Large conventional and inverse electrocaloric effects in PbMg0.5W0.5O3 multilayer capacitors above and below the Néel temperature

Bulk PbMg0.5W0.5O3 (PMW) is an antiferroelectric in which an electric field of 12 V μm−1 is sufficient to initiate a nominally reversible transition to a dipole-aligned (ferroelectric) phase if operating just below the Néel temperature T N, near room temperature (Li et al 2021 Adv. Funct. Mater. 31 2101176). Here we describe multilayer capacitors (MLCs) of PMW that permit 27 V µm−1 to be applied without breakdown. Below T N, nominally reversible driving of the partial (full) antiferroelectric–ferroelectric (AF–FE) transition over a wide (narrow) range of temperatures yields large inverse electrocaloric (EC) effects that peak at ΔTj ∼ –2.6 K when applying 25 V μm−1 at 293 K (ΔTj denotes directly measured temperature jumps). Above T N, nominally reversible driving of the partial (full) paraelectric–ferroelectric (PE–FE) transition yields large conventional EC effects that peak at ΔTj ∼ +5.2 K when applying 25 V μm−1 at 302 K. This good EC performance near room temperature implies that MLCs of PMW could be exploited in prototype EC coolers.


Introduction
Caloric effects [1][2][3][4][5][6][7][8] are nominally reversible thermal changes due to changes in ferroic order driven by changes of applied field. For example, EC effects arise when a change of electric field E drives a change of local polarization P, and EC effects can be large in FE materials above the Curie temperature T C when the PE-FE transition is electrically driven. Similarly, magnetocaloric, elastocaloric, and barocaloric effects are driven by changes of magnetic field, uniaxial stress, and pressure, respectively. Caloric effects are currently under intense study for applications, and the key specific goal of environmentally friendly solid-state heating and cooling [3] would avoid the problematic hydrochlorofluorocarbons and hydrocarbons of vapour-compression technology [9].
EC effects in bulk ceramics are relatively small [1], but large EC effects can be driven in thin ceramic [10] and polymer [11] films because thin films have large breakdown fields. Consequently, large EC effects can be driven in macroscopic working bodies by combining many thin films to form multilayer capacitators (MLCs) [12], in which EC effects are now heavily studied [13]. In [14], most of us and co-workers have shown that MLCs of the well known EC material PbSc 0.5 Ta 0.5 O 3 (PST) display large EC effects over a wide range of temperatures including room temperature, e.g. highly adiabatic field-driven temperature jumps of up to |∆T j | ∼ 5.5 K. These EC effects were shown in [14] to compare favourably with the magnetocaloric effects that arise in the Gd working bodies of magnetocaloric prototype coolers [15], and MLCs of PST have now been deployed in EC prototype coolers [16,17].
Although EC effects are typically studied in FE materials, EC effects in AF materials are attracting increased attention [18][19][20][21]. Unlike FE materials, AF materials can show field-driven phase transitions both  [24], drawn using VESTA software [27]. (c) Crudely sketched temperature-field (T-E) phase diagram based on [23] (solid lines) and selected data from figure 4(c) (dashed lines). The high-field region is labeled FE (ferroelectric) after [23], but some parts are not fully transformed.
above and below the zero-field transition temperature T N , as we now explain while ignoring the fact that a phase is not technically FE if the polarization arises only when an electric field is applied. Below T N , electrically driving the AF-FE transition [22] results in inverse EC effects, such that adiabatic field application (removal) decreases (increases) the temperature. Above T N , electrically driving the PE-FE transition results in conventional EC effects, such that adiabatic field application (removal) increases (decreases) the temperature. For example, bulk PbZrO 3 driven with 4.2 V µm −1 shows inverse (conventional) EC effects of |∆T j | = 3.5 K (|∆T j | = 5.5 K) below (above) T N ∼ 493 K [20], and bulk PMW driven with 12 V µm −1 shows inverse (conventional) EC effects of |∆T j | = 2.0 K (|∆T j | = 1.8 K) just below (above) a T N that lies near room temperature [23]. Note that we do not follow the common practice of describing inverse EC effects as negative EC effects because absolute sign depends on whether field is applied or removed, and whether one considers a temperature change in the adiabatic limit or an entropy change in the isothermal limit.
