III. RESULTS AND DISCUSSION

Figure 1 shows the XRD patterns of KNLN6–xBKT ceramics in the 2θ range of 20–70°. A stable solid solution among KNLN6 and BKT with a pure perovskite structure is formed in all compositions. The KNLN6 ceramic has a typical mixed phase of O–T, where the PPT temperature (T _O–T, see Fig. 3) is close to RT, which is in good agreement with the previous study.^{Reference Hao, Bai, Shen and Zhai20} As a certain BKT is added into KNLN6, the XRD patterns of BKT-modified samples keep almost unchanged, which seem to indicate the modified samples still maintain mixed phase of O–T. However, since the T _O–T is shifted down to below RT by the BKT addition as presented in Fig. 3, so we think T phase should be the RT dominant phase.^{Reference Cheng, Wu, Wang, Zhang, Zhu, Xiao, Wang, Lou and Liang21}

FIG. 1.

(a) XRD patterns of KNLN6–xBKT ceramics in the 2θ range of 20–70°. (b) The expanded XRD patterns of KNLN6–xBKT ceramics in the 2θ range of 42–49°.

Figure 2 provides the micrographs of KNLN6–xBKT (x = 0, 0.01, 0.03) ceramics microstructure observed on the (a–c) surface and (d–f) fracture surface of the compositions. It is found that all the indicated ceramics were well sintered and considerably dense (by applying the Archimedes method, all samples have a large relative density >96%) The average grain size decreases monotonically as the BKT content increases, implies that the role of BKT in KNN-based system is a grain growth inhibiter. Similar grain growth inhibition has also been reported for BNT or BKT-modified KNN-based ceramics.^{Reference Du, Zhou, Luo, Zhu, Qu, Li and Pei11,Reference Hao, Xu, Chu, Zhang, Chen, Li, Fu, Zang, Li and Yin17} In addition, it is noted that, for the pure KNLN6 ceramic, the grain sizes are not very homogeneous, which consists of coarse and fine grains, corresponding to about 10 and 2 µm, respectively. By contrast, BKT-modified samples have uniform grains with a relative small grain size, which is only about 3 and 2 µm, for samples with x = 0.01 and x = 0.03, respectively. For the microstructure observed on the fracture surface, it can be noted that the fracture surface is the typical transgranular fracture mode for the pure KNLN6 samples. Liquid phase sintering induced by LiNbO₃ should be responsible for the transgranular fracture. As BKT is added into KNLN6, the fracture surface becomes mainly intergranular fracture, which indicates that the BKT addition breaks the liquid phase sintering mechanisms.

FIG. 2.

The micrographs of KNLN6–xBKT (x = 0, 0.01, 0.03) ceramic microstructures observed on (a–c) the surface and (d–f) fracture surface of the compositions.

Figure 3 shows temperature dependence of dielectric permittivity ε_r and loss tangent tan δ for KNLN6–xBKT samples at a measuring frequency of 10 kHz. Two phase transitions, i.e., a Curie temperature T _C at the temperature of ∼460 °C and a polymorphic orthorhombic–tetragonal phase transition temperature T _O–T around RT are observed in pure KNLN6. With the addition of BKT, both the T _C and T _O–T are found to shift downward. For BKT-modified samples (x > 0.01), the T _O–T of the ceramics disappears (is shifted to below RT), and only the cubic–tetragonal phase transition is observed above RT, as shown in the inset of this figure. In this work, the addition of BKT in the ceramics decreases the T _C and shifts the PPT below RT. It is reasonable to believe that the temperature stability of KNLN6–xBKT ceramics could be improved due to the disappearance of PPT.^{Reference Zhang, Xia, Hao, Liu and Shrout13,Reference Wu, Xiao, Wang, Wu, Zhang, Li and Zhu15,Reference Hao, Xu, Chu, Zhang, Chen, Li, Fu, Zang, Li and Yin17} Moreover, it is found that the dielectric maxima drops down rapidly, and the dielectric peaks become extremely broad with the addition of BKT, which indicates that the relaxor behavior is enhanced by the addition of BKT. In addition, the dielectric loss tangent (tan δ) of the BKT-modified samples is lower than 5.5%, even at a temperature as high as 400 °C.

