ik1 ·Ri −ik2 ·Rj 1 X ik1 ·Ri −ik2 ·Rj
= e e hψnk1 |Ĥ|ψmk2 i = e e Enk1 δmn δk1 k2 (198)
Nkk Nkk
1 2 1 2
1 X ik1 ·(Ri −Rj )
= δmn e Enk1
N k
1
(c) If we allow mixing of different bands, the Wannier function with band index n at cell R is
defined by, s " #
V
Z
dke−ik·R
X
(k)
wnR (x) = Umn ψmk (x) (199)
(2π)3 BZ m
Note that the unit of Bloch wavefunction is [ψ(x)] = L−3/2 , U is dimensionless, and therefore
we also require that w(x) be of the same unit with ψ(x).
After we get the Wannier functions, we can calculate the hopping parameters between these
MLWFs, which are exactly the Hamiltonian matrix elements between these Wannier functions.
"Z # "Z #
V
Z
(k ) (k )
dk1 e−ik1 ·R1 dk2 e−ik2 ·R2
X X
hwmR1 |Ĥ|wnR2 i = dx Um01m ψm0 k1 (x) Ĥ(x) Un0 n2 ψn0 k2 (x)
(2π)3 m0 n0
V XZ Z
(k )∗ (k )
= dk1 dk2 eik1 ·R1 e−ik2 ·R2 Um01m Un0 n2 ∗
dxψm 0 k (x)Ĥ(x)ψn0 k2 (x)
(2π)3 m0 n0
1
V X
Z
(k )∗ (k )
= 3
dk1 dk2 eik1 ·R1 e−ik2 ·R2 Um01m Un0 n2 δk1 k2 δm0 n0 Em0 k1
(2π) m0 n0
V
Z
(k )∗ (k )
dk1 eik1 ·(R1 −R2 )
X
= Um01m Um01n Em0 k1
(2π)3 m0
V
Z
(k )† (k )
dk1 eik1 ·(R1 −R2 )
X
= Em0 k1 Umm1 0 Um01n
(2π)3 m0
= hwm(R1 −R2 )|Ĥ|wn0 i
(200)
which depends only upon (R1 − R2 ).
2~2 1
m∗ = (201)
a(E2 − E1 ) cos(ka)
37
(iii) Bloch oscillation,
eE
k(t) = k(0) + t (202)
~
a(E2 − E1 )
v(k) = sin(ka) (203)
Z 2~
x(t) = v(k(t))dt (204)
(E2 − E1 ) aeE
= cos( t) − 1 + x(0) (205)
2eE ~
2π~ (E2 −E1 )
So the period is T = aeE , range of distance is eE .
(b) In Wannier representation,
" #
∂ ∂ ∂
En (−i , −i ) + V (Rx , Ry ) ψ(Rx , Ry , t) = i~ ψ(Rx , Ry , t) (206)
∂Rx ∂Ry ∂t
In Bloch representation,
" #
∂ ∂ ∂
En (kx , ky ) + V (i ,i ) ψ(kx , ky , t) = i~ ψ(kx , ky , t) (207)
∂kx ∂ky ∂t
38
2 2
(b) v = − E0~a δk, m∗ = − E~0 a2 .
2
(c) v = − E~0 a , m∗ = − E0~δka3 .
~2
(d) v = − E~0 a , m∗ = E0 δka3
.
(a)
a a a
(k) = −2τ cos (kx + ky ) + cos (kx − ky ) + cos (kx + kz ) (214)
2 2 2
a a a
+ cos (kx − kz ) + cos (ky + kz ) + cos (ky − kz ) (215)
2 2 2
(b) (i)
1 ∂
vx = (216)
~ ∂kx
aτ aet aet
= sin (x + y ) + sin (x − y ) (217)
~ 2~ 2~
aet aet
+ sin (x + z ) + sin (x − z ) (218)
2~ 2~
so
2τ ae(x + y )t
Z
x(t) = v(t)dt = cos − 1 + other similar terms (219)
e(x + y ) 2~
"
2τ 1 1
= (cos A(x + y )t − 1) + (cos A(x − y )t − 1) (220)
e x + y x − y
1 1
+ (cos A(x + z )t − 1) + (cos A(x − z )t − 1) (221)
x + z x − z
(ii)
2τ ae
x(t) = y(t) = z(t) = (cos − 1) (222)
e ~
1 2 3 4 5 6
-0.5
-1.0
-1.5
-2.0
39
(iii) For the estimation, let’s take τ ∼
= 5 eV, and not worry about counting nearest neighbours.
The amplitude,
√
4τ 4 3 × 5eV
= (223)
e 1eV/mm
(a)
e~
H= σ · B(t) (224)
2mc !
e~|B(t)| cos(θ) e−iφ sin(θ)
= (225)
2mc eiφ sin(θ) − cos(θ)
e~|B(t)|
We use a trial eigenfunction ( in unit of 2mc )
! !
cos(θ/2) cos(θ) cos(θ/2) + sin(θ) sin(θ/2)
H iφ = iφ (226)
e sin(θ/2) e (sin(θ) cos(θ/2) − cos(θ) sin(θ/2))
!
cos(θ/2)
= (227)
eiφ sin(θ/2)
(b)
! !
1 − 21 sin(θ/2) 1 0
∇χ↑ (θ, φ) = 1 iφ θ̂ + φ̂ (229)
|B| 2 e cos(θ/2) |B| sin(θ) ieiφ sin(θ/2)
− sin2 (θ/2)
A(θ, φ) = φ̂ (230)
|B| sin(θ)
and therefore,
1
Ω(θ, φ) = ∇ × A(θ, φ) = B̂ (231)
2|B|2
(c)
1 |B|2
Z Z
γ↑ = dS · Ω = − dω (232)
2 |B|2
1
=− ∆ (233)
2
II.7. Hartree-Fock approximation for ferromagnetic electron gas.
40
(a) When we go from a paramagnetic to a ferromagnetic ground state, the rs parameter doesn’t
change, but the spin degeneracy factor gs changes from 2 to 1. Alternatively, we can write the
ground-state energy E solely in terms of the Fermi wavevector without any reference to gs ,
i.e., E = E(rs , gs ) = Ek (kF ). Now, all we have to do is to relate the Fermi wave vector from
paramagnetic case to the ferromagnetic case. We know that,
1/3 1/3
3 1
rs = , n = N/Ωxtal (234)
4π n
Also,
kF
X gs Ωxtal 3
N = gs (1) = k
k
6π 2 F
(235)
1/3
1 1
kF =
gs rs
The relationship between the paramagnetic and ferromagnetic Fermi wavevector is kFF =
21/3 kFP . So, the total energy is given by
E P = Ek (kFP ) = E P (rs )
(236)
E F = Ek (kFF ) = Ek (21/3 kFP ) = E F (rs )
We can rewrite the the ferromagnetic ground-state energy in terms of the paramagnetic one,
E F (rs ) = Ek (21/3 kFF ) = E P (2−1/3 rs ) (237)
0.02
rs
5 10 15 20
-0.02
Paramag.
Ferromag.
-0.04
-0.06
-0.08
-0.10
Note that the ferromagnetic state is lower in energy than the paramagnetic case for rs > 5.56.
This makes sense, as ferromagnetism can be though as a strongly-corollated phenomenum, so
it should take place as larger values of rs .
41
II.8. Exchange-correlation hole
2 2
(a) (i) There are 2 atoms per cell, so n = a3
. In Rydberg units, a = 7.94 a0 , so n = 7.943
, and
1/3
3 7.94
rs = 4π 21/3
= 3.91.
(ii) We i the exchange-correlation hole from the correlation function g↑↑ (r, 0) =
h can estimate
1 2 3
2 1 − (f (r)) , where f (r) = rkF j1 (rkF ). The first zero of the spherical Bessel function
j1 is at rkF = 4.5 = r 1.92
rs . So, the size of the xc hole is r ∼ 1.97 a0 .
(b) (i) The energy is roughly the Coulomb potential between the xc hole and the electron, so
it’s
e2 2
xc = − 2 = − 2 , in Ry (239)
z z
(ii) If we use the correct exchange-correlation hole, the charge density associated to the hole
would be farther away, so xc would be larger (i.e., less negative).
