See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331883752

Influence of dielectric constant on dielectric strength by defect discharge and

molecular polarization in solid insulation materials

Article in Journal of Applied Physics · March 2019

DOI: 10.1063/1.5086377

CITATIONS READS

5 870

6 authors, including:

Liu Jing Liang Zhao

Peking University Northwest Institute of Nuclear Technology
10 PUBLICATIONS 19 CITATIONS 84 PUBLICATIONS 621 CITATIONS

SEE PROFILE SEE PROFILE

Yuanrong Lu
Peking University
212 PUBLICATIONS 1,401 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

laser plasma ion acceleration View project

SSC-LINAC View project

All content following this page was uploaded by Liang Zhao on 21 April 2021.

The user has requested enhancement of the downloaded file.

Influence of dielectric constant on dielectric
strength by defect discharge and molecular
polarization in solid insulation materials
Cite as: J. Appl. Phys. 125, 115103 (2019); https://doi.org/10.1063/1.5086377
Submitted: 20 December 2018 . Accepted: 04 March 2019 . Published Online: 19 March 2019

Jing Liu , Jiancang Su, Liang Zhao, Rui Li, Yuanrong Lu, and Xiaolong Liu

ARTICLES YOU MAY BE INTERESTED IN

Numerical investigation on the effects of discharge conditions on a nanosecond pulsed

surface dielectric barrier discharge
Journal of Applied Physics 125, 113301 (2019); https://doi.org/10.1063/1.5052697

Electric field control of magnetism in Si3N4 gated Pt/Co/Pt heterostructures

Journal of Applied Physics 125, 114101 (2019); https://doi.org/10.1063/1.5083148

Phonon-limited mobility modeling of gallium nitride nanowires

Journal of Applied Physics 125, 114301 (2019); https://doi.org/10.1063/1.5072759

J. Appl. Phys. 125, 115103 (2019); https://doi.org/10.1063/1.5086377 125, 115103

© 2019 Author(s).
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

Influence of dielectric constant on dielectric

strength by defect discharge and molecular
polarization in solid insulation materials
Cite as: J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377
Submitted: 20 December 2018 · Accepted: 4 March 2019 · View Online Export Citation CrossMark
Published Online: 19 March 2019

Jing Liu,1,2,a) Jiancang Su,1 Liang Zhao,1 Rui Li,1 Yuanrong Lu,2 and Xiaolong Liu1

AFFILIATIONS
1
Science and Technology on High Power Microwave Laboratory, Northwest Institute of Nuclear Technology, Xi’an,
Shaanxi 710024, China
2
State Key Laboratory of Nuclear Physics and Technology, Peking University, Beijing 100871, China

a)
Electronic mail: [email protected]

ABSTRACT
Dielectric constant and dielectric strength are two intrinsic electrical parameters of solid insulating materials. In order to finally understand
the relationship between them, the one-sided influence of dielectric constant on dielectric strength is investigated. For theoretical analysis,
we propose that the influence is realized indirectly by two mechanisms: defect discharge in the micro level and molecular polarization in the
nanoscale. The transition phase is composed of the field enhancement related to permittivity and the breakdown of defect molecular bond.
While the local field around the bond is always higher than the cavity interior field, the breakdown field is a competition between the two
mechanisms. According to the simulation model of electrostatic field, the electric field around the cavity is significantly enhanced with a
larger dielectric constant. In view of the simulation result of partial discharge, the discharge intensity of dielectric and gas breakdown in the
cavity is promoted by the increase of material permittivity. To confirm the defect distribution and quantity, several prepared samples are
scanned with the ultrasonic microscope and the cavities are measured via image software. Based on the published experiment data from
oxide films and polymer bulks, a revised relationship is plotted and fitted for the dielectric strength and the dielectric constant. As molecular
polarization is applicable to oxide films with few defects and high permittivities, defect discharge is more effective for polymer bulks with
generous cavities and low permittivities. Since molecular polarization leads to breakdown by enhancing the local electric field, cavity dis-
charge is mainly due to the lower breakdown threshold of defect.

Published under license by AIP Publishing. https://doi.org/10.1063/1.5086377

I. INTRODUCTION solid insulation materials, few studies have been published on the
relationship between them.
Solid dielectric is widely used in electrical equipments, elec-
tronic components, pulsed power generators, and other fields for In the classical solid dielectric theory, breakdown is mainly
divided into four forms: electrical, thermal, electromechanical, and
insulation. The dielectric strength of polymers determines the
electrochemical. For the ideal solid dielectric, a proposed model
volume, life, and stability of insulators in the electric power indus-
suggests that the local electric field tends to weaken polar molecular
try and pulsed power technology.1,2 The breakdown strength of bonds and causes breakdown.5,6 To practical insulation materials,
thin oxide films has historically been of great reliability interest in defects such as cavities, impurities, and dislocation are usually
integrated circuit applications.3,4 There are a lot of related studies spread inside, which will obviously reduce the dielectric strength.7,8
on the breakdown characteristics and life evaluation of polymer If the dielectric is subjected to an electric field, a tiny leakage
insulators with a low dielectric constant. Besides, the breakdown current is generated and partial discharge will occur at interior
and the reliability of gate dielectrics with high permittivity in oxide defects.9 Because partial discharge is one of the main reasons for
films have been studied in detail. Although dielectric constant and dielectric deterioration and damage, we believe that defects should
dielectric strength are two intrinsically important parameters of be considered. When an external electric field is applied, all

