Zhilun
Lu‡
ab,
Ge
Wang‡
a,
Weichao
Bao‡
c,
Jinglei
Li‡
d,
Linhao
Li
a,
Ali
Mostaed
ae,
Huijing
Yang
af,
Hongfen
Ji
ag,
Dejun
Li
h,
Antonio
Feteira
i,
Fangfang
Xu
c,
Derek C.
Sinclair
a,
Dawei
Wang
*a,
Shi-Yu
Liu
*h and
Ian M.
Reaney
*a
aDepartment of Materials Science and Engineering, University of Sheffield, Sheffield, S1 3JD, UK. E-mail: dawei.wang@sheffield.ac.uk; i.m.reaney@sheffield.ac.uk
bThe Henry Royce Institute, Sir Robert Hadfield Building, Sheffield, S1 3JD, UK
cState Key Laboratory of High Performance Ceramics and Superfine Microstructure, Shanghai Institute of Ceramics, Shanghai, 200050, China
dElectronic Materials Research Laboratory, Key Laboratory of the Ministry of Education and International Center for Dielectric Research, Xi’an Jiaotong University, Xi’an 710049, Shaanxi, China
eDepartment of Materials, University of Oxford, Parks Road, Oxford, OX1 3PH, UK
fDepartment of Physics, Tangshan Normal University, Tangshan 063000, China
gLaboratory of Thin Film Techniques and Optical Test, Xi’an Technological University, Xi’an 710032, China
hCollege of Physics and Materials Science, Tianjin Normal University, Tianjin 300387, China. E-mail: shiyuliu@mail.tjnu.edu.cn
iMaterials and Engineering Research Institute, Sheffield Hallam University, Sheffield, S1 1WB, UK
First published on 21st August 2020
The Gerson–Marshall (1959) relationship predicts an increase in dielectric breakdown strength (BDS) and therefore, recoverable energy density (Wrec) with decreasing dielectric layer thickness. This relationship only operates however, if the total resistivity of the dielectric is sufficiently high and the electrical microstructure is homogeneous (no short circuit diffusion paths). BiFeO3–SrTiO3 (BF–ST) is a promising base for developing high energy density capacitors but Bi-rich compositions which have the highest polarisability per unit volume are ferroelectric rather than relaxor and are electrically too conductive. Here, we present a systematic strategy to optimise BDS and maximum polarisation via: (i) Nb-doping to increase resistivity by eliminating hole conduction and promoting electrical homogeneity and (ii) alloying with a third perovskite end-member, BiMg2/3Nb1/3O3 (BMN), to reduce long range polar coupling without decreasing the average ionic polarisability. These strategies result in an increase in BDS to give Wrec = 8.2 J cm−3 at 460 kV cm−1 for BF–ST–0.03Nb–0.1BMN ceramics, which when incorporated in a multilayer capacitor with dielectric layers of 8 μm thickness gives BDS > 1000 kV cm−1 and Wrec = 15.8 J cm−3.
