Sunteți pe pagina 1din 21

1

Foundations

Solutions to Exercises
Q1.1 Estimate the wavelengths (using = 2/k) associated with electrons in
typical solids. Some examples are Al (E = 11.7 eV; use the mass m 0 for
free electrons), highly doped n-GaAs (E = 50 meV and m = 0.067 m 0 ) and
a two-dimensional electron gas in GaAs (E = 10 meV). These lengths are
important because they set the scale for structures that must be fabricated
if we wish to manipulate the waves. The wavelength of electrons in an
electron microscope (E = 100 keV) sets the ultimate limit of resolution,
although aberrations of the lenses reduce this drastically in practice. Finally,
as a (very) classical object, consider the Clapham omnibus.

Write = 2/k = h/ p = h/mv = h/ 2m E where m, v and E are the mass,
velocity, and kinetic energy of the particle.
A free electron in Al with E = 11.7 eV and m = m 0 has wavelength = 0.36 nm,
comparable with the size of an atom.
An electron in highly doped n-GaAs with E = 50 meV and m = 0.067 m 0
has wavelength = 21 nm, which is much larger than the separation between
atoms. This makes it practicable to fabricate structures in GaAs on the scale of
the electronic wavelength, at least in one of the three spatial dimensions.
An electron in a two-dimensional electron gas in GaAs with E = 10 meV has a
yet longer wavelength of = 47 nm. This makes the system even more suitable
for observing the guidance and interference of waves.
Electrons in an electron microscope with E = 100 keV have wavelength =
0.004 nm. This is much smaller than the interatomic spacing, confirming that it
should be possible to resolve individual atoms with a sufficiently good lens.
Taking the mass of the Clapham omnibus to be 10 tonnes and its speed to be
50 km h1 (about 30 mph) gives a wavelength = 51039 m. This is clearly of
no practical significance and the driver need not be trained in quantum mechanics.

1
2 Foundations

Table Q1.1. Energies in meV of the first few bound states in an infinitely deep well, showing
the influence of the width a and effective mass m eff .

n electron heavy hole light hole


m eff = m e = 0.067 m hh = 0.5 m lh = 0.082
a= 10 nm 4 nm 10 nm 4 nm 10 nm 4 nm
1 56 351 8 47 46 287
2 225 30 188 184
3 506 68 423 413
4 120

Q1.2 Use the model of an infinitely deep potential well to estimate the first
few energy levels for an electron in GaAs in wells of width 10 nm and
4 nm. Remember the mass m in the equations should be replaced by m e m 0
where m e = 0.067 is the effective mass for electrons at the bottom of the
conduction band of GaAs. How accurate do you expect the results to be,
given that real wells in GaAs/AlGaAs are about 0.3 eV deep?
Repeat the calculation for holes, whose well is about 0.2 eV deep. These
come in two varieties, heavy and light, with effective masses of m hh = 0.5
and m lh = 0.082.
All we need for this is equation (1.12), which gives the energies of the bound states
for a particle of mass m in an infinitely deep well of width a.
h 2 2n 2
n = , (Q1.1)
2ma 2
with m = m 0 m eff and the appropriate effective mass m eff . The first few energy
levels for each case are listed in table Q1.1.
The infinitely deep well might give a reasonable approximation for the energy
of an electron in the lowest state of a 10 nm well, because this energy level is far
below the top of the real, finite well. It is doubtful for the next state, whose 225 meV
energy is not far below the 300 meV depth of the real well. The approximation is
useless for the 4 nm well because it predicts that the energy of the lowest state is
greater than the depth of the real well, and we shall see in section 4.2 that a one-
dimensional square well always has at least one bound state. Its validity is similar
for light holes, but the greater mass of heavy holes reduces their energy levels so
that the infinitely deep well works reasonably well even for the narrower 4 nm well.

Q1.3 Figure 1.4 shows experimental measurements of the energy levels in quan-
tum wells. The sample has four quantum wells of GaAs, of different thick-
ness a, grown in a thick layer of Al0.35Ga0.65As. The traces show the
Foundations 3

Table Q1.2. Observed and estimated energies for photoluminescence in a sample with four
quantum wells of different width.

experiment theory
wavelength energy assignment energy
peak (nm) (eV) (eV)
(a) 787 1.58 9 nm well 1.60
(b) 769 1.61 6 nm well 1.70
(c) 747 1.66 4 nm well 1.92
(d) 709 1.75 2 nm well 3.11
(e) 668 1.86 Al0.35Ga0.65As 1.96

intensity of the photoluminescence (PL) taken at a temperature of 12 K,


described in section 1.3.1. The energy of photons emitted is the difference
of the energies of the electron and hole in one of the wells. The narrower
wells have bound states of higher energy and therefore emit light of higher
frequency (shorter wavelength), giving rise to the set of four lines shown
on the PL trace. Given the thicknesses of the four wells, how well does
the particle-in-a-box model predict the energy of the PL? Assume that the
electrons and holes have effective mass m e 0.067 and m hh = 0.5 (heavy
holes). What might be responsible for peak (e)?

