Physica B 459 (2015) 105–109

Contents lists available at ScienceDirect

Physica B
journal homepage: www.elsevier.com/locate/physb

Modulus spectroscopy of grain–grain boundary binary system

Peng-Fei Cheng a,n, Jiang Song a, Sheng-Tao Li b, Hui Wang b
a
School of Science, Xi’an Polytechnic University, Xi’an 710048, China
b
State Key Laboratory of Electrical Insulation and Power Equipment, Xi’an Jiaotong University, Xi’an 710049, China

art ic l e i nf o a b s t r a c t

Article history: Understanding various polarization mechanisms in complex dielectric systems and specifying their
Received 6 October 2014 physical origins are key issues in dielectric physics. In this paper, four different methods for representing
Received in revised form dielectric properties were analyzed and compared. Depending on the details of the system under study,
2 December 2014
i.e., uniform or non-uniform, it was suggested that different representing approaches should be used to
Accepted 3 December 2014
obtain more valuable information. Especially, for the grain–grain boundary binary non-uniform system,
Available online 5 December 2014
its dielectric response was analyzed in detail in terms of modulus spectroscopy (MS). Furthermore, it was
Keywords: found that through MS, the dielectric responses between uniform and non-uniform systems, grain and
Modulus spectroscopy grain boundary, Maxwell–Wagner polarization and intrinsic polarization can be distinguished. Finally,
Non-uniform system
with the proposed model, the dielectric properties of CaCu3Ti4O12 (CCTO) ceramics were studied. The
Grain–grain boundary binary system
colossal dielectric constant of CCTO at low frequency was attributed to the pseudo relaxation process of
Maxwell–Wagner polarization
Intrinsic polarization grain.
& 2014 Published by Elsevier B.V.

1. Introduction which makes the further analysis of relaxation mechanism rather

Dielectric spectroscopy is a very powerful and widely used
method to explore the dielectric properties of dielectric. Through
dielectric spectra under various fields, including frequency, tem-
perature, direct current (DC) bias and so on, the polarization
mechanisms, conduction process, point defect structure, micro-
structure and other information can be obtained [1–9].
In fact, parameters obtained by dielectric spectroscopy are
overall apparent parameters, implying that the dielectric is as-
sumed to be a uniform system. Nevertheless, for a non-uniform
system such as grain–grain boundary binary system, a large dif-
ference between apparent dielectric parameters and those of each
component makes further analysis rather difficult. Especially for a
multi-component dielectric with complex microstructure, it is not
easy to know in advance whether its properties within a limited
frequency and temperature region originate from some certain
component or well-known polarization mechanisms. For example,
CaCu3Ti4O12 (CCTO) ceramics with giant dielectric constant have
very complex microstructure, including grain, grain boundary, [10,
11] nano-scale domains, [12–15] surface layer, [16, 17] and so on.
Although it is very clear that giant dielectric constant comes from
high frequency relaxation with activation energy of about 0.1 eV,
but the corresponding relaxation component is still not clear,
difficult. Therefore, representation methods of dielectric proper-
ties have to be studied in detail to reveal the relationship between
dielectric properties and microstructure.
Generally speaking, there are four methods for representing
dielectric properties, which are dielectric spectroscopy (DS), ad-
mittance spectroscopy (AS), electric modulus spectroscopy (MS),
and impedance spectroscopy (IS) [18–20]. Although these methods
are equivalent to each other theoretically, [21] they are different in
application because different microstructure information is con-
tained in them. In this paper, differences between the character-
izing of dielectric properties by these four methods were discussed
firstly, and then the theoretical analysis method of MS was ela-
borated. The similarities and differences between MS for uniform
and non-uniform systems were compared, and the essence of
Maxwell–Wagner (MW) polarization, the difference between MW
polarization and intrinsic polarization were investigated with a
simple grain–grain boundary binary system. Finally, as a real ap-
plication, the dielectric properties of CaCu3Ti4O12 (CCTO) ceramics
were investigated by MS and the results exhibit good agreements
with our proposed model.
2. Representation methods for dielectric properties

n
Corresponding author. For a uniform system with several relaxation processes, parallel
E-mail address: pfcheng@xpu.edu.cn (P.-F. Cheng). model is convenient and the equivalent circuit is shown in Fig. 1.

