Effect of doping on the structure, piezoelectric and electrical properties of Aurivillius type Ca_0.6(Na_0.5Bi_0.5)_0.4Bi₂Nb₂O₉ ceramics

Abstract

Aurivillius phase [Ca_0.6(Na_0.5Bi_0.5)_0.4]_1−x(LiCe)_x/2Bi₂Nb₂O₉ (CNBNLC-100x, x = 0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12) ceramics were prepared by the conventional solid state method. The LiCe substitution induced structural distortion and orthorhombic distortion of CNBNLC-100x ceramics which influenced piezoelectric activities and Curie temperature, respectively. The results implied that the more structural distortion of lattice leads to higher piezoelectricity. CNBNLC-8 ceramics with ultrahigh T_C of 904 °C had a high piezoelectric activity of 17.6 pC/N, even after annealing at 600 °C for 2 h, the d₃₃ value of CNBNLC-8 ceramics still retained 62.5% of its original value. Ce⁴⁺ as donor dopant, the dc resistivity of CNBN ceramics increased with the substitution of Ce⁴⁺ for Ca²⁺. At high temperature region, the extrinsic conductivity changes of the ceramics from p-type to n-type with the increase of Ce⁴⁺ doping amount. Further investigation demonstrated that the ac conductive process of ceramics referred to three different mechanisms. The main conduction mechanisms were considered as first-order ionization of oxygen vacancies, second- order ionization of oxygen vacancies and intrinsic dc conductivity at three temperature regions, respectively. All these good electrical properties provided the LiCe doped CNBN ceramics with a large potentiality for high-temperature piezoelectric applications.

1 Introduction

Aurivillius-type bismuth layer-structured ferroelectrics (BLSFs) have attracted lots of attention in recent years for their potential applications in ferroelectric/piezoelectric devices used in the high-temperature and high-frequency environment [1,2,3], since they have many unique properties, such as high Curie temperature, low dissipation factor and excellent fatigue endurance property [4, 5]. But the major defect of BLSFs ceramics is the low piezoelectric activity which limits their applications. A general formula of BLSFs is (Bi₂O₂)²⁺(A_m−1B_mO3_m+1)²⁻ where A is mono-, di-, tri-valent ion (or a combination of them) allowing dodecahedral coordination, e.g., Ca²⁺, Sr²⁺, Ba²⁺, Pb²⁺, Bi³⁺, etc., B is a combination of transition metal cations suited for octahedral coordination, like Nb⁵⁺, Ta⁵⁺, Ti⁴⁺, W⁶⁺, Zr⁴⁺, etc., m is the number of octahedral layers in the perovskite slab and its value varies from 1 to 6.

Both calcium bismuth niobate (CaBi₂Nb₂O₉, CBN) and sodium bismuth niobate (Na_0.5Bi_2.5Nb₂O₉, NBN) are BLSF members with m = 2. CBN has a high Curie temperature (T_C=940 °C) but a low piezoelectric coefficient (d₃₃ = 5.8 pC/N) [6, 7], while NBN possesses a lower Curie temperature (T_C=790 °C) but a higher piezoelectric coefficient (d₃₃ = 17 pC/N) [8]. Hence, if (Na_0.5Bi_0.5)²⁺ can enter A site to substitute for Ca²⁺, the formed solid solution Ca_1−x(Na_0.5Bi_0.5)_xBi₂Nb₂O₉ (CNBN) may be expected to keep a high Curie temperature (T_C) as well as gain an ideal piezoelectric constant (d₃₃), and its properties could be adjusted by the change of (Na_0.5Bi_0.5) content. An experiment shows that d₃₃ increases but T_C decreases with the increase of (Na_0.5Bi_0.5) content from 0 to 1 carried out by X. Tian [9]. Especially, a specific composition of Ca_0.6(Na_0.5Bi_0.5)_0.4Bi₂Nb₂O₉ has a desirable T_C and moderate d₃₃. In addition, to improve piezoelectric activity, A-site or B-site cations modified BLSFs have been investigated. Especially, co-substitution of alkali metal cations (Li⁺, K⁺, Na⁺, etc.) and lanthanon cations (La³⁺, Ce⁴⁺, Nd³⁺, Pr³⁺, etc.) at the A-site are proved to be more effective [10,11,12,13,14,15,16,17,18,19].

