I. INTRODUCTION
Electroceramics with ultrahigh dielectric constant have triggered an extensive research because of their potential application in miniaturized electronic components/device. Recently, much attention has been paid to a body-centered cubic CaCu3Ti4O12 (CCTO) as it exhibits high dielectric constant of 104–105 and good stability over a wide temperature range of 100–400 K (Subramanian et al., Reference Subramanian, Li, Duan, Reisner and Sleight2000; Homes et al., Reference Homes, Vogt, Shapiro, Wakimoto and Ramirez2001; Sinclair et al., Reference Sinclair, Adams, Morrison and West2002; Wang and Zhang, Reference Wang and Zhang2007). In addition, CCTO is a lead-free (Pb-free) compound and thus does not cause environmental pollution. Theoretical calculations reveal that CCTO provides a dielectric constant only in the range of 40–50 (He et al., Reference He, Neaton, Cohen and Vanderbilt2002). Impedance spectroscopy study on CCTO ceramics shows that they are electrically heterogeneous and consist of semiconducting grains with insulating grain boundaries. Therefore, an internal barrier layer capacitance (IBLC) model instead of an intrinsic property associated with the crystal structure (Capsoni et al., Reference Capsoni, Bini, Massarotti, Chiodelli, Mozzatic and Azzoni2004; Fang and Liu, Reference Fang and Liu2005; Zhang et al., Reference Zhang, Zheng, Wang, Zhao, Li and Wang2005; Wang and Zhang, Reference Wang and Zhang2007) is suggested to be responsible for the giant dielectric constant of CCTO. Although this model is now generally accepted, the origin of semiconductivity in grain and the composition of the insulating grain boundary phase in CCTO ceramics are controversial. Furthermore, the loss tangent of CCTO is rather high (~0.1–0.2 at 1 kHz) from an application viewpoint. Improvements are therefore needed to facilitate device implementation.
It is intriguing to note that substituting for Ca usually decreases the ultrahigh dielectric constant of CCTO (Subramanian and Sleight, Reference Subramanian and Sleight2002; Homes et al., Reference Homes, Vogt, Shapiro, Wakimoto, Subramanian and Ramirez2003; Liu et al., Reference Liu, Duan, Yin, Mei, Smith and Hardy2004, Reference Liu, Duan and Mei2005; Babu et al., Reference Babu, He, Zhang, Chen and Dhanasekaran2007; Ren et al., Reference Ren, Liang and Yang2010; Sebald et al., Reference Sebald, Krohnsa, Lunkenheimer, Ebbinghaus, Riegg, Reller and Loidl2010; Somphan et al., Reference Somphan, Thongbai, Yamwong and Maensiri2013). Therefore, investigating the dielectric properties of ACu3Ti4O12 in detail will help explore the underlying physics behind the dielectric behaviors of CCTO. In this paper, we investigated the microstructure, dielectric, and electrical properties of Eu2/3Cu3Ti4O12 (ECTO), Tb2/3Cu3Ti4O12 (TCTO), and Na1/2Eu1/2Cu3Ti4O12 (NECTO), together with the results of CCTO. Substituting for Ca decreased the dielectric constant and improved the loss tangent of CCTO. The possible reasons for the decreased dielectric response are discussed.
II. EXPERIMENTAL
ACu3Ti4O12 (A = Eu2/3, Tb2/3, and Na1/2Eu1/2) (ACTO) and CCTO powders were synthesized by the conventional solid-state reaction method. High-purity Eu2O3, Tb2O3, Na2CO3, CaCO3, CuO, and TiO2 were weighed according to the stoichiometric ratios and mixed thoroughly in an agate mortar. The mixed powders were calcined in air at 1000 °C for 12 h and at 1100 °C for 24 h with an intermediate grinding. The calcined samples were milled and pressed into pellets of 5 mm in diameter and approximately 0.8 mm in thickness. The pellets were then sintered in air at 1100 °C for 20 h.
