F E A T U R E A R T I C L E

Conduction and Polarization

Mechanisms and Trends in Dielectrics
Key Words: Conduction (ionic and dipolar), polarization (electronic, atomic, dipolar,
interfacial, dielectric constant, tangent δ, loss factor, model circuits, relaxation, Cole-Cole
plots

Introduction T.W. Dakin

T his article discusses the theoretically deduced mechanisms
for the magnitude and the trends of conduction and polar-
ization in electrical insulation (dielectrics) with molecular and
Formerly with Westinghouse Research Labs,
Pittsburgh, PA
physical structure, frequency, and temperature. The discussion
is intended to explain the atomic, electronic, molecular, and
ionic basis for these electrical properties, so that the reader can This tutorial article explains how
better understand why dielectrics behave the way they do.
App1ication of these principles should also guide the reader to dielectric data can be helpful in
make estimates or rough predictions of conductivity and dielec- interpreting molecular behavior and
tric constant levels, and their variations, from an examination
of the molecular/atomic composition of a material. structure.
Most of the discussion is more directly applicable to the con-
densed phases, solids and liquids. Gases, in principle, behave
the same, but they have important differences because of their
very much lower density and lack of restraint to molecular rota-
tion and translation. An overriding factor in the magnitude of perature and the AC conductivity with frequency; and the varia-
conduction and polarization is the number density of the par- tion is usually greater if the material has either ionic conductiv-
ticipating species: electrons, ions, or dipolar molecules. ity or dipolar molecules. Figure 1 shows the change in AC con-
Since electrical insulation is used primarily for its noncon- ductivity over 12 decades of frequency for insulation materials
ducting characteristic, the relatively low level of conductivity it for various levels of the loss factor, ε′ tan δ. The selected levels
does have can be an important limitation to its effective use. of ε′ tan δ, from 10-5 to 10, encompass almost the entire ob-
Therefore, it is very important to know what this level is and the served levels of this parameter. When tan δ exceeds 1, the mate-
factors that influence it. The conductivity or the AC loss tan- rial is more like a conductor than an insulator! Figure 1 empha-
gent, tan δ, or loss factor, ε′ tan δ, (ε′ is the relative dielectric sizes the general continuous increase of conductivity with fre-
constant) are commonly used as quality control parameters for quency; on this increase are superimposed variations associated
electrical insulation. Table 1 compares the DC resistivity·levels with dipolar and ionic-interfacial conduction and polarization,
of various types of materials. It is obvious from an examination which will be discussed later.
of Table 1 that there are more than 20 decades of difference in The metals and semiconductors are primarily electron con-
resistivity between the best metal conductors and the best insu- ductors with structures which permit extensive movement of
lating materials. The resistivity of various metals alone spreads electrons in conduction bands. “Hole” or P (positive) type semi-
over about three decade levels. Aqueous electrolyte solutions conductors also physically conduct by electrons moving from
range over about six decades of resistivity, and various insulat- hole to hole. Since a “hole” is a positive site where an electron
ing materials range over about six or more decades of resistiv- is missing, it seems as if positive charges were moving in the
ity, at levels above those of the aqueous electrolyte solutions. opposite direction. In contrast, except at very high electric fields
It is very important to recognize that the DC and AC conduc- or under ionizing radiation conditions, there are very few elec-
tivity of insulating materials may vary considerably with tem- trons in common insulating materials, which are not strongly

September-October 2006 — Vol. 22, No. 5 11

 Table 1. Comparative Resistivities.
Material Resistivity (ohm-m)
Copper 1.72 × 10 –8
Iron 10 –7
Nichrome alloy 10 –6
Graphite, pure 1.4 × 10 –5
10% NaCl in H 2 0 8.3 × 10 –2
10% acetic acid in H 2 0 6.5
Silicon, pure 10 3
Distilled H 2 O 10 4 – 10 5
1 Paper-phenolic laminate, XX grade 6 × 10 5
1 Paper-phenolic laminate, xxp grade 2 × 10 8
1 Glass-epoxy laminate, G10 grade 2 × 10 9
Polyvinyl chloride 10 9 – 10 12
Mica, Muscovite 10 14 – 10 15
Silica, pure 10 16
Polyethylene 10 16
1 Humidified 96 hours at 35°C and 90% R.H.

Figure 1. AC conductivity vs. frequency for varied levels of ε′

tan δ.
attached to an atom or molecule forming ions. Consequently
conductivity of dielectrics is usually due to the translation of
positive and negative charged ions of atomic or molecular size.
Short time DC or continuous AC conduction also occurs in di- tions. A special case is the “trapping” of electrons in vacant
electrics by the rotation of dipolar molecules or parts of mol- negative ion sites; these are known as “F” centers.
ecules. Dipolar molecules are asymmetric molecules having a In other cases, binding between atoms in molecules occurs
permanent separation of positive and negative charge internally; by sharing electrons between atoms forming some common elec-
they will be discussed in more detail later. tron orbits. This type of bonding is predominant in organic car-
bon compounds. Electrons in organic materials are usually at-
Electrical Conduction tached to particular molecules forming negative ions with an
The electron conductivity of solids is determined by their equal charge of positive ions so they do not contribute to elec-
atomic, crystal, or amorphous structure. Metals are good elec- tron conduction.
tron conductors. Their atoms all have in their outer shell one to Interatomic bonding is a mixture of complete sharing (cova-
three electrons that are free to move throughout the metals and lent) and partial or complete transfer or shift of electrons from
alloys in “conduction bands”, since the electrons in this state one of the atoms to the other which is usually the one 1acking
are not associated with any particular atom, pair or group of one to three electrons to fill its outer shell. The subject of chemi-
atoms. cal bonding is treated very thoroughly in a classic text by Linus
If the atoms of these same elements are combined into mol- Pauling [1] and in other texts on physical chemistry [2], [3].
ecules with different types of atoms which are lacking one to Electrons restricted to individual atoms or interatomic bonds
three electrons in their outer shell—such as oxygen, sulfur or are in the “band theory” representation of conduction said to be
the halogens (fluorine, chlorine, bromine, and iodine)—the elec- in the “valence band”. Sufficient thermal energy or ionizing ra-
trons from the outer shell of the metallic elements tend to trans- diation can lift some into the conduction band, but the required
fer completely to the outer shell of the combining elements, energy is high in the case of most insulating materials. There-
filling that shell and thereby forming compound molecules that fore, electron conduction is not common (except at very high
are ion pairs (for example: Na+C1– and Mg+2O+2). These com- electric fields) in conventional insulating materia1s.
pounds form solid ionic crystals where the electrons are tightly The energy gap between the valence band and the conduc-
bound in negative ions balanced by an equal total charge of tion band in metals is negligible, and the valence and conduc-
positive ions. These are not usually electron conductors, except tion bands are said to overlap. In intrinsic semiconductors, such
at very high electric fields or under ionizing radiation condi- as pure silicon and pure germanium, the conduction-valence

12 IEEE Electrical Insulation Magazine

band gaps are 1.1 to 1.2 and 0.72 to 0.78 eV (electron volts),
respectively; and these pure crystals are not very good electron Table 2. Electron Conduction Band Gaps. From Reference [26]
conductors compared to metals, as can be seen in Table 1. They
Compound Electron Volts
are much better conductors, however, than compounds where
covalent (shared electron) interatomic bonding occurs. A more GeB 0.01
thorough discussion of semiconductor conduction is given in GeP 0.013
[4]. In diamond, which is a fair insulator, the atoms are pre- SiP 0.044
dominantly bonded by electron sharing in individual carbon-
carbon atom bonds much like the carbon-carbon (single) bonds SiB 0.057
in most common organic compounds, the conduction-valence Sn (µ, grey) 0.08
band gap is 5.3 to 6 eV. The band gap is usually greater than 10 Ge, pure 0.66
eV in molecular compounds, so only a very-very few electrons
Si, pure 1.09
occur in the conduction band in such materials, and they are at
least potential insulators. In contrast, the electrons in the crystal P, (red) 1.5
plane of graphite have overlapping (pielectron) orbits permit- P (yellow) 2.7
ting the electrons to move relatively easily throughout the plane;
BN 4.6
therefore, graphite has a relatively high conductivity. This bond-
ing structure is similar to that which occurs within the molecule NaI ~7
of benzene and anthracene. Discussion of electron conduction KI ~7
in this type of organic compounds is presented later. NaCl ~10
The concentration of electrons thermally excited into the
conduction band is proportional to: KC1 ~10

