Enhancing electrical energy storage using polar polyimides with nitrile groups directly attached to the main chain

Article abstract of DOI:10.1039/c4ta03260h

A set of 12 new polyimides (PIs) with one or three polar CN dipoles directly attached to the aromatic diamine part were synthesized and their electric energy storage properties were studied using broadband dielectric spectroscopy (BDS) and electric displacement-electric field (D-E) loop measurements to determine their potential for high temperature film capacitors for aerospace applications. It was found that adding highly polar nitrile groups to the PI structure increased permittivity and thus electrical energy storage, especially at high temperatures, and 3 CN dipoles were better than 1 CN dipole. Below the glass transition temperature (T_g), a weak γ transition was observed around -100°C and a broad β transition was observed between 100 and 150°C. It was the β (i.e., precursor dipolar motion before long-range segmental motion, or glass transition), rather than the γ sub-T_g transition that substantially increased the permittivity of PIs. From the BDS results on PIs having 3 nitrile groups, the enhancement in permittivity from permanent dipoles decreased with dianhydride in the order of pyromellitic dianhydride (PMDA) > 4,4′-oxydiphthalic dianhydride (OPDA) > 1,1,1,3,3,3-hexafluoropropane dianhydride (6FDA) > 4,4′-benzophenonetetracarboxylic dianhydride (BTDA). Meanwhile, the increase in permittivity also decreased in the order of para-para, meta-para, and meta-meta linkage in the diamine, suggesting that the para-para linkage favored easier dipole rotation than the meta-meta linkage. From the D-E loop study, the PIs with a combination of PMDA dianhydride and a para-para linkage exhibited the highest discharged energy density and a reasonably low loss.

Full text of DOI:10.1039/c4ta03260h

Journal of
Materials Chemistry A
View Article Online
View Journal | View Issue
PAPER
Enhancing electrical energy storage using polar
polyimides with nitrile groups directly attached to
the main chain†
Cite this: J. Mater. Chem. A, 2014, 2,
20683
b
a
Imre Treufeld,a David H. Wang,bc Brian A. Kurish,bc Loon-Seng Tan and Lei Zhu
*
*
A set of 12 new polyimides (PIs) with one or three polar CN dipoles directly attached to the aromatic diamine part
were synthesized and their electric energy storage properties were studied using broadband dielectric
spectroscopy (BDS) and electric displacement–electric field (D–E) loop measurements to determine their
potential for high temperature film capacitors for aerospace applications. It was found that adding highly polar
nitrile groups to the PI structure increased permittivity and thus electrical energy storage, especially at high
temperatures, and 3 CN dipoles were better than 1 CN dipole. Below the glass transition temperature (Tg), a
weak g transition was observed around ꢀ100 ꢁC and a broad b transition was observed between 100 and 150
ꢁC. It was the b (i.e., precursor dipolar motion before long-range segmental motion, or glass transition), rather
than the g sub-Tg transition that substantially increased the permittivity of PIs. From the BDS results on PIs
having 3 nitrile groups, the enhancement in permittivity from permanent dipoles decreased with dianhydride
in the order of pyromellitic dianhydride (PMDA) > 4,40-oxydiphthalic dianhydride (OPDA) > 1,1,1,3,3,3-
hexafluoropropane dianhydride (6FDA) > 4,40-benzophenonetetracarboxylic dianhydride (BTDA). Meanwhile,
the increase in permittivity also decreased in the order of para–para, meta–para, and meta–meta linkage in
the diamine, suggesting that the para–para linkage favored easier dipole rotation than the meta–meta linkage.
From the D–E loop study, the PIs with a combination of PMDA dianhydride and a para–para linkage exhibited
the highest discharged energy density and a reasonably low loss.
Received 26th June 2014
Accepted 9th October 2014
DOI: 10.1039/c4ta03260h
www.rsc.org/MaterialsA
behavior showing a narrow hysteresis loop.15 As a consequence,
a high dielectric constant in the range of 30–70 can be obtained
1 Introduction
High dielectric constant (>5–10), high temperature (>150 ꢁC),
and low loss polymers are attractive dielectric materials for a
variety of practical applications1,2 such as lm capacitors for
power-conditioning,3–5 power electronics in hybrid electric
vehicles,6–8 pulsed power,9–11 and gate dielectrics for eld-effect
transistors.12,13 There have been several strategies to achieve this
goal. First, a signicant amount of research effort has been
dedicated to converting high dielectric constant ferroelectric
polymers such as poly(vinylidene uoride) (PVDF) and its
random copolymers into more or less linear dielectrics.14,15 It is
observed that decreasing the ferroelectric domain size to the
nanoscale (i.e., nanodomains) by repeat-unit crystal isomor-
phism can effectively achieve the so-called relaxor ferroelectric
with reasonably low dielectric loss. However, repeat-unit crystal
isomorphism introduces defects in the crystal lattice and tends
to decrease the melting temperature (Tm). For example, relaxor
ferroelectric
P(VDF-co-triuoroethylene-co-1,1-chlorouoro-
ethylene) [P(VDF-TrFE-CFE)] and P(VDF-co-triuoroethylene-co-
chlorotriuoroethylene) [P(VDF-TrFE-CTEF)] random terpoly-
mersꢁhave a Tm around 125 ꢁC, compared with the Tms of PVDF
(175 C) and P(VDF-TrFE) (ca. 155 C).15–18 Furthermore, nano-
ꢁ
sized ferroelectric domains decrease the ferroelectric-to-para-
electric (i.e., Curie) transition temperature (TC) to around room
temperature for both random terpolymers. Above room
temperature, the terpolymers are in the paraelectric phase,
where random dipoles highly interact with one another without
any existence of ferroelectric domains. In addition to the
decreased dipolar polarization for the high temperature para-
electric phase of P(VDF-TrFE), high crystalline and amorphous
dipole mobilities promote ionic and electronic mobility in the
material, resulting in high dielectric losses.15,19 Therefore, the
challenge lies in the effort of achieving high temperature relaxor
ferroelectric crystalline polymers.
aDepartment of Macromolecular Science and Engineering and Department of
Chemistry, Case Western Reserve University, Cleveland, Ohio 44106-7202, USA.