http://dx.doi.org/10.1016/j.physb.2014.12.001
0921-4526/& 2014 Published by Elsevier B.V.
106 P.-F. Cheng et al. / Physica B 459 (2015) 105–109

G1 modulus as follows:
k1ε1′ k2ε2′
C1 m′ = ωε0z″ = + = k1m1′ + k2m2′
ε′12 + ε1″ 2 ε′22 + ε2″ 2 (10)
G2
k1ε1″ k2ε2″
C2 m″ = ωε0z′ = + = k1m1″ + k2m2″
ε1′ 2 + ε1″ 2 ε1′ 2 + ε2″ 2 (11)
Fig. 1. Parallel equivalent circuit model for uniform system. From the analysis above, it can be seen that it is more con-
venient for uniform system to be represented with DS or AS. Under
The dielectric properties can be expressed with admittance as this condition, dielectric response can be exhibited separately ac-
follows cording to dielectric relaxations. While, IS or MS is more con-
venient for non-uniform system and dielectric properties can be
Y ⁎ = G1 + G2 + jω(C1 + C2) (1)
represented individually according to components. In the follow-
ing text, we will discuss representation of dielectric properties by
s s s
Y ′ = G1 + G2 = (g1 + g2) = ωε0(ε1″ + ε2″) = ωε0ε″ MS only.
l l l (2)

ωε0ε1′s ωε0ε2′s ωε0ε′s

Y ″ = ω(C1 + C2) = + = 3. Distinguish between uniform and non-uniform systems by
l l l (3)
modulus spectroscopy
where G is the conductance, C is the capacitor, ω is angle fre-
quency, s is electrode area, l is sample thickness, ε0, ε′ and ε′′ are 3.1. Modulus spectroscopy for uniform system
dielectric constant of vacuum, relative dielectric constant and
dielectric loss respectively. Removal of the geometric factors, To the first approximation, different polarization is considered
dielectric properties of uniform system can be represented as to be separated completely along the frequency axis. The effect of
follows DC conductance is much higher than polarization at low frequency
limit, and then the modulus induced by DC conductivity can be
Δε1′ Δε2′
ε′ = ε1′ + ε2′ = ε∞ + + expressed as
1 + ω2τ12 1 + ω2τ22 (4)
ω2ε02εs2/γ 2 ω2τp2
γ1 + γ2 m′ = =
Δε1′ωτ1 Δε2′ωτ2 εs(1 + ω2ε02εs2/γ 2) εs(1 + ω2τp2) (12)
ε″ = ε1″ + ε2″ = + +
ωε0 1 + ω2τ12 1 + ω2τ22 (5)
ωε0εs/γ ωτp
where γ is DC conductivity, τ is relaxation time, Δε is the variation m″ = =
of dielectric constant for a certain polarization from optical fre- εs(1 + ω2ε02εs2/γ 2) εs(1 + ω2τp2) (13)
quency to quasi-DC steady state. If a certain polarization originates
where ε0εs/γ is regarded as a new equivalent relaxation time τp.
from the ionization of point defect, for example electronic re-
The equations above are very similar to those of Debye equations
laxation, [1, 2] DC conductance and dielectric polarization will
and a “relaxation peak” will appear at low frequency limit in m″–f
have the same micro-origin.
curve also. However, the “relaxation peak” is not induced by di-
As for a non-uniform system, series model is more convenient
electric relaxation, but by DC conduction, so we call it a pseudo
as show in Fig. 2. The impedance of the system can be shown as relaxation. The pseudo relaxation is presented as an increasing line
follows: with a slope of 2 at low frequency side of m′–f curve, and then m′
1 1 R (1 − jωR1C1) R (1 − jωR2C2) turns to be the reciprocal of the dielectric constant with the in-
Z⁎ = Z1⁎ + Z2⁎ = + ⁎ = 1 + 2 crease of frequency, as shown in Fig. 3(a) for a uniform system
Y1⁎ Y2 1 + ω2R12C12 1 + ω2R22C22 (6)
without relaxations. If there are several relaxations in a uniform
n system, there will be several steps as shown in Fig. 3(b). Generally,
where Z and R are complex impedance and resistance respec-
tively. The corresponding specific impedance is the variation of dielectric constant induced by polarization is less
than 10 times, so there is very little difference between the
l1 ⁎ l2 ⁎ neighboring steps. In addition, the activation energy for pseudo
z⁎ = z + z2 = k1z1⁎ + k2z2⁎
l 1 l (7) relaxation should be same as DC conductance.