σe, conduction band ~exp (Egap/kT), (1)

where kT is 0.02518 eV at T=298 K. Table 2 lists a range of

conduction band gap values for different materials.
Intrinsic semiconductors, such as pure silicon and pure ger- effects. Also, electron semiconduction has been noted with in-
manium, are given enhanced electron conductivity by “doping” organic molecules with extended chains of alternating double
with phosphorous, arsenic, or antimony to give “n” (negative) and single carbon-carbon bonds. As in benzene, naphthalene,
conductivity or with boron, aluminum, or gallium to give “p” and anthracene, the pi orbit electrons associated with alternat-
(positive) conductivity. The n type dopants lie in the column ing single and double bonds are not associated with any par-
just above Si and Ge in the periodic table of elements and have ticular atom and are able to move around the ring or along mo-
one more electron in their outer shell than Si and Ge. This addi- lecular chains having “conjugated” double bonds. The energy
tional electron lies in an energy level just below (about 0.01 barrier between such internally semiconducting molecules is
eV) and close to the conduction band in the doped Si or Ge, so high, however, so electrons do not easily move from molecule
it increases very significantly (depending on the concentration to molecule, unless some doping agent is added to reduce this
of dopant) the concentration of electrons in the conduction band. barrier.
In contrast, p type dopants lie in the column just below Si There are numerous texts and reference works on solid state
and Ge in the periodic table, having one less electron in their semiconductor physics that deal with electron conduction in
outer shell than Si and Ge, thereby creating electron energy sites solids. References that can be consulted for more details are
close above the valence band of Si and Ge in the doped mate- [6], [7].
rial. A significant number of valence electrons in the doped Si
or Ge are thermally excited to become attached to the doping Ionic Conduction
atoms, thereby creating vacant electron sites, positive “holes” The phenomena of ionic conduction in dielectrics can per-
which can move about almost as readily as electrons in a con- haps be understood most easily by discussing the similarity to
duction band of these elements. (When an electron in the va- conduction in solutions of electrolytes, the most common of
lence band moves into an adjacent “hole”, it is equivalent to a which are water solutions. The difference between ionic con-
positive charge moving in the opposite direction.) duction in amorphous dielectrics compared to water is more a
Similar types of electron conduction processes to those dis- matter of degree than of mechanism. The concentration of unat-
cussed above for Si and Ge have not been commonly recog- tached ions is typically much greater in water, and their mobil-
nized to occur in conventional insulating materials, where the ity, µ+ and µ– is greater in water, whose viscosity (≅ 1 mPa·s (1
electrons are strongly attached to atoms or molecules forming centipoise) at 20°C) is relatively low compared to common in-
ions or are localized in interatomic bands. In some unusual or- sulating fluids (≅ 70 mPa·s for transformer oil at 20°C).
ganic chemical structures, electron conduction does occur. This Ionic conductivity can be expressed by (2), where n+ and n–
subject is reviewed in [5]. Experiments with polyethylene doped are the concentrations of the positive and negative ions, respec-
with iodine (not combined) have indicated electron conduction tively, per cubic meter, z+ and z– are the number of unit charges

September-October 2006 — Vol. 22, No. 5 13

 on the respective ions and e is the electron charge in coulombs. difficult to determine. Nevertheless, the principles outlined here
The mobilities have units of m/s/(V/m): are applicable and help us to understand the variations in con-
ductivity observed in insulating fluids as well as resins, due to
σ = (Σµ+n+z+ + Σµ–n–z–)e ohms–1m–1. (2) often unknown impurities.
It should be particularly noted that the dissociation constant
The mobilities of many ions in water solutions are available is strongly influenced by the dielectric constant of the solvent
from the extensive literature [2], [3] on electrolyte solutions (which could be the insulating fluid, resin, glass, etc.) and also
through the relation, (3), between the ionic equivalent conduc- the ion size. This is indicated by (5a), which shows the thermo-
tivities, Λeq and their mobilities. dynamic relation of the dissociation equilibrium constant to the
Gibbs Free Energy, ∆G. This free energy is also proportional to
µ = Λeq/Noe m2/V s, (3) the coulomb energy of attraction of the dissociated ions, as in-
dicated in:
where No is Avagadro’s number and Λeq is the equivalent ion
conductivity. Values of Λeq for individual ions, given in tables in ∆G = RTln(Kd) = Noq+q–/4πrεoε′, (5a)
many physical and electrochemical texts, see [2], [3] and in Kd = exp(∆G/RT) = exp(Noq+q–/4πrεoε′RT), (5b)
handbooks, are defined as the conductance (l/ohms) of a solu-
tion containing one equivalent of the ion per liter with a gradi- where q+ and q– are the coulomb values of the positive and nega-
ent of 1 V/cm. An equivalent is a molecular weight of the ion tive ion charges, r is the distance between their centers, ε’ is the
(in grams) divided by the ion unit charge. (For K+ this would be relative dielectric constant, and εo is the absolute dielectric con-
one, for Mg++ this would be two, etc.). Individual ions are usu- stant of vacuum. Equation (5b) indicates the exponential rela-
ally given for infinite dilution in water, since at higher concen- tion of the dissociation constant to the reciprocal dielectric con-
trations there is inter-ionic attraction of oppositely charged ions, stant of the medium and the Kelvin temperature. This very sig-
which results in a significantly less than proportional increase nificant relation indicates why water (ε′ ≈ 80) is such a good
in conductivity with increasing concentration. These equivalent ionic conductor and why transformer oil (ε′ = 2.2) is a good
ion conductivities are approximately additive, for example, the insulator.
molar conductivity of KC1 is the sum of the K and the Cl equiva- Figure 2 gives experimental data from several sources [9] for
lent ion conductivities. the dissociation constant, Kd, as a function of the solvent di-
The ionic conductivity of water solutions is much better docu- electric constant. The data are plotted according to (5) as log(Kd)
mented quantitatively than is possible for insulating liquids and vs. 1/ε′, and the other quantities such as T and R are assumed
resins, since the ionic concentrations and mobilities in water constant for the respective compounds. Figure 2 indicates a rough
can be determined more exactly. In very dilute water solution agreement with (5a) and (5b) over about 15 decades of the dis-
compounds like KC1 are completely dissociated into K+ and sociation constant. It also indicates the very important influ-
C1–, but compounds 1ike acetic acid have a small degree of ence of the relative dielectric constant on the conductivity of
dissociation, which is quantitatively expressed by a dissocia- insulating fluids, indicating why it is increasingly difficult to
tion (equi1ibrium) relation given in (4a) and (4b): achieve low conductivity with higher dielectric constant fluids.
Although insulation manufacturers try to avoid such com-
C+ C –/C1 = Kd , (4a) pounds as those salts and acids indicated in Figure 1, they or
α2Co/(1 – α) = Kd. (4b) similar materials may occur as chance impurities, which are
difficult to avoid, and are often initially removed by treating
In these equations, Kd is a dissociation constant relating the insulating fluids with adsorbent clay, etc.
concentration of the ions, C+ and C– to the concentration of the Figure 2 and (5a) and (5b) also indicate another basic point
undissociated neutral molecule, C1, α is the degree of dissocia- relating to the ionic conductivity of insulating fluids—the ef-
tion, = C+/Co or C–/Co, where Co is the concentration of the fect of ion size on the dissociation. The tetrabutyl ammonium
compound in the fluid. Equations (4a) and (4b) are correct for and tetraisoamyl ammonium positive ions are relatively large
dissociation to a single positive and a single negative ion. They ions, with relatively larger values of closest approach (sum of
become a little more involved for dissociation to combinations the ionic radii) to the negative nitrate ion. This is reflected in a
of singly and multiply charged ions. A reference book such as larger dissociation constant, as would also be expected accord-
that by MacInnes [8] on electrochemistry should be consulted ing to (5a) and (5b). Very small ions like H+ and Cl– tend to
for more details. A limited number of organic compounds dis- become attached to polar molecules, which gives them an ef-
sociate to ions; the more common ones are carboxylic acids fective larger radius. This may explain why the dissociation
(such as acetic acid) and salts formed from amines and various constant for HC1 is higher than might naively be expected.
mineral and organic acids. Gemant [10] has shown experimentally the importance of ion
In insulating fluids such as transformer oil, capacitor fluids, size on the ionic conductivity of hydrocarbons. Large complex
etc., the concentration of ions usually results from chance im- ions contribute importantly to the conductivity of hydrocarbon
purities and is typically orders of magnitude less than in water. transformer oil.
Furthermore, the particular type of compound and its ions, which Another important factor influencing ionic conductivity is
are responsible for the slight conductivity of insulating fluids is the ion mobility, as indicated by (2). Ion mobility in amorphous