PMW is structurally similar to PST because it is a perovskite with Pb on the A-sites, but unlike PST it is relatively easy to achieve good B-site order because the Mg 2+ and W 6+ valences differ by more than the Sc 3+ and Ta 5+ valences [23]. At zero field, the orthorhombic (Pmcn) AF phase exists below T N (figure 1(a)) and the cubic (Fm3m) PE phase exists above T N ( figure 1(b)) [24], if we ignore the small thermal hysteresis associated with the zero-field transition. The phase diagram in figure 1(c) compares the range of temperatures and fields explored here (dashed lines) with those explored in [23] (solid lines), where the maximum driving field (12 V µm −1 ) was only slightly larger than the minimum threshold field for starting the transition to the FE phase (∼10 V µm −1 ).
Here we report on MLCs of PMW, most of which can withstand 20 V µm −1 , some of which can withstand 27 V µm −1 . Adiabatic electrical polarization data are used to construct a P(T s ,E) phase diagram for PMW in which the onset (completion) of the FE phase is exceeded (achieved) at most (some) values of the set/starting temperature T s . Using a thermocouple to make direct measurements of temperature T # versus time t, we observe inverse (conventional) EC effects below (above) T N , as expected given the increase (decrease) of polarization on heating [3]. Between 238 K and a value of T N that lies near room temperature, the AF-FE transition yields inverse EC effects that reach |∆T j | ∼ 2.6 K at 293 K (fully driven transition). Between T N and 323 K, the partially driven PE-FE transition results in conventional EC effects that reach |∆T j | ∼ 5.2 K at 302 K.

MLC fabrication was as follows. Powders of high-purity Pb 3 O 4 and MgWO 4 (Kojundo Chemical Laboratory
Co., Ltd) were weighed, ball-milled for 16 h in distilled water (with balls of partially stabilized zirconia), dried, and then pulverized. Solid-state reaction was achieved by calcining the powder mixture at 1073 K for 4 h in air. Subsequent ball-milling with organic solvent and binder resulted in a slurry for tape-casting green sheets with the doctor blade technique. These green sheets were then electroded (by screen-printing Pt paste), stacked, pressed, and cut to obtain green MLCs. After burning off the binder at 773 K for 24 h, the proto-MLCs were sintered at 1223 K for 4 h in a Pb-containing atmosphere. Terminals were added by painting on a bespoke Ag paste, and then firing in air at 1023 K for 10 min.
Six similar MLCs were used in this study. MLC1-MLC4 were fabricated from Batch A of green sheets, and MLC5 and MLC6 were fabricated from the nominally equivalent Batch B. All MLCs had 19 active layers of thickness ∼40 µm (Batch A) or ∼36 µm (Batch B), Pt inner electrodes of thickness ∼2 µm, an active electrode area of ∼0.49 cm 2 per layer, and external dimensions of 10.5 × 7.4 × 0.8 mm 3 (simplified cross-sectional schematic in figure 2(a)). This geometry was employed previously for MLCs of PST [14], and it arose after optimizing the active and inactive volumes in other MLCs for thermocouple measurements of EC temperature change [25]. Green sheets from Batch A were also used to fabricate two 10 × 10 × 0.5 mm 3 plates of PMW for differential scanning calorimetry (DSC) and x-ray diffraction (XRD). These plates were made in the same way as the MLCs, but without Pt inner electrodes.
