You are on page 1of 29

Band-to-Band Transitions

Karl W. Böer and Udo W. Pohl

Contents
1 Optical Absorption Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 The Joint Density of States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Absorption Coefficient and Dielectric Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The Fundamental Absorption Edge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Direct and Indirect Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Indirect Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Allowed and Forbidden Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Band-to-Band Magnetoabsorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Transitions in Quantum Wells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.1 Energy Levels in Multiple Quantum Wells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Absorption in Quantum Wells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 Optical Bandgap of Amorphous Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1 Intrinsic Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Extrinsic Absorption in Glasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Abstract
Optically induced band-to-band transitions are resonance transitions and related
to the band structure by the momentum matrix-element and the joint density of
states. For transitions near the band edge, the theory of optical transitions
between the valence and conduction bands can be simplified with an effective-
mass approximation, assuming parabolic band shapes and arriving at

K.W. Böer
Naples, FL, USA
e-mail: karlwboer@gmail.com
U.W. Pohl (*)
Institut f€ur Festkörperphysik, Technische Universität Berlin, Berlin, Germany
e-mail: pohl@physik.tu-berlin.de

# Springer International Publishing Switzerland 2015 1


K.W. Böer, U.W. Pohl, Semiconductor Physics,
DOI 10.1007/978-3-319-06540-3_13-1
2 K.W. Böer and U.W. Pohl

quantitative expressions for the absorption as a function of the photon energy.


Depending on the conduction-band behavior, strong direct or weak indirect
transitions occur at the band edge. In addition, a contribution of forbidden
transitions modifies the absorption further away from the band edge. Deviations
from the ideal, periodic crystal lattice provide tailing states extending beyond the
band edge, usually as an Urbach tail which decreases exponentially with dis-
tance from the band edge. In quantum wells the two-dimensional joint density
of states leads to a steplike increase of the absorption for increasing photon
energy.

Keywords
Absorption coefficient • Band structure • Band tailing • Direct transitions •
Forbidden transitions • Indirect transitions • Joint density of states • Momen-
tum-matrix element • Optical transitions • Parabolic bands • Urbach tail • Van
Hove singularities

1 Optical Absorption Spectrum

The description of optical band-to-band transitions requires a quantum mechanical


analysis (Bassani 1966; Bassani and Pastori Parravicini 1975). For a simplified
treatment of transitions near the fundamental bandgap, an effective-mass model
can be used. This chapter focuses on the absorption process; the subsequent
radiative or nonradiative recombination processes are treated in chapter “▶ Carrier
Recombination.” The absorption spectrum is directly related to the imaginary
part of the dielectric function introduced in chapter “▶ Interaction of Light with
Solids.” One distinguishes direct band-to-band transitions, which are essentially
vertical transitions in E(k), and indirect transitions, which involve phonons, per-
mitting major changes of k during the transitions. We will first discuss the direct
transitions.

1.1 The Joint Density of States

Light of sufficiently short wavelength with

E ¼ A0 e exp½iðkr  ωtÞ (1)

initiates electronic transitions from one band to another. We have used A0 here as
the amplitude to avoid confusion with the energy of an optical transition, E. e is the
unit vector of the electric polarization and k is the wavevector of the light, traveling
in r direction.
The number of optically induced transitions at the same k (i.e., neglecting the
wavevector of the photon) between band μ and band ν is proportional to the square
of the momentum matrix elements given by
Band-to-Band Transitions 3

ð  
@
e Mμν ðkÞ ¼ e ψ μ ðk, rÞ iℏ ψ ν ðk, rÞ dr; (2)
V @r

with the integral extending over the crystal volume V. The proportionality factor for
a specific transition is

 
1 e A0 2  
Pμν ðkÞ ¼ δ Eμ ðkÞ  Eν ðkÞ  hν ; (3)
h m0 c

where δ is the Dirac delta function. The factor 1/h indicates the quantum nature of
the transition, e/(m0c) stems from the interaction Hamiltonian between light and
electrons, and A0 from the amplitude of the light (i.e., the value of the Poynting
vector A20). The delta function switches on this contribution when a transition occurs
from one state to another, i.e., when Eμ ðkÞ  Eν ðkÞ ¼ hν. No broadening of any of
these transitions is assumed. Close proximity to adjacent transitions and
Kramers–Kronig interaction make such broadening consideration unnecessary,
except for very pronounced features.
After integration over all states within the first Brillouin zone and all bands
between which the given photon hν can initiate transitions, we obtain for the
number of such transitions per unit volume and time Fermi’s golden rule:

Xð 2  2
W ðνÞ ¼ 3
Pμν e  Mμν ðkÞ dk
μ, ν ð2π Þ
BZ
  ð (4)
2 e A0 2 X    
e  Mμν ðkÞ2 δ Eμ ðkÞ  Eν ðkÞ  hν dk;
¼ 3
ð2π Þ h m0 c μ, ν BZ

where the integration is over the entire Brillouin zone (BZ). The factor Vunit cell/
(2π)3 normalizes the k vector density within the Brillouin zone; Vunit cell cancels
from the integration of Eq. 2; the factor 2 stems from the spin degeneracy.
Equation 4 follows directly from a perturbation theory. The matrix elements vary
little within the Brillouin zone; therefore, we can pull these out in front of the
integral. This leaves only the delta function inside the integral. The integral
identifies the sum over all possible transitions which can be initiated by photons
with a certain energy hν, and it is commonly referred to as the joint density of states
between these two bands. It is given by

ð
2  
J μν ðνÞ ¼ 3
δ Eμ ðkÞ  Eν ðkÞ  hν dk: (5)
ð2π Þ BZ

The argument of the delta function is a function of k. Since the photon


momentum is negligibly small, kinitial = kfinal applies for the transition (assumed
4 K.W. Böer and U.W. Pohl

 
to be direct), yielding for the delta function1 δ Eμ  Eν  hν ¼ δðkinital  kfinal Þ=
  
@=@k Eμ ðkÞ  Eν ðkÞ  . The joint density of states can then be written as
Eμ Eν ¼hν
an integral over a surface of constant energy instead of an integral in k space:
ð
2 dS
J μν ðνÞ ¼    ; (7)
ð2π Þ 3 @=@k Eμ ðkÞ  Eν ðkÞ 
Eμ Eν ¼hν

where dS is an element on the equal-energy surface in k space, with the surface


depicted by

Eμ ðkÞ  Eν ðkÞ ¼ hν; (8)

here μ stands for any one of the conduction bands and ν for any one of the valence
bands.
For each of the transitions, k is constant; it is a direct transition. When the slopes
of E(k) of both bands are different, Eq. 4 is fulfilled only for an infinitesimal surface
area. In contrast, a larger area about which such transitions can take place appears
near points at which both bands have the same slope, i.e., near critical points (van
Hove 1953 and Phillips 1956). Here

@Eμ ðkÞ=@k ¼ @Eν ðkÞ=@k ¼ 0 or @Eμ ðkÞ=@k  @Eν ðkÞ=@k ¼ 0; (9)

and the expression under the integral of Eq. 7 has a singularity. By integration this
results in a kink in the joint density of states. The singularity related to Eq. 7 is
referred to as a van Hove singularity.
The types of critical points are referred to as M0 . . . M3, P0 . . . P2, and Q0 or Q1 in
three-, two-, and one-dimensional E(k) representations, respectively; two- and
one-dimensional representations are applicable to superlattices or quantum wire
configurations (Sect. 3 of chapter “▶ Bands and Bandgaps in Solids”). They describe
the relative curvature of the lower and upper bands with respect to each other. In
a first-order expansion at a critical point kcp in k space, we can express this by

  X 3  2
Eμ ðkÞ  Eν ðkÞ ¼ Eμν, 0 kcp þ ai k  kcp ; (10)
i¼1

where the components of k and kcp are along the principal axes. Equation 10 may be
interpreted as a flat lower band (Eμν,0) and the appropriately changed curvature of
the upper band describing the energy difference between the original valence and

