Oxide‐Halide Perovskite Composites for Simultaneous Recycling of Lead Zirconate Titanate Piezoceramics and Methylammonium Lead Iodide Solar Cells

Global concerns over energy availability and the environment impose an urgent requirement for sustainable manufacturing, usage, and disposal of electronic components. Piezoelectric and photovoltaic components are being extensively used. They contain the hazardous element, Pb (e.g., in widely used and researched Pb(Zr,Ti)O3 and halide perovskites), but they are not being properly recycled or reused. This work demonstrates the fabrication of upside‐down composite sensor materials using crushed ceramic particles recycled from broken piezoceramics, polycrystalline halide perovskite powder collected from waste dye‐sensitized solar cells, and crystal particles of a Cd‐based perovskite composition, C6H5N(CH3)3CdBr3xCl3(1–x). The piezoceramic and halide perovskite particles are used as filler and binder, respectively, to show a proof of concept for the chemical and microstructural compatibility between the oxide and halide perovskite compounds while being recycled simultaneously. Production of the recycled and reusable materials requires only a marginal energy budget while achieving a very high material densification of >92%, as well as a 40% higher piezoelectric voltage coefficient, i.e., better sensing capability, than the pristine piezoceramics. This work thus offers an energy‐ and environmentally friendly approach to the recycling of hazardous elements as well as giving a second life to waste piezoelectric and photovoltaic components.


Introduction
Electroceramics that exhibit good dielectric, piezoelectric and ferroelectric properties are of great importance in the modern world DOI: 10.1002/smtd.202300830being employed as the crucial functional parts in energy storage and conversion components, such as capacitors and harvesters, as well as in data storage devices, e.g.[3] Electroceramics enjoy chemical stability and thus a long lifespan in service, however, they are also recycled less properly along with other electronic wastes compared to materials like metals and polymers.For instance, there are globally 620 000 tons of electroceramic wastes generated annually. [4]In addition, the production of electroceramics results in inevitable rejects (i.e., due to unwanted deformation, cracking, or inhomogeneous densification during the high-temperature sintering stage).The rejected materials, which do not meet specified standards or requirements of quality and are therefore deemed unacceptable for sale or use, end up being discarded without proper use, contributing to the increase of electroceramic wastes.From this aspect, there is a high demand for a proper waste handling approach to solve the issues of recycling and reuse of electroceramics. [4,5]mong all electroceramics, piezoelectric ceramics (piezoceramics) are highly versatile because of their exceptional electromechanical coupling capabilities which have found extensive applications in sensors, actuators, and transducers in structural health monitoring systems, microphones, buzzers, resonators, and filters. [3,6,7]The functional versatility and excellent electromechanical coupling effect of piezoceramics largely rely on the use of the element Pb.For instance, Pb(Zr,Ti)O 3 -based ceramics have dominated the market of piezoelectric components since their discovery in the 1950s due to their remarkable piezoelectric responses, as well as their chemical and functional stability over a wide range of temperatures.0] The environmental concerns over the use of Pb-containing piezoelectrics are not only raised by the toxic and hazardous nature of Pb itself, but also by the large energy and ecological footprint during mining, manufacturing, and disposal where Pb is usually leeched out into the surrounding environment and potentially demands immense amounts of energy for post-treatment and cleaning. [11]14] In terms of the recycling method, a recently developed concept of making upside-down ceramic-ceramic composites via the ultralow-temperature densification mechanism is a promising candidate.Such composite materials adopt exceptionally high filler content with multimodal particle sizes (the majority being >100 μm) and much lower binder (inversed matrix) content compared to that in a conventional composite, namely, "'upsidedown"'. [15]A PZT-LMO composite, where PZT refers to PZ29 (a commercial soft-type Pb(Zr,Ti)O 3 ) and LMO stands for Li 2 MoO 4 hereafter, has been introduced as the first upside-down piezoelectric composite.The sintered and crushed PZT ceramic particles as the filler were coated with LMO aqueous solution as the binder.The LMO-coated PZT particles were then pressed under 250 MPa at 120 °C, a temperature only a tenth of that needed for sintering a new PZT ceramic (≈1200 °C).Meanwhile, the high volume fraction of PZT particles, which was ≈75%, enabled a significant enhancement in the material's functional properties compared to other composites and ultralow-temperature fabricated counterparts. [16]The longitudinal piezoelectric charge coefficient (d 33 ) and voltage coefficient (g 33 ) were reported to be 84 pC N −1 and 33 mVm N −1 , respectively, for the PZT-LMO composite.
