Sol–gel synthesis and structural characterization of band gap engineered ferroelectric perovskite oxide potassium sodium barium nickel niobate

All chemicals used in this paper were purchased from Sigma Aldrich. Potassium nitrate (KNO₃, 99.5 %), sodium nitrate (NaNO₃, 99.5%), barium nitrate (Ba(NO₃)₂, >99%), nickel (II) nitrate hexahydrate (Ni(NO₃)₂·6H₂O, >99 %) and ammonium niobate (V) oxalate hydrate (C₄H₄NNbO₉·nH₂O, 99.99%) were used as reactants. EDTA (ethylenediaminetetraacetic acid, >99.4%) and CA (citric acid, >99.5%) were used as chelating agents. Figure 1 illustrates the synthetic procedure. First, the reactants with the correct stoichiometry were dissolved in de-ionized water or ethanol (with concentration of the solvent 1.2 M) at 100 °C and constantly stirred until a transparent, colorless solution was obtained. Subsequently, EDTA and CA were added into the solution with molecular ratio of EDTA:CA:Me = 1:2:1, where Me is the total metal ions. The pH value of the solution was adjusted to 9 with ammonia solution (NH₃·H₂O, 27–30 wt%, CO₂ traces < 0.002%). These steps were carried out in capped and sealed bottles.

It should be noted that after the above mentioned procedures, the aqueous solution was stable without precipitation after the stirring was stopped. The ethanol-based solution precipitated quickly in a few hours even if being constantly stirred. This solution could be stabilized through heat treatment at 90 °C in a sealed bottle placed in an oven for several days. Once the solution was stabilized, its color changed to khaki.

The stabilized solutions were transferred into an evaporation bowl, dried at 100 °C in an oven until dark goldenrod gels appeared, and finally fired at 450 °C with a heating rate of 5 °C/min until the gels turned to black flakes. The final products were obtained after grinding the flakes in mortar and calcining them at 750–850 °C. Thermal changes in the transitions from the dried gels to calcined powders were studied by carrying out DSC (differential scanning calorimetry) and TG (thermogravimetric) analyses (STA449, NETZSCH, Germany) in the temperature range of 25–900 °C with a heating rate of 3 °C/min. The phases of the synthesized powders were identified with XRD (X-ray diffractometer, D8 Discover, Bruker, Germany, step size 0.05°, scan speed 1 s/step; and PW 1800, Philips, Netherlands, step size 0.02°, scan speed 2 s/step) (with Cu Kα radiation at 40 kV and 40 mA). FE-SEM (field-emission scanning electron microscope, Sigma, Zeiss, Germany) was used to examine the morphology of the powders. Another FE-SEM (ULTRA plus, Zeiss, Germany) equipped with EDXS (energy dispersive X-ray spectrometer) was used for elemental analyses. The particle size distribution was measured both in SEM micrographs and by DLS (dynamic light scattering, Compact-Z, Photocor, Russian Federation). The zeta potential was measured with Zetasizer Nano (Malvern Instruments Ltd, UK). This Zetasizer was also used to double check thus confirm the particle sizes. Transmittance of the calcined powders was measured with a UV–Vis-NIR spectrophotometer (Cary 500 Scan, Varian, USA).

