Effect of the sintering technique on the ferroelectric and d33 piezoelectric coefficients of Bi0.5(Na0.84K0.16)0.5TiO3 ceramic

In the search of lead-free piezoelectric materials, ceramic processing techniques offer potential tools to increase the piezoelectric and ferroelectric properties in addition to new chemical compositions. Powders of pure BNKT16 (Bi0.5(Na0.84K0.16)0.5TiO3) phase were synthesized by sol-gel method with a low crystallization temperature (750 °C). Ceramic samples were sintered by pressureless sintering (PLS), sinter-forging (SF), and spark plasma sintering (SPS) techniques. Structural, morphological, and chemical characterizations were performed by XRD, Raman, EDS, and SEM. Sintered samples by PLS and SF exhibit rod-like grains associated to bismuth volatility. The highest remanent polarization (11.05 μC/cm2), coercive field (26.2 kV/mm), and piezoelectric coefficient (165 pC/N) were obtained for SF sample. The piezoresponse force microscopy (PFM) analysis shows that the crystallites at the nanoscale exhibit piezoelectric phenomenon and the highest piezoelectric response is reported for PLS sample. The presence of the rhombohedral phase, the increase in grain and crystallite size, and the oriented rod-like inclusions favoring the crystallographic texture are facts that enhance the piezoelectric coefficient for BNKT16 piezoceramics.


Introduction
The production of piezoelectric materials rapidly developed in the late 1950s due to the number of discovered piezoelectric ceramics, for example, the lead niobate and lead zirconate titanate (PZT) piezoelectric ceramics [1]. The PZT with high electromechanical coupling factor implies a high level of production with contaminants in the form of lead oxide (PbO) or lead zirconate titanate. This mainly occurs during calcination and sintering, where the PbO evaporates. As a consequence, in 2003 the European Union included PZT ceramics in the legislature as a toxic substance that needs to be replaced by safer materials [2]. Lead-free piezoelectric ceramics exhibit superior electromechanical responses near the morphotropic phase boundary (MPB). It is reported that their high piezoelectric response is due to the phase coexistence of rhombohedral and tetragonal phases at the optimal composition, i.e., the MPB [3][4][5][6][7].
NBT-based ceramics is a very intense field of research looking for lead-free piezoelectric ceramics [8,9]. Among them, the systems Bi 0.5 Na 0.5 TiO 3 (BNT) and Bi 0.5 K 0.5 TiO 3 (BKT), are two examples that have received considerable attention due to their excellent ferroelectric and piezoelectric properties near their rhombohedral-tetragonal composition (MPB) [5,10]. Bismuth sodium potassium titanate (BNKT) has been synthesized by sol-gel showing excellent dielectric, piezoelectric, and ferroelectric properties [5]. The synthesis of BNKT by chemical methods such as sol-gel has shown to be more effective because it is performed at a lower temperature than the solid-state reaction method, the microstructure is most homogeneous, and a better morphology control is achieved. Increasing grain size translates into excellent dielectric and piezoelectric properties [11]. The sinter-forging technique in ceramic materials induces a crystallographic texture with a preferred orientation [12]. The textured bismuth titanate (BiT) has better electrical properties along axis, i.e., in the direction of the preferred orientation [13]. Spark plasma sintering (SPS) allows very fast heating rates, very short holding time, and the possibility to obtain high dense samples at lower temperatures than the most sintering techniques [14]. Therefore, it is expected that the application of SPS technique in BNKT ceramics could suppress the volatilization of Bi during sintering [15,16]. In the present study, the synthesis of 0.84(Bi 0.5 Na 0.5 TiO 3 )-0.16(Bi 0.5 K 0.5 TiO 3 ) (BNKT16) was done using the sol-gel technique. The attention is focused on the effects of the pressureless sintering (PLS), sinter-forging (SF), and spark plasma sintering (SPS) techniques on structural and electrical properties.

