Energy storage properties of 0.87BaTiO3–0.13Bi(Zn2/3(Nb0.85Ta0.15)1/3)O3 multilayer ceramic capacitors with thin dielectric layers

Abstract: Multilayer ceramic capacitors (MLCCs) for energy storage applications require a large discharge energy density and high discharge/charge efficiency under high electric fields. Here, 0.87BaTiO₃-0.13Bi(Zn_2/3(Nb_0.85Ta_0.15)_1/3)O₃ (BTBZNT) MLCCs with double active dielectric layers were fabricated, and the effects of inner electrode and sintering method on the energy storage properties of BTBZNT MLCCs were investigated. By using the pure Pt as inner electrode instead of Ag_0.6Pd_0.4 alloys, an alternating current (AC) breakdown strength (BDS) enhancement from 1047 to 1500 kV/cm was achieved. By investigating the leakage current behavior of BTBZNT MLCCs, the Pt inner electrode and two-step sintering method (TSS) were confirmed to enhance the Schottky barrier and minimize the leakage current density. With relatively high permittivity, dielectric sublinearity, and ultra-high BDS, the Pt TSS BTBZNT MLCCs exhibited a surprisingly discharge energy density (U_dis) of 14.08 J/cm³. Moreover, under an operating electric field of 400 kV/cm, the MLCCs also exhibited thermal stability with U_dis variation < ±8% over a wide temperature (t) range from -50 to 175 °C and cycling reliability with U_dis reduction < 0.3% after 3000 charge-discharge cycles. These remarkable performances make Pt TSS BTBZNT MLCCs promising for energy storage applications.

Keywords: BaTiO₃; multilayer ceramic capacitor (MLCC); leakage current; energy storage

1 Introduction

Multilayer ceramic capacitors (MLCCs) have been widely investigated because of their high power density, fast charge–discharge capability, and long lifetime, compared with lithium-ion batteries, fuel cells, and electrochemical super capacitors ^[1–3]. They have potential applications in portable electronics, electric vehicles, medical devices, and pulsed power weapons; hence, low-cost and environmentally friendly MLCCs are urgently needed to fulfill these requirements ^[4,5]. Through the integration of multiple layers of ceramic capacitors, a large amount of energy can be stored and released by the MLCCs. The energy density (U) can be calculated by integrating the electric field (E) with the polarization (P) according to P–E hysteresis loops as shown in Eq. (1):  

Recently, BaTiO₃–BiMeO₃ (Me indicates the trivalent or meanly trivalent metallic cations) relaxor ferroelectrics have been of significant research interest ^[6–10]. Their markedly higher discharge/charge efficiency (E_eff) compared with ferroelectrics, allows for the realization of superior energy storage properties. In previous Ref. ^[11], 0.87BaTiO₃–0.13Bi(Zn_2/3(Nb_0.85Ta_0.15)_1/3)O₃ MLCCs with dielectric thickness (D) of 4.8 μm were fabricated using the Ag_0.6Pd_0.4 (AgPd) inner electrode and two-step sintering method (TSS) ^[12]. The submicron grains (mean grain size, G = 434 nm) of 0.87BaTiO₃–0.13Bi (Zn_2/3(Nb_0.85Ta_0.15)_1/3)O₃ (BTBZNT) MLCCs are much smaller than the micron grains of reported ceramics for energy storage applications ^[13–21], and the dielectric layers therefore contained over 10 grains to obtain stable dielectric property. Because of their thin dielectric layers ^[22–28], remarkable AC breakdown strengthen (BDS) of 1047 kV/cm and a maximum discharge energy density (U_max) of 10.1 J/cm³ were achieved. 

Such MLCCs with thin dielectric layers are typically subjected to high AC operating electric fields to store and release a large amount of energy, and the possible thermal breakdown may occur when heat generated from the leakage current exceeds the heat lost ^[29]. The leakage current density (J) of the MLCCs is reported strongly dependent on the thermodynamic temperature (T) and electric field (E), and the Schottky emission model ^[30] is used to describe these leakage current behaviors, as shown in Eq. (2): 

where e is the effective electron mass, m* is the effective Richardson constant, k is the Boltzmann constant, h is the Plank constant, φ_B is the Schottky barrier height, ε₀ is the vacuum permittivity, and ε_r is the relative permittivity at optical frequency. In an ideal Schottky barrier ^[31] (no electrode–dielectric interface states considered), the difference between the metal work function (φ_m, φ_Pt = 5.6 eV, φ_Ni = 5.1 eV ^[32], φ_Pd = 5.5 eV ^[33], m = Pt, Ni, Pd, etc.) and the semiconductor affinity (χ, χBaTiO₃ = 3.9 eV) is the Schottky barrier height of the electrode–dielectric interface ^[34], as shown in Eq. (3):

