Conduction in Metals
Conduction in Metals
Conduction in Metals
The fundamental equation for current transport is Ohm’s Law, taught in nearly every introductory E&M course,
and is given by
V = IR
where V is the voltage, I the current, and R is the resistance. In the SI system of units, voltage is an electromotive
potential EMF measured in volts (V), current is a charge flow per unit time measured in amps (A), and resistance
is the ratio of V/A as the derived unit Ohm with a symbol Ω.
Current is the charge moving past a fixed reference point per unit time, designed by the symbol Q with units
of coulombs (C). Charge itself is quantized (for our normal particles of electrons and protons) in units of the
fundamental electron charge
with 6.24 × 1018 electrons per coulomb. The symbol e is an unsigned quantity corresponding to this fundamental
charge constant. For signed charges, such as an electron or proton, we will use the symbol q where
The relationship V = IR is an extensive representation of Ohm’s Law. The term extensive refers to a property
that is dependent on the size of an object. For example, doubling the length of a wire doubles the resistance.
In the physical world, we are almost always concerned with these extensive properties, but for theory we work
in units and equations which are sample size independent, referred to as intensive properties. The intensive
⃗ and the property corresponding to the current I
property corresponding to a potential V is the electric field E,
is the current density J.
Consider a simple wire with uniform cross-section (a rectangular prism), which has a resistance (R) determined
as
L
R = ρ
A
where L is the length of the bar, A is the cross-sectional area, and ρ is the resistivity of the material. In SI units,
the resistivity would have units of Ω-m, but it is common to use the non-SI units of Ω-cm. The resistivity of
several common metals are given in Table 1.2, with the best metals having a resistivity approaching 1.5 µΩ-cm.
The form V = IR is clearly an extensive expression of Ohm’s law since all of the parameters depends on not only
material properties but also the physical dimensions of the sample. These physical dimensions can be removed by
shifting to the electric field E⃗ and the current density J.
⃗ Formally the electric field is the gradient of the potential
V , and the current density is just the area normalized current, and are given by
E⃗ = −∇V
I⃗
J⃗ =
A
For the simple rectangular “wire”, the electric field is just E = V /L, pointing in the direction from the positive
to the negative potential. Similarly, the current density is just I/A. Rearrangement of the extensive form Ohm’s
1
Table 1.2: Resistivity and atomic density at 25 o C for several pure,
large grain, metals and alloys. Data taken from the Handbook of
Chemistry and Physics, 95th Edition, 2014-2015.
Law yields
L
V = IR = I ρ
A
V I
= ρ
L A
E⃗ = ρJ⃗
For the majority of this course, we will think of E⃗ as a generalized force (analogous to the pressure in a water
pipe) and the current density J⃗ as the response to that force (the amount of water flowing). Thus it makes better
sense to rewrite this expression as
1 ⃗
J⃗ = E = σ E⃗
ρ
where σ is called the conductivity of the material. Numerically, the conductivity and resistivity are just inverses
of each other. This final form, (J⃗ = σ E),
⃗ is the intensive form of Ohm’s law written in terms of the local electric
field and current density.
Conductance, the inverse of resistance, is measured in SI units of Siemens (S), which is a derived quantity of
A/V. However, the unit is also often written as the mho (inverse ohm). Conductivity then has units of S·m−1
in SI units, or non-SI equivalents of mhos/cm, S/cm, (Ω-m)−1 , or (Ω-cm)−1 . The range of conductivities is
enormous, with very good metals approaching 106 mhos/cm and very good insulators exhibiting conductivity in
the 10−20 S/cm range.
There is no physical requirement that the current flow (density) be completely parallel to the direction of the
applied electric field. Indeed, in many materials, there is an anisotropy in the conductivity with electron flow
much easier in some crystal directions than in others. Due to this anisotropy, if the field is not aligned with one
of the principal axes of the material, the current density will not flow parallel to the applied field. For this reason,
the conductivity must be treated as a second rank tensor relating the vector electric field with the vector current
2
density. We denote the conductivity tensor as σ and Ohm’s Law becomes
J⃗ = σ E⃗
A complete treatment of tensors is beyond the scope of this text, but second rank tensors can be conveniently
represented in a matrix notation and is sufficient for this discussion. The vector quantities J, ⃗ E,
⃗ and σ exist
independent of any coordination system; specific coordinate representations are defined to enable the math. By
defining a set of basis vectors, we can represent any vector as a linear combination within that choice of a
coordinate system. In the following discussion, we follow the convention of Kusse and Westwig labeling these
unit vectors as êi , with three non-colinear vectors required to span 3D space. (Other authors use x̂i as these unit
vectors). The current density is then represented as
where the final representation without the explicit summation sign is referred to as the Einstein notation. The
Einstein notation implies summation over any index that is repeated in an expression so uijkl vk ml implies
summation over the k and l indices. The conductivity tensor in this notation is
The elements σji can be understood as being the coupling into a current in the êj direction from an electric field
in the êi director.
To emphasize, the second rank tensor σ is a property of the material independent of the coordinate system while
the elements σji are the specific representation within a specific coordinate system. Coordinate transformations
(ie. rotation) is a very specific operation since the underlying vectors must also be rotated. Indeed, in many texts,
a tensor is defined in terms of the way that it transforms under a coordinate change.
The tensor relationship between the field and the current density now becomes
J⃗ = σ E⃗
Jj = σji Ei
There is enormous power in this simplified notation between the coupling of a generalized response (J) ⃗ and a
⃗ especially when those elements are themselves higher order tensors. For example, stress
generalized force (E),
and strain are second rank tensors and the relationship between them is based on a fourth rank tensor. A vector
can be considered as a first rank tensor and a scalar (constant) is often referred to as a zero rank tensor.