(a) In a Wigner sphere (radius R = a0 rs ), ion charge is evenly distributed, and we can get the
electric field by using Gauss’s law,
Qr2
Z ∞ 1 3Q
− 3 , r<R
ϕ(r) = E · dl = 4π0 2R 2R
(240)
r Q , r>R
4π0 r
4π0 ~2
with R = a0 rs and a0 = me2
. In atomic units, we get,
3 r2
V (r) = − + 3 (242)
rs rs
42
(c) Total energy: E = − 1.8
rs +
3
3/2 . At equilibrium,
2rs
∂E 25
= 0 ⇒ rs = (245)
∂rs 16
4π 3
Then the volume per electron is Ve = 3 rs = 15.98 in units of a30 .
The bulk modulus is then given by,
∂ 2 E
B = −V = 4.005 × 10−3 (246)
∂V 2 r
25
s = 16
so
Z ∞
(t) = dteiωt (t) (250)
0
eiωt decays exponentially as a function of t when the imaginary part of ω is greater than zero. By
Cauchy’s theorem:
dω 0 (ω 0 )
I
(ω) = (251)
2πi ω 0 − ω − iη
assuming η is very small. In order for the Cauchy’s theorem to be valid, the function (ω) must
drop off faster than ω 0 for large ω 0 . Therefore,
dω 0 (ω 0 ) − ∞
I
(ω) − ∞ = (252)
2πi ω 0 − ω − iη
−∞ is the value of as ω goes to infinity. Now we can separate the equation into real and imagi-
nary parts.
We notice that when η goes to zero, the pole will move to the real axis, so we need to change
our contour to include a half-circuit around the pole at ω = ω 0 which passes below the pole. The
half-circuit part in the contour contributes half to the contour integral as there is only one pole
inside. If we want to discard the contribution from the half-circuit to make it a Cauchy principle
value, we must include a factor of 2 in the RHS to keep the equity .
dω 0 (ω 0 ) − ∞
I
∞
(ω) − =P (253)
πi ω 0 − ω
43
Therefore, using 1 and 2 for the real and imaginary part of we have
2ω 0 dω 0 2 (ω 0 )
I
∞
1 (ω) − =P (254)
π ω 02 − ω 2
2ω dω 0 1 (ω 0 ) − ∞
0
I
2 (ω) = −P (255)
π ω 02 − ω 2
II.11. Sum rules of dielectric function.
(a) In Drude model,
ωp2 τ 2
1 (ω) = 1 −
1 + ω2τ 2
(256)
ωp2 τ
2 (ω) =
ω(1 + ω 2 τ 2 )
ωp2 ωp2 τ −1
For τ → ∞, 1 (ω) → 1 − ω2
and 2 (ω) = ω τ −2 +ω 2 .
Use the identity,
ε
lim = πδ(x) (257)
ε→0 x2 + ε2
and we will have,
ωp2
2 (ω) = π δ(ω), as τ −1 → ∞ (258)
ω
and Z ∞ Z ∞
πωp2
ω2 (ω)dω = πωp2 dωδ(ω) = (259)
0 0 2
(b) Take inverse of dielectric function,
1 1 − i2 1 2
−1 (ω) = = 2 ⇒ Im[ ] = − 2 (260)
1 + i2 1 + 22 1 + 22
Use the Kramers-Krönig relation for −1 ,
ω 0 Im[1/(ω 0 )]
Z ∞
1 2
Re[ ]−1= P dω 0 (261)
(ω) π 0 ω 02 − ω 2
Consider the limit ω → ∞,
Z ∞
2
Re[1/] − 1 = − 2 P dω 0 ω 0 Im[1/(ω 0 )] (262)
πω 0
44
II.12. Induced charge in a metal.
Consider an impurity in a metal. Since there is no oscillation, we take ω = 0. We have Lindhard
dielectric function as,
ks2 F (q/2kF ) 1 1 − (q/2kF )2 1 + q/2kF
(q, ω = 0) = 1 + , F (q/2kF ) = ln (266)
q2 2 4(q/2kF ) 1 − q/2kF
And 2 (q, ω = 0) = 0. In this way, (q) = 1 (q) ∈ R.
From Maxwell’s equations,
∇ · D(r) = 4πρi (r)
(267)
∇ · E(r) = 4π(ρi (r) + ρs (r))
where ρi is impurity charge density and ρs is induced charge density.
With Fourier transform, we have,
iq · D(q) = 4πρi (q)
(268)
iq · E(q) = 4π(ρi (q) + ρs (q))
Impurity charge can be treated as a point charge in large r limit,
ρi (r) = Qδ(r)
Z
(269)
ρi (q) = ρi (r)e−iq·r dq = Q
The residue at the singular point z = iks is proportional to e−ks r /r. But we only consider r → ∞,
so it decays to zero.
With two branch cuts, x1 = ±1 ⇔ q = ±2kF , define two contours: C1 = {(x, y)|x = −2kF , y ∈
(+∞, η) ∪ (−η, −∞)} and C2 = {(x, y)|x = 2kF , y ∈ (+∞, +η) ∪ (−η, −∞)}.
" #
Q q2
Z Z
ρs (r) = 2 lim + qdq sin(qr) 2 −1
4π r η→0 C1 C2 q + ks2 F (q/2kF )
Q πks2 8kF3
Z ∞ Z ∞
= lim e−2ikF i du e−ur + e2ikF i vdv e−vr (273)
4π r 4kF (4kF + 21 ks2 )2 η→0
2 2 2
η η
45
II.13. Zeros and poles of dielectric function.
The interaction of light with a medium can be described by the dielectric function (k, ω), and the
Maxwell equations (in a non-magnetic medium):
k · (E) = 0 (278)
k·B = 0 (279)
ω
k×E = B (280)
c
ω
k × B = − (E) (281)
c
In general there are no longitudinal modes because,
k · (E) = 0 =⇒ E = 0 (282)
except when = 0. Therefore, longitudinal modes are only allowed for (ω, k) ∈ (ω, k) = 0.
Combine Eqn. (279) and (281), one gets the dispersion relationship for the transverse mode:
ω2 c2
= (283)
k2 (k, ω)
At high frequencies ≈ 1 then ω 2 /k 2 = c2 is just the free space dispersion, i.e. at high frequencies
the light does not interact strongly with the medium. There are also low frequency solutions to
Eqn. (283), which are the transverse normal modes of the medium. Since ω 2 is small and c2 is
large, Eqn. (283) is satisfied only if (k, ω) is very large, which occurs near the poles of . For
practical purposes, we can say that the transverse normal modes occur at the poles of .
46
Figure 20: Frequency-dependent reflectivity
Note that
∂ρ(r)
= |φj (r)| (286)
∂nj
So the derivative of each term is
∂Eext
Z
= d3 r vext (r)φ∗j (r)φj (r)
∂nj
∂T 1
Z
=− d3 r φ∗j (r)∇2 φj (r)
∂nj 2
∂Eee
Z
φ∗j (r)φj (r)ρ(r0 )
=e2
d3 r d3 r0 (287)
∂nj |r − r0 |
Z
= e2 d3 r vH (r)φ∗j (r)φj (r)
∂Exc
Z
= d3 r vxc (r)φ∗j (r)φj (r)
∂nj
47
We recognize the Kohn-Sham Hamiltonian as
1
H KS = − ∇2 + vext (r) + vH (r) + vxc (r) (288)
2
So,
∂E
Z
= d3 r φ∗j (r)H KS φj (r) = KS
j (289)
∂nj
This result is also known as Janak’s theory; a slightly more complete derivation including self-
consistent effects is found in PRB 18, 7165 (1978). In principle, this relation shows that there is a
connection between the eigenvalues of the Kohn-Sham equations and the energy to add or remove
an electron to the system. However, this relation alone is not enough to prove this property,
since this relation is only valid for infinitesimal changes in the number of electrons, which is not
physical. It can still be shown that, if the exact exchange-correlation is known, then the Kohn-
Sham eigenvalues for the highest occupied and lowest orbitals correspond to the ionization potential
and electron affinity of the system, respectively. See PRL 49, 1691 (1982).
Sec. III
III.1. Light absorption in InSb.
InSb is a direct band gap semiconductor. We ignore the heavy hole valence band for now. We use
the energy and momentum conservation relation in the optical absorption:
p2e p2h
2m∗c + 2m∗v = Ephoton − Egap (291)
pe = ph (292)
The electron and hole kinetic energies are 0.309 eV and 0.011 eV, respectively. If the photon
momentum is included, we have
Ephoton
pe = ph + (293)
c
The electron and hole momenta differ by approximately 0.8%.