We show optical images of MLC1 ( figure 2(b)) and MLC2 in cross-section (figure 2(c)), and we show the microstructure of PMW in MLC3 (figure 2(d)) as measured with a scanning electron microscope (SEM) (ERA-8900, ELIONIX Inc.) whose accelerating voltage was set to 15 kV. We crushed Plate 1 and used the resulting powder to measure the heat transfer rate dQ/|dT| ( figure 3(a)) with a DSC (DSC7000, Hitachi High-Tech Corporation). Plate 2 was affixed to a variable-temperature XRD sample stage using Apiezon-N grease, and after stabilizing at each measurement temperature for 10 min ( figure 3(b)), it was studied using Cu K α radiation in a SmartLab diffractometer equipped with a HyPix-3000 detector and a TTK-600 temperature chamber (Rigaku Corporation).
We measured the dielectric permittivity of MLC4 (figure 4(a)) with an LCR meter (E4980A, Keysight Technologies). We measured the 1 Hz unipolar polarization of MLC5 (figures 4(b) and (c)) with a ferroelectric tester (Precision Premier II, Radiant Technologies, Inc.). Our P(E) measurements therefore lie near the adiabatic limit given a thermal relaxation time of approximately ∼18 s for 1/e decay ( figure 5(a)). We drove the EC temperature change in MLC4 and MLC6 (figure 5) with a high-voltage source meter (6517B, Keysight Technologies). For the dielectric and polarization measurements, MLC set temperature was controlled using a variable-temperature stage (Linkam Scientific Instruments Ltd.), and measured using a small Pt100 sensor. For the measurements of EC temperature change, MLC set temperature was controlled using an equivalent variable-temperature stage with a small Pt100 sensor, but we instead report the very similar values of starting temperature T s that we measured using a ∼50 µm-diameter K-type thermocouple. This thermocouple was attached to MLC face centres by Kapton tape, and used to measure EC temperature change. XRD profiles measured on heating to 223 K (blue), 298 K (green) and 323 K (red), and simulations for the high-temperature cubic PE phase (purple) and the low-temperature orthorhombic AF phase (black). Inset: detail of the 298 K data (green) and the corresponding data after cooling from 323 K to 300 K (pink). Subscripts c and o denote cubic and orthorhombic reflections, respectively. The simulations employ data obtained from [24] via the Inorganic Crystal Structure Database (ICSD). Data in (a) for crushed Plate 1. Data in (b) for Plate 2.

Results and discussion
An optical image showing part of a polished MLC in cross-section (figure 2(c)) shows the 19 active layers, the two inactive outer layers, and the 20 inner electrodes. An SEM image that zooms in on one layer within a broken MLC (figure 2(d)) reveals a reasonably uniform grain size of ∼2 µm. By contrast, the grain size reported in bulk PMW is less uniform, and the average value of 2.32 µm is slightly larger [23].
Heating (cooling) plots of the heat transfer rate dQ/|dT| (figure 3(a)) reveal endothermic (exothermic) peaks associated with the zero-field AF-PE phase transition near room temperature. The observation of latent heat, and a small thermal hysteresis of 3 K imply that the transition is first order. Estimates of T N ∼ 303 K and latent heat ∼3.0 kJ kg −1 from the heating plot differ slightly from reported values of 309 K and 3.9 kJ kg −1 [23], possibly due to differences in fabrication and thus microstructure. Given that latent heat is measured at zero field, it cannot be compared with the electrically driven AF-FE and PE-FE transitions that we report later. The thermal hysteresis implies that T N cannot be identified as a single value, and in this paper it is sufficient to assume that T N takes a value near room temperature.
The zero-field AF-PE phase transition can also be identified from XRD profiles ( figure 3(b)), where peak splitting arises from the use of Cu K α1 and K α2 radiation. At 223 K (blue) and subsequently at 298 K (green), all reflections can be assigned to the low-temperature AF orthorhombic Pmcn structure (black) [24]. On further heating to 323 K (red), all reflections can be assigned to the high-temperature PE cubic Fm3m structure (purple) [24]. The inset of figure 3(b) shows that after heating to 298 K (green), the 024 o and 400 o reflections of the low-temperature AF orthorhombic phase are present, and the 400 c reflection of the high-temperature PE cubic phase is essentially absent. After subsequently cooling from 323 K to 300 K (pink), the 400 c reflection of the PE phase persists, and the 024 o and 400 o reflections of the AF phase reappear. The XRD data thus reveal phase coexistence in the vicinity of the Néel temperature, as expected given the thermal hysteresis ( figure 3(a)).