However, we limit ourselves here to the tensor of second rank and below. In this case, the expression Jj = σji Ei
can also be written as a matrix equation treating J⃗ and E⃗ as column vectors
J1 σ11 σ12 σ13 E1
J2 = σ21 σ22 σ23 E2
J3 σ31 σ32 σ33 E3
Note that the unit vectors ê1 , ê2 and ê3 need not necessarily be orthogonal to each other, nor need they be the
simple directions x̂, ŷ and ẑ. Of course, things are often simpler when they are indeed orthogonal.
While there are in general nine elements of the conductivity tensor, it can be shown that the tensor must be
symmetric with σij = σji . This reduces the number of independent elements from nine to six and ensures that
the matrix is Hermetian.
Considering σji as a matrix, we know that there is a coordinate transformation (rotation) that will diagonalize
the matrix. In this new coordinate system, referred to as the principal axes, the tensor has elements (possibly the
same or different) only along the diagonal. This is equivalent to the diagonalization of a matrix via the eigenvalue
3
and eigenvector construction. The eigenvectors correspond to the principal exes (in the original êi representation)
and the eigenvalues are the diagonal elements. Because the matrix itself is Hermetian, the eigenvectors will be
orthonormal (or in the case of degenerate eigenvalues, can be constructed to be orthonormal).
In the coordinate system based on these principal axes unit vectors, the conductivity tensor reduces to just the
eigenvalue matrix
σ11 0 0
σ = 0 σ22 0
0 0 σ33
where σ11 , σ22 and σ33 are the principal axes conductivities.
Finally most materials, and certainly all large scale polycrystalline materials, are isotropic and the conductivity
tensor reduces to a simple diagonal tensor with σ11 = σ22 = σ33 = σ and
σ 0 0
σ = 0 σ 0
0 0 σ
Consequently, for an isotropic material the scalar relationship J⃗ = σ E⃗ holds. This is certainly the most common
case.
While this discussion has focused on the tensor properties of the conductivity, Ohm’s Law can be equally written
E⃗ = ρ J⃗ in terms of the resistivity tensor ρ. For scaler properties, we know that σ = 1/ρ and a similar relationship
holds for the tensor representations
−1
σ11 σ12 σ13 ρ11 ρ12 ρ13 −1
σ = σ21 σ22 σ23 = ρ21 ρ22 ρ23 = ρ
σ31 σ32 σ33 ρ31 ρ32 ρ33
In the principal axes coordinate system, the matrix inverse is simple with σjj = ρ−1
jj but otherwise the full matrix
inverse must be computed. Values of the principal resistivities for a number of elements are given in Table 1.3.
In many cases, the symmetry of the crystal reduces the number of independent resistivities from three to two.
For example, in a hexagonal crystal, there are only two independent conductivities; one within the plane (the a
directions) and a second in the perpendicular direction (along the c-axis). In low symmetry crystals such as Ga
(an orthorhombic structure), all three conductivities are independent.
While elements have relatively small anisotropies and the effect can only be observed in single crystals, many
compounds and engineered materials can have extremely large anisotropies. The high temperature superconductor
YBCO (YB2 Cu3 O7−δ ) structurally consists of planes of copper oxide separated by the other ions. Conductivity
along these conducting Cu-O planes is much higher than electron transport across the planes. Graphite, consisting
of sheets of highly conductive graphene loosely stacked, exhibits a similar anisotropy. In polymers, it is possible
to create linear molecules that conduct charge only along the backbone. Aligning these molecules into a matrix
results in a true one-dimensional conductor with potential applications for interconnects. Aligned nanowires of
silver in an insulating matrix behave similarly.
Organic charge-transfer complexes can be intentionally designed for strong conductivity anisotropy. TCNQ (tetra-
cyanoquinodimethane), first reported in 1962, is an electron acceptor with high conductivity along the linear
molecule. When mixed with tetrathiafulvalene (TTF), a strong electron donor synthesized in 1970, the combina-
tion forms a strong charge-transfer complex referred to as TTF-TCNQ [ref]. The complex forms a well-structured
crystalline solid with almost metallic conductivity along the aligned stacks.
At this point, it is appropriate to simply define a metal as a material which conduct electricity well (large electrical
conductivity), and which are generally characterized by metallic bonding. The metallic bonding implies that some
fraction of the valence electrons are contributed to a delocalized “sea” of electrons that participate in the transport
4
Table 1.3: Principal axes resistivity of several ele-
mental metals (µΩ-cm).
properties. The number of electrons contributed can be roughly considered to be the valency. Group IA elements,
such as Li, Na and K, contribute one electron per atom. Similarly, Cu, Ag and Au also contribute one electron
per atom (Group IB) while Al and Ga contribute 3 electrons per atom (Group IIB). This sea of electrons move
in response to an applied electric field, ultimately becoming the current flow that we associate with conductors.
The free electrons in the “electron sea” are continuously in motion. Classically, we think of them as free particles
moving freely in the general potential wells created by the metallic atom cores, driven by thermal energy associated
with any finite temperature. Their velocity in this random motion is called the thermal velocity which, for electrons
in metals, is on the order of
Surprisingly, this velocity is almost independent of temperature and is an order of magnitude larger than expected
from simple classical mechanics arguments and a thermal energy on the order of kB T . 1
For metals, we must ultimately deal with electrons in the quantum mechanical limit. We will handle this properly
later as we discuss the quantum mechanical free electron model in the context of band theory. But briefly, while
classically all electrons can potentially have the same energy, quantum mechanics and Pauli’s exclusion principle
requires every electron to be in a different quantum state. As a result, all of the low energy states are filled and
the “mobile” electrons in a metal have an energy (the Fermi Energy) roughly two orders of magnitude greater
than the classical thermal energy (EFermi ≈ 5 eV versus kB T ≈ 0.026 eV at room temperature). Ultimately, the
thermal velocity of electrons in the metal is more precisely the Fermi velocity of the highest energy electrons in
a partially filled band.
There are two consequences to this quantum mechanical behavior. This first is that the thermal velocity is an
order of magnitude higher than expected. The second is that this velocity is almost independent of temperature.