E-mail: lxz121@case.edu
bAir Force Research Laboratory, Materials and Manufacturing Directorate,
Nanostructured and Biological Materials Branch, Wright-Patterson Air Force Base,
OH 45433, USA. E-mail: Loon.Tan@us.af.mil
cUES, Inc., 4401 Dayton-Xenia Road, Dayton, OH 45432, USA
The second strategy is to add permanent dipoles into a high
Tg amorphous polymer matrix while avoiding the formation of
ferroelectric domains. This is similar to dipolar glasses in
† Electronic supplementary information (ESI) available: Syntheses of monomers
and polymers, reection and transmission X-ray diffraction proles, details of
calculation, and BDS results for sample 2a. See DOI: 10.1039/c4ta03260h
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20683

View Article Online
Journal of Materials Chemistry A
Paper
ceramic materials,20,21 and analogous to spin glasses in hexauoropropane dianhydride (6FDA), 4,40-oxydiphthalic dia-
magnetic materials.22,23 The idea is to utilize the free volume in nhydride (OPDA), 4,40-benzophenonetetracarboxylic dianhy-
high Tg polymers to allow relatively free rotation of individual dride (BTDA), and pyromellitic dianhydride (PMDA) was
dipoles (i.e., sub-Tg transitions24) in order to enhance the conducted in N,N-dimethylacetamide (DMAc) at room temper-
dielectric constant while keeping dielectric loss low. Early work ature for 24 h to afford the corresponding poly(amic acid)s
utilized nitrile-containing styrenic and acrylic amorphous (PAA), which were subsequently imidized by stepwise thermal
polymers to achieve relatively high dielectric constant.25,26 For treatment up to 300 ꢁC to afford tough, creasable PI lms (20–40
example, poly(4-vinylbenzylcyanide) was reported to exhibit a mm thickness).
dielectric constant as high as 7.0 at 100 kHz. Recently, polar
Due to their polar nature, PIs tend to absorb a small amount
F–F27 and nitrile28 groups were added into bisphenol A of moisture, which may signicantly affect their dielectric
polycarbonate (PC) in order to increase its dielectric constant properties.38 Therefore, all PI lms were dried in a vacuum oven
(3r ¼ 2.95). For example, a uorinated tetraaryl bisphenol A PC, at 130 ꢁC for 24 h to remove water before any dielectric property
DiF p-TABPA-PC, exhibited a dielectric constant of 3.3,27 and a measurements. Aer drying, the lm samples were coated with
nitrile-modied bisphenol A PC exhibited a dielectric constant 100 nm silver electrodes on both sides by physical vapor
of 4.0 at room temperature.28
deposition (EvoVac Deposition System, Angstrom Engineering
Stimulated by the above promising work, we here focus on Inc., Kitchener, Ontario, Canada) to improve electrical contact.
high performance polyimides (PIs), which are attractive mate- The electrode areas were 0.785 cm2 for the BDS study and 0.0515
rials for use in highly aggressive environments because of their cm2 for the D–E loop study. The samples were then stored in a
high Tg, mechanical toughness, and resistance to solvents, desiccator lled with Drierite until dielectric measurements
radiation, heat and oxidation.29 These properties make PI lms were performed.
highly suitable for use in high temperature applications such as
aerospace power conditioning, electronics industry, and elec-
trical insulation materials. Previously, Kakimoto et al. reported
that attaching nitrile pendants into PIs would enhance their
dielectric constants,30 and two PIs from an unsymmetrical
diamine with nitrile groups showed poor solubility in aprotic
solvents and only became soluble aer the nitrile groups were
hydrolyzed.31 Their dielectric properties, however, were not
reported. Hybrid lms based on nitrile-containing PIs/inor-
ganic particles (pyrite ash, barium and titanium oxides) were
prepared and their nano-actuation was investigated.32,33 PIs
containing one and two nitrile groups per repeat unit were
synthesized and their piezoelectric behavior was analyzed. The
polymer containing two nitrile groups per repeat unit showed
higher remnant polarization than the ones containing only one
nitrile group per repeat unit due to its increase in polarity.
However, their values were low as compared with commercial
piezoelectric polymers such as PVDF.34
2.2. BDS measurements
Low-voltage BDS measurements were carried out using a
Novocontrol Concept 80 broadband dielectric spectrometer
(Montabaur, Germany). For temperature scans, temperature
was programmed to change linearly from ꢀ150 ꢁC to 190 ꢁC at a
rate of 2 ꢁC minꢀ1. A sine voltage of 1.0 Vrms (root-mean square
voltage) with frequency ranging from 107 Hz to 1 Hz was applied
across the sample lm during temperature ramping, and data
were recorded simultaneously. For frequency scans, samples
were held at a constant temperature and the same sine voltage
as above was applied with frequency ranging from 10ꢀ3 Hz to
107 Hz. High-voltage (HV) BDS experiments were carried out in
the same way, except that the frequency was scanned from 104
Hz to 1 Hz and the peak amplitude of the applied sinusoidal
electric eld was 10 and 50 MV mꢀ1
.
In this work, highly polar nitrile groups are directly attached
to the aromatic diamine part of various PI structures. In total, 12
new PIs, containing 1 or 3 CN dipoles, are synthesized. Their
2.3. D–E hysteresis loop measurements
dielectric properties and loss mechanisms are studied by D–E loop measurements were performed at 23, 100, and 190 ꢁC,
broadband dielectric spectroscopy (BDS) and electric displace- using a Premier II ferroelectric tester (Radiant Technologies,
ment–electric eld (D–E) loop measurements. Results show that Inc., Albuquerque, New Mexico, USA). The temperature was
3 CN dipoles are more effective than 1 CN dipole (i.e., number controlled by using a clean silicone oil bath equipped with an
density effect) in enhancing the dielectric constant and thus IKA RCT temperature controller (IKA Works, Inc., Wilmington,
electrical energy storage of PIs.
North Carolina, USA). Two consecutive sine-wave voltages with
equal amplitudes were applied to the lm sample and the
second loop was taken as the result to eliminate the transient
effect. The frequency of each wave was xed at 10 or 1000 Hz.
Stray capacitance (i.e., background capacitance from the xture
and high voltage cables) as a function of lm thickness was
2 Experimental
2.1. PI synthesis and lm preparation for dielectric
characterization
The polyimide lms were fabricated according to previous determined by measuring the electric displacement, D, of
publications.35–37 An example of PI synthesis is described in biaxially oriented polypropylene (BOPP, SB Electronics, Inc.,
Section I of the ESI.† The condensation polymerization of the Barre, Vermont, USA) lms with different thicknesses (8–55 mm)
diamines containing one or three nitrile dipoles with the at 4000 V. The stray capacitance was subtracted from the raw
dianhydrides being 2,20-bis(3,4-dicarboxyphenyl)-1,1,1,3,3,3- data of the lm sample according to its thickness.
20684 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
because the permittivity changes little (only 0.05 at 1 kHz) from
ꢀ150 ꢁC to 190 ꢁC, compared to samples 2b and 4c which
contain added nitrile groups. If we assume zero dipolar
contribution to permittivity, the difference between the
measured and calculated permittivity, namely 0.49, should arise
from vibrational polarization for sample 1, according to the
Clausius–Mossotti relations:39
2.4. Density measurement
Several pieces of lm samples were placed in a beaker lled with
pure water and some surfactant to make sure that they sank to
the bottom. Aer 5 min of ultrasonication to remove surface
bubbles, sodium iodide was gradually added until the lm
samples oated (or suspended) in the solution. Then, 10 mL of
the solution was sampled using a pipette and weighed to
determine its density. Finally, this density was taken as the
sample density.
3r ꢀ 1 MRU NA
¼ ðae þ avÞ
330
(2)
3r þ 2
r
where 3r has contributions from both ae and vibrational polar-
izability (av). In other words, the magnitude of vibrational
polarization is about 20% of that of the electronic polarization
for sample 1. This percentage is similar to what has been
reported for polymers in general.39,41 Since the structures of
samples 2b and 4c are similar to sample 1 except the nitrile
groups, we assume that the percentage of vibrational polariza-
tion relative to electronic polarization is also 20% for them.