Figure 2 shows the mean hydrodynamic diameter of the colloidal particles in the aqueous and ethanol-based sols (as made and after being stabilized by heat treatment as mentioned above) measured from DLS. Table 1 compares their mean particle size values and zeta potentials. The particles in the aqueous and stabilized ethanol-based sols were monodisperse with narrow size distributions between 0 and 11 nm (aqueous)/10 nm (stabilized ethanol based), and mean sizes of 5.6 nm (aqueous)/4.3 nm (stabilized ethanol based) in diameter. This makes both the sols suitable precursors for potential fabrication of thin films using methods like spin coating [20]. Thickness of the films is tunable in a broad range with fine interval thanks to the minimum achievable single-layer thickness, which is largely determined by the particle size. The particles in the as made ethanol-based sol demonstrated a broad size distribution (30-170 nm) with a mean value of 98 nm in diameter. This implies that the reactants and nanoparticles were more likely to form clusters in the ethanol environment, and, this explains that the as made ethanol-based sol tended to precipitate rapidly as presented above. Although the as made and stabilized ethanol-based sols showed similar average zeta potentials, that of the as made ethanol-based sol varied significantly as indicated by its deviation (Table 1). If taking the largest deviation into account, the zeta potential of the as made ethanol-based sol could reach a value as weak as −18 mV (−33 + 15 mV) (Table 1). This may be an indicative factor responsible for the instability of the as made ethanol-based sol. A part of the particles associated with relatively weak zeta potentials (e.g., −18 mV) may aggregate first due to lack of electrostatic repulsion between particles. The aggregated particles increase in size, thus the rest of the smaller particles need a stronger repulsion to confer stability, lifting the lower limit of zeta potential for resisting aggregation. This may further cause a domino effect, leading to a fast precipitation from the macroscopic view. In comparison, the deviations of zeta potentials for the aqueous and stabilized ethanol-based sols were much smaller, which may help to eliminate the triggering effect as may have happened in the as made ethanol-based sol. Table 1 also provides a hint that the electrostatic repulsion needed to stabilize the aqueous sol can be about only half of that needed to stabilize the ethanol-based sol. This may be attributed to the difference between the features of water and ethanol molecules where water is more chemically stable while ethanol may produce hydroxide (OH⁻) that can combine with the particles containing alkaline elements or can break the pH balance.

Figure 3 shows the XRD patterns of the KNN–xBNNO powders (after calcination at 750 °C) with x = 0, 0.02, 0.04, 0.06, 0.08, and 0.1, all made from the aqueous sols. It can be seen that all the powders formed orthorhombic perovskite phases, consistent with its parental composition (x = 0) with a space group of Amm2 [9, 21]. Additional reflections at about 37 and 43° indicate the presence of excess NiO that did not diffuse into the solid solutions [7, 9, 22]. It should be noted that the intensity of the NiO reflections in the x = 0.02 powder shown in Fig. 3 is hardly visible, indicating a relatively complete diffusion. The perovskite reflections for all the samples broadened with increase of the x value, implying a decreased crystallinity. In the meantime, the intensity of NiO reflections also increased with increased value of the compositional parameter x. According to the Scherrer equation [23], the broadened reflections corresponded to decreased crystallite size. From this perspective, the crystallite size of the KNN–xBNNO powders decreased with increased value of x. In addition, reflections of 111, 022/200, and 122/211 at about 32°, 46°, and 52°, respectively, shifted toward lower angles with increased value of x (Fig. 3). In KNN–xBNNO, two different transition metal cations were located at the B-site. The Nb⁵⁺ drives ferroelectricity together with A-site cations whilst Ni²⁺ reduces the band gap [7, 24]. An off-center lattice distortion and polarization in the classic ferroelectric perovskite KNbO₃ caused by Na⁺-doping [10] provides a favorable condition to introduce a combination of Ni²⁺ and O vacancy, which is responsible for the rise of electronic states in the gap of the KNbO₃ [7, 9, 22]. In this principle, the lattice parameters of KNN–xBNNO increased with increased x value in presence of Ni²⁺ ions, which has a larger ionic radius of 69 pm compared to Nb⁵⁺ (64 pm). This is the reason that the above mentioned reflections shifted to smaller angles (implying increasing lattice parameters) and at the same time broadened (implying smaller crystallite sizes).

Figure 4 shows the evolution of lattice parameters of the KNN–xBNNO calcined powders corresponding to the XRD patterns in Fig. 3. It can be seen that the incorporation of Ni induced the most obvious change on lattice parameter b. All lattice parameters as well as the trend of their evolutions with the Ni concentration are in good agreement with results obtained on sintered ceramics, which were made via solid-state route [9]. This indicates successful fabrications using the sol–gel method in this work. The intensity of the 102 reflection decreased with the increased value of x, which is a consistent observation to that of ceramic powders made via solid-state route [22]. The decreasing intensity of the 102 reflection may be caused by the change of particle orientation or due to the structure factor vanished thus prohibited diffraction of the 102 reflections.