1 Classically,electrons should have an average energy on the order of 32 kB T , where kB is Boltzmann’s constant (1.38 × 10−23 J/K
= 8.616 × 10−5 eV/K) and T is the absolute temperature. The “classical” thermal velocity is determined by equating the kinetic
energy of the electrons with the thermal energy. At room temperature
1 2 3
mvth = kB T
2 2
r r
3kB T (3)(1.38 × 10−23 J/K)(300 K)
vth = =
m 9.11 × 10−31 kg
= 1.17 × 105 m/s
which is an order of magnitude too small. The result, however, turns out to be approximately valid for electrons in another class of
materials – semiconductors.
5
In the classical limit, the thermal velocity would scale as the square root of temperature. But since the Fermi
Energy is so much higher than thermal energies, the thermal velocity is essentially constant . . . unless you can
find a way to work with metals at temperatures around 30,000 K.
For completeness, we should very briefly address the thermal velocity of electrons in semiconductors. The electron
density in a metal is on the order of 1023 e− /cm3 while in a semiconductor the density varies widely over orders of
magnitude with a maximum of typically 1020 e− /cm3 . With so many fewer electrons to deal with, Paul’s exclusion
principle is not nearly as imposing and electrons in a semiconductor behave much more classically. Indeed, the
thermal velocity (which will be dealt with as a saturation velocity) is on the order of 105 m/s for many materials.
Despite the very high thermal velocity, electrons do not travel very far before they are scattered off into a
new directions. The scattering time, whose average is designated at τ , is typically on the order of 20 fs
(20 femtoseconds = 2 × 10−14 s). These scattering events occur randomly and the probability of the time
between scattering can be approximated as a Poisson (exponential) probability distribution. Let p(t) dt be the
probability that an electron will be scattered in a window of time between t and t + dt. The Poisson distribution
with a mean scattering time of τ is given by
1 −t/τ
p(t) dt = e dt
τ
Between collisions, the electrons travels a distance ℓ known as the mean free path. On average, ℓ is just the
product vth τ and is on the order of 10 nm. We should compare this mean free path to the typical distance
between atoms in the condensed phase. The spacing d between atoms can be estimated from the atomic density
NA ≈ 5 × 1022 at/cm3 of condensed phase materials (see Table I).
1/3
1
d≈ ≈ 0.2 − 0.3 nm
NA
Thus an electron will travel on the order of 30-40 atoms before being scattered. This is not so short that we have
to painfully deal with the atomic granularity of materials, nor is it so large that we can completely ignore the
granularity.
In the absence of an applied electric field, this continuously scattered electrons thermal velocity undergoes random
motion with no net transport in any particular direction. But with an applied field E, ⃗ there is an additional drift
component to the electron’s trajectory, biasing the average motion in a definite direction and giving rise to a net
current. This biased motion is called the drift velocity and is designated as vd .
The force acting on a particle in an electric field is q E⃗ giving an acceleration a = q E/m
⃗ where q is the signed charge
on the particle and m is the mass. During the time between scattering events ti , the particle will experience a
net drift motion of 12 at2i . The average drift velocity vd can be computed as the total drift distance divided by the
total time
N
q E⃗ X 2
1 2 1 2 1 2 t
∆x at + at + at + . . . 2m i=1 i
vd = = 2 1 2 2 2 3 =
∆t t1 + t2 + t3 + . . . N
X
ti
i=1
q E⃗ ⟨t2 ⟩
=
2m ⟨t⟩
The expression ⟨t2 ⟩ is a shorthand for the average of scattering time squared.
6
By integrating over the probability distribution function, the two averages can be readily shown to be ⟨t⟩ = τ
and ⟨t2 ⟩ = 2τ 2 . 2 3 These results yield then a very simple classical expression for the drift velocity
! !
q E⃗ ⟨t2 ⟩ q E⃗ 2τ 2
vd = =
2m ⟨t⟩ 2m τ
qτ
= E⃗
m
= µ E⃗
where we have defined the mobility (µ) as the proportionality constant linking the drift velocity to the applied
electric field. The units of the mobility are m2 V−1 s−1 (in SI units) or, in more common use, cm2 /V-s. We can
roughly estimate the mobility of a good metal knowing a typical scattering time of τ = 20 fs.
This is indeed the scale of the mobility for a good conductor such as copper. But the value spans orders of
magnitude even for “good” electronic materials. Many organic semiconductors show mobilities on the order of
10−5 cm2 /V-s, silicon has a room temperature electron mobility of 1400 cm2 /V-s, and some high performance
semiconductors can have mobilities in excess of 100, 000 cm2 /V-s at low temperatures.
This is a remarkable result establishing a linear relationship between the drift velocity (response) and the applied
electric field (driving force), with Ohm’s Law critically dependent on this linearity. The one key assumption is
that the drift component of the velocity is small compared to the thermal velocity (vd < vth ). For metals with
a mobility of 20 cm2 /V-s and a thermal velocity of 2 × 106 m/s, non-linearities would not be expected until
electric fields of 10 MV/cm (109 V/m), much higher than one can realistically establish in a metal. However,
given the smaller thermal velocity in semiconductors, and the much higher mobilities, these non-linearities are
readily observed as a “saturation” in the drift velocity at high electric fields.
The drift velocity refers to the motion of a single electron. To get the total current, we have to address the
motion of all the electrons in the system. Let n be the density of conduction electrons in the metal. On average,
these electrons are traveling at a velocity µE⃗ and carrying a charge e (actually q but the sign of the carrier will
not matter). The current density J, ⃗ the total amount of charge flowing per unit area and per unit time, is then
2 Determining ⟨t2 ⟩ requires integrating all of the possible values of t weighted by the probability that the scattering time was t.
