Pb, Bi, and rare earth free X6R barium titanate–sodium niobate ceramics for high voltage capacitor applications

As the market for electric and hybrid vehicles expands, the demand for multilayer ceramic capacitors (MLCCs) that operate as energy storage devices is increasing. Ceramics are generally more stable than polymers at high temperature (>100 °C), making MLCCs especially suited for near engine applications as they do not require heat management systems.¹ Recently, research has focused on improving the energy storage performance of electroceramics for MLCCs to replace tantalum electrolytic and metalized film capacitors with a rated voltage at >400 V.² The relative permittivity (ε_r) of BaTiO₃ dielectric layers decreases sharply at high DC bias, and therefore, it is difficult using current commercial compositions to develop MLCCs with both a high capacitance and volumetric efficiency at high operating fields.^3,4

Based on Eq. (1), approaches to achieve a high W_rec, in, e.g., perovskites, include the following: (i) enhancing the dielectric breakdown strength (BDS) by optimizing the microstructure (eliminating pores, and controlling grain growth),¹¹ establishing a homogeneous electrical microstructure^12–14 through control of defect chemistry, and choosing materials with a large intrinsic bandgap such as NaNbO₃;¹⁵ (ii) increasing polarisability per unit cell volume by incorporating d₀ (Ti⁴⁺ and Nb⁵⁺) and electron lone-pair ions (Bi³⁺ and Pb²⁺) on the B- and A-sites, respectively;^16,17 and (iii) enhancing P_max–P_r by inducing relaxor behavior through alloying or doping to make pseudo-ternary and quaternary solid solutions.¹⁸ Based on Eq. (2), a higher η is achieved by either inducing relaxor behavior (narrower P–E hysteresis loop) or reducing the field difference between antiferroelectric and ferroelectric states in antiferroelectric compositions.^19,20

To design the composition presented in this study, both previous research and several practical considerations were taken into account, i.e., the avoidance of rare earth (RE) ions, Bi, and Pb. Previous research on high energy density ceramic capacitors has primarily focused on systems based on BaTiO₃ (BT), BiFeO₃ (BF), K_0.5Na_0.5NbO₃ (KNN), Na_0.5Bi_0.5TiO₃ (NBT), NaNbO₃ (NN), AgNbO₃ (AN), and lead-containing materials.^21–27 Among them, BT is currently most widely used as a base material for class II MLCCs and was thus selected for this study. NN, which has the highest bandgap among the common lead-free perovskites,¹⁵ was chosen as part of the solid solution to potentially enhance breakdown strength and improve E_max. Mg²⁺ was used as a dopant on the B-site to tune conductivity (p vs n) and achieve a higher reliability for the dielectric.^28,29 Bi³⁺ was not adopted as a dopant or alloying addition due to its thermodynamic incompatibility with Ni electrodes, which would necessitate the use of expensive Pt or Ag/Pd internal electrodes for MLCCs, making them less attractive from a practical perspective.^21,30,31 Pb²⁺ was excluded due to its negative health effects,³² and RE ions are not considered due to their cost, scarcity, and geopolitical uncertainty of their availability.³ 

The piezoelectric, dielectric, and ferroelectric properties of BT-NN solid solutions have been reported previously,^33–35 but to date its potential as a high voltage, temperature stable dielectric has not been considered due to its high conductivity. Here, we report the crystal structure, electrical microstructure, and dielectric and energy storage performance of BTNN-100xMg ceramics. Their potential for use in high voltage capacitors is discussed and the next steps toward commercialization assessed. BTNN-100xMg (x = 0, 0.005, 0.01, 0.02, 0.05) ceramics were prepared by conventional solid state reaction method. The fabrication and characterization methods are detailed in the supplementary material.