Experimental procedures
The precursors used in the present work were analytical grade bismuth nitrate pentahydrate (Bi(NO 3  For PLS green sample preparation, the BNKT powders were pulverized with a mortar and pestle adding 3 wt% aqueous polyvinyl alcohol (PVA) and pressed into pellets with a diameter of 6 mm under a uniaxial force of 6 kN. Sintering of the green body was performed at 1150 ℃ for 3 h in a Carbolite VST 1700 furnace. Bismuth volatilization restricts the range to select an effective sintering temperature. For a lower temperature, sintering is not effective enough and, for a higher one, bismuth volatilization is too high [5,17]. In general, the effect of volatile elements for lead-free piezoelectric ceramics is an issue of interest [18]. In the case of SF sample, the green body was placed between two platinum plates to prevent the sample reaction with the alumina pistons. After conventional firing (1150 ℃ ) for 1 h, the uniaxial stress was gradually applied at a 10 N/min rate until reaching a constant value of 7073.55 Pa (200 N). It was held for 1 h. Pressing was performed with an Instron 5967 tester system coupled to the furnace. Then, the applied stress was released and the 1150 ℃ temperature was held for 1 h annealing. The SPS process was carried out applying 500 A, pulse switch on/off 12/2, and 4 kN load on a 10 mm diameter graphite die in a Dr. Sinter facility. The densities were estimated using the Archimedes' method (ASTM C373).
The X-ray diffraction (XRD) patterns of the heattreated samples were recorded on a PANalytical X'pert PRO powder X-ray diffractometer equipped with an X´Celerator detector. A monochromatic X-ray source of Cu Kα radiation (λ ≈ 1.5406 Å) was used with 40 kV/30 mA of power. The XRD patterns were collected in the range of 10°-100° using the step-scanning mode with a step of 0.017° in 2θ and 60 s of step counting time. For the SPS sample, lattice parameters and the atomic positions were determined from the XRD patterns by the Rietveld refinement method using the FullProf Suite (version 2017) software package [19]. For the refinement, the following parameters were considered: (i) scale factor; (ii) zero displacement correction; (iii) unit cell and background parameters; (iv) peak profile parameters (full width at half maximum, FWHM and shape parameters), using a pseudo-Voigt function including the peak asymmetry.
The crystallite size was estimated by Scherrer's equation. For the morphological characterization, BNKT16 ceramics were sliced transversally, polished, and thermally etched at 1050 ℃ for 1 h. Then, they were analyzed by scanning electron microscopy (SEM) in a Hitachi SU 5000 equipped with energy dispersive spectroscopy (EDS), and the average of the grain size was estimated from SEM micrographs by an imageprocessing analyzing program (ImageJ 1.50i, National Institutes of Health, USA). Raman spectrums were performed with a WITec Focus Innovation.
Hysteresis loops were obtained using a Radian Technologies Precision Workstation (model P-WS) and high-power amplifier TREK (model 609A, Albuquerque, New Mexico, USA), with an applied electric field of 70 kV/cm over the pellets. Samples were polarized in a silicone oil bath at room temperature by applying a DC field of 40 kV/cm for 30 min. The piezoelectric constant d 33 was measured using a piezo d 33 meter (Pennebaker model 8000). The local piezoresponse was examined using switching spectroscopy piezoresponse force microscopy (SS-PFM) using an AFM system (MFP-3D, Asylum Research, Oxford Instruments).