Given the Poisson distribution, the mean square scattering time and the mean scatter time are given by Gamma functions
Z ∞ Z ∞
1 −t/τ
h i
⟨t2 ⟩ = t2 p(t) dt = t2 e dt
0 0
τ
Z ∞
= τ2 x2 e−x dx = τ 2 Γ(3) = 2τ 2
0
Z ∞ Z ∞
1 −t/τ
h i
⟨t⟩ = t p(t) dt = t e dt
0 0
τ
Z ∞
= τ x e−x dx = τ Γ(2) = τ
0
3 The Gamma function, an analytical extension of the factorial to complex arguments, is defined by the integral
Z ∞
Γ(z) = tz−1 e−t dt
0
While perhaps not as well known as sin(z), it is nonetheless common in engineering. Integrating by parts, you can readily show that
Γ(z + 1) = zΓ(z) and Γ(1) = 1. This then implies that Γ(n + 1) = n! for integers. We will use this function extensively.
7
simply
J⃗ = n e ⃗vd
= (n e µ)E⃗
= σ E⃗
where σ = neµ is the conductivity of the material. As mentioned previously, conductivity has units of S·m−1 in
SI units, or non-SI equivalents of mhos/cm, S/cm, (Ω-m)−1 , or (Ω-cm)−1 . The resistivity of the material and its
conductivity are just inverses of each other.
σ = neµ
1 1
ρ = =
σ neµ
We can now estimate the conductivity of a simple metal. Consider a metal with two valence electrons per atom
(such as Mg). With atomic densities on the order of 5 × 1022 at/cm3 , the electron density n will be on the order
of 1023 e− /cm3 . Given a mobility of 30 cm3 /V-s, we have
σ = neµ
= (1023 e− /cm3 )(1.6 × 10−19 C/e− )(30 cm2 /V-s)
=
480, 000 S/cm
1
ρ = = 2.1 × 10−6 Ω-cm
σ
= 2.1 µΩ-cm
This is a very reasonable estimate given that pure copper has a room temperature resistivity of 1.6 µΩ-cm.
We can combine our classical expression for the mobility with the expression for the conductivity to obtain
ne2 τ
σ = neµ =
m
As the conductivity depends only on the charge squared (e2 ), it really doesn’t matter if the charge carriers are
negative (electrons) or positive (ions). Indeed, this entire Drüde derivation applies equally to current transport
in metals, semiconductors and ionic materials (where charge is carried by mobile ions and charged crystal defects
rather than mobile electrons).
Finally, as the electron density changes only by a factor of 2-3 between different metals, the key factor determining
the conductivity (or resistivity) of materials is the mobility, or equivalently the scattering time τ .
While we introduced the concept of a scattering time (τ ) and a mean free path (ℓ = vth τ ), we have not yet
discussed what physically causes scattering of electrons. Electrons in crystalline metals must be properly treated
as waves traveling through a periodic array of the atoms on the lattice. As long as all the atoms are in their
places and of the correct type, this electron wave can travel freely much like an athlete running through the seats
in a stadium. Scattering occurs when this perfect structure is disturbed and the electron experiences an unusual
environment. Some of the potential origins for this scattering include:
Lattice scattering: Atoms on the lattice are continuously vibrating due to thermal energy and, if sufficiently
far out of place, will scatter the electron wave. The amplitude of these vibrations depends on temperature
and hence the mean scattering time decreases with increasing temperature. This will also often be referred
to as phonon scattering; phonons are a formal way to describe vibrations within a solid material.
Impurity scattering: Impurities, even at low concentrations, will disrupt the ordered structure and scat-
ter the electron wave. In this context, impurity scattering refers to elements which are present in small
concentrations.
8
Alloy scattering: In an alloy consisting of significant concentrations of one or more elements, the periodic
structure is severely compromised and the electron will undergo scattering on very short length scales.
Impurity and alloy scattering are essentially identical but refer to different regimes.
Surfaces scattering: The electron must remain in the solid and hence must be scattered back into the material
at any free surface or interface to a non-conducting material.
Grain Boundaries scattering: The periodic arrangement of atoms is broken at the boundary between indi-
vidual grains in the material. Transitioning from one grain to another generally results in scattering.
Dislocations scattering: Like grain boundaries, dislocations also disrupt the lattice periodicity near the core
resulting in scattering.
With the exception of the lattice (phonon) scattering, all of these mechanisms are related to the structure of the
metal sample and hence are, to first order, independent of temperature. Lattice scattering,
√ in contrast, is strongly
dependent on temperature as the vibrational amplitude increases as approximately T .
Each of the scattering mechanisms can be assigned its own τ . These individual term combine to give the average
scattering time in the sample; however, the shortest scattering times dominate and the times sum as the inverse:
1 1 1 1 1
= + + + + ...
τ τlattice τimpurities τgrain boundaries τsurfaces
Both the mobility and the conductivity scale proportional to the mean scattering time τ , but the resistivity scales
inversely with τ . We can write an expression for the resistivity as
m m 1
ρ = 2
=
ne τ ne2 τ
m 1 1 1 1
= + + + + ...
ne2 τlattice τimpurities τgrain boundaries τsurfaces
m m m m
= 2
+ 2 + 2 + 2 + ...
ne τlattice ne τimpurities ne τgrain boundaries ne τsurfaces
where the total resistivity of a material can be “considered” to be the sum of the resistivities of the individual
terms. In general, the only term with a temperature dependence is the lattice contribution and hence it is useful
to write this explicitly as
For a pure single-crystal metal, the dominant scattering mechanism is due to thermal vibration of the atoms
themselves. To understand the temperature dependence of this term, it is necessary to create a formalism for the
scattering probability and the mean scattering time. In scattering theory, the probability for scattering is treated
as a cross section σs 4 which is size (area) that a solid target would need to be to produce the same number of
scattering events. Cross sections are measured in units of an area (m2 for SI), but for atomic processes are often
given in units of the barn which equals 10−24 cm2 or 10−28 m2 .
4 The scattering cross section in electronics is at times designated as S to avoid confusion with the conductivity. But S can be
confused with the unit of conductance (Sieman) so isn’t really much better. The physcs community universally uses σ as a cross
section and here we just add the subscript to denote it as a scattering cross section.
