R?ssler U. Solid State Theory: An Introduction

Подождите немного. Документ загружается.

8.6 Cooper Pairs and the Gap 259

V ≠0

eff

∆

occupied

empty

Fig. 8.11. Mean ﬁeld (uppe r part) and eigenstates (lower part) in dependence on

the single-particle energy around E

with eﬀective interaction. Note the opening of

a gap at E

for which |H

| depends linea rly on the energy E(k), which is the energy dif-

ference of the single particle energy and the Fermi energy. This situation is

showninFig.

8.10 (upper part). The ﬁeld points in the positive or negative

z direction depending on the sign of E(k). The corresponding eigenstates are

pseudo-spin up (down) or occupied (empty) pair states below (above) E

, i.e.,

for negative (p o sitive) E(k) (see lower part of Fig.

8.10).

This picture changes for ﬁnite V

eﬀ

(Fig.

8.11). Now, at E

or E(k)=0

the ﬁctitious ﬁeld is ﬁnite, pointing somewhere in the xy plane. Its strength

depends on the expectation values of the pseudo-spin operators σ

′

and σ

′

But, away from the Fermi energy, this ﬁeld is turned into the z direction and

approaches the linear dependence as for the noninteracting case (upper part

of Fig.

8.11). The corresponding eigenstates are pseudo-spin vectors alig n ed

to this ﬁeld and turn around with H

. Their projection onto the z direction

is shown in the lower part o f Fig.

8.11.

In order to quantify the discussion, let us assume σ

′

 = 0, i.e., all

pseudo-spins together with the ﬁctitious ﬁelds are in the xz plane, and intro-

duce the energy parameter ∆ = V

eﬀ



′

σ

′

/2. The energy for exciting an

electron pair out of the correlated ground state is given by twice the modulus

of the ﬁeld

| = {E(k)

+∆

}

1/2

, (8.121)

where ∆ is the gap energy at E

(see Fig.

8.11). Due to the alignment of

the pseudo-spin along the ﬁeld H

(Fig.

8.12), the ratios of their x and z

components are equal and can be expressed by the angle θ

. As, in addition,

σ

′

 =sinθ

′

we may write

260 8 Electron–Phonon Interaction

< σ >

Fig. 8.12. Alignment of pseudo-spin and ﬁctitious ﬁeld in the xz plane

tan θ

eﬀ



′

sin θ

′

2E(k)

∆

E(k)

. (8.122)

This relation can be solved for ∆ by replacing

sin θ

′

∆

(∆

+ E

′

))

1/2

(8.123)

to obtain

∆=

eﬀ



′

∆

(∆

+ E

(k))

1/2

. (8.124)

The summation, limited to the range of the attractive interaction around the

Fermi surface, can again be performed as an integral, giving

eﬀ

D(E

)



+¯hω

−¯hω

(∆

+ E

(k))

1/2

= V

eﬀ

D(E

)sinh

−1

¯hω

∆

. (8.125)

This is easily solved to yield, as the BCS solution, for the gap parameter

∆=

¯hω

sinh(1/V

eﬀ

D(E

))

≃ 2¯hω

−1/V

eff

D(E

)

, (8.126)

where the last expression is obtained for 1 ≫ V

eﬀ

D(E

). According to this

formula, the gap energy (or the stability of the Cooper pairs) increases

with the cut-oﬀ phonon energy (or the Debye frequency) and the density

of single-particle states at the Fermi energy.

In order to ﬁnd an expression for the critical temperature, we have to ask

for the temperature dependence of the gap, which enters through the thermal

expectation values of the pseudo-spins

σ



=Tr

−β

=tanh

(8.127)

8.6 Cooper Pairs and the Gap 261

and (using again Fig.