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-1

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

molecules in the material are polarized synchronously and electrons emission (Fowler-Nordheim tunneling) begins to contribute a signifi-
are ionized, accelerated, or recombined. Since dielectric breakdown cant emission current13
is a phenomenon of electronic instability in the electrical break- pffiffiffiffiffiffiffiffiffiffi 3=2 !
down theory, polarization is another essential mechanism. 4 2qm* wB
J f ¼ qvR n exp
: (2)
In this paper, the influence of dielectric constant on dielectric 3 h E
strength is investigated, while the effect of dielectric strength on
dielectric constant is not mentioned. First, the defects breakdown Here, V a is the applied voltage and V t is the thermal voltage. wB is
mechanism, the cavity partial discharge, and the molecular polari- the barrier height, l is the barrier distance, and the average field E
zation process are discussed for theoretical analysis. Then, the sim- equals wB =l. The Richardson velocity with which the electrons
ulation results of the electrostatic field and the discharge pulse approach the barrier is obtained from
inside the cavity are presented with different material permittivities. rffiffiffiffiffiffiffiffiffi
kT
Finally, the ultrasonic inspection of the polymer and fitting of vR ¼ : (3)
experimental data are carried out to verify the existence of the 2πm
cavity and influence of permittivity. Equations (1) and (2) prove that the current is the product of the
carrier properties (charge q, velocity vR , and available density n) and
the emission probability of the exponential term. These mechanisms
II. THEORETICAL ANALYSIS of electron injection also exist at the interface of cathodes and dielec-
A. Local breakdown of defects trics.14,15 As the discharge of the metal impurity is essentially the
breakdown of the surrounding material, its initial electric field of dis-
Due to the air permeability of material, the cavity in the
charge is also larger than the gas impurity. If the dielectric strength
polymer is always filled with gas. As a result of the preparation
of the gas is considered lower than that of the liquid and solid, then
environment, the cavity in the oxide gate dielectric is vacuum. In
gas impurity is the most vulnerable to breakdown.10
addition to the cavity, there are a little moisture and metal impuri-
ties remaining in the dielectric generally.10 Unlike the cavity which
is difficult to eliminate, drying is effective in removing the moisture, B. Partial discharge of cavity
and metal detritus will be reduced by improved processing technic. Altogether, it is reasonable to use the cavity to represent all
Dislocation is another defect that exists widely in natural and artifi- defects in a quantitative breakdown model. For a typical gas-filled
cial solid dielectrics. Since dislocation changes the local structure of cavity, the initial electric field of partial discharge is determined by
the material, it is regarded as a nonmetallic impurity with different the breakdown threshold of gas. According to Paschen’s curve, the
physical properties from the dielectric.11 Thus, most of the defects breakdown electric field of gas is described by16,17
can be dealt with impurities of gas, liquid, and solid, in order to
make a comparison in a consistent model. Vgas Bp
Egas ¼ ¼ : (4)
The dielectric constant of gas impurities is very small and the d ln (Apd)
ln [ ln (1 þ 1=γ)]
increase of a local electric field by molecular polarization is negligi-
ble. But the polarized charges at the interface of the cavity and the Here, V gas is the gas breakdown voltage, d is the gap distance, p is
dielectric result in the electric fields in the gas to be greater than the pressure, γ is the secondary electron emission coefficient, A is
the surrounding material. The dielectric constant of liquid impuri- related to the saturation ionization, and B involves the excitation
ties (generally moisture) is higher than most solid dielectrics except and ionization energies.
for oxides. Although the overall electric field inside the liquid is not In fact, there are various shapes of cavities in materials, but all
apparently enhanced by polarized charges, the local electric field is of them can be approximated as ellipsoidal cavities.7,9 It is natural
greatly improved by molecular polarization. For dislocation which to define the electric potential at the origin of the ellipsoidal cavity
is handled by the nonmetallic solid impurity, its relative permittiv- as zero. The principal axes of the ellipsoidal cavity (a, b, c) are
ity and internal electric field are close to the original dielectric. assumed to lie on the Cartesian axis (x, y, z), respectively. In order
Generally speaking, the qualitative comparison shows that the elec- to analyze the effect of defects separately, the applied electric field
tric fields in the gas and liquid impurities are greater than the non- in the dielectric is supposed to be uniform. Generally, this exter-
metallic solid impurity and the dielectric. nally applied field ~ E0 equals the electrodes voltage U 0 divided by
But the influence of metal impurity is quite different from the the dielectric thickness d0 . Therefore, the electric potential distribu-
above types as it brings partial discharge around the impurity rather tion of a single cavity inside the dielectric is derived with Laplace
than its interior. On the one hand, the metal impurity distorts the equation and boundary conditions as
electric field nearby, and on the other hand, its free electrons ε2 E0x x ε2 E0y y ε2 E0z z
are injected into the dielectric. When strengths of the electric field are wcav ¼


,
ε2 þ (ε1
ε2 )na ε2 þ (ε1
ε2 )nb ε2 þ (ε1
ε2 )nc
lower than about 100 kV=mm
1 , thermal emission (the Richardson-
(5)
Dushman effect) is relatively accurate with a current of 12