9
15 50
Silver Tungsten
Copper Molybdenum
Gold Platinum
Aluminum 40 Strontium
Calcium Tantalum
Resistivity (µΩ-cm)
Resistivity (µΩ-cm)
Magnesium Lead
10
30
20
5
10
0 0
0 200 400 600 800 0 200 400 600 800
Temperature (K) Temperature (K)
Figure 1.5.2 Temperature dependent conductivity of elemental metals showing nearly linear dependence or resis-
tivity with temperature. (a) The high conductivity metals; (b) moderate conductivity metals.
The total probability for scattering per unit time (rate) is equal to the product of the size of the scattering objects
(σs ), the number of scattering sites that are present (NS ), and the velocity that the electron is traveling (vth ).
The mean lifetime τ is just the inverse of this scattering rate.
1
τ =
S Ns vth
Given a typical mean scattering time of 20 fs, we can estimate the cross section of thermal vibrations at room
temperature. NS is just the atomic density, which is typically 5 × 1022 at/cm3 , and the thermal velocity is
typically 1.5 × 106 m/s. The cross section is then
1 1
σs = =
τ NS vth (20 × 10−15 s)(5 × 1022 at/cm3 )(1.6 × 108 cm/s)
= 6.25 × 10−18 cm2 (6, 250, 000 barns)
This corresponds roughly to a disk with a diameter of 0.028 nm, which is reasonable for the vibrational amplitude
of atoms that are separated by 0.2-0.3 nm.
In a classical picture of the lattice, each of the atoms sits in a potential energy well generated by the neighboring
atoms. Vibrations are essentially harmonic oscillators within this well and we can estimate the temperature
dependence of the vibration amplitude from the equipartition theorem.5 If we let K be the effective spring
constant and a the vibrational amplitude, equipartition gives the amplitude of this potential energy as
1 1
Ka2 = kB T
2 2
and hence a2 ∝ T . Now, the cross sectional area will be proportional the πa2 and hence also proportional T . This
leads to the conclusion that the mean scattering time due to lattice vibrations scales inversely with temperature
and, since resistivity is inversely proportional to τ , the resistivity will scale proportional to temperature.
ρlattice ∝ T
5 The equipartition theorem states that, at high temperature, every quadratic term in an the energy Hamiltonian will have an
average energy of 12 kB T . For a harmonic oscillator, the Hamiltonian has two quadratic terms, one corresponding to the kinetic energy
of motion and one corresponding to the potential energy of the spring.
10
60
160
Iron
Nickel
50 Chromium 140
120
Resistivity (µΩ-cm)
Resistivity (µΩ-cm)
40
100
30
80
20 60
40
10
20
0 0
0 200 400 600 800 0 100 200 300 400 500 600 700
Temperature (K) Temperature (K)
Figure 1.5.2: Temperature dependent conductivity of elemental metals that show significant deviation from
linearity. (a) anomalous transition metals including magnetic iron and nickel; (d) strongly non-linear behavior of
manganese due to spin-disorder contributions.
Consequently, a very simple estimate for the resistivity of a pure material based on the known resistivity ρ0 at a
reference temperature T0 would be
T
ρ(T ) = ρ0
T0
Figure 1.5.2 shows the resistivity of a number elemental metals as a function of temperature. The linear increase
of resistance with temperature is remarkably good for temperatures from 100 K to 800 K for almost all metals.
However, our simple argument suggests that the resistivity only extrapolate to zero at 0 K, the data show the
resistivity tending to zero at a temperature of 50-100 K (on a linear scale).
As usual, at low temperature quantum mechanical effects become important and lattice vibrations become quan-
tum mechanically frozen out. When lattice scattering ceases to be important, the resistivity approaches a limiting
resistivity called the residual resistivity (ρr ) which depends critically on sample purity. Even isotopic variations
in an element contribute to the residual resistivity, acting comparable to impurity atoms in an otherwise homo-
geneous crystal lattice. For extremely pure aluminum, the residual resistivity can be as low as 10−4 µΩ-cm below
about 3 K (16,000 times more conductive than copper at room temperature).
Given the nearly linear relationship above some temperature between conductivity and temperature in pure
metals, the conductivity at moderate temperatures can be expressed as
where alpha is just a fitting parameter and T0 is the extrapolated temperature where the conductivity appears
to go to zero. Note that this α is not the same as the αTCR which is the temperature coefficient of resistivity.
However, there are exceptions to the rule that lattice scattering is linear with temperature. Our simple Drüde
picture of electron conductivity ignores all the potential interactions between different “types” of electrons. In
the ferroelectric elements such as iron and nickel, there are two distinct electron contributions that interact to
give rise to magnetism and a very non-linear temperature dependence for the resistivity, as seen in Figure 1.5.2.
In these materials, the resistivity is roughly quadratic with temperature below the Curie temperature.6 Above
the Curie temperature, the resistance returns to a nearly linear behavior as predicted above. One of the most
6 The Curie temperature is a critical temperature where the material transitions from ferromagnetic to paramagnetic. For Fe, Co
and Ni, the Curie temperatures are 1043, 1400 and 631 K respectively.
11
egregious violations of this simple model arises in manganese where there is a very large spin-disorder contribution
to the electrical resistivity. In manganese, the resistivity has a very weak temperature dependence above 70 K
before dropping precipitously at low temperatures.
