Electrocaloric effect in BaTiO3 multilayer capacitors with first-order phase transitions

Electrocaloric (EC) materials, presenting large adiabatic temperature change or isothermal entropy change under the application (or removal) of electric fields, offer an efficient alternative to caloric heat pumps for replacing hazardous gases used in traditional vapor-compression systems. Recently, a large EC temperature change of 5.5 K have been reported in Pb(Sc0.5Ta0.5)O3 multilayer ceramic capacitors (Nair et al 2019 Nature 575 468) thanks to its strong first-order phase transition and a temperature span of 13 K has been reported in a prototype based on these capacitors (Torelló et al 2020 Science 370 125). However, the toxicity of lead forces researchers to find eco-friendly materials exhibiting competitive EC performances. Here, we study the EC effect in lead-free BaTiO3 multilayer capacitors using an infrared camera. Unlike commercial BaTiO3 capacitors, we prepared our samples without sacrifying the first-order phase transition in BaTiO3 while a low amount of 0.2 mol% Mn was added as an acceptor dopant to improve electrical resistivity. Their EC adiabatic temperature variations show two peaks versus temperature, which match BaTiO3 two first-order phase transitions, as observed by differential scanning calorimetry. We measured a temperature drop of ∼0.9 K over a temperature range of 70 K under 170 kV cm−1, starting at 30 °C near the tetragonal-to-orthorhombic phase transition. Under the same electric field, a maximum temperature change of 2.4 K was recorded at 126 °C, at BaTiO3’s Curie temperature. Our findings suggest that further optimized BaTiO3 capacitors could offer a path for designing lead-free caloric cooling prototypes.


Introduction
Current vapor-cycle cooling technology causes around 12% of global greenhouse emissions in light of its electricity consumption and hydrofluorocarbon-based fluid emission [1]. A new refrigeration solution using the electrocaloric (EC) effect, therefore, has been considered as a green and efficient method to replace the above technologies [2,3]. The EC effect is not new to the scientific community as it was already predicted and defined by Lord Kelvin in 1878 [4]. It refers to reversible electric field-driven isothermal entropy change, ∆S or adiabatic temperature change, ∆T in dielectrics. The first recorded EC effect in Rochelle Salt (KNaC 4 H 4 O 6 · 4H 2 O) was reported by Kobeko and Kurtschatov in 1930 [5]. The EC effect was then found in Triglycine sulfate [6], Pb(Zr,Ti)O 3 [7], KTaO 3 [8], Pb(Zr,Sn,Ti)O 3 [9], and Pb(Sc 0.5 Ta 0.5 )O 3 (PST) [10], but the magnitude of those effects was small and only few of them were considered for practical cooling applications.
The turning point occurred when the giant EC effect with an adiabatic temperature change of 12 K was discovered and physically explained in PbZr 0.95 Ti 0.05 O 3 thin films in 2006 [11]. A similarly giant EC temperature change was subsequently observed in polyvinylidene fluoride (PVDF)-based organic ferroelectrics (FEs) [12]. After these two outstanding findings, giant or even colossal effects have been intensively reported in inorganic and organic materials, and organic-inorganic composites in the form of thin or thick films [13][14][15][16]. In practice, these film-form materials are more suitable for on-chip cooling rather than large-scale cooling due to their low effective active volume [2]. Finding large or giant EC effects in bulk materials could lead to the realization of EC refrigeration in large devices, in which ceramics and multilayer ceramic capacitors (MLCCs) are widely studied.
The low breakdown field of FE ceramics is one of the main obstacles to achieving large EC response. Multilayer capacitors, instead, appear to be an optimized solution capable of combining high breakdown field and large active volume [17,18]. A milestone has recently been achieved in PST MLCC, that is, an adiabatic temperature change of 5.5 K when applying an electric field of 290 kV cm −1 [19], resulting in the rapid development of EC cooling prototypes with PST MLCCs. Wang et al designed a solid-state-based EC cooler with a temperature difference of 5 K between the hot side and the cold side and obtained a very large heat flux (135 mW cm −2 ) [20]. Besides, Torelló et al showed an unprecedented 13 K-temperature difference in a prototype using the same PST MLCCs and a dielectric fluid for heat transfer [21].
Similarly to what was observed in PST MLCCs, enhanced EC responses are obtained in the vicinity of first-order phase transitions [19,22]. Besides, there is a need to find eco-friendly alternative materials to PST. In this study, the strategy is to make lead-free BaTiO 3 (BTO) MLCC with strong first-order phase transitions, which are intentionally avoided in commercial BTO MLCCs. A phase transition is accompanied by an abrupt change of properties, inferring non-linearity on the capacitance value. Commercial BTO MLCCs are made to be stable with voltage and temperature in a large temperature range. There are standards fixing the maximum acceptable capacitance variations, e.g. ∆C/C = +22/−82% between −30 • C and 85 • C in Y5V standard. Hence, the only way to respect it is to considerably reduce the phase transition of BTO in commercial MLCCs. Here we studied the EC effect in lead-free BTO MLCC using an infrared (IR) camera. We measured a temperature drop of at least 0.9 K under 170 kV cm −1 on a temperature range from 30 • C to 100 • C. Under the same electric field, the largest temperature change (2.4 K) was recorded at 126 • C. The adiabatic temperature changes versus temperature show two peaks matching the two first-order phase transitions of BTO in this temperature range, as observed by differential scanning calorimetry (DSC) and temperature-dependent dielectric permittivity. Our results are typically twice as large as the EC temperature changes obtained in commercial BTO MLCCs [17,23].