The piezoelectric properties of the composite have also been further improved by replacing LMO with TiO x as the binder.With a filler to binder volume ratio of 84:16 and after hot-pressing the material under 250 MPa at 275-350 °C, the TiO x binder transformed to amorphous TiO 2 (am-TiO 2 ) and settled in the intergranular regions.The PZT-am-TiO 2 composite exhibited larger d 33 and g 33 values of 150 pC N −1 and 53 mVm N −1 , respectively, attributed to the reduced mismatch of permittivity between the filler and binder materials and thus the enhancement of electric field coupling during the poling procedure. [17]he large filler particle size necessary for the fabrication of upside-down composites naturally benefits the recycling purpose.This is because the waste piezoceramics are typically either mechanically or electrically broken and therefore need reshaping to recover the mechanical or electrical properties.Crushing such piezoceramics provides ideally large particle sizes for the filler.However, neither LMO nor TiO x mentioned above is a sustainable option as the binder in the potential recycling process because new materials would need to be produced which would cause unnecessarily excessive consumption of energy and resources.For instance, the fabrication of LMO requires the precious resource of Li.TiO x requires an elevated temperature, as mentioned above, to settle in the composites.
Compared to LMO and TiO x , organometal halide perovskites may be ideal binders for the recycling of piezoceramics for several reasons.First, most organometal halide perovskites can be synthesized and crystallized at close to room temperature, offering energy-efficient processing of the recycling. [18,19]Second, their solubility in organic solvents enables effective coating of the filler particles via more controlled precipitation, promoting homogeneous densification of the upside-down composites.More importantly, the recently developed high-efficiency perovskite solar cells mostly use organometal halide perovskites as the absorbers. [20]These future solar cells, once retired, would provide abundant binder sources.This would become an added value for the recycling process since both waste piezoceramics and solar cells could be recycled simultaneously.In addition, as most piezoceramics are oxide perovskites, using halide perovskites as the binder may offer benefits on the compatibility of structural and functional development.For instance, the halide perovskites are structurally flexible due to their organic A-site, which may facilitate easier and more dense binding of the oxide perovskite particles than the LMO and TiO x can do, with the A-site organic groups possibly occupying A-site vacancies (if any) in the oxide perovskite unit cells (e.g., when having the same A-site valance such as for KNbO 3 -based compounds). [21,22]Some halide perovskites also show noticeable dielectric, ferroelectric, and piezoelectric properties, supporting the functional improvement of the composites in collaboration with the filler particles. [23]n order to demonstrate the recyclability of piezoceramics and perovskite solar cells, this work selects the organometal halide perovskite composition (i.e., grown crystal particles), (PTMA)CdBr 3x Cl 3(1-x) where PTMA is C 6 H 5 N(CH 3 ) 3 + , to be the binder.This composition has been first reported as a ferroelastic crystal that could be grown at room temperature and has exhibited a superior converse piezoelectric effect. [19]In this work, crushed PZT ceramic particles as the fillers are mixed with different types of crushed (PTMA)CdBr 3x Cl 3(1-x) crystal binders with x values of 0-0.9 to form upside-down composites.Moreover, MAPbI 3 (MA is CH 3 NH 3 + ), which is a cornerstone composition used in perovskite solar cells, [20] is chosen as the additive to the PZT-(PTMA)CdBr 3x Cl 3(1-x) composites.It is worth noting that, in this work, the MAPbI 3 polycrystalline powder is truly recycled from perovskite solar cells made in the laboratory.The objective of this study is to introduce a new piezoelectric composite with both recycled fillers and recycled additives and thus to pave the way toward the reduction of ecological footprints in the piezoelectric and photovoltaic industry via sustainable recycling and reuse of electronic components.

Microstructure and Dielectric Properties
Table 1 lists the compositions (including the developed particles and their roles in the composites) and relative densities of all the samples studied in this work.The samples are assigned codes to be referred to hereafter.Full datasets including values of the theoretical and measured densities can be found in Table S1 (Supporting Information).Briefly, there were four categories of samples: filler material (PZT ceramic), binder materials (A-C) which were (PTMA)CdBr 3x Cl 3(1-x) crystals with different Br/Cl ratios, additive material (D) which was MAPbI 3 powder recycled from solar cells, and composites.The composites were formed using the filler coated with different types and amounts (5-30 wt.%) of binders (PZT-A5, PZT-B5, PZT-B10, PZT-B20, PZT-B30, PZT-C5) as well as with 0.1-0.25 wt.% of additive (PZT-C5-D01, PZT-C5-D025).
