Carrier-Transport Equations: Karl W. Böer and Udo W. Pohl
Carrier-Transport Equations: Karl W. Böer and Udo W. Pohl
Contents
1 Carriers in Semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1 Bloch Electrons and Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 The Polaron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Conductivity and Mobility of Carriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.1 Electronic Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Electron Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3 Currents and Electric Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1 Drift Current in an Electric Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Diffusion Currents and Total Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.3 Electrochemical Fields and Quasi-Fermi Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4 Carrier Distributions in External and Built-In Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4 The Boltzmann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.1 The Boltzmann Equation for Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2 The Boltzmann Equation for Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.3 The Relaxation-Time Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.4 Carrier Scattering and Energy Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.5 The Mobility Effective Mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.6 Momentum and Energy Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.7 Phonon and Electron Drag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
K. W. Böer
Naples, FL, USA
e-mail: [email protected]
U. W. Pohl (*)
Institut für Festkörperphysik, EW5-1, Technische Universität Berlin, Berlin, Germany
e-mail: [email protected]
Abstract
The current in semiconductors is carried by electrons and holes. Their lattice
polarization modifies the effective mass, expressed as a change to polarons. While
for large polarons the effect is small, semiconductors with narrow bands and large
lattice polarization show a significant effect described by small polarons. The total
current is composed of a drift and a diffusion current of electrons and holes. The drift
current is determined by the electric field, and the energy obtained by carrier
acceleration is given to the lattice by inelastic scattering, which opposes the energy
gain, causing a constant carrier-drift velocity and Joule’s heating. The diffusion
current is proportional to the carrier gradient up to a limit given by the thermal
velocity. Proportionality factor of both drift and diffusion currents is the carrier
mobility, which is proportional to a relaxation time and inverse to the mobility
effective mass. Currents are proportional to negative potential gradients, with the
conductivity as the proportionality factor. In spatially inhomogeneous semiconduc-
tors, both an external field, impressed by an applied bias, and a built-in field, due to
space-charge regions, exist. Only the external field causes carrier heating by shifting
and deforming the carrier distribution from a Boltzmann distribution to a distorted
distribution with more carriers at higher energies.
The Boltzmann equation permits a detailed analysis of the carrier transport and the
carrier distribution, providing well-defined values for transport parameters such as
relaxation times. The Boltzmann equation can be integrated in closed form only for a
few special cases, but approximations for small applied fields provide the basis for
investigating scattering processes; these can be divided into essentially elastic pro-
cesses with mainly momentum exchange and, for carriers with sufficient accumu-
lated energy, into inelastic scattering with energy relaxation.
Keywords
Boltzmann transport equation Built-in electric field Carrier heating Collision
integral Conductivity Diffusion current Drift current Drift velocity
Effective mass Einstein relation Energy relaxation Fröhlich coupling
Inelastic and elastic scattering Joule’s heating Mean free path Mobility
effective mass Momentum relaxation Polaron Polaron mass Polaron self-
energy Quasi-Fermi level Relaxation time Relaxation-time approximation
1 Carriers in Semiconductors
Electrons and holes in the conduction and valence bands are quasi-free to move in space
and energy and to accept energy from an external field. Carrier transport in semi-
conductors with nonideal periodicity1 is subjected to scattering with a mean free path
1
A crystal with nonideal lattice periodicity is a solid which contains lattice defects (e.g., impurities)
and oscillatory motions (i.e., phonons).
Carrier-Transport Equations 3
λ between scattering events. Such scattering reduces the effective volume to a value of
λ3, in which there is coherence of the electron wave. Only within such limited volume
does any one electron experience lattice periodicity and is nonlocalized. Scattering
introduces a loss, i.e., a damping mechanism counteracting the energy gain from an
electric field. All scattering events change the carrier momentum, i.e., its direction of
motion. However, only some of them, the inelastic scattering events, significantly
change the energy of the carrier. Usually, several elastic scattering events are followed
by one inelastic event after the carrier has gained sufficient energy from the field to
permit inelastic scattering, usually by generating optical phonons. A large variety of
scattering events can be distinguished; they are listed in Sect. 4.4 and discussed in
chapter ▶ “Carrier Scattering at Low Electric Fields”. The sum of all of these determines
the carrier motion, its mobility, which will be defined later.
First a quasiclassical picture will be used in this chapter to describe the basic
elements of the carrier motion in a semiconductor. The quantum-mechanical part is
incorporated by using an effective mass rather than the rest mass, i.e., by dealing with
Bloch electrons or Bloch holes as quasiparticles. Before we discuss carrier transport, we
need to specify in a more refined model what we mean by carriers in semiconductors.
The excitation of electrons from the valence band (e.g., by absorption of photons)
creates a certain concentration n of electrons in the conduction bands. These
electrons interact with lattice imperfections, such as phonons, impurities, or other
deviations from an ideal periodicity. This interaction is termed a scattering event.
The scattering tends to bring the electrons into thermal equilibrium with the lattice
and, in doing so, to the lowest valley of the lowest conduction band.
Near the bottom of the conduction band, the electron is described as a Bloch
electron,2 i.e., as an electron with an effective mass given by the curvature of
E(k), as defined in Sect. 2.2 of chapter ▶ “The Origin of Band Structure”, Eq. 20.
In an analogous description, the hole is described as a Bloch-type quasiparticle,
residing near the top of the uppermost valence bands, with an effective mass given
by their curvatures – see Sect. 1.1.1 of chapter ▶ “Bands and Bandgaps in Solids”.
The band picture, however, which is the basis for this discussion, results from a series
of approximations listed in chapter ▶ “Quantum Mechanics of Electrons in Crystals”.
The most severe one is the adiabatic approximation, which limits the electron-phonon
interaction. Within the band model, the Bloch electron interacts with a static potential of
the nuclei, while only the electrons surrounding each nucleus are polarized dynamically.
With sufficient coupling between the moving electron and the lattice (i.e., a sufficiently
large coupling constant ac – see Sect. 3 1.2.1, this is no longer justified. For these cases,
2
We are adopting here the picture of a localized electron. Such localization can be justified in each
scattering event. In this model, we use a gas-kinetic analogy with scattering cross-sections, e.g., for
electron-phonon interaction. This is equivalent to a description of the interaction of delocalized
electrons and phonons when calculating scattering rates.
4 K. W. Böer and U. W. Pohl
therefore, the Bloch electron picture needs to be augmented. The corresponding quasi-
particle derived from a higher approximation is the polaron.
The previous discussion of free carriers with an effective mass was based on an ideal
periodic lattice. There are many reasons why a real crystal lattice shows deviations
from this periodicity. The perturbations, which cause major changes in the carrier
trajectories, can be described as local scattering centers discussed in chapter ▶ “Car-
rier Scattering at Low Electric Fields”.
Another interaction that steadily accompanies the carrier throughout its entire
motion,3 however, is better incorporated in the effective-mass picture. It also has an
influence on the scattering effectiveness and thus on the mobility, which will be
discussed in this section. This interaction involves lattice polarization and causes
deviations from the Bloch electron picture discussed above. Deviations are signifi-
cant in lattices with a large coupling of carriers to the lattice, such as in ionic crystals
(typically alkali halides), crystals with a strong ionic character and large bandgap
energy (e.g., transition-metal oxides), or organic crystals. Also many typical semi-
conductors contain at least a fraction of ionic bonding, especially compound semi-
conductors with higher iconicity (e.g., nitrides or II-VI compounds). Moreover, a
strong polarization of the lattice occurs also in the neighborhood of certain impurities
or in highly disordered semiconductors. In all these cases the interaction of electrons
with lattice polarization becomes important. Such interaction can be static (for an
electron at rest) or dynamic (accompanying a moving electron).
The interaction involves the part of the Hamiltonian not considered in the band
theory when describing the Bloch electron. This adiabatic approximation neglects
the interaction of electrons with the induced motion of lattice atoms. In order to
include the motion due to the lattice polarization by the electron, several approaches
can be taken, all of which relate to the coupling of the electron with the lattice. Such
coupling can be expressed as (Hayes and Stoneham 1984):
3
This interaction is also commonly described as an electron-phonon interaction, however, of a
different kind than that responsible for scattering. As a lattice deformation relates to phonons, the
interaction of electrons with the lattice causing a specific deformation can formally be described by
continuously absorbing and emitting phonons (see the following sections).
Carrier-Transport Equations 5
a b
E
Estrain
e ε
Epol
Eion
Fig. 1 (a) Schematic of the polarization cloud in the vicinity of an electron (e) in a ionic lattice.
Gray circles indicate regular lattice sites. (b) Polaron energy Epol given by the sum of elastic strain
energy Estrain and electrical polarization energy Eion
where e describes the relative displacements of the ions within the deformation
volume ΔV and the constant C1 is the energy density for a strain e = 1. This energy
cost is overbalanced by an energy gain Eion due to the polarization:
The constant C2 is an energy density, which describes the change of the charge
balance at a strain e = 1. The sum of Eqs. 1 and 2 yields the net energy gain Epol by
the formation of the polaron state as shown in Fig. 1b:
For a sufficiently large coupling to the lattice, the self-energy is large enough so
that the electron is actually trapped in its own potential well: it becomes a self-
trapped electron. This was suggested by Landau (1933) and Frenkel (1936) for bulk
crystals and later reported also in semiconductor confined systems such as quantum
wires (e.g., Muljarov and Tikhodeev 1997). Pekar (1954) pointed out, however, that
when the electron moves to a neighboring lattice position, the same trap level
appears. Thus, one can describe this as a continuous virtual level or a band below
the conduction band in which the electron can move. This picture applies for lattices
with smaller coupling constant.