1 ⎛ k1ε1″ k ε″ ⎞ 3.2. Modulus spectroscopy for non-uniform system

z′ = ⎜⎜ 2 2
+ 2 2 2 2 ⎟⎟ = k1z1′ + k2z2′
ωε0 ⎝ ε1′ + ε1″ ε2′ + ε2″ ⎠ (8)
Grain–grain boundary binary system is the simplest non-uni-
1 ⎛ k1ε1′ k ε′ ⎞ form system and it is also the basic microstructure of electronic
z″ = ⎜⎜ 2 2
+ 2 2 2 2 ⎟⎟ = k1z1″ + k2z2″ ceramics; therefore we adopt grain–grain boundary structure as
ωε0 ⎝ ε1′ + ε1″ ε2′ + ε2″ ⎠ (9) the typical non-uniform system in the following text.
where l1 and l2 are the thickness of the first and second compo-
nents, l is the thickness of the non-uniform system, k1 ¼l1/l, k2 ¼l2 3.2.1. Non-uniform system without polarization
/l. In the same way, series model can be expressed with specific At low frequency limit, dielectric polarization can be omitted
and MS can be expressed as
G1 G2 k1 ωε0ε1′ 2 k ⎛ ωε ε ′ ⎞
2

m10
′ ≈ ( ′ ≈ 2 ⎜⎜ 0 2 ⎟⎟
) , m20
C1 C2 ε1′ γ1 ε2′ ⎝ γ2 ⎠ (14)
where subscripts 1 and 2 represent the first component, grain
Fig. 2. Series equivalent circuit model for non-uniform system. boundary and the second component, grain respectively. Generally
 P.-F. Cheng et al. / Physica B 459 (2015) 105–109 107

a b
m’
1
m’ ε∞
1
ε
1 1
ε εs

Pseudo polarization Pseudo polarization

f f
Fig. 3. Modulus spectroscopy for uniform system (a) without polarization and (b) with polarizations.

γ2 c γ1, γ2/γ1 c d2/d1 and k2/ε2 c k1/ε1 for grain–grain boundary m’