In this work, LiCe-doped CNBN ceramics were prepared by the mixed oxide routes, which were aiming at attaining a high resistivity and a large piezoelectric activity as well as addressing the conduction mechanisms within BLSFs. The correlation between structural distortion and piezoelectric properties was also studied in detail to understand the effect of Li⁺ and Ce⁴⁺ co-substitution.

2 Experimental

Aurivillius type [Ca_0.6(Na_0.5Bi_0.5)_0.4]_1−x(LiCe)_x/2Bi₂Nb₂O₉ (CNBNLC-100x, x = 0, 0.02, 0.04, 0.06, 0.08, 0.10, 0.12) ceramics were prepared by a conventional solid-state reaction route. Bi₂O₃ (99%), CaCO₃ (99%), Na₂CO₃ (99.8%), Nb₂O₅ (99.5%), Li₂CO₃ (99.99%) and CeO₂ (99.99%) were used as starting raw materials. All powders were weighed according to the stoichiometric compositions and then mixed by ball milling in ethanol for 12 h. The mixtures were dried and then calcined at 850 °C for 3 h in an alumina crucible. The calcined powders were milled again in the same condition and granulated with polyvinyl alcohol (PVA) as a binder. The powders were finally pressed into disks of 10 mm in diameter and 1 mm in thickness by a steel die at a uniaxial pressure of about 10 MPa. After burning out the binder at 850 °C, the samples were sintered at 1080–1120 °C for 3 h in air. For the electrical measurement, silver electrodes were painted on both surfaces of the disks at 700 °C for 10 min. The samples were poled at 180 °C in a silicone oil bath for 20 min under dc electric field of 10–12 kV/mm and then cooled to room temperature.

The crystal structures of CNBNLC-100x ceramics were determined by X-ray diffractometer (DX2700, Dandong, China) employing Cu-K_α radiation (λ = 1.5418 Å) at a scan rate of 0.02°/s. The natural surfaces of sintered ceramics were observed by the scanning electron microscope (S-3400N, Hitachi, Japan). The temperature dependence of dielectric constant and dielectric loss of the ceramics were measured using a programmable furnace with an LCR analyzer (HP 4980A, Agilent, US). The piezoelectric constant d₃₃ were determined using a Berlincourt-type d₃₃ meter (ZJ-3A, Institute of Acoustics, Chinese Academy of Sciences, China). High-temperature electrical resistivity was tested by a resistance meter (Partul HRMS-900) at 10 kV from 200 to 710 °C. Besides, the thermal depoling experiments were conducted by annealing the poled samples for 2 h at each selected temperature, and measuring the d₃₃ value after they were cooled to room temperature.