Similar with ceramic powders made from solid-state route, calcination temperature also plays an important role in the sol–gel process chosen for this work. Figure 5 shows the TG and DSC curves measured with the dried goldenrod gels from both the aqueous and ethanol-based sols (x = 0.02). For both gels, the first significant weight loss (~40%) was observed in the temperature range of 150–200 °C. This should be assigned to the decomposition and volatilization of organic compounds such as EDTA and CA (NH₃·H₂O should have been removed during the drying process), which is supported by the endothermic behaviors on the DSC curves. The loss of organic compounds was followed by chemical reactions among the reactants and crystallizations of the particles. The weight losses observed at >200 °C should be resulted from the volatilization of N-containing compounds produced during the reactions as well as possible water of crystallization. For the aqueous sol, these procedures seemed to complete at 520 °C where both the weight loss on the TG curve and the major exothermic behavior on the DSC curve ended. In the temperature range of 800–850 °C, the DSC curve of the aqueous sol showed another smaller exothermic peak but no further weight loss was observed on the TG curve. This implies that an abnormal grain growth might happen at high temperatures. In comparison, for the ethanol-based sol the reactions did not finish until 800 °C with the evidences of consecutive weight losses on the TG curve and several exothermic peaks on the DSC curve in the temperature range of 200–800 °C. For this reason, the calcination temperature of the flakes made from the ethanol-based sols was chosen to be 850 °C to ensure complete chemical reaction and phase formation. In principle, 550 °C should be enough for the flakes made from the aqueous sols. However, in order to compare with the powders made from the ethanol-based sols, similar calcination temperatures were selected, i.e., 750–800 °C for the aqueous sols. The reason here was that a much lower calcination temperature (e.g., 550 °C) may suppress grain growth, generating misleading information about the grain and particle sizes, while a higher temperature (e.g., 850 °C) may lead to abnormal grain growth beyond the nanoscale. In addition, 450 °C was chosen to be the pyrolysis temperature for both the aqueous and ethanol-based sols because this was after the organic compound removal and before the major inorganic reaction (the highest exothermic peaks on the DSC curves) happened.

Figure 6 compares the XRD patterns and SEM micrographs of the x = 0.02 sample (made from the ethanol-based sol) measured after heat treatment at 450 °C (from gel to flake) and 850 °C (from flake to powder). As explained above with Fig. 5, the gel-flake evolution mainly involved the decompositions and volatilizations of organic compounds. As shown in Fig. 6, the black flakes already started to show some preliminary characteristics of the target perovskite structure, with most of the major reflections corresponding to the 850 °C calcined powder. However, it is obvious that the reaction was not complete at 450 °C. A significant amount of residual carbon in flakes is illustrated in Fig. 7a. The carbon source here was likely to be the residual carbon from the ethanol which formed carbon crosslinks between molecules. The residual carbon as well as the accumulation spots of certain elements like K explains the incomplete formation of the perovskite phase as indicated by XRD patterns in Fig. 6, i.e., additional reflections to the marked ones for the perovskite phase. This is supported by the evidence that on the DSC curve of the ethanol-based sol (Fig. 5) there were several exothermic peaks associated with consecutive weight losses on the corresponding TG curve at >450 °C until 800 °C, i.e., the transformation from flake to powder, indicating reactions of the residual K with the perovskite compound and phase formation. The incomplete phase formation at 450 °C may have been caused by residual carbon crosslinks prohibiting crystallization and/or may be because K needed a higher thermal energy to activate its reaction/diffusion in the structure. The subsequent calcination at a temperature of 850 °C successfully burned the carbon off (Fig. 7b), inducing a homogenous distribution of the elements. Large agglomerates of several microns were found with flakes, while for calcined powders the particle size decreased by an order of magnitude to ~100 nm (Figs. 6 and 10b). The large agglomerates were formed because of high surface energy yielded by nanoparticles in the procedure of sol–gel flake. Meanwhile, the large-sized particles on the black flakes were also likely due to the crosslinks formed around carbon atoms. Adding proper dispersant is a usual way to minimize the formation of agglomerates of nanoparticles, but this was not considered crucial in this work as the particle size was greatly reduced in the final product (the calcined powder). During the calcination, carbon was burnt off and thus the possible carbon crosslinks were destroyed, resulting in large particles and agglomerates breaking in to smaller particles. Compared to the 850 °C calcined powder, the XRD reflections of the 450 °C fired flake shifted toward lower angles (Fig. 6). This indicates that the lattice parameters of the flake tend to be larger than those of the powder, possibly due to inter-lattice strains induced by carbon crosslinks stretching the unit cells or the unit cell distortion/expansion caused by defects. For instance, the accumulated K spots (Fig. 7a) which did not diffuse into the A site of the ABO₃ perovskite unit cell imply a large amount of K⁺ vacancies in the flake, leading to variation of the lattice parameters.