MLCC fabrication
MLCCs of BTO were prepared by solid-state reaction and conventional tape casting methods. To obtain BaTi 0.998 Mn 0.002 O 3 powder, high purity BaCO 3 , TiO 2 , and Mn 3 O 4 were weighed, and ball milled in distilled water with partially stabilized zirconia balls for 16 h. Mn was added as an acceptor dopant to improve electrical resistivity under high electric field [24,25]. The slurry was dried and calcined at 1150 • C for 2 h. Calcinated powder was then ball milled in an organic solvent with binder for 16 h, and green sheets were fabricated by the doctor blade method. After screen printing Pt paste for the inner electrode, green chips of BTO MLCC were obtained by stacking, pressing and cutting the green sheets. The binder was burned at 500 • C for 24 h, and the MLCCs were sintered at 1350 • C for 4 h in air atmosphere. Ag paste was attached and fired at 750 • C to form external electrodes. The size of the fabricated MLCC is 5.5 × 6.7 × 0.5 mm 3 while each of the inner Pt electrodes has an area of 3.9 × 3.6 mm 2 .

Sample characterization
The structure and phase purity of these BTO MLCCs were characterized by x-ray diffraction (XRD, Bruker D8) using Cu-Kα radiation in the 2θ range from 15 to 70 • with steps of 0.02 • at room temperature. A BTO MLCC was crushed into powder using a mortar and pestle before the XRD measurement. The microstructure of the cross-section of BTO MLCC was observed by scanning electron microscope (SEM, Hitatchi SU-70). Heat flow dQ/|dT| (Q is heat) was collected with a heating/cooling rate of 5 K min −1 in a differential scanning calorimeter (DSC3 Mettler Toledo) with a sample mass of 10.4 mg. The temperature dependence of the dielectric constant was collected with an impedance spectrometer (Concept 40, Novocontrol) at 1 kHz with a scan rate of 0.5 K min −1 . The polarization-electric field (P-E) hysteresis loops under various temperatures were recorded at 10 Hz using a TF Analyzer 2000 (aixACCT) with a temperature controller. EC adiabatic temperature variations were measured with an IR camera (FLIR X6580sc) while the MLCCs were charged with a power supply (Keithley 2400) and their temperature controlled with a Linkam stage.   The red frame indicates the zoomed part of one dielectric and two electrode layers. A few pores are visible in the dielectric, exhibiting an average size smaller than 2 µm. The entire structure looks though uniform and dense. Figure 2 shows the heat flux dQ/|dT|, the dielectric and FE properties of one BTO MLCC over a wide temperature range (−20 • C to 180 • C). Two dQ/|dT| peaks evidence two first-order phase transitions (see figure 2(a)). A first peak at 21 • C on heating and 11 • C on cooling corresponds to the FE orthorhombic to FE tetragonal phase transition. A second peak, higher in amplitude, at 126 • C on heating with a hysteresis of 4 K on cooling is the result of the transition from the tetragonal phase to the paraelectric (PE) cubic phase. These values are slightly shifted (a few degrees upwards) compared with results in the literature on pure BaTiO 3 ceramics [26], which could be due to the 0.2% Mn doping of our MLCCs. The corresponding latent heats for FE-to-FE and FE-to-PE phase transitions are 0.3 kJ kg −1 and 1 kJ kg −1, respectively. These values match previously reported data in BTO single crystals [27,28] and ceramics [29]. Interestingly, the latent heat of the FE-to-PE phase transition is almost equal to the latent heat reported in PST ceramics (1 kJ kg −1 ) or MLCCs (1.2 kJ kg −1 ), which display a large EC effect of 5.5 K [19,30].