14 IEEE Electrical Insulation Magazine

 The increase of ionic conductivity of insulating materials with
temperature is a result of two superimposed effects: increase in
fluidity (inverse viscosity) and increase in dissociation to ions
with increasing temperature. The correlation of the trend of tan
δ of an insulating fluid (54% chlorinated diphenyl) with tem-
perature to the trend of the fluidity is illustrated in Figure 3.
The conductivity of a dielectric is proportional to its tan δ. The
ions in the insulating fluid of Figure 3 are impurities dissolved
in the liquid and are difficult to identify, as is the case in most
practical insulating materials. The trend toward higher conduc-
tivity in lower viscosity insulating fluids, as well as more flex-
ible resins with typical impurities, is well-known.
Before leaving the subject of ionic conduction, the case of
ionic crystals should be mentioned. For example, in the case of
crystalline salt, NaC1, which is a cubic crystal, all the atoms
are ions. But in a perfect crystal they cannot move at all easily,
being confined by the surrounding ions to their particular lat-
tice sites, and there is insufficient space to move between adja-
cent ions in the perfect crystal. Therefore dry crystalline NaCl
is a very high resistance insulation. If this or any crystal is im-

Figure 2. Variation in dissociation constant with solvent

dielectric constant.

media like liquids, amorphous resins, and glasses is roughly

inversely proportional to the effective internal viscosity of the
medium. This explains why (other factors being equal) the con-
ductivity of insulating fluids is usually lower for higher viscos-
ity fluids and still lower for resins. Mathematically, the relation
is indicated by Stokes law given in (6):

v = F/6πηr = zeE/6πεr, (6)

where η is the viscosity, r is the radius of a sphere moving with

a velocity v under a force, which is that of the electric field,
zeE. Since µ equals the ratio of the velocity to the electric force,
the mobility is given by (7):

µ = v/zeE = l/6πηr. (7)

This equation is a somewhat oversimplification of ion trans-

port, since molecules are not perfect spheres, and there may be
specific attractive forces between ions and the medium, which
consists of irregular molecules instead of a continuum as as-
sumed by the Stokes equation. However, it does indicate an
important trend that was recognized many years ago by electro-
lyte chemists, who know it as Walden’s rule, which states that
the ionic conductivity of solutions varies inversely as the fluid Figure 3. Fluidity and dissipation factor of viscous
viscosity. (see McInnes [8], p. 360). chlorinated insulating fluid.

September-October 2006 — Vol. 22, No. 5 15

 perfect however, this provides paths for conduction along inter static units for this quantity because of historic precedent, and 1
crystalline boundaries and through vacant lattice points left by × 10–18 statcoulomb cm is a Debye unit named after the cel-
too rapid crystal growth or by foreign ions such as Ca++ or Mg++, ebrated scientist who developed the theory of dipole molecular
which, by their double charge, create vacant lattice points in behavior [11]. One Debye equals 3.3356 × 10–30 coulomb meters.
which the singly charged Na+ ions would be. The number of To put this into a more comprehensible perspective, one Debye
positive ion charges must remain equal to the negative ion is equivalent to one electron negative charge separated from an
charges for the crystal to be neutral. Conduction in ionic crys- equal positive charge by 0.208 Angstroms. Since bond distances
tals and glasses occurs more by a series of discrete ion hopping in organic molecules are of the order of a few Angstroms and
steps from one lattice site to an adjacent empty site, rather than since bond dipole moments in organic molecules range from
by a diffusion flow as might be visualized in the case of liquids. 0.4 for carbon-hydrogen to about 4 for carbon-nitrile compounds,
In both cases, however, the ion conduction, through the mobil- with the group: –C–N, it must be concluded that the equivalent
ity factor, is governed by an activation mechanism so that the shift of the electron charge in polar bonds toward the atom of
variation with temperature is given by an equation similar to greater electron affinity is only a minor fraction of the inter-
(8): atomic bond length to establish the dipole moment in most or-
ganic molecules. However, in ionic compounds like NaC1, the
µ ~ exp(–∆F/RT), (8) shift of an electron from Na to C1 approaches the interatomic
distance in the molecule.
where ∆F is the activation energy for ion motion or hopping. In an electric field, E, as is indicated by Figure 4, there is a
Correlation with this relation has already been indicated in Fig- torque force on the dipole molecules. The torque on the dipolar
ure 3 for the tan δ (or conductivity) of a liquid. Glasses, ceram- molecules tending to turn them toward the electric field direc-
ics, and resins show similar correlation. Whenever there are in- tion is: Eqe sin(θ),where θ is the angle of the dipole axis to the
ternal boundaries or potential barriers, such as at the insulator/ electric field. This causes them to rotate. The rotation produces
conductor interface (electrodes), ionic conduction is interrupted, a small but significant current for a short duration, the time de-
and the conductivity varies with time with DC voltage, and with pending on the resistance to rotation presented by the medium
frequency with AC voltage. Such boundaries or barriers also in which the dipoles occur (the surrounding molecules). Since
lead to polarization which affects the apparent dielectric con- this charge movement is limited in extent to less than a molecu-
stant. Since the polarization, and conduction are so intercon- lar diameter, it also produces a polarization and is related to the
nected, AC conduction variations with time and frequency are dipole contribution to the dielectric constant, the quantitative
treated in the discussion on polarization. analysis of the AC dipole conductivity is combined in a suc-
Particular mention should be made here of the consequences ceeding section on dipolar polarization.
of ionic conduction with DC voltages over long periods of time.
Electrolysis can occur at the electrodes, producing gas (such as Polarization Mechanisms
H2 and O2) or other possibly deleterious compounds. Unless Polarization is responsible for the refractive index and di-
these can diffuse away, they can be harmful to the dielectric electric constant of materials. If there is no polarization, the
strength or insulation function. refractive index and dielectric constant are unity. This occurs
only with a vacuum. As is illustrated in Figure 4, there are four
Dipolar Conduction principle types of polarization: electron, ionic-interfacial,
Molecular dipoles are more commonly discussed in their re- atomic-ionic displacement, and dipolar. They will be discussed
lation to polarization and dielectric constant, but they also con- in that order.
tribute to conduction in insulation for short times and over lower As is illustrated in Figure 5, electron polarization persists
frequencies of AC voltage. Therefore, it is logical to discuss over the entire frequency range from DC up through optical
dipoles first along with conduction. frequency.
Permanent molecular dipoles occur to some extent in a vast Atomic-ionic displacement polarization extends from DC up
majority of molecules between different elements, since the at- through infrared frequencies. This type of polarization is the
oms of one element usually have a different electron affinity predominant form of polarization in inorganic crystals, glasses,
than those of another element; therefore, in the combined state and ceramics. It is the principle contributing mechanism to their
electrons are permanently shifted partially or completely from dielectric constant at a uniform level up to infrared frequencies.
one atom to the other. The extreme case of this has already been A special form of this polarization, namely “ferroelectric” (where
mentioned in the case of ionic compounds such as NaCl. polarization occurs collectively in domains), results in very high
Dipolar molecules occur as a major or minor constituent or effective dielectric constants, in analogy to ferromagnetic po-
impurity in both insulating liquids and resins. An electric di- larization. The high dielectric constant titanate ceramics are
pole consists of equal positive and negative charges separated examples of this. Ferroelectric polarization is limited to certain
at some distance. Its value, m, is the product of the charge, e, types of crystals and ceramics and to temperatures below their
and the spacing, s, so that m = e s. Table 3 gives values of a “Curie” temperature.
series of molecular dipole moments in units of statcoulomb cm. Dipolar polarization occurs from DC up to microwave fre-
The dielectric literature and even current texts use the electro- quencies, depending on the presence of dipolar molecules and

16 IEEE Electrical Insulation Magazine

 Table 3. Table of Dipole Moments and Dielectric Constants at 20ºC.

Compound Dipole Molecular. Density ηD η 2D ε′

Moment Weight g/cm 3
Hexane 0 86.18 0.659 1.375 1.89 1.89
Benzene 0 78.12 0.877 1.501 2.25 2.28
Tetrachlorethylene 0 165.8 1.62 1.505 2.265 2.28
1,4 Dichlor-benzene 0 147 1.244 1.528 2.33
Transformer oil 1 0.88 1.43 2.05 2.28
Toluene 0.36 92.15 0.865 1.496 2.24 2.3
Styrene 0.37 104.15 0.904 1.547 2.39 2.4
Diethyl Ether 1.2 74.12 1.35 1.831 1.831 4.33
Aniline (Amino-benzene) 1.53 93.13 1.02 1.582 2.516 6.89
Chloro-benzene 1.7 112.6 1.104 1.524 2.32 5.68
1,1, 1 Trichlor-ethane 1.78 133.4 1.336 1.438 2.067 5.66
Chlorinated (42%) Diphenyl 2 1.40 1.62 2.65 5.8
N Propyl Alcohol 1.66 60.11 0.801 1.385 1.918 19.3
Water 1.84 18.01 0.998 1.333 1.777 78.2*
Ethylene Glycol 2.25 62.07 1.086 1.432 2.05 41*
Acetone 2.85 58.08 0.778 1.359 1.847 20.9
Benzo Nitrile 2.8 103.13 0.99 1.529 2.337 25.2
Nitro Benzene 4.1 123.11 1.202 1.556 2.421 34.9*
* = Measured at 100 kHz.
1 = Mixture of mostly nonpolar hydrocarbons
2 = Formerly used power capacitor impregnant (a mixture of chlorinated diphenyls)