  ð1
w Va abc dμ
Jt ¼ qvR n exp
B exp

1 : (1) nv ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (v ¼ a, b, c),
Vt Vt 2 0 ( μ þ a ) þ ( μ þ b2 ) þ (μ þ c2 )(μ þ v2 )
2

1
For electric field strengths higher than 100 kV=mm , field electron (6)

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-2

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

where ε1 and ε2 are the dielectric constants of the cavity and mate- In the study of breakdown characteristics of polymer speci-
rial, E0x , E0y , and E0z are the components of uniform electric field mens, another breakdown criterion related to hole capture is pro-
~
E0 . Besides, nv (v = a, b, c) is the shape factor of the ellipsoidal posed.20 The free electron in the dielectric is accelerated by the
cavity, which is determined only by the length of three principal externally applied electric field within a distance of the mean free
axes. In particular, the shape factor equals 1/3 for a spherical cavity path λ. If the electron is captured by a hole and its obtained kinetic
(a ¼ b ¼ c), which happens to coincide with the common L ¼ 1=3 energy satisfies ΔH 0 , the molecular bond will break off. The differ-
for a Lorentz spherical cavity. For convenience, the applied electric ence between these two criteria is that p0 in Eq. (11) is replaced by
field ~ E0 is supposed to follow the direction of x axis and eλ. Hence, the dielectric strength decreases with the increase of the
E0y ¼ E0z ¼ 0. As the dielectric constants of the cavity and vacuum dielectric constant, which is verified in several polymer materials
are about the same, the electric field distribution inside the cavity is including PTFE, PMMA, Nylon, and PE.21
solved as9
~ 1 ~ D. Comparison of two mechanisms
Ecav ¼
rwcav ¼ E0 : (7)
1 þ (ε
1
r
1)na
Based on Eqs. (7) and (9), two field enhancement factors are
However, Eq. (7) is solved for a single cavity which is located in an defined as f cav ¼ Ecav =E0 and f loc ¼ Eloc =E0 , which are indepen-
infinite uniform electric field. Then, the details of the cavity size dent of the uniform applied electric field ~ E0 . As the factors are
and the dielectric thickness are lost. When the relative size of the expressed as functions of the relative permittivity εr , they are
cavity and dielectric is introduced later, Eq. (7) will be applicable to plotted with na ¼ L ¼ 1=3 in Fig. 1(a). Along with the increasing
both the cavity shape and size. Based on Eqs. (4) and (7), the crite- of relative dielectric constant εr , f cav gradually approaches the limit
rion for cavity partial discharge in a dielectric is given by of 1.5, while f loc grows continuously and linearly. Only if εr equals
1, the field enhancement factors f cav and f loc are the same. But it
Ebd ¼ [1 þ (ε
1
r
1)na ]Egas : (8) is unpractical for the natural and artificial solid insulation material.
The occurrence of cavity discharge is usually accompanied by the Thus f loc of molecular polarization is always beyond f cav of cavity
formation of small channels. Once the electrical tree structure ulti- discharge at any relative dielectric constant εr . When applied with
mately bridges the electrodes, the material loses its insulation the same external field ~ E0 , the electric field around polar molecular
capability. bond ~ Eloc in the nanoscale is much larger than the interior of the
cavity ~ Ecav in the micrometer level. It should be pointed out that
f cav in the actual cavity is directly related to the cavity size. Because
C. Polarization process of molecules no infinite dielectric really exists as the hypothesis in Eq. (7). Due
According to the Lorentz relation or the Mossotti field, the to the boundary conditions of dielectric, f cav is affected by the rela-
local electric field ~
Eloc in homogeneous and isotropic dielectrics is a tive size of the cavity and dielectric.
superposition of the externally applied field ~E0 and the dipolar field However, the breakdown electric fields of these two mechanisms
due to the polarization ~ P. In general, the local electric field is exhibit another trend as Fig. 1(b) shows. From Eqs. (8) and (11), it is
expressed as obvious that the breakdown field (dielectric strength) is influenced by
the dielectric constant. Besides, the breakdown criteria of cavity dis-
Eloc ¼ ~
~ E0 þ L(~
P=ε0 ) ¼ [1 þ (εr
1)L]~
E0 , (9) charge and molecular polarization are very similar in the mathemati-
cal form. With Egas ¼ 43:6 kV=mm at the inflexion of Paschen’s
where εr is the relative dielectric constant and L is the Lorentz
curve and ΔH 0 = p0 ¼ 2000 kV=mm for Pr2 O3 , the breakdown crite-
factor.18 Since the factor is initially calculated by a Lorentz spheri-
ria are compared. Because the increase of f cav is little when εr
cal cavity, L ¼ 1=3 has long been used for both cubic-symmetry
exceeds 10, the breakdown electric field stays near 30 kV/mm. The
and amorphous dielectrics. As the local electric field increases with
influence of εr on Ebd mainly operates at a low dielectric constant
the dielectric constant, the polar molecular bonds will be distorted
(,10) by the cavity discharge mechanism. Since f loc is proportional
and weakened in high permittivity materials, then becomes very
to εr , the breakdown electric field decreases sharply with the increase
susceptible to breakage by standard Boltzmann process.5 While the
of the dielectric constant. Then, εr owns an important effect on Ebd
bond strength in the absence of field is ΔH 0 , the enthalpy of activa-
via the molecular polarization mechanism to all dielectric constants.
tion (activation energy) ΔH in the presence of field is reduced by
A specific relative dielectric constant is defined as εc , where
ΔH ¼ ΔH0
p0 Eloc , (10) the influences of both mechanisms on breakdown field are the
same. Although ~ Eloc is far higher than ~ Ecav , the cavity is more easier
where p0 is the molecular dipole moment component opposite to to breakdown than the molecular bond if εr , εc . The reason is
~
Eloc . Bond breakage is expected when the activation energy ΔH that Egas is just about 1/40 of ΔH 0 = p0 , which counteracts the
tends to zero, which is one of the criteria for breakdown advantage of polarization over discharge in the electric field. The
ΔH0 breakdown threshold of the cavity and the molecular bond is more
Ebd ¼ [1 þ (εr
1)L]
1 , (11) decisive in this low permittivity region. When εr is over εc , Ebd of
p0
the molecular polarization process is smaller than cavity partial
where Ebd is the dielectric strength (breakdown field).19 Once the discharge. The mechanism of molecular polarization and the
molecular bond breaks, an associated defective site is thought to be influence of field enhancement begin to play a leading role. The
formed and a conductive percolation path of defect grows with time. factor f loc is strong enough to overcome the impact of low