Results and discussion
The phase transition temperatures deduced from the temperature-dependent dielectric constant curves in figure 2(b) are in line with the values from DSC. The maximum of the dielectric constant ε max , which is related to the PE to FE phase transition, occurs at 121 • C showing a value of 12 000. The heat flow and dielectric constant plots show that the FE-to-FE transition displays a larger thermal hysteresis and a smaller latent heat than the FE-to-PE phase transition.
The feature of the different FE phases can also be investigated through P-E loops (see figures 2(c) and (d)). From orthorhombic to tetragonal phases, the remanent polarization (P r ) decreases dramatically more than the maximum polarization (P max ). According to the so-called indirect method, this is in favor of a larger EC effect. Indeed, the simplest expression of the indirect method is deduced from the Landau theory of phase transitions [31], where β is the Curie constant and c is the specific heat capacity. Hence, a large P max and a low P r favor large ∆T, which happens specifically at the FE-to-PE phase transition where P r is almost zero. We then characterized the EC effect directly through an IR camera [22]. In figure 3(a), we show the IR images of one BTO MLCC and the thermal change under electric field, in which the frame with solid and dashed lines correspond to the whole and active areas, respectively.     peak maximum (see the inset in figure 3(b)), indicating a fast charging and discharging speed. Note that the camera frequency is 1.35 kHz, which enables reaching 0.75 ms between two successive points, ensuring adiabatic conditions. Any contribution related to Joule heating can be neglected because the temperature signal returns to the baseline when the field is still on and after having waited for the EC heat to be released in the surroundings. Besides, the exothermal and endothermal peaks are of similar magnitude under the same electric field after five cycles, which indicates a reversible and reproducible EC effect. Figure 4(a) shows the EC temperature change versus temperature at 170 kV cm −1 . As anticipated from the previous results, there are two temperature peaks. The sharpest and highest corresponds to the FE-to-PE phase transition, around 130 • C. The maximum EC temperature change (2.4 K) is obtained at 126 • C. A smaller peak attributed to the FE-to-FE transition appears at 35 • C. The latter contributes to a large temperature range where this BTO MLCC exhibits a ∆T of at least 0.9 • C at 170 kV cm −1 from 35 • C to 150 • C. Here, we define the temperature range (T ran. ) as a temperature interval of the EC effect with no less than 90% of its maximum value. Figure 4(b) shows the EC temperature change as a function of electric field at 35 • C from 85 to 240 kV cm −1 . A temperature change of up to 1.2 K at 35 • C can be achieved at 240 kV cm −1 .
We finally compare ∆T and T ran . in BTO single crystals, ceramics, thin films and MLCCs (commercial ones and ours) in table 1. Direct measurements in single crystal reveal a ∆T of 1 K in FE-to-FE and FE-to-PE transitions. However, T ran . in single crystals is between 2 and 4 K, which might limit its application in devices. Similarly, in ceramics, a ∆T of 1.5 K is reported with a T ran . less than 1 K. We see that strong first-order phase transition in these materials secures large ∆T, but, at the same time, low breakdown field prevents large T ran . In thin films and MLCCs, a high electric field guarantees both high ∆T and large T ran . Commercial BTO-based MLCCs, however, achieve large breakdown fields by sacrificing phase transitions that are crucial for the EC effect. Here, we demonstrate that a large EC effect with a large T ran . can be realized by keeping the first-order phase transition in our MLCCs.
These results are encouraging, though realistic applications would require the best results to happen closer to room temperature. In our case, it means that one has to find a way to bring the FE-to-PE transition close to room temperature. The best way to do so is probably to add Sr to BTO in order to obtain (Ba,Sr)TiO 3 . Indeed, Bai et al examined this idea by preparing a series of Ba 1 − x Sr x TiO 3 (x = 0.2-0.4) FE ceramics, in which phase transition temperatures decrease proportionally to Sr amount. Meanwhile, heat-flow and temperature dependence of permittivity curves indicate strong phase transitions in those ceramics [33].
An interesting way to enhance the EC effect further is to dope BTO. For instance, Qian et al observed a ∆T of 4.5 K in Ba(Zr 0.2 Ti 0.8 )O 3 ceramics near the so-called invariant critical point in the phase diagram [34]. Learning from those methods, we can probably improve our current result through doping while keeping first-order phase transitions.

Conclusion
In this study, large EC effects have been reported in lead-free BaTiO 3 MLCCs using an IR camera. We observed large EC effects around its first-order phase transitions. We found a maximum EC temperature change of 2.4 K at the FE-to-PE phase transition. In contrast, EC response in the FE-to-FE phase transition presents a low temperature change of 0.9 K, but with a large temperature range of more than 70 K. BTO first-order transitions contribute to improve the EC effect in these MLCCs. We suggest that doping BaTiO 3 should further enhance the EC performances of these MLCCs.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).