III.2. Exciton.
48
(a) The electron and hole effective masses for GaAs are 0.067m0 and 0.45m0 , respectively, where
m0 is the mass of a bare electron. The dielectric constant of GaAs is 14. The exciton binding
energy is,
me mh
Eb = × 13.6eV = 4meV (294)
(me + mh )m0 2
(b) Yes. An electron and a hole from an indirect band gap can still form a bound state. Such a
bound state may not be optically bright under the dipole approximation, since the electron
and hole have different momenta.
Ei + Er = Et (295)
Bi − Br = Bt (296)
Then we have, √
Ei − Er = Et (297)
Er
We eliminate Et and solve for Et .
√
√ √ Er −1
( − 1)Ei + ( + 1)Er = 0 ⇒ = −√ (298)
Ei +1
Recall the definition of reflectivity,
√
− 1 2
Er 2
R = = √ (299)
Ei + 1
and, √ √
= 1 + i2 = n + iκ ⇒ 1 = n2 − κ2 , 2 = 2nκ (300)
we get, r q
q
(n − 1)2 + κ2 21 + 22 +1− 2( 21 + 22 + 1 )
R= =q (301)
(n + 1)2 + κ2
r q
21 + 22 +1− 2( 21 + 22 + 1 )
ρext = 0 (302)
J ext =0 (303)
∂P
J ind = ∂t ⇒ ind (q; ω)
JG = (−iω)PG (q; ω) (304)
q+G
ρind = −∇ · P ⇒ ρind ind
G (q; ω) = −i(q + G) · PG (q; ω) = ω · JG (q; ω) (305)
E = −∇φ − 1c ∂A ω
∂t ⇒ EG (q; ω) = −i(q + G)φG + i c AG (306)
∇2 φ = −4πρ = −4πρind ⇒ (q + G)2 φG = 4πρind G (307)
2 1 ∂2A 1 ∂ 4π 4π ind
∇ A − c2 ∂ 2 t − c ∂t ∇φ = − c J = − c J (308)
∇ · A = 0 ⇒ (q + G) · AG = 0 (309)
49
where quantities with subscript G, such as φG actually means ρG (q; ω).
Take Div of Eq. 308,
1 ∂ 1∂ 2
LHS = ∇2 (∇ · A) − 2
(∇ · A) − ∇ φ
c ∂t c ∂t
1∂ 4π ∂ ind
=− −4πρind = ρ
c ∂t c ∂t (310)
4π
= RHS = − ∇ · J ind
c
⇒ iωρind
G = i(q + G) · JGind
, After Fourier transform
Therefore the longitudinal component of D is exactly zero, and we only have a transverse displace-
ment vector,
T
DG = DG (318)
50
(a) We use the energy and momentum conservation relation in the optical absorption.
1 1
(p21 c2 + m2 c4 ) 2 + Ephoton = (p22 c2 + m2 c4 ) 2 (319)
Ephoton
p1 + c = p2 (320)
p1 and p2 denote the momentum of an electron before and after absorbing a photon, respec-
tively. The above set of equations has no solution, unless E = 0.
(b) According to Drude model of metal
ωp2 τ 2
1 (ω) = 1 − 1+ω 2 τ 2
(321)
ωp2 τ
2 (ω) = ω(1+ω 2 τ 2 )
(322)
ω2 ωp2
(i) At ωτ 1, we have 1 (ω) = 1 − ωp2 , and 2 (ω) = ω3 τ
|1 |.
(ii) At ω < ωp , 2 (ω) |1 |. So R(w) ≈ 1.
ω2
(iii) At ω ωp , R(w) = ( 4ωp2 )2 .
~ω = Ecv ± Ep (323)
kc − kv = ∓Q (324)
51
Z ~ω∓Ep −Eig
Rind ∝ (Ec − Eig )1/2 (~ω ∓ Ep − Ec )1/2 dEc (329)
Ei g
Ec −Eg
Define x = ~ω∓Ep −Eig , then
Z 1
2
Rind ∝ (~ω ∓ Ep − Eig ) (Ec − Eig )1/2 x1/2 (1 − x)1/2 dx (330)
0
Z Z
Rind ∝ Dv (Ev )Dc (Ec )Ev δ(Ec − Ev − ~ω ± Ep )dEc dEv (333)
∝ (~ ∓ Ep − Eig )3 (334)
(
1
ω2
(~ω ∓ Ep − Eig )3 , ~ω ≥ Eig ± Ep
2 (ω) ∝ (335)
0, otherwise
A B
2 (ω) ∝ (~ω − Ep − Eig )3 [np (T ) + 1] + 2 (~ω + Ep − Eig )3 [np (T )] (336)
ω2 ω
III.7. Hydrogenic excitons with anisotropic perturbation
We introduce the relative and center-of-mass coordinates:
(
x = xe − xh
2 (ω) ∝ xe −mh xh (337)
X = mem e +mh
where mex = mey = m∗ke , mhx = mhy = m∗kh , mez = m∗⊥e , mhz = m∗⊥h .
52
In the center-of-mass reference system,
(0) µe4 1 1 1
E1s = E1 + h1s|H1 |1si = − 2 2
+ ( − )h1s|p2z |1si (342)
2 ~ 2 µ2 µ1
2µ1 (0)
By Virial theorem, we can get h1s|Pz2 |1si = 3 (−E1 ) ,so
(0) 4 µ1
E1s = E1 ( − ) (343)
3 3µ2
And
1 2µ1 (0)
h2s|p2z |2si = h2s|p2 |2si = (−E2 ) (344)
3 3
(0) 4 µ1
E2s = E2 ( − ) (345)
3 3µ2
For 2p0 , 2p±1 states:
1 Z − 2ar
ψ2p0 (r) = √ 3/2
e 0 (346)
4 2πa0 a0
1 1 1 ∂2
Z
∗
E2p0 = ( − ) drψ2p (r)(−~2 )ψ2p0 (r) (347)
2 µ2 µ1 0
∂z 2
and,
r − r
1 1 1 1 ∂2
Z
ψ2p±1 (r) = √ 3/2 e 2a0 sin θe±iφ E2p±1 = ( − ) ∗
drψ2p (r)(−~2 )ψ2p±1 (r) (348)
8 πa a0 2 µ2 µ1 ±1
∂z 2
0
(a) The neutron can emit two phonons, or absorb two phonons. Assume the energy change and
the momentum transfer of the neutron in the scatter process are ∆E and ∆P , respectively,
energy and momentum conservation dictates that, the total energy and momentum of the two
phonons should be ∆E and ∆P , respectively.
(b) In the two-phonon process, for a specific neutron energy change and observation angle, there
exist multiple combinations of different phonons which can satisfy the conservation of energy
and crystal momentum in the scattering process. However, in a one-phonon process, each neu-
tron energy change only corresponds to a specific phonon. As a result, two-phonon processes
do not produce sharp peaks in the number of scattered neutrons as a function of neutron
energy change during the scattering event for fixed observation angle.
53
(a) Electron-phonon interaction
M (q)ρ(q)[aq + a†q ],
P P †
Hep =
ρ(q) = ck+q ck
q q k (349)
M (q) = iC
~
|q|
2ρΩωq
c21 ~2m∗ X q
∆Ek = (357)
2pΩcs q 2k · q − q − 2m∗ cs q
2
R
→
P
We change the summation to an integral, D(E)dE. Using the single mode 3d Debye
q
3 = 6N π
model, qD Ω
c21 ~2m∗ q 3 dq
Z qD
∆Ek = (358)
2pπ 2 cs 0 2k · q − q 2 − 2m∗ cs q
For k = 0,
c2 ~m∗ q 2 dq
Z qD
∆E0 = − 1 2 (359)
2ρπ cs 0 q 2 + 2m∗ cs
Evaluating the integral we finally get
54
(b) The total number of phonons:
X |hk − q; q|Hep |k; 0i|2
hN i = (361)
q (Ek − Ekq − ωq )2
When k = 0,
c2 ~m∗
Z qD
qdq
hN i = − 1 2 (362)
2ρπ cs 0 (q 2 + 2m∗ cs )2
Evaluating the integral, we finally get
is the phonon destruction operator and ck the electron destruction operator. The perturbation is
the electron-phonon interaction,
X Mq a†q |k − q, 0i
|ψi = |k, 0i + (367)
q Ek − Ek−q
k 2 2 2
where Ek = ~2m ~
, Ek−q = 2n (k − q)2 + ~ωq . |k, 0i is the direct product of the free electron
wavefunction and the ground state of phonon |0i. Using,
we obtain,
X |Mk0 |2
hN iav = ~2 k02 ~2 k0 ·k 2
(369)
k0 (~ω + 2m − 2m )
55
(a) E = −α~ω = − 21 (0.01) = −0.005eV.