There are six possible regions that might generate photoluminescence. In order of
increasing energy, they are:

broad regions of GaAs, without quantum confinement, giving h = E gGaAs ;


the four quantum wells, each with a different energy;
the AlGaAs barriers, giving h = E gAlGaAs .

Start with GaAs, which is almost always seen in experiments because it provides
the substrate. Assume that the temperature of 12 K is low enough that we can use
the bandgap of GaAs at absolute zero, which is E gGaAs = 1.52 eV. It is important
to use the correct temperature because E gGaAs changes by 0.10 eV between 0 K
and 300 K. The wavelength for an optical transition with this energy is given by
= c/ = hc/E gGaAs = 815 nm. This wavelength is too long to be included in the
scan in figure 1.4 and the transition is therefore not seen.
Before looking at the quantum wells, check the energy of transitions in the
Al0.35Ga0.65As barriers. Here E gAl0.35 Ga0.65 As = 1.96 eV, using the data in appendix
3, giving = 634 nm. The peak at shortest wavelenth has = 668 nm giving
h = 1.86 eV. This is reasonably close; the smaller energy observed in the exper-
iment might be due to excitons (section 10.7) or defects.
4 Foundations

Next consider the quantum wells. The energy of the optical transition between
the lowest states for electrons and holes in infinitely deep wells is given by equation
(1.15),
h 2 2 h 2 2 h 2 2
h = E gGaAs + + = E GaAs
g + . (Q1.2)
2m 0 m e a 2 2m 0 m hh a 2 2m 0 m eh a 2
The optical effective mass, given by 1/m eh = 1/m e + 1/m hh , is m eh = 0.059 here,
dominated by the light mass of the electron.
Predictions using this equation are compared with the experimental results in
table Q1.2. The energy estimated for the 9 nm well agrees well with peak (a),
and peak (b) is not too far away. Unfortunately the approximation of an infinitely
deep well is poor for the 4 nm well and hopeless for the 2 nm well, because we
know that the energy of the transition between the bound states cannot exceed
E gAl0.35 Ga0.65 As = 1.96 eV. A better calculation will follow in section 4.2.

Q1.4 The current and charge densities derived in section 1.4 are guaranteed to
conserve charge provided that V (x , t) is real. Show that particles will be
created or absorbed if V is complex. This trick is sometimes used for a
source or sink of particles in modelling.

The current and charge densities were deduced from the pair of equations (1.26)
and (1.27),
h 2 2
9 9 + 9 V 9 = i h 9 9, (Q1.3)
2m x 2 t
h 2 2
9 2 9 + 9V 9 = i h 9 9 . (Q1.4)
2m x t
Proceed as before by subtracting these equations, but allow V (x ) to be a com-
plex function. This leads to an additional term on the left of (V V )|9|2 =
2i|9|2 Im V . The other terms can be manipulated as before, to give
  
h q 2q
9 9 9 9 = (q|9|2 ) Im V |9|2 . (Q1.5)
x 2im x x t h
If we define the current and charge densities in the usual way, this becomes
J 2
+ = Im V . (Q1.6)
x t h
The usual result has zero on the right.
Consider a translationally invariant system to reveal the effect of the extra term.
In this case J / x = 0 and /t = (2/h ) Im V . Thus a positive imaginary
potential causes the charge density to grow, and such a potential behaves as a
Foundations 5

source of particles. Likewise, a potential with a negative imaginary part acts as a


sink that absorbs particles.
A quick way to verify the sign is to add an imaginary part to the energy of a
stationary state and consider its dependence on time, exp(i Et/h ). Then
   
i Et E + Re V + i Im V
exp exp i t
h h
   
E + Re V Im V
= exp i t exp t . (Q1.7)
h h
The real exponential causes the wavefunction to grow if Im V > 0 and decay if
Im V < 0.
Q1.5 Construct a wavefunction to describe a uniform beam of free electrons mov-
ing along +x at 105 m s1 and carrying a current density of 1 A mm 2.
Repeat this for a current density of 104 A m2 carried by electrons in
GaAs (m = 0.067m 0 ) travelling at 105 m s1.
A uniform beam of electrons may be described by a single plane wave, 9(x , t) =
A exp[i(kx t)]. It was shown in section 1.4 that this carries a current density J =
qv|A|2 with velocity v = h k/m and charge q = e. Substituting v = 105 m s1
and J = 1 A mm 2 = 106 A m2 gives |A|2 = 6.2 1013 m3 . This is the
density of electrons per unit volume, and has the appropriate dimensions. Thus
|A| = 7.9 106 m3/2, which sets the dimensions of the wavefunction; the phase
of A cannot be determined because it has no effect on the current. We also need k =
mv/h = 8.6 108 m1 = 0.86 nm1 and = E/h = 12 mv 2 /h = 4.3 1013 s1.
The numbers for GaAs lead to n = |A|2 = 6.2 1017 m3 , k = 0.058 nm1 and
= 2.9 1012 s1.
Q1.6 Derive the current density for counter-propagating travelling and decaying
waves, equations (1.34) and (1.36).
The current density is given generally by equation (1.32),
 
h q
J (x , t) = 9 9 9 9 . (Q1.8)
2im x x
The counter-propagating travelling waves have
9(x , t) = [A+ exp(ikx ) + A exp(ikx )] exp(it), (Q1.9)
and we just need to substitute this into the expression for the current, taking care
with the complex conjugate. For example,
9
9 = ik[|A+ |2 |A |2 + A+ A exp(2ikx ) A A+ exp(2ikx )]. (Q1.10)
x
6 Foundations

