Layer modulating and electrical properties in acceptor-doped Sr₅Nb₅O₁₇ ceramic with perovskite layer structure

Synthesis

Samples of Sr₂Nb_2-2xGa_xMo_xO_7-0.5x (x ≤ 0.2) were prepared by a conventional solid-phase reaction method with SrCO₃ (99.0%), Nb₂O₅ (99.99%), Ga₂O₃ (99.99%), and MoO₃ (99.90%) as starting materials. After being weighed and repeatedly ground in an alcoholic medium for 2 h, followed by air drying, the well-mixed raw materials were pre-calcined at 1,200 °C for 12 h in the air atmosphere with a heating and cooling rate of 5 °C/min. The pre-calcined powder was then milled and pressed into pellets (diameter: 20 mm, thickness: 4-5 mm) and then sintered at 1,450 °C for 12 h. Next, the calcined powder samples were processed with dilute sulfuric acid solution, then filtered, thoroughly washed with water and dried at room temperature under ambient conditions. The resulting products were reground and pressed into pellets (diameter: 10 mm, thickness: 1-2 mm) and sintered at 1,450 °C for 12 h at a heating and cooling rate of 5 °C/min.

Characterization

X-ray diffraction (XRD) was performed on a Panalytical X’pert Pro X-ray diffractometer with a Cu-Kα radiation source. High-quality laboratory XRD data for Rietveld refinement were collected over a 2θ range from 10° to 80° using an interval of 0.02°. The Pawley and Rietveld refinements of the XRD data were carried out by the Topas-Academic software^[29]. The Microstructure and energy disperse spectroscopy (EDS) were performed on a Hitachi (Tokyo, Japan) S4800 scanning electron microscope (SEM). The morphology and internal structure of the prepared samples were further checked by transmission electron microscopy (TEM), selected area electron diffraction (SAED), and high-resolution transmission electron microscopy (HRTEM) using a JEOL JEM-2100F electron microscope operated at a voltage of 200 kV. Elemental mapping of the materials was obtained through an EDAX detector attached to JEOL-2100F. The electrochemical impedance spectroscopy (EIS) was recorded on the Bio-Logic VSP instrument over a 10⁶ to 10^-1 Hz frequency range. Both surfaces of the samples were coated with platinum paste and fired at 850 °C for 30 min to remove any organic components from the electrodes before EIS measurement. For second harmonic generation (SHG) measurements, laser pulses from a Nd:YAG laser system with a wavelength of 1,064 nm were used as the fundamental light and the second harmonic light with a wavelength of 532 nm was detected in a specular reflection geometry. The samples were irradiated with laser pulses of approximately 10 ns duration at a repetition frequency of 40 Hz. The laser was focused to the diffraction limit using an objective lens with a numerical aperture of 0.2 mm. Polarization-electric field (PE) loops were measured using a ferroelectric tester (TF Analyzer 2000, aixACCT, Germany) (150 °C, 10 Hz frequency).

Phases analysis and purification

X-ray diffraction (XRD) experiments were first performed on the as-made samples from solid-state reaction to investigate the phase compositions for Sr₂Nb_2-2xGa_xMo_xO_7-0.5x (x = 0.1, 0.2). As shown in Supplementary Figure 1A, it revealed that there are two main phases, i.e., 4-layered Sr₂Nb₂O₇ (PDF#70-0114) phase and 5-layered Sr₅Nb₅O₁₇ phase (PDF#83-0603) for the sample Sr₂Nb_1.8Ga_0.1Mo_0.1O_6.95 (SNGM10). While for the Sr₂Nb_1.6Ga_0.2Mo_0.2O_6.9 (SNGM20) sample, besides the main phase of 5-layered Sr₅Nb₅O₁₇, a secondary phase of SrMoO₄ (PDF#85-0809) was observed. Moreover, the HAADF-STEM image and energy dispersive X-ray spectroscopy (EDX) elemental maps of SNGM20 clearly show the segregation of the Mo element, as shown in Supplementary Figure 1B. In the Mo segregated area, much lower Nb and Ga concentrations were detected, while both Sr and O elements showed a homogeneous distribution, suggesting that both the main phase and secondary phase contain Sr and O elements, and the Mo element is exclusively distributed in the secondary phase. HRTEM images are shown in Supplementary Figure 1C. It can be seen that the interplanar spacings in the pink squared area of 0.283 and 1.625 nm can be indexed to the (0 -2 0) and (200) crystal planes of Sr₅Nb₅O₁₇ (PDF#83-0603), respectively. The green squared area displays a 0.116 nm d-spacing associated with the (219) plane of SrMoO₄ (PDF#85-0809), further confirming the Mo-rich area of the SNGM20 sample to be the SrMoO₄ phase.