Based on bond refractions,39,40 the molar refraction for sample
2b is calculated to be 152.2 cm3 molꢀ1. Taking into account the
20% vibrational polarization, the permittivity 3r from electronic
and vibrational polarizations can be calculated to be 3.15.
Compared with the experimental 3r0 of 3.30, the contribution
from dipolar polarization must be 0.15 for sample 2b at ꢀ150
ꢁC. Following the same calculation method, sample 4c has a
molar refraction of 212.9 cm3 molꢀ1 and a contribution to
permittivity of 0.50 from dipolar polarization. This makes sense
because 4c has 3 CN dipoles per repeat unit as opposed to 1 CN
dipole for sample 2b, and a contribution of 0.50 from dipole
motions for 4c is approximately 3 times higher than that of 0.15
for sample 2b. We can therefore conclude that the permittivity
increase at ꢀ150 ꢁC from sample 1 to 2b, and nally to 4c
originates from dipolar polarization.
3 Results and discussion
Totally, thirteen PIs were synthesized and their dielectric
properties were measured. A summary of the chemical struc-
tures and the measurement results is presented in Table 1.
3.1. Effect of the number of –CN groups in the diamine part
of PI
To investigate the effect of the number of nitrile groups in the
diamine part of the PIs on dielectric properties, we selected
three PIs from the list in Table 1. They are 1, 2b, and 4c, which
contain 0, 1, and 3 nitrile groups, respectively. Due to limitation
of available PIs, the dianhydride part in sample 1 is different
from those in samples 2b and 4c. Fig. 1 shows the real (3r0) and
imaginary (30r0) parts of relative permittivity and the dissipation
factor (tan d) as a function of temperature for samples 1, 2b, and
4c. At ꢀ150 ꢁC (see Fig. 1A, D, and G), 3r0 increased as the
number of nitrile groups increased, from 2.93 for 1 to 3.30 for
2b, and to 3.60 for 4c. We shall now try to determine the origins
of this increase in 3r0. Even though it might seem as if the
increase in 3r0 arises exclusively from the added nitrile groups,
this has to be veried because 1, 2b and 4c all have slightly
ꢁ
Upon increasing temperature from ꢀ150 ꢁC to 190 C, both
ꢁ
different structures. At such a low temperature (ꢀ150 C), the
3r0 and 30r0 increased for all three samples. These increases could
be attributed to the enhanced dipole motion as temperature
increased. Because the glass transition temperatures (Tgs) for
these samples are above 200 ꢁC (see Table 1), contribution from
impurity ion migration to this increase in permittivity might be
ignored for frequencies above 10 Hz. From the 30r0 plots for CN-
containing samples (Fig. 1E and H), weak g transition peaks
were noticed at around ꢀ85 ꢁC at 1 Hz. Similar relaxation peaks
were also seen for other CN-containing samples (see Fig. 4, 6,
and 8 later). These peaks could arise from absorbed water in the
samples; although our samples had been dried in a vacuum
oven at 130 ꢁC for at least 24 h before testing, it was fairly
difficult to completely eliminate trace amounts of absorbed
moisture because all samples contained polar –CN groups. A
similar phenomenon was also reported for Kapton, namely, a
trace amount of absorbed moisture produced a relaxation peak
around ꢀ100 ꢁC.38 For low frequency 30r0 and tan d plots (Fig. 1E/F
and H/I), another relatively broad relaxation peak was observed
around 150–180 ꢁC for samples 2b and 4c (and all other
CN-containing samples; see Fig. 4, 6, and 8 later). This peak
could be assigned to the b relaxation, which is oen due to the
onset of dipole motions that are precursors to the long-range
segmental motions in the a relaxation (or glass transition). To
contribution from impurity ion migration to the relative
permittivity could be ignored. Therefore, 3r0 itself should only
originate from electronic, atomic (or vibrational), and dipolar
polarizations. It is possible to accurately (usually within 1%
error) predict the electronic component of the permittivity, 3r,e
using the Lorentz–Lorenz relationship that relates molar
refraction RM to permittivity:39
,
3r;e ꢀ 1 MRU NA
RM
¼
¼
330
ae
(1)
3r;e þ 2
r
where MRU is the molar mass of the repeat unit, r is the bulk
density of the PI, NA is the Avogadro constant, 30 is the vacuum
permittivity, and ae is the electronic polarizability. In this case,
the molar refraction is that of the repeat unit of the PI. Molar
refraction is obtained by adding up molar refractions of indi-
vidual bonds (i.e., bond refractions39,40) in the repeat unit. The
molar refraction for sample 1 is calculated to be 158.5 cm3
molꢀ1 and the permittivity obtained for sample 1 from elec-
tronic polarization is 2.44 (see Section III of the ESI† for
calculations). This calculated value, however, is somewhat lower
than the experimental value of 2.93, because it does not include
vibrational and dipolar polarizations. From Fig. 1A, we can see
that the contribution from dipolar polarization is insignicant,
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20685

View Article Online
Journal of Materials Chemistry A
Paper
Table 1 Summary of measurement results of dielectric properties of the PIs
b
c
d
Tg Density RU density mRU Predicted Pdip D3r0
Exp. Pdip Pdip
Group Sample
1
Statea (ꢁC) (g cmꢀ3
)
(ꢂ1027 mꢀ3
)
(D)
(mC mꢀ2
)
(BDS) (mC mꢀ2
)
ratio
Am
199 1.433
1.23
0.28 0.0056
0.054
—
—
Am
Am
255 1.367
240 1.367
1.39
1.39
5.32 2.27
5.32 2.27
0.635 0.36
0.413 0.32
0.16
0.14
2
Cryst
(high)
341 1.357
305 1.367
1.11
1.12
17.1 17.9
17.1 18.0
1.372 0.87
0.644 0.85
0.048
0.047
3
Cryst
(high)
Cryst
(low)
235 1.342
0.98
16.0 13.9
1.139 0.83
0.060
Cryst
(low)
4
229 1.342
216 1.342
0.98
0.98
16.0 13.9
16.0 13.2
0.850 0.46
0.761 0.12
0.033
0.008
Am
Am
244 1.388
0.87
15.1 11.1
0.850 0.60
0.054
5
Am
Am
232 1.400
226 1.400
0.88
0.88
15.1 11.2
15.1 11.2
0.921 0.64
0.486 0.49
0.058
0.044
6
Am
238 1.357
0.98
14.1 10.9
0.583 0.49
0.045
20686 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
Table 1 (Contd.)
b
c
d
Tg Density RU density mRU Predicted Pdip D3r0
Exp. Pdip Pdip
Group Sample
Statea (ꢁC) (g cmꢀ3
)
(ꢂ1027 mꢀ3
)
(D)
(mC mꢀ2
)
(BDS) (mC mꢀ2
)
ratio
Am
218 1.344
0.97
14.1 10.8
0.427 0.25
0.023
a
The state of PI lms was determined by X-ray diffraction (XRD) and comparing the transmission- and reection-mode XRD spectra, which are
b
shown in Section II of the ESI. The dipole moment of the repeat unit (mRU) is calculated by the absolute value of diamine dipole moment
minus dianhydride dipole moment. The dipole moments of diphenyl ether, 1,1,1,3,3,3-hexauoropropane, benzophenone, and benzonitrile are
c
1.14, 2.0, 2.96, and 4.18 D, respectively. D3r0 is the difference between the real part of relative permittivity at 190 ꢁC and 1 kHz and that at
ꢀ150 ꢁC and 1 kHz. Since at ꢀ150 ꢁC only electronic and atomic polarizations contribute to permittivity while at 190 ꢁC dipole vibration also
contributes, the difference of relative permittivity between the two temperatures can be attributed solely to the dipolar polarization.