Details of the SS-PFM experimental procedure can be found elsewhere [20]. Figure 1 shows the XRD pattern for the BNKT16 powders obtained by sol-gel. The XRD pattern is associated to the heat-treated powders obtained at 750 ℃. It suggests a perovskite structure with a single pseudo-cubic phase. It is known that the A sites are occupied by Bi, Na, and K [1]. Figure 2 shows the XRD patterns and the Rietveld refinement results for the BNKT16 SPS, PLS, and SF samples where the coexistence of tetragonal and rhombohedral phases of the BNKT, labeled as "t" and "r" respectively is seen. Herein, for the Rietveld refinement, the coexistence of the rhombohedral and tetragonal phases is considered as it was previously reported by Kreisel et al. [21]. Table 1 summarizes the lattice parameters, the phase volume fraction obtained from Rietveld refinement process for the compositions obtained by the three different sintering techniques, and each global χ 2 adjustment quality parameter. Table   2 reports the atom normalized coordinates, occupancy, and the Wyckoff positions for each spatial group. The atomic positions for the tetragonal phase with P4mm spatial group were refined for the SPS sample, and the atomic positions for the tetragonal P4bm and rhombohedral R3c phases were refined for the PLS sample. The BNKT16-R3c is also present for SPS and SF, BNKT16-P4bm for SF. However, it is not necessary to refine the atomic position again, because no improvement is obtained for χ 2 adjustment parameter.    For the SPS sample, the split peak at 2θ ≈ 46° and the unique peak at 2θ ≈ 40° can be observed as evidence of the tetragonal BNKT phase ((001) T and (100) T peaks, P4mm space group, C 4v , Fig. S1(a) in the Electronic Supplementary Material (ESM)) [22]. The other way around (unique peak for 2θ ≈ 46° and split peak for 2θ ≈ 40°) is a sign of the BNKT rhombohedral phase ((006) R and (202) R peaks, Fig.  S1(c) in the ESM). For the SPS sample, the preponderant phase is the tetragonal (97.44%) as can be observed in Fig. 2(a) and reported in Table 1. For the sample obtained by the conventional PLS method, it can be noticed that the split peaks are located between 39° and 41°, which suggests the presence of a rhombohedral phase with a R3c space group (C 3v , Fig.  S1(c) in the ESM) as can be expected for this composition [23]. The Rietveld adjustment suggests that the rhombohedral is the preponderant phase with a concentration of 85.59% (Table 1). The XRD pattern for the sample obtained by SF method maintains the same split peaks for 39° < 2θ < 41°. It can also be pointed out that these split peaks are better defined for SF than for PLS. It seems that split peaks are also present for 22° < 2θ < 24°, which may suggest the presence of the tetragonal phase. Therefore, the coexistence of rhombohedral and tetragonal phases [24] can be expected. In fact, the Rietveld adjustment is reporting the coexistence of tetragonal (2.48%) and rhombohedral (97.52%) phases. As can be noticed, the rhombohedral fraction increases in the following sequence: SPS, PLS, and SF. For the PLS and SF samples, the best adjustment is obtained considering the space group P4bm instead of P4mm, which also corresponds to the tetragonal crystal system. Figure 3 shows the Raman spectra analysis of the BNKT16 powders and the sintered samples obtained by PLS, SF, and SPS methods. Raman analysis for all samples considered the criterion of Gaussian peak fitting. Table 3 reports the wave number of each Raman fitted peak shown in Fig. 3 and its relative intensity. The Raman active modes can be estimated with the help of the Bilbao crystallographic server [25,26]. The presence of 12 active modes are predicted for the BNKT tetragonal phase P4mm, Γ Raman = 5A 1 + B 1 + 6E ((S.3) in the ESM); 31 active modes for the BNKT tetragonal phase P4bm, Γ Raman = 7A 1 + 3B 1 + 6B 2 + 15E ((S.6) in the ESM); and 19 active modes for the BNKT rhombohedral R3c, Γ Raman = 6A 1 + 13E ((S.9) in the ESM).

2 Raman spectroscopy