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-3

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 2. Geometry and mesh of electrostatic field model via the CST EM Studio:
(a) polymer sample and Bruce electrodes, (b) section view from y direction, and
(c) meshed spherical cavity.

compared with the calculation result in Fig. 1(a). During the elec-
trostatic field simulation, the physical parameters of dielectrics
(conductivity, inductivity, loss angle, etc.) are consistent except per-
mittivity. But the molecular polarization of the nanoscale is too
small to be simulated with current electromagnetic software.
Moreover, the influence of permittivity on the breakdown field is
analyzed experimentally in Sec. IV, which corresponds to the theo-
retical calculation in Fig. 1(b).
The simulation model is reconstructed in equal proportion
according to our improved system for the cavity experiment. The
geometry of the electrostatic field model is composed of two Bruce

FIG. 1. Comparison of cavity discharge and molecular polarization at different

relative dielectric constant: (a) field enhancement factor and (b) breakdown elec-
tric field.

breakdown threshold Egas within a large permittivity region. The

critical dielectric constant is supposed to be the order of hundred
which varies directly with the values of Egas and ΔH 0 = p0 . In Fig. 1(b),
the critical value is suggested as εc  200.

III. SIMULATION RESULTS

A. Electrostatic field model
The distribution and strength of the applied electric field are
decisive for the specific breakdown location and insulation life of
uniform dielectrics. As the cavity field is directly influenced by
material permittivity, the variation of the electrostatic field is simu-
FIG. 3. Electrostatic field distribution along the x axis and variation with five
lated with the model in Fig. 2. The distribution and strength of relative dielectric constants.
simulation result are presented in Fig. 3 and Table I, which

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-4

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

TABLE I. Field enhancement factor of cavity with different relative permittivities.

εr 1 2 3 4 5
f acav 1 1.2 1.28571 1.33333 1.36364
f bcav 1 1.18087 1.25664 1.29830 1.32465
f ccav 1 1.1998 1.2855 1.3330 1.3633
f dcav 1 1.1999 1.2855 1.3331 1.3634
a
Calculated by Eq. (7).
b
Simulated by CST EMS for electrostatic field model.
c
Simulated by Comsol for partial discharge model (AC).
d
Simulated by Comsol for partial discharge model (pulse).

electrodes and a polymer sample. While the electrode diameter

is about 42.8 mm, the electrode surface in Fig. 2(b) consists of a
straight line, a sinusoidal curve, and an arc. Because of the
special shape of electrode surface, the electric field is uniform in
the middle flat area with the absence of the cavity. In order to
simulate the impact of sample permittivity on the cavity field, a
spherical cavity is constructed and specially meshed at the
center of sample. Considering the ultrasonic inspection result
and the element mesh size, the diameter of the cavity is chosen
to be 200 μm. For the sample thickness of is 2 mm, the cavity is
only 1/10 of the sample along the z axis, which can hardly
change the overall electric field.
If the relative dielectric constant of sample equals 1, it is the
same with the gas filled cavity. The electric field is the highest within
the middle region from x ¼
7 mm to x ¼ 7 mm. Breakdown is
more likely to occur in this circular area and the field enhancement
factor of cavity f cav is 1. With the increase of permittivity εr from
1 to 5, the factor f cav raises synchronously from 1 to 1.3246. The
difference between the calculation and simulation is always within
3% as Table I shows. The field distribution around the cavity is sig-
nificantly distorted from x ¼
0:5 mm to x ¼ 0:5 mm, which may
promote the partial discharge of the cavity. Therefore, breakdown
occurs more easily in materials with larger dielectric constant as the
field enhancement of the cavity is more obvious.