The dependence on time cancels between 9 and 9 . Subtracting the second term
in the current causes the spatial dependence to cancel too, and we are left with
h k
J =q (|A+ |2 |A |2 ) = qv(|A+ |2 |A |2 ). (Q1.11)
m
This is just the difference of the current carried by the two waves travelling in
opposite directions.
The calculation for the counter-propagating decaying waves,
9(x , t) = [B+ exp( x ) + B exp( x )] exp(it), (Q1.12)
proceeds in the same way. The difference is that the complex conjugation has no
effect on the spatial exponential functions, so
9
9 = [|B+ |2 exp(2 x ) |B |2 exp(2 x ) + B+ B B B+ ]. (Q1.13)
x
The difference gives
h h
J =q (B+ B B+ B ) = q 2 Im(B+ B ). (Q1.14)
im m
Neither component would carry a current by itself because the wavefunction is real
(or can be transformed to be real), but the superposition carries current if there is a
difference in phase between the components in the two directions.
Q1.7 Show that the root-mean-square width 1x of an electron in state n of an
infinitely deep square well of width a is given by
 
a2 6
(1x )2 = 1 2 2 . (E1.1)
12 n
Classically the particle is equally likely
to be found anywhere in the well
(unless it is at rest). Show that 1x = a/ 12 in this case, and note that the
quantum mechanical result approaches this value for high energies (high
n).
Take the well to lie in the range 0 < x < a. The root-mean-square width 1x is
given by equation (1.47), (1x )2 = hx 2i hx i2. The tedious part is the evaluation
of the integrals for the two expectation values.
Z Z
2 a n x
hx i = n (x ) x n (x ) dx = x sin2 dx
a 0 a
Z
2a
= sin2 n d
2 0
 
a 2 2n sin 2n + cos 2n
=
2 2 4n 2 0
Foundations 7

= 1
2
a. (Q1.15)
This is of course obvious by symmetry.
Z Z
2 a 2 2 n x
hx 2i = n (x ) x 2 n (x ) dx = x sin dx
a 0 a
Z
2a 2 2 2
= sin n d
3 0
 
a 2 3 (2n 2 2 1) sin 2n + 2n cos 2n
=
3 3 4n 3
  0
1 1
= a2. (Q1.16)
3 2n 2 2
Combining these gives the desired result,
 
a2 6
(1x )2 = hx 2i hx i2 = 1 2 2 . (Q1.17)
12 n
A classical particle moves with constant velocity between the two barriers and is
therefore equally likely to be found at any point between them. This can be described
by a probability density function P(x ) = 1/a. The corresponding expectation
values are
Z Z
1 a
hx i = x P(x ) dx = x dx = 12 a, (Q1.18)
a 0
Z Z
1 a 2
hx 2i = x 2 P(x ) dx = x dx = 13 a 2, (Q1.19)
a 0
(1x )2 = hx 2i hx i2 = 12 a .
1 2
(Q1.20)
The difference between the classical and quantum-mechanical results for the ex-
pectation values vanishes as the energy of the particle increases, a general result
known as the correspondence principle. The minimum spread 1x occurs occurs
in the ground state, n = 1, when the quantum-mechanical density is least like the
classical distribution.
Q1.8 Consider the propagation of Gaussian wavepackets of initial width d =
25 nm and d = 50 nm, that represent electrons in GaAs (m = 0.067m 0 ).
Sketch 1x (t) for the two packets. How long does it take before they double
in length?
The width of a Gaussian wavepacket as a function of time is given by equation
(1.66),
s s
 2  2

h t t
1x (t) = d +
2 =d 1+ . (Q1.21)
2md
8 Foundations
200

1x / nm

100

25 nm
initial width d =
50 nm

0
0 2 4 6 8 10
t / ps

Figure Q1.1. Spread as a function of time of two Gaussian wavepackets representing elec-
trons in GaAs, with initial widths of 25 nm and 50 nm.