Figure 1a shows the STEM (scanning transmission electron microscopy) micrograph and the EDX (energy dispersive X-ray spectroscopy) elemental mapping results for the crushed PZT particles coated by the binder phase C. Here, the organometal halide composition is shown to effectively coat the PZT particles despite some minorly uneven distribution of the organic (C, N) and inorganic (Cd, Cl) elements.This was expected as the binder phase was introduced in a deionized water-acetonitrile solution in which the organic and inorganic elements existed as individual ions.These ionic precipitations would redissolve in the transport liquid which could stimulate the rearrangement of the recrystallized particles during hot pressing at a later stage, and tended to result in a higher composite density compared to the case without such a coating. [15]igure 1b shows the FESEM (field emission scanning electron microscopy) micrograph and the EDX elemental mapping results for the PZT-C5 composite sample.FESEM micrographs and EDX elemental maps for other composite samples with A and B binder phases and with D as the additive are given in Figures S1-S7 (Supporting Information).From these figures, it is obvious that the PZT filler particles were densely packed with the binder phases and pores were hardly visible.This implies an ideal, very high level of densification.Indeed, the relative densities of the composites reached 95% of the ceramic density, i.e., nearly 93% for PZT-A5, PZT-B5 and PZT-C5 compared to ≈98.8% for high-temperature sintered PZ29 ceramics supplied by the commercial manufacturer (Table 1 and Table S1, Supporting Information). [24]This is to say, all samples with 5 wt.% (PTMA)CdBr 3x Cl 3(1-x) where x equals to 0, 0.5, and 0.9, respectively, were sufficiently densified as a composite.
The concentration of the binder phase was further increased beyond 5 wt.% using the B phase to investigate the influence of the volume fraction of binder on densification as well as on functional properties of the upside-down composites.As the density of the binder phase was much lower than that of the PZT filler (Table S1, Supporting Information), both the theoretical and measured densities of the composites decreased with the increase of the binder volume fraction for samples PZT-B5, PZT-B10, PZT-B20, and PZT-B30 (Table S1, Supporting Information), as expected.However, the relative densities of the composites went beyond 100% when the binder concentration was larger than 5 wt.% (Table 1).In principle, relative density cannot exceed 100%.The >100% relative density shown in Table 1 was because the measured densities of the samples were higher than the corresponding theoretical densities (Table S1, Supporting Information).
To discover the underlying reason, Figure S8 (Supporting Information) provides the results of DSC/TGA (differential scanning calorimetry/thermogravimetric analysis) carried out with the pure binder B and composite sample PZT-B20.An endothermic peak appeared at ≈200 °C with a correspondingly drastic decrease of the weight in the range of 200-300 °C for both the crystal and composite samples.This could correlate to the loss of certain organic compounds.During the fabrication of the composites, a mildly elevated temperature of 150 °C and an increased pressure of 250 MPa (compared to the air pressure of only 101 kPa during the DSC/TGA) were applied to the samples.It is reasonable to assume that the high pressure made up the missing endothermic force of the ≈50 °C difference between the DSC/TGA measurement temperature and the fabrication temperature, and then promoted the organic compound to leave the samples through evaporation while being dissolved in the solvent during the hot-pressing procedure.
Despite the loss of organic compounds, all the composite samples showed a dominant distribution of the filler particles with minor binders and additives filling the intergranular areas and with seamless adherence at the interfaces (Figure 1; Figures S1-S7, Supporting Information).This indicates that the oxide-halide perovskite composites had an excellent wettability between the filler and binder phases that was beneficial for achieving low porosity and thus high relative density.In comparison, pores and lower relative densities have been observed in the above-mentioned PZT-LMO and PZT-am-TiO 2 composites reported in previous works. [16,17]Such a difference implies that the organometal halide perovskite binder might be advantageous over other types of binders for oxide perovskite fillers due to an easier migration of the organic A-site groups and hence a better ability to redistribute on surfaces and among interfaces of oxide perovskite particles.
Figures S9 and S10 (Supporting Information) show the XRD (X-ray diffraction) patterns of the filler and composites, which confirmed the coexistence of the PZT filler and (PTMA)CdBr 3x Cl 3(1-x) binder phases (Figure S9, Supporting Information), despite the fact that the 5 wt.% binder phase was below the detection limitation of the XRD (Figure S10, Supporting Information).