8.12) deﬁnes ∆(T )bytherelation

tan θ

eﬀ



′

sin θ

′

2E(k)

tanh

′

∆(T )

E(k)

. (8.128)

Replacing sin θ

′

=∆(T )/{E

(k)+∆

(T )}

1/2

gives

eﬀ



′

tanh

(k)+∆

(T )}

−1/2

=1. (8.129)

This equation can be solved for ∆(T ). For T = T

the gap vanishes and

| = E(k), which simpliﬁes the relation to

eﬀ



′

E(k)

tanh

E(k)

=1. (8.130)

Extending this result, which is valid for the adopted spin model related to

pair states, to include single particle excitations, T

has to b e replaced by

. The factor 2 accounts for the doubling of the entrop y due to a doubling

of the possible excitations (see [

4]). The sum over k is again performed as an

energy integral, giving the BCS result for the critical temperature

eﬀ

D(E

)



¯hω

/2k

tanh(x)

dx. (8.131)

For T

≪ Θ

,whereΘ

is the Debye temperature, this can be evaluated to

yield

=1.14

−1/V

eff

D(E

)

. (8.132)

This result immediately explains the isoto pe eﬀect, because the Debye tem-

perature is inversely proportional to the square root of the ion mass. With the

help of (

8.126), we may also write the direct relation between the gap energy

and the critical temperature

2∆ = 3.52k

. (8.133)

The va li d i ty of this relation is demonstrated in Table

8.1 for some representa-

tive metals.

The principal correctness of the mean-ﬁeld approximation is shown in

Fig.

8.13 by plotting the reduced value of the gap ∆(T )/∆(0) as the order

parameter for diﬀerent superconductors versus the reduced temperature T/T

The experimental data for diﬀerent materials deﬁne a universal curve which

is well described by the graph from BCS theory. Note the similarity with

Fig.

6.10.

Similar to the Fr¨ohlich polaron, a Coop er pair is a comp osite particle.

However, the statistical properties of the polaron are those of the electron

262 8 Electron–Phonon Interaction

Table 8.1. Debye temperature Θ

, critical temperature T

,andratio2∆(0)/k

for diﬀerent metals, from [

31].

Metal Θ

[K] T

[K]2∆(0)/k

An 235 0.9 3.2

Cd 164 0.56 3.2

Hg 70 4.16 4.6

Al 375 1.2 3.4

Sn 195 3.75 3.5

Pb 96 7.22 4.3

1.0

0.8

0.6

0.4

0.2

0.2 0.4 0.6 0.8 1.0

∆(T) / ∆(0)

T/T

Fig. 8.13. Dependence of the reduced gap energy on the reduced temperature for

diﬀerent superconductors (symbols)andtheBCSresult(solid line) (after [30])

because the phonon cloud does not change the spin, whereas the Cooper pair,

which is a particle with integer spin composed of two electrons, is a boson.

Due to their bosonic nature, a system of Cooper pairs can undergo Bose–

Einstein condensation to reach the superconducting state. Depending on the

total spin of the two electrons in a Cooper pair, one distinguishes singlet and

triplet superconductors.

Bose–Einstein condensation is possible not only for paired electrons in a

solid. By cooling atoms conﬁned in electromagnetic traps, degenerate Fermi

gases can be realized (see Chap.

4), which consist of atoms. For these Fermi

systems, one has, recently, also studied the formation of molecules due to

inter-atomic forces. A system of such bosonic molecules can make a phase

transformation into the supraﬂuid state corresponding to the superconducting

state of the electrons [

258, 259].

Problems 263

Problems

8.1 Derive the expression fo r the electron–phonon interaction in a Bravais

lattice without ma king use of the single-band approximation. In this case

the operator form should be



†

(−q)+a

(q)



†

′

. (8.134)

Derive a n expression for the interaction matrix element, which leads to a

relation between k, k

′

and q.

8.2 Express the divergence of the continuous lattice displacement ﬁeld u(r)

by the components of the strain tensor ﬁeld ǫ(r) and realize that only

longitudinal phonons contribute.

8.3 Show that charge carriers in an energy ba nd deriving from p states can

interact via deformation potential coupling with b oth longitudinal and

transverse phonons.