Here d is the initial width of the packet, which spreads on a timescale = 2md 2 /h .
This timescale is = 0.72 psand 2.9 ps for d = 25 nm and 50 nm. The wavepackets
double in length when t = 3 , which is 1.3 ps and 5.0 ps respectively.
Plots of 1x (t) are shown in figure Q1.1. The wavepacket with the smaller initial
width broadens more rapidly as a function of time and soon becomes longer than the
other wavepacket. The wavepackets broaden linearly for large times, as expected
when they are dominated by the range of velocities within them.
Q1.9 Using [x, p] = i h , show that [x , p 2 ] = 2i h p.
(Hint: add and subtract
For the common form of Hamiltonian H = p 2 /2m + V (x),
p x p.) show
that [x , H ] = (i h /m) p.
This result will be useful when treating optical
response in chapter 8.
Writing out the commutator, taking care to preserve the order of the operators, gives
[x , p 2 ] = x p p p p x
= x p p p x p + p x p p p x
= (x p p x)
p + p(
x p p x)

= [x , p]
p + p[
x , p]
= 2i h p.
(Q1.22)
The potential in the Hamiltonian H = p 2 /2m + V (x ) obviously commutes with x,

so we find
 
p 2 1 i h
H ] = x,
[x, = [x , p 2 ] =
p. (Q1.23)
2m 2m m
Foundations 9

Q1.10 Orbital angular momentum is defined in classical mechanics by L = r p.


Similarly, in quantum mechanics, the operator L x = y p z z p y and so
on. Show that different components of L do not commute with each other;
instead [ L x , L y ] = i h L z , which shows that no two components of the
angular momentum can be measured simultaneously.
The total angular momentum is defined by L 2 = L 2x + L 2y + L 2y . Show
that [ L x , L 2 ] = 0, which means that it is possible to specify both the total
angular momentum and any one component at the same time.

Expand the expressions that appear in the commutator, and examine each term to
bring out factors that commute. The only pairs that do not commute are those with
a position and momentum operator for the same coordinate; others can be freely
re-ordered. Thus

[ L x , L y ] = [ y p z z p y , z p x x p z ]
= [ y p z , z p x ] [ y p z , x p z ] [z p y , z p x ] + [z p y , x p z ]
= y p x [ p z , z ] 0 0 + x p y [z , p z ]
= i h (x p y y p x ) = i h L z . (Q1.24)

The calculation of [ L x , L 2 ] = [ L x , L 2x ] + [ L x , L 2y ] + [ L x , L 2z ] is very similar to


the previous exercise, and the square of the operator is treated in the same way.
The first of the three terms is trivial, [ L x , L 2x ] = 0. The next give the following
commutators,

[ L x , L 2y ] = L x L 2y L 2y L x
= L x L y L y L y L x L y + L y L x L y L y L y L x
= [ L x , L y ] L y + L y [ L x , L y ] = i h ( L z L y + L y L z ). (Q1.25)

Likewise

[ L x , L 2z ] = [ L x , L z ] L z + L z [ L x , L z ] = i h ( L y L z L z L y ). (Q1.26)

The sum of these gives [ L x , L 2 ] = 0 as required.

Q1.11 The eigenstates in a parabolic potential (section 4.3) can be written in the
form n (x ) Hn1 (x ) exp( 12 x 2) where Hn (x ) is a Hermite polynomial.
The first few of these are H0(x ) = 1, H1 (x ) = 2x , and H2(x ) = 4x 2 2.
Verify that these states are orthogonal (symmetry helps).

The eigenstates extend over all space, so we can use their symmetry about x = 0 to
eliminate unnecessary integrals. The Hermite polynomials are alternately even and
10 Foundations

odd, which shows immediately that 2 (x ) is orthogonal to both 1 (x ) and 3 (x ).


The remaining integral to verify is
Z Z
(4x 2 2)ex dx
2
3 (x ) 1(x ) dx


= 4 21 2 = 0. (Q1.27)

This employs the standard integrals


Z Z

et dt = , t 2 et dt =
2 2 1
2
. (Q1.28)

Q1.12 A triangular potential has V (x ) = eF x for x 0 and is impenetrable


for x < 0. The wavefunction 1 (x ) x exp( 12 bx ) is often used as
an approximation to its lowest eigenstate (section 7.5.2). Occasionally
an approximation is needed for the second state. Write it in the form
x (c bx ) exp( 12 bx ) and choose c to make it orthogonal to the lowest
state. Normalize both states.
The approximate wavefunctions contain a factor of x to ensure that they vanish at
the impenetrable wall at x = 0. Calculations using this wavefunction (known as
the FangHoward approximation; see section 9.3.3) usually reduce to the standard
integral
Z
t n et dt = n! 0(n + 1). (Q1.29)
0

Here we need
Z Z
0 = 2 (x ) 1(x ) dx x 2(c bx )ebx dx
Z
0
1 y 1
= y (c y)e
2
dy = (2c 6), (Q1.30)
b3 0 b3
whence c = 3.
For normalization we need the following integrals:
Z Z
2 bx 1 2 y 2
x e dx = 3 y e dy = 3, (Q1.31)
0 b 0 b
Z Z
2 bx 1 6
x (3 bx ) e
2
dx = 3 (9y 2 6y 3 + y 4 ) ey d y = 3 . (Q1.32)
0 b 0 b
The normalized, orthogonal wavefunctions are therefore
1 (x ) = ( 12 b)1/2bx exp( 12 bx ), (Q1.33)
2 (x ) = ( 16 b)1/2bx (3 bx ) exp( 12 bx ). (Q1.34)
Foundations 11

These approximate wavefunctions will be used to calculate scattering between the


subbands of a two-dimensional electron gas in chapter 9.