binary system, so we have τp1 ¼ ωε0ε′1/γ1 c τp2 ¼ ωε0ε′2/γ2 and
k2
k2
ε ∞2
m′10 »m′20 , which implies that modulus at lower frequency region k1 ε s2
mainly corresponds to grain boundary. At high frequency limit k1 ε∞1
modulus can be expressed as m′1∞ = k1/ε1 { m′2∞ = k2/ε2 for k1 {k2, ε s1
thus modulus at high frequency region mainly corresponds to Dielectric relaxation
grain. The main features of MS for grain–grain boundary binary MW polarization
system without relaxations can be shown as Fig. 4. Dielectric relaxation
In fact, Fig. 4 and Fig. 3(b) are similar apparently, therefore
pseudo relaxation of grain can be regarded as an equivalent po- Pseudo relaxation
f
larization, corresponding to MW polarization in a non-uniform
system. For MW polarization, dielectric constant changes from ε2 Fig. 5. Modulus spectroscopy for non-uniform system with polarizations.
/k2 at high frequency to ε1/k1 at low frequency. As for grain–grain
boundary binary system, k1 { k2, so apparent dielectric constant of m’
the system increase quickly from ε2 to ε1/k1 with the decrease of
frequency. Considering k1/ε1 { k2/ε2, we have k2ε2(ωε0/γ2)2 ¼ I J
k2/ε2, namely ωε0ε2/γ2 ¼1. Loss peak with relaxation time τ in k2 / ε s 2
G H
dielectric spectroscopy corresponds to ωτ ¼ 1, while loss peak with B C
k1 / ε ∞1
relaxation time τs in MS corresponds to ωτs ¼ εs/ε1 ¼(ε1/k1)/
k1 / ε s1 E F
(ε2/k2), then τs for pseudo polarization in MS can be expressed as A D
τs ¼ ε0ε1d2/γ2d1. Relaxation time τs for MW polarization is d2/d1 τ0
times longer than pseudo relaxation in MS, however the most f
probable frequency of MW polarization is the same as pseudo O
relaxation of grain. The results above can be extended to any non-
Fig. 6. The relationship between pseudo polarization and intrinsic polarization in
uniform system. So it can be concluded that MW polarization in
non-uniform system.
non-uniform system is a pseudo relaxation induced by DC con-
ductivity of the high frequency component.
polarization
3.2.2. Non-uniform system with polarizations in grain or at grain For a binary system with polarizations, the MS can be shown as
boundary Fig. 6, where OA and EG represent the effect of DC conductance of
Considering that geometry size of grain is more than 103 times grain boundary and grain respectively. Generally, DC conductance
larger than grain boundary and the change in dielectric constant is of grain boundary and grain depends on Schottky barrier and the
less than 10 times after a common polarization, the change in ionization of point defects respectively. Since Schottky barrier
modulus by polarization is much less than the change by DC height is much higher than ionization energy of point defect,
conductance. Therefore, little steps in MS correspond to dielectric pseudo polarization of grain boundary must locates at the lowest
relaxation, while huge steps correspond to pseudo relaxation or frequency compared with others. In other words, intrinsic polar-
MW polarization, as show in Fig. 5. ization and MW polarization locates on the right side of line OA as
shown in Fig. 6.
3.2.3. The relationship between pseudo relaxation and intrinsic If intrinsic polarization enters into the region of AE, then
modulus curve is shown as OADCEGH in Fig. 6. Intrinsic polar-
m’ ization appears in the first component and MW polarization re-
presented as EGH appears at higher frequency. If intrinsic relaxa-
k2 tion locates on the right side of point E, then the modulus curved
ε2 is shown as OBEGHIJ in Fig. 6. For this situation, intrinsic polar-
k1
ization appears in the second component and MW polarization
ε1
locates at lower frequency than intrinsic polarization.
For a common electric ceramic without ferroelectricity, in-
MW polarization
trinsic polarization at grain boundary comes from electronic re-
laxation. At a certain temperature, DC conductance of grain is
Pseudo relaxation
determined mainly by the point defect with the lowest activation
f
energy. If there are n kinds of point defects in grain, pseudo po-
Fig. 4. Modulus spectroscopy for non-uniform system without polarization. larization of grain boundary, n  1 kinds of electronic relaxations at