8.4 Derive the expression for the piezoelectric electron–phonon coupling (

8.37)

by performing the same steps as outlined for the polar optical electron–

phonon coupling. Which acoustic branches contribute to the coupling?

8.5 Given the scattering rates 1/τ of an electron with energy ǫ due to defor-

mation potential (∼Tǫ

1/2

) and piezoelectric c oupl ing (∼Tǫ

−1/2

)forlarge

T compared with the phonon energies, calculate the temperature depen-

dence of the contributions of both scattering processes to the mobility of

electrons. Compare with Fig.

8.5.

8.6 Calculate the number of phonons ¯n in a composite particle due to defor-

mation potential coupling with acoustic p honons by using ﬁrst order

perturbation theory. Note that the q dependence of this coupling diﬀers

from that of the Fr¨ohlich coupling and that the q integration changes.

Discuss the r esult in dependence on the material parameters. Estimate ¯n

if the carrier would be an ion (e.g., a proton) moving in the lattice.

8.7 Evaluate the commutator between the operators H

el−ph

and S.Discuss

the approximation made in writing the expression given in the text.

8.8 Calculate the ground state energy of the BCS Hamiltonian (

8.120)relative

to the normal ground state. Note that only electrons close to E

are

relevant.

Defects, Disorder, and Localization

The crystalline o rder of a solid, characterized by the conﬁguration {R

} of all

atoms with their equilibrium positions at lattice points, has been anticipated

throughout this book so far. The essential consequence of the lattice period-

icity, deriving from this conﬁguration, is that all states be it of phonons, of

electrons, or of magnons follow Bloch’s theorem with a modulus that rep eats

with the lattice period, e.g.,

|ψ

(r + R

)| = |ψ

(r)| , for all R

. (9.1)

Because of this property, Bloch states are extended states. In this chapter, devi-

ations from the crystalline order and their consequences shall be considered for

the electrons (but they are essentially valid also for the collective excitations

of the solid). This is done ﬁrst for individual point defects or impurities,which

have to be classiﬁed, and can lead to discrete bound states localized in the

vicinity of the defect. They are described by wave functions that decay (expo-

nentially) away from the defect. A more general deviation from the crystalline

structure is c ompositional or structural disor d er, which will be exp erienced by

the electrons as a random potential. The energy spectrum of the electrons

can s till exhibit bands with continuous density of states but with a modi-

ﬁcation of the electronic wave functions depending on their energy: Toward

the band edges, the electrons are more strongly inﬂuenced by the potential

ﬂuctuations and get conﬁned in deep local minima. Such localized states have

the characteristic form

ψ(r) ∼ e

−|r|/λ

, (9.2)

where λ is the lo calization length. Thus, disorder causes localization and

becomes resp onsible for dramatic changes in the transport properties. They

show up at low temperatures at which – as one of the prominent features – a

metal–insulator transition (MIT) can take place due to disorder. The experi-

mental investigations of the MIT and the conceptual theoretical work carried

out to understand this phenomenon belong to the outstanding achievements

made in solid state physics during the last decades.[

275–279]

U. R¨ossler, Solid State Theory,

DOI 10.1007/978-3-540-92762-4

 Springer-Verlag Berlin Heidelberg 2009

266 9 Defects, Disorder, and Localization

Fig. 9.1. Representation of point defects in a binary solid (open and ﬁlled symbols

represent two diﬀerent atomic species)

9.1 Point Defects

Deviations from the periodic structure can be classiﬁed according to their

dimension. Point defects represent zero-dimensional perturbations of the peri-

odic lattice structure. Diﬀerent cases are depicted in Fig.