the resistance to molecular rotation presented by the material’s larly displaced or oriented charges internally in the material.
internal structure. These charges are associated with all the atoms and molecules
Interfacial polarization, involving a longer range ion move- and any grosser internal barrier features of some insulation. This
ment, is observed usually only at lower frequencies. section dea1s with the electrons within the atoms. When an atom
Ionic-interfacial and dipolar polarization are the most im- or molecule is placed in an electric field, the electrons in the
portant for most organic liquid and resin electrical insulation, atom are displaced very slightly toward the positive direction of
since they are associated with conduction and dielectric loss at the field. This occurs almost instantly within times less than a
commonly used frequencies. A form of polarization and associ- half cycle of optical frequency.
ated conduction like ionic-interfacial is also observed in inor- Electron polarization is responsible for the optical refractive
ganic glasses and ceramics at lower frequencies and increases index, η, and is part of the dielectric constant of all dielectric
with increased temperature. materials. This is illustrated in Figure 5. The magnitude of the
Each of these types of polarization will now be discussed in refractive index and its contribution to the dielectric constant,
more detail. The magnitude of each type of polarization de- ε′∞ = η2, is essentially constant from zero frequency up to the
pends primarily on the number density of the participating spe- optical frequency range where dispersion (a slight increase) and
cies and the resistance to motion presented by the medium in resonance absorption, attenuation, occurs at specific frequen-
the case of ionic-interfacial and dipolar types. cies. This type of polarization does not contribute to conduction
or dielectric loss in conventional electrical insulating materia1s
Electron Polarization at frequencies common1y used in insulation app1ications.
Polarization has the net effect of placing opposite charges in Table 3 gives values of ε′ at 60 Hz and ε′∞ = η2 (η values
a material at the face adjacent to the electrodes of a capacitor. reported for the Sodium d line optical frequency in handbooks)
These apparent charges are the cumulative effect of many simi- for various types of dielectric materials. The values of η for

September-October 2006 — Vol. 22, No. 5 17

 Figure 5. Frequency trend of polarization effects.

like heptane and benzene. Since ionic-interfacial and dipolar

polarizations are associated with conduction, a low level of these
polarizations also indicates a low conduction level of the di-
electric. A low dielectric constant also indicates a minimal ten-
dency to dissolution or dissociation of impurities. It should be
cautioned in drawing conclusions from differences between the
dielectric constant and the refractive index squared of resins
and viscous liquids, since the dielectric constant of these mate-
rials may be quite temperature dependent. As will be discussed
in more detail later, increased temperatures are likely to increase
the lower frequency dielectric constant quite appreciably, if the
material has polar constituents, which are likely to be more
mobile at higher temperature. The refractive index, however,
changes very little, decreasing as the material density decreases
with increasing temperature.
Although the overall range for the refractive index is not large,
the values can be measured so precisely that they are often used
Figure 4. Principle types of polarization. for identification and analysis. For example, the percentage of
aromatic (benzene like) hydrocarbon in mineral transformer oil
is indicated by its refractive index, together with its density and
viscosity (ASTM D2140).
dielectric liquids range from about 1.2 for fluorocarbon liquids The relative polarizability of individual molecu1es in a di-
up to about 1.6 for chlorinated hydrocarbon liquids. η increases electric is related to the refractive index by (9):
with the electron (and atomic) density and with aromatic (ben-
zene like) molecular structures, which have a greater electron
polarizability. For example, η for benzene is 1.498 compared η2 – 1 ε∞ – 1 ρNoαo
= = , (9)
with 1.372 for hexane. Organic resins have values from about η2 + 2 ε∞ + 2 M3εo
1.3 for fluorocarbon resins to about 1.6 for those high in oxy-
gen, chlorine, or aromatic content. Inorganic glasses have η where M is the molecu1ar weight, ρ is the density, and No is
values typically in the range from about 1.458 (for fused quartz) Avogadro’s number (number of molecules/gm molecular
to 1.89 for the heaviest lead glass. weight). (Note that ρNo/M is the number of molecules per cu-
The difference between the η2 value and the dielectric con- bic centimeter, a factor which appears in several other equa-
stant at 60 Hz for the same materia1 indicates the contributions tions relating molecular parameters to material parameters in
of ionic-interfacial, dipolar, and atomic-ionic polarizations in this article.) Equation (9) is derived assuming that the internal,
this material. This can be a useful guide in initial material se- local, electric field, EL, seen by individual molecules is that
lection. For example, it can be seen in Table 3 that there is very proposed in [12] (EL = (ε′ + 2) E/3, where E is the external
little difference between η2 and ε′ for the hydrocarbon liquids applied field. A rather simple derivation illustrates the logic of

18 IEEE Electrical Insulation Magazine

going from molecular parameters to macroscopic dielectric val- Interfacial polarization was one of the first types of polariza-
ues. It goes as follows: tion to be recognized before 1900, and it is known as Maxwell-
The dielectric macroscopic polarization, P, is: Wagner polarization by the authors of the first papers on the
subject [11]. It is the principle one noted with DC voltage and
P = (ε′ – 1)Eεo. (10) quite low frequency AC, since it is typically the slowest to de-
velop. It may sometimes, however, be confused with slow di-
This should equal the sum of the contribution of all the indi- pole polarization, which is discussed next. The frequency trend
vidual molecules, whose individual molecule electron polariz- of the conductance and polarization due to ionic-interfacial or
ability is αe. The latter, multiplied by the number of molecules/ dipolar effects are similar, but the time scale is much shorter for
cm3 and by the local electric field should equal the macroscopic the dipolar polarization. The conduction leading to polarization
polarizability, as in (11): is most conveniently understood by reference to electric circuit
analogies shown in Figure 6. The simplest analogy is shown in
Figure 6(a) where there is a resistor, RS, in series with a capaci-
ε′ + 2 tor, CS, and a capacitor in parallel with the series combination.
Pe = (Noρ/M)αeEL = (Noρ/M)αe E. (11)
3 The parallel capacitor, CP, represents the high frequency polar-
izability associated with the high-frequency dielectric constant
Rearranging (11) gives the expression for the molar refrac- component, ε∞, which is constant with frequency through the
tion, Rm, which is used by physical chemists in studying mo- time and frequency range where the ionic-interfacial or dipolar
lecular structure [2]:

Rm =
( )( )
M
ρ
η2 – 1
η2 + 2
=
No α e
3εo
. (12)

It has been shown that the molecular refraction is approxi-

mately equal to the sum of the refraction values for individual
interatomic bonds in organic compounds, and the sum of the
values for individual ions in ionic crystals. This relation per-
mits the estimation of the approximate refractive index of ma-
terials where values may not be available. A table of ionic re-
fractions is given in [3] (p. 616).
Finally, regarding the electron polarization contribution to
the dielectric constant, it is noted that this type of polarization
is present in all materials and, in some such as nonpolar organic
liquids and resins, it contributes the most significant part of the
dielectric constant up to optical frequencies. This type of polar-
ization does not contribute to conductivity or dielectric loss in
most dielectrics. An exception could be a dielectric with semi-
conductor inclusions or a material having electronic semicon-
ductor structure as discussed previously.

Ionic-Interfacial Polarization
This type of polarization occurs in dielectrics having in their
composition two, or more, different materials, which are not
intimately dissolved one in the other. Examples of this are mica
in resin, glass fibers in resin, or a solid insulation barrier, such
as paper in a liquid insulation. Materials having a matrix of
amorphous material with embedded crystalline material might
also be expected to fall into the category of heterogeneous sub-
stances, which could have ionic-interfacial polarization. Ceram-
ics fall into this category. Usually the different materials in the
composite will have different electrical (usually ionic) conduc-
tivities, and this characteristic leads to a buildup of charge at
the interface between them when voltage is applied to the mate-
rials in series. One element of such a material is represented by
the equivalent circuit of two series capacitors, each with a resis-
tor in parallel with it, as shown in Figure 6(c). Figure 6. Circuit analogies of electrical insulation.