(b) m∗ = 0.6m
1− α = 0.6
1−1/12 m = 0.65m.
6
56
at temperature T = 0 ,
|k0 i = |di,
(
f0 (En0 k0 ) = 1
(377)
|ki = Planewave, f0 (Enk ) = 0
Then
πe2 ~2 X
Re[σij ] = hd|∇j |kihk|∇i |diδ(Enk − End − ~ω) (378)
m2 ω k
Where
Because hd|∇j |di = 0 as, |di is spherical symmetric but ∇j is spherical antisymmetric.
Then
~2 k 2 ~2 α2
Z
Re[σij ] ∝ dkk 4 δ( + − ~ω)
2m 2m
(384)
4 δ(k − k0 )
Z
= dkk ~2 k0
m
where k0 = ( 2mω 2
~ − α ) As a result,
1 1
Z
2
dkk 3 2
Re[σij ] ∝ δ(k − k0 )δij
ω (α + k 2 )2
(385)
2mω 1
∝( − α2 )3/2 5 δij
~ ω
Sketch: see Fig. 21
57
s
w
Figure 21: Sketch
(a)
~c
r0 = kF = 6.25 × 10−5 cm (386)
eH
(b)
2
eH
Ak = Ar (387)
~c
(c)
1 −e
k̂ × k̂ = [p × A + A × p] (388)
~2 c
Apply on a function f ,
e ie ie
(k̂ × k̂)f = − [(p × A)f − A × (pf ) + A × (pf )] = (∇ × A) = B (389)
~2 c ~c ~c
III.15. Magnetic properties of materials in classical physics
where particles interact with each other according to an arbitrary potential U that depends
only on their positions.
N
Y N
Y
dΩ = dωi = dxi dyi dzi dPxi dP yidP zi (392)
i=1 i=1
P
−β N
i Y
1
Z Z
ZN = ... e i dωi (393)
N !h3N i=1
N
1 Y 1
Z
−βi
= e dωi (394)
N ! i=1 h3
zN
= (395)
N!
58
(b) Consider spatially uniform static magnetic field we have,
B = Bk
E=0 ⇒ A = (0, Bx, 0) (396)
B = ∇ × A
z̃i = zi . The partition function is invariant under this coordinate transformation, therefore
cannot affect any thermodynamic variable. There cannot be any magnetization within the
classical theory.
III.16. Magnetoresistivity.
Suppose we have a standard geometry of Hall bar and apply a B field in z direction, and an
external E field in x direction. In steady state, the total E field is in x − y plane.
E = Exi + Ey j, B = Bk (400)
The drift velocity v of each species of carrier can be treated using the Relaxation Time Approxi-
mation,
dv v
m( + ) = q(E + v × B) (401)
dt τ
where m is isotropic and energy-independent effective mass.
(
m( dv
dt +
x vx
τ ) = q(Ex + vy B)
dv vy (402)
m( dty + τ ) = q(Ey + vx B)
Then,
q2B 2 2 qτ eBτ
vy (1 + 2
τ )= (Ey − Ex ) (405)
m m m
59
Consider an arbitrary number of carrier types,
qj2 B 2 2 qj τj qj Bτj
vyj (1 + 2 τj ) = (Ey − Ex ) (406)
mj mj mj
Assume q1 = −q2 = e, m1 = m2 = m, τ1 = τ2 = τ .
Define ξ = eτmB . We have
1 2 eBx 2
H= (px + (py + ) + p2z ) − eψ(r) (415)
2m c
We take
60
we will solve the eigenvalue problem HΨ(r) = E(r),
" #
1 2eBx e 2 B 2 x2
p2x + p2y + py + + p2z Ψ(r) = Eei(py y+pz z)/~ χ(x) (417)
2m c c2
1 eBx 2 1 2
p2x + (py + ) χ(x) = (E − p )χ(x) (418)
2m c 2m z
1 2 1 1 2
px + mω 2 (x − x0 )2 χ(x) = (E − p )χ(x) (419)
2m 2 2m z
eBx 2 e2 B 2 cpy 2
where 21 mω 2 (x − x0 )2 = 1
2m (py + c ) = 2mc2
(x + eB )
e2 B 2 eB
ω2 = ⇒ ω= (420)
m2 c2 mc
cpy
x0 = − (421)
eB
cp
so it is a harmonic oscillator in x direction with the equilibrium position at x0 = − eBy so the
energy level:
p2z 1
E(n,pz ) =
+ hω(n + ) (422)
2m 2
so the energy level are discrete for each fixed value of Pz , and energy separation is in accordance
with the cyclotron formula.
The wavefunctions:
2 (x−x 2 /2
Ψpz ,n (x) = ei(py y+pz z)/~ e−α 0)
Hn (α(x − x0 )) (423)
q
mω
where α = ~ and Hn (x) is Hermite function.
For the gauge A = (− 21 By, 12 Bx, 0)
1 eB 2 eB 2
H= (px − y) + (py + x) + p2z (424)
2m 2c 2c
1 2 e2 B 2 2 eB 1 2
= (px + p2y ) + (x + y 2 ) + (xpy − ypx ) + p (425)
2m 8mc 2 2mc 2m z
1 eB ∂
define H0 = 2m (p2x + p2y ) + 12 mωL2 (x2 + y 2 ), ωL = 2mc , and we notice that Lz = xpy − ypx = −i~ ∂φ .
Assume in cylindrical coordinate
Ψ(ρ, φ) = R(ρ)eimφ , m = 0, ±1, ±2... (426)
" #
~2 d2 1 d 1
− ( + ) + mωL2 ρ2 R(ρ) = (E − m~ωL )R(ρ) (427)
2M dρ2 ρ dρ 2
By solving the equation, we have,
En = (2np + |m| + m + 1)~ωL (428)
The wavefunction is given by,
2 ρ2 /4
Rnρ +m+1 ∝ ρ|m| F (−nρ , |m| + 1, α2 ρ2 /2)e−α (429)
q
eB
where α = 2~c , and F is confluent hypergeometric function. So both gauges give us discrete
eigenvalues.
61
III.18. Bohr-Sommerfeld quantization. Principles of the Theory of Solids, J. M. Ziman, Sec.
9.6-9.7.
III.19. Cyclotron orbit.
K-space orbit: eclipse motion,
s
E
kx (t) = cos ωc (t − t0 ) (430)
A
s
E
ky (t) = sin ωc (t − t0 ) (431)
B
kz (t) = const. (432)
Real-space orbit: similar eclipse motion as in k-space, but rotated by 90◦ , timed ~c
eH , and centered
at (x0 , y0 ).
III.20. Conduction band electrons in a magnetic field.
For H = 0,
~2 (kx − k0 )2 ky2 + kz2
" #
E(k) = Ec (k0 ) + + (433)
2 ml mt
where ml is the longitudinal mass and mt the transverse mass. If we now switch on H 6= 0,
(px − eAx /c)2 (py − eAy /c)2 + (pz − eAz /c)2
Hef f = + (434)
2ml 2mt
The vector field is chosen to be,
A = Hy(− sin θ, 0, cos θ) (435)
such that H is in the x − z plane. With some manipulation,
p2y
2 " #
1 eH sin2 θ cos2 θ 2 ~2 kx2 ~2 kz2 eH kx sin θ kz cos θ
H= + + y + + + ~y −
2mt 2 c ml mt 2ml 2mt c ml mt
(436)
p2y 1
= + mt ω ∗ 2 (y − y0 )2 + A
2mt 2
r
~2 kx2
∗
where ω ∗ = eH
m∗ c , 1/m∗ = 1
mt
sin 2θ
ml + cos 2θ
mt , y0 = − mmt ω~∗ kx sin θ
ml − kz cos θ
mt , and A = 2ml +
2
~2 kz2
1 ~2 m∗2 kx sin θ kz cos θ
2mt − 2 mt ml − mt .
62
III.22. Mott formula for thermopower.