108 P.-F. Cheng et al. / Physica B 459 (2015) 105–109
grain boundary and MW polarization (grain pseudo polarization),
corresponding to the point defect with the lowest activation en-
ergy in grain, will appears one by one along the frequency axis.
For point A in Fig. 6, it is regarded that DC conductance effect is
almost the same as dielectric polarization, so we have
⎛ ωε ⎞2 k ωε0εs1
k1εs1⎜⎜ 0 ⎟⎟ = 1 ⇒ =1
⎝ γ1 ⎠ εs1 γ1
If we define τs1 ¼ ε0εs1/γ1, then τs1 represents relaxation time of
pseudo polarization induced by DC conductance of the first com-
ponent. The activation energy for this pseudo polarization is equal
to Schottky barrier height at grain boundary. In the same way, it
can be obtained from point B
⎛ ωε ⎞2 k ωε0ε∞1
k1ε∞1⎜⎜ 0 ⎟⎟ = 1 ⇒ =1
⎝ γ1 ⎠ ε∞1 γ1
If we define τ11 ¼ ε0ε11/γ1, then τ11 represents another re-
laxation time of pseudo polarization induced by DC conductance
of the first component. Obviously, for a certain system, modulus
curve is either OADC or OABC. The difference between these two
conditions is that pseudo relaxation represented with the former
is accompanied with electronic relaxation at grain boundary, while
the latter accompanied without any intrinsic polarization in the
region of BCE.
Similarly, point C represents dielectric relaxation in the first
component and the corresponding activation energy is the ioni-
zation energy of intrinsic point defect. Since ionization energy of
point defect is less than Schottky barrier height, the frequency for
C moves to higher frequency compared with A.
Point G represents pseudo polarization of the second com-
ponent, and it can be obtained from G in the same way that
ωε0εs2/γ2 ¼1. If it is regarded as an equivalent polarization, the
effective relaxation time becomes εs/ε1 times longer in MS. At this
time εs ¼ ε11/k1 and ε1 ¼ εs2/k2, so relaxation time of MW
a b
c d
Fig. 7. Dielectric properties of CaCu3Ti4O12 ceramics, (a) ε′–f, (b) ε″–f, (c) m′–f (d) m″–f.
polarization is to be τs ¼ ε0εs2ε11d2/γ2εs2d1 E ε0ε11d2/γ2d1. Al-
though relaxation time for MW polarization is d2/d1 times longer
than pseudo relaxation of the second component, the most
probable frequencies are the same.
4. Dielectric properties of CaCu3Ti4O12 ceramics
(15)
CaCu3Ti4O12 (CCTO) ceramics were prepared by traditional so-
lid phase sintering methods and the dielectric properties were
measured with Novocontrol wide band dielectric spectrometer.
Measuring frequencies varied from 0.1 Hz to 107 Hz and measuring
temperatures changed from  100 °C to 100 °C. Dielectric proper-
ties of CCTO ceramics are shown in Fig. 7. As shown in Fig. 7(a),
dielectric constant of CCTO at low frequency is more than 105,
while it is just about 100 at high frequency. Two relaxation peaks
(16) at middle and high frequency were observed in ε″–f curves as
shown in Fig. 7(b), and the corresponding activation energies are
0.5 eV and 0.1 eV respectively. On the other hand, from Fig. 7(c),
three steps in m′–f curves were observed and the corresponding
activation energies are 0.65 eV, 0.5 eV and 0.1 eV respectively. The
slope of m′ at low frequency in Fig. 7(c) is 1.8, which is close to the
theoretical slope of pseudo relaxation. This implies that the extra
peak at low frequency in Fig. 7(d) corresponds to DC conductance
of grain boundary. The increase of m′ at high frequency is about
1000 times, which corresponds to the size ratio of grain and grain
boundary, therefore high frequency dielectric response corres-
ponds to DC conductance of grain. In addition, since the relaxation
frequency is between the frequencies of pseudo polarizations of
grain boundary and grain, it can be confirmed that middle fre-
quency relaxation originates from electron relaxation of traps
formed by intrinsic point defects at grain boundary. In fact, oxygen
vacancy is the main intrinsic point defect with activation energy of
0.1 eV for monovalent ionization and 0.5 eV for bivalent ionization
respectively [7, 12]. Obviously, DC conductance of grain is de-
termined by monovalent ionization of oxygen vacancy, while
 P.-F. Cheng et al. / Physica B 459 (2015) 105–109 109