Electrical measurements are presented in figure 4. The zero-field AF-PE phase transition is observed in the real part of the relative dielectric permittivity measured at 1 kHz (ε, figure 4(a)). Values of ε are similar to values obtained for bulk PMW [23], and consistent with the low (high) low-field gradient of P(E) below (above) T N ( figure 4(b)). Heating and cooling sweeps show that the thermal hysteresis in ε of ∼1 K (inset, figure 4(a)) is of similar magnitude to the DSC value of ∼3 K ( figure 3(a)); the small discrepancy may arise due to different temperature sweep rates (5 K min −1 for ε, 10 K min −1 for dQ/|dT|). A loss tangent as low as tan δ ∼ 10 −3 at a temperature as high as 450 K evidences high electrical resistivity, implying negligible Joule heating during EC cycles ( figure 5(a)).
The P(E) measurements shown in figure 4 were highly adiabatic and unipolar. The data were collected out to 27 V µm −1 in 238 K ⩽ T ⩽ 328 K at 136 values of starting temperature T s measured every ∼0.7 K on cooling. Data for six starting temperatures appear in figure 4(b). We see that the AF-PE transition (PE-FE transition) can be driven to a large (small) extent below (above) T N , and that any substantial degree of completion results in a small field hysteresis. Specifically for temperatures that lie near and below T N , we see that P(E) is large and single-valued at high field, implying that the transition is driven in full (black and orange data, figure 4(b)).
The field-application branches of our unipolar P(E) measurements at all 136 starting temperatures yield a P(T s ,E) phase diagram ( figure 4(c)). The vertical trajectories are isentropic, and so this phase diagram is best understood if one conceptualizes conversion to a P(S,E) map, where entropy S increases monotonically with increasing T s via S(T s ) =´T s T0 [c(T ′ s )/T ′ s ]dT ′ s (T 0 → 0 if S denotes absolute entropy, c denotes specific heat capacity) [26]. The low-field AF-PE border remains near T N at all fields as expected ( figure 1(c)), and the temperature-dependent threshold field (light blue in figure 4(c)) takes a minimum value of 12 V µm −1 near T N . Inverse (conventional) EC effects below (above) T N are observed via direct measurements of EC temperature change. Figure 5(a) shows the time-dependent temperature T # that is measured when switching on and off a field of 25 V µm −1 at two representative starting temperatures. The high-field return to the starting temperature evidences negligible Joule heating, as required for applications. Figure 5(b) shows the temperature jump ∆T j that is observed when switching on and off fields of different magnitude in a range of starting temperatures. Using 25 V µm −1 , the largest inverse EC effect is |∆T j | = 2.6 K at T s = 293 K, and the largest conventional EC effect is |∆T j | = 5.2 K at T s = 302 K.

Summary
By repeatably applying fields as large as 27 V µm −1 to MLCs of PMW without breakdown, we have expanded the bulk PMW phase diagram along the axes of both temperature and field ( figure 4(c)). Below the Néel temperature T N that lies near room temperature, electrically driving the AF-FE transition yields inverse EC effects of up to |∆T j | = 2.6 K, exceeding the value of |∆T j | = 2.0 K for bulk PMW [23]. Above the Néel temperature, electrically driving the PE-FE transition yields conventional EC effects of up to |∆T j | = 5.2 K, exceeding the value of |∆T j | = 1.8 K for bulk PMW [23]. These conventional EC effects are similar to the large EC effects of |∆T j | = 5.5 K for MLCs of PST [14] and bulk PbZrO 3 [20], and we suggest that MLCs of PMW could be exploited in prototype EC coolers. By further optimizing both MLC geometry and the fabrication process, even larger breakdown fields could be achieved. This would yield larger EC effects by permitting the AF-FE and PE-FE transitions to be driven in full over a wider range of temperatures, and by permitting the FE phase to be more highly polarized [14].

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.