1 −1
Q= L L1 (439)
eT 0
Within the effective mass approximation, assume a low temperature, we have
∂f 0 σ
Z
L0 = R(E)(E − µ)0 (− )dE = 2 (440)
∂E e
and
∂f 0 ∂R(E) ∂f 0 1
Z Z
L1 = 1
R(E)(E − µ) (− )dE = (E − µ)2 (− )dE = 2 π 2 (kB T )2 σ 0 (441)
∂E ∂E ∂E 3e
As a result,
1 2 2 σ0
Q= π kB T (442)
3e σ
Sec. IV
IV.1. Green’s function of a free Fermi gas.
(
−i(1 − np )e−iεp t , t>0
G(p, t) = (443)
inp e−iεp t , t≤0
1
G(p, p0 ) = (444)
p0 − εp + iη sgn(pF − p)
IV.2. Momentum distribution.
(a)
G(p, 0− ) = inp , G(p, 0+ ) = −i(1 − np ) (445)
(b)
G(p, 0+ ) − G(p, 0− ) = −i (446)
IV.3. Interacting Green’s function.
(a) The poles in the lower half-plane (Γp > 0) determine the quasi-particle (electron) energies for
|p| > pF , while the poles in the upper half-plane (Γp < 0) determine the quasi-particle (hole)
energies for |p| < pF . Using the definition of,
1
Z
Np = dε G(p, ε) (447)
2πi C1
where the contours are shown in Fig. 22. Note that the contour C2 is above the real axis since
1 R∞
this expression comes from G(p, 0− ) = iNp and G(p, t) = 2π −∞ dε G(p, ε)e
−iεt .
63
Im[ω]
C2
C1 Re[ω]
(b) Consider p1 (|p1 | > pF ) scattering off of p2 (|p2 | < pF ) into two states p3 , p4 (both > pF ). Since
Γp is proportional to the amount of phase space available for this process, we will calculate
the phase space in the following.
First recall the conservation of energy:
and define,
∆pn ≡ |pn − pF | (451)
which leads to,
∆p1 − ∆p2 ≈ ∆p3 + ∆p4 (452)
where we have dropped second order terms.
Since ∆p1 − ∆p2 > 0[∆p3 , ∆p4 > 0], we will have 0 < ∆p2 < ∆p1 so the phase space available
for p2 is proportional to 4πp2F (p1 − pF ). And, given a certain p1 , p2 and ∆p1 − ∆p2 , then
0 < ∆p3 , ∆p4 < ∆p where ∆p ranges from 0 to ∆p1 . Your choice of ∆p3 (given ∆p2 ) fixed ∆p4 ,
so the phase space associated with the energy distribution in the final state is proportional to
∆p1 = (p1 − pF ). Also the directions of p3 , p4 are severely restricted for p1 near the Fermi
surface, so we neglect that effect on Γp . In the end, we have,
The initial unperturbed state is the Fermi sea with one electron at p: c†p |0i. The second-order shift
in energy is,
X X |gpp0 |2 |hψj |c†p0 cp (aqλ + a†−qλ )|ψp i|2
∆Ep = (455)
λp0 j
Ep − Ej
64
(a) The electron at p may emit and re-absorb a virtual phonon:
X |gpp0 ,λ |(1 − n0p )(N−q + 1) X Z
d3 p0 |gpp0 ,λ |2 (1 − np0 )(N−q + 1)
∆E1 = = P (456)
λp0
εp − εp0 − Ω−qλ λ
(2π)3 εp − εp0 − Ω−qλ
where the principal integral is taken in the continuous limit, and Ep − Ej = εp − (εp0 + Ω−qλ ).
(b) At the same time, electrons in the Fermi sea cannot (via virtual phonons) make a transition
to p. This involves an energy shift −∆E2 :
qT2 F m2 n 2 2
o
2 (q, ω) = (4F q − (q + ω) )(−π) + (4F q − (q − ω) )(π)
2q 2 2kF q 3
2 2
(461)
e m
= 2ω( 3 )
q
ii) q + qvF > ω > qvF − q > 0, and therefore q + qvF + ω > 0, q − qvF + ω > 0,
q + qvF − ω > 0, and q − qvF − ω < 0. In this way,
qT2 F m2 2 m2 e2 2 m2
2 (q; ω) = (4F q − (q − ω) )π = [kF − (q − ω)2 ] (462)
2q 2 2kF q 3 q3 q2
65
i) q − qvF > ω > 0, we have 2 (q, ω) = 0.
ii) q + qvF > ω > −qvF + q > 0,
m2 e2 2 m
2 (q, ω) = [kF − ( )2 (q − ω)2 ] (463)
q3 q
Similarly,
! !2
ω − − qv 2qvF −2qvF 1 2qvF
q F
In = In 1 − = − + · · · (466)
ω − q + qvF ω − q + qvF ω − q + qvF 2 ω − q + qvF
Plug Eq. (465) and (466) into Eq. (459) and we will find out that 1 (q, ω → ∞) = 1 + O( ω1 ),
which means the real part 1 always approaches 1 at large ω.
(d) From (a), it’s obviously that 2 (q, ω → 0)
i = 0, and the static (q) is just
2
qT
h
1 1+x q
1 (q, 0) = 1 + F
1+ 2x (1 − x2 )In 1−x , where x = .
2q 2 2kF
(e) We can simplify the real part 1 of Eq. (IV.2) as,
a/2 + 1 − x
2 c 1 a
.
1 (q, ω) = (a, x) = 1 + α 1 + [1 − (− + x)2 ]In
π vF 2a 2 a/2 − 1 − x
(467)
1 a a/2 + 1 + x
+ [1 − ( + x)2 ]In
2a 2 a/2 − 1 + x
. . mω
where α is fine-structure constant, c the speed of light, a = kqF and x = qk F
. From Page 38 in
Solid State Physics by Ashcroft&Mermin, we know that the Fermi velocities of representative
metals (such as Na, Ag, Au, Fe, etc.) are around 0.8 × 106 ∼ 2.0 × 106 m/s. In this way, we
fix a and vF , and plot 1 (x) as a function of x in Fig. 23. As we can see, for representative
metals, when q < kF , we will have negative 1 for intermediate value of ω.
(f) We need to show that Eq. (IV.3) gives the same results of Eq. (IV.2) within all the ranges of
q and ω. Define v = cos θ, and note that,
Z 1
pqv m . m
dvδ(q + − ω) = Θ(p − p1 ), p1 = |ω − q | (468)
−1 m pq q
and similarly, Z 1
qpv m . m
δ(q + + ω) = Θ(p − p2 ), p2 = |ω + q | (469)
−1 m pq q
66
a=1 a = 0.8
6 6
vF= 1.0x10 m/s vF= 1.0x10 m/s
ε1
a=1 a=1
6 6
vF= 0.8x10 m/s vF= 2.0x10 m/s
ε1
x x
Figure 23: Lindhard dielectric function 1 for representative metals
And therefore,
me2 h 2 2 2 2
i
2 (q, ω) = Θ(kF − p1 )(kF − p1 ) − Θ(kF − p 2 )(kF − p 2 ) (470)
q3
For example, in the case of q < 2kF and qvF −q > ω > 0, we have kF −p1 > 0 and kF −p2 > 0,
and therefore,
me2 2m2 e2
2 (q, ω) = 3 (kF2 − p21 − kF2 + p22 ) = ω (471)
q q3
which is exactly the result of Eq. (461). The reader should confirm other cases just like above.