Broader contextSuccessful decarbonisation of the world economy will be underpinned by efficient energy harvesting and storage technologies. Batteries exhibit high energy density but low power density, making them unsuitable for applications requiring fast charge–discharge rates. In contrast, dielectric capacitors widely used in pulsed power electronics have intrinsically higher power density. The next generation of dielectric capacitors for energy storage applications will be required to have higher energy density, discharge efficiency, and breakdown strength for higher voltage applications. In addition, finding less toxic, lead-free materials has been a major scientific and technological challenge. Here, we demonstrate tailored dopant strategies in lead-free BiFeO3–SrTiO3 (BF–ST) based energy storage capacitors to increase resistivity, promote electrical homogeneity and reduce polar coupling without decreasing the average ionic polarisability. By employing both strategies, a record high energy density in BF–ST is attained in multilayer ceramic capacitors. This work sheds new light on designing ultrahigh energy storage materials and can be employed in most high polarisability oxide-based materials. |
The recoverable energy storage density of dielectric capacitors (Wrec) can be obtained using the following equation:
(1) |
In general, conductivity σ is given by:
σ = nqμ, | (2) |
Aliovalent doping has long been an effective way to tailor the electrical performance of oxides. Acceptor (e.g. Ca2+ for Ti4+) doping in BaTiO3 facilitates the formation of oxygen vacancies, prevents the reduction of Ti4+ when sintered in low oxygen partial pressure atmospheres and therefore, reduces leakage current.22,23 In contrast, donor doping BaTiO3 (e.g. La3+ for Ba2+) creates metal (Ti) vacancies that result in the loss of oxygen (δ) from the lattice when sintering in air above 1300 °C: Ba1−xLaxTi1−x/4O3−δ ceramics are n-type semiconducting for low levels of x.24,25 In some cases, there can be significant levels of oxide-ion conductivity such that, ionic as opposed to electronic conduction, dominates the electrical properties. For example, oxygen vacancies can occur in Na1/2Bi1/2TiO3 (NBT) due to acceptor doping and/or Bi2O3-loss during processing and leads to high levels of oxide-ion conductivity.26,27 Donor doping with low levels of Nb5+ for Ti4+via the mechanism in eqn (3) is an effective method to compensate for the formation of oxygen vacancies during processing and suppresses ionic conductivity such that dielectric behavior is restored.26
(3) |
BiFeO3 (BF)-based ceramics are a promising ‘next generation’ of lead-free electroceramics for applications as high energy density capacitors, piezoelectrics, ferroelectrics, electrocalorics and multiferroics.28–38 However, the low BDS (high leakage current), resulting from the volatilisation of Bi2O3 and consequent oxidation of some Fe3+ to Fe4+ ions is one of the main obstacles to their commercial usage.39,40 In this work, BiFeO3–SrTiO3 (BF–ST)-based ceramics are used as a case study to illustrate how materials can be systematically modified to create exceptional dielectric properties suitable for electroceramic applications.
0.4BF–0.6ST has been previously proposed as a novel and high temperature lead-free ceramic achieving high energy density of Wrec = 6.0 J cm−3 at 420 kV cm−1 in multilayer ceramic capacitors (MLCCs).41 We propose that BF-rich compositions, which were not reported by the authors of ref. 41, could potentially exhibit higher energy density due to their higher ionic polarisability per unit volume which leads to the appearance of spontaneous polarisation. BF-rich compositions are typically p-type semiconductors39 and too leaky for high energy density capacitor applications. Moreover, they are predominantly ferroelectric rather than relaxor and hence in their initial state have intrinsically low Wrec.42 A systematic study was therefore undertaken of 0.6BF–0.4ST (reported as a nominal morphotropic phase boundary, MPB, composition)43 using (i) doping with Nb on the B-site to control σ and induce electrical homogeneity and (ii) alloying with a third end-member, BiMg2/3Nb1/3O3 (BMN) to promote a relaxor state without decreasing the average ionic polarisability. Record-high Wrec of 8.2 and 15.8 J cm−3 for lead-free materials were achieved in 0.5BF–0.4ST–0.03Nb–0.1BMN bulk ceramics and multilayer capacitors, respectively. This work, therefore, defines clear engineering guidelines for designing ultrahigh energy storage materials and can readily be adapted for most high polarisability oxide-based systems.