The state of the electron with its surrounding polarization cloud can then be
described as a quasiparticle, the polaron. The distinguishing parameter between the
self-trapped and the mobile polaron is the strength of the electron-lattice interaction,
which also can be related to the polaron size. The small polaron is tightly bound and
self-trapped: it moves via hopping between neighboring ions. The large polaron
moves much like an electron described by the Boltzmann equation with scattering
events, but with a larger effective mass caused by the polarization cloud carried
along by the polaron (for a review, see Christov 1982). One can also describe the
polaron as an electron surrounded by a cloud of virtual phonons, which represents its
surrounding polarization. One refers to these polarons as phonon-dressed electrons,
since attached to them is the “fabric” of the surrounding lattice. They can be
described by a Fröhlich Hamiltonian, which explicitly accounts for the electron-
phonon (LO phonons at k = 0) interaction (Fröhlich et al. 1950, advanced by Lee
et al. 1953). Alternatively, it can be described with the Feynman Hamiltonian
(Feynman path integral, Feynman 1955), which simulates the virtual phonons by
a fictitious particle that interacts with the electron via a harmonic potential (Peeters
and Devreese 1984).
The size of the polaron is measured by the extent of the lattice distortion caused
by the electron. Large polarons extend substantially beyond nearest-neighbor
distances; small polarons do not. For a review, see Velasco and García-
Moliner (1997).
The net difference unaccounted for in the Bloch electron picture describes the
Coulomb energy of the polaron:
Carrier-Transport Equations 7
Epol/e
e2 1 1 1
Epol, Coul ¼ with ¼ : (5)
8π e e 0 r pol e eopt estat
The resulting net Coulomb potential of the polaron is shown in Fig. 2. Within this
potential funnel, the polaron self-energy can be expressed as the 1s quasi-hydrogen
ground-state energy
e 2 m
stat pol
Epol ¼ EqH : (6)
e mn
It is distinguished from the quasi-hydrogen ground-state energy by replacing estat
with e , and the electron effective mass with that of the polaron.
The radius of the polaron can be defined as the corresponding 1s quasi-hydrogen
radius
e m n 4π e e0 ℏ2
r pol ¼ aqH ¼ : (7)
estat mpol mpol e2
When the polaron eigenfunctions are overlapping, the polaron self-energy (Eq. 6)
broadens from a sharp level into a polaron band within which polarons can move
through the crystal. The width of this band can be estimated from the uncertainty by
absorbing or emitting virtual LO phonons as ℏωLO. The corresponding uncertainty
radius of such a polaron is given by the uncertainty distance of finding a particle that
interacted with an LO phonon: it has an energy ℏ2k2/(2mpol) with uncertainty ℏωLO.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Therefore, it has an uncertainty in wavenumber of 2mpol ωLO =ℏ, the reciprocal of
which is its corresponding uncertainty in position
ℏ
r pol ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (8)
2mpol ℏωLO
which is also used as the radius of a large polaron. For typical compound semi-
conductors, the polaron radius is on the order of several lattice constants.
as the ratio of the Coulomb energy of a polaron, which describes the electron-phonon
interaction, to the energy of the LO phonon, i.e., the predominantly interacting
phonon:
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
u
u1 e2
u
t2 4π e e0 r pol
αc ¼ : (9)
ℏωLO
Entering the expression for rpol Eq. 7 into Eq. 9 and replacing e* from Eq. 5, we
obtain for the Fröhlich coupling constant αc:
rffiffiffiffiffiffiffiffiffiffiffi
e2 2 mpol 1 1
αc ¼ ; (10)
8π e0 ℏ ℏωLO eopt estat
which is below unity for good semiconductors. Here αc can be interpreted as twice
the number of virtual phonons surrounding – that is, interacting with – a slowly
moving carrier in the respective band.
Semiconductors without ionic bonding (Si, Ge, α-Sn) have eopt = estat and
consequently αc = 0. Fröhlich coupling constants are therefore much larger for
ionic than for covalent semiconductors and increase with increasing effective charge
and decreasing strength of the lattice binding forces.4 Some values of αc for typical
semiconductors are given in Table 1.
Polaron Energy and Effective Mass Using αc, we can express the polaron energy
as a fraction of the LO phonon energy5:
p2
Assuming a small perturbation of the parabolic band, Lee et al. (1953) obtain for the
energy dispersion within the band
ℏ2 k2 ℏ2 k2 ℏ2 k2
Eð k Þ ¼ αc ℏωLO þ þ . . . ffi αc ℏωLO þ ; (12)
2 mn 12 mn 2 mpol
4
A decrease of lattice-bonding forces corresponds also to a decreased Debye temperature; see also
Sect. 1.1.2 of chapter ▶ “Phonon-Induced Thermal Properties”.
5
This result is obtained from Fröhlich et al. (1950) and with a variational method from Lee et al.
(1953). It can be used
up
to αc 1. Earlier results from Pekar, using an adiabatic approximation,
yielded Epol ¼ α2c =3π ℏωLO , which gives a lower self-energy than Eq. 11 in the range of
validity αc < 1.
Carrier-Transport Equations 9
mn
mpol, large ffi : (13)
1 α6c
Equation 13 may be used for small αc, i.e., a weak electron-phonon interaction. For
larger values αc 1 , the following approximation is used for the polaron mass
(Feynman 1955):
16 α4c mn
mpol, large ffi : (14)
81π 4
Materials with large polarons that show a significant increase in the effective mass
are silver halides. These have an intermediate coupling constant (1.60 and 1.91 for
AgBr and AgCl, respectively).
The variational method of Lee et al. (1953) is used to compute intermediate
coupling. Many II-VI and some III-V compounds have αc values that make these
large polarons sufficiently distinct from electrons (see Table 1 and Evrard 1984).
Their effective mass can be determined by cyclotron resonance (Peeters and
Devreese 1984). The Landau levels appearing in a magnetic field are shifted by ΔE
ffi αc ℏωLO for αc 1; when the cyclotron resonance frequency approaches ωLO,
this level splits into two peaks indicating the strength of the electron-lattice interac-
tion. For a more exact approximation using a path integral formulation, see Feynman
(1955). For reviews, see Kartheuser et al. (1979), Bogoliubov and Bogoliubov Jr.
(1986), and Devreese (1984).
In Table 2, the properties of large polarons for a number of crystals are listed. We
note the inverse relation of rpol and Epol; for III-V semiconductors, the polaron mass
is close to the electron rest-mass, while it increases for II-VI semiconductors and gets
large for halides. Correspondingly mn/mpol reflects basically the effective electron
mass in typical semiconductors.
In the presence of a magnetic field, the cyclotron behavior of polarons within the
corresponding Landau levels is observed instead of the bare Bloch band mass of carriers.
For a transition from Landau level n to n + 1, Bajaj (1968) obtained for polarons
αc 3 ω2 ℏk2
ωc, pol ¼ ωc 1 αc c þnþ1 ; (15)
6 20 ωLO 2 mn ωc
10 K. W. Böer and U. W. Pohl
Epol,opt,small, because of the strong lattice coupling; see Sects. 1.2.1 and 1.2.2 of chapter
▶ “Optical Properties of Defects”. For further discussion of Epol, see the review of
Devreese (1984).
The effective mass of small polarons is given by (see Appel 1968):
mn α2c
mpol,small ffi : (17)
48
Examples for materials with small polarons are narrow-band semiconductors with
large αc values (~10), e.g., transition-metal oxides such as NiO or the molecular
crystals described in Sect. 3 of chapter ▶ “The Origin of Band Structure” with their
Carrier-Transport Equations 11
particularly small band dispersion (see Sect. 4.3 of chapter ▶ “Quantum Mechanics of
Electrons in Crystals”). Due to this narrow bandwidth, the band conductivity is usually
disturbed by phonon scattering in organic crystals at room temperature, and hopping
mechanisms may prevail even if the material has crystalline structure; see Sects. 4.3
and 5.1.2 of chapter ▶ “Carrier Transport Induced and Controlled by Defects”. A
review for small polarons is given by Emin (1973); for a summary of polaron mobility
due to various scattering mechanisms, see Appel (1968) and Evrard (1984).
When the coupling to the lattice is strong enough so that the polarizing electron
produces a significant Coulomb funnel, a second electron with opposite spin may
be trapped within the same funnel (Chakraverty and Schlenker 1976). This
bipolaron again can move through the lattice at an energy below that of a free
electron and with an effective mass somewhat larger than that of two free
electrons (Böttger and Bryksin 1985). For bipolarons in organic polymers, see
Brazovskii et al. (1998). A significant difference, compared with two free elec-
trons or two independent polarons, is the fact that the new quasiparticle bipolaron
has zero spin and consequently acts as a boson. It is also referred to as a Cooper
pair, and its formation is used to explain superconductivity; see Sect. 1.2 of
chapter ▶ “Superconductivity”.
Existence Criteria for Different Polarons There are three different energies, the
relative magnitudes of which determine the preferred existence of large polarons,
small polarons, or bipolarons. These are6
6
More precisely, these three energies are: Erelax, the energy given to phonons during lattice
relaxation; J, the bandwidth of a band created by free, uncoupled carriers of the given density;
and U, the energy necessary to put two carriers with opposite spin on the same lattice site.
12 K. W. Böer and U. W. Pohl
Carrier transport proceeds under external forces,7 resulting in drift, and under
internal quasiforces, resulting in diffusion. This may involve different charged
particles which contribute additively to the current or have an indirect effect when
it involves neutral particles, e.g., excitons (see Sect. 1.2 of chapter ▶ “Excitons”).
Exciton diffusion plays a major role in devices fabricated from organic semicon-
ductors; for a review see Mikhnenko et al. (2015).