The XRD patterns of BTNN-100xMg ceramics are shown in Fig. 1, which exhibit a mixture of perovskite phases, with a secondary Ba₅Ti₁₄Nb₂O₃₉ (PDF 00-038-1431) phase detected for x < 0.02, Ba₁₁Ti₂₈O_66.5 (PDF 01-073-5502) for x = 0.02, and Ba₄Ti₁₁O₂₆ (PDF 01-083-1459) for x = 0.05. The changes in XRD profile of the secondary phase are shown in Fig. S1. The peaks of perovskite phases, as exemplified by (110) in Fig. 1(b), shift to higher angles with x ≤ 0.02, then back to lower 2θ at x = 0.05. Effectively, the unit cell volume initially shrinks and then expands. The appearance of a Nb containing second phase for x < 0.02 is consistent with volatilization of Na which leaves Nb in excess, compensated by the formation Ba₅Ti₁₄Nb₂O₃₉ second phase. As Mg concentration increases, however, the second phase for x ≥ 0.02 no longer contains Nb suggesting that it is retained within the perovskite matrix. The simplest explanation is that Mg enters in small concentrations onto the perovskite A-site, compensating in part for Na-ion vacancies and allowing Nb to be retained. However, we note that Mg²⁺ in 12-fold (A-site) coordination is unknown since its ionic radius is normally considered too small to occupy the cuboctahedral volume. Mg²⁺ in small concentrations on the A-site, however, would explain the reduction in unit cell volume for x ≤ 0.02. The expansion of the lattice for x > 0.02 may relate to the substitution of Mg²⁺ for Ti⁴⁺ ions onto the B-site, which is compensated by O-vacancies giving rise to a small repulsive force on the B-site sublattice. Mg²⁺ is typically known to occupy the B-site in perovskite titanates.^36,37 The substitution of Mg onto the B site is consistent with changes to the A_1g vibration mode shown at ∼770 and ∼830 cm⁻¹ in the Raman spectra in Fig. S3.³⁸ 

Rietveld refinements were performed on the XRD data to analyze the crystal structure of the samples. A single or combination of tetragonal $P4mm$ and cubic $Pm 3 ¯m$ space group was adopted in the fitting with the best result obtained with coexistence of the two symmetries as shown in Fig. S2. The refinements confirm that the unit cell volume of these two phases decreases slightly and then increase with x, as shown in Fig. S4. The phase fraction of $P4mm$ and $Pm 3 ¯m$ is ∼50:50 for x ≤ 0.02, but a significant increase in the proportion of cubic phase is observed for x = 0.05. This trend is confirmed by the disappearance of the B₁ mode at ∼301 cm⁻¹ in the Raman spectra (Fig. S3).³⁹ Based on the ionic displacements resolved in the Rietveld refinement, the spontaneous polarization of the BTNN-100xMg samples is estimated using the Shimakawa model, with details shown in the supplementary material.^40,41 As shown in Fig. S2(d), the total spontaneous polarization decreases from ∼6 to ∼1 μC/cm² with increasing dopant concentration.

The microstructure of BTNN-100xMg ceramics display dramatic changes with x, as shown in Fig. S5. For x < 0.02, the average grain size is ∼0.5 μm, with some extremely small grains and a few that are larger but with none >2 μm. For x = 0.02, the grain boundaries become distinct accompanied by an increase in the average grain size to ∼1.5 μm. For x = 0.05, the average grain size increases dramatically to >10 μm. The increase in grain size is commensurate with the onset of acceptor Mg²⁺ doping onto the B-site which has been shown in BT based materials to enhance mass transfer.⁴² 

The permittivity temperature plots of BTNN-100xMg ceramics are shown in Fig. 2(a). In the undoped BT10NN sample, the broad phase transition suggests that the transformation from cubic to tetragonal phase is not at a distinct Curie temperature, but rather over >100 K. This is also the case for x = 0.005 and x = 0.01 and is attributed to an inhomogeneous composition and microstructure. However, a single, broad, and large peak in permittivity is observed for x = 0.02, which decreases in temperature with increasing dopant concentration. The high dielectric loss of x = 0.05 suggests that the high permittivity is not entirely due to the contribution of the polar phase but also due to space charge from clusters of defects, most likely O vacancies. As shown in Fig. S6, the best dielectric temperature stability is achieved in x = 0.01, which meets the commercial specification of X6R MLCC (ΔC/C₀ ≤ ±15% from −55 to 105 °C). For x = 0.01, an additional dielectric anomaly is detected at around 32 K in Fig. 2(b), which may be associated with a transition to a lower symmetry phase.

The effect of Mg²⁺ dopant on the conduction mechanism of BTNN-100xMg ceramic is shown in Fig. S7. As the Mg²⁺ concentration increases, the conduction shifts from n- to p- then back to n-type for x = 0.05. The shift from n to p is due to acceptor doping. A possible mechanism for the shift back to n is the excessive grain growth that limits the level of grain boundaries in the ceramic, which means there is less channels for oxygen to enter the samples during the cooling process from sintering. This would leave more oxygen vacancies for x = 0.05, which makes the sample n-type and more conductive. The total resistivity at 550 °C in air increases, then drops at x = 0.05, which may be due to an increase in the concentration of oxygen vacancies with acceptor doping.