FIG. 3.

Temperature dependence of dielectric permittivity ε_r and loss tangent tan δ for KNLN6–xBKT samples at a measuring frequency of 10 kHz (inset figure gives the temperature dependence of ε_r in the temperature range of −100 °C to RT).

To further characterize the dielectric diffuseness and the relaxor behavior of the phase transition of KNLN6–xBKT system, quantitative characterizations as described in the following have been done. It is known that for a normal ferroelectric, the dielectric constant above the Curie point follows the Curie–Weiss law described by:

(1)$${\varepsilon _{\rm{r}}} = C{(T - {T_{\rm{0}}})^{ - 1}}(T > {T_{\rm{0}}})\quad ,$$

where C is the Curie–Weiss constant, and T ₀ is the Curie–Weiss temperature. The Curie–Weiss constant C reflects the nature of the ferroelectric transition. In KNLN6–xBKT system, all C are 10⁵ °C order, indicating that the high-temperature paraelectric phase is driven by a displacive transition.^{Reference Nakamura, Mitsui and Furuichi22} To make the value of T ₀ certain, the plot of the inverse dielectric constant versus temperature at 10 kHz is provided in Fig. 4(a). It is found that ε_r of all samples deviates from the Curie–Weiss law above the Curie temperature, and as the BKT content increases, the deviation ΔT _m as defined by the following [Eq. (2)] increases continually as shown in the inset of Fig. 4(a),

(2)$$\Delta {T_{\rm{m}}} = {T_{{\rm{CW}}}} - {T_{\rm{m}}}\,\left( {^\circ {\rm{C}}} \right)\quad ,$$

where T _CW represents the temperature from which ε_r starts to follow the Curie–Weiss law, and T _m denotes the temperature at which ε_r value reaches the maximum. This result indicates that the diffuse phase transition behavior of KNLN6 is obviously enhanced by BKT addition, which corresponds to the broader and diffused dielectric peak in BKT-modified samples.

FIG. 4.

(a) Inverse dielectric permittivity at 10 kHz as a function of temperature for KNLN6–xBKT ceramics. (b) Plot of log(1/ε − 1/ε_m) as a function of log (T − T _m) for KNLN6–xBKT ceramics at 10 kHz [symbol: experimental data; solid line: fitting to Eq. (3)].

Furthermore, a modified empirical expression^{Reference Uchino and Nomura23} was used in the present work to further characterize the dielectric dispersion and diffuseness of the phase transition.

(3)$${1 \mathord{\left/ {\vphantom {1 {{\varepsilon _{\rm{r}}}}}} \right. \kern-\nulldelimiterspace} {{\varepsilon _{\rm{r}}}}} - {{1 \mathord{\left/ {\vphantom {1 {\left( {{\varepsilon _{\rm{r}}}} \right)}}} \right. \kern-\nulldelimiterspace} {\left( {{\varepsilon _{\rm{r}}}} \right)}}_{\rm{m}}} = {{{{(T - {T_{\rm{m}}})}^\gamma }} \mathord{\left/ {\vphantom {{{{(T - {T_{\rm{m}}})}^\gamma }} C}} \right. \kern-\nulldelimiterspace} C}\quad ,$$

where γ and C are all constants, and the γ value is between 1 and 2. The limiting values γ = 1 and γ = 2 obey the equation to Curie–Weiss law, which are the character for the case of normal ferroelectric and for an ideal relaxor ferroelectric, respectively. The logarithmic plots related to this equation for the KNLN6–xBKT samples are shown in Fig. 4(b). A linear relationship is obtained for these samples. The slope of the fitting curves used to determine the γ value of the KNLN6–xBKT ceramics increases gradually with the increasing BKT content, and the fitting value γ reaches 1.878 for samples with x = 0.05, suggesting the occurrence of a typical relaxor behavior. The enhanced relaxor behavior in BKT-modified KNLN6 should be ascribed to the disorder in the perovskite unit cell induced by BKT substitutions.^{Reference Guo, Kakimoto and Ohsato24}