Endeavor to synthesize pure 5-layered Sr₅Nb₅O₁₇ phase by adjusting the sintering temperature or the proportion of Mo failed, as shown in Supplementary Figure 2 for the products of the sample SNGMO20 prepared under some other conditions. Thus, to get pure 5-layered Sr₅Nb₅O₁₇ phase samples, a chemical purification method was then applied. For this method, the SNGM20 sample was simply fully washed with dilute sulfuric acid (in which the SrMoO₄ was soluble while the niobate insoluble) and followed with liquid-solid separation by filtering three times. The subsequent XRD and TEM/EDX measurements confirmed the pure 5-layered Sr₅Nb₅O₁₇ phase nature of the yielded dried powders. As shown in Figure 2A and B, all the diffraction peaks in the XRD pattern of the washed sample can be well indexed by the 5-layered Sr₅Nb₅O₁₇ phase (PDF#83-0603), and no extra peaks from the secondary phase were identified. In addition, obvious segregation of the Mo element was no longer disclosed by the TEM/EDX characterizations on many different areas of the sample, although there was still very weak signal from the Mo element indicating trace concentration in the sample, as shown in Figure 2C for a selected area. Such trace Mo element may originate from the light surface contamination of the insoluble 5-layered Sr₅Nb₅O₁₇ phase powders with the SrMoO₄ solution. Such surface contamination is inevitable and the surface trace SrMoO₄ is, therefore, hard to remove completely. Nevertheless, this slight surface contamination was thought to have little, if any, impact on the bulk properties of the 5-layered Sr₅Nb₅O₁₇ phase sample. Selected area electron diffraction (SAED) further confirmed the 5-layered structure of the purified sample. As demonstrated in Figure 2D, the electron diffraction pattern from the [001] projection direction can be well indexed to the noncentrosymmetric 5-layer structured Sr₅Nb₅O₁₇ phase with space group Pnn2. The measured interplanar spacings in Figure 2E correspond to the (200) and (010) crystallographic planes of Sr₅Nb₅O₁₇(ICSD#79699), respectively. A comparison of HRTEM images of the 5-SNGO and 4-SNO samples [Supplementary Figure 3] clearly revealed the characteristic lattice spacings: ~1.630 nm of the (200) plane for the 5-layered structure and 1.342 nm of the (020) plane for the 4-layered structure, respectively.

Figure 2. (A) XRD patterns of SNGM20 samples before and after washing; (B) Local enlargement of the XRD spectrum in (A); (C) High-angle annular dark field (HAADF) images and energy dispersive spectroscopy (EDS) mapping of SNGM20 powder after washing; SAED (D) and (E) HRTEM images of the SNGM20 powder after washing.

As the Mo element failed in the substitution for Nb and existed in the form of SrMoO₄, a more exact composition of the purified 5-layered Sr₅Nb₅O₁₇ phase sample may be written as Sr_1.8Nb_1.6Ga_0.2O_6.1 (viewed as the consequence of subtracting the Mo-containing part of Sr_0.2Mo_0.2O_0.8 from the designed initial composition of Sr₂Nb_1.6Ga_0.2Mo_0.2O_6.9) and normalized as Sr₅Nb_4.444Ga_0.556O_16.944 (or Sr₂Nb_1.78Ga_0.22O_6.78). This chemical composition was further supported by the Rietveld refinements for the high-resolution XRD data. For the Rietveld refinements, we used the initial orthorhombic 5-layered Sr₅Nb₅O₁₇ structure model reported by Schmalle et al. with the space group Pnn2 and containing three Sr sites (two 4c sites and one 2b site), three Nb sites (two 4c sites and one 2a site), and nine oxygen sites^[8]. The refinement results revealed that the Ga ions were preferred distributed on the Nb1 site (2a) with an occupancy of about 0.549, with the sum of the occupancy of Nb and Ga constrained to be 1.0 on this site. The other Nb sites and Sr sites are all exclusively occupied by Nb and Sr, respectively, with occupancies close to or even larger than 1.0. For the oxygen sites, although their positions were freely refined, the occupancies of these sites were simply set to be one without being refined, due to their weak scattering power to X-ray. The fitting plots are shown in Figure 3 and the final refined structural parameters are summarized in Supplementary Table 1. The refined cation composition of Sr₅Nb_0.452(2)Ga_0.548(2) further confirmed our above speculation for the actual composition of the purified 5-layer Sr₅Nb₅O₁₇ phase sample (denoted as 5-SNGO hereafter). However, when we tried to synthesize the Mo-free and Ga-doped Sr₂Nb_2-yGa_yO_7-y samples, although single-phase products can be obtained for the samples x ≤ 0.15, they crystallized in the 4-layer Sr₂Nb₂O₇ structure, as shown in Supplementary Figure 4.