In addition to thermal vibrations scattering the electron wave, any lattice discontinuity will similarly cause
scattering. Impurity atoms, in particular, not only distort the lattice in the vicinity of the atom, but also may
have be in a charged state. At low concentrations, we can argue that the impurities act separately with a cross
section per impurity related to the specific combination of impurity and matrix. The inverse scattering time,
proportional to the resistivity contribution, is given by
1
ρimpurities ∝ = σs NI vth
τ
where NI is the number of impurities in the lattice. We define XI to be the molar (atomic) fraction of impurities
in the lattice. The proportionality allows us to write for impurity resistivity as
ρimpurity = C XI
where C is called Nordheim’s constant. The extension of this rule to non-dilute binary alloys is known as
Nordheim’s rule where the alloy contribution to the resistivity is given by
ρalloy = C XA XB = C XA (1 − XA )
where XA and XB are the atomic concentrations of each element in the binary alloy. Using a simple linear scaling
for the lattice contribution, the resistivity of an alloy can be approximated as
ρ = XA ρA + XB ρB + C XA XB
where ρA and ρB are the resistivities of the pure elements (lattice scattering). This relationship must be modified
if the number of electrons contributed by the two elements are different since the linearly extrapolation is only
valid for 1/τ and not the electron density.
Figure 1.5.3 shows the impact of impurities on the conductivity of high purity copper at low (impurity) con-
centrations. Nordheim’s constant C is essentially the slope of these curves and varies enormously depending on
the ionic size, structural compatibility, and electronegativity. Impurities such as Ti dramatically increase the
resistivity even at extremely small concentrations while Ag and In have relatively minor impact. Values for a few
impurities in Cu and in Au are given in the table below.
At higher concentrations, impurity scattering becomes more alloy scattering and corresponds to large composition
changes. Figure 1.5.3 show the variation of the resistivity of several alloys as a function of composition across the
entire composition range. As expected, the maximum resistivity occurs for alloys near 50%. The temperature
dependence for alloy scattering is almost non-existent and hence resistivity curves are parallel for temperatures
from 100 K to 400 K.
12
Figure 1.5.3 Effect of impurities on the resistivity of (a) tough pitch copper pure copper and (b) oxygen-free
copper as a function of the concentratin of various impurities. At low concentrations, the increase in resistivity
is linear though the slope varies enormously depending the atomic size and electronegativity. Data from . . .
60 16
Resistivity [µΩ-cm]
40
10
30 8
6
20
10
2
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Atomic percent of alloy element Atomic percent Au
Figure 1.5.3 (a) Resistivity of alloys as a function of composition for various systems. (b) Resistivity of the Cu-Au
system as a function of composition at various temperatures.
13
Figure 1.5.3 Both impurities and dislocations decrease the mean scattering time with comparable impact. (a)
Resistivity of aluminum alloys. Al 1100 is commercially pure while 3003 is alloyed with 1.2% Mg. (b) Resistivity
of Cu-Sn alloys as a function of both the Sn concentration and with cold work. The effects are additive. Figure
from “An Introduction to Electronic and Ionic Materials”, Wei Gao and Nigel Sammes, World Scientific.
Figure 1.5.3 shows the effect of alloying on copper alloys at low temperature. As predicted by Matthiessen’s rule,
the addition of impurities simply shifts the curves upward. Dislocations behave very similarly to impurities due
to scattering from the dislocation cores. Effects of cold-work and alloy are additive as seen in the Cu-Sn system.
The Nordheim relationship assumes that the impurities are randomly positioned on lattice sites in a homogeneous
alloy. Ordered alloys and inhomogeneous alloys violate this assumption. In a few binary systems, there are
specific compositions that form ordered compounds. For example, Cu and Au form ordered alloys near Cu3 Au
and CuAu. For CuAu, the lattice nearly alternates between Cu for compositions near 50/50. Annealed samples
at this composition show a marked reduction in resistivity as the impurity effect disappears.
The second violation occurs for any system in a two phase equilibrium. In such systems, the sample consists of
independent regions of materials with different conductivities. The “average” resistivity in such cases depends
critically on the microstructure and volume fraction of the two phases.
Grain boundary and surface scattering are similar in their behavior; both occur as the electron experiences
scattering as the crystal lattice “terminates”. These become critical only as the dimensions approach the mean
free path of electrons, typically 10-20 nm at room temperature but potentially much longer at low temperatures.
These effects become significant in semiconductor processing where films on these dimensions are common.
The issue of scattering at grain boundaries and surfaces is further complicated by the potential for “coherent”
scattering that does not randomize the electron’s velocity and cancel out the accumulated drift velocity component.
For example, at a free surface, the electron can be specularly scattered and continue on its trajectory as if hitting
a mirror. These specular reflections do not contribute to the reduction in the mean scattering time τ . Indeed,
for ultrahigh quality films such as produced by Molecular Beam Epitaxy (MBE), specular reflections are highly
probable. If p is the probability for specular reflection, the resistivity of thin films is given by the Fuchs-Sondheimer
equation which can be simplified as
3 ℓ
ρ = ρ0 1 + (1 − p)
8 t
where ρ0 is the resistivity of the pure bulk crystal and ℓ/t is the ratio of the mean scattering length (ℓ) to the
film thickness (t). In thin films, the effect of surface scattering becomes evident at thicknesses of 50-100 nm and
14
can represent an order of magnitude increase as thicknesses decrease below 10 nm. Potentially all four surfaces
of a rectangular wire would contribute to increasing the resistivity (though with potentially different p), which is
becoming a major challenge as Cu interconnects in microelectronics move to sub-20 nm dimensions.
Similar arguments can be made for grain boundary scattering. However, rather than specular scattering, there
is a some probability for the electron to be reflected at the interface, or to be partially scattered traversing the
boundary. The resistivity due to the grain boundary scattering is often represented by the Mayadas-Shatkez
formula where
4 ℓ R
ρ = ρ0 1 +
3 d 1−R
where ρ0 is the resistivity of the pure bulk crystal, ℓ/d is again the ratio of the mean scattering length to the
average grain diameter, and R is a reflection probability coefficient. As R goes to zero, the grain boundaries have
no effect on the resistivity while as R goes to unity, the electrons remain trapped in their individual grains, can
never cross boundaries, and the resistivity approaches infinity.