Carrier transport occurs in bands near the band edges, i.e., near Ec for electrons
and near Ev for holes. For materials with a sufficiently large defect density, carrier
transport may proceed also via tunneling between trapping states. It may also involve
carriers hopping from traps into the band or hopping of self-trapped small polarons.
Trapped carriers travel a short distance in the band and later are recaptured, then
reemitted, and so on; an analogous process occurs with small polarons – see chapter
▶ “Carrier Transport Induced and Controlled by Defects”.
All of these processes add up to produce the total current and usually have vastly
different magnitudes. Ordinarily, only one transport process predominates in homoge-
neous semiconductors. In nonhomogeneous materials, however, at least two and fre-
quently four contributions are important in different regions of the devices. These are
drift and diffusion currents of electrons and holes. First, a rather simple picture of the
carrier transport is presented, which serves as guidance for a more sophisticated
approach in later chapters.
7
Strictly speaking, steady-state carrier transport is due to external forces only. The diffusion current
originates from a deformed density profile due to external forces and is a portion of the conven-
tionally considered diffusion component. The major part of the diffusion is used to compensate the
built-in field and has no part in the actual carrier transport: both drift and diffusion cancel each other
and are caused by an artificial model consideration – see Sect. 3.4.
Carrier-Transport Equations 13
vn /vmp
0.5
0
0 50 100 150 200 250
vn (km/s)
At finite temperatures, carriers are found above the edge of the respective band
according to their statistical distribution function. In semiconductors, they usually
follow the Boltzmann distribution function within the bands when they are not
degenerate, i.e., when the carrier densities are below 0.1Nc or 0.1Nv (Eqs. 18 and
22 of chapter ▶ “Equilibrium Statistics of Carriers”). Their thermal velocity, the root
mean square velocity,8 is obtained from the equipartition principle, i.e., kinetic
energy = ½kT per degree of freedom:
hence mn 2 3
v ¼ k T; (18)
2 2
pffiffiffiffiffiffiffiffi 3k T
hv2 i ¼ vrms ¼ ; (19)
mn
or
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
m0 T ðKÞ
vrms ¼ 1:18
107 ðcm=sÞ: (20)
mn 300
8
The rms (root mean square) velocity vrms, which is commonly used, should be distinguished from
the slightly different average velocity vav ¼ hjvji and from the most probable velocity vmp. Their
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
ratios are vrms : vav : vmp ¼ 3=2 : 4=π : 1 ¼ 1:2247 : 1:1284 : 1, as long as the carriers follow
Boltzmann statistics. For a distinction between these different velocities, see Fig. 3.
14 K. W. Böer and U. W. Pohl
which visualizes the filling of space with atomic spheres and expects only very limited
possibilities for an electron traversing between these spheres without being scattered.
The introduction of phonons and crystal defects provides centers for scattering. In
this way, a carrier motion results, which can be described as a Brownian motion9 with a
mean free path commensurate with the average distance between scattering centers.
This distance is several hundred angstroms in typical crystalline semiconductors, i.e.,
the mean free path extends to distances much longer than the interatomic spacing.
When a carrier responds to an external field, it is accelerated in the direction of the
electric field. Many important features of the carrier motion can be explained by
assuming only inelastic scattering. Since it takes place at defects, which themselves
are in thermal equilibrium with the lattice, the carrier tends to lose the excess energy
gained between the scattering events. Figure 4 illustrates the typical motion of a
carrier with and without an external electric field, assuming for both cases identical
scattering events. The changes due to the field are exaggerated; under normal
external fields, the changes from the random walk without field are very small
perturbations barely being visible in the scale of Fig. 4.
dV dψ
F¼ ¼ : (21)
dx dx
With Ec ¼ jej ψ þ const; it yields a positive field when the band slopes downward
from the cathode toward the anode, giving the visual impression that electrons “roll
9
This motion resembles a random walk (Chandrasekhar 1943).
In chapters dealing with carrier transport, F is chosen for the field, since E is used for the energy.
10
Carrier-Transport Equations 15
downhill” and holes “bubble up.” We will use the proper signs in the following
sections, however, reverting back to the commonly used e ¼ jej later in order to
avoid confusion in comparison to familiar descriptions.
When expressing forces, we need to distinguish the sign of the carrier; therefore,
for an accelerating force we have
dv
mn ¼ e F: (23)
dt
During a free path, the electron gains an incremental velocity, for an arbitrarily
chosen field in x direction F = (Fx, 0, 0):
e
Δvx ¼ Fx τsc : (24)
mn
After averaging the incremental velocity between collisions and replacing the scat-
tering time τsc with the average time τ between scattering events, we obtain the drift
velocity vd:
e
vd ¼ τ Fx : (25)
mn
With an electron density n and a charge e, we obtain for the current density for
electrons
j n ¼ e n vd ; (26)
e2
jn ¼ σ n F, with σ n ¼ τn n : (27)
mn
In a homogeneous semiconductor, the external field is given by the bias V, divided
by the electrode distance d, yielding Ohm’s law
16 K. W. Böer and U. W. Pohl
V V
jn ¼ σ n F ¼ σ n ¼ ; (28)
d AR
with A as the area of the semiconductor normal to the current and the resistance
ρn d d
R¼ ¼ ; (29)
A A σn
where ρn is the specific resistivity ρn ¼ 1=σ n :
Joule’s Heating The additional energy gained by the accelerated electrons from the
external electric field is delivered to the lattice during inelastic collisions, generating
phonons. This Joule’s heating is given by the power density p which n electrons
transfer to the lattice while travelling at the average constant drift velocity vd despite
the accelerating force – eF; applying the sign convention used above we obtain
j2n
p ¼ n e vd F ¼ jn F ¼ σ n F2 ¼ : (30)
σn
The quantity
e
τ n ¼ μn (31)
mn
is the electron mobility, since carriers are more mobile when they experience
less scattering, i.e., the time between collisions is larger, and when their effective
mass is smaller, i.e., they can be accelerated more easily. With e for electrons and
+e for holes, the mobility is negative for electrons and positive for holes, while
Carrier-Transport Equations 17
the conductivity (/ e2, see Eq. 27) is always positive. Conventionally, however,
μn ¼ jμn j is used, and we will follow this convention here.
The electron conductivity and hole conductivity are given by
σ n ¼ e μn n and σ p ¼ e μp p: (32)
Gas-Kinetic Model for Electron Scattering Different types of lattice defects are
effective to a differing degree in carrier scattering. In a simple gas-kinetic model,
scattering centers have a well-defined scattering cross-section sn; a scattering
event, i.e., a marked deflection11 from an otherwise straight carrier path with
exchange of momentum and/or energy, takes place when the carrier approaches
the scattering center within its cross section. A mean free path λn can then be
derived by constructing a cylinder of cross section sn around an arbitrary straight
carrier path and computing the average distance from the last scattering center to
which this cylinder will extend until it incorporates the centerpoint of the next
scattering center. At this length the cylinder volume λn sn equals the average
volume 1/Nsc (cm3) that one of these centers occupies, where Nsc is the density
of scattering centers; hence
1
λn ¼ : (33)
sn N sc
Consequently, the time between scattering events is given by
1
τsc ¼ : (34)
vrms sn N sc
This time is used to obtain an estimate for the carrier mobility in the Drude
approximation:
e e λ
μn ¼ τsc ¼ ; (35)
mn mn vrms
or
m0
μn ¼ 1:8
1015 τsc ðsÞ cm2 =Vs ; (36)
mn
or, using the expression of Eq. 19 for vrms:
Often a minimum scattering angle of 90 is used to distinguish scattering events with loss of
11
a 5 b5
0.4 0.4 1
1
3 0.7 3 0.7
2 2
1 1
–14 –13 –12 –11 –10 1 2 3 4 5
log (τsc/s) log (λ/Å)
Fig. 5 Carrier mobility as a function of (a) the time between scattering events τsc and of (b) the
mean free path λ, with the effective electron mass mn as family parameter
3=2
m0 300 1=2 2
μn ¼ 1:5
λðÅÞ cm =Vs : (37)
mn T ðK Þ
The dependence of the carrier mobility on the mean scattering time and mean free
path is shown in Fig. 5. The application of this simple gas-kinetic model, however,
has to be taken with caution because of its simplified assumptions. Generally, it
yields too large densities of tolerable scattering centers.
The drift current is the product of the elementary charge, the carrier mobility, the
single carrier density (derived in chapter ▶ “Equilibrium Statistics of Carriers”), and
the electric field. For electrons or holes, it is
E d
Ec
eV
EF
Ev
V
F¼ : (39)
d
For the field concept to apply, the distance d between electrodes must also be
large compared to the interatomic spacing. The field can then be expressed by the
macroscopic sloping of the bands12:
1 d Ec 1 d Ev
F¼ ¼ : (40)
e dx e dx
It is also given, and more importantly so, by the slope of the Fermi potential (see Sect.
3.3) which, within the homogeneous material, is the same as the slope of the bands:
1 d EF
F¼ : (41)
e dx
The bias is expressed as the difference of the Fermi levels between both electrodes;
compare Fig. 6.
When using the electrostatic potential ψ n with
d ψn
¼ F; (42)
dx
the drift current can be expressed as a product of the electrical conductivity and the
negative gradient of this potential:
12
As a reminder: here and in all following sections, |e| is used when not explicitly stated differently.
20 K. W. Böer and U. W. Pohl
Ev
u
1 0
0 x d
d ψn d ψp
jn,drift ¼ σ n or jp,drift ¼ σ p : (43)
dx dx
For reasons to become apparent below, two electrostatic potentials
are introduced:
ψn
and ψ p for conduction and valence bands, respectively, with e ψ n ψ p ¼ Eg .