d
Experimental polarization only due to dipole orientation. This is calcꢀu1lated as: D–E loop polarization at 190 ꢁC and 1 kHz minus polarization
from BDS at ꢀ150 ꢁC and 100 kHz, extrapolated to a eld of 100 MV m
.
avoid signicant deformation of samples, the upper tempera- D3r0 ¼ 3r0 (190 ꢁC) – 3r0 (ꢀ150 ꢁC). Since electronic and vibrational
polarizations did not change much with temperature,39,42 D3r0
ꢁ
ture limit was chosen as 190 C.
0
ꢁ
ꢁ
Fig. 2A compares the increase of 3r from ꢀ150 C to 190 C could be attributed to the change in dipolar polarization only.
for samples 1, 2b, and 4c at different frequencies, i.e., Clearly, adding more CN dipoles into the PI structure increased
Fig. 1 3r0 (A, D and G), 30r0 (B, E and H), and tan d (C, F and I) as a function of temperature for samples 1, 2b, and 4c, which contain 0, 1, and 3 –CN
groups, respectively.
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20687

View Article Online
Journal of Materials Chemistry A
Paper
from dipolar polarization (Table 1), D3r0 indeed decreases in the
order of 3a > 4a > 5a > 6a. Therefore, it is reasonable to assume
that the diamine and dianhydride parts are antiparallel in the
samples.
Presently, there is no explicit theoretical prediction of
dipolar polarization for main-chain polymers containing
permanent dipoles, and here we resort to a freely rotating dipole
model. Assuming the net dipole moment of each repeat unit,
mRU, can freely rotate as a single dipole, the predicted polari-
zation from dipole orientation (Pdip) is:39
Fig. 2 (A) Increase of the real part of permittivity, D3r0, and (B) tan d at
190 ꢁC at 103, 104, and 105 Hz for samples 1, 2b, and 4c.
ꢀ
ꢁ
1
mRUE
Pdip ¼ NmRU coth u ꢀ
; here u ¼
(3)
D3r0. For example, sample 1 did not contain any CN dipoles and
its D3r0 was only about 0.05 at 1 kHz. Sample 2b contained 1 CN
dipole and its D3r0 increased to 0.41 at 1 kHz. Sample 4c con-
tained 3 CN dipoles and its D3r0 increased to 0.76 at 1 kHz.
The effect of added CN dipoles could be better understood by
predicting the polarization from permanent dipoles in the
sample. Assuming each repeat unit adopts an anti-parallel
conguration, i.e., dianhydride and diamine dipoles orient in
the opposite directions (see the chemical structures in Table 1;
the dianhydride dipole moment is pointing down and the
diamine dipole moment is pointing up), without external elec-
tric eld, the net dipole moment of each repeat unit (mRU) can be
calculated by subtracting the dianhydride dipole moment from
the diamine dipole moment. According to the literature, dipole
moments of diphenyl ether,43 1,1,1,3,3,3-hexauoropropane,44
benzophenone,45 and benzonitrile46 are 1.14, 2.0, 2.96, and 4.18
Debye (D), respectively. The dianhydride dipole moments
increase in the order PMDA (0 D) < BTDA (14.14 D) < 6FDA
(15.10 D) < OPDA (15.96 D). Samples 3a, 4a, 5a, and 6a contain
these dianhydrides in this exact order (see results later), and
they all have identical diamine parts, meaning that their total
dipole moments would decrease from 3a–6a as the dianhydride
dipole moment increases. From the permittivity increase D3r0
u
kBT
where N is the repeat unit density, mRU the dipole moment of a
repeat unit, E the applied electric eld, kB the Boltzmann
constant, and T the absolute temperature. Supposedly, E should
be the local electric eld for the freely rotating dipole. However,
there have been debates on the prediction of local eld for polar
materials with freely rotating permanent dipoles.47 It is
considered that the local eld should be inappreciably different
from the external electric eld. Therefore, we use the applied
electric eld here. On the basis of these assumptions, the pre-
dicted polarizations from freely rotational permanent dipoles
were calculated to be 0.0056, 2.27, and 13.9 mC mꢀ2 for samples
1, 2b, and 4c, respectively (see Table 1). Obviously, adding more
CN dipoles into PI samples increased the dipolar polarization,
and thus D3r0 followed exactly the trend of dipolar polarization
for samples 1, 2b, and 4c. Nonetheless, as the permittivity
increased, the dielectric loss (tan d) at 103 to 105 Hz also
increased proportionally (Fig. 2B). For polar polymers, loss
mechanisms include dipole relaxation loss, impurity-ion
migration loss, and electronic (or DC) conduction loss. Below,
we will discuss the loss mechanisms in these PI samples.
Fig. 3 Frequency-scan 3r0 and 3r00 plots for samples 1, 2b, and 4c at ꢀ150 ꢁC (A and D), 25 ꢁC (B and E), and 190 ꢁC (C and F), respectively.
20688 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
Fig. 3 shows frequency-scan plots of 3r0 and 3r00 for samples 1, dianhydride parts for groups 3–6 are PMDA, OPDA, 6FDA,
2b, and 4c at ꢀ150 ꢁC, 25 ꢁC, and 190 ꢁC. At ꢀ150 ꢁC (Fig. 3A and and BTDA, respectively. The diamine (DAm) part contains a
D), 3r0 increased with added –CN groups and remained nearly para–para (pp), meta–para (mp), or a meta–meta (mm) linkage
independent of frequency for all samples. Meanwhile, 30r0 stayed to the dianhydride part. Below, we will compare the dielectric
relatively low, between 10ꢀ3 and 10ꢀ2 (i.e., tan d between 3 ꢂ properties of PI samples having the para–para linkage in the
10ꢀ4 and 3 ꢂ 10ꢀ3). However, these tan d values were still higher diamine part.
than the dissipation factor for BOPP, which is around 1 ꢂ 10ꢀ5
Fig. 4 shows temperature-scan BDS spectra for samples
at ꢀ150 ꢁC.48 At such low temperatures, both impurity-ion 3a, 4a, 5a, and 6a. For these samples, the 3r0 slightly uctu-
migration loss and electronic conduction loss could be negli- ated around 3.5 at ꢀ150 ꢁC, likely due to different responses
gible. Generally speaking, no dipole ipping should occur from different dianhydride parts in the samples. Upon
below the lowest sub-Tg transition temperature. However in this raising the temperature from ꢀ150 ꢁC to 190 ꢁC, 3r0 showed a
case, molecular motion could not be completely frozen at ꢀ150 signicant increase above ca. 50 ꢁC. However, the amounts of
ꢁC and it was still possible for dipoles to wiggle even at such a increase, D3r0, at 103, 104, and 105 Hz are different for these
low temperature, causing higher loss than in nonpolar, linear samples (see Fig. 5A). Basically, D3r0 decreased consistently
dielectric polymers such as BOPP.
from 3a to 6a, which again could be attributed to the
Regarding sample 1 which contained no –CN dipoles, decrease of the predicted Pdip (see Table 1), which in turn
when the temperature increased from ꢀ150 ꢁC to 25 ꢁC and to could be associated with the rigidity of the polymer chain.
