3.1. X-Ray Phase and Structural Analysis
The crystal structure of the powdered ceramics with a Bi
6Fe
2Ti
3O
18 composition were studied by the X-ray diffraction method at room temperature. A line profile analysis (LPA) was used to determine the microstructural parameters [
31]. For this purpose, the X-ray diffraction pattern, shown in
Figure 1b, was analyzed with the assumption of the pseudo-Voigt profile function. The following parameters of fitting were achieved: R
expected = 4.44%; R
profile= 3.25%, R
weighted profile = 4.20%, and GOF = 0.947. The quality of the fitting procedure, as evidenced by R-parameters, was acceptably good, especially considering the complex system studied.
To retrieve microstructural parameters such as the average crystallite size and the microstrain, the simplified Williamson–Hall analysis was used [
28]. The Williamson–Hall method demonstrates that the approximate formulas for the size broadening and strain broadening of the diffraction peak profile vary differently with respect to the Bragg angle, θ. The formula for size broadening varies as (cosθ)
−1, whereas the formula for strain broadening varies as ~tanθ [
32]. By plotting the product of the total broadening (of the diffraction line) and the cosine of the Bragg angle θ as a function of sinθ, the average strain component is obtained from the slope of the straight line (calculated by linear regression) and the average crystallite size component is obtained from the intersection of the straight line with the ordinate axis [
31]. The relevant Williamson–Hall plot is given in
Figure 1c. According to the performed LPA, it was found that the average crystallite size was 150 Å and the average crystallite (rms) strain was 0.1%.
The qualitative phase analysis of the X-ray diffraction pattern of Bi
6Fe
2Ti
3O
18 ceramics was performed with the Match! (Crystal Impact) software [
24]. The analysis showed that, apart from the expected Bi
6Fe
2Ti
3O
18 phase, there are diffraction peaks that could be assigned to bismuth-based layer-structured Aurivillius phases with the number of layers in the slab other than
m = 5. The results are shown in
Table 1.
One can see from the
Table 1 that the Aurivillius phases retrieved from the X-ray data bases exhibited an orthorhombic structure with the following space group numbers: 35 and 42. At the same time, it should be noted that the orthorhombic distortion of the tetragonal unit cell (i.e., the ratio of the base edge lengths of the cuboidal unit cell) characteristic for the Aurivillius phases included in
Table 1 was close to 1. The figure of merit (FoM) is also shown in
Table 1.
It is worth noting that the parent structure for a compound exhibiting
m = 5 perovskite-like layers described by Aurivillius [
33] was found to adopt the tetragonal structure (
I4/mmm, Space Group No. 139). Due to the fact that Aurivillius phases exhibit small amounts of orthorhombic distortion, this structure was adopted to search for the unit cell of the Bi
6Fe
2Ti
3O
18 compound and its further refinement. The results are shown in
Figure 1d. The following lattice parameters were obtained:
a0 =
b0 = 3.8625Å,
c0 = 49.585Å. The refined average crystallite size was 156Å and the average crystallite (rms) strain was 0.41%. The parameters of the Rietveld refinement were as follows: R
expected = 4.47%; R
profile= 9.45%, R
weighted profile = 12.02%, and GOF = 8.49.
3.2. EBSD Studies of Ceramics
Electron backscatter diffraction (EBSD), the primary tool for lattice orientation determinations, was utilized in the present study to understand the complex microstructures in the Aurivillius-type layer-structured bismuth titanium oxides. The data obtained by EBSD mapping were analyzed using the EDAX OIM Analysis™ software, and the results are shown in
Figure 2,
Figure 3,
Figure 4 and
Figure 5.
All the EBSD data analyses were performed using the EDAX OIM Analysis™.
Figure 2a presents an inverse pole figure map obtained using electron backscatter diffraction (EBSD), which reveals the crystallographic orientation in the Aurivillius-type layer-structured bismuth titanium oxides. The orientation of the grains is represented by a color-coded legend (as shown in
Figure 2c), which corresponds to specific crystallographic directions according to the standard inverse pole figure key. The black regions visible in the map correspond to non-indexed areas, indicating voids within the microstructure (e.g., pores) or regions with insufficient diffraction data due to the surface quality or geometrical factors.