1
The rewriting of Eq. 5 can be understood from the behavior of the Dirac delta function
ð  
d f ðxÞ1
gðxÞ δ½f ðxÞ dx ¼ gðx0 Þ   : (6)
dx x¼x0
Band-to-Band Transitions 5

J(E ) 3D 2D 1D
M0 M1 M2 M3 P0 P1 P2 Q0 Q1

Fig. 1 Joint density-of-state representation near critical points (schematic) in three (Mi), two (Pi),
and one (Qi) dimensions of E(k) with subscripts identifying the number of negative signs of the
factors ai in Eq. 10

conduction bands. Here positive quantities ai are proportionality factors measuring


the relative band curvatures, and the combination of their signs identifies the type of
critical point, 2 with a schematic representation shown in Fig. 1.
The integrand in Eq. 7 can be written
ð π dk3/a1, yielding for the joint density of
states the expression J μν ðνÞ ¼ 8π
const
2a
1
dk3 (Cohen and Chelikowsky 1988). Jμν is thus
proportional to the extent of the surface in k3 direction. At an M0 critical point, the
  1=2  1=2
surface extends k3 ¼  ω  ωcp =a3 , leading to J μν ðνÞ / ν  νcp rising
from hνcp. The structure of other critical points follows similar geometrical argu-
ments (Cohen and Chelikowsky 1988). In optical spectra of αo, R, or e00 , the critical
points lead to sharp van Hove singularities, albeit the matrix elements (Eq. 2) effect
some deviation from the ideal structure.
Matrix elements are mainly influenced by selection rules, which are similar to
those in atomic spectroscopy. The symmetry properties of ψ μ and ψ ν in Eq. 2 – for
example, whether even or odd under reflection or inversion and whether they are the
same or different from each other – determine whether the matrix element has a
finite value or vanishes; judgment can be rendered on group theoretical arguments
(see Bassani and Pastori Parravicini 1975). Comparing the results of such an
analysis with experimental observation of strong or weak optical absorption pro-
vides quite convincing arguments by assigning certain features of the absorption
spectrum to the appropriate critical points of the E(k) behavior.
The photon which is absorbed in the band-to-band transition can be released in a
radiative recombination process. Such luminescence – referred to as spontaneous
emission – is important for optoelectronic devices. Luminescence is time-reversed
to optical absorption; therefore, the same matrix elements and selection rules apply for
both processes. The luminescence intensity is obtained from a product of the matrix
element with the joint density of states and the distributions of the excited carriers.
Radiative recombination is discussed in chapter “▶ Carrier Recombination.”

2
0, 1, 2, or 3 negative factors ai in Eq. 10 yield for M0 to M3 a maximum, saddle point, saddle point,
or minimum, respectively, in three dimensions; 0, 1, or 2 negative factors for P0 to P2 a maximum,
saddle point, or minimum, respectively, in two dimensions; and 0 or 1 negative factors for Q0 or Q1
a maximum or minimum, respectively, in one dimension.
6 K.W. Böer and U.W. Pohl

1.2 Absorption Coefficient and Dielectric Function

The optical absorption coefficient αo (Sect. 1.1.3 of chapter “▶ Interaction of Light


with Solids”) is proportional to the number of optical transitions per unit volume
and unit time. The coefficient can be calculated from simple optical principles
(Bassani and Pastori Parravicini 1975): it is given by the absorbed energy per unit
time and volume, hv  W(v), divided by the energy flux 2πA20ν2e0nr/c. The energy
flux is equal to the optical energy density given by the square of the wave amplitude
per wavelength interval 2πA20ν2n2r /c2, divided by the light velocity within the
semiconductor c/nr. With Eq. 4 we now obtain

2 e2 Xð    
αo ðνÞ ¼ e  Mμν ðkÞ2 δ Eμ ðkÞ  Eν ðkÞ  hν dk: (11)
3 e n cm2 2πν
ð2π Þ 0 r 0 μ, ν BZ

For a further analysis of the integral, we need more knowledge about E(k). This will
be done near the fundamental absorption edge in Sect. 1.3 where some simplifying
assumptions for the valence and conduction bands are introduced. For a short
review, see Madelung (1981).
With the availability of synchrotron radiation sources, light sources of sufficient
power and stability have become available to perform absorption and reflection
spectroscopy for higher bands in the vacuum UV and soft x-ray range. Since the
optical absorption here is very strong, requiring extremely thin crystals for absorp-
tion spectroscopy which are difficult to obtain with sufficient crystal perfection, it is
preferable to obtain such spectra in reflection.
A more sensitive set of methods relates to modulation spectroscopy (Sect. 1.2.3
of chapter “▶ Interaction of Light with Solids”) where a crystal variable – e.g., the
electric field, temperature, or light intensity – is modulated and the changing
reflection signal is picked up by look-in technology. To compare the optical
properties with the band structure, one must now express the transitions in terms
of the complex dielectric constant e ¼ e0 þ ie00 ð¼ e1 þ ie2 Þ. The optical absorption
constant αo is then obtained from e00 ¼ αo nr c=ð2πνÞ (see Eqs. 22 and 24 of chapter
“▶ Interaction of Light with Solids”), with
2 e2 Xð    
e00 ¼ e  Mμν ðkÞ2 δ Eμ ðkÞ  Eν ðkÞ  hν dk: (12)
3 2
ð2π Þ e0 m20 ð2πνÞ μ, ν BZ
This representation has the advantage of not depending on the index of refraction,
which causes a distortion of the calculated e00 (ν) distribution. With the first
Kramers–Kronig
  relation (Eq. 20 of chapter “▶ Photon-Phonon Interaction”) and
ωa ¼ Eμ  Eν =ℏ, we obtain for e0 from Eq. 12 after conversion of the δ function
(Bassani and Pastori Parravicini 1975)

ð  
2 e2 X 2 e  Mμν ðkÞ2 dk
0  
e ¼1þ   :
π m20 μ, ν BZ ð2π Þ3 Eμ ðkÞ  Eν ðkÞ =ℏ Eμ ðkÞ  Eν ðkÞ 2 =ℏ2  ω2
(13)
Band-to-Band Transitions 7

a E2
M1 M2
40 Ge Χ4 → Χ1 ∑4 → ∑1

35 exp.
theory
30
E1
25
M1 M
e”

20 Λ3 → Λ1 L3’ →3 L
3

15

10
M0
5 M0 Γ25’ → Γ15
L3’ → L1
0
1 2 3 4 5 6 7 8
hν (eV)
b
5 L3
Ge
4 Γ15 Γ15

3
M3 ∆1 ∑1
2 Λ 1 M0
Γ2’
E (eV)

L1 Γ2’ Χ1
1 M1
M0 M2 M0
0
M0 M1 Γ25’
–1 Λ3 Γ25’
L3’
–2
∆5
∑4
Χ4
l/2,l/2,l/2 0,0,0 l,0,0 l/2,l/2,0 0,0,0
k 3/4,3/4,0

Fig. 2 (a) Optical absorption spectrum with critical points and (b) energy bands with
corresponding transitions for Ge (After Brust et al. 1962). The dashed curve is the joint density
of states, weighted by oscillator strength f, computed by the empirical pseudofunctional method

As an example, the e00 (ν) spectrum is given in Fig. 2 for Ge. It shows several
shoulders and spikes that can be connected to critical points in the corresponding
theoretical curve, such as the L, Γ, and X transitions (see Sect. 4.2 of chapter
“▶ Quantum Mechanics of Electrons in Crystals”). The theoretical spectrum was
obtained from an empirical pseudopotential calculation (Phillips 1966) by numer-
ically sampling Eμ ðkÞ  Eν ðkÞ in the entire Brillouin zone at a large number of
8 K.W. Böer and U.W. Pohl

a , E2
60 E0
50 Si E1
40 EgX
30 e1 ,
20 e2 E1
e

10
0
–10
–20
b 3.5 1.0
3.0 Si Exp. , E2 ,
E0 E1 E1 0.8
a o (106 cm–1)