Then, four regions can be described in the Raman spectra, which are: 100-180, 180-430, 430-680, and 680-930 cm -1 [25]. The first band region, in the range of 100-180 cm -1 describes the A-site vibrations associated to the Bi-O, Na-O, and K-O bonds which is dominated by an intense band centered at 125 cm -1 assigned to a Raman active mode A 1 . According to Kreisel et al. [21], who studied a similar composition BNT-BKT, the Raman shift observed in this band is  sensitive to the pressure conditions and by another type of cation substitution. Almost no difference can be observed between the powder sample and the sintered ones. The second region is located in the range of 180-430 cm -1 and is attributed to the vibration of the Ti-O bond. The peak centered at 265 cm -1 (Raman active mode A 1 ) is associated to the tetragonal phase in the perovskite structure. Deconvolution shows one peak for the powder sample and two peaks for the sintered ones. It can also be observed in Table 3 that peak shift wave number is sensitive to the sintered technique. It means that in all samples, there is a coexistence of tetragonal phase with rhombohedral phase as it was discussed in the XRD section. The third region, located between 430 and 680 cm -1 , is attributed to the oxygen displacements and is interpreted in terms of the polyhedral TiO 6 vibrations [21]. The peak at 532 cm -1 for the powder sample shifts to values that run from 510 to 558 cm -1 , then this peak turns out to be sensitive to the sintered technique. The broad band observed in the fourth region could be assigned to the A 1 (LO) and E(LO) overlapping bands [27]. The peak analysis interpretation for the PLS sample suggests an important contribution of the rhombohedral phase over the tetragonal phase. The spectral shape of the SF sample shows a peak broadening in the second region that could be attributed to a splitting into two Raman bands that are associated to an important presence of a tetragonal phase as discussed in the XRD section. According to previous reports [28,29], other Raman bands appear when a tetragonal symmetry is more evident as noticed in the XRD patterns for the PLS and SPS methods. The SPS spectral shape shows a broadened peak in the second region. This can be interpreted like an important contribution from a tetragonal phase. The variation of the Raman spectrum shape for this compound obtained by different processes is a clear demonstration of changes in the contribution of rhombohedral and tetragonal symmetry in the structure. In the region of 680-930 cm -1 , the presence of broader peaks is related to the oxygen vacancy presence [27]. Figure 4 shows the secondary electron-SEM micrographs of the BNKT16 sintered by (a) PLS, (b) SF, and (c) SPS, and their respective grain size distribution. Only nearly spherical grains were considered for the grain size determination. The BNKT16 samples sintered by PLS and SF show grains with two well-differentiated morphologies: nearly spherical grains commonly seen in most ceramics, and the rod-like grains which can be associated with the rhombohedral phase. Average grain sizes of 1.81 and 1.89 µm were obtained for PLS and SF samples, respectively. The SPS sample shows a lower average grain size of 1.76 µm and it does not exhibit rod-like grains. This can be explained by the short time needed to accomplish the sintering process, thus reducing the volatilization of alkaline elements and in consequence inhibiting the formation of the rod-like grains which did not have sufficient time to form. Rod-like grains are generally larger than spherical ones and they are more frequently seen in SF sample. The Na/K ratio is very important for the piezoelectric response in the BNKT systems [30,31]. To obtain the semi-quantitative Na/K ratios and atomic percentage, energy dispersive spectroscopy (EDS Bruker 60) analysis was done in all BNKT16 samples manufactured by different processes. Figure 5 illustrates the BSE-SEM images for the microstructure regions of the BNKT16 (a) PLS, (b) SF, and (c) SPS ceramics. The grain matrix (marked regions as a.2, b.1, and c.1) has a nominal composition ratio Na/K ratio 9.41, 1.21, and 2.8 for PLS, SF, and SPS samples, respectively. This nominal ratio is higher in the matrix when the powders are sintered by PLS, and is lower when the powders are sintered by SF. Table 4 summarizes the EDS results for all samples; the table clearly reveals that the element concentration is highly affected by the used sintering technique. A clear tendency related to Bi volatilization can be clearly observed. In the rods, lower atomic percentage of Bi is detected, while the atomic percentage of K exhibits higher values than those detected in the grain matrix. Hence, considering rod composition (marked as regions a.1 and b.2 in Fig.  5), a question arises about whether a secondary phase is formed or not, because of the high degree of Bi volatilization. From the XRD patterns (Fig. 2), no secondary phase is detected. It is worthy to mention that BNKT rod-like grains have been previously reported only by two other research works [3,30] and they have considered the rods as a secondary phase based on local Raman measurements. At higher temperature, they also visualized the secondary phase by XRD.