67
Using the Dirac identity,
1 1
lim =P ∓ iπδ(x) (475)
η→0 x ± iη x
we can get, (
1 −πF (q, q0 ), q0 > 0
Im = (476)
(q) πF (q, q0 ), q0 < 0
and
− 1 Im 1
, q0 > 0
F (q, q0 ) = 1 π (q,q
1
)
0
(477)
Im
π , (q,q0 ) q0 < 0
IV.7. Behavior of self energy. Migdal’s theorem is an approximation to the exact relation,
d4 q
Z
Σph (p) = i ḡq G(p + q)Γ(p, q)D(q) (478)
(2π)4
where Γ(p, q) is the vertex. Migdal’s asserts that the correction from the second and higher-order
m 1/2
ḡq ∼ 10−2 ḡq and are negligible. As a proof, consider the second-order
terms is of the order M
correction. To simplify things, assume that the electron-phonon matrix element m depends only
on the momentum transfer then,
d4 k
Z
0
Γ2 (q, q ) = i |gq0 |2 D(q 0 )G(k)G(k + q) (479)
(2π)4
The range of energy that is relevant is such that k0 , q0 ≤ ωD , and |p||q| ∼ pF . Now D(q) ∼ q0−2
for large q0 > ωD . So there is a natural cutoff in the integration at ωD . We can cutoff the electron
momentum at pF when the kinetic energy dominates the weak ionic potential inside the core, and
also replace the electron-phonon coupling by |g|2 /ωD , an average value. Then,
|q|2 d3 k
Z ωD
dk0 1 1
Z
Γ2 ≈ i 3
(480)
ωD −ωD 2π (2π) k0 − k + iδk0 k0 + q0 − k+q + iδk0 +q0
(a) Time-ordered Green’s function G0T (p, t − t0 ) for free electron gas,
i h
i
0
G0T (p, t −t)= − Θ(t − t0 )hφN 0 |cp (t)c† 0
p (t )|φ N
0 i − Θ(t − t 0
)hφ N † 0
|c
0 p (t )c p (t)|φ N
0 i
~
i
0
h i
= − e−ip (t−t )/~ Θ(t − t0 )hφN 0 |cp c † N
|φ
p 0 i − Θ(t − t 0
)hφ N †
|c
0 p c p |φ N
0 i (481)
~
i
0
e−ip (t−t )/~ Θ(t − t0 )Θ(p − kF )) − Θ(t − t0 )Θ(kF − p)
= −
~
68
Retarded Green’s function G0R (p, t − t0 ) for free electron gas,
i
h i
G0R (p, t − t0 ) = − Θ(t − t0 ) hφN0 p|c (t)c † 0
p (t )|φ N
0 i + hφ N † 0
|c
0 p (t )cp (t)|φ N
0 i
~
i
0
h i
= − Θ(t − t0 )e−ip (t−t )/~ hφN |c c
0 p p 0
† N
|φ i + hφ N †
|c c
0 p p 0 |φ N
i (482)
~
i
0
= − Θ(t − t0 )e−ip (t−t )/~
~
(b) Time-ordered Green’s function G0T (p, ω) for free electron gas,
Z ∞ Z ∞ Z 0
i
G0T (p, ω) = dt eiωt G0T (p, t) = −
Θ(p − kF ) dt ei(ω−p )t/~ − Θ(kF − p) dt ei(ω−p )t/~
−∞ ~ 0 −∞
Θ(p − kF ) Θ(kF − p)
= +
~ω − p + i η ~ω − p − i η
1
=
~ω − p + i sgn(p − kF )η
(483)
where η → 0+ .
Retarded Green’s function G0R (p, ω) for free electron gas,
Z ∞ Z ∞
i
G0R (p, ω) = dt eiωt G0R (p, t) = − dt ei(ω−p )t/~
−∞ ~ 0
(484)
1
=
~ω − p + i η
(d) Obvious.
(e) Z ∞
dE 1 1
RHS = 2πδ(E − p ) = βp = np = LHS (487)
−∞ 2π βE +1 e +1
(f)
ImΣ
A(p, ω) = −2 (488)
[~ω − (p − µ) − ReΣ]2 + (ImΣ)2
(g) Recall the definition of generalized power function: w(z) = z c = ecInz , z 6= 0.
• If c = k ∈ Z, we have w = z k = |z|k eikarg(z) = ekInz , which is single-valued.
• If c = 1/k, k ∈ Z, we have w = |z|1/k eiθ/k ei2nπ/k , which is multi-valued.
We define the principle value of z c = ecInz , z 6= 0. In this way, w(z) is multi-valued for any
non-integer value of c, with a branch cut on positive real axis.
If we have a self-energy in the form of ΣR (p, z) = C[f (p) − z]1/2 ,
• If f (p) − E > 0, then ΣR (p, E) = C[f (p) − E]1/2 = C[a]1/2 = CeIna/2
69
π
• If f (p) − E < 0, then ΣR (p, E) = C[aei(−π) ]1/2 = CeIna/2 e−i 2
For electron-hole excitation, we study the electron-hole pair excited at the same time, that is,
t01 = t1 , t02 = t2 . Write explicitly the expression of G1 (1, 20 ) and G1 (2, 10 ):
i X i
0 0
0
L (1, 1 ; 2, 2 ) = (− )2 φj (1)φ∗j (20 )φk (2)φ∗k (10 ) exp − (j − k )(t1 − t2 )
~ jk ~
× [Θ(t1 − t2 )haj a†j i − Θ(t2 − t1 )ha†j aj i] × [Θ(t2 − t1 )hak a†k i − Θ(t1 − t2 )ha†k ak i]
1 X i
= 2 φj (1)φ∗j (20 )φk (2)φ∗k (10 ) exp − (j − k )(t1 − t2 )
~ jk ~
70
(a) In the S phase, the magnetic B field is always 0, and in the N phase, M is 0. At a given T,
from Hc where S and N have the same free energy, to H where S is lower in free energy than
N, we have the free energy difference,
H 1
Z Z
∆G = M dH = − dH = [H 2 − Hc2 ] (494)
4π 8π
(b)
∂∆G Hc ∂Hc
∆S = −( )H = ( )H (495)
∂T 4π ∂T
(c) At the phase transition, Hc = 0.
∂S Tc ∂ 2 Hc H0
∆c = T = 2
=− (496)
∂T 4π ∂ T 2πTc
Integrate the terms with δψ ∗ and δA by parts and assume that ψ and its derivatives vanish
at the boundaries, and we will get,
1
Z
δFs = dr δψ ∗ αψ + β|ψ|2 ψ + (−i~∇ + e∗ A)2 ψ + c.c.
2m∗
(499)
∇×B e∗ ∗
Z
+ dr δA · + ψ (−i~∇ + e∗ A) ψ + c.c.
2µ0 2m∗
At the minimum δFs = 0, and therefore the coefficient of δψ ∗ should vanish, which gives first
GL equation,
1
(−i~∇ + e∗ A)2 ψ + αψ + β|ψ|2 ψ = 0 (500)
2m∗
And the second GL equation can be derived by equating the coefficient of δA to zero,
1 e∗
∇ × (∇ × A) = − ∗ ψ ∗ (−i~∇ + e∗ A) ψ + c.c. (501)
µ0 2m
71
The GL equation is aΨ + b|Ψ|2 Ψ∗ − c∇2 Ψ = 0 and the natural boundary condition is that
n · c∇Ψ is continuous across the boundary. To linearize it, we ignore higher order terms, and
it becomes:
aΨ − c∇2 Ψ = 0 (503)
i.e.
d2 Ψ
aΨ − c =0 (504)
dx2
so
Ψ(x) = Aeκx + Be−κx
s
r
a a T (505)
κ= = ( − 1)
c c Tc
(c) On the left normal metal side,
Ψ(x) = Aeκ1 x
s
a1 T (506)
κ1 = ( − 1)
c Tc1
According to GL equation, we have Ψ∗ ∇Ψ − Ψ∇Ψ∗ = 0. If Ψ = |Ψ|eiθ , then we get ∇θ = 0.
θ is constant in space, then we can choose it to be 0. Therefore A is real.
On the right SC side,
∆Ψ(x) = C sin κ2 x + D cos κ2 x
s
a0 T (507)
κ2 = (1 − )
c0 Tc0
First GL equation is that
a(Ψ0 + ∆Ψ) + b(Ψ0 + ∆Ψ)3 − c∇2 (Ψ0 + ∆Ψ) = 0 (508)
with Ψ0 being a local minimum. It gives us a∆Ψ − c∇2 (∆Ψ) = 0 and the solution above.
Boundary condition at x = 0 plane.
From above, we have
Ψ− (x) = Aeκ1 x (509)
Ψ+ (x) = Ψ0 + C sin κ2 x + D cos κ2 x (510)
72
So
Ψ0
A=B= (517)
2 cosh(κ1 d)
c1 κ1 Ψ0
C cos(κ2 d + Ψ+ ) = −D cos(κ2 d − Ψ− ) = tanh(κ1 d) (518)
c0 κ2
Therefore, the phase difference can be tuned by the normal metal layer in between.
2.0
1.8
1.6
1.4
1.2
1.0
0.8
-3 -2 -1 0 1 2 3
(a) Start with an expression for the difference in the Helmholtz free energy density between su-
perconducting and normal phases.