2.5 MDF
2.0 EgX 0.6
R

R
1.5 0.4
1.0
0.2
0.5
a
0.0 0.0
0 1 2 3 4 5 6
hν (eV)
Fig. 3 (a) Real and imaginary parts of the dielectric function as well as (b) absorption coefficient
and normal-incidence reflectivity of Si at 300 K as a function of the photon energy (After Adagi
2005). Circles and solid lines are, respectively, experimental data and calculations based on a
model dielectric function (MDF) with a given number of interband critical points

k values. By ordering and adding the transitions that occur at the same value of hv,
possibly involving different band combinations, and repeating for other values of
hv, we obtain the joint density of states. These yield the given e00 (ν) as shown in
Fig. 2. The biggest peaks relate to ranges in E(k) where both bands involved in the
transition run parallel to each other for an extended k range.
The spectral distribution between 0 and 6 eV of e0 and e00 (also written e1 and e2)
and their relation to the experimentally obtained optical constants (see Eqs. 22 and
24 of chapter “▶ Interaction of Light with Solids”) for Si are shown in Fig. 3.
Advances in ab initio calculations – mostly within density-functional theory and
local-density approximation (DFT-LDA, Sect. 2.2.2 of chapter “▶ Quantum
Mechanics of Electrons in Crystals”) – made it possible to compare the optical
absorption spectrum calculated from first principles with experiment. Commonly
LDA eigenvalues and eigenfunctions are used to calculate e00 . Several corrections
are, however, to be made to obtain quantitative agreement with experiment. These
are self-energy corrections (Sect. 2.2.3 of chapter “▶ Quantum Mechanics of
Electrons in Crystals”; Hybertsen and Louie 1985; Aulbur et al. 2000) for
correcting the underestimation of the energy gap in LDA and excitonic effects
associated with Coulomb interaction between electron and hole (Albrecht
et al. 1998).
Band-to-Band Transitions 9

GaAs E1 + Δ1 ,E
E0 2
E1
22 K
20 253 K
e2 504 K
754 K
10

0
E1 E1 + Δ1
GaAs 22 K
253 K
20
E0 504 K
, 754 K
E0
10
e1
E2
0

–10

1 2 3 4 5 6
hν (eV)

Fig. 4 Real and imaginary parts of the dielectric constant of GaAs as a function of the photon
energy with the temperature as the family parameter (After Lautenschlager et al. 1987b)

The influence of temperature on the principal bandgap was discussed in Sect. 2.2
of chapter “▶ Bands and Bandgaps in Solids.” An example for changes within
higher bands is shown in Fig. 4 for the real and imaginary parts of the dielectric
constant for GaAs (for Si see Lautenschlager et al. 1987a; Shkrebtii et al. 2010).

Absorption from Core Levels


When the excitation takes place from a sharp core level rather than from a wide
valence band, the structure of the band to which this transition occurs, the conduction
band, can be obtained directly. When comparing excitation spectra from core levels
with those from the valence band, however, a difference occurs because of the
different relaxation of the excited state which is stronger when the hole is localized
at the core than when it is more widely spread out in the valence band (Zunger 1983).
The joint density of states involving excitation from a core level is identical with
the density of states in the conduction band. A typical example for a core-to-
conduction-band spectrum of the reflectivity is given in Fig. 5. The experimental
curves usually show less detail than those obtained from band-structure calculations
(Martinez et al. 1975). This may be caused by electron–hole interaction and local
10 K.W. Böer and U.W. Pohl

Δ(6),P(6)
1

d3/2 Σ(6)
d 2R / dλ2 (arb. units)
PbTe Exp.

3/2
d
0

d3/2 Σ’(6),Λ(6)
d3/2 L(6)
d5/2 Σ(6)

Λ(7,8)

d3/2 Δ”(6)
d3/2 Π (6)
d3/2 Δ’(6)
|Σ’(7)
d5/2 Σ(7)
–1 d 5/2

0.04
PbTe Exp. (arb. units)
Theory

0.03
Reflectivity

0.02

0.01
Threshold
Energy

1 2 3 4 5 6 7
0.0
18 20 22 24 26
hν (eV)
Fig. 5 Reflectivity of PbTe in the conduction-band range with excitation from the core levels d3/2
and d5/2 (distance from the conduction-band edge 18.6 eV; see shifted scale.) The upper part of
the figure shows the second derivative of the reflectivity and exposes more structure of the
reflection spectrum. The transitions to the corresponding critical points are identified (After
Martinez et al. 1975)

field effects which are insufficiently accounted for in the single-electron approxi-
mation used for band calculation (Hanke and Sham 1974). These effects can be
included following methods similar to those applied in valence–conduction-band
transitions (Aulbur et al. 2000; Albrecht et al. 1998).

1.3 The Fundamental Absorption Edge

Attention will now be focused on the energy range near the threshold for
valence–conduction-band transitions, the fundamental absorption edge. Usually,
several excitation processes are possible between different subbands for a given
Band-to-Band Transitions 11

Fig. 6 Optical vertical kz


excitation transitions for
monochromatic photons at
the band edge of GaAs near
k = 0 from the heavy hole
(hh), light hole (lh), and
spin–orbit split-off (so) kx
valence bands into the
conduction band. The upper Γ6 cb
diagram indicates the
resulting electron distribution E(k)
in k space. The flattened outer
rings indicate the strongly
warped hh and the lesser Eg
warped lh valence bands
(Sect. 1.2 of chapter kx
“▶ Bands and Bandgaps in
Solids”); the split-off valence Γ8 ∆0 hh
band is not warped (After Γ8 lh
Lyon 1986)

Γ7 so

optical excitation energy. For a photon energy slightly exceeding the bandgap,
transitions from different valence bands into different E(k) values of the conduction
band are possible near k = 0, as shown in Fig. 6. The distribution of these states is
represented by nearly spherical shells (upper part of Fig. 6). Deviations from
spherical shells occur because of the warping of the valence bands. Thermalization
will average the excited electron distribution to approach a Boltzmann distribution
(chapter “▶ Equilibrium Statistics of Carriers”).
The transition between the top of one valence band to the bottom of the lowest
conduction band when both lie at k = 0 is discussed first. This transition is
responsible for the direct optical absorption edge. The joint density of states is
estimated from a parabolic approximation of both bands:

ℏ2 k2 ℏ2 k2 ℏ2 k2
hν ¼ Ec ðkÞ  Ev ðkÞ ¼ Eg þ þ ¼ Eg þ (14)
2mn 2mp 2μ
with Eg as the bandgap energy and μ as a reduced carrier mass

1 1 1
¼ þ : (15)
μ mn mp

Recognizing that Eq. 5 yields


 
2 d 4π 3
J cv ¼ k
ð2π Þ3 dðhνÞ 3
12 K.W. Böer and U.W. Pohl

and using Eq. 14, we obtain for the joint density of states

 3=2
1 2μ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
J cv ¼ hν  Eg : (16)
2π 2 ℏ2

After introducing Jcv into Eq. 11, we obtain for the optical absorption coefficient
near the band edge (Moss et al. 1973):

2π e2 ð2μÞ3=2 je  Mcv j2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi


αo, cv ¼ hν  Eg : (17)
3 m20 e0 nr c h3 ν

The momentum matrix element can be approximated for transitions from the three
valence bands in zincblende-type semiconductors across the bandgap (near Eg)
from the Kane estimate (Kane 1957) when using


m 2
0
je  Mcv j2 ffi P2 ; (18)

where P is the interband (momentum matrix) parameter, which can be obtained from
the k  p perturbation theory (Sect. 1.2.8 of chapter “▶ Bands and Bandgaps in Solids”):
 