3 Results and discussion

Figure 1 shows the XRD patterns of the CNBNLC-100x ceramics in the 2θ range of 20°–70°, which do not find obvious secondary phase. It indicates that the doping ions successfully diffused into the perovskite-like structure and formed a solid solution. As presented in Fig. 1(a) and (b), the diffraction peaks of CNBNLC-100x ceramics shift slightly to higher angles with the value of x varying from 0 to 0.06, which should be ascribed to the lattice shrinkage caused by the substitution of smaller Li⁺ and Ce⁴⁺ (Li⁺~ 0.76 Å, Ce⁴⁺~ 0.87 Å) [18] for larger Ca²⁺, Na⁺ and Bi³⁺ ions (Ca²⁺~ 1.34 Å, Na⁺~ 1.02 Å, Bi³⁺~ 1.36 Å) [20]. With LiCe content further increasing, the diffraction peaks shift gradually to the lower angles. Because the valent state of Ce cation changes from 4 + to 3 + and the larger Ce³⁺ diffused into the lattice. The shift of diffraction peaks implies that the LiCe doping leads to lattice distortion. In order to characterize the structures of CNBNLC-100x ceramics, XRD Rietveld refinements were carried out by using MAUD program [21]. Figure 2 shows the refined factors and lattice parameters of CNBNLC-100x ceramics. The refined factors indicate a good match between the observed and calculated patterns. Meanwhile, the atomic positions are listed in Table 1. According to the refinement result, perovskite unite, Bi₂O₂ layers structures and bond distances of CNBNLC-100x (x = 0, 0.04, 0.08, 0.12) ceramics are shown in Fig. 3. For CNBNLC-100x ceramics, the tilt angle along c axis of NbO₆ octahedron has a growing trend with x increasing from 0 to 0.08 (11° for CNBNLC-0, 13° for CNBNLC-4, 20° for CNBNLC-8) owing to the substitution of smaller ions (Li⁺, Ce⁴⁺) for large ions (Ca²⁺, Na⁺, Bi³⁺) in A-site. Then the tilt angle decreases with x further increasing to 0.12 (9° for CNBNLC-12) due to the variety of Ce valence [22]. In order to characterize the distortion of crystal structure, the sum of the squared deviations of bond lengths from the average is defined as structural distortions Δ [23]:

Fig. 1

XRD patterns of CNBNLC-100x ceramics in the 2θ range of 20–70°

Fig. 2

Refined factors and lattice parameters of CNBNLC-100x (x = 0, 0.04, 0.08, 0.12) ceramics

Table 1 Atomic coordinates of CNBNLC-100x (x = 0, 0.04, 0.08, 0.12) ceramics at room temperature

Fig. 3

NbO₆ octahedron and Bi₂O₂ layer crystal structure of a CNBNLC-0; b CNBNLC-4; c CNBNLC-8; d CNBNLC-12

$$\Delta =\frac{1}{Z}{\sum {\left( {\frac{{R - \bar {R}}}{{\bar {R}}}} \right)} ^2}$$

(1)

where Z is the coordination number, R is the individual bond length, and $\bar {R}$ is the average bond length. The structural distortion of CNBNLC-100x (x = 0, 0.04, 0.08, 0.12) ceramics are shown in Fig. 4. The distortion of the perovskite unit changes gradually but that of the Bi₂O₂ layer increases first and then decreases with the enhancement of LiCe contents. Both NbO₆ octahedron and Bi₂O₂ layer of CNBNLC-8 ceramics have the largest structural distortion.

Fig. 4

Structural distortion of CNBNLC-100x (x = 0, 0.04, 0.08, 0.12) ceramics

Figure 5 shows SEM images of the natural surface and the average grain sizes of CNBNLC-100x ceramics sintered at 1100 °C. The micrographs with the presence of well-packed and pore-free microstructure indicate that dense samples were achieved by this ceramic process [24]. Further, the average grain size values of CNBNLC-100x ceramics are shown in Fig. 5h. It can be clearly seen that the grain size of CNBNLC-8 ceramic is significantly bigger compared with other compositions. Its grain size is found to be 12 µm in length scale and 3 µm in thickness scale. The corresponding aspect ratio (l/t) is thus determined as 4. However, with increasing contents of LiCe, the aspect ratio (l/t) of samples decline significantly, implying that excess LiCe dopants delays the diffusion mechanisms of a-b plane and decreases the anisotropic feature of microstructure [25]. The element analysis of CNBNLC-8 ceramics is obtained by EDS shown in Fig. 5j. All elements are found except Li element. The Li atoms cannot be detected due to their rather light atomic mass. The nominal Ca/Ce ratio in CNBNLC-8 ceramics 13.8 while the measured Ca/Ce ratio is 14.45, which slightly deviates from the stoichiometry.