X-ray powder diffraction (XRD) data were recorded on an X-ray diffractometer (XRD-7000, SHIMADZU Limited) with CuKα radiation and a diffracted-beam graphite monochromator operated at 40 kV and 40 mA. The XRD data were indexed using the DICVOL91 program (Boultif and Louёr, Reference Boultif and Louër1991). The surface morphologies were obtained using an FEI, XL-30 scanning electronic microscopy (SEM) coupled with energy-dispersive spectrometer (EDS). To measure the dielectric properties, silver electrodes were painted on the samples' surfaces. Dielectric and complex impedance data were collected using an Agilent-4294A impedance analyzer with an AC voltage of 0.5 V. The measurements were performed between 293 and 473 K over the frequency range of 40 Hz–2.5 MHz.
III. RESULTS
The XRD patterns of CCTO and ACTO are shown in Figure 1. All the patterns can be indexed to a body-centered cubic cell (space group 
$Im\bar{3}$). This finding indicates that the doped ions do not cause any significant changes in the crystal structure. Notably, the diffraction peaks of ACTO shift toward high angles compared to CCTO (see the inset of Figure 1), indicating the change in the lattice constant (Chen et al., Reference Chen, Liang and Wang1995; Chen and Eysel, Reference Chen and Eysel1999). The indexed lattice parameters of ACTO are 7.38914(15) Å (ECTO), 7.38406(10) Å (TCTO), and 7.39140(10) Å (NECTO), which shrink compared to that of CCTO (7.39397(11) Å); this is due to the smaller ionic radii of Eu3+ (r = 0.95 Å), Tb3+ (r = 0.92 Å), and Na+ (r = 0.95 Å) than Ca2+ (r = 0.99 Å), and vacancies existed on the Ca site in ECTO and TCTO (the A sites are one-third vacant in order to achieve charge neutrality). Table I lists the observed and calculated d spacings and 2θ angles for CCTO and ACTO. In addition, no second phase is detected in the XRD patterns of ACTO. All these results show that Eu3+, Tb3+, or Na+ and Eu3+ together enter the lattice and substitute for Ca2+ (Chen et al., Reference Chen, Bauernfeind and Braun1997).
XRD patterns for CCTO, ECTO, TCTO, and NECTO. The inset is the expanded view of the XRD patterns in the 2θ range of 48.5°–50°.
List of indexes, d values, and 2θ of CCTO, ECTO, TCTO, and NECTO.
The microstructures of CCTO were examined with the SEM technique (see Figure 2). SEM showed the formation of irregularly shaped crystallites. Some grains of CCTO grew rapidly to sizes around 50 µm, indicating abnormal grain growth (Adams et al., Reference Adams, Sinclair and West2002; Chung et al., Reference Chung, Kim and Kang2004; Feng and Shiau, Reference Feng and Shiau2004). Moreover, a secondary liquid phase that crystallized between the grains was detected. To confirm the elemental compositions of the grains and the secondary liquid phase at grain boundaries, EDS spectra were acquired from the positions marked as #1 and #2 with two boxes. The EDS spectra of the grains showed the peaks corresponding to Ca, Cu, Ti, and O elements. The EDS spectra of the secondary liquid phase show that Cu peaks of the secondary liquid phase are much stronger than those of the grains. The analysis results of these two EDS spectra are listed in Table II. This finding indicates that the atomic percentage of the grains is nonstoichiometric, and Cu and O are deficient in the grain of CCTO. In the selected area of the secondary liquid phase, copper (48.46 at%) and oxygen (45.56 at%) were detected as the major components. It is reasonable to assume that this intergranular secondary liquid phase is CuO (or rich in CuO). However, we did not detect the CuO phase in XRD data; this may be because the amount of CuO was too little to be detected by the XRD technique. Further, a little amount of Ti (4.12 at%) and Ca (1.85 at%) was found to co-exist with copper and oxygen in the selected area of the secondary liquid phase.
The SEM micrograph of CCTO and the corresponding EDS spectra.
Atomic percentages (at%) of CCTO grain and the secondary phase obtained from their EDS spectra.