Figure 2b shows the orientation distribution with respect to the sample normal direction [001]. The presented distribution did not exhibit significant clustering, suggesting the absence of a strong preferred crystallographic orientation along this direction.
Further insight into the material texture is provided by the pole figures shown in
Figure 3, which correspond to the {001} and {100} crystallographic planes. The color intensity represents the multiple of uniform distribution (m.u.d.), a quantitative measure indicating how much the observed orientation distribution deviates from a random texture. While there was some variation in intensity, the multiple of uniform distribution value was ~2.67, indicating the presence of a weak texture.
Image quality (IQ) maps are a grayscale representation, where the pixel intensities correspond to the quality of the Kikuchi diffraction patterns. Bright regions indicate high-quality patterns, usually obtained from well-crystallized and correctly prepared (well-polished and strain-free) surfaces, while darker regions suggest a poor diffraction quality caused by strain occurrence, crystallographic disorder, or porosity.
Figure 4a presents an IQ map for the Bi
6Fe
2Ti
3O
18 ceramics. The grain boundaries are clearly visible and the black regions, in this case, correspond to pores. The microstructure shows relatively fine grains of a polycrystalline ceramic material, some with an elongated shape. This correlates with the grain size distribution plot (
Figure 4b), where the grain diameter is shown in relation to the number of grains. The peaks around 0.3 to 0.4 µm indicate that most of the grains fell within this range. For sizes above 0.6 µm, the number of grains significantly decreased.
In order to complement the microstructural observations, an SEM (scanning electron microscope) image of a bigger area was introduced, as shown in
Figure 5a. The observations were made using the secondary electron (SE) mode in high-vacuum mode. The presented image reveals a relatively high porosity of the Bi
6Fe
2Ti
3O
18 material. Energy-dispersive X-ray spectroscopy (EDS) measurements collected from the whole presented surface (
Figure 5b) confirmed the chemical composition stability. The presence of the silicon in the spectrum can be explained by residues of silica in the pores after the polishing process. The carbon presence is explained by the sample’s surface coating.
3.3. Impedance Spectroscopy Measurements of Ceramics
Impedance spectroscopy (IS) is a specific technique that can be considered a subset of broadband dielectric spectroscopy (BBDS). While broadband dielectric spectroscopy covers a wide frequency range from 10
−6 to 10
12 Hz and includes various methods for studying the dielectric properties of materials [
34], impedance spectroscopy specifically focuses on measuring the impedance of a material as a function of the frequency [
20,
29].
A reliable analysis of impedance spectroscopy data requires verification of the data quality and consistency. One widely accepted method for such validation is based on the Kramers–Kronig (K–K) relations, which establish a mathematical link between the real and imaginary parts of a frequency-dependent complex function. In the context of impedance, this means that the real part of the impedance spectrum, ZRe(ω), can be calculated from the imaginary part ZIm(ω), and vice versa. This property enables internal consistency checks of the measured data.
The K-K relations can be expressed in the integral form as follows [
30,
35]:
where R
∞ = Z
Re(ω→∞);
ZRe,i + j
ZIm,i represent the measured impedance at frequency ω
i. These relations are used to confirm that the measured impedance spectrum obeys the principles of causality and linearity.
The calculations necessary for performing the K-K test were accomplished with a computer program according to the methodology described in scientific publications [
30,
35]. In the present study, the impedance data were found to be consistent with the K–K framework, validating the reliability of the measurements prior to further dielectric analysis.
In an ideal case, the results of impedance spectroscopy measurements over a wide range of frequencies can be presented by semicircles in a complex Z″-Z′ plane (Nyquist plot). Each semicircle represents the contribution of a particular process (electrodes and contacts, grain boundaries, and grain interior) to the total impedance of the sample.
Nyquist plots for the Bi
6Fe
2Ti
3O
18 ceramics within the temperature range from −30 °C to +200 °C are shown in
Figure 6.
One can see from
Figure 6 that the measured values of the two components of complex impedance—resistance (real part, Z’) and reactance (imaginary part, Z″)—presented in the form of Nyquist plots rarely take the shape of perfect semicircles (while the isotropic scale is preserved on both axes). They are often described as depressed or deformed semicircles, with their center lying below the x-axis (
Figure 6a,b). At low temperatures, the deviation from the semicircular shape is substantial (
Figure 6c).