9.1 for a solid

containing two sublattices with atoms A and B, respectively. In a substi-

tutional impurity C

(or C

), an atom C diﬀering from those of the host

material A (or B) replaces the host atom at a lattice site. An interstitial

impurity I means an atom (of the host material, but also diﬀerent from it)

taking a position in some free space between the lattice points. A vacancy V

(or V

) refers to an empty lattice point. A Frenkel

defect, V

− I

,isthe

combination of a vacancy with a nearby interstitial. It can be created by dis-

placing a host atom from its lattice site to the interstitial position. In binary

solids AB, especially in compound semiconductors, an anti-site defect A

possible with an atom A at the regular site of a B atom. This anti-site defect

is more likely, the closer A and B are in the periodic table.

In semiconductor s, isolated impurities can change the electronic spectrum

by causing states in the fundamental energy gap from which, at elevated tem-

perature carriers a re released into the nearby band edges. This is the concept

of doping which is of great importance in device applicatio n s. Therefore, the

theory of impurities within the eﬀective-mass approximation originates from

the early days of semiconductor physics [

260], but was reﬁned and extended

following the progress of research [

261–268].

For an isolated substitutional impurity in an otherwise crystalline solid,

we may write

Yakov Ilyich Frenkel 1894–1954

9.1 Point Defects 267

H = −

¯h

∆ + V

eﬀ

(r)+U(r) (9.3)

with the periodic eﬀective potential V

eﬀ

(r)=



v(r − R

) introduced in

Sect.

5.1 and the impurity potential U(r)=v

(r) − v(r), where the lattice

site of the substitution is taken as the origin. Without U(r), the eigenvalue

problem of H leads to band structure E

(k) and Bloch functions ψ

(r)which

represent the solution of the unperturbed problem. Including the impurity

potential, the eigenvalue problem

Hφ = Eφ (9.4)

can be solved by expanding φ(r)=



(k)ψ

(r)asawavepacketof

Bloch waves. The variational principle for the expectation value of the energy

leads to the set of coupled linear equa tions



(k) − E



(k)+



′

nk|U |n

′

f

′

) = 0 (9.5)

for the expansion coeﬃcients f

(k). Using the Fourier transform U(r)=



exp(iq·r) and expanding the product of the periodic parts of the Bloch

functions in a Fourier series, u

∗

(r)u

′

(r)=



nkn

′

(G)exp(−iG · r),

the matrix element of the impurity potential U (r) can be expressed as

nk|U |n

′

 =



nkn

′

(G)δ

q,k−k

′

. (9.6)

For G = 0, one has

nkn

′

(0)=



∗

(r)u

′

(r) ≃ δ

′

r, (9.7)

(which is exact for k = k

′

)andwrites



(k) − E



(k)+



′

k−k

′

)



′



G=0

k−k

′

nkn

′

(G)f

′

)=0. (9.8)

The last term in this equation, describing the interband coupling due to the

impurity potential, can be neglected if for G = 0,therelation

k−k

′

k−k

′

nkn

′

(G) ≪ 1 (9.9)

holds, which leads to the one-band approximation. Taking the Fourier trans-

form with f

(k)=

exp(−ik · r)f

(r)d

r/V and by writing

(k →

∇) ≃ E

(0) −

¯h

∗

∆, (9.10)

268 9 Defects, Disorder, and Localization

we ﬁnd the eﬀective-mass equation for shallow impurities in a simple band

[

260]



−

¯h

∗

∆ + U(r)



(r)=(E − E

(0))f

(r). (9.11)

The condition (

9.9) is fulﬁlled for |k −k

′

|≪|G| and applies if the expan-

sion coeﬃcients f

(k) diﬀer from zero only for small k in the vicinity of the

band edge. This is the case for charged impurities as, e.g., a substitutional

Si (or Ge) at a Ga site in GaAs, for which the impurity potential is essen-

tially a screened Coulomb potential with a short-range correction term U

(central-cel l correction) to account for the chemical nature of the impurity

U(r)=U

(r) −

4πε

εr

. (9.12)