Impedance spectroscopy data for 0.6BF–0.4ST ceramics with 0, 1 and 3% Nb are shown in Fig. 1. A Z* plot for 0.6BF–0.4ST ceramic (Fig. 1a) exhibits a distorted and broadened arc response at room temperature (RT), with a total resistivity of ∼180 kΩ cm. This is consistent with 0.6BF–0.4ST ceramics being leaky dielectrics at RT. The corresponding Z′′ and M′′ spectroscopic plots (Fig. 1b) show clear evidence of large Debye peaks with the corresponding frequencies for Z′′ and M′′ maximum, fmax, separated by ∼2 orders of magnitude (Fig. 1b). The data are analysed using an equivalent circuit based on two parallel resistor–capacitor (RC) elements connected in series. R and C values are extracted from the Z′′ and M′′ spectra as described previously.11 The extracted C values are ∼1 × 10−10 F cm−1 (based on M′′ data, component 2) and ∼4.5 × 10−10 F cm−1 (based on Z′′ data, component 1) which indicates they are bulk (grain) and grain boundary responses, respectively.46 The full width half maximum (FWHM) of the Z′′ and M′′ peaks exceeds 2 orders of magnitude in the frequency spectrum suggesting there is considerable electrical heterogeneity associated with 0.6BF–0.4ST ceramics.32
For 1 and 3% Nb doped 0.6BF–0.4ST, ceramics are too resistive to measure by impedance spectroscopy at RT and higher temperatures are required to obtain useful data. The Z* plots and Z′′, M′′ spectra for these samples at 703 K are shown in Fig. 1c–e. The most obvious difference in the data is the presence of a single arc in the Z* data, Fig. 1c and a single Z′′ and M′′ Debye peak that occurs at the same frequency, Fig. 1d and e. The FWHM of these peaks are ∼1.2 orders of magnitude, closer to the 1.14 decades for an ideal Debye response. The data are analysed using a single parallel RC element associated with a bulk response and clearly shows that 1–3% Nb doping produces a dramatic increase in electrical resistivity and improves the electrical homogeneity of the ceramics. An Arrhenius plot of the bulk conductivity (extracted from M′′ spectra) for all ceramics in this series is given in Fig. 1f and shows that Nb doped ceramics are excellent dielectrics with bulk conductivity <25 μS cm−1 at 703 K in contrast to undoped ceramics with a bulk conductivity that exceed 1 μS cm−1 at RT. In addition to the decrease in conductivity, there is a dramatic increase in the Ea associated with the bulk conductivity from ∼0.37 to ∼1.18 eV. This clearly indicates a change in the mechanism. ∼0.4 eV is consistent with p-type electronic conduction reported in undoped BF and suggests the presence of mixed oxidation states of Fe3+ and Fe4+ ions.45 The band gap of BF is reported to be ∼2.09 to 2.32eV whereas for ST it is ∼3.2 eV.47,48 Although the exact band gap of BF–ST in this study is unknown, based on the end-members, an Ea of ∼1.2 eV is reasonably close to half the optical band gap, suggesting that the electronic conduction for the Nb-containing ceramics is close to the intrinsic (band gap) mechanism and beneficial for BDS.49,50 We conclude therefore, that p-type conduction mechanism has been successfully suppressed by Nb doping in BF–ST in accordance with eqn (3).
Impedance data for 0.54BF–0.4ST–0.06BMN ceramics sintered in air and N2 are shown in Fig. 2. Air sintered ceramics are leaky dielectrics at RT with a broadened and non-ideal arc in Z* with a total resistivity of ∼4 MΩ cm, Fig. 2a. The Z′′ spectrum displays a single, resistive element (component 1), R1 ∼ 3.5 MΩ cm and C1 ∼ 2.3 × 10−10 F cm−1 whereas the M′′ data shows the presence of a second element (component 2) that is more conductive, R2 ∼ 72 kΩ cm and C2 ∼ 1.1 × 10−10 F cm−1 and evidence for a third (yet more conductive) element with a peak maximum >1 MHz, Fig. 2b. These ceramics are clearly electrically heterogeneous with resistive and conductive components. The M′′ spectrum for component 2 has a bulk-related C value and an Arrhenius plot of the conductivity (1/R2) for this component is given in Fig. 2c and shows an Ea of ∼0.4 eV. This is consistent with the p-type mechanism obtained for 0.6BF–0.4ST ceramics, Fig. 1f.