1 1 3ℏ2 Eg þ Δ0
P ¼
2
 Eg (19)
m0n m0 2 3Eg þ 2Δ0
(compare with Eq. 33 of chapter “▶ Bands and Bandgaps in Solids”). Here Δ0 is
the spin–orbit splitting energy and m0n the effective mass at the bottom of the
conduction band. With known mn, Eg, and Δ0, we obtain a numerical value for P,
and therefore for the matrix element,3

3m0   Eg þ Δ0
je  Mcv j2 ¼ m0  m0n Eg : (21)
2m0n 3Eg þ 2Δ0

When this is introduced into Eq. 17, we obtain for hν ffi Eg and m0n ffi mn :

3
The momentum matrix-element with dimension W2s4cm2 should not be confused with the often
used oscillator strength

2
f cv ¼ je  Mcv j2 ; (20)
3m0 hν
which is dimensionless and of the order of one, while the matrix element is not. The factor 1/3 in
 2
Eq. 19 is due to averaging, with jMx j2 ¼ My  ¼ jMz j2 ¼ 1=3jMj2; factor 2 accounts for the spin.
Band-to-Band Transitions 13

pffiffiffiffiffiffiffiffi !3=2
2π e2 2m0 mn mp m0  mn  
αo, cv ffi   f Eg (22)
e 0 nr h2 c m0 mn þ mp mn

with

  Eg þ Δ0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f Eg ¼ hν  Eg ; (23)
3Eg þ 2Δ0

which increases proportional to the square root of the energy difference from the
band edge:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
αo, cv ¼ αo, dir / hν  Eg : (24)

2 Direct and Indirect Transitions

The optical transitions discussed in the previous section are direct transitions; they
involve a direct optical excitation from the valence to the conduction band, using
only photons. This is represented by a vertical transition in the reduced E(k)
diagram, since the momentum of a photon, which is added to or subtracted from
the electron momentum when a photon is absorbed or emitted, is negligibly small:
the ratio of photon to electron momentum can be estimated from pph =pel ¼ ðhν=cÞ
=ðℏkÞ which, at the surface of the Brillouin zone, is ffi ðh=λÞ=ðℏπ=aÞ ¼ 2a=λ ffi
103 for visible light.

2.1 Indirect Transitions

Nonvertical transitions are possible when, in addition to the photon, phonons are
absorbed or emitted during the transition. These transitions are indirect transitions.
The phonons can provide a large change in the electron momentum k (the phonon
momentum is identified by q):

k0 ¼ k  q: (25)

Because of the necessity of finding a suitable phonon for the optical transition to
obey energy and momentum conservation for the transition, the probability of an
indirect transition is less than that of a direct transition by several orders of
magnitude. An estimate of the matrix elements for phonon-assisted electron tran-
sitions follows the corresponding selection rules including the symmetry of the
phonon (Lax and Hopfield 1961).
Indirect transitions identify the bandgap for materials where the lowest energy
minimum of the conduction band is not at the same k as the highest maximum of the
14 K.W. Böer and U.W. Pohl

a b c
E E E

-ħω Ec
Ec +ħω Ec

Ev Ev Ev

kx kx kx

Fig. 7 (a) Direct and (b) indirect transitions with absorption and (c) with emission of a phonon

valence band. For instance, the highest maximum of the valence band is at k =
0, but the lowest minimum of the conduction band is at k 6¼ 0 (Fig. 7). These
materials are indirect bandgap semiconductors; Si, Ge, GaP, AlAs, and AlSb are
examples. Most other semiconductors are direct gap semiconductors.
When, together with the photon, a phonon is absorbed, we obtain for the
absorption constant
ðð
e2 f BE jMj2 2dk1 2dk2
αabs
o ¼ δðEc ðk2 Þ  Ev ðk1 Þ  hν þ ℏωÞ; (26)
e0 nr cm20 2πν BZ ð2π Þ3 ð2π Þ3

where fBE is the Bose–Einstein distribution function for the phonon and M is the
matrix element for the simultaneous absorption of a phonon and a photon (Bassani
and Pastori Parravicini 1975). Using a parabolic band approximation and integrat-
ing over the delta function yields
 3=2  
e2 f BE jMj2 1 2mp 2mn 3=2  2
αabs ¼ hν  Eg þ ℏω : (27)
o
e0 nr cm20 2πν 8ð2π Þ3 ℏ2 ℏ2

A similar value for the absorption constant is obtained when a phonon is emitted,
except þℏω is replaced by ℏω and fBE is replaced by 1  fBE. The total optical
absorption of the indirect transition results as the sum

αo, ind ¼ αabs


o þ αo :
emi
(28)

It has a quadratic dependence on the photon energy (for more details, see Moss
et al. 1973):
 2  2
αo, ind / f BE hν þ ℏω  Eg þ ð1  f BE Þ hν  ℏω  Eg ; (29)

where ℏω is the energy of a phonon of proper momentum and energy. The two
branches of indirect transitions (Eq. 29 and Fig. 8) show a different temperature
Band-to-Band Transitions 15

a b
2
ao √a o
direct indirect

phonon
emission
phonon
absorption

Eg hν Eg-ħω Eg Eg+ħω hν

Fig. 8 Simple theoretical behavior of the absorption coefficient for (a) direct and (b) indirect
transitions

dependence. The phonon absorption branch vanishes at low temperatures, when the
phonons are not thermally excited (frozen-out).
Weak indirect transitions, resulting in a factor 103 lower absorption, can be
observed only in a wavelength range where they do not compete with direct
transitions. Indirect transitions are followed at higher energies by direct transitions
with a secondary band edge. Figure 9 shows an example for Ge where the indirect
band edge is preceding the direct edge.

2.2 Allowed and Forbidden Transitions

The matrix element (Eq. 2) contains two terms; the second one was neglected in the
previous discussions. The one described before relates to an allowed transition and
has a value close to 1 for its oscillator strength; here the selection rules are fulfilled.
The other term relates to a forbidden transition; it represents a transition in which
the Bloch function of the electron wave is orthogonal to the electric polarization of
the light. The forbidden transition becomes finite but small when the electron
momentum changes slightly during a transition from the ground to the excited
state, which is caused by the finite, small momentum of the photon. It becomes
important for the agreement between theory and experiment for the absorption
spectrum further away from the band edge (Johnson 1967). This is indicated in
Fig. 10b for the example of InSb.
The matrix element for such a forbidden transition may be calculated from the
k  p perturbation theory with

Mcv ðkÞ ¼ ðk  k0 Þ j@=@kðMcv ðkÞÞjk¼k0 : (30)

This transition-matrix element increases for larger values of k and therefore pro-
vides a larger contribution to the optical absorption (see Fig. 10b).
16 K.W. Böer and U.W. Pohl

a 4
b
10 Ge Ge
3 indirect

a 2o (105 cm–2)
10 phonon

(cm-½)
emission direct
a o (cm–1)

2 300 K 10 2
10
8

√a o
1 77 K 6
10 1
4
absorption √a o a 2o
0
10
2
0
0.6 0.7 0.8 0.9 1.0 0.6 0.7 0.8 0.9
hν (eV) hν (eV)
Fig. 9 (a) Absorption coefficient near the band edge for Ge (After Dash and Newman 1955). (b)
Analysis of the lower (101 . . . 101 cm1) and upper (101 . . . 5 102 cm1) part of the absorption
range at 300 K, giving evidence for indirect and direct absorption edges following each other at
0.66 and 0.81 eV, respectively (After Bube 1974)

a b
290 K
249 K
195 K
K

InSb
333

9 Si
8 104
ao hν (eV/cm)

7
(cm-½)