The microstructures of ACTO are shown in Figure 3. The microstructure of ACTO was found to be different from that of CCTO as their grains have smooth faces associated with cubic appearance without the segregation of the Cu-rich phase at the grain boundaries. Moreover, the abnormal growth of the grain stops, and the grains become apparently uniform and fine. The largest grain of ACTO was about 10 µm. All these results indicate that Eu, Tb, and Na occupying the Ca site can induce significant changes in the microstructure. Such microstructure changes are considered to correlate closely with the absence of the CuO secondary phase (as discussed below).
The SEM micrographs of ECTO (a), TCTO (b), and NECTO (c).
The dielectric responses of CCTO and ACTO were investigated. The results are shown in Figure 4. Figure 4(a) presents the frequency-dependent dielectric responses at room temperature. It is apparent that the dielectric constants decreased after substitution on the Ca site. Notably, the loss tangent of the substituted samples also largely decreased compared to that of CCTO. The dielectric constants and loss tangents of the four samples at 1 kHz were about 10 400 and 0.182 (CCTO), 1200 and 0.058 (ECTO), 1500 and 0.042 (TCTO), and 2600 and 0.062 (NECTO), respectively. We can see that substituting on the Ca site reduces the loss tangent by more than 65% at 1 kHz. Although the dielectric constant also decreases, it still maintains a high magnitude of 103. The temperature-dependent dielectric responses were investigated in the temperature range of 293–473 K [see Figure 4(b)]. The measurement frequency was 1 kHz. The CCTO sample was found to exhibit good temperature-independent dielectric constants below 360 K. Above 360 K, the dielectric constant rapidly increases with increasing temperature. The large increase in dielectric constant is accompanied with a loss tangent peak, indicating high-temperature relaxation was present in this sample (Grubbs et al., Reference Grubbs, Venturini, Clem, Richardson, Tuttle and Samara2005; Lei and Chen, Reference Lei and Chen2007; Li et al., Reference Li, Liu and Li2017). Compared to CCTO, the dielectric constants of ACTO only slowly increased as the temperature increased, thus exhibiting good temperature-independent dielectric constants in the measured temperature range. The increments of dielectric constants in the temperature range of 293–473 K were about 380%, 250%, 170%, and 150% for CCTO, ECTO, TCTO, and NECTO, respectively. In addition, no peaks were found in the loss tangent plots of ACTO. Thus, the samples of ACTO do not exhibit high-temperature dielectric relaxations. According to Grubbs et al. (Reference Grubbs, Venturini, Clem, Richardson, Tuttle and Samara2005) and Lei and Chen (Reference Lei and Chen2007), the high-temperature dielectric relaxation of CCTO is related to the oxygen vacancies or the oxygen vacancy-related point defect located at the grain boundary. The reduced high-temperature dielectric relaxation indicates the decrease in oxygen vacancies in the samples of ACTO.
The frequency-dependent dielectric responses measured at room temperature (a) and temperature-dependent dielectric responses measured at 1 kHz (b) of CCTO, ECTO, TCTO, and NECTO.
To investigate the electrical properties of grain and grain boundary for CCTO and ACTO, the complex impedance plots measured at room temperature are presented in Figure 5. Steep inclines with nonzero intercepts at high frequency on the Z′-axis (see the inset of Figure 5) were observed for all the samples. The steep inclines are parts of semicircles representing the low-frequency grain boundary responses. It was reported that the diameters of the low-frequency semicircle correspond to the grain boundary resistances, whereas the high-frequency intercepts with the Z′-axis correspond to grain resistances. At room temperature, R gb ≫ 5 MΩ, ω max < 40 Hz, and therefore, only a small section of the grain boundary semicircle is observed, as a steep incline at low frequencies. It is apparent that both the grain resistance and the grain boundary resistance increase by substituting on the Ca site. The increased grain boundary resistance may originate from the small grains of ACTO possessing large grain boundary area and/or the absence of the CuO secondary phase located at the grain boundary.
The complex impedance plots of CCTO, ECTO, TCTO, and NECTO measured at room temperature. The inset shows the expanded view of the high-frequency impedance data close to the origin.