To investigate the electrical microstructure of the BTNN-100xMg ceramics, impedance data were collected on samples between 450 and 650 °C. The contribution of different components could be distinguished by analyzing the frequency spectrum of the imaginary part of impedance (Z″) and imaginary part of the modulus (M″), as shown in Figs. 3(a) and 3(b). It is clear from Fig. 3(b) that the M″ peaks of x ≤ 0.01 are broad, consisting of at least two Debye peaks with very similar peak frequencies, which makes it difficult to identify the contributions from individual electrical components. However, the M″ peaks begin to split at higher dopant levels. Peak frequencies differ by at least two orders of magnitude, providing an opportunity for more detailed analysis. Thus, data from x = 0.02 are selected to illustrate how the electrical microstructure of these ceramics may be resolved from their impedance responses. The capacitances of the electrical components of x = 0.02 are extracted from the M″ peaks as shown in Fig. 3(c), and their Curie–Weiss behavior is fitted in Fig. 3(d). Details of the analyses are in the supplementary material.

From Fig. 3, we conclude that an electrical core–shell structure is shown in the impedance spectra of BTNN-100xMg ceramics. At low dopant levels, the core and shell regions share a similar capacitance of ∼35 pF/cm (as corrected for overall sample geometry), and a similar f_max at 10–100 kHz. For x > 0.01, the grains grow significantly, and the resistance of the shell becomes much larger than that of the core, which separates the characteristic frequency, allowing a more detailed analysis of those regions. f_max for the core component shifts to much higher frequency (off scale) with x increased to 0.05, indicating a significantly more conductive core region in the sample, which could be further investigated.

The conductivity and activation energy of the charge carriers in the ceramic and the core–shell components correspond to the response of R₁ and R₂ in the equivalent circuit, respectively. As indicated in Figs. 4(a) and 4(b), the shell components have, overall, a lower conductivity and higher activation energy than the core components. When comparing Figs. 4(a) and 4(b) with Fig. S8, it is evident that the resistive response of the ceramics is dominated by the shell component. Since no dielectric anomaly was detected in the dielectric temperature spectrum above ∼100 °C, an extrapolation of Arrhenius plot was performed to deduct the conductivity closer to room temperature. As shown in Figs. 4(c) and 4(d), x = 0.02 have the most resistive shell component and highest activation energy for both the core and shell components, which potentially improves the breakdown strength under applied field. However, this is not the case with the P–E measurement of the samples. This may due to the grain growth, which increased the volume fraction of the electrically more conductive core region, based on M″-f spectra.

The spontaneous polarization of BTNN-100xMg ceramics decrease with increasing dopant level, as shown in Fig. S2(d). This trend could also be observed in the P–E loops shown in Fig. 5(a), with polarization measured at 380 kV/cm decreasing from ∼28 to ∼24 μC/cm² for x = 0.00 to 0.02. However, the measured polarization value is significantly larger than the calculated spontaneous polarization (∼5 μC/cm²). The difference in maximum polarization measured in P–E loop and spontaneous polarization is often the case for relaxor materials. A crossover from dielectric to semiconductor behavior is seen for x > 0.02. As shown in Fig. S9, the I–V curve of Au-5Mg-Au device is typical for a double Schottky barrier device.⁴³ 

In BTNN-100xMg ceramics, the optimum energy storage performance was achieved for x = 0.01, which has W_rec = 3.4 J/cm³ with η of 82.6% at 400 kV/cm. As shown in Fig. 5(b), E_max and energy storage density of the samples increased, whereas the efficiency of energy storage dropped with increasing dopant concentration. The enhanced energy storage performance may associate with a more resistive core region with elevated activation energy. Additionally, x = 0.01 shows a broad plateau in permittivity as a function of temperature. In Figs. 5(c) and 5(d), the energy discharge performance of x = 0.01 is evaluated in an underdamped and overdamped R–C circuit. In both scenarios, the sample is able to discharge the electrical energy to the resistor within 1 μs.

In this paper, the criteria for evaluating energy storage dielectric materials are briefly discussed. The crystal structure of BTNN-100xMg ceramics were analyzed through Rietveld refinement of the x-ray data to obtain the phase assemblage and the microscopic spontaneous polarization of the ceramics. The electrical microstructures of the Mg-doped BTNN ceramics were analyzed with impedance spectroscopy data and ε_r vs T assessed. With an optimized W_rec of 3.4 J/cm³, η of 82.6% at 400 kV/cm, and meeting the X6R specification, Mg-doped BTNN ceramics are a good candidate for high voltage MLCCs. However, 1440 °C is too high to use Ni internal electrodes in the fabrication of MLCCs and glass additions would be required to reduce the sintering temperature and, thus, potentially make compositions compatible with base metal technology. Preliminary results with 1% glass addition have suggested a reduction in sintering temperature is achievable while retaining a usable dielectric response but further work is required to develop reliable MLCC formulations. Nonetheless, these compositions show promise and represent a sensible direction for research into high voltage/energy storage capacitors where there is a clear commercial need for Pb, Bi, and RE free dielectric compositions.