6
5
90 K 103
78 K
20 K Exp.
√ao

4 A
3 102 B
2 C
1 D
0 101
1.0 1.1 1.2 1.3 1.4 0.0 0.2 0.4 0.6 0.8
hν (eV) hν (eV)
Fig. 10 (a) Absorption coefficient of Si with temperature as the family parameter indicating
decreasing contribution of the phonon absorption branch with decreasing T (After MacFarlane and
Roberts 1955). (b) Absorption coefficient near the band edge of InSb; (A) direct allowed transi-
tions; (B) direct forbidden transitions; (C) as curve (A), however, corrected for nonparabolic bands;
(D) as curve (C), however, corrected for the k dependence of matrix elements (After Johnson 1967)

Forbidden transitions are always additional components for both direct and
indirect transitions. The absorption coefficient for these transitions varies as a
function of photon energy as (Bardeen et al. 1956)
 3=2
αo, dir, forb ¼ Adf hν  Eg : (31)
 3
αo, ind, forb ¼ Aif hν  ℏω  Eg : (32)
Band-to-Band Transitions 17

The proportionality coefficient for forbidden direct transitions is given for parabolic
bands as

8π 3 e2 ð2μÞ3=2
Adf ¼ j@=@kðMcv Þj2k¼k0 (33)
3e0 nr cm20 h5 ν

(Moss et al. 1973).

Transitions from Different Valence Bands


Transitions from the two upper valence bands, which are split at k = 0 in aniso-
tropic lattices, can be separated by using polarized light. For instance, the band-
edge transitions from the two valence bands, which are separated by crystal-field
splitting in CdS (Sect. 1.2.3 of chapter “▶ Bands and Bandgaps in Solids”), can be
identified. When polarized light with its electric vector perpendicular to the c axis is
used, transitions from one band are allowed, whereas transitions from the other
band are only allowed when the electric vector is parallel to the c axis (Fig. 11).
Transitions from the spin–orbit split-off band overlap and are more difficult to
separate because they cannot be turned off individually.
In anisotropic semiconductors, certain band degeneracies in valence and con-
duction bands are removed and the optical absorption depends on the relative
orientation of the crystal with respect to the light beam and its polarization. Only
such transitions which have a component of the dipole orientation in the direction of
the electrical vector of the light can be excited. Consequently, the absorption
spectrum becomes angle dependent: it is anisotropic. An example for such anisot-
ropy related to valence-band splitting was given above.

2.3 Band-to-Band Magnetoabsorption

In a magnetic field, band degeneracies are lifted. Band states are split into different
Landau levels. Such splitting is proportional to the magnetic induction B. It is
described in more detail in Sect. 2 of chapter “▶ Carriers in Magnetic Fields and
Temperature Gradients.” Direct band-to-band transitions at k ffi 0 in a magnetic
field are then given by transitions between the different Landau levels of the
valence and conduction bands.
For a simple two-band model, the splitting of the bands is shown in Fig. 12a. The
allowed transitions near k = 0 are – for the field Bz along the z axis – given by

  ℏ2 k 2
ΔE ¼ Eg þ nq þ 12 ðℏ ωe þ ℏ ωh Þ þ Δmj ðge þ gh Þ μB Bz þ
2μ (34)
  ℏ e Bz ℏ 2 2
k
¼ Eg þ nq þ 12 þ Δmj gtot μB Bz þ , nq ¼ 0, 1, 2, 3, . . . ;
μ 2μ

where nq is the quantum number denoting the Landau levels (equal in the valence and
conduction band for allowed transitions),ℏωe, h ¼ eB=me, h is the cyclotron-resonance
18 K.W. Böer and U.W. Pohl

Fig. 11 Absorption edge of λ (nm)


CdS measured in polarized 520 500 480
light at various temperatures
105
(After Dutton 1958)
CdS
487.4 nm
484.4 nm
104

ao (cm–1) 103

202 K
342 K

102 90 K
300 K

2.4 2.5 2.6


hν (eV)

nq
a b c d
E E 2 E
cb mj
1 j=1/2
+1/2
0 s-type -1/2
l=0
ħωe ½ħωe s=1/2 s- σ+
3 1 2

Eg(B=0) Eg(B) Eg(B=0) +3/2


+1/2
½ħωh -1/2
-3/2
ħωh nq hh j=3/2
0 lh j=3/2 +1/2
p-type -1/2
1
kz kz 2 l=1 so k
j=1/2
s=1/2

Fig. 12 (a) Band structure near k = 0 for single bands (a) without and (b) with a magnetic
induction B; B splits the bands into equidistant Landau levels. (c) Band structure for zincblende
semiconductors. (d) Allowed transitions for spin-split Landau levels of one pair of valence and
conduction bands with equal quantum number nq. Positive g factors are assumed for the splitting;
numbers at the transitions indicate the absorption-intensity ratio

energy for electrons or holes, and Δmj = Δ(ml + sz) is the difference in the respective
azimuthal plus spin quantum numbers of the two bands; μB is the Bohr magneton, and

1
μ is the reduced effective mass m1 n þ m1
p . Here g is the respective Landé
g factor (see also Sect. 2 of chapter “▶ Carriers in Magnetic Fields and Temperature
Gradients”), which is derived from the g tensor (Yafet 1963); it depends on the angle θ
between the magnetic induction and the principal axis of the band ellipsoid:
Band-to-Band Transitions 19

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
g¼ g2k cos2 θ þ g2⊥ sin2 θ, where gk ¼ gzz and g⊥ ¼ gxx ¼ gyy .

The transitions are subject to selection rules which depend on the polarization of
the radiation. These selection rules are

For EkB : Δkz ¼ 0, Δnq ¼ 0, Δmj ¼ 0


(35)
For E⊥B : Δkz ¼ 0, Δnq ¼ 0, Δmj ¼ 1;

for right and left circular polarized light with Δmj = +1 and 1 indicated as σ + and
σ , respectively. The intensity ratio of the different transitions is indicative of the
original degree of spin orientation. For instance, for equidistributed spin of conduc-
tion electrons, the intensity ratio of absorption for right polarized light (32 ! 12 to
12 ! þ12) is 3:1.
In typical semiconductors, the valence band is p-like and the conduction band is
s-like (Fig. 12c). There are hence four possibilities at k = 0 for the angular
momentum quantum number mj = ml + sz of the valence band: mj = 3/2, 1/2,
1/2, and 3/2. Those of the conduction band are 1/2 and +1/2. For E⊥B and sz =
1/2, we have transitions from mj = +1/2 and 3/2; for sz = 1/2, we have
transitions from mj = +3/2 and 1/2. The permitted transitions are shown in
Fig. 12d; such scheme applies for each pair of Landau levels with equal quantum
numbers nq for valence and conduction band.
The general behavior can be seen in Fig. 13a with Landau levels of Ge at room
temperature identified. They shift linearly with the magnetic induction (Fig. 13b)
and permit a measurement of the reduced effective mass μ1 from the slopes, and
the bandgap from an extrapolation4 to B ! 0. At low temperatures, the transitions
from the light and heavy-hole bands and from the split-off band are resolved
(Zawadzki and Lax 1966; Roth et al. 1959). Oscillatory magnetoabsorption was
observed for many semiconductors, e.g., for GaAs (Vrehen 1968), InN (Millot
et al. 2011), InP (Rochon and Fortin 1975), GaSe (Watanabe et al. 2003), or InSe
(Millot et al. 2010). An analysis of such spectra also provides information about
band anisotropies (Sari 1972). For a review, see Mavroides (1972). Interband
transitions between Landau levels were also observed in emission spectra (Gubarev
et al. 1993).