B. Partial discharge model

The influence of relative dielectric constant εr on dielectric
strength Ebd is composed of two mechanisms: cavity discharge and
molecular polarization. Drawing on the research methods of Illias,7 FIG. 4. Simulation results of cavity discharge process for typical relative
partial discharge inside the cavity is simulated with the variation of dielectric constant as AC voltage is applied: (a) electric field against time, (b)
εr . The geometry of the sample and the cavity is consistent with discharge quantity against time, and (c) cavity temperature against time.
the previous electrostatic field model. The electric field, the dis-
charge quantity, and the temperature within the cavity are pre-
sented in Fig. 4 and Table II. The f
q
n information is plotted as
Fig. 5 shows, which is a three-dimensional pattern of PRPD ( phase Detailed parameters of the applied voltage are U 0 ¼ 40 kV,
resolved partial discharge) plot. However, no appropriate simula- f ¼ 50 Hz, T ¼ 20 ms, and cycle ¼ 10. All electric field parameters
tion method is found for the bond breakage by molecular polariza- of the 200 μm cavity (applied Ecav0 , real-time Ecav and polarized Es )
tion. Because the bond dimension exceeds the microscopic limit of increase steadily with the dielectric constant. While E0 (t) is the
the mesh grid in finite element codes. applied field, Ecav0 (t) is the cavity field in the absence of surface
Simulations of the sinusoid AC voltage and the sawtooth pulse charge, Ecav (t) is the instantaneous electric field in the cavity center.
voltage are carried out with five typical relative dielectric constants. Besides, Es (t) is the polarized electric field generated by free charge

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-5

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

TABLE II. Partial discharge of cavity with different relative permittivity.

AC (Sinusoid) Pulse (Sawtooth)

εr 1 2 3 4 5 1 2 3 4 5
Max Ecav0 (kV/mm) 20.000 23.997 25.710 26.661 27.266 19.938 23.923 25.631 26.580 27.184
Max Ecav (kV/mm) 22.799 25.580 27.964 26.661 27.266 19.937 23.922 25.041 26.577 27.181
Max Es (kV/mm) 19.381 23.193 25.075 25.962 26.525 17.707 23.219 23.332 23.823 25.810
Total Qcharge (pC) 44.445 87.997 176.90 225.38 347.99 9.6940 24.756 50.280 62.178 73.729
Mean Qmax (pC) 1.2012 2.9621 4.7474 6.6241 8.4273 0.8742 2.4170 3.6981 5.4043 6.5059
Mean Qmin (pC) 0.3260 0.7137 1.0402 1.3625 1.8975 0.5514 1.2333 2.2276 2.8836 3.7905
Max T cav ( C) 23.153 30.268 35.903 41.934 48.695 21.904 27.100 30.980 38.599 39.794
Mean T cav ( C) 20.031 20.069 20.118 20.158 20.199 20.009 20.025 20.059 20.075 20.094

which accumulated on the cavity surface. Since Einc is the initial elec-
tric field of gas in the cavity like Egas , the change is not obviously
compared with other parameters. In particular, the factor fcav
obtained by Ecav0 =E0 in the partial discharge model is identical to
the electrostatic field model as Table I shows. Meanwhile, the dis-
charge quantity parameters (total for 10 cycles Qcharge and mean for
per cycle Qmax , Qmin ) are also increased gradually. Besides, the max
and mean cavity temperature T cav shows a similar trend for both AC
and pulse voltage. As a result, cavity partial discharge becomes more
intense if the relative permittivity increases. While cavity is the main
source of material partial discharge, the dielectric is more likely to
breakdown when the dielectric constant gets larger.

IV. EXPERIMENTAL DATA

A. Ultrasonic inspection
In order to assess the relationship between cavity parameters
and insulation performance, the characterization of cavities is nec-
essary for dielectric. While cutting may destroy the cavity morphol-
ogy, translucency precludes the optical microscope. Besides, the
partial discharge test is insufficient to identify the cavity precise
location and size. The X-ray method and electron beam technique
are hardly to provide images with specific cross sections and
harmful to materials. In contrast, the ultrasonic microscope is com-
petent for the nondestructive inspection of dielectric, because cavi-
ties with large attenuation are sensitive to ultrasound.
Though the electric field of the cavity is increasing with the
dielectric constant of the material, cavities have little effect on the
intrinsic permittivity of the pure material. It is inconvenient and
unclear to study the effect of the dielectric constant on dielectric
strength indirectly by defect characterization. Because our theoretical,
simulation, and experimental analysis are all based on the assumption
of the cavity, it is a crucial problem whether the cavity exists in solid
dielectrics. Therefore, the ultrasonic testing is used as a visual means
to confirm the cavity distribution and quantity in this paper.
FIG. 5. Simulation results of cavity partial discharge for the typical relative Moreover, the influence of cavities on dielectric insulation is pre-
dielectric constant as the AC voltage is applied: (a) f
q
n plot at εr ¼ 1 and
(b) f
q
n plot at εr ¼ 5.
sented in another paper.22 As an electrochemical process, the
influence of cavities on breakdown is a gradual accumulation process

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-6

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

FIG. 6. Images of polyetherimide

samples with generous defects (identi-
fied by Scandium software): (a) sample
with a cavity ratio of 0.89% and (b)
sample with a cavity ratio of 1.73%.