In a homogeneous semiconductor in steady state and with vanishing space
charge, these drift currents are the total currents, and the slopes of both potentials
are the same. There are special cases, however, in which the band edges of the
valence and conduction bands are no longer parallel to each other. One of these will
be mentioned briefly in the following section.
13
The atoms in the alloyed lattice (or here in the anion sublattice) are statistically arranged. Strictly
speaking, this causes random fluctuation of the composition in a microscopic volume element of the
crystal and results in a local fluctuation of the bandgap energy due to a locally varying parameter x;
as a result, extended or localized states with energies Ec,v(x) close to the band edge Ec, v ðxÞ of the
mean composition x are formed, leading to some tailing of the band edges (Sect. 3.1 of chapter
▶ “Optical Properties of Defects”). The composition parameter x used in the text actually refers to
the mean composition, and the effect of band tailing is neglected here.
14
Major deviations from linearity of Eg with composition are observed when the conduction-band
minimum lies at a different point in the Brillouin zone for the two end members. One example is the
alloy of Ge and Si. Other deviations (bowing – see Sect. 2.1 of chapter ▶ “Bands and Bandgaps in
Solids”) are observed when the alloying atoms are of substantially different size and electronegativity.
Carrier-Transport Equations 21
chapter ▶ “Bands and Bandgaps in Solids”. ZnSe and ZnS are both n-type materials.
Depending on doping, the Fermi level in ZnSe can be shifted easily between 0.8 and
0.2 e V below Ec and in ZnS between 1.0 and 0.4 eV. Depending on the doping profile
in the mixed-composition region, a wide variety of relative slopes, including non-
monotonic slopes, of valence and conduction bands can be designed for a
vanishing bias, i.e., for a horizontal Fermi energy EF. In Fig. 7, an example with
opposite and linear sloping of Ec(x) and Ev(x) is shown, resulting effectively in a
built-in field (see Sect. 3.4) of opposite sign for electrons and holes. In thermo-
dynamic equilibrium, however, there is no net current in spite of the sloping
bands. This is accomplished by exact compensation of finite drift currents with
opposing diffusion currents, which self-consistently determine the slopes of the
bands. An example for the application of such graded AluGa1-uAs composition is
given in Horio et al. (1999).
The change in the bandgap energy can be expressed as
and
The corresponding built-in fields for electrons and holes are given by
@ΔEg ðxÞ @ψ
Fn ¼ AE ¼ n (47)
e@x @x
and
@ΔEg ðxÞ @ψ p
Fp ¼ ð AE 1 Þ ¼ ; (48)
e@x @x
justifying the introduction of separate electrostatic potentials for electrons and holes:
with ΔEg ðxÞ 6¼ 0 and AE 6¼ 1/2, we have Fn 6¼ Fp.
Carrier diffusion by itself can be observed when the external field vanishes and a
concentration gradient exists. An example in which these conditions are approxi-
mately fulfilled is the diffusion of minority carriers created by an inhomogeneous
22 K. W. Böer and U. W. Pohl
x0-dx/2 x0 x0+dx/2 x
optical excitation (see, e.g., Najafi et al. 2015). The diffusion current is proportional
to the diffusion coefficient D and to the carrier-density gradient; for electrons or
holes, it is
dn dp
jn,diff ¼ e Dn or jp,diff ¼ e Dp : (49)
dx dx
The negative sign of the hole current is due to the fact that in both equations e
¼ jej is used. The diffusion current can be derived as the difference between two
currents caused by a completely random motion of carriers originating in adjacent
slabs with slightly different carrier densities (Fig. 8). The current originating at x0 +
dx/2 and crossing the interface at x0 from right to left, is caused by the Brownian
motion of electrons of a density n0 + dn/2. It is given by
dn v2rms τn
j n,diff ¼ e n0 þ : (50)
2 3 dx
The current crossing the boundary from left to right is accordingly given by
! dn v2rms τn
j n, diff ¼ e n0 : (51)
2 3 dx
The current is proportional to the carrier velocity vrms and the carrier mean free
path λn. In turn, λn is given by vrmsτn. The factor 13 arises from gas-kinetic arguments
when the root mean square velocity is obtained from an isotropic velocity distribu-
tion: v2 ¼ v2rms ¼ v2x þ v2y þ v2z ; with v2x ¼ v2y ¼ v2z , we obtain the relation v2x ¼ 13v2rms
for the x-component used in Eq. 50.
The difference of both currents (Eqs. 50 and 51) is the net diffusion current
Carrier-Transport Equations 23
! v2rms τn dn
jn, diff ¼ j n, diff j n, diff ¼ e ; (52)
3 dx
with the diffusion coefficient given by
v2rms τn
Dn ¼ : (53)
3
By using v2rms ¼ 3k T=mn (Eq. 19), we obtain the more commonly used equation
for the diffusion current
dn dp
jn,diff ¼ μn k T and jp,diff ¼ μp k T : (54)
dx dx
Both diffusion currents for electrons and holes have the same negative sign for a
positive gradient of n(x) or p(x) when recognizing that μn is negative and μp is
positive. However, since the conventional notation with μn ¼ jμn j is used, the
difference in signs appears.
Maximum Diffusion Currents As the gradient of the carrier density increases, the
diffusion current increases proportionally to it (Eq. 49). However, this proportion-
ality is limited, when the density gradient becomes so steep that the reverse current
(Eq. 51) becomes negligible compared to the forward current (Eq. 50).
When increasing the distance dx to the mean free path λn, we obtain from Eq. 52
!
with j n, diff j n, diff for the maximum possible diffusion current through a planar
surface
v2rms τn
jn,diff ,max ¼ e n0 ; (55)
3 λn
or, for carriers following Boltzmann statistics and within a device with planar
geometry,
en
jn,diff ,max ¼ pffiffiffiffiffi vrms : (56)
6π
x0 x0+Δx x
V
V0
ΔV=FΔx
x0 x0+Δx x
The Einstein Relation Comparing the diffusion equations, Eq. 49 with Eq. 54, we
obtain a relation between the diffusion constant and the carrier mobility:
μn, p k T
Dn, p ¼ ; (57)
e
which is known as the Einstein relation and holds for systems that follow Boltzmann
statistics. This can be seen from the following arguments. In thermal equilibrium, the
total current, as well as each carrier current, vanishes: j
jn
jp
0. The electron
current is composed of drift and diffusion currents – see Eq. 61; hence,
dn
μn nF þ Dn ¼ 0; (58)
dx
which can be integrated to yield
eF Δx
nðx0 þ ΔxÞ ¼ nðx0 Þexp : (60)
kT
Total Currents The total current is given as the sum of drift and diffusion currents.
For electrons, we have
dn
jn ¼ jn,drift þ jn,diff ¼ e μn n F þ Dn ; (61)
dx
j ¼ jn þ jp : (63)
For evaluating the total current, we need to consider the gradient of ψ(x) for the drift
component and the gradient of n(x) for the diffusion component. It is instructive to
deduce the electron and hole current for thermal equilibrium. In thermal equilibrium,
n is given by the Fermi distribution (Eq. 10 of chapter ▶ “Equilibrium Statistics of
Carriers”). When the Fermi energy is separated by several kT from the band edge, we
can disregard the 1 in the denominator of Eq. 10 in chapter ▶ “Equilibrium Statistics
of Carriers” and approximate this equation with the Boltzmann distribution, yielding
26 K. W. Böer and U. W. Pohl
Nc
Ec ðxÞ EF ðxÞ ¼ k T ln : (64)
nð x Þ
Replacing Ec(x) with eψ ðxÞ þ c and differentiating both sides of Eq. 64 with
respect to x, we obtain after division by e
dψ n 1 dEF k T 1 dn
¼ : (65)
dx e dx e n dx
After multiplying both sides with σ n ¼ e μn n and rearranging, we obtain
1 dEF dψ dn
σn ¼ σ n n þ μn k T : (66)
e dx dx dx
The right-hand side is the total electron current; thus, the left-hand side must also be
equal to jn:
1 dEF
jn ¼ σ n : (67)
e dx
Since in thermal equilibrium, i.e., with vanishing external field, the Fermi level must
be constant (horizontal in an E(x) presentation), we conclude that the electron and
hole current must vanish separately in equilibrium:
dEF
0 ) jn
jp
0: (68)
dx
Quasi-Fermi Levels In steady state, e.g., with a constant external excitation (chap-
ter ▶ “Carrier Generation”), the electron and hole densities deviate from thermody-
namic equilibrium values. Nevertheless, we may use the Fermi distribution to
describe their density in the bands, using the quasi-Fermi levels EFn and EFp
according to the definition equations
1
n
Nc ; (69)
Ec EFn
exp þ1
kT
and
1
p
Nv ; (70)
EFp Ev
exp þ1
kT
with EFn 6¼ EFp. This rather useful approximation introduces errors which may or
may not be acceptable depending on the cause for deviation from the thermal
equilibrium. In general, the error is quite small for optical excitation and for low
Carrier-Transport Equations 27
EFp
Ev
external fields. For high external fields, the distribution function is substantially
deformed (see chapter ▶ “Carrier Scattering at High Electric Fields”), and a more
sophisticated approximation is required.
With optical or field-induced carrier generation discussed in chapter ▶ “Carrier
Generation”, n and p are increased above their thermodynamic equilibrium value;
hence, EFp < EF < EFn, resulting in a decreased distance of both quasi-Fermi levels
from their corresponding bands; see Fig. 10. In certain cases, the recombination may
be increased above the equilibrium value, as, for instance, in a pn junction in reverse
bias; here, EFn can drop below EFp.
In using the same algebraic procedure as described in the previous section, we
have for the total electron current in steady state:
1 dEFn
jn ¼ σ n ; (71)
e dx
this means, the total electron current is proportional to the negative slope of the
quasi-Fermi potential, as the drift current is proportional to the negative slope of the
electrostatic potential (Eq. 43). For both currents, the conductivity is the proportion-
ality constant.