3 Transitions in Quantum Wells

One of the most direct confirmations of the simple quantum-mechanical model is


obtained from the measurement of the optical spectra in low-dimensional structures
(quantum wells, quantum wires, or quantum dots). Since the limited interaction

4
For semiconductors without inversion symmetry (e.g., GaAs), the valence-band extrema do not
occur exactly at k = 0, and the extrapolations of the measured lines in Fig. 13b do not precisely
meet in one point at B = 0; a detailed analysis is given by Zwerdling et al. (1957).
20 K.W. Böer and U.W. Pohl

a 0.45 b 0.88
0.40
Ge
0.87 7

0.35 0.86 6
Transmission

B (T) 0.85

hν (eV)
0.30 5
0.44 0.84 4
0.25
3.0 0.83
3
0.20 3.6 0.82 2
0.15 0.81 1
direct energy gap
Eg
0.78 0.80 0.82 0.84 0.86 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
hν (eV) B (T)

Fig. 13 (a) Transmission spectrum of Ge with minima at absorbing transitions. (b) Energy of
principal minima in panel (a) as a function of the magnetic induction (After Zwerdling et al. 1957)

volume of a single structure leads to only weak absorption, usually either absorp-
tion spectra of multiple similar structures5 or photoluminescence excitation spectra
are measured.

3.1 Energy Levels in Multiple Quantum Wells

Optically induced absorption transitions in multiple quantum wells give an instruc-


tive picture of the energy-level structure and reflect selection rules.6 A single
quantum well contains a set of energy levels that can approximately be calculated
from a square-well potential with infinite barrier height, yielding

ℏ2 π 2 2
En ¼ n , n ¼ 1, 2, 3, . . . : (36)
2mn l2QW

See Eq. 54 of chapter “▶ Bands and Bandgaps in Solids.”


When multiple wells in a periodic superlattice are separated by barriers with a
small thickness, their states interact and the energy levels start to split. Eventually,
the interaction of many wells leads to mini-bands (see Sect. 3.1.2 of chapter
“▶ Bands and Bandgaps in Solids”). A development of mini-bands is shown in
Fig. 14. Two quantum wells are formed by quantum well (QW) layers of GaAs and
are separated by one barrier (B) layer of Al0.35Ga0.65As. In a sequence of experi-
ments, the well or barrier width was varied. For each well the transitions show a
shift toward higher energies according to Eq. 36 and wider spacing of the lines with

5
Such structures are multiple quantum wells or arrays of quantum wires with a sufficient thickness
of the separating barrier material or ensembles of quantum dots.
6
For a study of quantum wires, see, e.g., Ihara et al. (2007).
Band-to-Band Transitions 21

a b c
lQW E
2 QWs 2 QWs
lB = 10 ML lQW = 10 ML
Ec
10
lB
20 3
Transmissivity T

5
30 7
50 10
15
20

lQW lB lQW

0.00 0.08 0.16 0.24 0.32 0.00 0.08 0.16 0.24 0.32 x
Energy (eV) Energy (eV)

Fig. 14 Optical transmission spectrum of two GaAs quantum wells coupled by one
Al0.35Ga0.65As barrier layer with (a) variation of the well thickness lQW and (b) variation of the
barrier thickness lB expressed in numbers of monolayers (ML; 10 ML ffi 28 Å) and showing the
split into two levels for each quantum state. (c) Schematics of the well and barrier geometry for a
30/10/30 ML structure, the third curve in (a) (After Torabi et al. 1987)

decreased well width lQW. In addition, the lines split and the splitting increases with
reduced barrier width lB, as shown in Fig. 14b. With decreasing barrier width, these
lines also become broader (Schulman and McGill 1981); here, the interaction
between the wells becomes more probable (see Sect. 3.1.2 of chapter “▶ Bands
and Bandgaps in Solids”). The observation agrees with the calculation of a simple
one-dimensional square well/barrier potential.
When more wells in a periodic superlattice structure can interact, the lines
spread to mini-bands: due to the interaction of a sufficient number of layers, the
discrete features shown in Fig. 14 are broadened into continuous mini-bands, as
indicated in a sequence of curves in Fig. 15 and discussed in the following section.
The mean free path of electrons, however, gives their coherence length across
which superlattice periodicity is recognized.

3.2 Absorption in Quantum Wells

When the well width is very large, the optical absorption spectrum for superlattices
is identical to that of the bulk (well) material (Fig. 15, uppermost curve). With
decreasing well width, the onset of optical absorption shifts to higher energies, and
relatively sharp lines appear within the band of the well material, as shown in the
second and third curves. The lines correspond to the eigenstates of conduction
electrons in this well, which for a simple square well are given in Eq. 36, indicating
a 1/l2QW dependence (Dingle 1975; Miller et al. 1980; Pinczuk and Worlock 1982).
The bound eigenstates penetrate only about 25 Å into the AlGaAs barriers, which
are 250 Å thick and hence separate the wells. The doublets shown in this figure are
22 K.W. Böer and U.W. Pohl

Fig. 15 Optical absorption


spectrum of GaAs/GaAlAs GaAs/AlGaAs SL lQW
superlattices with different
lattice constants. The doublets 4000 Å
are due to transitions from
light and heavy hole mini-
bands. The lQW = 4000 Å n=1
n=2 n=3 n=4
sample shows bulk properties
hh lh
with the exciton peak at hh lh hh lh
hh lh 192 Å
1.513 eV – see Fig. 8 of

αo (arb. units)
chapter “▶ Excitons” (After
Dingle et al. 1974)

n=1
n=2
hh
lh
hh lh
116 Å
n=1
hh lh

50 Å

1.50 1.55 1.60 1.65 1.70 1.75


Energy (eV)

due to transitions from the light- and heavy-hole mini-bands which were omitted in
Fig. 16 for clarity. The lines become broader and mini-bands develop as the wells
and the barrier become thinner (lowest curve in Fig. 15).
The envelope functions of the different eigenstates of valence and conduction
electrons within the well are shown in Fig. 16a for a deep, isolated quantum well.
For the optical transitions, the figure indicates maximum overlap between
eigenfunctions of the same subband index n, while transitions between subbands
of different indices have rather small transition probabilities. This is the basis for
the corresponding selection rules.
The two-dimensional joint density-of-state function (Fig. 16b) is steplike
for quantum wells (or superlattices with wide barriers). These steps become
smooth when mini-bands develop and broaden with decreasing barrier width –
see Fig. 38 of chapter “▶ Bands and Bandgaps in Solids.” Due to excitonic
contributions to absorption (see Sect. 2.1 of chapter “▶ Excitons”), peaks occur
slightly below each subband threshold, which causes a modified, increased absorp-
tion at each lower edge shown by the solid curve of Fig. 16b. This is in general
agreement with the experiment – see Fig. 15 and also Sect. 2.1 of chapter
“▶ Excitons.”
Band-to-Band Transitions 23

a b
E E,hν

e3

ΔEc e2
Jcv,2D(E) ao(hν)
e1

lQW

h1
ΔEv h2
h3

x Jcv,2D(E), αo(hν)

Fig. 16 (a): Envelope functions of three eigenstates in a single quantum well. (b) Corresponding
joint density of states from valence- to conduction-band transitions (dashed curve) and absorption
coefficient including a simple excitonic feature preceding each step (solid curve)

The confinement of the electron eigenfunctions in narrow mini-bands yields high


oscillator strength for optical transitions. The optical absorption into the mini-bands
is a direct one, even though the well material may have an indirect bandgap.