Figure 2 shows the real part of relative permittivity (ɛ') and dielectric loss (tan ) of unpoled composite samples measured at frequencies of 1, 10, and 100 kHz.The full spectra of the dielectric measurement for both unpoled and poled samples are provided in Figures S11-S13 (Supporting Information).For comparison, the crushed (PTMA)CdBr 3x Cl 3(1-x) crystals were also made into polycrystalline pellets using the same method for making composites but without the use of filler.Their dielectric properties are listed in Table S1 (Supporting Information).With 5 wt.% of the binder, all the composites exhibited similar ' values in the range of 200-250 at 1 kHz (PZT-A5, PZT-B5, PZT-C5 and PZT-C5-D01), except for the outlier of PZT-C5-D025 that showed a larger ' value of >300 (Figure 2a).These values decreased slightly with the increase of the measurement frequency (Figure 2b,c; Figures S11a,S12a, and S13a, Supporting Information), but it was conclusive that the permittivity of the composites was independent of the Br/Cl ratio in the binder phase.Compared to the high-temperature sintered PZ29 ceramics, [24] the permittivity of the composites was only ≈10% of that of the PZ29 ceramics.This was expected due to the composite effect of the significantly different permittivity values between that of the filler phase (≈2800 according to the supplier) and that of the binder phase (as low as 5-6, Table S1, Supporting Information).
Such an effect also caused a decreasing trend in the ' values with the increase of the volume fraction of the binder phase (PZT-B5, PZT-B10, PZT-B20 and PZT-B30), i.e., from ≈200 for PZT-B5 to only ≈50 for PZT-B30.In the composite effect, the measured permittivity of the composite sample is determined by the volume fraction of each phase. [25]The fraction volumes of the low-permittivity binder in samples PZT-A5, PZT-B5, PZT-C5 and PZT-C5-D01 were comparable (Table 1), resulting in the similar permittivity values for the corresponding composites as discussed above.However, the much lower density of the binder compared to that of the filler (Table S1, Supporting Information) induced a rapid increase in the volume fraction of the binder phase even though the weight percentage was only moderately increased (from 5 wt.% to 30 wt.%).Consequently, the resultant decrease in permittivity also appeared exaggerated (Figure 2).

Influence of the MAPbI 3 Additive
The special MAPbI 3 additive, which is a high-permittivity as well as a high-conductivity semiconductor compound, [22] resulted in some intriguing phenomena.For instance, in Figure 2, it can be noticed that the influence of the 0.1 wt.% additive (MAPbI 3 ) on permittivity was negligible by comparing the values for PZT-C5 and PZT-C5-D01.However, even a very minor increase of the additive (from 0.1 wt.% to 0.25 wt.%) induced a clear increase in ' values, despite the fact that both the concentrations were below the detection limit of EDX (i.e., not visible in Figures S6 and S7, Supporting Information).This was also believed to be attributed to the composite effect, as the ' values of the MAPbI 3 crystals could reach the magnitude of 10 3 , [22] comparable to that of the filler.Such a large permittivity was able to influence the composite even though the volume fraction was only as small as 0.4% (Table 1).It could then be estimated that the threshold of volume fraction for the additive to be able to influence the permittivity of the composite was between 0.2%-0.4% (Table 1, Figure 2).This threshold also determined whether the composite could be effectively poled to show a piezoelectric response, meaning that sample PZT-C5-D025 was too conductive due to the above-threshold volume fraction of the conductive additive, and, during poling, the sample could not sustain the high poling voltage and thus the sample did not show a macroscopic piezoelectric response.On the contrary, the below-threshold PZT-C5-D01 could be successfully poled, as will be discussed below.
All the composite samples without the additive were much more lossy (tan  values of 0.05-0.1 at 1 kHz, Figure 2a) than the PZ29 ceramics (tan  ≈ 0.02). [24]Although the dielectric loss was significantly reduced to 0.02-0.05at higher frequencies (Figure 2b,c; Figure S11b and S12b, Supporting Information), the composite samples with additives always showed much smaller tan  values compared to those of samples without additives at lower frequencies (Figure 2a; Figure S13b, Supporting Information).Impressively, the dielectric loss of samples PZT-C5-D01 and PZT-C5-D025 was similar to that of the PZ29 ceramic at all frequencies (Figure 2; Figure S13b, Supporting Information).