In order to emphasize this similarity, we define the electrochemical potentials for
electrons and holes:
1 1
φn ¼ EFn and φp ¼ EFp : (72)
e e
The total currents can now be expressed as
@φn @φp
jn ¼ σ n and jp ¼ σ p : (73)
@x @x
For homogeneous semiconductors with homogeneous generation of carriers, these
currents become the drift currents, and Eq. 73 becomes equal to Eq. 43.
In steady state, the total current is divergence-free, i.e., jn+ jp = const (for
included carrier generation see chapter ▶ “Photoconductivity”). Therefore,
28 K. W. Böer and U. W. Pohl
@φn @φp
σn þ σp
const (74)
@x @x
or
@φn @φp
μ n nð x Þ þ μ p pð x Þ
const: (75)
@x @x
Since a semiconductor is predominantly either n- or p-type, except for the inner part
of a junction, we usually can neglect one part of the sum. For example, for the n-type
region,
@φn
nð x Þ
const; (76)
@x
i.e., if there is a gradient in the carrier density, then the highest slope in φn(x) is
expected where the carrier density is lowest for an inhomogeneous n(x) distribution.
From Eq. 75, we also conclude that for vanishing currents in steady state, the
slopes of the quasi-Fermi potentials must be opposite to each other. The lower the
corresponding carrier densities, the higher the slopes:
@φn @φp
μ n nð x Þ ¼ μp pðxÞ : (77)
@x @x
@φn @φp
jn ¼ σ n , jp ¼ σ p ;
@x @x
dψ n dψ p
jn, drift ¼ σ n , jp, drift ¼ σ p ; (78)
dx dx
ð φn ψ n Þ φp ψ p
jn, diff ¼ σ n , jp, diff ¼ σ p :
@x @x
An external and internal field of equal magnitude results in the same slope of the
bands. Therefore, this distinction between internal and external fields is usually not
made, and the subscripts at the fields are omitted. There are, however, drawbacks in
such a general description of fields, which can best be seen from carrier heating in an
electric field. Carrier heating is used to describe the field dependence of the mobility
(see Sects. 2 and 3 of chapter ▶ “Carrier Scattering at High Electric Fields”) in a
microscopic model: accelerated carriers are shifted up to higher energies within a
band; consequently, their effective mass and the scattering probability change. Usually
the effective mass increases and the creation of phonons becomes easier. The mobility
hence becomes field dependent and usually decreases with increasing field.
Such carrier heating is absent in thermal equilibrium: the carrier gas and the
lattice with its phonon spectrum are in equilibrium within each volume element;
thus, carrier and lattice temperatures remain the same (Stratton 1969). No energy can
be extracted from an internal field, i.e., from a sloped band, due to a space charge in
equilibrium.16 This situation may be illustrated with an example replacing electrical
with gravitational forces: a sloping band due to a space-charge region looks much
like a mountain introduced on top of a sea-level plane, the Fermi level being
equivalent to the sea level. As the introduction of the mountain does little to the
distribution of molecules in air, the introduction of a sloping band does little to the
distribution of electrons in the conduction band. Since there are fewer molecules
above the mountain, the air pressure is reduced, just as there are fewer electrons in a
band where it has a larger distance from the Fermi level as illustrated in Fig. 11.
However, when one wants to conveniently integrate overall altitudes (energies) in
order to arrive at a single number – the air pressure (or electron density) – one must
consider additional model consequences to prevent winds from blowing from the
valleys with high pressure to the mountain top with low pressure by following only
the pressure gradient. Neither should one expect a current of electrons from the
15
pn junctions are the best studied intentional space-charge regions. Inhomogeneous doping
distributions – especially near surfaces, contacts, or other crystal inhomogeneities – are often
unintentional and hard to eliminate.
16
This argument no longer holds with a bias, which will modify the space charge; partial heating
occurs, proportional to the fraction of external field. This heating can be related to the tilting of the
quasi-Fermi levels.
30 K. W. Böer and U. W. Pohl
EF
1/2
1 x
f (E)
regions of a semiconductor with the conduction band close to the Fermi level, which
results in a high electron density, to a region with low electron density in the absence
of an external field. To prevent such currents in the electron-density model, one uses
the internal fields, i.e., the built-in fields, and balances the diffusion current with an
exactly compensating drift current. The advantage of this approach is the use of a
simple carrier density and a simple transport equation. The penalty is the need for
some careful definitions of transport parameters, e.g., the mobility, when comparing
external with built-in fields and evaluating the ensuing drift and diffusion currents
when the external fields are strong enough to cause carrier heating.
17
With bias, the Fermi level in a junction is split into two quasi-Fermi levels which are tilted,
however, with space-dependent slope. Regions of high slope within the junction region will become
preferentially heated. The formation of such regions depends on the change of the carrier distribu-
tion with bias and its contribution to the electrochemical potential (the quasi-Fermi level). Integra-
tion of transport, Poisson, and continuity equations yields a quantitative description of this behavior
(Böer 1985b).
Carrier-Transport Equations 31
a b
E E
E1
ħω
Ec
EF EF
x x1 x2 x
Fig. 12 Sloping band due to (a) an internal (built-in) field with horizontal Fermi level EF and (b)
due to an external field with parallel sloping of both, bands and Fermi level. The electron
distribution is indicated by a dot distribution, and the action of field and scattering by arrows
(After Böer (1985a))
dependence of the lower integration boundary, while the energy distribution of the
carrier n(E) remains independent in space:
ð1
nð x Þ ¼ nðEÞ dE: (81)
Ec ðxÞ
This is similar to the velocity distribution of air molecules, which is the same at any
given altitude, whether over a mountain or an adjacent plane; whereas the integrated
number, i.e., the air pressure near the surface of the sloping terrain, is not. This does
not cause any macroscopic air motion, since at any stratum of constant altitude, the
molecular distribution is the same; hence, the molecular motion remains totally
random.
In a similar fashion, electrons at the same distance above the Fermi level are
surrounded by strata of constant electron density; within such strata their motion
must remain random. During scattering in thermal equilibrium, the same amounts of
phonons are generated as are absorbed by electrons, except for statistical fluctua-
tions: on the average, all events are randomized. Electron and hole currents both
vanish in equilibrium for every volume element. Figure 12a illustrates such a
behavior.
In an external field, however, bands and Fermi level are tilted parallel to each
other; this means, with applied bias, the carrier distribution becomes a function of the
spatial coordinate as illustrated in Fig. 12b. When electrons are accelerated in the
field, they move from a region of higher density nðE1 EF Þx1 to a region of lower
density nðE1 EF Þx2. These electrons can dissipate their net additional energy to the
lattice by emitting phonons and causing lattice (Joule’s) heating. In addition, while in
32 K. W. Böer and U. W. Pohl
net motion, electrons fill higher states of the energy distribution, thereby causing the
carrier temperature to increase. The carrier motion in an external field is therefore no
longer random; it has a finite component in field direction: the drift velocity vd =
μFext and the collisions with lattice defects are at least partially inelastic. A net
current and lattice heating result.
where g(v) is the density of states in the conduction band per unit volume of velocity
space and d3v is the appropriate volume element in velocity space. If Fx is the built-in
field Fint, then the distribution function is the Boltzmann function fB(v). If Fx is the
external field Fext, the distribution function is modified due to carrier heating
according to the field strength f Fext ðvÞ – see Sects. 4.4 and 4.6. The averaging process
involves the distribution function, which is modified by both scattering and
effective-mass contributions. For a review, see Nag (1980); see also Seeger (1973)
and Conwell (1967). In contrast, when only a built-in field is present, the averaging
must be done with the undeformed Boltzmann distribution, since lattice and electron
temperatures remain the same at each point of the semiconductor.
A more detailed discussion of differences in carrier transport for external, built-in,
and mixed fields is postponed to chapter ▶ “Carrier Scattering at High Electric
Fields” after an explicit introduction of carrier heating.
Second, the interaction with a scattering center often depends on the energy of the
electron, e.g., the scattering cross-section of ions is energy dependent, and only more
energetic electrons can be scattered inelastically. Therefore, the assumption of a
constant, energy-independent time between scattering events needs to be refined. In
this chapter, we will introduce the basic Boltzmann equation which permits an
analysis of a more advanced description of carrier scattering.
df @f @f @f @f
¼ þ k_ þ r_ ¼ ; (83)
dt @t @k @r @t coll
where the index coll indicates the collisions. The first term accounts for the local
change of the distribution in time, the second term for the change in momentum
space, and the third term for the change of the distribution in real space. The sum of
these changes must be equal to the changes of the distribution caused by collisions.
This simplified Liouville equation is called the Boltzmann equation.
In steady state (@f/@t
0), the Boltzmann equation reads
18
The most severe approximation is the linear relation in time, which eliminates memory effects in
the Boltzmann equation (Nag 1980).
34 K. W. Böer and U. W. Pohl
@f ðr, k, tÞ @f @f
¼ k_ þ r_ (84)
@t coll @k @r
with the first term determined by the forces acting on free electrons, and k_ given by
e
k_ ¼ F: (85)
ℏ
Here F is the electric field. The second term is proportional to the spatial gradient of
the carrier distribution and to the group velocity
1 @EðkÞ
r_ ¼ ¼ v: (86)
ℏ @k
This basic Boltzmann equation contains all the dependences necessary for analyzing
carrier transport.19 Some of these dependences, such as the temperature dependence,
are contained implicitly. The important part for the carrier transport is the innocent-
looking left-hand side of the Boltzmann Eq. 84, which contains the contribution of
the more or less inelastic collisions that provide the “friction” for the carrier
transport.