4 Optical Bandgap of Amorphous Semiconductors

4.1 Intrinsic Absorption

The optical bandgap of amorphous semiconductors is less well defined than in a


crystalline material (Sect. 4.2 of chapter “▶ Bands and Bandgaps in Solids”), due to
substantial tailing of defect states into the bandgap (Mott and Davis 1979;
Sa-yakanit and Glyde 1987; see also Sect. 1 of chapter “▶ Defects in Organic and
Amorphous Semiconductors”). Usually, an effective bandgap is taken from the
optical absorption spectrum plotted in an (αo hv)1/2 versus (hvEg) presentation and
extrapolated to αo = 0 or, for practical purposes in thin films, presented as E04 – the
energy is used at which αo reaches a value of 104 cm1. Plotting the absorption
coefficient versus bandgap energies in a semilogarithmic graph, we obtain a straight
line and deduce from its slope the Urbach parameter E0:
 
hν  Eg
αo ðνÞ ¼ αo, 0 exp ; (37)
E0
24 K.W. Böer and U.W. Pohl

106
α-Si:H

105
a o (cm–1)

104
Ttreat (K)

0K
60 0K
55 0K
3 50
3K

10 Tmeas (K)
29

1K
.7K
15
12

102
1.4 1.6 1.8 2.0 2.2
hν (eV)
Fig. 17 Optical absorption of α-Si:H as a function of the photon energy with measurement
temperature Tmeas or treatment temperature Ttreat indicated, pointing to an additive effect of
temperature and structural disorder (After Cody et al. 1981)

which gives the steepness of the level distribution near the band edge.7
A particularly striking example of band tailing of α-Si:H is given in Fig. 17
which demonstrates the relation between an increased severity of disorder and the
slope of the band tailing (Cody et al. 1981). Such increased severity is caused by a
heat treatment of α-Si:H at temperatures above 700 K, when hydrogen is released:
thereby dangling bonds are created which act as major defects (see Sect. 2 of
chapter “▶ Defects in Organic and Amorphous Semiconductors”). The tailing,
described by Eq. 37, increases from E0 = 50–100 meV with increasing treatment
temperature. Recent modeling indicates that the bandgap energy is not fixed but
depends somewhat on hydrogen saturation (Legesse et al. 2014).
Another example for hydrogenated (or fluorinated) amorphous Si and Si1xGex
alloys is shown in Fig. 18. The α-Si:H shows a relatively steep absorption edge,
indicating direct absorption; the slope decreases with alloying.
The hydrogenated α-Si1xGex alloys show little bowing in spite of the large lattice
mismatch between Si and Ge. This is characteristic for one-atomic semiconductors which
show less restraint against lattice deformation than two-atomic or higher-atomic lattices.

7
An exponential Urbach tail (with Urbach energy in the meV range) was also observed in
absorption spectra of high-quality GaAs/AlGaAs quantum wells (Bhattacharya et al. 2015). The
broadening is assigned to disorder originating from the electric field of zero-point oscillations of
LO phonons in polar semiconductors, yielding a fundamental limit to the Urbach slope.
Band-to-Band Transitions 25

a 105 hydride fluoride Ge%


b 2.0
0, 0
26, 32 1.8
104 37, 38
E04
53, 51
1.6
103 Si1-xGex:H:F
1.4 α-Si1-xGex:H
Eg
≈ α-Si1-xGex:H:F ≈
ao (cm–1)

102

E (eV)
0.07
101 E0
0.06
100

0.05
10–1

10–2 0.04
0.7 0.9 1.1 1.3 1.5 1.7 1.9 0 0.2 0.4 0.6
hν (eV) Si x Si0.4Ge0.6

Fig. 18 (a) Optical absorption spectra of α-Si1xGex:H and α-Si1xGex:F. (b) Variation of the
bandgap energies with composition (After Mackenzie et al. 1988)

4.2 Extrinsic Absorption in Glasses

The absorption of amorphous materials below the band edge decreases rapidly,
following the Urbach tail, and reaches values that can be much lower than those for
high-purity crystals. It is difficult to make crystals with very low defect densities,
while in amorphous structures, most defects, i.e., wrong bonds or impurities, are
incorporated into the host material and yield a much lower optical absorption far
from the absorption edge. This makes glasses good materials for optical fibers in
communication systems.
Fused silica (SiO2) is used at a length of up to 30 km between repeater stations
with an absorption of 0.16 dB/km,8 corresponding to an absorption constant of
3.7 107 cm1, compared to the lowest absorption of high-purity GaAs of 2
103 cm1 (Lines 1986). Optical absorption and pulse spreading because of finite
dispersion (@nr/@λ) are the limiting factors for the length of an optical transmission
line (Lines 1986). Absorption losses are usually caused by larger inclusions,
imperfections, and impurities and require purification in the range 1013 cm3, i.e.,
in the parts per billion (ppb) range. The remaining losses in SiO2 fibers are due to
water contamination, which is difficult to remove.

8
1 dB (decibel) is equal to the logarithm (base ten) of the ratio of two power levels having the value
0.1, that is, log10( pl/p0) = 0.1 or pl/p0 = 1.259. For the example given above of 0.16 dB/km, one
measures a reduction of the light intensity by 3.8 % per km. The intensity is reduced to 33 % after
30 km.
26 K.W. Böer and U.W. Pohl

Fig. 19 Attenuation of light λ (μm)


in intrinsic glasses as a 10 5 3 2 1.5 1.2 1 0.8
function of the wavelength for
4
a hypothetical glass (After
Lines 1986)

log αo (dB/km)
2

1
Rayleigh
multiphonon Urbach
0

–1
0 0.5 1.0 1.5
1/λ (mm–1)

When fibers are prepared well enough so that only intrinsic phenomena limit the
absorption, then three effects have to be considered:

1. The Urbach tail, responsible for electronic transitions, and extending in energy
from the bandgap downward
2. Multiphonon absorption, extending from the Reststrahl frequency upward
3. Rayleigh scattering (see Sect. 3.3.1 of chapter “▶ Photon-Phonon Interaction”),
due to density fluctuation in the midfrequency range, as shown in Fig. 19

The lowest intrinsic absorption values predicted for SiO2 are 0.1 dB/km at λ =
1.55 μm with 90 % of the attenuation stemming from Rayleigh scattering, 10 %
from multiphonon absorption, and negligible contribution from the Urbach tail.

5 Summary

The theory of optical transitions between the valence and conduction bands was
sketched in rather general terms. Fermi’s golden rule yields the transition proba-
bility from the momentum matrix-element and the joint density of states. The joint
density of states has characteristic kinks or maxima at critical points; they occur at
energies where the k-depending slopes of the contributing bands are equal. For
transitions near the band edge, the theory can be simplified by assuming parabolic
band shapes. This effective-mass approach yields quantitative expressions for the
dielectric function and the absorption as a function of the photon energy. In
quantum wells, the two-dimensional joint density of states leads to a steplike
Band-to-Band Transitions 27

increase of the absorption for increasing photon energy, modified by additional


excitonic absorption at the step edges.
Depending on the conduction-band behavior, one has strong direct or weak
indirect transitions at the band edge; the latter require the participation of a phonon
to fulfill energy and momentum conservation. Contributions of forbidden transi-
tions additionally modify the absorption further away from the band edge.
Depending on the slope of the optical absorption in the intrinsic range near the
band edge, direct or indirect transitions can be identified unambiguously, and, with
the help of an external magnetic field, the effective mass can be determined from
the period of magneto-oscillations of the near-edge absorption due to Landau
splitting of the band states.
Deviations from the ideal, periodic crystal lattice provide tailing states extending
beyond the band edge, usually as an Urbach tail which decreases exponentially with
distance from the band edge. The resulting tailing of the absorption coefficient
below the bandgap energy is particularly pronounced for amorphous
semiconductors.