190 ꢁC, there was no obvious increase in 3r0 at low frequencies For example, Pdip decreased from 17.9 mC mꢀ2 for 3a to 13.9
(<1 Hz; see Fig. 3A–C). This differed from samples 2b and 4c mC mꢀ2 for 4a, to 11.1 mC mꢀ2 for 5a, and nally to 10.9 mC
0
that contained CN dipoles. At 25 C, 3r for samples 2b and 4c mꢀ2 for 6a.
ꢁ
showed a slight increase in the low frequency region (<1 Hz),
For samples 3a through 6a, g transitions were observed
while for all samples, 30r0 continuously increased with around ꢀ100 ꢁC in 30r0 and tan d plots in Fig. 4. This could again
decreasing frequency, going from 0.002 to 0.03 and the loss be attributed to either dipolar group relaxation in the polymer
mechanism was again primarily attributed to dipole motion. or to the absorbed moisture. Meanwhile, a broad b relaxation
00
0
As the temperature increased to 190 ꢁC, 3r for samples 2b and shoulder peak was observed in 3r at 1 Hz around 100–150 C,
ꢁ
4c had a step-wise decrease and 30r0 showed a weak and broad and with increasing the frequency it moved to higher
dipole relaxation peak around 103 Hz with an upturn at temperatures.
frequencies below 1 Hz. Since the low frequency upturn has a
slope close to ꢀ1, we attribute it to the impurity-ion migration 3a–6a is shown in Fig. 5B. As we can see, a higher tan d value in
loss.49,50 Fig. 5B also coincided with a higher D3r0 value in Fig. 5A. From
A comparison of tan d values at 103 to 105 Hz for samples
From the above results, we conclude that adding CN dipoles the above discussion, we understood that larger dipole motions
to the PI main chain primarily gives rise to two effects. First, it resulted in a higher D3r0, which in turn also resulted in a higher
introduces dipolar motion (e.g. wiggling) into the PI sample. tan d, because the CN dipoles were directly attached to the PI
Because the –CN groups are directly attached to the main chain main chain.
in a 90ꢁ conguration, this motion seems to be hindered and
will cause signicant friction with randomly packed neigh-
boring chains. As a result, together with an increase in 3r0, there
is also an increase in 30r0 due to the dipoles trying to rotate the
main chain. This explains the trend in Fig. 2 that both 3r0 and 3r00
increase with adding –CN groups into the PI structure. Second,
by adding –CN groups, the PI samples become more polar and
are easily contaminated with impurity ions with enhanced ion
mobility. At high enough temperature and low enough
frequency, impurity-ion migration loss becomes signicant (see
Fig. 3F). For the low-eld BDS spectra in Fig. 1 and 3, the
applied voltage is so low (1 Vrms) that electronic conduction may
be ignored.51
3.3. PI samples with a meta–para linkage in the diamine
Fig. 6 shows the temperature-scan dielectric spectra of 3r0, 3r00, and
tan d for samples 4b and 5b with a meta–para linkage in the
diamine. Similar results were obtained compared with the
above PI samples with a para–para linkage in the diamine. At
ꢀ150 ꢁC, 3r0 was about 3.63 and 3.25 for samples 4b and 5b,
0
ꢁ
respectively. With increasing the temperature up to 190 C, 3r
increased continuously. In the 30r0 and tan d plots, the g relaxa-
ꢁ
tion peak was seen at ꢀ100 C and the b shoulder was seen at
100–150 ꢁC at a frequency of 1 Hz. With increasing the
frequency, both relaxation peaks shied to higher tempera-
tures. Both D3r0 and tan d were similar for 4b and 5b, as shown in
Fig. 7.
3.2. PI samples with a para–para linkage in the diamine
3.4. PI samples with a meta–meta linkage in the diamine
From the above dielectric property measurements, we
learned that PI samples with 3 nitrile groups showed high Fig. 8 shows the temperature-scan dielectric spectra of 3r0, 3r00,
permittivity due to dipolar polarization. Below, we will focus and tan d for samples 3b, 4c, 5c, and 6b with a meta–meta
on the PI samples containing 3 nitrile groups (see Table 1). linkage in the diamine. Again, similar results were obtained
These samples are divided into four groups based on the compared with PI samples with para–para and meta–para
dianhydride part of the PI repeat unit, so each group only linkages in the diamine. The 3r0 values at ꢀ150 ꢁC were around
contains PIs with the same dianhydride part. The 3.5 for these samples, except for sample 5c, whose 3r0 was ca.
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20689

View Article Online
Journal of Materials Chemistry A
Paper
Fig. 4 Temperature-scan plots of 3r0, 3r00, and tan d for samples 3a, 4a, 5a, and 6a with a para–para linkage in the diamine part.
ꢁ
seen around 100–150 C for the 1 Hz curve. With increasing
the frequency, both relaxation peaks shied to higher
temperatures.
Fig. 9 shows a comparison of D3r0 and tan d for samples 3b,
4c, 5c, and 6b. The D3r0 decreased from sample 3b to 6b. This
was understandable because the predicted Pdip decreased from
18.0 mC mꢀ2 for 3b to 13.2 mC mꢀ2 for 4c, to 11.2 for 5c and
nally to 10.8 for 6b. Meanwhile, tan d also decreased from 3b
to 6b, because the higher predicted Pdip from CN dipoles
directly attached to the main chain would result in a higher loss.
Fig. 5 (A) Increase of the real part of permittivity, D3r0, and (B) tan d at
190 ꢁC at 103, 104, and 105 Hz for samples 3a, 4a, 5a, and 6a with a
para–para linkage in the diamine.
3.5. Effect of diamine linkage type on D3r0
3.15. With increasing temperature from ꢀ150 ꢁC to 190 ꢁC, 3r0
increased continuously. From the 30r0 and tan d plots, the g
relaxation peak was seen at ca. ꢀ80 ꢁC and the b shoulder was
From the above results, we learned that the dipole moment of a
repeat unit played an important role in enhancing the D3r0.
Meanwhile, we noticed that the type of linkage in the diamine
20690 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
Fig. 6 Temperature-scan plots of 3r0, 3r00, and tan d for samples 4b and 5b with a meta–para linkage in the diamine.
dielectric properties from D–E loops. In this study, for example,
the slope of the loop (i.e., Dmax/Emax) gave the apparent dielectric
constant; the charged and discharged energy densities (Ue) were
Ð
obtained from Ue ¼ EdD of the respective charging and dis-
charging curves;52,53 the loop area represented the energy lost.