Fig. 5

SEM images of natural surface and the average grain sizes of CNBNLC-100x ceramics sintered at 1100 °C: a CNBNLC-0; b CNBNLC-2; c CNBNLC-4; d CNBNLC-6; e CNBNLC-8; f CNBNLC-10; g CNBNLC-12; h the average grain sizes of CNBNLC-100x ceramics; i The section of CNBNLC-8 ceramics; j the EDS analysis of CNBNLC-8 ceramics

Figure 6a and b presents the relative permittivity ε_r and dielectric loss tanδ measured at 1 MHz as a function of temperature for the CNBNLC-100x ceramics. The dielectric peak of ferro-paraelectric phase transition slightly shifts to a higher temperature at x = 0.04 (T_C=905 °C) and then shift to lower temperature with x increasing shown in inset Fig. 6(a). It is known that the value of Curie temperature T_C is correlated with the pseudo-perovskite distortion. In consequence, the orthorhombicity of the ceramics defined as 2(a − b)/(a + b), where the a and b are the lattice parameters [26]. The relationship between T_C and orthorhombicity is shown in Fig. 6c and both of them exhibit the same variation trend. In Fig. 6b, it is clear that the loss of CNBNLC-100x ceramics rises with the increase of LiCe contents at high temperature. This phenomenon indicates that the massive LiCe dopants could reduce resistivity.

Fig. 6

Temperature dependence of a dielectric constant (ɛ_r) and b dielectric loss (tanδ) of the CNBNLC-100x ceramics measured at 1 MHz; Inset (a) shows the change trends in T_C and structural orthorhombicity [2(a − b)/(a + b)] of the samples with an increase in the x value

The piezoelectric coefficient d₃₃ values of CNBNLC-100x ceramics are shown in Fig. 7a. The piezoelectric activity of the ceramics was improved by increasing LiCe content but decreased by excess LiCe. It can be clearly seen that d₃₃ value and structural distortion of CNBNLC-100x ceramics have the similar trend. The d₃₃ of CNBNLC-100x ceramics get the maximum value of 17.6 pC/N at x = 0.08, which is consistent with structural distortion. That is to say, the introducing of LiCe dopants induces structural distortion and leads to larger piezoelectric activity. In addition, it was reported that the domain motion is much easier in coarse grains than that in fine grains [25] which means the increasing grain size will help improve piezoelectric properties. It can be seen in Fig. 5h that CNBNLC-8 ceramics with the largest grain size gain the highest d₃₃ value. Figure 7b shows the thermal stability of relative d₃₃ of CNBNLC-100x ceramics and the inset Fig. 7b presents thermal stability of the piezoelectric coefficient d₃₃ of CNBNLC-100x ceramics. For CNBNLC-0 and CNBNLC-2 ceramics, the d₃₃ value changes slightly with the annealing temperature raising. The degradation of d₃₃ at low temperature is due to the thermally activated reverses switching of the non-180° ferroelectric domains. As seen clearly from these d₃₃-T curves, undoped CNBNLC-0 ceramics and the lowly-doped CNBNLC-2 ceramics have better thermal stability than those of highly-doped ones. The phenomenon could be attributed to the significant depoling of the space polarization (such as point defect, etc.) within doped BLSFs ceramics generated by the thermal activation [27, 28]. However, even after annealing at 600 °C for 2 h, the d₃₃ value of CNBNLC-8 ceramics could retain ~ 11 pC/N (62.5% of its original values), proving that the material is suitable for working at high temperatures.