September-October 2006 — Vol. 22, No. 5 19

 polarization is changing. When the voltage is first applied, it constitutes conduction for short times of the order of an AC
will appear across the resistance, RS (and, of course, across the cycle at lower frequencies. This rotation or orientation produces
parallel capacitor CP), but the voltage division will gradually polarization, whose magnitude can be theoretically related to
change until most of the voltage is across CS. In this analogy, the molecular dipole moment, m. Thus, the contribution of di-
the resistor represents the physical resistance of ions moving up polar polarization to the macroscopic dielectric constant can be
to an interfacial barrier. With this type of polarization, the ap- theoretically related to the molecular dipole moment. The cor-
parent capacitance is higher at lower frequency and declines rectness of this relation is quite good for the gas phase and for
with increasing frequency due to the insufficient time during dilute solutions of polar molecules in nonpolar solvents, where
the half cycle to charge the capacitor. the mutual interaction of the dipoles is minimal, but the relation
There is usually also in most insulations a conductance, rep- is more complicated and less correct for condensed phases of
resented by RP, in parallel with the other components, as shown polar molecules. Nevertheless the modifications of the theory
in Figure 6(b). This conductance is responsible for the DC con- [13], [14] to account for the mutual interaction of dipoles in
ductivity, but it also contributes to the AC conductivity at a level condensed phases has produced some quite good agreement of
which decreases inversely as the frequency. Also included in theory and experiment for dipolar liquids.
the circuit in Figure 6(b) is a second series R-C circuit consist- Similar to the case of electron polarization, there is an indi-
ing of RSD in series with CSD to represent dipolar polarization, vidual molecule polarizability due to its dipole moment, m. This
which is discussed in the next section. Interfacial polarization was first calculated by Professor Peter Debye [12], who was
usually has a much longer RC time constant and occurs some- awarded the Nobel Prize for this and related research. He showed
what independently of the dipolar polarization. To extend the that:
analogy to heterogeneous materials having two components in
series, each with different parallel capacitance and resistance αd = m2/3kT, (14)
characteristics, an additional series resistance and capacitance
are added in Figure 6(c). The circuit analogy of Figure 6(c) is where k is the Boltzman (gas) constant and T is the Kelvin tem-
the one usually used to represent Maxwell-Wagner polarization perature. This relation can be obtained by deriving the statisti-
and the frequency dependence of the loss factor, ε″, and dielec- cal mean amount of orientation of the dipole in an electric field,
tric constant ε′ is derived from it. The series RC circuits of Fig- E, from the ratio of the electric energy, mEcos(θ) of the indi-
ure 6(c) can still be represented by a single time constant equal vidual dipoles to the thermal energy, kT, where θ is the angle
to: between the dipole axis and the electric field direction. Since
the ratio, mEcos(θ)/kT is much less than unity for practical elec-
Maxwell-Wagner Time constant = R1R2 (C1 + C2)/(R1 + R2). tric fields, the mean amount of orientation is rather small and
(13) the polarization is approximately proportional to the applied
electric field. Only if the electric field exceeds about 1 kV/m
The frequency dependence of the capacitance (and real di- does the polarization start to approach being complete (θ =0),
electric constant) due to interfacial polarization as represented and the polarization departs from being proportional to the elec-
by circuits like Figure 6(c) is similar to that for dipolar polar- tric field. This is a qualitative explanation of this derivation,
ization discussed in the next section. The overall trend of ε′ and which can be found in many references, for example [13], [14].
ε″ with frequency is illustrated by the curves in Figure 9(a). The In a similar manner to that used for electron polarization, the
basis for the maximum in ε″ and the decline of ε′ toward higher relation of m to P is reached:
frequencies is similar to that observed in dipolar polarization.
Since the conductivity typically increases (resistivity de-
No ρ No ρ
creases) with increasing temperature, this affects the trend of
the real dielectric constant, ε′ due to interfacial polarization much
like an increase in frequency, since the RC time constant is re-
Pd =
( )
M
αdEL =
M ( ) m2
3kT
EL, (15)

duced Also, the upturn in conductivity toward lower frequency where P is the dipolar polarization, N0 is Avogadro’s number, M
due to the reduced resistance in parallel with the capacitors will, is the molecular weight, ρ is the density, and EL is the local
of course, be increased. electrical field, as influenced by the medium surrounding each
molecule. According to the Lorenz-Lorentz relation for the in-
Dipolar Polarization ternal electric field (which was also used in the case of electron
Dipolar polarization is primarily of importance in insulating polarization): EL = E(ε′ + 2)/3, where E is the applied external
organic liquids and resins where dipolar molecules are suffi- field and ε′ is the dielectric constant.
ciently mobile to be oriented by the electric field. Although di- Now, inserting the value for the local electric field into (15),
pole molecules occur in most inorganic glasses, ceramics, and an adjusted value for P is obtained:
crystals, the molecules are not sufficiently mobile or unique
entities at the temperatures at which these substances are used Noρ m2 ε′ + 2
to rotate as individual molecules in electrical insulation. Dipo-
lar molecules have already been discussed in connection with
conduction mechanisms, since their rotation in an electric field
Pd =
( ) ( )
M 3kT 3
E. (16)

20 IEEE Electrical Insulation Magazine

 Then, combining the dipole polarization with the electron sity, refractive indices, and dipole moments. Data in Table 3 are
polarization and equating the sum to the macroscopic polariza- limited to liquids where the dipoles are able to rotate easily. In
tion, an equation relating the dipole moment and the electron polymers where the dipolar groups are rigidly attached to a poly-
polarizability to the observed dielectric constant results: mer chain, the dipoles are usually unable to rotate until the tem-
perature is raised toward the softening point in which the chain
P = Pe + Pd = (ε – 1)εoE, (17a) can twist (although in a few cases polar groups not rigidly at-
tached to polymer chains can rotate at lower temperatures). The
P=
( ) ( ( ) ( ) ( ))
Noρ
M
αe
ε+2
3
+
m2
3kT
ε+2
3
E.(17b) values for the liquids could be expected to decrease abruptly
with temperature when the 1iquids crystallize during freezing
or more slowly if they become gradually more viscous. Values
Combining (17a) and(17b) gives: in Table 3 have been assembled from [13], [14], [22], [23],
which present more extensive data.
The compounds listed in Table 3 are representative of almost

( ) ( )( )
ε–1
ε+2
=
No ρ
3εoM
αe +
m2
3kT
. (17c) the entire range of dipole moments and liquid dielectric con-
stants. For compounds with zero or very low dipole moments, it
can be noted that the square of the refractive index (η2) is nearly
This relation is reasonably accurate for dipolar gasses and equal to the dielectric constant (ε′).
slightly less so for dilute solutions of dipolar molecules in non- The dipole moment of organic compounds is usually domi-
polar solvents (such as hydrocarbons). It has been used to mea- nated by particular polar groups, usually involving atoms other
sure and calculate the dipole moments of many molecules. But than C or H as is indicated by an inspection of Table 3. When
it is not accurate for dipolar liquids. these polar groups are attached to a rigid molecule like benzene
The reason for the incorrectness of (17) for the condensed the dipole moments will add vectorially. For example, the di-
phase of dipolar molecules lies primarily in the error of the as- pole moment of 1,4-dichlorobenzene, where the –C–C1 groups
sumed value of the internal electric field which is affected by are directly opposite each other, is zero.
the proximity of surrounding molecules and by specific electro- It can be noted in Table 3 that the dielectric constants of the
static attractions between some types of molecules such as those compounds with –OH groups (namely propyl alcohol, water,
with –OH groups, which occur in organic alcohols and acids and ethylene glycol) are higher than would be expected from
and water. In those compounds, a secondary intermolecular “hy- their molecule dipole moments. This is attributed to the spe-
drogen bond” occurs with a specific orientation between mol- cific attraction and orientation of two or several molecules to
ecules. A more correct equation relating the dielectric constant each other through hydrogen bonding. This may also be true to
of dipolar liquids to their dipole moments has been derived by a lesser extent with amino benzene (aniline). This intermolecu-
Prof. Onsager in [13], [14]. It was assumed that the local inter- lar orientation makes them behave as if their effective dipole
nal field was that for an empty spherical cavity in a continuous moment was higher. This has been theoretically analyzed by
fluid with a dielectric constant, ε′. This relation is given in (18): Kirkwood and colleagues [20].
Compounds with high dielectric constants (because the solu-
bility and dissociation constant of ionic materials is greater in
EL = E
( ) 3ε′
2ε′ + 1
. (18) them) are difficult to purify from ionic impurities. These ionic
impurities raise the conductivity so much as to make it difficult
to measure their capacitance and dielectric constant correctly at
The same equation is also used in calculating the electric low frequency. At higher frequencies, however, the loss tangent,
field in gas bubbles in insulation. Onsager’s relation, see (l9), which decreases inversely with frequency for constant conduc-
calculates only the difference between the overall dielectric tivity, is sufficiently low that dielectric constant measurements
constant, ε′, and η2, that is due to the refractive index plus atomic are easier. Note that the values for the highest ε′ liquids are
polarization, which is usually quite small in organic molecules reported at 100 kHz in Table 3.
in contrast to most inorganic solids: The group dipole moments of the groups attached to poly-
mer chains can usually be expected to be the same as in small
individual molecules; but they do not contribute significantly to
(ε′ – η2)(2ε′ + η2) 1 No ρ m 2 the resin dielectric constant, except when the polymer chain or
= . (19)
ε′(η2 + 2) 3εo M 3kT side groups can move easily. This typically occurs at elevated
temperatures as the resin approaches its physical softening point.
The major part of the variation in the left term is in the factor: Thus, viscous liquids and polymers with dipoles often show sig-
(ε′ – η2); this quantity is approximately proportional to m2/M nificant increase in dielectric constant with increasing tempera-
since the other quantities in the right term are nearly constant at ture in the range where the dipolar molecules become suffi-
constant temperature. ciently mobile. A good correlation between the dielectric fre-
Table 3 gives the dielectric constants of different types of quency/temperature characteristics and the mechanical flexibility
molecular liquids together with their molecular weights, den- has been shown for polar polymers.