Figure 5 gives the complex impedance spectra of the KNLN6–xBKT ceramics with x = 0, 0.01, 0.03, and 0.05 in the frequency range of 0.01 Hz–20 MHz over 350–500 °C. Commonly, electrode, grain boundary, and grain components contribute to the conductivities of the polycrystalline materials, and their effects on conduction can be segregated into three parallel RC elements connected in series in an equivalent circuit, from low frequency to high frequency.^{Reference Irvine, Sinclair and West25} For the KNLN6–xBKT system, the Debye-like semicircles at variable temperatures become distorted, and the distorted semicircles at each temperature can be segregated into two natural electrical components corresponding to the two fitting semicircles in the impedance plots [Figs. 5(a)–5(f)]. The bigger semicircle in the higher frequency range is ascribed to the grain effect, while the grain boundary response contributes to the smaller semicircle in the lower frequency range.^{Reference Atamanik and Thangadurai26}

FIG. 5.

(a)–(d) The complex impedance spectra of the KNLN6–xBKT ceramics with x = 0, 0.01, 0.03, and 0.05 in the frequency range of 0.01 Hz–20 MHz over 350–500 °C. (e), (f) The fitting semicircles in the impedance plots of the KNLN6–xBKT ceramics (x = 0 and 0.05).

Figure 6 represents the imaginary part of impedance as a function of frequency at different temperatures, for the KNLN6–xBKT ceramics. It is worth noting that two peaks appeared in the imaginary part of impedance spectrum for all compositions, which correspond to the two Debye-like semicircles. This result further confirms that two components (grain boundary and grain) contribute to the conductivities of the materials. In addition, the appearance of the peaks in the spectrum is an indication of the beginning of electrical relaxation in the material at different temperatures. The peaks that appear are asymmetric, and their positions appear to be shifting toward the higher frequency side with increasing temperature. The asymmetric broadening of the peaks confirms the presence of electrical processes in the material with a difference of relaxation time.^{Reference Boukriba, Sediri and Gharbi27} Furthermore, it can also found that as the temperature reaches up to a certain value, the peak remains unmoving or even shifts to the lower frequency side. This can be ascribed to the cubic–tetragonal phase transition near the Curie temperature at which the relaxation time increases.^{Reference Cheng, Shi, Qi, Chen and Yu28}

FIG. 6.

The imaginary part of impedance as a function of frequency at different temperatures, for the KNLN6–xBKT ceramics.

Moreover, the total conductivities (σ) of the ceramics can be detected from the complex impedance data, and variable-temperature conductivities of the KNLN6–xBKT ceramics are shown in Fig. 7(a). It is observed that the BKT substitution leads to the increased conductivities of the KNLN6 ceramics. Besides, the activation energy (E _a) for electrical conduction can be detected by using the Arrhenius equation:

(4)$$\sigma = {\sigma _0}\exp \left( {{{ - {E_{\rm{a}}}} \mathord{\left/ {\vphantom {{ - {E_{\rm{a}}}} {kT}}} \right. \kern-\nulldelimiterspace} {kT}}} \right)\quad ,$$

where σ and σ₀ are the dc conductivity and the pre-exponent constant, respectively, E _a is the dc activation energy, k is the Boltzmann's constant, and T is the absolute temperature. The values of E _a of the KNLN6–xBKT ceramics are in the range of 0.79–1.09 eV, which are closely related to the oxygen ion/vacancy migration.^{Reference Li, Fan, Long, Dong and Sun29} From Fig. 7(a), it is clear that the introduction of the BKT causes a decrease in E _a, and the decreased E _a of the BKT-substituted KNLN6 ceramics should be attributed to an increase in the number of oxygen vacancies in the lattice positions due to replacement of Nb⁵⁺ by Ti⁴⁺ at the B-site, which can be described as: $2{\rm{TiO}}_2 \buildrel {{\rm{Nb}}_2 {\rm{O}}_5 } \over \longrightarrow 2{\rm{Ti\prime}}_{{\rm{Nb}}} + {\rm{\ddot V^{\bullet\bullet}}}_{\rm{o}} + 4{\rm{{\rm O}}}_{\rm{o}}$.