Q1.13 Derive the expectation value of x for the wavefunction (1.54), which con-
tains a mixture of two states in an infinitely deep quantum well that lies
between 0 and a. First, show that the coefficients A1 and A2 must obey
|A1 |2 + |A2 |2 = 1 for the state to be normalized (remember that the func-
tions n (x ) are orthonormal). Take the coefficients to be real, because
complex values change only the phase of the oscillation. For t > 0, the
wavefunction is given by

9(x , t) = A1 1 (x ) ei1 t /h + A2 2 (x ) ei2 t /h . (E1.2)

Calculate the expectation value of x (t) using


Z a
hx (t)i = 9 (x , t) x 9(x , t) dx . (E1.3)
0

This gives four terms when the wavefunction (E1.2) is inserted. Fortunately
those in which both wavefunctions are the same give 12 a by symmetry. This
leaves only the cross-terms, which can be evaluated with the aid of the
integral
Z
2 a x 2 x 16a
x sin sin dx = 2 . (E1.4)
a 0 a a 9
The final result is quoted as equation (1.55).

We need the probability density produced by the wavefunction (E1.2), which is

|9(x , t)|2 = |A1 |2 |1 (x )|2 + |A2 |2 |2 (x )|2


+ A1 A2 1 (x )2 (x ) exp[i(1 2 )t] + c.c., (Q1.35)

where c.c. means the complex conjugate of the previous term, and ensures that
the sum is real. In this case A1 , A2 , 1 (x ), and 2 (x ) are all real. The individual
states are assumed to be orthonormal, which means that
Z Z Z
|1 (x )| dx = |2 (x )| dx = 1,
2 2
1 (x ) 2(x ) dx = 0. (Q1.36)

Normalization of the mixture then requires


Z
1 = |9(x , t)|2 dx = |A1 |2 + |A2 |2 ; (Q1.37)

the time-dependent terms with A1 A2 vanish by orthogonality of 1 and 2 .


12 Foundations

The expectation value of x is then given by


Z a Z a
hx (t)i = |A1 | 2
1 x dx + |A2 |
2 2
22 x dx
0
Z a 0

+ 2A1 A2 cos(1 2 )t 1 (x ) x 2 (x ) dx . (Q1.38)


0

The cosine comes from the combination of the two terms in the last line of equation
(Q1.35). The densities of both 1 and 2 are symmetric about the middle of the
well, so each integral gives 12 a. The integral with the cross-term was given in the
question (equation E1.4), so the result is
32a
hx (t)i = 1
2
a(|A1 |2 + |A2 |2 ) A1 A2 cos(1 2 )t
 9 2 
a 64
= 1 A1 A2 cos(1 2)t . (Q1.39)
2 9 2
This shows explicitly that a mixture of states is needed to produce a statewhere hx i
changes with time. The strongest possible mixture has A1 = A2 = 1/ 2, which
gives hx (t)i 12 a[1 0.36 cos(1 2 )t]. The maximum possible excursion,
exhibited by a classical particle bouncing back and forth in the well, would have
unity instead of 0.36.

Q1.14 The dispersion relation (1.94) is used to model nonparabolicity in the con-
duction band of GaAs. What are the limiting forms of the velocity at small
and large energies, and what happens to the density of states in these two
limits?
Evaluate the velocity and density of states exactly, and plot them with
the density of states for this nonparabolic band and the usual parabolic
approximation that holds at low energy.

The approximate dispersion relation (equation 1.94) is

h 2 K 2
E(1 + E) = . (Q1.40)
2m 0 m e
For small energy, where E  1, we can drop the correction to leave the standard
parabolic form, E h 2 K 2 /2m 0 m e . The usual results can be taken over unchanged;
the velocity v(K ) = (1/h )d E/d K h K /m 0 m e and the density of states is given

by equation (1.93), n(E) = (m 0 m e / 2 h 3 ) 2m 0 m e E. In the opposite limit we

can drop unity in comparison with E, and find E h K / 2m 0 m e . This yields

a constant velocity of vmax = 1/ 2m 0 m e , showing that the velocity saturates,
rather than continuing to rise without limit as in a parabolic band. The density of
Foundations 13
2.0 2.0

vmax nonparabolic
1.5 1.5
v / 106 m s1

E / eV
parabolic
1.0 1.0

0.5 0.5

0.0 0.0
2 1 0 1 2 0 0.4 0.8
1 1 3
k / nm n(E) / eV nm

Figure Q1.2. Velocity v(K ), energy E(K ), and density of states n(E) for a nonparabolic
band. The dispersion relation is E(1 + E) = h 2 K 2 /2m 0 m e with = 0.6 eV1 and
m e = 0.067, which models the conduction band in GaAs.

states can be calculated in the same way that led to equation (1.93), and is found to
be
K2 dK K2 (2m 0 m e )3/2 2
n(E) = 2 = 2 = E . (Q1.41)
dE h v(E) 2 h 3
The reduction in the velocity raises the density of states, which grows as E 2 rather
than E 1/2.
Exact expressions for the velocity and density of states can be found by parametric
differentiation of equation (Q1.40). The results are