September-October 2006 — Vol. 22, No. 5 21

 Frequency and Temperature Variations with Prof. A. K. Jonscher [15] has used this as a basis for a differ-
Interfacial and Dipolar Polarization ent empirical approach to the frequency dependence of the di-
The above discussion on interfacial and dipolar polarization electric constants, which will be discussed later. The more fa-
has indicated qualitatively that their contributions to the dielec- miliar circuit analogy of an exponentially declining current with
tric constant and loss vary with temperature and frequency. In time as in (20) leads to the classical Debye frequency depen-
this section the quantitative magnitude of this variation will be dence of polarization and gives a more easily comprehended
discussed. Both of these processes depend on the effective in- picture of the frequency and temperature dependence of the di-
ternal viscosity presented by the material to the movement of electric constants, so it will be discussed first.
ions or rotation of dipole molecules, respectively. The longer Most investigations of dielectric relaxation have been made
range movement of ions up to barriers typically takes a longer by measurements at individual frequencies and extending over
time to develop polarization than the simple rotation of dipoles. a frequency range; but basically the phenomena are in a time
Therefore, ionic interfacial polarization is observed at lower fre- frame, essentially a result of time delay of ion or molecule re-
quencies, where the half cycle is of sufficient time. Dipolar po- sponse to the applied electric field. Figure 7 presents a repre-
larization usually persists up to higher frequency than ionic in- sentative graph of the current through an insulation when a step
terfacial polarization in the same viscosity material, although DC voltage is applied. The initial current is a pulse of very short
the two processes may overlap. The circuit analogy of Figure duration, which would be very difficult to measure or even de-
6(b) is representative of the interfacial (subscript I) and dipolar tect due to its short duration. It is the current to the parallel
(subscript D) polarization and loss, respectively. It should be capacitor, CP, in the circuits of Figure 6, and is the current asso-
particularly noted that the temperature dependence of these po- ciated with electronic or “atomic” polarization, which develops
larizations is quantitatively related to the frequency dependence. in times of the order of a half cycle of optical or infrared fre-
It is easier to understand the frequency (and temperature) quencies. Next, if the insulation is polar in nature, there is an
dependence by the very appropriate analog to the series resistor initial high level of current, which declines in time at a measur-
and capacitor circuits in Figure 6. The equation for the current,
I, resulting from a step voltage, V, applied to a series RC circuit
is, of course, familiar to any one exposed to simple circuit theory:

I = Ioexp(–t/RC) = (Vo/R)exp(–t/RC), (20a)

where t is the time after application of voltages Vo, R is the

resistance, and C is the capacitance. Initially all of the voltage
appears across the resistance, R. Current flows through the re-
sistor until the capacitor is charged up to the full voltage, V. The
voltage across the resistor and the current decline exponentially
in time as the voltage on the capacitor increases. This repre-
sents the polarization process. This simple circuit is analogous
to both ionic-interfacial and dipolar polarizations, but with dif-
ferent RC time constants, respectively. Even the temperature
dependence is indicated, since the resistance, R, value corre-
sponds to the viscous resistance to ion movement or dipolar
molecule rotation, and viscosity decreases with increasing tem-
perature. It is also obvious why the ionic-interfacial polariza-
tion should have a longer time constant than the dipolar polar-
ization, since the “resistor” is “longer” in the former because
the ion movement is typically greater than a molecular rotation.
In the dielectric literature, the time constant of a material is
usually referred to as the relaxation time and the frequency de-
pendent polarization as dielectric relaxation phenomena.
The simple RC circuit analogy discussed above is an ideal-
ized model of the dielectric which assumes there is only a single
time constant that represents the dielectric. Practical dielectrics
are usually more complex, and the DC current empirically of-
ten declines not exponentially but as an inverse power of the
time as in (20b):

I = Iot–n. (20b)
Figure 7. Typical DC current response of insulation to a step
voltage application.

22 IEEE Electrical Insulation Magazine

able rate as in Figure 7. This current is associated with the di- which can also be integrated easily by consulting a table of defi-
pole orientation polarization I (RSD), which declines most rap- nite integrals to give:
idly, or the I (RSI) associated with the ionic-interfacial polariza-
tion, which decays more slowly because of its longer time con-
stant. With proper instrumentation, both of these current steps V ω VC(RC)ω
Ireal(ω) = = . (25)
are measurable and differentiated. Conventional recorders, how- R ((1/RC)2 + ω2) 1 + ω2(RC)2
ever, usually measure only the slower polarization current, much
like the current curve in Figure 7. A much faster polarization After putting RC = τ the relaxation time constant, and εo′ =
current could be hidden in the initial pulse. C/Co, the frequency function of the in phase (loss) current re-
Eventually, the insulation current levels off, as illustrated in sults:
Figure 7, at a level determined by the longtime DC conductiv-
ity of the insulation, which is represented by RP in the circuit
analogies of Figures 6(b) and6(c). Routine DC insulation resis- ωτεo′
Ireal(ω) = VωCo = VG. (26)
tance measurement standards arbitrarily establish 1 minute as 1 + ω2τ2
the time at which the DC insulation resistance should be re-
ported. This practice will sometimes include some polarization The in-phase current component is also plotted in Figure 8(b),
current and not be a true DC resistance. also normalized by dividing by 2πfCo.
It is possible to deduce from time measurements of current, If there were some parallel conductance, as in Figure 6(c),
as in the above case of a step voltage applied to an RC circuit, which occurs often in insulation, there would be an additional
what the frequency dependence of the circuit or material is. G term that would cause an upturn of the loss factor at the low
Indeed, time response is used in practical dielectric measure- frequency end of the curve on Figure 8(b). (This effect is ne-
ments to deduce frequency response. Time and frequency are glected in Figure 8(b), but is illustrated in Figure 12).
exactly related through a Fourier transform or integral. The Fou-
rier cosine integral of the time function, f(t), of the current, I,
gives the out of phase (imaginary) frequency dependence of the
current (exclusive of the effect of parallel capacitance and con-
ductance shown in Figure 6:
∞
I′(ω) = jω 兰ο f(t) cos (ωt)dt, (21)

where ω = 2πf and [assuming an exponential decline in current

from equation (20)] f(t) = (V/R)exp(–t/RC), which is plotted in
Figure 8(a) with an assumed time constant of 500 ms.
The above integration can be made simply by referring to a
handbook table of definite integrals. It yields:

ωV(1/RC) ωCV
I′(ω) = j = j . (22)
R((1/RC)2 + ω2) 1 + ω2(RC)2

Now replacing the time constant, RC, with τ and C/Co with
εo′, where Co is the capacitance at zero frequency, the following
relation results:

ωCoε′ο
I′(ω)/V = j = jB. (23)
1 + ω2τ2

This derived imaginary (out of phase) current, normalized to

unity by dividing by 2πfCo is calculated and plotted in Figure
8(b).
Correspondingly, the Fourier sine integral gives the frequency
function of the current in phase with the applied voltage:
Figure 8. Calculated current response of a series RC circuit
∞
to a step DC voltage. (a) I vs. time. (b) I vs. frequency.
Ireal(ω) = ω
V
R
兰
ο exp(–t/RC)sin(ωt)dt, (24)

September-October 2006 — Vol. 22, No. 5 23

 Equations (23) and (26) can be combined to give the overall factor, ε″, or in-phase current, is a maximum equal to (1/2) (εo –
current as a function of frequency: ε∞) for an ideal, single relaxation time. In fact such an equiva-
lence is a good test for a single relaxation time. With a distribu-