FIG. 7.

(a) Arrhenius plots of dc electrical conductivity (σ) of KNLN6–xBKT ceramics with various BKT amounts. (b) Temperature dependence of logarithmic conductivity for KNLN6–xBKT (x = 0, 0.05) ceramics in the temperature range of 300–500 °C.

Figure 7(b) shows the logarithmic variable-temperature conductivity of the KNLN6–xBKT (x = 0, 0.05) ceramics in the temperature range of 300–500 °C. The curve could be fitted with two linear lines at low- and high-temperature ranges. The obtained corresponding activation energies E _a of KNLN6 were 0.60 and 1.09 eV for two ranges. The activation energy 0.60 eV for the dc electrical conductivities suggested an extrinsic conduction mechanism, which was most probably p-type conduction. As the temperature increased, the activation energy increased to 1.09 eV, which was associated with a change from extrinsic to intrinsic conductivity.^{Reference Zhang, Yan and Reece30}

Figure 8 shows the (a) P–E hysteresis loops, (b) bipolar S–E curves, (c) ε_r–E curves, (d) unipolar S–E curves, and (e) I–E curves of KNLN6–xBKT ceramics as a function of BKT content, measured at RT and 10 Hz. A typical ferroelectric loop has been demonstrated in all samples. The P–E and bipolar S–E loops present fleet changes of polarization and strain in the region near the coercive field E _c [Figs. 8(a) and 8(b)], which is the electric field corresponding to the current peak [Fig. 8(e)], and the current peak is coincident with the fields corresponding to P = 0, suggesting that all samples are under complete saturation conditions.^{Reference Viola, Saunders, Wei, Chong, Luo, Reece and Yan31,Reference Yan, Inam, Viola, Ning, Zhang, Jiang, Zeng, Gao and Reece32} With increasing BKT content, the loops of the ceramics become much narrower, suggesting that BKT addition hinders the domain-wall switching,^{Reference Cheng, Wu, Wang, Zhang, Zhu, Xiao, Wang, Lou and Liang21} evident by the much broader current peak (domain switching peak). Change of the shape in P–E loops from “well saturated” to “slim” implies that the domain structure of the ceramics undergoes a macrodomain–microdomain transition with the addition of BKT,^{Reference George33} which also confirms the appearance of a typical relaxor behavior at high content of BKT.^{Reference Guo, Kakimoto and Ohsato24,Reference Tan and Viehland34} The normal to relaxor ferroelectric transformation induced by BKT can also be well confirmed by the bipolar S–E curves. For samples with low-BKT content (x = 0, 0.01), they show “butterfly”-shaped strain hysteresis loops with relative large negative strain S _neg, which is typical for normal ferroelectrics. With further increasing BKT content, the S–E loop exhibits the so-called “sprout shape” (with a diminished S _neg), recently defined for relaxor lead-free ceramics related to the presence of near nonpolar phases at zero electric field.^{Reference Viola, Saunders, Wei, Chong, Luo, Reece and Yan31} For the unipolar strain curve as shown in Fig. 8(d), all samples show small hysteresis, which indicates both the piezoelectric effect, and the domain wall motions contribute to the strain level. The large signal d ₃₃* (S _max/E _max) determined from the unipolar strain curve deceases linearly from 300 pm/V to 127 pm/V with the addition of BKT, implying that the addition of BKT weakens the piezoelectricity of the ceramics. The in situ measurements of relative permittivity ε_r as a function of bias field are shown in Fig. 8(c), compositions with x ≤ 0.01 exhibit notable hysteretic behavior, which is closely connected to switching of ferroelectric domains. By contrast, xBKT samples (x > 0.03) display indiscernible hysteretic behavior, also confirming BKT addition hinders the domain-wall switching.^{Reference Dittmer, Anton, Jo, Simons, Daniels, Hoffman, Pokorny, Reaney and Rödel35}

FIG. 8.