Figure 3. The Rietveld fitting plots of the SNGM20 powder after washing.

Therefore, it seems that the addition of Mo element is beneficial for the formation of the 5-layer Sr₅Nb₅O₁₇ phase. In fact, as had been well documented, the crystallized structure type of the SrNbO_x system is closely related to the oxygen content x, e.g., SrNbO_3.5 and SrNbO_3.4 for the 4-layered Sr₂Nb₂O₇ and 5-layered Sr₅Nb₅O₁₇, respectively. Herein, for the prepared Ga single-doped materials Sr₂Nb_2-yGa_yO_7-y, the lowest oxygen content for the sample y = 0.15 was SrNbO_3.425, higher than the up-limit (SrNbO_3.42)^[8] for a stable 5-layered structure. Hence, only a 4-layered Sr₂Nb₂O₇ phase was identified for these Ga single-doped Sr₂Nb_2-yGa_yO_7-y materials. However, for the 5-SNGO (Sr₅Nb_4.444Ga_0.556O_16.944) sample, the oxygen content was SrNbO_3.39, even lower than that (SrNbO_3.4) for the parent Sr₅Nb₅O₁₇ and therefore forming the 5-layered structure. This is interesting and the Mo element, therefore, seems to facilitate the formation of a 5-layer phase by improving the Ga-doping level and accordingly decreasing the oxygen content in the SrNbO_x system, although the exact mechanism of how the Mo element improved the Ga-doping level remained unclear. Additionally, it is worth noting that although the existence of Mo promotes the formation of the 5-layer phase, single Mo doping in Sr₂Nb_2-xMo_xO_7+0.5x(x = 0.1, 0.2) led to a 4-layer phase (with a minor impurity of SrNb₈O₁₄ in both samples, as shown in Supplementary Figure 5) but not a 5-layer structure, which may be attributed to the fact that Mo donor doping at Nb sites is expected to introduce excess oxygen rather than oxygen vacancies into the structure, thereby preventing the formation of an oxygen-deficient 5-layer phase. Nevertheless, the reason for this intriguing Mo doping behavior in the SrNbO_x system remains unknown.

Defect formation energy calculations

To get a fundamental understanding of the solid solution behavior of Ga doped at the Nb site, defect formation energies were calculated using static lattice atomistic simulations based on the General Utility Lattice Program (GULP)^[30,31]. The Buckingham potential function was used to model the interaction between ions with the shell model to describe the electronic polarizability for the structural modeling. The potential parameters used for the atomistic simulation are listed in Table 1 and the well-reproducing experimentally observed crystal structures (see Table 2) validate the feasibility of these interatomic potential parameters.

Then, the energetics of substitutions was calculated for the single Ga doping, as well as the Ga and Mo co-doping on the Nb site, with the creation of oxygen vacancies as charge compensating defects. This doping process of Ga for Nb and Ga/Mo co-doping for Nb can be represented by the following defect reactions (1) and (2), respectively:

The energies of this solid solution reaction can be evaluated by combining the appropriate point defect energy and lattice energy terms. The point defect energies and lattice energies are summarized in Table 3. We can see that the lowest oxygen vacancy formation energy was calculated to occur at the O5 position, suggesting the most likely position for oxygen vacancy. The solid solution energies of Ga substitution for Nb, as well as Ga and Mo co-doping for Nb, in the 5-layered Sr₅Nb₅O₁₇ structure, are summarized in Table 4, with the oxygen vacancy supposed to be created at the O5 position due to its lowest point defect formation energy as shown in Table 3. The results show that the solid solution energies of Ga doping at the Nb1, Nb2, and Nb3 sites are -15.8931, -1.4693, and -11.6292 eV, respectively, indicating the Ga atoms are preferred to occupy the Nb1 position, which has the highest symmetry among the three Nb sites, and this is in good agreement with the results from Rietveld refinements for the XRD data. In contrast, for Ga and Mo co-doping for Nb1, Nb2 and Nb3 sites, the final calculated solid solution energies are 17.6358, 19.8798, and 18.0536 eV, respectively, which is apparently higher than that of the single Ga doping, which well explains experimentally observed behaviors that the Ga and Mo co-doping failed to obtain single phase and only Ga succeeded in substitution for Nb atoms in the 5-layered Sr₅Nb₅O₁₇ phase.

Nonlinear optical properties

An optical second harmonic generation (SHG) measurement method was then utilized to investigate the symmetry of the purified sample 5-SNGO. The SHG has a high sensitivity to symmetry breaking and can be used to study the symmetry, ferroelectricity, and ferromagnetism of materials and to determine the phase transition temperature^[32,33]. Non-centrosymmetry (NCS) is the prerequisite for an SHG process because the SHG coefficient is zero (χ² = 0) in all centrosymmetric materials. Therefore, ferroelectric states that do not have inversion symmetry usually provide strong second-harmonic signals, which display second-order nonlinear optical behavior. The SHG plot of the 5-SNGO sample is shown in Figure 4. It can be seen that as the incident light power increases, it needs to pass through a threshold power, which corresponds to the power point where the signal begins to significantly enhance. When the incident optical power squared is greater than 50 mW², the intensity of the SHG is proportional to the incident light power squared and exhibits second-order nonlinear optical behavior, consistent with the noncentrosymmetric structure with the space group Pnn2.

Figure 4. SHG intensity as a function of the incident light power squared in the 5-SNGO sample.

Ferroelectric measurements

Figure 5A and B shows the ferroelectric hysteresis loops of Sr₂Nb₂O₇ (4-SNO) and 5-SNGO measured at 150 °C, 10 Hz frequency, under different electric fields. Figure 5C and D shows the polarization-electric field (P-E) hysteresis loops and charging current density-electric field (I-E) curves for 4-SNO and 5-SNGO ceramics at an electric field of 150 kV/cm. For the 4-SNO sample, saturated P-E hysteresis loops and obvious peaks in the I-E curves are observed, confirming its ferroelectricity. When the coercive field (E_c) is applied, the I-E peak appears, indicating ferroelectric domain switching. The 4-SNO sample has a spontaneous polarization strength (P_s) of 4.09 μC/cm², a remanent polarization (P_r) of 1.17 μC/cm², and a corresponding coercive field (E_c) of 33.52 kV/cm. For the 5-SNGO sample, no I-E peak is detected, indicating that it exhibits paraelectric behavior.

Figure 5. Polarization-electric field hysteresis loops (P-E) of (A) 4-SNO and (B) 5-SNGO measured at 150 °C, 10 Hz frequency, under different electric fields. Polarization-electric field (P-E) hysteresis loops and charging current density-electric field (I-E) curves for (C) 4-SNO and (D) 5-SGNO samples at 150 °C, 10 Hz frequency.

Dielectric studies

Figure 6 displays the frequency dependence of the dielectric constant (ε′) and dielectric loss (tanδ) for samples 4-SNO and 5-SNGO at room temperature. We can see that as the frequency increases from 10² to ~10⁵ Hz, the dielectric constant decreases sharply (from ~280 to ~90) for the 4-SNO sample, indicating dispersive behavior. Then, the ε′ reaches a constant value at higher frequencies. However, for the 5-SNGO sample, only a slight decrease in ε′ (from ~100 to ~60) was observed within a much more narrow frequency range of 10² to ~10³ Hz, and then it kept stable. Usually, for a given sample, the relatively higher values observed at low frequencies may be attributed to the space charge polarization, such as the Maxwell-Wagner effect. The dielectric constant decreases with increasing frequency can be explained by the fact that in a high-frequency electric field, the polarization response could not keep up with the change in the electric field direction, resulting in a decrease in the degree of polarization. For dielectric loss, similar behaviors were also observed that at the low-frequency range, the tanδ of the 5-SNGO sample showed a sharper decrease from 0.60 to 0.006 within the frequency range of 10² to ~10⁵ Hz, while the 4-SNO sample decreased from 0.20 to 0.001 within almost the same frequency range. At higher frequencies, the tanδ kept stable for both samples. The higher dielectric losses at low frequencies are related to the grain boundary defects^[34]. Thus, generally speaking, the ε′ and tanδ of the 4-SNO sample are higher than those of the 5-SNGO sample, especially at the low-frequency range, but the 5-SNGO sample displayed better frequency stabilities compared to the 4-SNO sample. Polarization studies show that 5-SNGO has a slightly lower dielectric constant and dielectric loss than 4-SNGO, possibly related to the concentration of oxygen vacancies created by the substitution of Ga for Nb ions. The formation of oxygen vacancies reduces dielectric loss and improves frequency stability. Therefore, the more flattened dielectric constant and lower dielectric loss of the 5-SNGO sample offer better possibilities for applications compared to the classical 4-SNO sample.