( )
tion of relaxation times, the maximum would be lower and
ωτεo′ εo′ broader, and the decline in real dielectric constant more gradual.
I(ω) = VωCo +j . (27)
1 + ω2τ2 1 + ω2τ2 This is often seen with polar polymers. Curves for ε′ and ε″ for
a single relaxation time and an often observed type of deviation
This has the same form as the Debye relation for the fre- having a maximum ε″ value 50% of that for a single relaxation
quency dependence of the “real” and “imaginary” dielectric time and indicating a symmetric distribution of relaxation times
constants with an assumed single relaxation time, τ. In the Debye are given in Figure 9(a). (The shape of the curve corresponds to
approach, the principle current is assumed to be capacitive, so the Cole-Cole arc relation between ε″ and ε′ for ∝ = 0.36 as in
the j factor (for out of phase current) appears in front of the (30b) and Figure 10(b) which will be discussed later.) The AC
expression, and the in-phase current component is assigned a conductivity, 2πfε″ curves corresponding to the same param-
negative j factor to make it “real” and to indicate that it lags the eters are plotted in Figure 9(b).
applied voltage. The AC conductivity always increases with frequency, and it
Since the above derivation applies only to the polarization should be clearly understood that the frequency variations due
over and above the electron polarization (or any other high fre- to the various conduction-polarization processes discussed here
quency polarization which is constant through the frequency are merely “waves” superimposed on the steady upward trend
range of the interfacial or dipolar polarization being consid- of the AC conductivity with increasing frequency. Although the
ered) the εo′ term of (27) is replaced by the difference between conductivity appears to be leveling off in Figure 9(b), these
the low-frequency and high-frequency dielectric constants, εo –
ε∞, to give equations (28a) and (28b):

)
(a)
I(ω) = jVCo
((εo – ε∞)
1 + ω2τ2
–j
(εo – ε∞)ωτ
1 + ω2τ2
, (28a)

I(ω) = jVCo(ε′(ω) – jε′′(ω)), (28b)

where Co is the geometric, vacuum, capacitance.

In Figure 8(b) the variation of out-of-phase current with fre-
quency (B/2τfCo) corresponds to the variation in the real di-
electric constant for a single relaxation time. If there were an-
other relaxation process with a sufficiently different (single) re-
laxation time, it would follow a similar curve form at a lower
frequency, or a higher frequency, and would add to this.
The dielectric literature presents the complex dielectric con-
stant, ε*, which is given in (29a) in several different equivalent (b)
ways as summarized in (29a), (29b), (29c) and (29d).

ε*(w) = ε′(ω) – jε(ω), (29a)

ε*(w) = ε′(ω) – jε′(tan δ(ω), (29b)
εo – ε∞
ε′ = ε∞ + , (29c)
1 + ω2τ2

ε′ = (εo – ε∞)ωτ . (29d)

1 + ω2τ2

In the real world, materials do not always behave as this

simple theory of a single relaxation time assumes. More often
than not, there is a distribution of relaxation times, or some
mechanism that gives a frequency dependence which can be
interpreted as a distribution of relaxation times. A distribution
of relaxation times is reasonable to expect from some heteroge- Figure 9(a). Dielectric constant, εr′, and loss factor, εr″, for a
neity in the inter-molecular structure, particularly of polymers. Debye-type polarization. (b) AC conductivity, σ, due to a
Note that in Figure 8(b) and equation (28) when ω = 1/τ the loss Debye-type polarization.

24 IEEE Electrical Insulation Magazine

 (εo – ε∞)
ε* = ε∞ + , (30a)
(1 + jωτ)

(εo – ε∞)
ε* = ε∞ + , (30b)
(1 + jωτ)1–α

(εo – ε∞)
ε* = ε∞ + . (30c)
(1 + jωτ)β

Several types of deviations from the simple semicircle (which

are presumed to be due to a distribution of relaxation times)
have been observed with polar liquids and polymers. Two of
these are indicated by the curves of Figures 10(b) and10(c). One
of these [Figure 10b and equation (30b)] is a semicircle with its
center below the zero axis and was reported in [18]. The sec-
ond, reported in [19], has a low frequency part like a quarter
circle, but it approaches the zero axis at the high-frequency end
at an acute angle [Figure 10c and equation (30c)]. The former
deviation suggests a range of high-frequency relaxation times
superimposed on a single relaxation time at lower frequency.
Such relations at present remain empirical and have no detailed
theoretical support, but they are useful in analyzing frequency
variations of the real and imaginary dielectric constants.
Another more general way of relating the loss factor, ε″, to
the changes in the real dielectric constant; ε′, is through the
basic Kronig-Kramer equations, (31a) and (31b), which are in-
dependent of the particular algebraic relation of ε″ or ε′ to fre-
quency or distributions of relaxation times:

2 ∞ xε″(x) 2 ∞ ε″
Figure 10. Cole type plots of ε″ vs. ε′. ε′ = 兰ο dx or 兰ο dx, (31a)
π x2 – ω2 π x
at ω = 0

curves apply only to these dipolar or ionic interfacial conduc- 2ω ∞ ε′(ω)

tion effects. There is some unidentified mechanism of dielec- ε″ =
π
兰ο x2
dx, (31b)
tric loss or conduction which maintains the loss tangent for all
liquids and solids at a minimum in the range of about 10–5 to
where (31a) predicts that the integral of the curve of the loss
10–4, as is indicated in Figure 1.
factor versus the loge frequency is proportional to the real di-
Another important relation to be considered in connection
electric constant at ω = 0. Equation (31a) can be easily con-
with the frequency variation of ε′ and ε″ is the relation of these
firmed by applying it to (29d). This indicates that changes in
quantities to each other. Since each is a function of frequency,
real dielectric constant ε′ above zero frequency are always asso-
they are functions of each other as the frequency is varied. This
ciated with conduction, as has been discussed qualitatively ear-
was first studied by Cole and Cole [16] who reported that a plot
lier.
of ε″ vs. ε′ on the abscissa over a range of frequency was, for a
Another empirical method of analyzing the frequency de-
material having a singe relaxation time, a semicircle with its
pendence of dielectric response has been suggested by A. K.
center on the ε″ = 0 axis, as is illustrated in Figure 10(a). Each
Jonscher [15], who has published an extensive analysis of the
point on the curve is for ε″ and ε′ at one particular frequency.
field of dielectric relaxation [15]. A Laplace transform of equa-
The semicircle intersected the zero axis at εo′ on the low fre-
tion (20b) for the inverse power dependence of the current ver-
quency (high ε′) end, and at ε∞ on the high frequency (low ε′)
sus time yields the corresponding frequency dependence of the
end. It is possible to derive algebraically the semicircle relation
complex dielectric constant, (32).
between ε″ and ε′ by eliminating ω from (29c) and (29d). This
is equiva1ent to the more general complex equation (30a) de-
ε∗ = (jω)k–1. (32)
rived by Cole and Cole [16], [17]:

September-October 2006 — Vol. 22, No. 5 25

 This yields a double power relation, (33), for the loss factor,
ε″, dependence on frequency:

1
ε″ = . (33)
(ω/ωp)–m + (ω/ωp)1–n

The exponents m and (1 – n) fall in the range between 0.1 and 1

for all the many materials in the literature whose data Jonscher
analyzed. He recommends plotting AC loss factor data vs. fre-
quency on log-log scales, which, according to the above equa-
tion yields two lines of different slopes at high and low fre-
quency, intersecting with a gradual change in slope in the vicin- Figure 11. Comparison of tan δ and loss factor, ε″,curves for
ity of the frequency of the loss factor maximum. Debye type polarization, (εo = 6, ε∞ = 2, α = 0).
The theoretical interpretation of dielectric data by Jonscher
rejects the traditional concept of a distribution of relaxation times
and supports the plotting of both dielectric loss factor and di-
electric conductivity on log-log scales. This has the effect of Long chain polymers with dipolar groups attached to the chain
compressing the variation of these quantities, but it may be use- at periodic intervals show similar types of curves of ε′ and ε″
ful in systematizing data on many different types of materials. versus frequency or temperature as do dipolar liquids, but at
More details on this approach can be found in [15]. much lower frequencies or at higher temperatures since the in-
ternal viscosity of the resin is much greater. Increased viscosity
shifts the loss factor, ε″, maxima and related dielectric constant,
Temperature Variations with Interfacial and
ε′ dispersion region to lower frequencies at the same tempera-
Dipolar Polarizations ture, or conversely. Data on the dielectric constant and loss fac-
The variation in loss factor ε″ and real dielectric constant, ε′, tor are often plotted as a function of frequency at several fixed
with temperature can be understood most easily through the ef- temperatures, as in the curves in Figure 12 for a polyvinyl ac-
fect of temperature on the time constant, τ, or primarily, R, the etate resin. These curves, from [24], illustrate the interrelation
resistance, in the series RC circuit analogy of a dielectric, as in of temperature and frequency on the trend of ε′ and ε″. It is
equations (20a), (20b), and the following equations. noteworthy also that the heights of the loss factor maxima are
The equations (29c) and (29d) for the real and imaginary decreased with increased temperature. This is correlated with a
dielectric constants involve the time constant, τ, multiplied by decreased low frequency dielectric constant, εo′, [refer to equa-
the frequency, ω. Therefore, a decrease in time constant is equiva- tion (29d)]. The low-frequency dielectric constant is decreased
lent to an increase in frequency for the same level of ε′ or ε″. according to the temperature influence on the dielectric con-
The time constant can be assumed to be roughly proportional to stant indicated in (l9d), where the temperature appears in the
the effective internal viscosity of the dielectric, which resists denominator of the dipole moment term and also in the density
the ion translation or the dipole rotation by the electric field. In term, since the density decreases with increasing temperature.
turn the viscosity is often found to be a decreasing function of Polyvinyl acetate structures include an ester group attached
the temperature according to an exponential activation energy to the main polymer chain with one dipole component rigidly
dependence. Thus, τ is a function of temperature according to attached to the chain so it cannot move unless the chain moves,
(34): and a second dipole component which can oscillate a little in-
dependently of the chain. Thus, there is at least one very high-
frequency polarization process which is not delineated by the
τ = A exp
( )
–Q
RT
. (34) data plotted in Figure 12. This is indicated more clearly by a
plot of ε″ vs. ε′ like those in Figure 10. The major portion of the
similar type of curve for the polyvinyl acetate resembles the
Therefore, plots of ε′ and ε″ versus increasing temperature Cole-Davidson skewed arc type of curve [Figure 10(c)] but with
will show a trend similar to plots of ε′ and ε″ versus increasing a slightly depressed circle center. Also notable is the fact that
frequency, but reversed. the ε′ data for polyvinyl acetate extrapolate at the higher fre-
Some of the frequency and temperature dependencies of ε′ quency end to a value significantly above the square of the re-
and ε″ presented above in equation or idealized graph form are fractive index, thereby indicating a dipole which can easily os-
now illustrated with plots of some experimental data. cillate at quite high frequencies, but its contribution to the di-
Sometimes in the literature, the loss tangent rather than the electric constant is not large. This mobile dipole probably in-
loss factor is plotted versus frequency. The effect of this is illus- volves the >C=O group in the ester group attached through a
trated in Figure 11. The maximum of the loss factor curve is rotatable carbon-carbon bond to the main polymer chain.
shifted a little in temperature or frequency from the correspond- If the log of the frequencies at the maxima of the loss factor,
ing curve for loss tangent. ε″, from Figure 12 are plotted—according to (34)—vs. the re-