φs − φn 4π
f≡ H2
= ψ04 − 2ψ02 − 2 V µ · H0 (519)
( 8πcb )V Hcb
2 2 2
where ψ02 = ψψ2 , ψ∞
2 = 4πe δ , µ = 1 R
mc2 4π (H − H0 )dV . Hcb is the bulk critical field, δ is the
∞
London penetration depth.
73
We want to find f (ψ0 , H0 ), so we need to compute µ. This means we should compute the
magnetic field in the slab. The solution to this problem is,
∂f
We determine ψ0 as a function of H0 by minimizing f , ∂ψ0 = 0.
√
d 5
(b) If δ < 2 , then ψ02 goes to zero smoothly when H0 = Hc . Thus the phase transition is first
√ √
d 5 d 5
order when δ > 2 , and second order when δ < 2 .
∂2B 1 ∂B B
+ − 2 =0 (522)
∂ρ2 ρ ∂ρ λ
This is the modified Bessel’s equation which has the solution:
ρ ρ
B(ρ) = c1 I0 ( ) + c2 K0 ( ) (523)
λ λ
For x → 0, K0 (x) diverges while I0 (x) is finite and well behaved. Therefore, we must have c2 = 0.
The boundary condition at ρ = R is:
H0
B(R) = H0 → c1 = (524)
I0 (R/λ)
cH0 I1 ( λρ )
J =− φ̂ (527)
4πλI0 ( Rλ)
and,
cH0 I1 ( R
λ)
Jmax = R
(528)
4πλI0 ( λ )
74
Figure 25: BJ
75
For the numerator we use a trick,
!
0 ik0 ·r
X
∇k Ψ(r) = 0 = ∇k a(k )e = eik·r [∇k a(k) + ira(k)] (534)
k0
Finally, 2
R∞
dN () ∂a ∂
0 ∂ ∂k
hr2 i ≈ R∞ (538)
0 d N ()|a|2
Using the approximation,
V
Vk,k0 = − for k , k0 < D (539)
Ω
we get,
2
C ∂a 4C 2 C2
a(k) = , and = , a2 = (540)
E − 2k ∂ (E − 2)4 (E − 2)2
2
∂ 2~2 F
Also ∂k 0 = m = ~2 vF2 , and then,
R∞ ∞
d 1
2 2 0 (E−2)4
3
2 2 6(E−2) 0
4 ~2 vF2
hr2 i = 4~ vF R ∞ d = 4~ vF ∞ = (541)
0 (E−2)2
1 3 E2
2(E−2) 0
and,
2 ~vF
2 2
q ~vF ~vF N (0)V
hr2 i =√ =√ e N (0)V − 1 ≈ √ e for N (0)V 1 (542)
3 |E| 3D 3D
so,
q ~vF
hr2 i ∼ ∼ ξ0 = coherence length (543)
∆
(b) Let ~kcm denote the center of mass momentum. Then the Schrodinger equation gives:
" #
~2
− (∇2 + ∇22 ) + V Ψpair = (E + 2F )Ψpair (544)
2m 1
or, X
[Ek+kcm /2 + E−k+kcm /2 ]ak + ak0 Vkk0 = (E + 2F )ak (545)
k0
~2 k2
where E = binding energy of the pair and Ek = 2m . Define k = Ek − F . Then we obtain,
P
ak0 Vkk0
k0
ak = (546)
E − k+kcm /2 − −k+kcm /2
76
Assuming Vkk0 = − VΩ for 0 ≤ E ≤ ~ωD , we obtain the Cooper pair equation,
V X 1
1=− (547)
Ω k0 E − k+kcm /2 − −k+kcm /2
2
~
but (k+kcm /2 + −k+kcm /2 = 2k + 4m 2 ), so, to first order in k , the denominator is given
kcm cm
by E − 2k , which is the same as that with kcm = 0.
effect of kcm 6= 0 is on the boundary condition when we replace k0 → dk0 →
P R
The
R 2 0
1 d dΩ N (0). Before 1 = 0 and 2 = ~ωD . Now we have,
~2 2 2
1 = [k + 2kF · kcm /2 + kcm ] − F ≈ ~vF kcm (cos θ)/2 (548)
2m F
q . It turns out that 1 = ~vF kcm | cos θ|/2. The upper limit of the integral
to first order in kcm
is such that k2 = kF2 + kD 2 − k | cos θ|/2 which leads to,
cm
since ~ωD is the dominating term. Keeping the kcm term in the upper limit would contribute
2 in the final expression of E. Therefore, we have,
only with a term like kcm
D
Z π/2 Z D Z π/2
V d N () V 1
1=− dθ = − 4πN (0) dθ − ln(E − 2)
Ω 0 ~vF kcm cos θ/2 E − 2 Ω 0 2 ~vF kcm cos θ/2
1 N (0)V 2~ωD
≈ ln 1 −
2 Ω E − ~vF kcm
(550)
For the weak coupling limit we have,
2Ω
Ekcm = −2D e N (0)V + ~kcm vF (551)
(c) If we just consider kcm = 0, the single pair model has an energy gap, but if we consider
kcm 6= 0, then the single pair model has a continuous spectrum (corresponding to various
values of kcm ) because as kcm increases from zero, the pair binding energy decreases linearly.
2Ω
If E0 = 2D e N (0)V = ~kcm vF then,
kB Tc kB Tc 1
kcm = ∼ kF ∼ 10−4 cm−1 ∼ (552)
~vF F ξ0
(d) For a triplet state the spatial wave function Ψ(R = r − r 0 ) has to be antisymmetric under
exchange of coordinates r and r 0 . This means Ψ(r) = −Ψ(−R) which implies that Ψ has
an angular dependence that contains at least the first odd spherical harmonics Y1m . So, if we
consider the same model as in the singlet case where Vkk0 = − VΩ and as no angular dependence
we get (in the development of the Bethe-Goldstone equation):
X X
(2k − E)ak = V ak 0 = V [ak0 + a−k0 ] = 0 (553)
0<k0 <k D k0 in half sphere
So, there is no nontrivial solution for this equation and there is no binding energy. The triplet
state can be favored if the potential has a more attractive component with angular dependence
Vl=1 . This is the case in He3 superfluid.
77
IV.18. BCS model I.
(a) The Bardeen-Pines interaction at finite temperatures can be considered by using the phonon
occupation number. The emission terms change by a factor of (1 − n(q))2 , and the absorption
terms change by a factor of n2 (q).
(b) Screened Coulomb potential Vc (r) = e −ks r ⇒ Vc (q) = e 1
4π0 r e 0 q 2 +ks2 . And the Bardeen-Pine
interaction is given by,
Also for q ∼ ks , we have Vc (q) ≈ e0 2k12 . If we take n = 1029 m−3 ⇒ ks ∼ 1010 m−1 , which leads
s
to,
N (0)Vc (q) ≈ 9, N (0)VBP ≈ 0.1 (555)
(c) For the wavefunction ψ = Πk [uk + vk b†k ]φ0 , we calculate various energy contributions
(d) The gap equation can be derived by minimizing hHBCS i with respect to vk .
q
(e) Minimum energy is ∆2 + ξl2 . In tunneling and photo emission experiments, the measured
gap is ∆. In heat capacity and optical absorption experiments, the measured gap is 2∆.
(b) Z q
X
D(E) = δ(E − Ek ) = N (0) dk δ(E − 2k + ∆2 )
k
√ √
(561)
δ(k + E 2 − ∆2 ) δ(k − E 2 − ∆2 ) 2N (0)|E|
Z
= N (0) dk √ + √ =√
E 2 −∆2 E 2 −∆2 E 2 − ∆2
|E| |E|
78
Figure 26: DOS
(c)
[bk , b†k0 ]− = δk,k0 (1 − nk↑ − n−k↓ ) (562)
b2k = 0 (565)
2
b†k = 0 (566)
∆l
∆k X |ξl | |ξl |
=− λ 2 1/2 (567)
|ξk | l 2 1 + ∆ξll
Note that if ∆k /|ξk | = α, then every term related to ∆k and ∆l will disappear and we can
have an answer to the equation:
X α|ξ|l
α=− λ
l
2(1 + α2 )1/2
2(1 + α2 )1/2 X
Z ~ωD
⇒− = |ξl | = N (0) dξ |ξ| = N (0)~2 ωD
2 (568)
λ l −~ωD
p λ
⇒ 1 + α2 = − N (0)(~ωD )2
2 79
In order for α and ∆k to be real, − λ2 N (0)(~ωD )2 > 1 and therefore λN (0)(~ωD )2 < −2,
q
λ2 2 4
α= 4 N (0)(~ωD ) − 1, ∆k = α|ξk |.