The loss mechanisms again include electronic conduction,
impurity ion migration, and dipole motion. From a recent
report,54 we understand that electronic conduction shis up the
upper half of the D–E loop, and impurity ionic migration and
dipole switching broaden the loop symmetrically and increase
the slope as well. Therefore, the loss from electronic conduction
can be deconvoluted. However, it is fairly difficult to deconvo-
lute contributions from impurity ions and dipoles in one D–E
loop. One way to circumvent this difficulty is to run D–E loops at
a high enough frequency (e.g., 1 kHz or above) so that the
contribution from impurity ion migration could be minimized,
because relaxation for amorphous dipoles usually occurs above
1 kHz whereas impurity ions in polymers are oen slow, e.g.,
below 1 kHz. Examples of dipolar and impurity ion losses have
been discussed in Fig. 3C and F.
Fig. 7 (A) Increase of the real part of permittivity, D3r0, and (B) tan d at
190 ꢁC at 103, 104, and 105 Hz for samples 4b and 5b with a meta–para
linkage in the diamine.
also affects the D3r0. From Fig. 10, the general trend for groups 3–
6 was that the D3r0 decreased from para–para to meta–para, and
to meta–meta regarding the type of linkage in the diamine,
regardless of whether the frequency was 1 kHz or 100 kHz. This
indicates that the para–para linkage in the diamine containing
3 CN dipoles could rotate more easily than the meta–para and
meta–meta linkages.
From Fig. 11, PI samples without any CN dipoles, such as 1,
Kapton, and Ultem, exhibited narrow loops, no matter what the
frequencies and temperatures. This is the advantage of linear
dielectric polymers. On the other hand, PI samples with CN
dipoles exhibited broader loops. Below, we will discuss the
different loss mechanisms for CN-containing PI samples.
Careful inspection of D–E loops for CN-containing PI samples
indicated that electronic conduction loss was not a signicant
loss mechanism for the PI samples because most loops (except
that of 2b, which may result from defects such as pin-holes in
the sample) were symmetrical about the electric eld axis. We
consider that electron traps in these PI samples were effective in
preventing the hopping mechanism of electron conduction55
3.6. Electrical energy storage studied by D–E loops
In the above studies, we have focused on linear dielectric
properties at low electric elds. In practice, polymer dielectrics
are oen used at high electric elds. For example, BOPP lm
DC-link capacitors in electric vehicles oen operate at 200 MV
mꢀ1 and at a frequency of 10–20 kHz.8 Therefore, it is highly
desired to investigate high-eld dielectric properties of the PI
samples in Table 1. Here, we employ both D–E hysteresis loop
and high-eld BDS studies. Fig. 11 shows typical D–E loops
under a poling eld of 100 MV mꢀ1 for various PI samples in
Table 1, together with Kapton and Ultem for comparison. In
general, one could obtain useful information on high-eld
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20691

View Article Online
Journal of Materials Chemistry A
Paper
Fig. 8 Temperature-scan plots of 3r0, 3r00, and tan d for samples 3b, 4c, 5c, and 6b with a meta–meta linkage in the diamine.
Fig. 9 (A) Increase of the real part of permittivity, D3r0, and (B) tan d at
190 ꢁC at 103, 104, and 105 Hz for samples 3b, 4c, 5c, and 6b with a
meta–meta linkage in the diamine.
Fig. 10 Increase of the real part of permittivity, D3r0, at (A) 1 kHz and (B)
100 kHz for groups 3–6 with different types of linkage in diamines:
para–para (ppDAm), meta–para (mpDAm), and meta–meta
(mmDAm). Connecting lines are shown to guide the eye.
when the temperature was below the Tg. At 100 ꢁC, all D–E loops
at the poling frequency of 10 Hz appeared relatively narrow,
indicating that electron traps were effective at a temperature of
at least 100 ꢁC below the Tg. At 190 ꢁC and 1 kHz, the D–E loops
appeared slightly broader than those at 100 ꢁC and 10 Hz.
Considering the dipole relaxation peak at 190 ꢁC around 1 kHz
(Fig. 3F), the loss mechanism at 190 C and 1 kHz should be
ꢁ
attributed to dipole wiggling in the CN-containing PI samples.
ꢁ
However, the loops measured at 190 C and 10 Hz (shown in
20692 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
Fig. 11 Bipolar D–E hysteresis loops for various polyimide samples in Table 1, together with Kapton and Ultem, measured at 10 and 1 kHz (where
noted).
red) had a steeper slope than those measured at 190 C and 1 the slope of the D–E loops.54 On the basis of our previous
ꢁ
kHz. This could be attributed to additional impurity ion reports, less than ppm level of impurity ions could cause
migration in the samples because migration of impurity ions signicant impurity-ion migration loss.51
increases the loss (i.e., broader loops) and the accumulation of
To understand the effect of dipolar polarization, we had to
these ions at both electrodes increases the capacitance and thus exclude the loss mechanism due to impurity ions. As we
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20693

View Article Online
Journal of Materials Chemistry A
Paper
Fig. 12 Discharged energy density of different PIs from DE-loop
measurements at 100 MV mꢀ1 at different temperatures and
frequencies.
Fig. 14 3r0 and 30r0 versus temperature for samples 3b and 5a under a
high electric field with a peak amplitude of 10 MV mꢀ1 for 3b and 50
MV mꢀ1, for 5a.
addition, introducing more CN dipoles onto the PI main chain
further increased the polymer rigidity.
The discharged energy densities and loss% at an applied
eld of 100 MV mꢀ1 were obtained from D–E loops for all
samples, and results are summarized in Fig. 12 and 13. Samples
3a and 3b exhibited the highest discharged energy density with
reasonably low loss%.
PI samples without any nitrile groups had the lowest
loss%; however, their discharged energy densities were also
the lowest. For example, the discharged energy density for 3a
ꢁ
at 190 C and 10 Hz was about 1.9 times those of samples 1,
Kapton, and Ultem (see Fig. 12). From Fig. 13, the closer the
Fig. 13 Hysteresis loop loss% from D–E loop measurements at 100
MV mꢀ1 at different temperatures and frequencies for different PI
samples.
ꢁ
ꢁ
heights of the red (190 C and 10 Hz) and blue (190 C and 1
kHz) bars, the less ionic and electronic conduction loss
ꢁ
because the loss% at 190 C and 10 Hz included losses from
impurity ion migration and dipole motion, whereas the loss%
at 190 ꢁC and 1 kHz should be exclusively due to dipole
motion.
mentioned above, running D–E loops at 1 kHz could possibly
eliminate the inuence from impurity ions because their
motion is slower than that of dipoles. From the D–E loops at 190
ꢁC and 1 kHz, an apparent dielectric constant could be
obtained. If we assume that the permittivity at ꢀ150 ꢁC and 100
kHz in the BDS study originated solely from deformational (i.e.,
electronic and vibrational) polarization, the polarization from
dipoles could be obtained from the difference between the D–E
3.7. High voltage BDS study
In addition to D–E loops, high-eld dielectric properties were
also studied using HV BDS, and the results for samples 3b and
5a are shown in Fig. 14. These results look similar to the low-
eld BDS spectra; however, there are differences. First, the 3r0
values at ꢀ150 ꢁC for both samples were around 3.5 and
increased with increasing temperature. The difference was
that the D3r0 at 1 kHz was higher for HV BDS than for normal
BDS. For example, D3r0 was 1.25 and 1.06 in HV BDS, whereas it
was 0.62 and 0.87 in low-eld BDS for samples 3b and 5a,
respectively. Second, g and b relaxation peaks were also
observed in HV BDS 30r0 plots for samples 3b and 5a. However,
the 30r0 values were much higher in HV BDS than those in low-
eld BDS. This could be attributed to enhanced electronic
conduction, impurity ion migration, and dipole motion under
high electric elds.