Figure 6. Frequency dependence of the real (ε’) parts of dielectric permittivity and dielectric loss (tan) of 4-SNO and 5-SNGO ceramics at room temperature.

Complex impedance

To investigate the contribution of defective dipole/oxygen vacancies to the electrical properties of 5-SNGO, the complex impedance spectroscopy technique was also applied. Figure 7A shows the complex impedance spectra of the 5-SNGO ceramics, measured over the temperature range of 750 to 900 °C. It can be observed from the figure that only one semicircular arc is observed in the impedance spectrum, which indicates that the ceramic is mainly affected by the grain response and that the resistance of the ceramic decreases with increasing temperature, corresponding to the negative temperature coefficient effect of the ceramic resistor^[35], and that the increase in the AC conductivity suggests the thermally activated process of the defective carriers. The DC resistivities of both the 4-SNO and 5-SNGO samples in the range of 600-900 °C are shown in Figure 7B, from which it can be clearly observed that the 5-SNGO ceramic has a resistivity approximately two orders of magnitude higher than that of the 4-SNO, e.g., 1.95 × 10⁵vs. 1.73 × 10³ Ω·cm at 900 °C, for 5-SNGO and 4-SNO, respectively. Thus, 5-SNGO material has an apparently enhanced insulating property compared to that of the 4-SNO material. A further study of the dielectric relaxation process and activation energy changes in the 5-SNGO ceramic can be achieved through the analysis of the normalized imaginary impedance Z″/Z″_max [Figure 7C]. The 5-SNGO sample exhibits only one peak in the investigated frequency and temperature range and shows excellent symmetry at each temperature. Meanwhile, at high temperatures, the carrier hopping movement is accelerated and the peaks are shifted toward the high-frequency region, which can be attributed to the thermally activated jumps of the carriers (oxygen vacancies). Arrhenius fitting of the peaks was performed using the following equation^[36]:

Figure 7. (A) Complex plane impedance plots of 5-SNGO recorded at temperatures 750-900 °C; (B) direct current (DC) resistivity of 4-SNO and 5-SNGO ceramics; (C) frequency dependence of normalized impedance Z″/Z″_max of 5-SNGO; (D) the relationship of Z″/Z″max and M″/M″max at 800 °C of 5-SNGO.

where ω₀, E_a, k_B, and T denote the pre-exponential factor, the relaxation activation energy of imaginary impedance Z″, the Boltzmann constant (8.617 × 10^-5 eV/K), and the Kelvin temperature, respectively. The relaxation activation energy E_a can be obtained from the data points in the plot of ln(ω) versus 1,000/T by linear fitting, as shown in the inset of Figure 7C. The E_a of SNGMO ceramics is 1.79 eV at 750-900 °C. Ang et al. concluded that the high-temperature relaxation in many perovskite-like structures is caused by the long-range migration of oxygen vacancies^[37]. In contrast, for high-temperature insulating materials, the higher relaxation activation energy contributes to the insulation stability at elevated temperatures, prevents excessive charge-carrier mobility, and improves the potential for applying this ceramic at extreme temperatures. Figure 7D shows the Z″/Z″_maxvs. f and M″/M″_maxvs. f curves for 5-SNGO ceramic at 800 °C. It is noticed that the Z″/Z″_max and M″/M″_max peaks do not completely overlap but are very close to each other, suggesting contributions from both long-range and localized relaxation effects^[38]. Comparison of the impedance and modulus data allows us to rationalize the subject response on the basis of local (dielectric relaxation) and non-local (ionic conductivity) relaxation processes.