To highlight the characteristic depression of impedance semicircles, the experimental Nyquist plots were analyzed using circular arc fitting, analogous to the Cole–Cole approach for non-ideal dielectric relaxation.
Figure 6d–f present fitted impedance spectra at selected temperatures. The key fitting parameters are summarized in each figure.
As the temperature decreased, the intercept of the arc on the real axis increased, indicating an increase in bulk resistance from ZRe = 1.44 × 107 Ω at 135 °C to ZRe = 8.46 × 107 Ω at 95 °C, and further to ZRe = 6.41 × 108 Ω at 55 °C. Concurrently, the depression angle β (indicative of the arc’s deviation from a perfect semicircle) and the characteristic frequency (ωmax) decreased:
At 135 °C: β = 13.85°, ωmax = 481 rad/s;
At 95 °C: β = 12.98°, ωmax = 83 rad/s;
At 55 °C: β = 12.49°, ωmax = 11 rad/s.
These depressed arcs are indicative of non-Debye relaxation behavior, as also suggested by the stretching exponent β extracted from the KWW fits. The centers of the fitted arcs lie below the real axis, with coordinates (in Ω) as follows:
(7.18 × 106, 1.78 × 106) at 135 °C;
(4.22 × 107, 9.76 × 106) at 95 °C;
(3.20 × 108, 7.11 × 107) at 55 °C.
This analysis reinforces the evidence for distributed relaxation mechanisms and heterogeneity in the BFTO ceramics.
This phenomenon, called non-Debye relaxation, is attributed to the distribution of Debye relaxations with different time constants.
The Debye response is given by the following:
where
Zs and
Z∞ stand for the static and high-frequency-limiting values of the impedance, respectively. Equation (3) can be generalized to obtain the Cole–Cole equation [
36]:
which is identical to the Debye case when α = 1, and is broader the lower the exponent α is. The Cole–Cole equation mathematically expresses the distribution of relaxations with different time constants.
Spectroscopic plots of the imaginary part of complex impedance (Z″) are depicted in
Figure 7. This representation is characterized by a peak, signifying the presence of dielectric relaxation in the sample. The observed peak provides insights into the dynamics and behavior of the relaxation mechanism, allowing for a more detailed analysis of the dielectric response within the material.
Figure 7 presents the imaginary part of impedance (−Z″) as a function of the angular frequency (ω) across several temperatures. The curves exhibit peaks that are noticeably broader and asymmetric compared to those predicted by the ideal Debye model, which assumes a single relaxation time. As the temperature increases, the peak frequency shifts toward higher values, reflecting thermally activated relaxation processes.
To quantify this deviation from ideality, the experimental data were fitted using a single-time-constant Debye model. The fitting results are shown for selected temperatures in
Figure 7d–f. At each temperature, the experimental peaks are clearly broader and exhibit flatter tops than the corresponding Debye curves, confirming the presence of a distribution of relaxation times.
This non-Debye behavior is characteristic of complex dielectric systems, such as BFTO ceramics, where structural or compositional heterogeneities lead to a broad range of relaxation dynamics. These findings are consistent with the stretched-exponential behavior (via KWW fits,
Figure 8) and with the depressed semicircular arcs in the complex impedance plane (
Figure 6), both indicative of Cole–Cole-type relaxation.
A number of empirical relaxation functions have been used to describe the imaginary part of the general response function
Z″(ω). A commonality of the functions mentioned above is that they are characterized by power laws far away from intermediate frequencies around the peak frequency ω
p. When viewed versus a logarithmic frequency scale such as abscissa (like in
Figure 7), the slope at the high- and low-frequency limits of the Debye plot is 1 and −1 at a low ω and a high ω, respectively. For the Cole–Cole equation, the slopes are α and -α at a low and high angular frequency, respectively.
To model the behavior of the imaginary component of the impedance, the alternative equation for the susceptibility functions was used [
18]. A three-parameter formula for relaxation in the frequency domain given by the equation
was applied to the analysis of the normalized amplitude (scaled) imaginary part of the impedance
Z″/Z″
max(
ω). In Equation (5),
Z″ represents the current value of the imaginary part of the complex impedance, Z″
max and
ωmax define the height and position of the peak, and “
b” is an internally independent shape parameter for high frequencies.