ꢁ
loop apparent dielectric constant at 190 C and 1 kHz and the
BDS permittivity at ꢀ150 ꢁC and 100 kHz. Table 1 lists the
experimental Pdip values for different PI samples. The ratio
between experimental Pdip and predicted Pdip could be an
indicator of dipole wiggling in the PI samples at 190 ꢁC. For
samples containing 1 nitrile group, the ratio was about 0.15,
whereas for samples containing 3 nitrile groups, the ratio was
only around 0.05. This suggested that xing CN dipoles onto
rigid PI main chains signicantly decreased their mobility,
compared with the freely rotating model (see eqn (3)). In
20694 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014

View Article Online
Paper
Journal of Materials Chemistry A
As we can see from Table 1, the fraction of dipole ipping for permittivity and electrical energy storage while maintaining a
most of the samples is only about 0.05. This can be explained by low dielectric loss.
the rigid structure of the PIs, which does not allow dipoles to
turn. This rigid structure of course is what imparts great
thermal stability to PIs. Also, from Fig. 2B, dielectric losses
Acknowledgements
increase with the addition of CN dipoles. Since the CN dipoles The authors acknowledge AFRL Materials and Manufacturing
are attached to the main chain of the PIs in all samples Directorate for partial nancial support. We thank Zhongbo
studied, the chain needs to turn with the CN dipoles when an Zhang for performing the reection-mode XRD measurements
electric eld is applied. The result is obvious: the dipoles are and Michelle Song for helping with the D–E loop measure-
difficult to turn because they need to turn part of the rigid ments. We are also grateful for helpful discussion with Dr
main chain. However, small amplitude turning may still Elshad Allahyarov and Professor Donald Schuele at Case
occur, but is hampered by strong frictional forces giving rise Western Reserve University. IT is grateful to the US Department
to relatively high losses. To solve this problem of low dipole of State for the International Fulbright Science and Technology
turning and high loss, the dipoles should not be attached Award.
directly to the PI main chain. Bendler et al.28 reported a study
attaching cyanomethylene groups to bisphenol
A poly-
Notes and references
carbonate (PC), having a CH2 group between the main chain
and the nitrile group so that the dipole could rotate more
freely in the side group. The CN-modied PC achieved a
polarization fraction of 0.10 of the theoretical prediction and
a dielectric constant of 4.0. In future, we plan to attach
–CH2CN groups to PIs in order to enhance the dipole mobility
but keep the dipolar loss low.
1 W. J. Sarjeant, J. Zirnheld and F. W. MacDougall, IEEE Trans.
Plasma Sci., 1998, 26, 1368–1392.
2 Q. Tan, P. Irwin and Y. Cao, IEEJ Trans. Fund. Mater., 2006,
126, 1152–1159.
3 N. Venkat, T. D. Dang, Z. W. Bai, V. K. McNier, J. N. DeCerbo,
B. H. Tsao and J. T. Stricker, Mater. Sci. Eng., B, 2010, 168, 16–
21.
4 J. Ho and T. R. Jow, Characterization of High Temperature
Polymer Thin Films for Power Conditioning Capacitors, Army
Research Laboratory, Adelphi, MD, 2009.
5 W. J. Sarjeant, I. W. Clelland and R. A. Price, Proc. IEEE, 2001,
89, 846–855.
6 H. Wen, W. Xiao, X. Wen and P. Armstrong, IEEE Trans. Veh.
Technol., 2012, 61, 2950–2964.
7 C. Zou, Q. M. Zhang, S. H. Zhang, D. Kushner, X. Zhou,
R. Bernard and R. J. Orchard, J. Vac. Sci. Technol., B:
Nanotechnol. Microelectron.: Mater., Process., Meas.,
Phenom., 2011, 29, 061401.
4 Conclusions
In this work, we systematically investigated a series of PI
samples containing highly polar nitrile groups. BDS and D–E
loop results indicated that adding CN dipoles into PI struc-
tures increased the permittivity. Especially on the diamine
unit basis, 3 nitrile groups were more effective than 1 nitrile
group in increasing the permittivity. Among polar PI samples
having 3 nitrile groups, the dipole moment of the dianhy-
dride also played an important role. With increasing the
dipole moment of the dianhydride part in the order of
PMDA, OPDA, 6FDA, and BTDA, the enhancement of
permittivity gradually decreased. This suggested that the
dianhydride and diamine should arrange in an opposite
direction with dipole moments cancelling each other on the
molecular scale. In addition, para–para linkage in the
diamine yielded a higher permittivity than meta–para and
meta–meta linkages between dianhydride and diamine parts
due to easier rotation around the para–para bonds. From the
8 D. Montanari, K. Saarinen, F. Scagliarini, D. Zeidler,
M. Niskala and C. Nender, Film Capacitors for Automotive
and
Industrial
Applications,
CARTS
Proceedings,
Jacksonville, FL, USA, April 2009.
9 S. H. Zhang, B. Zellers, J. Henrish, S. Rockey, D. Anderson,
C. Zou and Q. M. Zhang, High Energy Density Film
Capacitors, IEEE Pulsed Power Conference, 2009, vol. 1 and
2, pp. 776–780.
high-eld D–E loop study, discharged energy density and 10 Y. Wang, X. Zhou, Q. Chen, B. Chu and Q. M. Zhang, IEEE
loss% were obtained. Sample 3a exhibited the highest dis-
Trans. Dielectr. Electr. Insul., 2010, 17, 1036–1042.
charged energy density due to its high dipolar polarization. 11 F. MacDougall, J. Ennis, X. H. Yang, K. Seal, S. Phatak,
From the HV BDS study, even higher permittivities were
observed.