Fig. 7

a The piezoelectric coefficient d₃₃ of CNBNLC-100x ceramics; b thermal stability of relative d₃₃ of CNBNLC-100x ceramics, annealed at various temperatures for 2 h

Besides piezoelectric activity, the electric conductivity is another key factor which should be considered for piezoceramics applied in high temperature. Figure 8a shows the resistivity of the CNBNLC-100x ceramics as a function of temperature from 200 to 500 °C. As the same as other Aurivillius type ceramics, the electrical conduction mechanism of CNBN ceramics is also a mixed p-type electrical conduction related to the volatilization of bismuth element during sintering. However, in the cooling process, oxygen atoms in the atmosphere tend to enter into the crystal lattices of ceramics again, generating the holes (h˙) by the compensating mechanism of unbalanced valence. When the high-valent donor Ce⁴⁺ substitutes for Ca²⁺ in CNBNLC-100x ceramics, there will be two redundant electrons created at A-site for per substitution. Thus the oxygen vacancies resulted from the bismuth volatilization can be eliminated by the neutralization of those electrons with the help of holes. Finally, the concentration of charge carriers in CNBNLC-100x ceramics decreases. Based on this conduction mechanism, the resistivity of CNBN ceramics at high temperature can be improved by doping Ce⁴⁺. The donor doping can be described by the following Kroger-Vink defect-chemical equations:

Fig. 8

a Resistivity of CNBNLC-100x ceramics as a function of measuring temperature; b Resistivity as a function of x for CNBNLC-100x ceramics at 200, 300, 400, and 500 °C

$${\text{Ce}}{{\text{O}}_{\text{2}}}+{\text{V}}_{{\text{O}}}^{{ \cdot \cdot }}+{\text{1}}/2{{\text{O}}_{\text{2}}}\xrightarrow{{{\text{C}}{{\text{a}}^{{\text{2}}+}}}}{\text{Ce}}_{{{\text{Ca}}}}^{{ \cdot \cdot }}+{\text{3}}{{\text{O}}_{\text{O}}}$$

(2)

The resistivity at different temperatures as a function of x for CNBNLC-100x ceramics is shown in Fig. 8(b), and the dc resistivity at 500 °C and 600 °C of all the sample have the same order of magnitude as shown in Table 2. The CNBNLC-2 ceramic reaches the highest resistivity of 4.82 × 10⁶ Ω cm at 500 °C and 5.33 × 10⁵ Ω cm at 600 °C among all the samples. The improvement in resistivity of CNBNLC-100x ceramics is attributed to the decreased concentration of charge carriers including holes and oxygen vacancies by doping LiCe. The CNBNLC-2 sample reaches the highest resistivity indicates its concentration of charge carriers has been decreased to the minimum value by doping LiCe with an appropriate content [29]. However, if the doping content of LiCe continually increases, the ceramics will be charged to an n-type semiconductor by the overmuch electrons created by the LiCe substitution for Ca²⁺, thus the resistivity exhibits a continuous downtrend as increasing from 0.04 to 0.12, which is described in Fig. 8b.

Table 2 Activation energy (E_a) obtained from the Arrhenius plots, the resistivity at 500 °C and 600 °C of CNBNLC-100x ceramics and the transition temperature T_change corresponding to the changes of E_a

Furthermore, the behavior of temperature dependent conduction follows the Arrhenius relationship:

$$\sigma ={\sigma _0}\exp ( - {E_{\text{a}}}/{k_B}T)$$

(3)

where σ₀ is the pre-exponential factor, E_a is the activation energy of the charge carriers, k_B is the Boltzmann’s constant and T is the absolute temperature. The Arrhenius plots of σ versus 1/T of the CNBNLC-100x ceramics are shown in Fig. 9. All conduction behavior can be described as two different temperature regions except CNBNLC-0 ceramics. According to the linearly fitting for the curve, the activation energy E_a is estimated to be from 1.21 to 1.36 eV at high temperature as shown in Fig. 9. The similar values of E_a, which seems to be independent of the LiCe dopant amount, which can further prove that the intrinsic conduction is predominant for CNBNLC-100x ceramics at high temperature. It has been reported that the intrinsic electronic conductivity activation energy is equal to half of the band gap E_g [30]. At low temperature, the activation energy is found to be 0.76–0.98 eV. It is close to the value of oxygen vacancies (~ 1 eV) involved in the extrinsic electrical conduction [31]. As a consequence, the conductivity mechanism at low temperature is considered as an ionic-electronic mixed conduction. In Fig. 9, it also shows that the transition temperature from extrinsic conduction to intrinsic conduction increases with increasing LiCe donor dopants. Because of the contribution from the electrons created by excessive Ce⁴⁺ amount, the low-temperature range dominated by the extrinsic conduction becomes wider [32].