(b) Ek2 = ∆2k + ξk2 = (α2 + 1)ξk2 , when |ξk | < ~ωD .
Ns (E) dξ √ 1 , |ξ | < ~ω
α2 +1 k D
= = (569)
N (0) dE 1, |ξk | > ~ωD
∆ ∆ ∂N
# of Cooper pairs kF 4πkF2 = ∂ = N (0)∆ (570)
EF ∂k
∂k
Then,
1 1
rp3 ∝ ∝ 1/2 (571)
N (0)∆ F ∆
also,
1/2
vF
ξ∝ ∝ F (572)
∆ ∆
which leads to,
2/3
rp ∆
∼ 1 (573)
ξ F
where ∆ ∼ kB Tc ∼ meV for most conventional superconductors like Al.
1 Kc
ln(kB Tc ) = ln(1.141 ) − Z1 , where Z1 = , and Kc∗ = (574)
Kp − Kc∗ 1 + Kc ln(2 /1 )
80
Finding the extrema,
∂ 1 ∂Z1 ∂Kc∗ 1 ∗ 2
2 (Kc )
ln(KB Tc ) = 0 = − = − Z 1 (575)
∂1 1 ∂Kc∗ ∂1 1 1
which means,
Z12 Kc∗ 2 = 1
Kc∗ 2 = (Kp − Kc∗ )2
Kc∗ 2 = Kp2 − 2Kp Kc∗ + Kc∗ 2 (576)
0 = Kp (Kp − 2Kc∗ )
Kc∗ = Kp /2
So, at the maximum,
2Kc 1 1 2
Kp = 2Kc∗ = , and Z1 = ∗
= ∗ = (577)
1 + Kc ln(2 /1 ) Kp − Kc Kc Kp
c†lσ
P ik·l †
= √1 e ckσ (582)
N
k
P −ik·l
clσ = √1 e ckσ (583)
N
k
81
And therefore,
X † X 1 X
ei(k1 −k2 )·l c†
X †
clσ clσ = k1 σ ck2 σ = ckσ ckσ , (584)
l k1 k2
N l k
−2µ2B ρ(ε) ∂f
R
M (T, B) ∂ε dε χp (T )
χ0 (T ) = = = (587)
B U ∂f
1 − 2NUµ2 χp (T )
R
1 + N ρ(ε) ∂ε dε
B
82
The average of the z-component of the magnetic moment is given by,
1 1 2π 2π 1
Z Z Z
hµz i = d cos θ dϕµz eβµB cos θ = d(cos θ)µ cos θeβµB cos θ
Z −1 0 Z −1
Z 1
2π ∂
= d(cos θ)eβµB cos θ (592)
ZB ∂β −1
1 ∂ ln Z kB T 2 ∂ ln Z µB kB T
= =− = µ coth − = µL(µB/kB T )
B ∂β B ∂T kB T µB
where L(x) is the Langevin function,
1
L(x) = coth x − (593)
x
(b) For µB/kB T 1, making use of coth(x) ≈ 1/x + x/3 for |x| 1, we have the Curie’s law.
µ2 B C
hµz i = = (594)
3kB T T
µ2 B
where C = 3kB .
83
After solving the tight-binding model, we get the band structure as,
s
√ q √ ky a
a
−i 12 ( 3kx +3ky ) i a2 ( 3kx +ky ) −i k√x a i k√
xa
E(kx , ky ) = ±te 1 + eiky a + e e 3 + 2e 2 3 cos (599)
2
where vF = 3ta2 .
Recall that a massless Dirac fermion has dispersion relation E = pc. Here we have low-energy
effective Hamiltonian, Ĥ(k̃) = vF p̂, and therefore the low-energy carriers in graphene behave
like a 2D massless Dirac fermions.
(b) Use Cn = n1 a1 + n2 a2 to denote the lattice vector of a nanotube. From Bloch theorem,
ψk (r + Cn ) = eik·Cn ψk (r)
(601)
ψk (r + Cn ) = ψk (r)
which means,
k · Cn = 2mπ (602)
Take k = αb1 + βb2 , we have,
1
αn1 + βn2 = m ⇒ α = (m − βn2 ) (603)
n1
Note that,
2π 1
b1 = ( √ i + j)
a 3
(604)
2π 1
b2 = ( √ i − j)
a 3
and therefore,
1
k= (m − βn2 )b1 + βb2
n1
(605)
2π 1 n2 m n2 m
= (β( √ − √ )+ √ )i + (β(−1 − ) + )j
a 3 3n1 3n1 n1 n1
where β is continuous. To be explicit,
m
2π n2
ky = (−1 − )[β − n1 n2 ]
a n1 1 + n1
√m (606)
2π 1 n2 3n1
kx = (√ − √ )[β + ]
a 3 3n2 √1 − √n2
3 3n1
And therefore,
ky = γkx + mb (607)
√ 1/n1 1/n1
with γ = 3 nn21 −n
+n2
1
and b = 2π n2
a (1 + n1 )[ 1−n2 /n1 + 1+n2 /n1 ].
If n1 = n2 = n, then kx = √2π m, ky = − 4π [β − m ] defines
the cutting lines. If these cutting
3n a 2n
lines come across K± , then we will have a metal; otherwise, an insulator.
84
IV.28. Carrier properties of graphene.
4π
k̃x = kx −
3a (610)
k̃y = ky
then,
√
a 3 2π ak̃x 2π k̃x a
HAB (k) ≈ (−t)[1 + ik̃y √ ][1 + 2(1 − i ak̃y )(cos cos − sin sin )]
3 2 3 2 3 2
√ √
a 3a 3a (611)
= (−t)(1 + i √ k̃y )(− k̃x + i k̃y )
3 2 2
∂ ∂
= (−i)~v0 ( −i )
∂x ∂y
where we have used, √
3
~v0 = ta
2
∂
k̃x = −i (612)
∂x
∂
k̃y = −i
∂y
Take zero onsite energy, HAA = HBB = 0. And therefore the total Hamiltonian is given by,
∂ ∂
!
0 ∂x − i ∂y ∂ ∂
H = −i~v0 ∂ ∂ = −i~v0 (σx + σy ) (613)
∂x + i ∂y 0 ∂x ∂y
(b) Take, !
1 1
|ski = √ iθ eik·r , s = ±1 (614)
2 se k
Then,
∂ ∂
! !
0 − i ∂y 1 1
H|ski = −i~v0 ∂ ∂
∂x √ iθ eik·r
∂x + i ∂y 0 2 se k
! (615)
i~v0 s(cos θk + i sin θk )(ikx + ky ) ik·r
=−√ e
2 ikx − ky
85
kx ky
We have cos θk = k , sin θk = k .
In this way, !
1 s
H|ski = ~v0 k √ iθ eik·r (616)
2 e k
• s = +1 ⇒ H| + ki = ~v0 k| + ki.
• s = −1 ⇒ H| − ki = −~v0 k| − ki.
|ski is an eigenstate of the Hamiltonian H, and that the energy eigenvalue corresponding to
this eigenstate is given by,
sk = s~v0 k (617)
(c)
Z k
X 2A
N =2 1= 2πkdk
k
(2π)2 0
A (618)
⇒ N (k)dk = kdk
π
A
⇒ N ()d = 2 2 d
~ v0 π
Consider that there are two Dirac cones in BZ, K± . We double the density of states and get,
2A
Ñ ()d = d (619)
~2 v02 π
(d) Consider a weak short-range potential V (r) = V0 δ(r). We know that the scattering amplitude
is proportional to transition matrix elements,
Calculate the matrix element of V (r) between initial state |ski and final state |sk0 i,
V0 X
Z
0 0 ik00 ·r 00 0
dr(1 + e−i(θk0 −θk ) )ei(k +k−k )·r
X
hsk |V (r)|ski = V0 hsk | e |ski =
k00
2 k00
(621)
V0
= [1 + e−i(θk0 −θk ) ]
2
For chiral electron, the scattering amplitude is angular-dependent.
On the other hand, for free electrons,
00 ·r
hk0 |V (r)|ki = V0 hk0 |
X
eik |ki = V0 (622)
k00
86