B. Spinks, N. Keller, C. Naruo and T. R. Jow, Large High
Energy Density Pulse Discharge Capacitor
Although enhanced permittivity was achieved for these CN-
containing PI samples, dielectric losses, i.e., tan d and loss%,
Characterization, IEEE International Pulsed Power
Conference, Monterey, CA, 2005, pp. 1215–1218.
were also increased. This could be attributed to the fact that CN 12 A. Facchetti, M. H. Yoon and T. J. Marks, Adv. Mater., 2005,
dipoles were directly attached to the PI main chain. In a recent 17, 1705–1725.
study,28 the nitrile group was attached to the bisphenol-A unit of 13 J. Veres, S. Ogier, G. Lloyd and D. de Leeuw, Chem. Mater.,
PC via a methylene group. The rotation of the acetonitrile side 2004, 16, 4543–4555.
group became easier than the nitrilephenyl main chain in our PI 14 L. Zhu and Q. Wang, Macromolecules, 2012, 45, 2937–2954.
samples. In future, we will synthesize PI samples with a polar- 15 L. Yang, X. Li, E. Allahyarov, P. L. Taylor, Q. M. Zhang and
izable side group such as –CH2CN in order to achieve higher
L. Zhu, Polymer, 2013, 54, 1709–1728.
This journal is © The Royal Society of Chemistry 2014
J. Mater. Chem. A, 2014, 2, 20683–20696 | 20695

View Article Online
Journal of Materials Chemistry A
Paper
16 T. C. Chung and A. Petchsuk, Macromolecules, 2002, 35, 36 D. H. Wang, J. K. Riley, S. P. Fillery, M. F. Durstock, R. A. Vaia
7678–7684.
and L. S. Tan, J. Polym. Sci., Part A: Polym. Chem., 2013, 51,
4998–5011.
37 D. H. Wang, B. A. Kurish, I. Treufeld, L. Zhu and L.-S. Tan, J.
Polym. Sci., Part A: Polym. Chem., 2014, DOI: 10.1002/
pola.27445.
17 Y. Lu, J. Claude, B. Neese, Q. Zhang and Q. Wang, J. Am.
Chem. Soc., 2006, 128, 8120–8121.
18 Y. Lu, J. Claude, L. E. Norena-Franco and Q. Wang, J. Phys.
Chem. B, 2008, 112, 10411–10416.
19 R. Su, J. K. Tseng, M. S. Lu, M. Lin, Q. Fu and L. Zhu, Polymer, 38 J. Melcher, Y. Daben and G. Arlt, IEEE Trans. Electr. Insul.,
2012, 53, 728–739. 1989, 24, 31–38.
20 G. A. Samara, J. Phys.: Condens. Matter, 2003, 15, R367–R411. 39 A. R. Blythe and D. Bloor, Electrical Properties of Polymers,
21 A. A. Bokov and Z. G. Ye, J. Mater. Sci., 2006, 41, 31–52. Cambridge University Press, New York, 2nd edn, 2005.
22 J. L. Dormann, D. Fiorani and E. Tronc, Adv. Chem. Phys., 40 A. I. Vogel, W. T. Cresswell, G. H. Jeffery and J. Leicester, J.
1997, 98, 283–494. Chem. Soc., 1952, 514–549.
23 D. L. LesliePelecky and R. D. Rieke, Chem. Mater., 1996, 8, 41 S. Milleori, A. Alparone, A. Milleori and A. Vanella,
1770–1783. Biophys. Chem., 2008, 132, 139–147.
24 J. R. Fried, Sub-Tg transitions in Physical Properties of 42 M. A. Buldakov, I. I. Matrosov and V. N. Cherepanov, Opt.
Polymers Handbook, Springer, New York, 2007, ch. 13, pp.
217–232.
Spectrosc., 2000, 89, 37–41.
43 I. E. Coop and L. E. Sutton, J. Chem. Soc., 1938, 1869–1872.
25 F. Bai, T. D. Jones, T. W. Kelley, T.-C. Lee, K. M. Lewandowski 44 A. R. H. Goodwin and J. B. Mehl, Int. J. Thermophys., 1997, 18,
and D. V. Muyres, Organic polymers, electronic devices and
methods, US Pat., US2004/0222415, 11 November, 2004.
26 F. Bai, T. D. Jones, T. W. Kelley, T.-C. Lee, K. M. Lewandowski
795–806.
45 C. W. N. Cumper and A. P. Thurston, J. Chem. Soc., Perkin
Trans. 2, 1972, 106–111.
and D. V. Muyres, Organic polymers, laminates and 46 K. Wohlfart, M. Schnell, J. U. Grabow and J. Kupper, J. Mol.
capacitors, US Pat., US2005/0019592 A1, 27 January, 2005. Spectrosc., 2008, 247, 119–121.
27 J. T. Bendler, C. A. Edmondson, M. C. Wintersgill, 47 J. S. Blakemore, Solid State Physics, Cambridge University
D. A. Boyles, T. S. Filipova and J. J. Fontanella, Eur. Polym.
J., 2012, 48, 830–840.
28 J. T. Bendler, D. A. Boyles, C. A. Edmondson, T. Filipova,
J. J. Fontanella, M. A. Westgate and M. C. Wintersgill,
Macromolecules, 2013, 46, 4024–4033.
29 M. K. Ghosh and K. L. Mittal, Polyimides: Fundamentals and
Applications, Marcel Dekker, New York, 1996.
30 L. Li, R. Kikuchi, M. A. Kakimoto, M. Jikei and A. Takahashi,
High Perform. Polym., 2005, 17, 135–147.
31 H. A. Kang, I. S. Chung, M. A. Kakimoto and S. Y. Kim,
Polymer J., 2001, 33, 284–289.
Press, New York, 2nd edn, 1985.
48 A. Kahouli, O. Gallot-Lavallee, P. Rain, O. Lesaint,
C. Guillermin and J.-M. Lupin, Relationship between
structure and dielectric properties of bi-oriented isotactic
polypropylene lms for capacitors, IEEE international
conference on solid dielectrics, Bologna, 2013, pp. 1068–1071.
¨
49 F. Kremer and A. Schonhals, Broadband Dielectric
Spectroscopy, Springer, New York, 2003.
50 L. F. Chen, Microwave Electronics: Measurement and Materials
Characterization, Wiley, Hoboken, NJ, 2004.
51 M. Mackey, D. E. Schuele, L. Zhu and E. Baer, J. Appl. Phys.,
2012, 111, 113702.
52 F. Guan, J. Wang, L. Yang, J. K. Tseng, K. Han, Q. Wang and
L. Zhu, Macromolecules, 2011, 44, 2190–2199.
32 C. Hamciuc, E. Hamciuc, M. Ignat and G. Zarnescu, High
Perform. Polym., 2009, 21, 205–218.
33 E. Hamciuc, C. Hamciuc and M. Ignat, High Perform. Polym.,
2010, 22, 225–236.
53 B. Chu, X. Zhou, K. Ren, B. Neese, M. R. Lin, Q. Wang,
F. Bauer and Q. M. Zhang, Science, 2006, 313, 334–336.
34 B. Gonzalo, J. L. Vilas, T. Breczewski, M. A. Perez-Jubindo,
M. R. De La Fuente, M. Rodriguez and L. M. Leon, J. 54 L. Yang, E. Allahyarov, F. Guan and L. Zhu, Macromolecules,
Polym. Sci., Part A: Polym. Chem., 2009, 47, 722–730. 2013, 46, 9698–9711.
35 J. D. Jacobs, M. J. Arlen, D. H. Wang, Z. Ounaies, R. Berry, 55 K.-C. Kao, Dielectric Phenomena in Solids: with Emphasis on
L. S. Tan, P. H. Garrett and R. A. Vaia, Polymer, 2010, 51,
3139–3146.
Physical Concepts of Electronic Processes, Elsevier Academic
Press, Boston, 2004.
20696 | J. Mater. Chem. A, 2014, 2, 20683–20696
This journal is © The Royal Society of Chemistry 2014