The collision term, also referred to as the collision integral, describes the transi-
tion of an electron from a state Ek, k to a state Ek0 , k0 . This can be expressed as the
difference between electrons scattered from the state k, occupied according to the
Fermi-Dirac distribution function fFD(k), into the state k0 , unoccupied according to
1 f FD ðk0 Þ, minus the reverse process, and integrated over all possible states k0 , into
and from which such scattering is possible:
ð
@f ðkÞ V
¼ ff FD ðkÞ½1 f FD ðk0 ÞSðk, k0 Þ f FD ðk0 Þ½1 f FD ðkÞSðk0 ,kÞg dk0 ;
@t coll ð2π Þ3
(87)
where V is the crystal volume and S is the scattering probability
2π
Sðk, k0 Þ ¼ jMðk, k0 Þj δðEk Ek0 ΔEÞ:
2
(88)
ℏ
19
Here discussed for electrons, although with a change of the appropriate parameters, it is directly
applicable to holes, polarons, etc. The influence of other fields, such as thermal or magnetic fields, is
neglected here; for such influences, see chapter ▶ “Carriers in Magnetic Fields and Temperature
Gradients”.
Carrier-Transport Equations 35
M(k, k0 ) is the matrix element for the scattering event and ΔE is the fractional
change in electron energy during the partially inelastic scattering. The matrix
elements can be expressed as
ð
Mðk, k0 Þ ¼ ψ q0 , k0 ΔVψ q, k dv (89)
v
where ΔV is the perturbation potential inducing the scattering event, dv is the volume
element, and ψ q, k , ψ q0 , k0 are the wavefunctions before and after scattering. The
perturbation potential depends on the type of scattering event and could be the
deformation potential for scattering on acoustic or optical phonons; see chapter
▶ “Carrier Scattering at Low Electric Fields”:
Ξ c ð@=@rÞ u acoustic phonons
ΔV ¼ (90)
D0 u optical phonons
A Boltzmann-type equation similar to Eqs. 83 and 84 can be set up for the phonon
system, which interacts with the electron system. Since the only driving forces for
the phonon system are those of diffusion due to thermal gradients (neglecting drag
effects discussed in Sect. 4.7), we obtain for steady-state conditions
@f ðqÞ @f ðqÞ @ωðqÞ
¼ r_ with r_ ¼ ; (91)
@t coll @r @r
here r_ is the group velocity of phonons, i.e., the sound velocity in the low
q acoustic branch. The gradient of the phonon distribution function f (r, q, T(r),
t) contains the thermal gradient. If undisturbed, the phonon distribution
is described by the Bose-Einstein function fBE(q) – see Eq. 12 of chapter ▶ “Equi-
librium Statistics of Carriers”. The collision term contains all phonon-phonon and
phonon-electron interactions. We will regard the former as less important to the
present discussion. Interaction of optical with acoustic phonons, however, can
become quite important, e.g., for cooling of a heated electron ensemble – see chapter
▶ “Dynamic Processes”. In a fashion similar to that given for the electron collision
term (Eq. 87), we obtain the transition rate by taking the product of the densities of
the occupied and the empty states and the matrix element for each transition,
integrated over all possible transitions for absorption and a similar term for emission
of phonons,
36 K. W. Böer and U. W. Pohl
ðð
@f ðk, qÞ V
¼ fSðk þ q, kÞ½1 þ f BE ðkÞ f BE ðk þ qÞ g
@t coll ð2π Þ3
Sðk, k þ qÞ ½1 þ f BE ðk þ qÞ f BE ðkÞ
dk dq : (92)
Assuming that this collision term is linear in the deviation from the unperturbed
distribution f0, we obtain
@f f f0
¼ : (94)
@t coll τm
Equation 93 can then be integrated. It yields an exponential return from the steady-
state, perturbed function f = f0 + δ f (with a time-independent δ f ) to the undisturbed
distribution in equilibrium f0 with the momentum relaxation time τm as the charac-
teristic time constant (see Sect. 4.6.1):
Carrier-Transport Equations 37
t
f ðtÞ f 0 ¼ δf exp : (95)
τm
In this linearized form, the deformed distribution function will be used first. In the
following sections, an example with zero magnetic field and vanishing gradients in
n and T is discussed. Here the second term of Eq. 84 vanishes.
In a homogeneous semiconductor with a force produced by a constant electric field
F, we obtain from Eqs. 94 and 84 for a small perturbation of the distribution function
eF @f δf
¼ (96)
ℏ @k τm
or, with ℏk ¼ mn v, hence @=@k ¼ ðℏ=mn Þ @=@v, we obtain
e @f
τm F ¼ δ f: (97)
mn @v
This shows that the change in the distribution function is proportional to the drift
velocity vd = (e/mn) τmF. With f0 given by the Boltzmann distribution
m n v2
f 0 / exp 2
; (98)
kT
@f @f mn v
ffi 0¼ f ; (99)
@v @v kT 0
and with Eq. 97, we have the following equation as the final result for the deformed
Boltzmann distribution due to an external field, in the relaxation time approximation:
e
f ¼ f0 1 τm F v : (100)
kT
For a homogeneous semiconductor, @f =@r 0 holds, and we obtain from Eqs. 83 and 87
@f ðkÞ e F @f ðkÞ
¼
@t ℏ @k
ð
V
3 ff ðkÞ½1 f ðk0 Þ Sðk, k0 Þ f ðk0 Þ½1 f ðkÞ Sðk0 , kÞg dk0 ; (101)
8π
with the electric field F producing a deformation from the equilibrium distribution
and the collision integral, i.e., the second term in Eq. 99, counteracting this defor-
mation. For an analysis, see Haug (1972).
In equilibrium, the solution of the Boltzmann equation is the Boltzmann or Fermi
function f0(k). With an applied electric field F, the distribution is shifted by the drift
velocity (see Fig. 13) and is slightly deformed. If the effective mass mn is isotropic,
the scattering probability S (Eq. 88) generally depends only on the magnitude
jk k0 j and hence on the angle included by the vectors, but not on the individual
orientations of k and k0 . The distribution function is then conveniently expressed by
a series development using Legendre polynomials (Pn):
X
1
f ðkÞ ¼ f 0 ðkÞ þ f n ðkÞ Pn ð cos θÞ; (102)
n¼1
where θ is the angle between F and k. After introducing f (k) into Eq. 101, we obtain
for the steady-state (@f =@t
0) a set of n equations to determine f (k).
ky
kx
δkx
Fig. 13 At an applied electric field, the Fermi sphere of occupied states in k space (green-filled
circles) is shifted by an amount δkx, yielding the sphere indicated by the red circle; magenta filling
indicates states occupied due to the applied field. Relaxation to equilibrium occurs by inelastic
scattering into unoccupied states (red arrow) of the equilibrium sphere (gray circle)
Carrier-Transport Equations 39
For small fields, only the first two terms of the development of Eq. 102 are taken
and yield
The perturbation term f1(k) of the distribution function is often expressed in terms of
a function ϕ(E)
eℏ @f 0 ðEÞ
f 1 ðkÞ ¼ Fk ϕðEÞ; (104)
mn @E
with E = E(k). This permits a simplified expression for the collision integral (Nag
1980)
ð
@f ðkÞ eℏ V
@t ¼ Fk f ð EÞ ½ 1 f 0 ð E0 Þ
coll mn k T 8π 3 0
(105)
k0 cos θk
ϕ ð EÞ ϕðE0 Þ Sðk, k0 Þ dk0 ;
k
where θk is the angle between k and k0 and E0 is the energy corresponding to the
wavevector k0 .
Any further simplification of the collision integral requires assumptions of the
specific scattering event, which will be listed in Sect. 1 of chapter ▶ “Carrier
Scattering at Low Electric Fields” and dealt with sequentially in the following
sections. However, some general remarks here will assist in categorizing the differ-
ent scattering types.
Elastic Scattering Elastic scattering keeps the electron energy during the scattering
event unchanged: E0 ¼ E. This simplifies Eq. 105 to
@f ðkÞ eℏ
¼ Fk
@t coll mn k T
ð
V
f 0 ð EÞ ½ 1 f 0 ð EÞ 3 ð1 cos θk Þ Sðk, k0 Þ dk0 : (106)
8π
Elastic scattering events are:
The total scattering term is given as the sum over the different scattering types:
X @f
@f
¼ : (107)
@t coll i
@t coll, i
The Carrier Current The current of carriers can be obtained from the
deformed Boltzmann distribution Eq. 100 by summation over all nel carriers
and velocities v,
X
nel X Xnel X
v F v f 0 τm
j¼e e v δf ¼ e2 : (108)
i¼1 v i¼1 v
kT
Assuming a spherical equi-energy surface for E(k), the summation over v F v can be
carried out,20 using for v v the averages vx vy ¼ vy vz ¼ hvz vx i ¼ 0 and v2x
D E
¼ v2y ¼ v2z ¼ v2 =3; this yields
e2 X 2
j¼ v τm f 0 F: (109)
3k T v
Considering that
X
nel X
n¼ f 0; (110)
i¼1 v
we obtain
X
v2 τ m f 0
eF v e 2
j ¼ en X ¼ en v τm F; (111)
3k T f0 3k T
v
20
The quantities τm and f0 in Eq. 108 are functions of E only and hence do not change over a
constant-energy surface.