of damage by cavity discharge. The damage caused by cavity dis- ΔH0

charge to materials is mainly in the electrical, thermal, and chemical Ebd ¼ [εr þ (L
1
1)]
1 ¼ A2 (εr þ B2 )
1 : (13)
p0 L
forms. This damage process is applicable to both large and small cav-
ities, which widely spread inside the practical dielectrics.
As Fig. 6 shows, cavities are widely spread inside the polyetheri- With the fitting result in Fig. 7(a), the parameters of the cavity
mide (PEI) samples. The specifications of samples are thin wafers with discharge and molecular polarization are calculated as na ¼ 0:8204,
a diameter of 40 mm and a thickness of 2 mm. While the samples are Egas ¼ 46:88 kV=mm, L ¼ 0:2460, and ΔH 0 = p0 ¼ 32:26 kV=mm.
scanned from the ultrasonic microscope of PVA SAM300E model, the Due to the particularity of the inflection point on Paschen’s curve,
output image resolution is 86:6 μm at 30 MHz. With the pulse-echo it is the most sensitive cavity size to the breakdown voltage. The
backscattered amplitude integral (BAI) mode imaging technique, minimal breakdown voltage in air at the standard atmospheric
cavities are indicated by the contrast of gray level and depth of the pressure is 327 V at a distance of 7:5 μm. The difference between
field. Through the image analysis software of Scandium, the cavities the corresponding breakdown field and the fitting result is only
are identified, marked, and measured subsequently. The cavity about 7%. As the mean free path λ is 0:5
2 nm and the bond
ratios of these two samples are 0.89% and 1.73%, which denote the strength ΔH 0 . 4 eV, ΔH 0 =eλ is at least 62 times higher than the
ratio of the total defect area to the sample area. Specifically, 83% of fitting value. Consequently, the cavity discharge model is more con-
all counted cavities are approximated to 200 μm in an average diam- sistent with the polymer bulk in electric accessory. This is mostly
eter and 4 pixels in the mean area. However, the cavities smaller due to the low relative permittivity (2
9) and large thickness
than the scanning resolution cannot be imaged or measured. As a (. 1 mm) of the polymer bulk. The enhancement of low dielectric
result, cavity discharge takes account of the micrometer level, while constant on the local electric field is poor, when cavity becomes the
molecular polarization is concerned with the nanoscale. main source of partial discharge inside the dielectric.
By repeating similar calculations, the parameters are determined
B. Quantitative fitting with Fig. 7(b) as na ¼ 0:9615, Egas ¼ 4608 kV=mm, L ¼ 0:1554,
and ΔH 0 = p0 ¼ 1710 kV=mm. The breakdown field at the knee
Considering that various dielectrics have been tested in history, point of Paschen’s curve is 43.6 kV/mm, it is two orders smaller
we prefer to use published data in the literature rather than to than the fitted Egas . For theoretical ΔH 0 = p0 of Pr2 O3 is about
repeat experiments independently. Based on the electrical parame- 2000 kV/mm, the difference is about 15% with the fitting result.
ters of 29 polymer bulks in Table III and 31 oxide films from Ref. When considered of SiO2 which is 3000 kV/mm, the fitted ΔH 0 = p0
23, cavity discharge and molecular polarization are fitted and com- is in the same order. Accordingly, the molecular polarization model
pared in Fig. 7. An approximate relation between the dielectric is more suitable for the oxide films of gate dielectric. This is mainly
strength and dielectric constant of Ebd
(εr )
1=2 is found by determined by the high relative permittivity (4
183) and very thin
McPherson.23 However, we suggest that the dielectric strength thickness (, 1 μm) of the oxide film. The local electric field at the
should not be fitted directly into a power function of the dielectric molecular scale is greatly enhanced by the high dielectric constant,
constant. Because the power of relative permittivity εr is fixed in while the cavities in the micrometer scale hardly exist in oxide films.
Eqs. (8) and (11), while the physical parameters of materials are var- Based on numerous experiments and mature theories, the
iables. Therefore, the dielectric strength and dielectric constant minimum breakdown voltage of air is a constant at standard atmo-
should be applied to the following fitting forms: spheric pressure. In this paper, the voltage is fitted and calculated
  in another independent way. The results obtained by these two
Ebd ¼ Egas na (n
1
1
a
1) þ εr ¼ A1 þ B1 ε
1
r , (12) irrelevant methods happen to be the same within a difference of

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-7

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

TABLE III. Dielectric constant and dielectric strength of some polymers from
Ref. 24.