The KWW equation describing the relaxation function
ϕ(
t) [
16,
17] is expressed as follows:
where
β is the stretching parameter,
τ is the relaxation time, and
f is a measure of the fraction of the experimental quantity being investigated (the amplitude parameter) that is relaxed via α-relaxation.
The relationships providing a bridge between the parameters used in the analysis of the frequency domain representation (Equation (5)) and the parameters characterizing the time-domain relaxation function (Equation (6)) are given by the following:
where
Γ is the gamma function [
37]. These relationships facilitate the translation of the findings between the time and frequency domains, aiding in understanding the relaxation behavior within the material [
38].
The outcomes of modeling the normalized (in amplitude) imaginary part of impedance (Z″/Z″
max) with a normalized frequency (ω/ω
max) for Bi
6Fe
2Ti
3O
18 ceramics at different temperatures, carried out according to the function exhibiting the skewed shape given by Equation (5) (modified KWW formula), are depicted in
Figure 8.
A visual examination of
Figure 8 reveals that the experimental data align well with the model. The quality factor
R2 was 0.991–0.999.
The identification of non-Debye-type relaxation phenomena was possible, as evidenced by the analysis of the stretching parameter β of the KWW function, within temperatures from 20 °C to 200 °C. Despite the BBDS measurements being performed for temperatures lower than 20 °C, it was found that the relaxation peak on the spectroscopic dependence of the imaginary part of impedance shifted to the lower frequency and “disappeared” from the measuring frequency window. It was found that the β parameter of the KWW function ranged from ~0.72 to 0.82 within the temperature range from 200 °C to 20 °C (in the impedance formalism). A lower β value indicates a more stretched relaxation function.
The temperature dependence of the relative permittivity (ε
r) and the dielectric loss tangent (tan δ) for Bi
6Fe
2Ti
3O
18 ceramics is presented in
Figure 9a and
Figure 9b, respectively. The experimental data were obtained over a measurement field frequency range of 100 kHz to 900 kHz. As shown in
Figure 9a, the relative permittivity exhibited a distinct maximum within the temperature range of 600 K to 700 K. This peak shifted toward higher temperatures and decreased in magnitude with an increasing frequency, indicating a frequency-dispersive dielectric response. In contrast, the dielectric loss tangent (
Figure 9b) initially increased slowly with temperature, reaching a value of approximately 1 between 450 K and 550 K, followed by a rapid increase at higher temperatures, exceeding a value of 5 for T > 700 K.
The experimental data were obtained over a measurement field frequency range of 100 kHz to 900 kHz. As shown in
Figure 9a, the relative permittivity exhibited a distinct maximum within the temperature range of 300 °C to 400 °C. This peak shifted toward higher temperatures and decreased in magnitude with an increasing frequency, indicating a frequency-dispersive dielectric response. In contrast, the dielectric loss tangent (
Figure 9b) initially increased slowly with temperature, reaching a value of approximately 1 between 100 °C and 300 °C, followed by a rapid increase at higher temperatures, exceeding a value of 5 for T > 450 °C.
This behavior is characteristic of relaxor ferroelectric materials, where diffuse phase transitions and the strong frequency dispersion of the dielectric peak are commonly observed. The shift of the ε
r maximum to higher temperatures with an increasing frequency suggests the presence of thermally activated dielectric relaxation processes, likely driven by local structural distortions and compositional inhomogeneities inherent to the layered Aurivillius structure of Bi
6Fe
2Ti
3O
18 [
10,
11].
These findings correlate well with the microstructural features revealed by the EBSD and BBDS analyses. In particular, EBSD mapping (
Figure 2,
Figure 3,
Figure 4 and
Figure 5) demonstrated significant orientation-dependent contrast and the presence of ferroelastic domains, which are known to contribute to local polarization fluctuations and dielectric dispersion. The BBDS results further confirmed the existence of domain wall activity and polar anisotropy at the sub-micron scale [
20,
29]. Such domain-related phenomena can give rise to a broadened dielectric response and relaxor-like behavior, as observed in the temperature-dependent permittivity measurements.
Additionally, the pronounced increase in dielectric loss (tan δ) above 700 K (
Figure 7b) may be attributed to an increased electrical conductivity at elevated temperatures [
13]. This is likely related to thermally activated charge carriers, such as oxygen vacancies or hopping between Fe
3+/Fe
2+ centers, as suggested by the defect chemistry typical of Bi-based Aurivillius oxides.