Carrier-Transport Equations 41
e 2
μn ¼ v τm : (112)
3k T
With hEi ¼ ð3=2Þ k T ¼ mn v2 =2; we can replace 3 kT by mnhv2 i, yielding
e v2 τ m e hE τ m i
μn ¼ ¼ : (113)
mn hv2 i mn hEi
This result replaces the average time between scattering events obtained from the
Drude theory with the energy-weighted average of the relaxation time τm. Dropping
the requirement of spherical equi-energy surfaces, the end result (Eq. 113) remains
the same, except that mn is replaced by the anisotropic mobility effective mass; for
more detail, see Conwell (1982).
Following external forces, the carriers are accelerated proportionally to their effec-
tive masses (Eq. 23). Their anisotropy is taken into consideration by introducing a
mobility effective mass. For a three-axes ellipsoid, this effective mass is given by the
inverse average of the effective masses along the main axes:
1 1 1 1 1
¼ þ þ : (114)
mn, μ 3 m1 m2 m3
In general, a mobility tensor is introduced for each of the νd satellite valley E(k)
ellipsoids, identified by the index i:
0 1
@ 2 EðiÞ @ 2 EðiÞ @ 2 EðiÞ
B C
B @k2x @kx @ky @kx @kz C
B 2 ðiÞ C
e τm B @ E @ 2 EðiÞ @ 2 EðiÞ C
μðiÞ ¼ 2 B B
C: (115)
ℏ B @ky @kx @k2y @ky @kz CC
B 2 ðiÞ C
@@ E @ 2 EðiÞ @ 2 EðiÞ A
@kz @kx @kz @ky @k2z
0 1
1 2 1
þ 0 0
B 3 mn, t mn, l C
B C
e τm B 1 2 1 C
μtot ¼ 2 B 0 þ 0 C; (116)
ℏ B 3 mn, t mn, l C
B C
@ 1 2 1 A
0 0 þ
3 m n, t m n, l
1 1 2 1
¼ þ : (117)
mn, μ 3 mn, t mn, l
The hole mobility effective mass can be derived in a similar fashion. Assuming
spherical E(k) surfaces around k = 0 (for warped bands, see below) and disregarding
the deeper spin-orbit band, one obtains
1 1 1 1
¼ þ ; (118)
mp, μ 2 mp, lh mp, hh
here, mp,lh and mp,hh represent the light and heavy hole masses in the corresponding
bands.
m0
m p, ¼ 0 1 þ 0:333 Γ þ 0:0106 Γ þ . . .
2
(119)
AB
for the mobility effective mass. The variable m represents the light or heavy
hole mass if the upper or lower sign, respectively, is used in Eq. 119. The values
of the constants A, B, and C can be obtained from the Luttinger parameters
(Eq. 16 of chapter ▶ “Bands and Bandgaps in Solids”), which are given in
Table 5 of chapter ▶ “Bands and Bandgaps in Solids” for a number of typical
semiconductors.
Carrier-Transport Equations 43
With each scattering event, momentum is exchanged; the carrier changes the direc-
tion of its path. In addition, more or less energy is exchanged, with the carrier losing
or gaining energy from the scattering center. There are different rates for momentum
and energy relaxation, and the relaxation times depend on the specific scattering
mechanism.
It can be obtained from the net increment of the electron momentum, which is
proportional to the average drift velocity,
ð
v x f ð v Þ gð v Þ d 3 v
vx ¼ ð ; (121)
f ðvÞ gðvÞ d 3 v
where g(v) is the density of states. Assuming only small changes from the thermal
distribution, g(v) can be expressed as the effective density of states at the edge of the
band (Eq. 18 of chapter ▶ “Equilibrium Statistics of Carriers”) and cancels out in
Eq. 121. This yields
ð1
v4 τm f 0 ðvÞ dv
e Fx 0 e Fx 2
vx ¼ ð1 ¼ v τm ðvÞ : (122)
3kT 3kT
v2 f 0 ðvÞ dv
0
Using the equipartition law for a Boltzmann gas of electrons m2 v2 ¼ 32k T, we obtain
for the average drift velocity
e Fx v 2 τ m ð v Þ e
vx ¼ ¼ h τ m i Fx : (123)
3kT hv2 i mn
This result is closely related to the Drude equation, however, having replaced τ in
Eq. 25 with the average momentum relaxation time
44 K. W. Böer and U. W. Pohl
2
v τ m ðvÞ
hτm i ¼ : (124)
hv2 i
For an evaluation of Eq. 124, we need the distribution function f0 and the actual
scattering mechanism to determine τm(v) – see chapter ▶ “Carrier Scattering at Low
Electric Fields” and Seeger (1973).
After a collision, the electron path changes by an angle θ, and the fractional
change of angle per collision is on the average h1 cos θi . The momentum
relaxation time is the time after which the electron path is totally randomized, i.e.,
its “memory” is lost; hence,
τsc
τm ¼ ; (125)
h1 cos θi
where τsc is the average time between two collisions. Scattering with θ 90 is
memory erasing (Sect. 3.1 of chapter ▶ “Carrier Scattering at Low Electric
Fields”). Only the collisions in which all angles θ are equally probable result
in hcos θi = 0 and, therefore, yield τm = τsc. For small-angle scattering events,
one needs several scatterings before the momentum is relaxed, leading to τm >
τsc.
The Mean Free Path of Carriers Between collisions, the carrier traverses one free
path. The mean free path is obtained by averaging
2
v λðvÞ
λ¼ : (126)
hv 2 i
2 rffiffiffiffiffi
v 3π
λ ¼ hτ m i ¼ hτm ivrms : (127)
hvi 8
d hEi h Ei E L
¼ e F h vi ; (129)
dt τe
with EL the equilibrium energy at lattice temperature,
ð ð
E f dk v f dk
h Ei ¼ ð , and hvi ¼ ð : (130)
f dk f dk
The energy relaxation time is then obtained from Eq. 127 after switching off the
field, yielding
hEi EL
τe ¼ ; (131)
d hE i
dt
this means, τe is given by the ratio of the average surplus energy to the rate of energy
loss due to scattering and is a function of E; the energy-loss rate is not a simple
exponential function. It shows a maximum when the electron energy equals the
optical phonon energy.
The rate of momentum or energy loss depends on the actual scattering mecha-
nism. From gas-kinetic arguments, one obtains for collisions between an electron
and a lattice defect of mass M an energy-exchange rate of
τe mn
¼ : (132)
τsc ion M
The energy loss is negligible in one scattering event if the scattering center is a defect
atom, since M mn.
The fraction of energy lost by an electron in a collision with acoustic phonons can
also be obtained from an effective-mass ratio. Using the equivalent phonon mass
kT
mphonon ¼ ; (133)
v2s
which is on the order of 103. In other words, only 0.1% of the electron energy can
be lost to an acoustic phonon during any one-scattering event.
In contrast, the ratio of energy relaxation time to scattering time for optical
phonons is
46 K. W. Böer and U. W. Pohl
τe ℏω0
¼ ; (135)
τsc opt phonon kT
which is on the order of 1 at room temperature. This means that many scattering
events usually pass before the accumulated energy obtained from the field can be
dissipated to the lattice by emitting one optical phonon, while the momentum is
relaxed after one or only a few collisions. This modifies the rather crude model
given in Sect. 2.1 by introducing the momentum relaxation time for evaluating the
mobility and the energy relaxation time for Joule’s heating.21 A more detailed
discussion is given in chapters ▶ “Carrier Scattering at Low Electric Fields” and
▶ “Carrier Scattering at High Electric Fields”, when an analysis of the different
scattering mechanisms and a better estimate of the magnitude of energy obtained
from the field is given.
Interacting electrons and phonons exchange energy and momentum. A drift motion
superimposed on the random motion of one ensemble transfers part of the net
momentum to the other ensemble during scattering. This means that electrons
drifting in an external field tend to push phonons in the same direction, which causes
a slight temperature gradient in the field direction, superimposed on the homoge-
neous Joule’s heating. This process is called electron drag (Hubner and Shockley
1960).
Similarly, a temperature gradient tends to push electrons from the warm to the
cold end of a semiconductor. This is known as phonon drag. The drag effect can be
quite large, e.g., up to a factor of 6 compared to simple thermopower in p-type Ge at
20 K, as shown by Herring (1954).
When phonons propagate as acoustic waves, ac-electric fields can be induced; or,
vice versa, when sufficiently high electric fields are applied, coherent phonon waves
can be generated when the drift velocity of electrons surpasses the (sound) velocity
of the phonon waves (McFee 1966). These acousto-electric effects have technical
application for creating current oscillators (Bray 1969).
5 Summary
Electrons and holes are carriers of the current in semiconductors. Their effective
mass is modified by polarizing the lattice, but this influence, expressed as a change to
polarons, is negligible for most semiconductors. It is contained in the effective mass
21
In two-dimensional structures, significant momentum relaxation is found due to near-surface
acoustic phonon scattering; see Pipa et al. (1999).
Carrier-Transport Equations 47
obtained by cyclotron resonance and is commonly listed as the effective mass of the
carriers, i.e., of electrons or holes. Only for small polarons, which occur in semi-
conductors with narrow bands and large lattice polarization, the effective mass is
significantly affected and self-trapping occurs.
The current through a semiconductor is composed of a drift and a diffusion
current of electrons and holes. In homogeneous semiconductors, only one of these
four components is dominant. The drift current is determined by the electric field,
which acts as a slight perturbation of an essentially random walk of carriers, except
for very high fields. The additional energy obtained by carrier acceleration from the
field is given to the lattice by inelastic scattering, causing Joule’s heating. The
diffusion current is proportional to the carrier gradient up to a maximum diffusion
current, which is limited by the thermal velocity of carriers. Proportionality factor of
both drift and diffusion currents is the carrier mobility, which is proportional to a
relaxation time and inversely proportional to an effective mass tensor, the mobility
effective mass.
Drift and total currents are proportional to negative potential gradients: the first
one being the electrostatic potential and the second the electrochemical potential.