εr Ebd
Material Type (106 Hz) (kV/mm)
Polyethylene Low density 2.3 22.8
Polypropylene Pure 2.3 30
Polymethyl Pure 2.95 19.65
methacrylate
Polymethyl α-Methylstyrene 3.03 18.7
methacrylate
Polymethyl ACS 3.01 21.7
methacrylate
Polyvinyl chloride Hard 2.325 26.5
Polystyrene General 2.525 23.6
Polycarbonate Pure 2.54 19.5
Polycarbonate 25% glass fiber 3.17 22
Polyformaldehyde Pure 3.56 18.6
Chlorinated Pure 3.1 18.3
polyether
Polyamide resin Nylon 1010 3.1 20
Polyamide resin Nylon 6 3.4 20
Polyamide resin Nylon 610 2.98 20
Polyamide resin MC-Nylon 3.7 19.1
Polyamide resin 40% glass fiber 3.1 23
Fluorinated resin Polytetrafluoroethylene 2.1 32.5
Fluorinated resin Polytrifluorochloroethylene 2.5 19.7
Phenolic resin Wood powder filling 9 12.95
Urea α-cellulose filling 8.25 13.75
formaldehyde
resin
Alkyd resin Inorganic filling 6.3 15.7
Alkyd resin Glass fiber filling 6.6 15.3
Diallyl phthalate Inorganic filling 5.2 16
Diallyl phthalate Glass fiber filling 4.45 16.6
Unsaturated Casted and hard 3.65 17.25
polyester FIG. 7. Variation of the dielectric strength on relative dielectric constant for
Unsaturated Glass fiber filling 4.85 16.65 different solid materials: (a) polymer bulks and (b) oxide films.
polyester
Unsaturated Glass cloth filling 4.8 16.65
polyester
Polyimide Glass fiber filling 3.8 19.6
Polyurethane Pure 5.75 15.7 While the correlations of three fitting methods are matched,
the coefficient of cavity discharge is 0.76 for polymer bulks. In
nearly two decades of dielectric constant, the fitting correlation
coefficients of oxide films are all over 0.83. Theoretical dielectric
strength is the intrinsic property of solid insulation materials. In
only 7%. This consistency demonstrates that the cavity is an impor- practice, the breakdown electric field is related to field, electrode,
tant factor of the breakdown process and dielectric strength for environment, etc. The insulation performance decreases with the
polymers. Because the cavity electric field is related to relative per- increase of sample size, working temperature, AC frequency, and
mittivity, the dielectric constant determines the dielectric strength ambient humidity. Therefore, the breakdown of the solid material
of polymers indirectly. Meanwhile, the practical ΔH 0 = p0 of oxide is such a complex physical phenomenon. There are numerous
films coincides with the fitted one at a difference of 15%. Since concepts, theories, and experimental studies. As cavity discharge
ΔH 0 = p0 is also obtained from two irrelevant ways, molecular polari- and molecular polarization play a part in dielectric strength, the
zation contributes to the dielectric strength of oxides. As the polariza- correlation coefficients are reasonable. Since the dielectric cons-
tion degree is determined with the dielectric constant, the dielectric tant stands out from many factors as Fig. 7 shows, it has an
strength of oxides is indirectly affected by relative permittivity. important influence on dielectric strength.

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-8

Published under license by AIP Publishing.
 Journal of ARTICLE scitation.org/journal/jap
Applied Physics