26 IEEE Electrical Insulation Magazine

 tration and movement, the effect predominates in insulations
with higher dielectric constants, which dissolve and dissociate
more ionic impurities and lower viscosity materials, which per-
mit higher ionic mobilities.
It should be noted, however, that the loss factor of the paral-
lel conductance decreases inversely with the frequency (ε″ = G/
2πfC), so the loss factor will be influenced most at low fre-
quency. The effect of parallel conductance (occurring together
with dipolar conduction and polarization) on ε″ is illustrated by
the up turn of the ε″ curves for 90ºC and 100ºC at low frequency
in Figure 12. This is a typical indication of parallel connduction
in the materia1. (The magnitude of the effect would be related
to the DC conductivity.)

Conclusions
This article shows how dielectric data can be helpful in in-
terpreting molecular behavior and structure, and conversely how
a knowledge of molecular structure can aid in predicting di-
electric properties where these are critical for an application.

References
[1] L .C. Pauling, The Nature of the Chemical Bond. Ithaca, NY: Cornell
Univ. Press, 1939.
[2] P. W. Atkins, Physical Chemistry, San Francisco, CA: W. H. Free-
man & Co., 1978.
[3] G. W. Castellan, Physical Chemistry, Reading, MA: Addison Wesley,
1971.
[4] C. Kittel, Introduction to Solid State Physics, New York: Wiley, 1956.
[5] A. Rembaum, J. Moacanin and H. A. Pohl, in Progress in Dielec-
trics, vol. 6, J. B. Birks and J. Hart, Eds. London: Heywood, 1956,
Figure 12. Loss factor, ε″ , and relative dielectric constant of pp. 41, 102.
polyvinyl acetate 5. [6] A. J. Dekker, Solid State Physics, Engelwood Cliffs, NJ: Prentice
Hall, 1959.
[7] N. F. Mott and R. W. Gurney, Electronic Processes in Ionic Crystals,
ciprocal Kelvin temperature, the slope of the line yields the ac- Oxford Clarendon, 1940.
tivation energy for the polarization process. This activation en- [8] D. A. MacInnes, Principles of Electrochemistry, New York: Reinhold,
ergy has been compared by theorists to the energy (heat) of 1939.
vaporization of the part of the polymer participating in the di- [9] R. M. Fuoss and C. Akraus, J. Am. Chem. Soc., vol. 55, pp. 476,
electric orientation like that of Figure 12. From this the number 1019, 1933. Also: J. Chem. Phys., vol. 2, p. 386, 1934.
of units of the polymer chain that are participating in the dielec- [10] A. Gemant. Ions in Hydrocarbons, New York: Interscience, 1962.
tric orientation and relaxation is inferred; [25] may be consulted [11] A. von Hippel, Dielectrics and Waves, Cambridge, MA: M. I. T.
for a review of the theory of dielectric relaxation of polar poly- Press, 1954. Also: J. Volger, Progress in Semiconductors, London:
Heywood and New York: Wiley, p. 211, 1960.
mers.
[12] P. Debye, Polar Molecules, Chemical Catalog, Co., 1929; reprint by
Nonpolar polymers such as polyethylene, polypropylene, and New York: Dover, 1945.
polystyrene have no dipolar groups (except possibly as impuri- [13] C. P. Smyth, Dielectric Behavior and Structure, New York: McGraw
ties) so their dielectric properties vary little with temperature Hill, 1955.
and frequency. In these materials, there is usually little or no [14] C. J. F. Bottcher (revision by O. C. van Belle, P. Borewijk, and R. A.
significant ionic-interfacial or dipolar polarization represented Rip), Theory of Electric Polarization. Amsterdam, The Netherlands:
by the series resistance in Figure 6. Elsevier , 1973.
A major effect of temperature on inorganic glasses and ce- [15] A. K. Jonscher, J. Material Science (G. B.) vol. 16, pp 2037-2060,
ramics, as well as many plastics and insulating liquids, is to 1983; and Dielectric Relaxation, Chelsea Dielectrics Press, London,
1983.
decrease the parallel resistance, RP, indicated in the circuit anal-
[16] K. S. Cole and R. H. Cole, J. Chem. Phys. vol.19, p. 1484, 1951.
ogy of Figure 6(b) and (c). This effect may occur alone or in [17] R. H. Cole, pp. 47–101, Progress in Dielectrics, vol. 3, J. B. Birks
addition to the effects of dipolar and ionic-interfacial polariza- and J. Hart, Eds., New York: Wiley, 1961.
tion on the conductivity. For all insulations, the internal resis- [18] D. W. Davidson and R. H. Cole, J. Chem Phys. vol. 19, p. 484, 1951.
tance to parallel ionic movement decreases as the temperature [19] L. Onsager, J. Am. Chem. Soc., vol. 58, p. 1486, 1936.
climbs, causing the electrical conductance, 2πfε″εo, to increase [20] J. B. Kirkwood and G. Oster, J. Chem. Phys. vol. 11, p. 175, 1943.
with temperature, Since this conduction is due to ionic concen- [21] L. G. Wesson, Ed. and Compiler, Tables of Electric Dipole Moments,
Cambridge, MA: M. I. T. Press, 1948.

September-October 2006 — Vol. 22, No. 5 27

 [22] P. Debye and H. Sack, Dielectric Constants and Dipole Moments, Editors Note:
(Part of Ann. Tables of Constants), Published by Hermann et Co., This article is a chapter written by Dr. T. W. Dakin that was
Paris, 1937.
to be part of a book on dielectrics. Although colleagues of Dr.
[23] A. von Hippel, Ed., Tables of Dielectric Materials in Dielectric Ma-
terials and Applications, Cambridge MA: M. I. T. Press, 1954 (Note: Dakin think that other chapters were written, no other chapters
Some of these data are reproduced in recent editions of the Hand- have been found. Dr. Dakin wrote this chapter probably between
book of Chemistry and Physics, Chemical Rubber Co. Press, Cleve- 20 and 30 years ago while he was working at the Westinghouse
land OH). Research laboratories in Pittsburgh. In spite of the article’s age—
[24] T. W. Dakin, Trans. Electrochem. Soc., vol. 83, pp. 175–85, 1943. or maybe because of its age—it provides an excellent tutorial
[25] W. Kauzmann, Rev. Modern Physics, vol. 14, p. 12, 1942. on the subject of conduction and polarization mechanisms and
[26] J. J. Lander, Semiconductors, N. B. Hannay, Ed., Nwe York: their trends in dielectrics. As it is unlikely that the other chap-
Reinhold, 1959, Chapter 2. ters will be found or published, some colleagues of Dr. Dakin
thought the chapter should be published and the Electrical In-
sulation Magazine, a source of tutorial-type articles, was a logi-
cal choice.

28 IEEE Electrical Insulation Magazine