The temperature dependence of the resistivity (or resistance) is very often characterized by a parameter αT CR
referred to as the TCR or temperature coefficient of resistivity. It may help to think of it as the electrical analog
to the coefficient of thermal expansion. Formally, the TCR is defined as
1 dR 1 dρ
αT CR = =
R dT ρ dT
Materials with a very small TCR are often used to make circuit devices such as resistors since their values do
not change appreciably with temperature swings. In contrast, a material with a large TCR might be used as a
temperature sensor directly.
For pure elements, the resistivity is given by just the lattice contribution (and residual resistivity) which is
approximately proportional to temperature. In a material where ρ = AT , the TCR is just the inverse temperature.
1 dρ 1
αT CR = = (A)
ρ dT AT
1
=
T
Of course, the resistivity is not directly proportional to temperature (an offset) and so the TCR is not exactly
the inverse temperature. However, it remains a good approximation. At 273 K, αCu = 1/232, αAl = 1/233,
αAg = 1/244, and αAu = 1/251.
The addition of alloying impurities, surfaces and grain boundaries all serve to reduce the TCR since these
contributions to the resistivity are roughly independent of temperature.
The simple resistor is probably the oldest electronic device in existence, serving only to restrict the current flow
for a given voltage. In many (or most) applications, it is desirable to maintain a constant resistance independent
of the environmental temperature, thus requiring a very low TCR. Very low TCR resistors are based on alloys
like manganin or nichrome.
Bulk metal foil resistors have TCRs of 0.2 ppm/K, high precision thin film are 5 ppm/K, precision thick film are
50 ppm/K and inexpensive wirewound resistors can vary from 100 to 1000 ppm/K.
In practice, wirewound resistors suffer from parasitic inductance and capacitance from the windings. Both of
these issues can be mitigated at the expense of more complex manufacturing. (Remind prof that he has more to
write here)
15
Table 1: List of common alloys for use in precision wire-wound resistors. Many are based on very low TCRs to
minimize variation with temperature.
Hall Effect:
Interaction of magnetic fields with electron flow.
σ = neµ
Here, n and µ occur only as the product nµ. How do we measure these independently?
[INSERT MAGIC HALL EFFECT FIGURE HERE]
F⃗ = q E + ⊑§ ⃗
⃗ B
Fmag ⃗
= q ⃗v xB
⃗ is up, so q ⃗v xB
⃗v xB ⃗ is down.
EH is the Hall field induced by carrier accumulation on the edges of the sample.
Force in the y-direction must be balanced (= 0).
Fy = 0 = q (EH + ⊑D · B‡ )
|EH | = |vD | · |Bz |
|EH | = µ|E§ | · |B‡ |
EH 1
µ = ·
E§ |Bz |
16
No worry about n. Then
σ
n =
eµ
EH = µE§ B‡
σ
= E§ B‡
ne
Jx Bz
=
ne
where Jx = σE§
EH 1
RH = =
Jx Bz ne
More to come.
As you probably have experienced personally, metals that are good conductors of electricity are also very good
conductors of heat. Copper and aluminum are used to coat the bottom of high quality cookware, with the purpose
of spreading the heat uniformly across the cooking surface. However, not all metals are created equally. The
handle of a cast-iron skillet is also metallic, but can be handled without gloves on the campfire. These properties
are closely related to the electrical conductivity of the material.
In most regimes, heat flow is described by Fourier’s law where
dT
Γ = −κ∇T = −κ
dx
where Γ is the heat flux in units of W/m2 or W/cm2 , ∇T is the gradient of the temperature through a material,
and κ is called the thermal conductivity. Units of thermal conductivity are W m−1 K−1 (in SI units) or W/cm-K.
The SI unit is increasingly common in use today, though older literature is typically W/cm-K. The minus sign in
the equation just denotes that heat flows down the temperature gradient from areas of high temperature toward
the lower temperature. Typical values are given in table 1.8.
Like Ohm’s Law, the expression is not difficult to justify physically. We intuitively expect that the heat transported
is proportional to the temperature difference, and we simply assign a proportionality constant called the thermal
conductivity. Like electrical conductivity, thermal conductivity can vary over orders of magnitude.
Two distinct processes contribute to thermal transport in materials. The first, active in all materials, is the
lattice contribution. Atoms on the “hot” side vibrate with more energy than those on the cold side. But these
vibrations are coupled and the vibrational energy is transmitted from hot atoms to colder atoms though phonons.
The ability of these phonons to carry heat can be described in many of the same terms as electron transport,
but we will leave that for another time. The very high thermal conductivity of diamond is linked to this phonon
transport.
The second mechanism is heat transport through the electron systems. Again, electrons at the hot end have a
higher thermal energy as those in the cold region. But through their random thermal velocity, these electrons
move from the hot area to the cold transporting the energy with them. While the absolute thermal velocity of
electrons in a metal is approximately independent of temperature at ≈5 eV, the electrons do have an additional
17
1.8 Thermal conductivity and resistivity of various metal elements
and selected alloys at 25 o C showing the Wiedemann-Franz rela-
tionship. Data taken from the CRC.
18
1000
8
6
5
thermal conductivity [Wm K ]
Ag
-1
4 Cu
3 Au
-1
Al
2
W
Mo Na
Mg Figure 1: Scatter plot of the thermal
Yellow Brass Zn conductivity versus the electrical resistiv-
100 Cr Ni
8 Bronze Fe In ity for various elemental metals and se-
6 Sn Pt lected alloys at 25 o C. The solid line is
5 1020 Steel the Wiedemann-Franz law κ = CWFL σT .
4
Pb The relationship is valid in regimes where
3
U the thermal conductivity is dominated by
2 Ti electron transport with minimal phonon
304 Stainless transport.
10
8 Hg
6
2 3 4 5 6 8 2 3 4 5 6 8
.01 0.1 1
-1 -1
conductivity [µΩ cm ]
energy proportional to kB T . Scattering collisions between the electrons and the lattice result in extremely rapid
equilibration between the lattice and electron temperatures (on times scales of picoseconds).