φ_B= φ_m  - χ   (3) 

However, the electrode–dielectric counter-diffusion is inevitable during the sintering process, and as-formed interfacial alloys ^[35–37] of minor work functions (φ_Ba = 2.7 eV, φ_Ti = 4.3 eV ^[31]) lower the Schottky barrier height. An effective approach to depress the formation of alloy is using a smart sintering method ^[38]. By increasing the heating rate from 200 to 3000 ℃/h, the interfacial alloy became thinner, and the Schottky barrier height was improved from 1.14 to 1.27 eV ^[39] in BaTiO₃ MLCCs. Therefore, this indicates that by using inner electrodes with large work function and a suitable sintering method, the Schottky barrier of MLCCs can be heightened, and thus, the leakage current density can be minimized. 

In this study, BTBZNT MLCCs with thin dielectric layers using different inner electrodes and sintering methods were fabricated. The effects of inner electrode and sintering method on the energy storage property of BTBZNT MLCCs were investigated. 

2 Material and methods 

2. 1 Powder synthesizing 

BTBZNT powders were synthesized via conventional solid-state reactions. Starting powders of BaTiO₃ (99.9% purity), Bi₂O₃ (analytical reagent, AR), ZnO (AR), Nb₂O₅ (99.99% purity), and Ta₂O₅ (99.9% purity, Ningxia Orient Tantalum Industry Co., Ltd., China) were first stoichiometrically weighed, and then ballmilled for 24 h in an isopropanol (AR) medium. The well-milled slurry was separated and dried at t = 80 ℃ for 12 h. The obtained powders were calcined at t = 900 ℃ for 4 h. All the powders except for Ta₂O₅, were purchased from Sinopharm Chemical Reagent Co., Ltd., China. 

2. 2 MLCCs processing 

The calcined powders were mixed in a solution with 27 wt% ethyl acetate (AR), 27 wt% ethyl alcohol (AR), and 1 wt% triglycerides as dispersant. The suspension solution was then ball-milled using zirconia balls for 24 h. The organics, including 10 wt% polyvinyl butyral (aerospace grade) as binder, 2 wt% polyethylene glycol (AR) and 2 wt% butyl benzyl phthalate (AR, Aladdin Biochemical Technology Co., Ltd., China) as plasticizer, were added, and the slurry was then ball-milled for an additional 24 h. All the organics except butyl benzyl phthalate were purchased from Sinopharm Chemical Reagent Co., Ltd., China. The well-milled slurry was separated and left to stand for 24 h to remove air. After roll-to-roll tape casting with a test coater (CMD-S1.7/D/4.OH/3-778, Yasui Seiki Corp, Japan), continuous tapes with thickness variation < ±0.2 μm were formed. The Pt paste (MC-Pt100, GRIKIN Advanced Material Co., Ltd., China) was then screen printed on the green tapes with a size of 2.7 mm × 3.8 mm to serve as the inner electrode with an automated printing machine (NPM-1Y01, Yodogawa NCC Co., Ltd., Japan). Electrode patterned tapes were stacked with a 0.3 mm off-set from each other so that the effective electrode area was 2.7 mm × 3.2 mm. The active dielectric layer quantity and dielectric thickness were controlled using an automated laminating machine (NSM-1Y01, Yodogawa NCC Co., Ltd., Japan). After these processes, the BTBZNT MLCCs with double active dielectric layers (D ≈ 5 μm) were fabricated via the two-step sintering method (TSS, held at t = 1185 ℃ for 1 min and then at t = 1015 ℃ for 3 h) or the one-step sintering method (OSS, held at t = 1115 ℃ for 2 h) in air. 