Dielectric loss can be affected by many factors, and it is also highly dependent on microstructure. [26]Although the tan  values for both the filler phase (≈0.02) and the binder phase (<0.01) were reasonably small, there was a considerable mismatch of the dielectric behaviors between the two phases.As the filler phase was ferroelectric, its dielectric behavior was contributed by not only electric field-induced dipoles but also permanent dipoles, i.e., spontaneous polarizations existing in domains.While the binder phases tended to be less ferroelectric but more ferroelastic, [19] their dielectric behavior was mainly contributed by electric-field induced dipoles.During the dielectric measurement, the majority of the applied electric field locally concentrated on the low-permittivity binder phases, [25,27] but the domains in the filler phase were still able to interfere to some extent.As the spontaneous polarizations in the domains of the polycrystalline filler particles did not necessarily align in the same direction of the electric field-induced dipoles in the intergranular binder, an increased level of phase lag during the measurement occurred and this effect was more obvious at lower frequencies of the applied electric field.Such a conflict between the dielectric behaviors of the filler and binder phases faded with the increase of the measurement frequency since the domain contribution was frequency dependent and the polarization-induced dielectric response was diminished at high frequencies. [2]ompared to the PZT-LMO and PZT-am-TiO 2 composites made with the same method, the PZT-5 wt.% (PTMA)CdBr 3x Cl 3(1-x) composites in this work generally showed higher and similar permittivity, respectively.The dielectric loss of the PZT-5 wt.% (PTMA)CdBr 3x Cl 3(1-x) composites, on the other hand, was generally higher. [16,17]ome "'lubricating"' effect on the low-frequency dielectric conflict between the filler and binder phases was introduced when the composite samples contained the additive.The additive, MAPbI 3 , was special due to its simultaneous high permittivity and high conductivity, compared to the filler phase which yielded high permittivity but low conductivity, and to the binder phases that had low permittivity and low conductivity.More importantly, tetragonal MAPbI 3 has been proven to be ferroelectric. [22]Such a relatively conductive, ferroelectric compound between the filler and binder phases acted as a dielectric lubricant by allowing the externally applied electric field during measurement to be more evenly distributed from the binder phase to the filler phase while buffering the lag from the filler's ferroelectric domains to the binder's ferroelastic lattice through intermediate ferroelectric conductive channels.This local effect was well reflected in the macroscopic dielectric properties with the drastically reduced dielectric loss of the composite samples with the additive, compared to those without the additive, at low frequencies (Figure 2a; Figure S13b, Supporting Information).
It can also be noticed that, after poling, the permittivity of all the samples increased by ≈20% in average (Figures S11a,S12a,S13a, Supporting Information) compared to the values before poling.This was expected because a tetragonalrich perovskite phase, whose permittivity usually increases after poling, [28] was identified in the XRD pattern for the filler (Figure S9, Supporting Information).This was also unambiguous evidence that all the composite samples, except the PZT-C5-D025 which was too conductive to sustain high voltage, could be effectively poled.

Ferroelectric and Piezoelectric Properties
To examine the ferroelectric properties of the composites, polarization and strain were measured on the samples as a function of electric field.Measurement conditions were screened beforehand by varying the frequency of the electric field from 1 Hz to 1 kHz as well as by testing the maximum allowed electric field that could be applied without breaking down the samples.The breakdown electric field was found to be just above 50 kV cm −1 , and therefore, the maximum applied electric field was ≈50 kV cm −1 for each sample.
Figure 3a shows the resultant polarization (P)-electric field (E) hysteresis loops (widely known as the P-E loops) for the composite samples as well as for the pressed crushed crystals as references.The pure binder phases hardly exhibited any noticeable ferroelectric behavior while ferroelectric switching was observed for all the composite samples.According to literature, [19] the (PTMA)CdBr 3x Cl 3(1-x) crystals have been able to show a moderate ferroelectric behavior, although much weaker than that of the PZT ceramics.The reported ferroelectric switching in the (PTMA)CdBr 3x Cl 3(1-x) crystals was caused by the swing-type reorientation of the spontaneous polarization, meaning that the polarization only slightly rotated by an angle of a few degrees during the ferroelectric switching.The typical >90°reorientation of polarization commonly seen in conventional ferroelectrics was forbidden in the (PTMA)CdBr 3x Cl 3(1-x) crystals due to crystallographic orientation. [19]However, in this work, the pressed crushed crystals were in fact polycrystalline, which eliminated observable P-E loops at the macroscopic scale.