Additionally, the thickness is one of the key factors of solid 8

C. Forssén and H. Edin, “Partial discharges in a cavity at variable applied fre-
breakdown characteristics.25,26 The breakdown field generally quency part 2: Measurements and modeling,” IEEE Trans. Dielectr. Electric.
decreases with the increase of dielectric thickness. In contrast, cavity Insul. 15, 1610–1616 (2008).
9
K. C. Kao, Dielectric Phenomena in Solids: With Emphasis on Physical Concepts
discharge and molecular polarization only aim at breakdown and the
of Electronic Processes, 1st ed. (Elsevier Academic Press, London, 2004), Section
discharge threshold. The localized breakdown and discharge by these 8.2, p. 528.
two mechanisms are just the starting point of the whole process.27,28 10
E. Kuffel, W. S. Zaengl, and J. Kuffel, High Voltage Engineering: Fundamentals,
The residual charge from breakdown and discharge will undergo a 2nd ed. (Newnes, Oxford, 2000), Section 5.4, p. 324.
series of process in the material, such as migration, multiplication, 11
N. E. Smirnov and P. I. Pronin, “Dielectric and magnetic properties of contin-
and recombination.9,29 When the electrodes are totally connected by uum with dislocations,” Phys. Lett. A 382, 2245–2250 (2018). Special Issue in
the growing electrical branches, the dielectric completely loses its memory of Professor V.B. Braginsky.
insulation capacity. Hence, the effect of dielectric constant on dielec-
12
S. Coulombe and J. L. Meunier, “Thermo-field emission: A comparative
tric strength mainly determines the threshold of local breakdown, study,” J. Phys. D Appl. Phys. 30, 776 (1997).
not the development process of global breakdown.
13
A. Heeger, I. Parker, and Y. Yang, “Carrier injection into semiconducting
polymers: Fowler-Nordheim field-emission tunneling,” Synthetic Met. 67, 23–29
(1994).
14
L. S. Hung, C. W. Tang, and M. G. Mason, “Enhanced electron injection in
V. CONCLUSION organic electroluminescence devices using an Al/LiF electrode,” Appl. Phys. Lett.
On balance, the influence of dielectric constant on dielectric 70, 152–154 (1997).
strength does exist, as proved by the backward derivation of Egas ,
15
H. B. Wu, F. Huang, J. B. Peng, and Y. Cao, “High-efficiency electron injection
ΔH 0 = p0 , and the fitted correlation coefficients. However, the cathode of Au for polymer light-emitting devices,” Org. Electron. 6, 118–128
(2005).
influence is realized by cavity discharge and molecular polarization 16
P. Osmokrovic, I. Krivokapic, and S. Krstic, “Mechanism of electrical breakdown
in an indirect way. In addition, permittivity is only one of the many left of Paschen minimum,” IEEE Trans. Dielectr. Electr. Insul. 1, 77–81 (1994).
factors affecting dielectric strength, but it is the key to local break- 17
M. Klas, Š. Matejčik, B. Radjenović, and M. R. Radjenović, “The breakdown
down. A combination of discharge and polarization mechanisms voltage characteristics and the secondary electron production in direct current
coexists in a given solid dielectrics, while only one dominates typi- hydrogen discharges for the gaps ranging from 1 μm to 100 μm,” Phys. Lett. A
cally. Since cavity discharge is more important for polymer bulks of 376, 1048–1052 (2012).
low permittivity, molecular polarization plays a leading role in oxide 18
J. W. McPherson, “Polarity dependent thermochemical E-model for describing
films with high permittivity. As cavity discharge is mainly due to the time dependent dielectric breakdown in metal-oxide-semiconductor devices with
lower threshold of defect breakdown, molecular polarization leads to hyper-thin gate dielectrics,” J. Appl. Phys. 120, 104102 (2016).
19
J. W. McPherson, J. Y. Kim, A. Shanware, and H. C. Mogul, “Thermochemical
breakdown by enhancing the local electric field.
description of dielectric breakdown in high dielectric constant materials,” Appl.
Phys. Lett. 82, 2121–2123 (2003).
20
L. Zhao, “Research on breakdown characteristics of solid insulation materials
ACKNOWLEDGMENTS under nanosecond pulses,” Ph.D. thesis (Xi’an Jiaotong University, 2012).
This work was supported by the National Natural Science
21
L. Zhao, J. C. Su, Y. F. Pan, and X. B. Zhang, “The effect of polymer type on
Foundation of China (NNSFC) under Grant No. 51377135. electric breakdown strength on a nanosecond time scale,” Chin. Phys. B 21,
033102 (2012).
22
J. Liu, J. C. Su, R. Li, L. Zhao, Y. R. Lu, and X. L. Liu, “Characterization of cav-
ities in polyetherimide and the influence on insulation performance of ultra-high
REFERENCES
voltage pulse,” Mater. Lett. 240, 153–156 (2019).
1 23
L. Zhao, J. C. Su, X. B. Zhang, Y. F. Pan, L. M. Wang, X. Sun, and R. Li, Proposed Universal Relationship Between Dielectric Breakdown and Dielectric
“Research on reliability and lifetime of solid insulation structures in pulsed Constant, edited by J. W. McPherson, J. Y. Kim, A. Shanware, H. C. Mogul, and
power systems,” IEEE Trans. Plasma Sci. 41, 165–172 (2013). J. Rodriguez, Digest. International Electron Devices Meeting, IEEE (IEEE,
2
V. A. Zakrevskii, N. T. Sudar, A. Zaopo, and Y. A. Dubitsky, “Mechanism of San Francisco, 2002).
24
electrical degradation and breakdown of insulating polymers,” J. Appl. Phys. 93, M. B. Yang and Z. Y. Tang, China Materials Engineering Canon: Polymer
2135–2139 (2003). Materials Engineering (Part I), 1st ed. (Chemical Industry Press, Beijing, 2006),
3
G. S. Haase, E. T. Ogawa, and J. W. McPherson, “Reliability analysis method Section 4.8, p. 54.
for low-k interconnect dielectrics breakdown in integrated circuits,” J. Appl. 25
L. Zhao, G. Z. Liu, J. C. Su, Y. F. Pan, and X. B. Zhang, “Investigation of thick-
Phys. 98, 034503 (2005). ness effect on electric breakdown strength of polymers under nanosecond
4
H. C. Mogul and J. W. McPherson, “Electrical stability of a 65 nm technology pulses,” IEEE Trans. Plasma Sci. 39, 1613–1618 (2011).
Ni-silicided junctions under accelerated bias stressing,” Appl. Phys. Lett. 90, 26
J. S. Huang, S. Shian, R. M. Diebold, Z. G. Suo, and D. R. Clarke, “The thick-
162110 (2007). ness and stretch dependence of the electrical breakdown strength of an acrylic
5
J. W. McPherson, R. B. Khamankar, and A. Shanware, “Complementary model dielectric elastomer,” Appl. Phys. Lett. 101, 122905 (2012).
for intrinsic time-dependent dielectric breakdown in SiO2 dielectrics,” J. Appl. 27
Y. Sekii, “Initiation and growth of electrical trees in LDPE caused by impulse
Phys. 88, 5351–5359 (2000). voltage,” IEEE Trans. Dielectr. Electr. Insul. 5, 748–753 (1998).
6
J. W. McPherson, V. K. Reddy, and H. C. Mogul, “Field-enhanced Si-Si bond- 28
X. R. Chen, Y. Xu, X. L. Cao, S. J. Dodd, and L. A. Dissado, “Effect of tree
breakage mechanism for time-dependent dielectric breakdown in thin-film SiO2 channel conductivity on electrical tree shape and breakdown in XLPE cable insu-
dielectrics,” Appl. Phys. Lett. 71, 1101–1103 (1997). lation samples,” IEEE Trans. Dielectr. Electr. Insul. 18, 847–859 (2011).
7
H. A. Illias, G. Chen, and P. L. Lewin, “Modeling of partial discharge activity in 29
A. V. Gurevich, Y. V. Medvedev, and K. P. Zybin, “Thermal electrons and elec-
spherical cavities within a dielectric material,” IEEE Electr. Insul. Mag. 27, 38–45 tric current generated by runaway breakdown effect,” Phys. Lett. A 321, 179–184
(2011). (2004).

J. Appl. Phys. 125, 115103 (2019); doi: 10.1063/1.5086377 125, 115103-9

Published under license by AIP Publishing.

View publication stats