2. 3 Characterization and testing method 

The micro-structure was characterized via scanning electron microscopy (SEM; Supra 40/40vp, Carl Zeiss Corp, Germany) at an operating voltage of 15 kV and high-resolution transmission electron microscopy (HRTEM; JEM-2010F, JEOL Ltd., Japan) at an operating voltage of 200 kV. The phase structures were detected via X-ray diffraction (XRD; D8 advance-A25, Bruker Co., Ltd., Germany). The temperature dependence of ε_r and dielectric loss (tanδ) were measured via an impedance analyzer (HP4278A; Hewlett-Packard, USA) with a temperature controller (Delta Design 9023, Cohu Semiconductor Equipment Group, USA). Hysteresis loops were measured using a ferroelectric measuring system (TF ANALYZER 2000E, aixACCT SystemsGmbH, Germany) at a frequency ( f ) of 10 Hz. The ε_r versus E curves were measured using a power device analyzer/curve tracer (Agilent B1505A; Agilent Technologies, USA) with an applied maximum direct current (DC) voltage (E = 400 kV/cm) and AC bias voltage (E = 10 kV/cm) at f = 10 Hz. The AC BDS at f = 10 Hz was measured using a high-voltage tester (YD2670B, Yangzi Electronic Co., Ltd., China). Unless specifically mentioned, all characterization and testing methods were carried out at room temperature (RT, t = 25 ℃). 

3 Results and discussion 

Figure 1 shows the micro-structure of the Pt BTBZNT MLCCs. The two-step sintering method ^[40], a universal method for preparing nanocrystalline ceramics, helped to decrease G of the Pt BTBZNT MLCCs from 346 to 271 nm. 

Fig. 1 SEM images of (a) Pt OSS and (c) Pt TSS BTBZNT MLCCs surface. Grain size distribution of (b) Pt OSS and (d) Pt TSS BTBZNT MLCCs. 

Figure 2 shows the XRD patterns of the BTBZNT powders and Pt MLCCs. Perovskite phase structures with no split of the (002) and (200) peaks were detected, implying that all the dopants diffused into BaTiO₃ lattices forming a homogenous perovskite solid solution with symmetrical crystalline structure. The broader (002) peak of the Pt TSS MLCCs (full width at half maximum, 2θ_FWHM = 0.142°) than the Pt OSS MLCCs (2θ_FWHM = 0.104°) implied finer grains according to Debye–Scherrer equation, which is consistent with SEM results. 

Fig. 2 XRD patterns of the BTBZNT powders and Pt BTBZNT MLCCs, and magnified patterns from 2θ = 45–45.5°. 

Figure 3 shows the temperature dependence of the ε_r, dielectric loss, and temperature capacitance coefficients of the BTBZNT MLCCs. Because of the existence of PNRs in the grains ^[11], the BTBZNT MLCCs have a relatively high permittivity (ε_r ≈ 1200) and a low dielectric loss (tanδ < 0.5% at f = 100 kHz) at RT. Broad Curie peaks can be observed in all MLCCs, and they arise from the random fields created by cation disorder because of the size and charge differences that disrupt the development of long-range polar ordering ^[41,42]. The Curie temperature (T_C) shifts to a higher value with increasing frequency, exhibiting the relaxor phenomenon with a strong diffuse phase transition ^[43]. The modified Curie–Weiss law ^[44]is used to describe the diffuse phase transition and is given by Eq. (4): 

where ε_m is the maximum εr and t_m is t corresponding to εm. C' is the modified Curie–Weiss constant and γ is an indicator of the diffuseness degree. By changing the temperature dependence of ε_r at f = 1 kHz into the forms ln(1/ε – 1/ε_m) and ln(t – t_m), linear regressions with excellent correlations (goodness of fit, R² > 0.999) could be realized, as shown in Fig. 3 (insets). The large γ > 1.7 indicates the strong diffuse phase transition in the BTBZNT MLCCs. Because of the broad Curie peaks around RT, the temperature capacitance coefficients (TCCs) between 100 and 1 kHz are maintained at < ±15 % in the range of t = (–55) – 125 ℃, which meets the Electronics Industries Association (EIA) X7R specification. 

Fig. 3 Temperature dependence of the ε_r and dielectric loss of (a) Pt OSS, (c) Pt TSS, and (e) AgPd TSS BTBZNT MLCC. Insets are the modified Curies–Weiss law fitting. Temperature capacitance coefficient curves of (b) Pt OSS, (d) Pt TSS, and (f) AgPd TSS BTBZNT MLCC.  