3 Scanning electron microscopy and energy dispersive spectrometer
In summary, evidence of phase coexistence can be observed by means of XRD and Raman analysis. Rodlike grains can be considered as a secondary phase in the sense that they have a remarkable difference in the Bi composition. However, evidence of a different crystal structure is not detectable. Henceforth, local Table 4 Composition of the regions analyzed by EDS in Fig. 5, which represents the atomic percentages of elements and Na/K ratios properties at this level can be different in comparison to nearly circular grains. Figure 6 shows the amplitude and phase response as a function of applied DC bias voltage from -40 to +40 V at a drive frequency of 300 kHz applied between the tip and the bottom electrode. This causes the sample to vibrate at the same frequency due to the converse piezoelectric effect [32,33]. The amplitude (pm) versus voltage (V) exhibits the butterfly loops that describe the local displacement which are characteristic of the piezoelectric response in ferroelectric media.  Table 5 shows the PFM  Table 4.   Figure 7 shows the global hysteresis loops, collected at room temperature, of the BNKT16 ceramic samples sintered by PLS, SF, and SPS. The hysteresis loops for the PLS (Fig. 7(a)) and SPS (Fig. 7(c)) are not in complete agreement with the PFM results previously discussed, where all samples exhibit similar and good local ferroelectric−piezoelectric properties. The elliptic shape and the inclination of the global loops in the Figs. 7(a) and 7(c), and the decrease of the polarization for the maximum value of the applied electric field, more pronounced for the SPS sample, can be associated with microstructural defects, spinning of domain walls, and dielectric loss due to electrical conduction along and across to the grain boundaries, among others. With this, the maximum (P MAX ) and remanent (P r ) polarization as well as the coercive electric field (E c ) are strongly affected and not reveal the real microscopic ferroelectric properties. Similarity in shape to the electric hysteresis loops of Figs. 7(a) and 7(c) for BNKT ceramic family has been previously reported [5]. However, the loops of the BNKT16 ceramic sample sintered by SF ( Fig.  7(b)) exhibit a hysteresis behavior that looks closer to a hard ferroelectric material. In fact, the inflection point between remanent polarization and its corresponding coercive electric field can be clearly observed. SF samples are the most compacted ones with the highest density.

5 Electrical analysis
Values of d 33 , P r , E c , average grain sizes, and density of all ceramic samples are listed in Table 6. It can be observed that d 33 , P r , E c increase in the ordered sequence: SPS, PLS, and SF, which is the same order as the rhombohedral phase volume fraction increases (Table 1). It can also be noticed that grain size increases in the same sequence as well as the density (Table 6). SF samples have the largest grain size (1.89 µm) and thus the highest d 33 (165 pC/N) which is one of the top values reported in the literature [3,30]. The tendency in the grain size-piezoelectric propery relation of Table 6 has been reported in previous works [5,30] and is also observed herein. However, the BNKT composition has other important factors to consider and it is the presence of grains with different morphologies (Fig. 3) and different stoichiometric composition ( Table 4).
The tendency of the observed properties in the sintered samples (Table 6) is associated with the morphology of SF specimen (Fig. 3(b)) that can be easily differentiated from PLS ( Fig. 3(a)) and SPS (Fig. 3(c)) ones. For SF, the rod-like grains grow perpendicular to the direction of the uniaxial applied stress, and they are embedded in the gain matrix. The interaction between nearly spherical and rod-like grains, which can also be observed as a secondary phase, is expected to play a key role in the improvement of piezoelectric response as reported by Ramajo et al. [30] for their PLS sample. In the present work, this interaction is also presented for the PLS sample in lower amounts. For PLS sample, the volume fraction of rod-like grains is lower than the one of SF sample, and they are widely spread not showing a preferred orientation.
Crystallographic texture is expected to be another important factor that affects final piezoelectric properties. Table 7 reports the modified March's function G 1 parameter   [34,35]. Hence, the lower the value between 0.0 and 1.0 for G 1 , the formation of a crystallographic texture is more likely to occur. It can be observed from Table 7 that the texture for the PLS and SPS samples are quite similar and the SF experiments a remarkable higher texture. This result is congruent with the fact that a uniaxial stress applied during sintering favors the crystallographic orientation distribution along the stress direction. For the SPS sample, a stress is applied along the axial direction, but two facts need to be mentioned: SPS processing time is very short and the powders are confined inside a mold, then the stress applied along the axial sample direction produces a stress distribution [36]. As a result, not such a preferred orientation is developed like in the SF sample. Then, the stronger the texture, the higher the piezoelectric coefficient. Finally, it is worth to mention that for BNKT-R3c phase for the SF sample, a combination of three preferred orientations is found. It contributes to a better Rietveld refinement. SEM micrograph for SF ( Fig. 4(b)) depictures a congruent statement with this Rietveld finding. For this microstructure, it can be expected that, at least, there exist two different preferred orientations, one for the nearly circular grains and the other one for the rodlike grains. The highest piezoelectric coefficient globally measured is for SF sample. However, the PFM analysis reports that the highest piezoelectric response is for PLS sample. This does not have to be a contradicting result. The contributions for the piezoelectric coefficient can be found at different structural levels from the unit cell to the grain level. For example, SF sample can be described as a ceramic-ceramic composite whose nearly spherical grain matrix has well oriented rod-like Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.