Compared to the pure KNN (x = 0), the BNNO doped compositions have a broader window of calcination temperature. It is well known that the calcination and sintering temperature of pure KNN ceramics must be carefully chosen as a lower temperature may lead to incomplete phase formation while a higher temperature (e.g., >800 °C) may cause volatilization of the alkaline elements [11, 21]. Figure 8 shows the XRD patterns of KNN–xBNNO powders for x = 0 and x = 0.1 (made from the aqueous sols) calcined with different profiles. A better crystallinity was found for pure KNN when calcined at 650 °C compared to 600 °C (Fig. 8a, b) where the reflection splitting can be more clearly recognized at 650 °C. This is consistent with previous experience and the same trend was found on the KNN–xBNNO (x > 0) samples. In comparison, higher temperature was needed to calcine the KNN–xBNNO (x > 0) samples. In other words, the doping of BNNO into KNN increased the difficulty of phase formation. This should have brought a disadvantage to the KNN–xBNNO (x > 0) compositions as higher temperature tends to cause loss of alkaline elements, leading to the stoichiometry shifting away from the designed values. However, alkaline cations seemed to be immobilized inside the lattice of KNN–xBNNO thus were more stable than that in pure KNN. This statement was supported by the fact that strong volatilization of potassium hydroxide was detected in the pure KNN precursor calcined at 850 °C while it was not observed in KNN–xBNNO (x = 0.1) calcined at the same temperature (850 °C). This helped to successfully synthesize KNN–xBNNO at necessarily higher temperatures to form proper perovskite phases. The evidence can also be seen in Fig. 8c, d where higher calcination temperature did not lead to significant change of reflection positions and shapes representing variations of lattice parameters. This is consistent with the observations in corresponding ceramics made via the solid-state route [22].

Figure 9 shows the SEM micrographs of the calcined KNN–xBNNO powders with different values of x (made from the aqueous sols). Cubic grains were grown in all samples, corresponding to the orthorhombic shape of the perovskite phase. The grains became more homogeneously distributed and more agglomerates were formed with increased value of x. To determine the grain sizes, the lengths of 100 grains were measured with the software ImageJ for powders with x = 0, 0.02, and 0.04. Grain sizes of the other powders were too small (<70 nm) to be accurately measured. Figure 10 shows the grain length distribution with log-normal distribution fitted curves. The average grain size was 679 ± 40 nm, 123 ± 2 nm and 106 ± 1 nm for the compositions with x = 0, 0.02, and 0.04, respectively. The decreasing trend of particle (grain) size with the increased value of x is consistent with the results expected from the XRD patterns (see discussions related to Fig. 3), which was likely to be due to Ni²⁺–O vacancy defect dipoles acting as centers to suppress grain growth. With the increase of x value, an increasing concentration of the defect dipoles may play a more significant role in grain growth suppression and hence the insufficiently grown, small sized grains looked more homogeneously distributed. The smaller grains were associated with higher surface (grain boundary) energy, leading to more severe agglomeration.

Figure 11 shows the absorption behaviors in the visible light range for KNN–xBNNO powders made from the aqueous sols and calcined at 750 °C with different x values. Compared to the pure KNN (x = 0), all other powders containing Ni showed an additional absorption edge at the wavelength of about 650–700 nm, corresponding to a photon energy of 1.8–1.9 eV. This qualitatively confirms the effect of Ni doping on band gap reduction, indicating successful fabrications of the KNN–xBNNO powders via the sol–gel method.