The dielectric nonlinearity, which is the nonlinear reduction of the permittivity with increasing electric field, frequently degrades the energy storage properties of ceramics. The ε_r versus electric field and electric field capacitance coefficient (ECC) curves of the BTBZNT MLCCs are shown in Fig. 4. The gradual variation in ε_r with increasing E in the range of t = (–50) – 150 ℃ indicates the weak dielectric nonlinearity. The Johnson’s phenomenological expression based on the Devonshire’s phenomenological theory ^[45] is adopted to evaluate the dielectric nonlinearity, as shown in Eq. (5):

where ε_r(E) is the εr at E, ε_r(0) is the εr at E = 0 kV/cm, and αε_r(0)³ is an indicator of the degree of dielectric nonlinearity. By changing ε_r versus E curves into the forms (ε_r(0)/ε_r(E))3 and E², linear regressions (R² > 0.996) were achieved, as shown in Fig. 4. The αεr(0)₃ is one order lower than the SrTiO₃-surface-modified BaTiO₃ relaxor ferroelectrics (αε_r(0)³ ≈ 9×10^-4) ^[46] with an approximate similar G, demonstrating the dielectric sublinearity of the BTBZNT MLCCs. Moreover, a further reduction in αε_r(0)³ with increasing temperature significantly benefits the MLCC’s high-temperature energy storage properties, as shown in Fig. 4 (insets). 

Fig. 4 ε_r versus the electric field curves of (a) Pt OSS, (d) Pt TSS, and (g) AgPd TSS BTBZNT MLCC. Electric field capacitance coefficient curves of (b) Pt OSS, (e) Pt TSS, and (h) AgPd TSS BTBZNT MLCC. (c) Johnson’s phenomenological expression fitting of (g) Pt OSS, (h) Pt TSS, and (i) AgPd TSS BTBZNT MLCC. Insets show the temperature dependence of αε_r(0)³. 

Because all the MLCCs exhibited similar dielectric property, the effects of inner electrode and sintering method were negligible, considering of the uncertain effective electrode area and dielectric thickness after sintering. Figure 5 shows the complex impedance spectra and three series resistor–capacitance (3RC) equivalent circuit fitting of the BTBZNT MLCCs. By fitting the experimental data with 3RC equivalent circuit ^[47,48] as shown in Fig. 5, which were assigned to grain, grain-boundary, and electrode–dielectric interface. The Arrhenius relationship is adopted to evaluate the activation energy (E_a) of the insulation resistance (R_ins),as shown in Eq. (6):

where R₀ is the characteristic resistance. By changing the Arrhenius relationship into the forms ln(R_ins) and 1000/T as shown in Fig. 5. The Ea corresponding to the interface insulation resistance of the Pt TSS BTBZNT MLCC (E_a = 2.85 eV) was higher than that of the Pt OSS BTBZNT MLCC (E_a = 2.26 eV) and the AgPd TSS BTBZNT MLCC (E_a = 2.70 eV). The insulation resistance of grain (R_g), grain-boundary (R_gb), and interface (R_int) of the BTBZNT MLCCs at various temperatures are listed in Table 1. 

Table 1 Comparison between the insulation resistances of the BTBZNT MLCCs 

Fig. 5 Complex impedance spectra of experiment data and 3RC equivalent circuit fitting of (a) Pt OSS, (d) Pt TSS, and (g) AgPd TSS BTBZNT MLCC. Insets are magnified complex impedance spectra at higher temperatures. Arrhenius relationship fitting of the activation energy of (d) Pt OSS, (e) Pt TSS, and (f) AgPd TSS BTBZNT MLCC. 

To further investigate the conduction mechanisms, current–voltage measurements were performed. Figure 6 shows the leakage current curves and the Schottky thermionic emission model fitting of the BTBZNT MLCCs. Despite the approximate maximum leakage current density (J_max = 2.15 μA/cm²) at T = 348 K, the higher J was detected for the AgPd TSS MLCC (J_max = 44.5 μA/cm²) than the Pt OSS MLCC (Jmax = 27.8 μA/cm²) and the Pt TSS MLCC (Jmax = 24.2 μA/cm²) at T = 423 K, implying different φ_B according to Eq. (1). Leakage current data under E > 250 kV/cm were adopted and linear correlations of ln(J/T²) and E^1/2 could be realized, guaranteeing the reliability of the Schottky thermionic emission model fitting results. The φ_B of the AgPd TSS, Pt OSS, and Pt TSS MLCC, evaluated from the plots of extrapolated values of ln(J/T²) for E→0 versus 1000/T in Fig. 6 (insets), were 0.407, 0.431, and 0.489 eV, respectively. 