In contrast, ceramics sintered in N2 are electrically insulating at RT and useful impedance data could only be obtained at >573 K. A Z* plot for a sample measured at 613 K is shown in Fig. 2a and displays a single, near ideal arc with a total resistivity of ∼0.16 MΩ cm. Z′′, M′′ spectra (not shown) displayed a single Debye peak with a maximum at the same frequency with C ∼ 1 × 10−10 F cm−1, confirming a bulk-type response. Processing in N2 dramatically improves the electrical homogeneity of the ceramics and suppresses the conductivity. An Arrhenius plot of the conductivity associated with the M′′ data is shown in Fig. 2c and displays much lower conductivity compared to air-fired ceramics with Ea ∼ 0.86 eV. Processing in N2 therefore, suppresses p-type conductivity but 0.54BF–0.4ST–0.06BMN ceramics are ∼one order of magnitude more conductive than Nb-containing 0.6BF–0.4ST ceramics at ∼623 K (1000/T = 1.6), Fig. 1f. This suggests that Nb-doping is more effective at suppressing p-type conductivity than processing in N2.
The appearance of a core–shell microstructure results from microchemical segregation, driven by immiscibility, into Bi, Fe and Sr, Ti rich regions during slow-cooling from sintering.31,32,37 Although the microstructure arises from chemical heterogeneity, electrical homogeneity and low electrical conductivity have been demonstrated in previous studies through the use of dopants which modifies the defect chemistry of the core and shell such that they have similar resistivity based on impedance spectroscopy data.19 The elimination of conducting cores through electrical homogeneity enhances BDS and facilitates the fabrication of thinner dielectric layers suitable for MLCCs which do not short circuit under large applied field (>500 kV cm−1).19
The Seebeck coefficient (S) of BF–ST–BMN–xNb ceramics as a function of temperature is shown in Fig. 3c. The positive S for x ≤ 0.02 indicates p-type conduction but the magnitude of S becomes smaller as x increases, presumably due to a decrease in n.52 For x ≥ 0.03, S changes from positive to mixed negative and positive as a function of temperature, indicating a shift from p- to n-type conduction.53 Typical Z* plots and combined Z′′ and M′′ spectroscopic plots at 703 K of BF–ST–BMN–xNb ceramics are provided in Fig. S3a and b (ESI†). For x > 0 compositions, only one semicircle is observed in the Z* plot (Fig. S3a, ESI†), with a single Debye peak in both Z′′ and M′′ spectroscopic plots located at the same frequency (Fig. S3b, ESI†), indicating that all Nb doped ceramics are electrically homogeneous with only one parallel RC element.21 The total electrical resistivity of the Nb doped ceramics is several orders of magnitude higher than undoped. The absence of a semicircle/peak in the impedance data between 12 and 273 K (Fig. S3c and d, ESI†) provides strong evidence that there are no high conductivity regions in the Nb doped samples. The incline in M′′ is not a conductive component but is due to relaxor-type behavior, as shown in Fig. S4 (ESI†). Nb Doping is therefore effective at suppressing the p-type conductivity and increasing the electrical homogeneity of BF–ST–BMN. The Ea increases from ∼0.41 eV for x = 0 to ∼1.20 eV for x = 0.04 and eventually drops to ∼1.13 eV for x = 0.05 (Fig. 3d), indicating near intrinsic behaviour (band gap conduction) at x = 0.03 and 0.04, which is conducive to high BDS.49,50
The unipolar PE loops at maximum electric fields (Emax) of BF–ST–BMN–xNb (x = 0.01, 0.02, 0.03, 0.04, 0.05) are shown in Fig. 3e. x = 0 is too conductive to be measured, however, Emax is significantly enhanced for x > 0 in agreement with impedance data (Fig. S3, ESI†). The calculated energy storage properties are plotted in Fig. 3f, with maximum Wrec = 6 J cm−3 and η = 74.6% at 360 kV cm−1 for x = 0.03.