The proportionality factor of both is the conductivity. In spatially inhomogeneous
semiconductors, both an external field, impressed by an applied bias, and a built-in
field, due to space-charge regions, exist within the semiconductor. The external field
causes carrier heating by shifting and deforming the carrier distribution from a
Boltzmann distribution to a distorted distribution with more carriers at higher
energies within the band. In contrast, the built-in field leaves the Boltzmann distri-
bution of carriers unchanged; the carrier gas remains unheated at exactly the same
temperature as the lattice at every volume element of the crystal, except for statistical
fluctuations. A consequence of the difference between external and built-in fields is
the difference in determining the field dependence of the mobility, which requires an
averaging over carriers with different energies within the band. For a built-in field,
the averaging follows a Boltzmann distribution; in an external field, there are more
electrons at higher energies, and the distribution is distorted accordingly. This can
have significant impact for the evaluation of device performances when high fields
are considered.
The Boltzmann equation permits a sophisticated analysis of the carrier transport,
including changes in the carrier distribution. Such change in the distribution sub-
stantially influences the averaging, which is necessary to arrive at well-defined
values for a number of transport parameters – most importantly the relaxation
times. Significant differences can be defined between the time between two scatter-
ing events τsc, the momentum relaxation time τm, and the energy relaxation time τe.
Although the Boltzmann equation cannot be integrated in closed form except for a
few special cases, the deformed distribution function can be approximated for small
applied fields. It provides the basis for the investigation of various scattering
processes, which can be divided into essentially elastic processes with mainly
momentum exchange and, for carriers with sufficient accumulated energy, into
inelastic scattering with energy relaxation. The latter becomes more prevalent at
elevated temperatures and higher electric fields.
48 K. W. Böer and U. W. Pohl
References
Appel J (1968) Polarons. In: Seitz F, Turnbull D, Ehrenreich H (eds) Solid state physics, vol 21.
Academic Press, New York, pp 193–391
Bajaj KK (1968) Polaron in a magnetic field. Phys Rev 170:694
Böer KW (1985a) High-field carrier transport in inhomogeneous semiconductors. Ann Phys
497:371
Böer KW (1985b) Current-voltage characteristics of diodes with and without light. Phys Status
Solidi A 87:719
Bogoliubov NN, Bogoliubov NN Jr (1986) Aspects of polaron theory. World Scientific Publishing,
Singapore
Böttger H, Bryksin VV (1985) Hopping conduction in solids. Phys Status Solidi B 96:219
Bray R (1969) A Perspective on acoustoelectric instabilities. IBM J Res Dev 13:487
Brazovskii S, Kirova N, Yu ZG, Bishop AR, Saxena A (1998) Stability of bipolarons in conjugated
polymers. Opt Mater 9:502
Chakraverty BK, Schlenker C (1976) On the existence of bipolarons in Ti4O7. J Physique (Paris)
Colloq 37:C4–C353
Chandrasekhar S (1943) Stochastic problems in physics and astronomy. Rev Mod Phys 15:1
Chattopadhyay D, Rakshit PC, Kabasi A (1989) Diffusion of one- and two-dimensional hot
electrons in semiconductor quantum-well structures at low temperatures. Superlattice Micro-
struct 6:399
Christov SG (1982) Adiabatic polaron theory of electron hopping in crystals: a reaction-rate
approach. Phys Rev B 26:6918
Cohen MH, Economou EN, Soukoulis CM (1983) Electron-phonon interactions near the mobility
edge in disordered semiconductors. J Non-Cryst Solids 59/60:15
Comas F, Mora-Ramos ME (1989) Polaron effect in single semiconductor heterostructures. Physica
B 159:413
Conwell EM (1967) High-field transport in semiconductors. Academic Press, New York
Conwell EM (1982) The Boltzmann equation. In: Paul W, Moss TS (eds) Handbook of semi-
conductors vol 1: band theory and transport properties. North Holland Publishing, Amsterdam,
pp 513–561
Devreese JT (1984) Some recent developments on the theory of polarons. In: Devreese JT, Peeters
FM (eds) Polarons and excitons in polar semiconductors and ionic crystals. Plenum Press,
New York, pp 165–183
Drude P (1900) Zur Elektronentheorie der Metalle. Ann Phys 1:566. (On the electron theory of
metals, in German)
Dushman S (1930) Thermionic emission. Rev Mod Phys 2:381
Emin D (1973) On the existence of free and self-trapped carriers in insulators: an abrupt
temperature-dependent conductivity transition. Adv Phys 22:57
Esaki L, Tsu R (1970) Superlattice and negative differential conductivity in semiconductors. IBM J
Res Dev 14:61
Evrard R (1984) Polarons. In: Di Bartolo B (ed) Collective excitations in solids. Plenum Press,
New York, pp 501–522
Ferziger JD, Kaper HG (1972) Mathematical theory of transport processes in gasses. North Holland
Publishing, Amsterdam
Feynman RP (1955) Slow electrons in a polar crystal. Phys Rev 97:660
Frenkel JI (1936) On the absorption of light and the trapping of electrons and positive holes in
crystalline dielectrics. Sov Phys J 9:158
Fröhlich H, Pelzer H, Zienau S (1950) Properties of slow electrons in polar materials. Philos Mag
(Ser 7) 41:221
Haug A (1972) Theoretical solid state physics. Pergamon Press, Oxford
Hayes W, Stoneham AM (1984) Defects and defect processes in nonmetallic solids. Wiley,
New York
Carrier-Transport Equations 49
Herring C (1954) Theory of the thermoelectric power of semiconductors. Phys Rev 96:1163
Holstein T (1959) Studies of polaron motion Part 1: the molecular-crystal model. Ann Phys 8:325
Horio K, Okada T, Nakatani A (1999) Energy transport simulation for graded HBT’s: importance of
setting adequate values for transport parameters. IEEE Trans Electron Devices 46:641
Hubner K, Shockley W (1960) Transmitted phonon drag measurements in silicon. Phys Rev Lett
4:504
Kan EC, Ravaioli U, Chen D (1991) Multidimensional augmented current equation including
velocity overshoot. IEEE Electron Device Lett 12:419
Kartheuser E, Devreese JT, Evrard R (1979) Polaron mobility at low temperature: a self-consistent
equation-of-motion approach. Phys Rev B 19:546
Landau LD (1933) Electron motion in crystal lattices. Sov Phys J 3:664
Landsberg PT (1952) On the diffusion theory of rectification. Proc R Soc Lond A Math Phys Sci
213:226
Landsberg PT, Cheng HC (1985) Activity coefficient and Einstein relation for different densities of
states in semiconductors. Phys Rev B 32:8021
Lee TD, Low F, Pines D (1953) The motion of slow electrons in a polar crystal. Phys Rev 90:297
Leo K (1996) Optical investigations of Bloch oscillations in semiconductor superlattices. Phys
Scripta T68:78
Liouville J (1838) Note sur la théorie de la variation des constantes arbitraires. J Math Pures Appl
3:349. (Note on the theory of the variation of arbritrary constants, in French)
Lorentz HA (1909) The theory of electrons. Teubner Verlag, Leipzig
McFee JH (1966) Transmission and amplification of acoustic waves. In: Mason W (ed) Physical
acoustics, vol 4A. Academic Press, New York, pp 1–44
Mikhnenko OV, Blom PWM, Nguyen T-Q (2015) Exciton diffusion in organic semiconductors.
Energy Environ Sci 8:1867
Muljarov EA, Tikhodeev SG (1997) Self-trapped excitons in semiconductor quantum wires inside a
polar dielectric matrix. Phys Status Solidi A 164:393
Nag BR (1980) Electron transport in compound semiconductors. Springer, Berlin
Najafi E, Scarborough TD, Tang J, Zewail A (2015) Four-dimensional imaging of carrier interface
dynamics in p-n junctions. Science 347:164
Peeters FM, Devreese JT (1984) Theory of polaron mobility. In: Ehrenreich H, Turnbull D (eds)
Solid state physics, vol 38. Academic Press, Orlando, pp 81–133
Pekar SI (1954) Untersuchungen über die Elektronentheorie der Kristalle. Akademie Verlag, Berlin.
(Investigations on the electron theory of crystals, in German)
Pipa VI, Vagidov NZ, Mitin VV, Stroscio M (1999) Momentum relaxation of 2D electron gas due to
near-surface acoustic phonon scattering. Physica B 270:280
Seeger K (1973) Semiconductor physics. Springer, Wien/New York
Shah J (1994) Bloch oscillations in semiconductor superlattices. Proc SPIE 2145:144
Simoen E, Claeys C, Czerwinski A, Katcki J (1998) Accurate extraction of the diffusion current in
silicon p-n junction diodes. Appl Phys Lett 72:1054
Sommerfeld A (1928) Zur Elektronentheorie der Metalle auf Grund der Fermischen Statistik. Z
Phys 47:1 (On the electron theory due to Fermi’s statistics, in German)
Stratton R (1969) Carrier heating or cooling in a strong built-in electric field. J Appl Phys 40:4582
Tjablikov SV (1952) Zh Eksp Teor Fiz 23:381
Toyozawa Y (1981) Charge transfer instability with structural change. I. Two-sites two-electrons
system. J Phys Soc Jpn 50:1861
Velasco VR, García-Moliner F (1997) Polar optical modes in semiconductor nanostructures. Surf
Sci Rep 28:123
Waschke C, Leisching P, Haring Bolivar P, Schwedler R, Brüggemann F, Roskos HG, Leo K,
Kurz H, Köhler K (1994) Detection of Bloch oscillations in a semiconductor superlattice by
time-resolved terahertz spectroscopy and degenerate four-wave mixing. Solid State Electron
37:1321