electronic relaxation at grain boundary is controlled by bivalent
ionization of oxygen vacancy. This point defect structure of CCTO
coincides with the results of MS. In conclusion, two components
can be distinguished by MS, one is grain boundary, and the other is
grain; giant dielectric constant of CCTO ceramics originates from
pseudo relaxation of grain.
Plan Projects of Education Department of Shaanxi Province of
China (Nos. 12JK0434, 2010JK573) and the Doctoral Scientific Re-
search Foundation of Xi’an Polytechnic University (No. BS0814).
References

[1] G. Vincent, D. Bois, P. Pinard, J. Appl. Phys. 46 (1975) 5173.

5. Conclusions [2] P.F. Cheng, S.T. Li, L. Zhang, J.Y. Li, Appl. Phys. Lett. 93 (2008) 012902.
[3] A.K. Jonscher, Nature 253 (1975) 717.
[4] A.K. Jonscher, Nature 267 (1977) 673.
In this paper, four different methods for representing dielectric [5] P.F. Cheng, S.T. Li, J.Y. Li, C. Ding, Y. Yang, Chin. Phys. B 21 (2012) 097201.
properties were analyzed. It was found that dielectric spectroscopy [6] J.C. Dyre, T.B. Schroder, Rev. Mod. Phys. 72 (2000) 873.
[7] J.Y. Li, X.T. Zhao, S.T. Li, M.A. Alim, J. Appl. Phys. 108 (2010) 104104.
(DS) and admittance spectroscopy (AS) are convenient for uniform [8] P.F. Cheng, S.T. Li, J.Y. Li, Adv. Mater. Res. 393-395 (2012) 24.
system, while impedance spectroscopy (IS) and modulus spec- [9] P.F. Cheng, S.T. Li, H. Wang, Physica B 449 (2014) 160.
troscopy (MS) are more convenient for non-uniform system. In [10] D.C. Sinclair, T.B. Adams, F.D. Morrison, A.R. West, Appl. Phys. Lett. 80 (2002)
2153.
addition, MS of uniform and non-uniform systems was studied in
[11] T.B. Adams, D.C. Sinclair, A.R. West, Adv. Mater. 14 (2002) 321.
detail and the difference between intrinsic polarization and MW [12] M.A. Subramanian, L. Dong, N. Duan, A.W. Reisner, J. Solid State Chem. 151
polarization was discussed as well. It was found that MW polar- (2000) 323.
ization originates from pseudo polarization of the second com- [13] J. Li, A.W. Sleight, M.A. Subramanian, Solid State Commun. 135 (2005) 260.
[14] T.T. Fang, H.K. Shiau, J. Am. Ceram. Soc. 87 (2004) 2072.
ponent (high frequency component). Finally, with the proposed [15] T.T. Fang, C.P. Liu, Chem. Mater. 17 (2005) 5167.
model, the dielectric properties of CCTO ceramics were in- [16] P. Lunkenheimer, S. Krohns, S. Riegg, S.G. Ebbinghaus, A. Reller, A. Loidl., Eur.
vestigated by using MS. It was proposed that the colossal dielectric Phys. J. Spec. Top. 180 (2010) 61.
[17] J. Sebald, S. Krohns, P. Lunkenheimer, S.G. Ebbinghaus, S. Riegg, A. Reller,
constant of CCTO originates from pseudo relaxation of grain. A. Loidl, Solid State Commun. 150 (2010) 857.
[18] F. Kremer, A. Schonhals, Broadband dielectric spectroscopy, Springer, New
York, 2003.
[19] C.J.F. Bottcher, P. Bordewijk, Theory of electric polarization, Elsevier, Am-
Acknowledgement sterdam, 1978.
[20] A.K. Jonscher, J. Phys. D: Appl. Phys. 32 (1999) R57.
This work is supported by the National Natural Science Foun- [21] J. Ross Macdonald, Phys. Rev. B 49 (1999) 9428.

dation of China (Nos. 51277138, 50972118), the Scientific Research