Fig. 4 (a) The XRD patterns, (b) temperature dependent dielectric permittivity (εrvs. T) and loss (tanδ vs. T) data at 100 kHz, (c) Arrhenius plots of conductivity obtained from M′′ peaks, (d) unipolar P–E loops under Emax and (e) energy storage properties, Wrec and η for BF–ST–Nb–yBMN (y = 0.02, 0.04, 0.06, 0.08, 0.10 and 0.12). (f) A comparison of Wrec for y = 0.10 in this work with other lead-free bulk ceramics.11,12,55–89 |
ε r and tanδ vs. temperature for BF–ST–Nb–yBMN ceramics at 100 kHz (Fig. 4b) show all compositions exhibit a broad εr maximum (εmax) which is frequency dependent (Fig. S6, ESI†), consistent with relaxor behaviour. εmax decreases and broadens with increasing BMN concentration. The Z* plots obtained from the ceramics at 703 K exhibit a single semicircle for each composition (Fig. S7a, ESI†) and co-incident Debye peaks in Z′′ and M′′ spectroscopic plots (Fig. S7b, ESI†). All data can be modelled on a single parallel RC element where bulk-type behavior dominates and therefore samples remain electrically homogeneous.21 With increasing BMN concentration, ceramics become more resistive with a small increase in Ea (Fig. 4c), suggesting potential further enhancement in BDS.
W rec and η for BF–ST–Nb–yBMN (y = 0.02, 0.04, 0.06, 0.08, 0.10 and 0.12) ceramics extrapolated from unipolar P–E loops under Emax are shown in Fig. 4d and e. The substitution of BMN leads to slimmer and relaxor-like P–E loops with higher η and higher BDS. Wrec reaches a maximum of 8.2 J cm−3 at 460 kV cm−1 with η of 74.1% for y = 0.10, which is the highest Wrec reported to date for lead-free bulk ceramics (see Fig. 4f).11,12,55–89
A two-phase (rhombohedral and cubic, R + C) coexistence model is initially proposed in which the nano-polar regions (NPRs) with R structure are randomly nucleated and embedded in a C non-polar matrix, Fig. 5a. According to the local random field (LRF) model,92–95 the polar regions are oriented along different polarisation directions, are isotropic and macroscopically paraelectric (with zero macroscopic polarisation), in agreement with a superparaelectric model.96 The schematic picture of a three-dimensional (3D) Landau free-energy F() profile of polarisation rotation for BF–ST–Nb–BMN is given in Fig. 5b. F() of rhombohedral (R), tetragonal (T) and orthorhombic (O) phases follows the sequence, R < T < O, indicating that R-phase is the only polar phase retained at RT at zero field. However, the energy barrier to polarisation rotation between 〈111〉R and 〈100〉T is small, as shown in the two-dimensional (2D) Landau free-energy curve, Fig. 5c, which suggests the polar phase may easily undergo polarisation rotation between 〈111〉 and 〈100〉 directions under an external E. The 2D F() curve of polarisation extension for BF–ST–Nb–BMN is shown in Fig. 5d, where a small energy barrier between the paraelectric (C) and ferroelectric (R) phases is found. With an external E, the F() of the relaxor is lowered, as shown in Fig. 5d, which indicates a large polarisation response is induced under a modest E. Consequently, the NPRs in BF–ST–Nb–BMN are easy to re-orientate and re-arrange in the direction of E, Fig. 5e, due to the small energy barriers associated with polarisation rotation and extension.97
According to the above, as E increases, Pmax for BF–ST–Nb–BMN is expected to keep increasing by both polarisation rotation and extension due to the small values of F(). As a result, provided a high BDS is maintained, a continuous increase of P with E is predicted leading to high Wrec and large η.
An SEM image and corresponding energy-dispersive spectroscopy (EDS) elemental maps acquired from a cross section of a multilayer are illustrated in Fig. S8 (ESI†). The thickness and active electrode area of one dielectric layer are ∼8 μm (Fig. S8, ESI†) and 5 mm2, respectively. The interface between the Pt electrode and BF–ST–Nb–0.1BMN ceramic is characterised using (scanning) transmission electron microscopy ((S)TEM) with images, electron diffraction patterns and associated EDS maps shown in Fig. 6(a–f). According to the selected area diffraction (SAD) patterns in Fig. 6b and c, the {011} atomic planes of the ceramic grain displayed in Fig. 6a are ∼7° off parallel to Pt{011} which is considered as a nominally random arrangement. However, a coherent 10–20 nm intermediate layer is observed at the surface of Pt. We propose that Bi3+ in the BF-based ceramic is reduced by residual carbon from the binder at the Pt/dielectric interface and forms an alloy with the Pt.99 The EDS maps in Fig. 6e confirm the presence of Bi in the intermediate layer. In addition, discrete Bi–Pt phases, marked by α and β in Fig. 6d are observed, the former of which based on Fig. 6e and f, is indexed as Bi2Pt.