2m 0 m e E 1 + E
v(E) = , (Q1.42)
m 0 m e 1 + 2 E

m 0 m e 2m 0 m e E
n(E) = (1 + 2 E) 1 + E. (Q1.43)
2 h 3
In both cases the first quotient is the standard result for a parabolic band, and the
additional terms give the effect of nonparabolicity. These curves are plotted in
figure Q1.2 with the corresponding curves for a parabolic band. The saturation
in the velocity, linearity of E(K ), and growth of density of states as E 2 are all
clear for the nonparabolic band. A worrying feature is that the deviations from
the behaviour of a parabolic band become visible around E = 0.3 eV. We often
consider electrons with such an energy, because it is around the height of wells and
barriers in the GaAsAlGaAs system. Although most calculations are based on a
14 Foundations

parabolic approximation to the bands, this must be questioned if accurate results


are needed.
Q1.15 Show that the density of states for free electrons in two dimensions (the
simplest case) is
m
n 2D(E) = 2(E). (E1.5)
h 2
(Again, the result is more complicated for a realistic dispersion relation,
with further singularities inside the band.)
The general formula shows that the density of states in k-space is given by n 2D (k) =
2/(2 )2 per unit volume of real space. Assume that the energy (k) is a function
only of the magnitude of k. Then the region between two lines of energy and
+ is an annulus of area 2 k k, which contains (k/ )k states. The density of
states in energy for E 0 is then
k dk k
n 2D (E) = = (Q1.44)
dE h v(E)
k m m
= = . (Q1.45)
h k
2
h 2
The first line holds for any band with circular symmetry, but the constant density
of states in the second line is restricted to a parabolic band.
Q1.16 Derive the effective density of states for aRthree-dimensional system (equa-

tion 1.119). You will need the integral 0 x 1/2ex dx = 12 ! = 0( 32 ) =

1
2
.
The density of electrons is given in general by equation (1.111),
Z
n = n(E) f (E, E F , T ) d E. (Q1.46)

Substitute the density of states for a parabolic band in three dimensions with its
bottom at E c and effective mass m e . Assume that we can take the Boltzmann limit
(1.108) of the FermiDirac function, which will be valid if (E c E F )  kB T . The
exponential decay allows us to set the upper limit of integration to infinity. Thus
Z  
(2m 0 m e )3/2 E EF
n (E E c ) exp
1/2
d E. (Q1.47)
2 2h 3 Ec kB T
The substitution x = (E E c )/kB T leads to
  Z
E c E F (2m 0 m e kB T )3/2 1/2 x
n exp x e dx (Q1.48)
kB T 2 2h 3 0
   
m 0 m e kB T 3/2 Ec EF
= 2 exp . (Q1.49)
2 h 3 kB T
Foundations 15

The factor in front of the exponential defines the effective density of states,
 
(3D) m 0 m e kB T 3/2
Nc = 2 , (Q1.50)
2 h 3
which confirms equation (1.119).

Q1.17 Combine the semiconductor equation with that for neutrality to show that
the density of electrons is given by
 q 
n = 2 (ND NA ) + (ND NA ) + 4n i
1 2 2
(E1.6)

in a classical semiconductor. Check its limits for n-type material where


ND NA  n i , p-type material where NA ND  n i , and very lightly-
doped material where |ND NA |  n i . What are the corresponding results
for p?

We need the semiconductor equation (1.122),


 
Eg
np = n 2i = Nc Nv exp , (Q1.51)
kB T
and the condition for neutrality, p + ND = n + NA . Eliminating p between these
leads to a quadratic equation for n,

n 2 + (NA ND )n n 2i = 0, (Q1.52)

whose solution gives equation (E1.6). The other root is negative and can therefore
be discarded.
Now look at the limits. The condition ND NA  n i usually holds in n-type
material, and the binomial expansion gives
n 2i
n ND NA + . (Q1.53)
ND NA
In practice one can often take n = ND . In the opposite limit of p-type material
where NA ND  n i, the electrons are minority carriers and
n 2i
n (Q1.54)
NA ND
to lowest order. This is consistent with p NA ND and the semiconductor
equation np = n 2i . Finally, very lightly doped material may be in the intrinsic limit
where |ND NA |  n i. In this case the binomial expansion gives

n n i + 12 (ND NA ). (Q1.55)
16 Foundations

The corresponding result for holes can be found in the same way,
 q 
p = 2 (NA ND ) + (NA ND ) + 4n i .
1 2 2
(Q1.56)

The limits are also obvious.