The complex impedance plots of ACTO were also investigated at different temperatures. The results are shown in Figure 6. With the increase in temperature, both the grain resistance and grain boundary resistance decreased. Conduction data of grain, σ, where σ = 1/R, were obtained by the nonzero intercepts of the complex impedance arcs and plotted against reciprocal temperature in Arrhenius format:
$$\sigma = \sigma _0{\rm exp}\lpar {- E_{\rm a}/k_{\rm B}T} \rpar $$where σ 0 is the prefactor, E a is the activation energy for the relaxation, k B is the Boltzmann constant, and T is the absolute temperature. Figure 7 shows the plots of log σ versus 1/T, in which the solid lines are fitted results using Eq. (1). From the slopes of the fitted straight lines, we obtain the activation energy values of 0.45(5), 0.45(7), and 0.41(5) eV for the grain of ECTO, TCTO, and NECTO, respectively. These values were much larger than the reported values of 0.05–0.1 eV for the grain of CCTO (Homes et al., Reference Homes, Vogt, Shapiro, Wakimoto and Ramirez2001; Chiodelli et al., Reference Chiodelli, Massarotti, Capsoni, Bini, Azzoni, Mozzati and Lupotto2004; Zhang and Tang, Reference Zhang and Tang2004), indicating the different conduction mechanism for the grain of ACTO. The activation energy of 0.05–0.1 eV for the grain of CCTO indicates that the electrons hopping accounts for grain conduction (Zhang and Tang, Reference Zhang and Tang2004). In our case, the activation energies of 0.41–0.45 eV for the grain of ACTO are in accordance with the ion jumping activation energy (0.2–1.0 eV) (Bidault et al., Reference Bidault, Maglione, Actis and Kchikech1995). Therefore, substituting on the Ca site induces a change in the conducting mechanism of grain from electron hopping to ion jumping.
The complex impedance plots (1) and the expanded views of the high-frequency impedance data close to the origin (2) of ECTO (a), TCTO (b), and NECTO (c) measured at different temperatures.
Arrhenius plots of grain conduction data for ECTO, TCTO, and NECTO.
IV. DISCUSSION
The high-temperature dielectric relaxation illustrates the presence of oxygen vacancies in CCTO, which is confirmed by the EDS results that the grain of CCTO is oxygen-deficient. Oxygen vacancies arise from a small amount of oxygen loss during high-temperature sintering in the air, which is a common phenomenon in titanate-based perovskite at ≥1000 °C in the air (Morrison et al., Reference Morrison, Sinclair and West2001). Oxygen loss from the lattice results in the generation of electrons according to:
$${\rm O}_0\Leftrightarrow {\rm V}_0^{\cdot \cdot} + 2e + \displaystyle{1 \over 2}{\rm O}_2$$For carrier compensation, Cu ions are separated from the lattice to compensate the conduction electrons contributed by oxygen vacancies according to:
$${\rm C}{\rm u}_{{\rm Cu}} + 2e\Leftrightarrow {\rm V}_{{\rm Cu}}^{\rm \prime\prime} + {\rm Cu}$$The departure of Cu ions leads to the formation of Cu vacancies in the lattice. These Cu ions mostly aggregate at grain boundaries and finally oxidize to CuO on cooling. The CuO phase in the oxide powder allows liquid-phase sintering and facilitates the grain growth (Sulaimain et al., Reference Sulaimain, Hutagalung, Ain and Ahmad2010; Thongbai et al., Reference Thongbai, Putasaeng, Yamwong and Maensiri2012; Jesurani et al., Reference Jesurani, Kanagesan, Hashim and Ismail2013 ; Yuan et al., Reference Yuan, Wu, Liu, Luo and Li2013a, Reference Yuan, Luo and Wang2013b; Senda et al., Reference Senda, Rhouma, Torkani, Megriche and Autret2017). This finding is in accordance with our experimental results that CCTO with the CuO secondary phase located at the grain boundary possesses larger grains. However, for the samples of ACTO, the high-temperature dielectric relaxation is largely reduced, indicating the decrease in oxygen vacancies in the substituted samples. Furthermore, we did not observe any secondary phase located at the grain boundary. The absence of CuO secondary phase induces the decrease in the grain sizes of the substituted samples. According to the internal barrier layer capacitor (IBLC) model, the effective dielectric constant ε eff can be represented by the following equation (West et al., Reference West, Adams, Morrison and Sinclair2004; Ni et al., Reference Ni, Chen, Liu and Hou2006; Yuan et al., Reference Yuan, Luo and Wang2013b):