Fig. 9

Arrhenius relationship between dc conductivity and temperature for the CNBNLC-100x ceramics

The temperature dependence of the σ_ac can be predicted by the dielectric properties of CNBNLC ceramics [33]:

$${\sigma _{ac}}=\omega {\varepsilon _0}{\varepsilon _r}\tan \delta$$

(4)

where the σ_ac is the total conductivity, where ω is angular frequency, ɛ₀ is the permittivity of vacuum, ɛ_r is the relative permittivity, tanδ is the dielectric loss. Figure 10 shows the σ_ac of CNBNLC-8 ceramics as a function of temperature at different frequencies. Obviously, the ac conduction behavior of the ceramics can be divided into three temperature regions at different frequencies. According to the slope of three fitting linear at the three regions, the values of activation energy E_a are calculated to be 0.1 eV, 0.6 eV, and 1.66 eV respectively. In region-I, the σ_ac values are weakly temperature dependent and with the increase of test frequency, the ac conductivity increases by the order of magnitude. The obtained value of activation energy (~ 0.1 eV) indicates that the free conduction electrons resulted from the first-order ionization of oxygen vacancies. In region-II, the σ_ac values increase rapidly above 460 °C at low frequency and all curves tend to merge. The observed E_a value (~ 0.6 eV) is very close to 0.7 eV which is the half of second-order ionization energy of oxygen vacancies. With further increasing temperatures (region-III, T > 690 °C), however, the E_a value (~ 1.66 eV) is equal to half the energy of the band gap, indicating that the conduction mechanism is referred to the intrinsic dc conductivity [34].

Fig. 10

Temperature dependence of ac conductivity of CNBNLC-8 ceramics at selected frequencies

Figure 11 shows the ac conductivity of CNBNLC-8 ceramic as a function of measured frequency between 450 and 650 °C. It is found that the ac conductivity can be separated into two regions. The σ_ac values remain constant approximately at low frequency region which implies that the dc conductivity dominates this process; Meanwhile, at high frequency region, the conductivity starts to increase non-linearly with increasing frequency. The frequency dependence of σ_ac can be described by Jonscher’s power law:

Fig. 11

a Frequency dependence of the ac conductivity of the CNBNLC-8 at several temperatures. b Temperature dependence of the exponent value of n. The inset shows temperature dependence of ɛ_r and tanδ during 400–700 °C of the CNBNLC-8 ceramics

$$\sigma (\omega )={\sigma _{dc}}+A{\omega ^n}$$

(5)

where the σ (ω) is the total conductivity, the σ_dc is the dc conductivity, the coefficient A and exponent n (0 < n < 1) are constants depend on temperature and material intrinsic property. After fitting curves at varies temperature, exponent n shows a minimum around 550 °C which is consistent with the temperature range of dielectric loss anomaly [35, 36].

4 Conclusions

In summary, the LiCe-doping CNBN-based ceramics have a promising application for piezoelectric sensors under high temperature. CNBNLC-8 ceramic has both of high piezoelectric coefficient (17.6 pC/N) and Curie temperature (904 °C) which are owing to the structural distortion and orthorhombic distortion of the lattice. Even annealing at 600 °C for 2 h, the d₃₃ value of CNBNLC-8 ceramics retained 62.5% of its original value. All compositions of CNBNLC-100x ceramics have high resistivity. The dc conductivity decreases with the amount of donor dopant increasing. With further increase of LiCe dopant, the dc conduction mechanism of the ceramics changes from p-type to n-type. The ac conduction mechanism of CNBNLC-8 ceramics below 690 °C is attributed to first-order ionization and second-order ionization oxygen vacancies and the main mechanism above Curie temperature is intrinsic dc conductivity.