Q1.18 Show that the Fermi wavevector K F = (3 2n 3D)1/3 in three dimensions
and derive the corresponding result for one dimension.
The general result is that the number of states inside the Fermi (hyper)sphere must
equal the number of particles (per unit volume of real space). The number of states
is just the product of the volume and the density of states in k-space, 2/(2 )d in d
dimensions. Thus in 1, 2, and 3 dimensions we find
2
2kF(1D) = n 1D, kF(1D) = 12 n 1D (Q1.57)
2
2
[kF(2D) ]2 = n 2D, kF(2D) = (2 n 2D)1/2 (Q1.58)
(2 )2
2
4
[kF(3D) ]3 = n 3D, kF(3D) = (3 2n 3D )1/3 (Q1.59)
3
(2 )3
A general formula can also be written down. The volume of a unit hypersphere is
d/2 /( 12 d)! = d/2 / 0(1 + 21 d) in d dimensions, whence

kF(d) = 2 1/2[ 12 ( 12 d)!n d ]1/d . (Q1.60)


The individual cases are considerably simpler!
Q1.19 For the following examples, calculate the Fermi temperature and determine
whether the system is degenerate at room temperature, in liquid nitrogen
(77 K), and in liquid helium (4.2 K).
Al has 18.1 1028 m3 electrons, about the maximum for a common
metal, and an effective mass near unity. Cs is a less typical metal with a
much lower density, 0.91 1028 m3 .
Highly doped n-GaAs has 5 1024 m3 electrons with m e = 0.067.
One could also dope it p-type with a higher effective mass, m h 0.5 for
heavy holes. Lightly-doped material might have something like 1021 m3
carriers.
A two-dimensional electron gas in GaAs has (1 10) 1015 m2 elec-
trons (dont forget the dimensions!).
The general relation is
h 2 kF2 h 2 (3 2n 3D)2/3 h 2 n 2D 2 h 2 n 21D
E F0 = kB TF = = = = . (Q1.61)
2m 0 m eff 2m 0 m eff m 0 m eff 8m 0 m eff
Foundations 17

Table Q1.3. Fermi temperatures of various systems.

material density effective mass Fermi temperature


Al 18.1 1028 m3 1 140 000 K
Cs 0.91 1028 m3 1 18 000 K
n-GaAs 5 1024 m3 0.067 1800 K
n-GaAs 1 1021 m3 0.067 6.3 K
p-GaAs 5 1024 m3 0.5 250 K
p-GaAs 1 1021 m3 0.5 0.85 K
2DEG in GaAs 10 1015 m2 0.067 410 K
2DEG in GaAs 1 1015 m2 0.067 41 K

Numerical results using these equations are listed in table Q1.3. Electrons in both
metals are clearly degenerate at any reasonable temperature. Highly-doped n-type
GaAs also contains degenerate electrons at room temperature and below, but elec-
trons in the lightly-doped material is degenerate only in liquid helium. The heavier
effective mass of holes in p-GaAs makes it harder to keep them degenerate; they
are losing degeneracy at room temperature even with the higher doping although
the holes are degenerate in liquid nitrogen or cooler. The lightly p-type material has
nondegenerate holes at all temperatures considered. The practical range of densities
for a 2DEG has a significant effect on its Fermi temperature; the 2DEG is just about
degenerate at room temperature when its density is high, but at low density it is
nondegenerate in liquid nitrogen.
Q1.20 Show that the average energy of electrons in a three-dimensional electron
gas at zero temperature is E 3D = 35 E F0 , with E 2D = 12 E F0 in two dimen-
sions. Show also that the corresponding results are 32 kB T and kB T at high
temperature, where the electrons are nondegenerate.
Define zero energy to be the bottom of the band. At zero temperature all states are
occupied whose energy is below the Fermi level E F0 , and those with higher energy
are empty. Thus the average energy is given by
Z E0 ,Z 0
F EF
E = E n(E) d E n(E) d E . (Q1.62)
0 0

The constant factors in the density of states n(E) cancel, so we need only consider its
dependence on energy. Consider n(E) E p to get a general result. The integrals
are trivial, and the denominator becomes
Z E0 Z E0
F F (E 0 ) p+1
n(E) d E E p dE = F . (Q1.63)
0 0 p+1
18 Foundations

The numerator is no more difficult, and we find


p+1 0
E = E . (Q1.64)
p+2 F
The density of states for free electrons in d = 1, 2, and 3 has index p = 12 , 0, and
1
2
. Thus E 1D = 13 E F0 , E 2D = 12 E F0 , and E 3D = 35 E F0 .
We must include a Boltzmann factor in the integral if the electrons are nonde-
generate, and extend the upper limit to infinity. Then the denominator of equation
(Q1.62) is replaced by
Z  
E EF
n(E) exp d E. (Q1.65)
0 kB T
The prefactor again cancels and takes the Fermi level with it. Now the denominator
is proportional to
Z
E p exp(E/kB T ) d E = p!(kB T ) p+1. (Q1.66)
0
The numerator is similar and the factorial cancels in the quotient to leave
E = ( p + 1)kB T . (Q1.67)
These give E 1D = 12 kB T , E 2D = kB T , and E 3D = 32 kB T . These are standard
classical results from the equipartition of energy in a system with only kinetic
energy.
Q1.21 Plot the average occupation hni of a donor described by equation (1.128) as
a function of , to show the effect of the Hubbard U . Take U = 100 meV
and T = 77 K, and plot the curve for U = 0 for comparison. Although it
seems obvious that U should be positive, there are other systems where it
behaves as though it were negative. What effect does U = 100 meV have
on the occupation as a function of ? Provide a suitable approximation
for hni in this limit, the analogue of equation (1.129).
It would be useful to explain a bit more about the calculation before describing the
plots.
According to quantum statistical mechanics, the probability of occupation for
a quantum state that contains N electrons with total energy E is proportional to
exp[(E N )/kB T ], where is the chemical potential. This applies to a grand
canonical ensemble, where the small system under study is weakly coupled to a
large reservoir with which it can exchange both energy and particles. Formally
the chemical potential enters as a Lagrange multiplier that constrains the average
number of particles in the small system; it sets the energy required for particles to
move between the system and reservoir.
Foundations 19