Fig. 6 Leakage current curves of (a) AgPd TSS, (b) Pt OSS and (c) Pt TSS BTBZNT MLCC. Schottky thermionic emission model fitting of (d) AgPd TSS, (e) Pt OSS, and (f) Pt TSS BTBZNT MLCC. Insets show the Schottky barrier height fitting. 

The higher φ_B was obtained for the Pt TSS than AgPd TSS MLCC. It can be explained by the difference of the work function between Pt (φ_m = 5.6 eV) and Ag_0.6Pd_0.4 alloy (φ_m = 4.6 eV), which was evaluated by the Freeouf empirical model ^[49] as given by Eq. (7):

where φ_M and φ_m are the work functions of metal M and m, respectively, and x is the concentration of the metal. In addition, a higher φ_B was also obtained for the Pt TSS than Pt OSS MLCC, indicating that the two-step sintering method suppressed the formation of the interface (the thickness decreased from 35 to 30 nm as shown in Fig. 7), and thus heightened the Schottky barrier. 

Fig. 7 TEM images of the electrode–dielectric interface of (a) Pt OSS and (b) Pt TSS BTBZNT MLCC. Elements line scanning of the electrode–dielectric interface of (c) Pt OSS and (d) Pt TSS BTBZNT MLCC.

Figure 8 shows the hysteresis loops and the energy storage property of the BTBZNT MLCCs. The Pt TSS BTBZNT MLCC had an exceptional E_max = 1500 kV/cm, which was higher than the Pt OSS (E_max = 1301 kV/cm) and AgPd TSS (E_max = 1047 kV/cm) BTBZNT MLCC. Thus it was confirmed that using Pt and TSS were effective approaches minimizing the leakage current density, and thus enhancing the breakdown strengthen. The Weibull distribution was used to evaluate the AC BDS data, as given by Eq. (8):

where P is the cumulative probability of failure, α is a scale parameter characterizing the breakdown strength, β is a shape parameter indicating the dispersion of the data, and E is the critical electric field above which breakdown occurs. The BDS data are ranked in ascending order and the values of P are estimated by Eq. (9): 

where i is the rank and N is the total number of samples. α and β are then determined via linear regression of ln[–ln(1–P)] and lnE, as shown in Fig. 8(d) (inset). The large β > 8 guaranteed the reliability of the Weibull analysis, and α was in good agreement with the E_max. Despite of lowered Eeff under increasing E, the U_dis of the BTBZNT MLCC was compensated by the increasing maximum polarization, showing linear correlations of E and Udis in Fig. 8(d). Because of the enhanced Emax, the U_max of the TSS BTBZNT MLCC was improved from 10.1 to 14.1 J/cm³ using Pt instead of AgPd inner electrode. 

Fig. 8 Hysteresis loops of (a) AgPd TSS, (b) Pt OSS, and (c) Pt TSS BTBZNT MLCC. (d) Energy storage property of the BTBZNT MLCCs. Inset shows the Weibull distribution fitting results of AC BDS. 

The Pt TSS BTBZNT MLCC also had an excellent thermal stability and the cycling reliability of over 3000 charge–discharge cycles, as shown in Fig. 9, which were generally superior to the reported energy storage MLCCs ^{[6,9,50–54]}, as listed in Table 2. 

Fig. 9 (a) Hysteresis loop–temperature evolution. (b) Temperature dependence of the discharge energy density and discharge/charge efficiency. (c) Hysteresis loop–cycling evolution. (d) Cycling reliability of the discharge energy density and discharge/charge efficiency of the Pt TSS BTBZNT MLCC. 

Table 2 Comparison between the energy storage property and the thermal stability of the reported MLCCs 

4 Conclusions 

BTBZNT double-layer ceramic capacitors with D ≈5 μm using the Pt or AgPd inner electrode were successfully fabricated via a one-step or two-step sintering method, and the Schottky thermionic emission was confirmed to dominate the conduction mechanism of the BTBZNT MLCCs. By investigating the leakage current behaviors of the BTBZNT MLCCs, using the Pt and TSS were proved to be effective approaches heightening φ_B, minimizing J, and thus enhancing BDS. The Pt TSS BTBZNT MLCC had the remarkable energy property with an E_max = 1500 kV/cm and a U_max = 14.1 J/cm³, and an excellent thermal stability with variation < ±8% at t = (–50) – 150 ℃ and E = 400 kV/cm. The Pt TSS BTBZNT MLCC was generally superior than recently reported energy storage MLCCs, and the improvements in their electric properties further indicate that they have high potential for energy storage applications.

References: omitted