References
Adagi S (2005) Properties of group-IV, III-V and II-VI semiconductors. Wiley, Chichester
Albrecht S, Reining L, Del Sole R, Onida G (1998) Ab initio calculation of excitonic effects in the
optical spectra of semiconductors. Phys Rev Lett 80:4510
Aulbur WG, Jonsson L, Wilkins JW (2000) Quasiparticle calculations in solids. Solid State Phys
54:1
Bardeen J, Blatt FJ, Hall LH (1956) Indirect transition from the valence to the conduction band. In:
Breeckenridge R, Russel B, Halm T (eds) Proceedings of photoconductivity conference.
Wiley, New York, p 146
Bassani GF (1966) Methods of band calculations applicable to III–V compounds. In: Willardson
RK, Beer AC (eds) Semiconductors and semimetals, vol 1. Academic Press, New York, p 21
Bassani GF, Pastori Parravicini G (1975) Electronic states and optical transitions in solids.
Pergamon Press, Oxford
Bhattacharya R, Mondal R, Khatua P, Rudra A, Kapon E, Malzer S, Döhler G, Pal B, Bansal B
(2015) Measurements of the electric field of zero-point optical phonons in GaAs quantum wells
support the Urbach rule for zero-temperature lifetime broadening. Phys Rev Lett 114:047402
Brust D, Phillips JC, Bassani F (1962) Critical points and ultraviolet reflectivity of semiconduc-
tors. Phys Rev Lett 9:94
Bube RH (1974) Electronic properties of crystalline solids. Academic Press, New York
Cody GD, Tiedje T, Abeles B, Brooks B, Goldstein Y (1981) Disorder and the optical absorption
edge of hydrogenated amorphous silicon. Phys Rev Lett 47:1480
Cohen ML, Chelikowsky JR (1988) Electronic structure and optical properties of semiconductors.
Springer, Berlin
Dash WC, Newman R (1955) Intrinsic optical absorption in single-crystal germanium and silicon
at 77
K and 300
K. Phys Rev 99:1151
Dingle R (1975) Confined carrier quantum states in ultrathin semiconductor hetero-structures. In:
Queisser HJ (ed) Festkörperprobleme, vol 15, Advances in solid state physics. Pergamon/
Vieweg, Braunschweig, p 21
Dingle R, Wiegmann W, Henry CH (1974) Quantum states of confined carriers in very thin
AlxGa1xAs-GaAs-AlxGa1xAs heterostructures. Phys Rev Lett 33:827
Dutton D (1958) Fundamental absorption edge in cadmium sulfide. Phys Rev 112:785
28 K.W. Böer and U.W. Pohl

Gubarev SI, Ruf T, Cardona M, Ploog K (1993) Resonant magneto-luminescence of high quality
GaAs. Solid State Commun 85:853
Hanke W, Sham LJ (1974) Dielectric response in the Wannier representation: application to the
optical spectrum of diamond. Phys Rev Lett 33:582
Hybertsen MS, Louie SG (1985) First-principles theory of quasiparticles: calculation of band gaps
in semiconductors and insulators. Phys Rev Lett 55:1418
Ihara T, Hayamizu Y, Yoshita M, Akiyama H, Pfeiffer LN, West KW (2007) One-dimensional
band-edge absorption in a doped quantum wire. Phys Rev Lett 99:126803
Johnson EA (1967) Absorption near the fundamental edge. In: Willardson RK, Beer AC (eds)
Semiconductors and semimetals, vol 3. Academic Press, London
Kane EO (1957) Band structure of indium antimonide. J Phys Chem Solids 1:249
Lautenschlager P, Garriga M, Vina L, Cardona M (1987a) Temperature dependence of the
dielectric function and interband critical points in silicon. Phys Rev B 36:4821
Lautenschlager P, Garriga M, Logothetidis S, Cardona M (1987b) Interband critical points of
GaAs and their temperature dependence. Phys Rev B 35:9174
Lax M, Hopfield JJ (1961) Selection rules connecting different points in the Brillouin zone. Phys
Rev 124:115
Legesse M, Nolan M, Fagas G (2014) A first principles analysis of the effect of hydrogen
concentration in hydrogenated amorphous silicon on the formation of strained Si-Si bonds
and the optical and mobility gaps. J Appl Phys 115:203711
Lines ME (1986) Ultralow-loss glasses. Annu Rev Mater Sci 16:113
Lyon SA (1986) Spectroscopy of hot carriers in semiconductors. J Lumin 35:121
Macfarlane GG, Roberts V (1955) Infrared absorption of silicon near the lattice edge. Phys Rev
98:1865
Mackenzie KD, Burnett JH, Eggert JR, Li YM, Paul W (1988) Comparison of the structural,
electrical, and optical properties of amorphous silicon-germanium alloys produced from
hydrides and fluorides. Phys Rev B 38:6120
Madelung O (1981) Introduction to solid state theory. Springer, Berlin/New York
Martinez G, Schl€uter M, Cohen ML (1975) Electronic structure of PbSe and PbTe II – optical
properties. Phys Rev B 11:660
Mavroides JG (1972) Magneto-optical properties. In: Abeles F (ed) Optical properties of solids.
North Holland, Amsterdam
Miller RC, Kleinman DA, Nordland WA Jr, Gossard AC (1980) Luminescence studies of optically
pumped quantum wells in GaAs-AlxGa1-xAs multilayer structures. Phys Rev B 22:863
Millot M, Broto J-M, George S, González J, Segura A (2010) Electronic structure of indium
selenide probed by magnetoabsorption spectroscopy under high pressure. Phys Rev B
81:205211
Millot M, Ubrig N, Poumirol J-P, Gherasoiu I, Walukiewicz W, George S, Portugall O, Léotin J,
Goiran M, Broto J-M (2011) Determination of effective mass in InN by high-field oscillatory
magnetoabsorption spectroscopy. Phys Rev B 83:125204
Moss TS, Burrell GJ, Ellis B (1973) Semiconductor optoelectronics. Wiley, New York
Mott NF, Davis EA (1979) Electronic processes in non-crystalline materials. Claredon Press,
Oxford, UK
Phillips JC (1956) Critical points and lattice vibration spectra. Phys Rev 104:1263
Phillips JC (1966) The fundamental optical spectra of solids. Solid State Phys 18:55
Pinczuk A, Worlock JM (1982) Light scattering by two-dimensional electron systems in semi-
conductors. Surf Sci 113:69
Rochon P, Fortin E (1975) Photovoltaic effect and interband magneto-optical transitions in InP.
Phys Rev B 12:5803
Roth LM, Lax B, Zwerdling S (1959) Theory of optical magneto-absorption effects in semi-
conductors. Phys Rev 114:90
Sari SO (1972) Excitonic effects in Landau transitions at the E1 edges of InSb and GaSb. Phys Rev
B 6:2304
Band-to-Band Transitions 29

Sa-yakanit V, Glyde HR (1987) Urbach tails and disorder. Comments Condens Matter Phys 13:35
Schulman JN, McGill TC (1981) Complex band structure and superlattice electronic states. Phys
Rev B 23:4149
Shkrebtii AI, Ibrahim ZA, Teatro T, Richter W, Lee MJG, Henderson L (2010) Theory of the
temperature dependent dielectric function of semiconductors: from bulk to surfaces. Applica-
tion to GaAs and Si. Phys Status Solidi B 247:1881
Torabi A, Brennan KF, Summers CJ (1987) Photoluminescence studies of coupled quantum well
structures in the AlGaAs/GaAs system. Proc SPIE 0792:152
van Hove L (1953) The occurrence of singularities in the elastic frequency distribution of a crystal.
Phys Rev 89:1189
Vrehen QHF (1968) Interband magneto-optical absorption in gallium arsenide. J Phys Chem
Solids 29:129
Watanabe K, Uchida K, Miura N (2003) Magneto-optical effects observed for GaSe in megagauss
magnetic fields. Phys Rev B 68:155312
Yafet Y (1963) g factors and spin-lattice relaxation of conduction electrons. In: Seitz F, Turnbull D
(eds) Solid state physics, vol 14. Academic Press, New York, p 1
Zawadzki W, Lax B (1966) Two-band model for Bloch electrons in crossed electric and magnetic
fields. Phys Rev Lett 16:1001
Zunger A (1983) One-electron broken-symmetry approach to the core-hole spectra of semicon-
ductors. Phys Rev Lett 50:1215
Zwerdling S, Lax B, Roth LM (1957) Oscillatory magneto-absorption in semiconductors. Phys
Rev 108:1402

You might also like