(a) P–E hysteresis loops, (b) bipolar S–E curves, (c) ε_r–E curves, (d) unipolar S–E, (e) I–E curves of KNLN6–xBKT ceramics as a function of BKT content, measured at RT and 10 Hz.

Since the addition of BKT in the ceramics shifts the PPT below RT. It is expected that the temperature stability of KNLN6 ceramics could be improved by BKT doping. Figure 9 shows the P–E (insets: I–E loops) hysteresis loops and unipolar S–E curves of KNLN6–xBKT ceramics with x = 0, 0.02, and 0.04 at different temperatures (25–150 °C). For pure KNLN6 samples, P–E loops change notably with elevating measuring temperature, and both the remnant polarization P _r and unipolar strain S exhibit a linear decrease in this temperature range. By contrast, for 0.02BKT samples, the P–E loops keep a typical ferroelectric shape with elevating measuring temperature. Meanwhile, samples have a small P _r and S deviations in the temperature range. For the 0.04BKT composition, no significant deformation of the P–E loops and unipolar strain curves is observed in the whole range of the measuring temperatures. These results indicate that the modified KNLN6 ceramics exhibit relatively good temperature stability in the measured temperature range.^{Reference Hao, Xu, Chu, Zhang, Chen, Li, Fu, Zang, Li and Yin17,Reference Wu, Xiao, Wu, Sun, Lin, Zhu, Yu, Zhuang and Wei36} The results can be seen more clearly in Fig. 10, which gives the remnant polarization P _r and large signal d ₃₃* (S _max/E _max) as a function of temperature for pure and modified KNLN6 materials. KNLN6 ceramics showed a higher P _r and d ₃₃* at RT. In the case of BKT-modified samples, despite the lower initial P _r and d ₃₃* at RT, they demonstrate relatively stable behavior with temperature. The P _r and d ₃₃* at 150 °C relative to those at RT for samples of x = 0.03–0.05 are all above 90%, while the two values for unmodified ceramic are both only about 35%.

FIG. 9.

The P–E hysteresis loops (insets: I–E loops) and unipolar S–E curves of KNLN6–xBKT ceramics with x = 0, 0.02, and 0.04 at different temperatures (25–150 °C).

FIG. 10.

(a) The remnant polarization P _r and (b) large signal d ₃₃* (S _max/E _max) as a function of temperature for KNLN6–xBKT ceramics.

The above results show that the temperature stability of KNLN6 ceramics has been greatly improved by the addition of BKT due to phase transition from O–T mixed phase to a single tetragonal phase and the shift of PPT down to below RT. It further reveals that samples with a single tetragonal phase favor good resistance to thermal cycles. To further study the temperature stability of the system and give a detailed understanding of temperature resistance mechanism, we chosen two compositions, KNLN6 with O–T mixed phase and 0.04BKT-modified samples with single T phase, and carefully studied the phase transition behavior induced by the temperature. Figures 11(a) and 11(b) give the XRD patterns measured at different temperatures (20–500 °C) of the two samples. As shown in this figure, the XRD patterns of KNLN6 ceramics show the coexistence of orthorhombic and tetragonal phases at 20 °C, tetragonal phase at 80–450 °C, and cubic phase at 500 °C. For 0.04BKT-modified samples, a stable tetragonal phase exists until 300 °C, following a tetragonal–cubic phase transition at 400 °C. These phase transition points reflected by the high-temperature XRD measurements are consisted with that determined by the dielectric permittivity peaks, as disclosed in Fig. 3. Figure 11(c) gives the variation of the tetragonality c/a with temperature, for KNLN6 and 0.04BKT-modified samples. It is evident that the c/a ratio of 0.04BKT-modified samples keeps a stable value of 1.013 in the temperature range from RT to 300 °C, and then it drops rapidly to 1 near the Curie temperature. By contrast, KNLN6 material shows an obvious decease of c/a ratio in the whole temperature range from RT to 500 °C. Based on the above, it can be concluded that the excellent temperature stability of BKT-modified properties should be attributed to the stable tetragonal phase over a wide temperature range.

FIG. 11.