Fig. 6 (a) TEM micrograph obtained from an interface between a BF–ST–Nb–0.1BMN grain and a Pt grain (electrode). Here the Pt grain is close to its [001] zone axis and the BF–ST–Nb–0.1BMN grain is ∼10° off from its [311] zone axis. (b) SAD pattern obtained from the area marked by a blue circle shown on (a) i.e. interface between BF–ST–Nb–0.1BMN and Pt. Here, few reflections related to BF–ST–Nb–0.1BMN are marked by red circles. (c) SAD pattern obtained from the area marked by an orange circle shown on (a), i.e. Pt. (d) High resolution TEM micrograph obtained from the interface between ceramic and Pt shown in (a). Here different phases (e.g. α and β) are observed at the surface of Pt. (e) A bright-field STEM image and corresponding chemical EDS elemental maps obtained from the interface between ceramic and Pt shown in (a). (f) A bright-field STEM image obtained from the interface between Pt and α shown in (d). (g) Unipolar P–E loops. (h) The energy storage properties, Wrec and η, for BF–ST–Nb–0.1BMN multilayer. (i) A comparison for Wrec of the multilayer in this work with other lead-free multilayers reported in ref. 10, 12, 13, 21, 56 and 100–109. |
The unipolar P–E loops and energy storage properties for the BF–ST–Nb–0.1BMN multilayer are shown in Fig. 6g–i. With increasing E, Wrec increases until breakdown at >1000 kV cm−1, thereby illustrating the enhancement in BDS in accordance with the Gerson–Marshall relationship.98 The multilayer achieved the highest recorded Wrec of 15.8 J cm−3 with η = 75.2% for random polycrystalline lead-free MLCCs. Given the high BDS, it is not clear how the formation of the Bi–Pt alloy affects the high field electrical properties but it is likely that reduction will locally affect Pmax in the vicinity of the interface by decreasing the concentration of the Bi-rich phase. Whether this also creates a local space charge layer, more conducting than the bulk of the dielectric within the multilayer, remains to be elucidated.
X-ray diffraction (XRD) on ceramics was performed using a Bruker D2 phaser. The operating voltage and current were 40 kV and 30 mA, respectively. The polished surface morphology of ceramic samples was examined using a FEI Inspect F50 scanning electron microscope (SEM) with BSE and EDS detectors. Samples for (scanning) transmission electron microscopy ((S)TEM) were mechanically grounded to reach a thickness of ∼50 μm. Further polishing was conducted at liquid nitrogen temperature by Ar+ ion milling using a Gatan Precision Ion Polishing System PIPS (II). (S)TEM data were acquired with a JEOL JEM 2100F (JEOL, Tokyo, Japan) operated at 200 kV. Impedance spectroscopy (IS) data were collected using an Agilent E4980A (Agilent Technologies Inc., Palo-Alto, CA) from 20 Hz–2 MHz and from 12 to 873 K with an ac voltage of 100 mV. Resistivity of samples was obtained by fitting the experimental IS data using ZView software (Scribner Associates, Inc., Southern Pines, NC). The temperature dependence of the dielectric properties (permittivity and tanδ) were measured using an Agilent 4184A precision LCR meter from room temperature (RT) to 550 °C at 1, 10, 100, 250 kHz and 1 MHz. IS and LCR data were corrected by a geometric factor (thickness/surface area). For ferroelectric polarisation–field (P–E) loops measurements, ceramics were ground to a thickness of ∼0.15 mm and then gold sputtered. Bipolar P–E loops were obtained using an aixACCT TF 2000E ferroelectric tester at 1 Hz.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ee02104k |
‡ Z. Lu, G. Wang, W. Bao and J. Li contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2020 |