$$\varepsilon _{{\rm eff}} = \varepsilon _{\rm r}\lpar {t_{\rm g}/t_{{\rm gb}}} \rpar $$where t g is the grain average size, t gb is the average thickness of the grain boundary layer, and ε r is the relative dielectric constant of the grain boundary layer. Therefore, smaller grain average size will induce a lower dielectric constant. Because the grain sizes of ACTO largely decrease compared to CCTO, this may be a reason for the decrease in the dielectric constant of ACTO. Furthermore, the activation energy results indicate that the conducting mechanism for the grain of ACTO is ion jumping rather than electron hopping (which is for the grain of CCTO). The different carrier in grain may cause the different polarization form of the grain, which may also account for the decreased dielectric constant. In addition, the grain boundary plays an important role in the ceramic dielectric response due to giant barrier layer effects (Kim et al., Reference Kim, Lee, Lee, Kim, Riu and Lee2008; Liu et al., Reference Liu, Fan, Chen and Fang2009). The different grain boundary properties (e.g., absence of CuO secondary phase) may be another reason for the decreased dielectric constant of ACTO.
Therefore, the decreased dielectric constant of ACTO may be related to the decreased grain size, the different carrier in the grain, the different grain boundary properties, or a combination of these factors. Copper plays an important role in the decreased grain size and the different grain boundary properties, as the departure of copper from the lattice leads to the formation of the CuO secondary phase at the grain boundary and further induces abnormal grain growth. However, the departure of copper relies on the formation of oxygen vacancies according to Eqs. (2) and (3). Thus, both the factors of decreased grain size and the different grain boundary properties are associated with oxygen vacancies. According to Eq. (2), one oxygen vacancies can create two electrons, which contribute to the grain conduction of CCTO (Wang et al., Reference Wang, Zhang, Xu, Li, Zhou and Chen2002; Li et al., Reference Li, Feteira, Sinclair and West2006). For the samples of ACTO, the amount of oxygen vacancies decreases. This may induce a different conduction mechanism for the grain. Furthermore, the decreased oxygen vacancies further decrease the content of Cu vacancies in the grain of ACTO [according to Eqs. (2) and (3)], which may also account for the different conduction mechanism for the grain of ACTO. In a word, the deficiency of oxygen vacancies may be a key reason for the decreased dielectric constant for the samples of ACTO. However, they remain high at values of 103. Further, the decrease in loss tangent may originate from the increase in the grain boundary resistance.
V. CONCLUSION
The microstructure, dielectric, and electrical properties of ACu3Ti4O12 (A = Eu2/3, Tb2/3, and Na1/2Eu1/2) ceramics were investigated systematically. The results show that substituting for Ca decreases the loss tangent by more than 65% at 1 kHz, while largely improving the loss tangent of CCTO. Although the dielectric constant also decreases, it remains at a high magnitude of 103. The SEM measurements reveal that the mean grain size largely decreases by substituting for Ca and there is no CuO secondary phase located at the grain boundary in the samples of ACTO. The complex impedance measurements indicate that both the grain resistance and grain boundary resistance increase because of substituting for Ca. The activation energies of 0.41–0.45 eV for the grain of ACTO indicate that the conducting mechanism is ion jumping rather than electron hopping for the grain of CCTO. We speculate that the reduced dielectric constant of CCTO may be related to the decreased grain size, the different carrier in the grain, the different grain boundary properties, or a combination of these factors. All these factors are associated with the deficiency of oxygen vacancies in the samples of ACTO. The decrease in the loss tangent may be due to the increase in the grain boundary resistance.
SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at https://doi.org/10.1017/S0885715619000769.
FUNDING INFORMATION
This research was funded by Natural Science Basic Research Plan in Shaanxi Province of China (Program Nos.: 2019JQ-097 and 2018JQ1026), Scientific Research Program Funded by Shaanxi Provincial Education Department (Grant No. 17JK0548), and the Xi'an University of Technology.