Table Q1.4. The four possible states of an impurity that can hold zero, one or two electrons,
showing their total energies and the probability of being found in each state.

state energy probability


proportional to
empty 0 1
one electron, 1 exp[( )/ kB T ]
one electron, 1 exp[( )/ kB T ]
two electrons, 2 2 + U exp[(2 + U 2)/ kB T ]

This general results allows us to calculate the probability for the system to be
found in each of its four states, and the results are listed in table Q1.4. The proba-
bilities should be normalized by dividing by the sum of the last column of the table,
and can then be used to calculate averages. For example, the average number of
electrons is given by

0 1 + 1 e()/ kB T + 1 e()/ kB T + 2 e(2+U 2)/ kB T


hni = . (Q1.68)
1 + e()/ kB T + e()/ kB T + e(2+U 2)/ kB T

This reduces to equation (1.128). Other statistics, such as hn 2i, can be deduced in
the same way.
Plots of the average occupation according to equation (Q1.68) are shown in figure
Q1.3. The total occupation is plotted, with the contributions from impurities that
hold one and two electrons. The Hubbard U is positive, and much larger than
kB T , in figure Q1.3(a). The average occupation rises in two steps as a function of
chemical potential. The first, around = , occurs when the first electron enters the
impurity. Almost all impurities hold a single electron in the range < < + U .
It becomes favourable for a second electron to occupy the impurities for > +U
and the number of singly-occupied impurities drops as increases.
The occupation is plotted in figure Q1.3(b) for the case U = 0 where there
is no repulsion between electrons. It was shown in section 1.8.5 that the overall
occupation is given by twice the Fermi function. The plot shows that there is a small
region around = where the impurity may be singly occupied, but outside this
range the occupation is either 0 or 2.
Finally, the effect of a negative Hubbard U is shown in figure Q1.3(c). Electrons
are attracted to one another if U < 0 so the impurity is either empty or doubly
occupied the curve for single occupation barely creeps off the axis. The attraction
U is shared between a pair of electrons, so the rise in occupation occurs around
= + 12 U . Another feature of this transition is its steepness: it is clearly sharper
than the plot for U = 0 in figure Q1.3(b). Expansion of the general result (1.128)
20 Foundations
2
(a) U = 100 meV

1
average number of electrons on impurity

0
2
(b) U = 0

0
2
(c) U = 100 meV

1
electrons on singly occupied sites
electrons on doubly occupied sites
total number of electrons
0
-100 -50 0 50 100 150
( ) / meV

Figure Q1.3. Occupation of an impurity that can hold zero, one or two electrons as a function
of chemical potential . The temperature T = 77 K, the energy of a single electron on the
impurity is , and the Hubbard U takes the values (a) 100, (b) 0, and (c) 100 meV in the
three plots.

around this region gives


2
hni . (Q1.69)
exp[( + 1
2
U )/ 12 kB T ] + 1
The behaviour is like a FermiDirac function with half the actual temperature.
It might seem that the direct Coulomb repulsion between electrons should ensure
that U is always positive, but the total interaction between a pair of electrons may
be strongly modified by their environment. This may lead to a negative effective
Foundations 21

value of U . Chemical bonds, for example, almost always contain pairs of electrons,
which implies U < 0. More complicated effects occur in solids, where electrons
can also interact indirectly by distorting the lattice around them. The DX centre, to
be described in section 9.2.1, is generally believed to have a negative U . This is an
ion of Si+ , which acts as a donor in the semiconductor Alx Ga 1x As. It behaves as
a normal donor for a single electron, which is trapped weakly. A second electron,
however, permits a distortion of the lattice to occur around the Si and release
energy. Thus the second electron is more tightly bound than the first, which is
another way of describing the effect of a negative U .
As these examples show, even the simple problem of the number of electrons
on an impurity is far more than a trivial application of the FermiDirac function if
interactions are taken into account.

Further reading
There is an excellent textbook on quantum mechanics, omitted from the list of ref-
erences in my book, by Herbert Kroemer, Quantum mechanics: for engineering,
materials science, and applied physics, Englewood Cliffs, New Jersey, USA: Pren-
tice Hall (1994). It goes far beyond my book and provides much insight into aspects
that I was not able to include.

S-ar putea să vă placă și