Since both electrical conductivity and thermal conductivity depend on the random motion of the electrons and
the scattering time τ , it is no surprise that these two properties are directly proportional. This is manifested as
the Wiedemann-Franz Law (or Wiedemann-Franz-Lorentz Law) where
κ π 2 kB
2
≈ = L0 = 2.44 × 10−8 WΩK−2
σT 3e2
where L0 (or CWFL ) is known as the Lorentz constant. The original observation in 1853 by Wiedemann and Franz
was the proportionality relationship between the thermal and electrical conductivity. Lorentz in 1872 added the
critical temperature dependence and determined the numerical constant.
The Wiedemann-Franz law has proved to be extremely robust since its initial discovery in 1883. The actual
proportionality constant varies with temperature and material system, but never by more than 50%. However,
in 1996 C.L. Kane and Matthew Fisher predicted theoretically that the limit could be bypassed if transport were
limited to one-dimensional chains in a 3D lattice. Coupling of spin and charge gives rise to two types of transport
with charge flow blocked by impurities but heat transport allowed. This was experimentally verified by Nigel
Hussey in “purple bronze” where electrons indeed travel along atomic chains in the lattice. Heat is conducted
about 105 times faster than would be expected.7 This gives rise to materials with excellent thermal conduction
but minimal electrical conduction.
For thermoelectric research, the holy grail is the inverse. The ideal material would conduct electricity readily while
blocking thermal transport. While there are numerous research concepts to achieve this material by nanoscale
engineering, no real breakthroughs have been reported.
7 Gross violation of the Wiedemann-Franz law in a quasi-one-dimensional conductor, Nicholas Wakeham, Alimamy F. Ban-
gura, Xiaofeng Xu, Jean-Francois Mercure, Martha Greenblatt and Nigel E. Hussey, Nature Communications 2, 396 (2011).
doi:10.1038/ncomms1406
19
1.9 Summary Equations
Thermal velocity: Average random motion velocity of electrons in a material. For metals,
τ = 2 × 10−14 s = 20 fs
Mean Free Path: The average distance an electron moves in any given direction before undergoing some sort
of scattering event. The mean free path is given as
ℓ = vth τ
⃗vd = µE⃗
Mobility: The proportionality factor relating the drift velocity and the applied electric field
⃗vd = µ E⃗
Conductivity: Proportionality constant in the intrinsic form of Ohm’s Law. It is the product of the number
of charge carriers flowing, the amount of charge each carriers, and the mobility (velocity).
σ = neµ
J⃗ = σ E⃗
Residual resistivity: The resistivity of a material as the temperature approaches 0 K. As lattice scattering
is eliminated, the residual resistivity is a measure of all other scattering mechanisms and can be used to
measure the purity and quality of a metal.
Matthiessen’s Rule: As the scattering times for multiple processes add in inverse, the resistivity of a material
is equal to the sum of the contributing factors.
1 1 1 1 1
= + + + + ...
τ τlattice τimpurities τgrain boundaries τsurfaces
ρ = ρlattice + ρimpurities + ρgrain boundaries + ρsurfaces + . . .
20
Nordheim’s Rule: At low concentrations, the contribution of impurities in a metal to the resistivity is pro-
portional to the concentration of those impurities. Given Xi the molar (atomic) fraction of the impurities,
in the dilute limit
ρimpurities ≈ Ci Xi
where Ci is Nordheim’s constant. In a binary alloy where the two components have comparable electron
densities, the contribution is approximately
ρalloy ≈ Ci Xi (1 − Xi )
Grain Boundary and Surface Scattering: As either grain size or sample dimensions approach the electron
mean free path, the resistivity of a material will increase. The resistivity of a fine-grain polycrystalline
material is approximately
ρ 4 l R
≈ 1+
ρ0 3 d 1−R
where ρ0 is the resistivity of the pure material, d is the average grain diameter, l the electron mean free
path, and R is the probability an electron is reflected at the grain boundary. For thin samples, increase in
resistivity due to surface scattering is approximated as
ρ 3 l
≈ 1+ (1 − p)
ρ0 8 t
and has units of inverse temperature (K−1 ). This is also the fractional change in the resistance of a sample
as long as changes in the dimensions of the sample can be neglected. For metals whose conductivity is
dominated by lattice scattering, αT CR is approximately 1/T .
Wiedemann-Franz Law: The electronic contribution to the thermal conductivity of a metal is proportional
to the electron conductivity. This is expressed as the Wiedemann-Franz Law where
κ
≈ L0 = 2.44 × 10−8 WΩK−2
σT
where L0 is also often written as CW F or CW F L .
While exact values can be readily obtained from many sources today (internet), it remains important to have a
“feeling” for the size of physical properties. The table below lists several parameters that most engineers know
off the top of their head for doing “back of the envelope” calculations. Metallic values are based on relatively
good metals and common metal alloys in the large grain, large sample limit.
21
Typical values for metallic-like materials
Property SI units common units Pure metal Non-dilute alloy Ion / polymer
Atomic spacing m nm 0.1 - 0.3 0.1 - 0.3 0.1 - 0.3
Atomic density at/m3 at/cm3 5 × 1022 5 × 1022 5 × 1022
Carrier density e− /m3 e− /cm3 0.5 - 2 × 1023 0.5 - 2 × 1023 1019 - 1023
Resistivity Ω-m µΩ-cm 2 - 10 20 - 1000 0.1 - 1000 Ω-cm
Conductivity S/m mhos/cm or S/cm 0.1 - 0.5 × 106 1 - 50 × 103 10 − 3 -10
Mobility m2 /V-s cm2 /V-s 20 - 40 0.1 - 5 10−10 - 0.1
Scattering time s fs 10 - 20 0.1 - 5 < 0.1
Mean free path m nm 5 - 30 0.5 - 5 <0.5
Thermal conductivity W/m-K W/m-K 10 - 400 1 - 10 10−3 - 0.1
W/cm-K 0.1 - 4 0.01 - 0.1 10−5 - 0.001
TCR K−1 K−1 0.003 0 - 0.001 .005 - 0.5
22