This leads to a modiﬁed hydrogen problem with bound states below the band

minimum. The characteristic units of energy and length, the eﬀective Rydb erg

constant Ry

∗

=(m

∗

/mε

)Ry, and the Bohr radius a

∗

=(εm/m

∗

, resp ec-

tively, scale according to the material parameters from the atomic values. For

GaAs, one has Ry

∗

≃ 10

−3

Ry ≃ 5meVanda

∗

≃ 100 a

≃ 10 nm. The bound

states for hydrogen-like impurities can be classiﬁed by the angular momentum

quantum numbers. Only the 1s state is essentially modiﬁed by the central-cell

correction. Figure

9.2 draws attention to these bound states at the band edge

in k and in real space. In k space, the 1s wa ve function is represented by its

Fourier transform f

(k) which extends over a width of the inverse eﬀective

Bohr radius around the band minimum. In real space, it is a wave function

(r) extending over a distance of about the eﬀective Bohr radius around

the impurity site. The real space localization of the state increases with the

strength of the impurity potential. Shallow acceptor states have a more com-

plex structure, which derives from the eﬀective-mass Hamiltonian of p-like

valence band s [

261].

In contrast to (hydrogen-like) shallow impurities, whose spectrum is dom-

inated by the long-range Coulomb interaction with only minor modiﬁcations

conduction band

f (k)

f (r)

energy

Fig. 9.2. Impurity states at a band edge in k space (left) and in real space (right)

9.1 Point Defects 269

due to the central-cell correction (especially for the ground state), the situa-

tion is completely reversed for deep impurities. Here, the energy spectrum is

determined by the short-range central-cell potential, and the inﬂuence of the

long-range Coulomb potential (if present at all) is considered as a correction.

Consequently, the deep impurity states are strongly localized to the neighbor-

hood of the impurity site accompanied by lattice distortions. This situation

cannot be well described by a sup erposition of extended Bloch states. Here

it is more appropriate to use atomic orbitals or concepts of scattering theory

with a localized basis.

Starting from the eigenvalue problem of (

9.3) with the impurity potential

U(r) now dominated by the central-cell correction, we use the Ansatz for the

wave function

ψ(r)=



α,R

(R)φ

(r − R) (9.13)

with atomic orbitals φ

(r −R) localized at R. While in Sect.

5.4,whenintro-

ducing the LCAO method, we have constructed Bloch functions out of the

atomic orbitals, we now remain in the localized representation. The variational

principle for the energy leads to the set of coupled linear equations



′

αR,α

′

− ES

αR,α

′

) = 0 (9.14)

for the expansion coeﬃcients c

(R) and energy eigenvalues following from the

secular problem

H

αR,α

′

− ES

αR,α

′

 =0. (9.15)

Here H

αR,α

′

and S

αR,α

′

are the matrix elements of the Hamiltonian and

of the overlap between atomic orbitals (see Sect.

5.4), respectively.

Some aspects of deep impurities, in particular the chemical trends, can be

understood in this model as exempliﬁed here for an isoelectronic impurity,

e.g., GaP:N (meaning that P in GaP is substituted by N) [

261, 269]. In

the picture of atomic orbitals, the valence and conduction bands of intrin-

sic semiconductors with tetrahedral coordination are formed by the bonding

and anti-bonding states, respectively, of the s and p orbitals. This is sketched

for the host atoms in the left hand side of Fig.

9.3, where the shaded regions

indicate the continua of the energy bands. Replacing one host atom P by the

impurity N leads to diﬀerent pairs of s and p bonding and anti-bonding states

localized at the impurity site (right hand side of Fig.

9.3): in particular, the

s-bonding state becomes a resonant impurity level deep in the valence band,

while the s-anti-bonding state o f the Ga-N pair is lower e d with respect to that

of the Ga-P pair of the host crystal, thus forming a deep trap in the energy

gap. Corresponding impurity states (not shown in the ﬁgure) are formed out

of the p orbitals.

Chemical trends can be discussed by simulating a continuous change of the

impurity (X) with a change of its atomic level energy E

. For the bonding