XRD patterns of KNLN6–xBKT ceramics with (a) x = 0 and (b) x = 0.04 in the 2θ range of 20–70° at different temperatures. (c) Variation of the tetragonality c/a with temperature, for KNLN6 and 0.04BKT-modified samples.

Suitable materials for piezoelectric applications not only show excellent piezoelectric properties and good temperature stability but also have to be reliable under long-term electric cycling.^{Reference Yao, Glaum, Wang, Jo, Rödel and Li37} For electromechanical applications, such as sensors or ultrasonic devices, a high-piezoelectric coefficient d ₃₃ is often the desired quality. For actuator applications, however, the normalized strain d ₃₃* is a more important feature.^{Reference Yao, Glaum, Wang, Jo, Rödel and Li37–Reference Jo, Dittmer, Acosta, Zang, Groh, Sapper, Wang and Röodel39} However, for KNN-based materials, this property and its fatigue behavior did not get much attention so far,^{Reference Zhang, Xia, Hao, Liu and Shrout13,Reference Yao, Wang and Li14,Reference Yao, Glaum, Wang, Jo, Rödel and Li37} even though it is a crucial aspect for industrial implementation. Figure 12 shows the unipolar fatigue behavior of (a) KNLN6 and (b) 0.05BKT-modified ceramics for 10⁵ cycles, and the unipolar strain levels at different unipolar fatigue cycles are summarized in Fig. 12(c). It can be observed that the KNLN6 material exhibits obvious fatigue behavior, whereupon the unipolar strain is found to decrease by 50% after 10⁵ switching cycles. For the BKT-modified materials, however, no fatigue behavior up to 10⁵ switching cycles is observed.

FIG. 12.

The unipolar fatigue behavior of (a) KNLN6 and (b) 0.05BKT-modified ceramics for 10⁵ cycles, (c) the unipolar strain levels of the two compositions at different unipolar fatigue cycles.

To understand the behavior of the unipolar strain during electrical cycling more in detail, bipolar hysteresis loops of polarization P, strain S, permittivity ε_r, and piezoelectric coefficient d ₃₃* at different fatigue cycles were recorded. The evolution of polarization P–E, bipolar strain S–E, d ₃₃*−E and ε_r−E hysteresis loops of KNLN6 and 0.05BKT-modified ceramics is displayed in Fig. 13. Similar to the fatigue behavior reflected in the unipolar strain loops, pure KNLN6 ceramics exhibit significant fatigue behavior, the remnant polarization P _r is found to decrease by ∼45% after 10⁵ bipolar fatigue cycles, with a corresponding decrease in bipolar strain S levels, as depicted in Figs. 13(a) and 13(b). Furthermore, d ₃₃*−E and ε_r−E hysteresis loops provided in Figs. 13(c) and 13(d) also show an obvious degradation with bipolar electric cycling. By contrast, KNLN6 ceramics with 0.05BKT addition exhibit fatigue-free behavior: no decrease of the P _r and the corresponding S, d ₃₃*, and ε_r values are observed as represented in Figs. 13(e)–13(h), even as the switching cycles reaching up to 10⁵. In the present work, 90° domain switching occurs much easier in the KNLN6 ceramics when compared to the BKT-modified samples, owing to the coexistence of tetragonal and orthorhombic phases.^{Reference Zhang, Xia, Hao, Liu and Shrout13,Reference Li, Rogan, Ustundag and Bhattacharya40} It is widely accepted that bipolar fatigue generates microscopic defects and/or defect clusters in bulk ceramics, effectively pinning 90° domain wall motion, resulting in the loss of switchable polarization during continuous electric field switching.^{Reference Zhang, Xia, Hao, Liu and Shrout13} However, for 0.05BKT-modified ceramics, 90° domain switching occurs much harder. As a result, the polarization is found to maintain the same value with increasing switching cycles, and no fatigue is observed in the modified KNLN6 materials.

FIG. 13.

The evolution of (a), (e) polarization P–E, (b), (f) bipolar strain S–E, (c), (g) d ₃₃*−E, (d), (h) ε_r−E hysteresis loops with different fatigue cycles for KNLN6 and 0.05BKT-modified ceramics.