The remanent polarization (P r ) of the composites decreased with the increase of the binder volume fraction as shown by comparing the P-E loops of samples PZT-B5, PZT-B10, PZT-B20, and PZT-B30.This was simply because the binder phase was non-ferroelectric in the long-range at a macroscopic scale while only the fillers provided the ferroelectric properties.As a result, the composite samples with the smallest binder volume fractions, i.e., PZT-A5, PZT-B5, and PZT-C5, showed the largest polarization values (maximum polarization (P m ) of 4-7 μC cm −2 at 50 kV cm −1 and P r of 3-5 μC cm −2 ).The difference in the polarization values among these PZT-5 wt.% (PTMA)CdBr 3x Cl 3(1-x) composites was thought to be attributed to the microstructure and complex electrical interaction between the filler and binder phases.Despite these differences, the polarizations of all the composite samples in this work were superior to those for the PZT-LMO (P r <1 μC cm −2 ) and PZT-am-TiO 2 (P r ≈ 2 μC cm −2 ) composites. [16,17]The larger polarizations were believed to be derived from the ferroelastic nature of the binder phases which eased the domain pinning caused by the LMO or am-TiO 2 particles between the PZT particles.
From the P-E loops, it can be seen that all composites had the same coercive field of ≈30 kV cm −1 .This coercive field value is relatively high and thus these composites could be considered as hard ferroelectrics.Hard ferroelectrics have a greater tendency to maintain their long-range ordering and are less easily switchable.As a result, they are particularly suitable for applications  that require high levels of electrical excitation and/or mechanical stress, such as high voltage or high power generators and transducers. [29]he existence of true ferroelectric switching was also evidenced in the strain (S)-E hysteresis loops (known as butterfly loops), as shown in Figure 3b.The butterfly shape of the bipolar loops is the result of the contribution of three types of effects, including the converse piezoelectric effect of the lattice, non-180°domain switching, and other domain wall motion. [30]herefore, the evolution of the maximum strains with different types and volume fractions of the binder phases generally agreed to the trend of the polarizations for all the composite samples (Figure 3a).Nevertheless, the maximum strain level (0.03%-0.04%) exhibited by the PZT-5 wt.% (PTMA)CdBr 3x Cl 3(1-x) composites was only a tenth of those exhibited by the ceramic counterparts. [31]This might disadvantage the recycled materials from being reused in actuators.
After poling, the piezoelectric properties were evaluated for all the samples.Figure 4 compares the piezoelectric charge coefficient (d 33 ) and voltage coefficient (g 33 ) of different samples, where g 33 = d 33 /(ɛ 0 •ɛ′) (ɛ 0 is vacuum permittivity).Table S1 (Supporting Information) lists the full dataset of these piezoelectric properties which also includes the piezoelectric energy harvesting figure of merit (FOM), where FOM = d 33 •g 33 , a factor indicating the sensitivity of piezoelectric sensors.FOM is positively related to the energy density in the electromechanical coupling process and thus reveals the generic capability of the piezoelectric energy harvesting process. [32]he pressed crushed (PTMA)CdBr 3x Cl 3(1-x) pellets did not show any piezoelectric response (Table S1, Supporting Information), which was expected from the correspondingly negligible ferroelectric behavior.The PZT-C5 showed the largest d 33 value of 90-100 pC N −1 among all the composite samples (Figure 4a and Table S1, Supporting Information).Yet, the value was still only ≈20% of that for the PZ29 ceramics. [24]The degraded d 33 values compared to those of the ceramic counterparts seemed to be inevitable for all upside-down composites, as has also been observed with PZT-LMO and PZT-am-TiO 2 composites. [16,17]This was because grains and grain boundaries play a crucial role in achieving excellent d 33 values. [2]In the composites, the lowpermittivity, non-piezoelectric binder phases partially occupied the grain boundaries, leading to ineffective poling as the poling electric field was locally concentrated on the binder phases.
The introduction of only 0.1 wt.% MAPbI 3 additive further lowered the d 33 value by ≈50% (Figure 4a), which was likely because the semiconducting MAPbI 3 made the intergranular area more conductive and thus screened the local charges and interrupted the long-range, macroscopic piezoelectric interaction.Increasing the additive to 0.25 wt.% made the sample too conductive to be properly poled as the sample was short-circuited during poling.This could be because the volume fraction of the semiconducting additive was sufficient to create a conduction path a) Density values were relative to the high-temperature sintered ceramic values rather than to the theoretical densities.
for electrons injected by the poling field.Therefore, as discussed above, the threshold of volume fraction for the additives was determined to be in the range of 0.2%-0.4% (Table 1).To ensure the recycled materials are able to exhibit proper piezoelectricity, the volume fraction of the additive should be maintained below such a threshold.The d 33 values of the composite samples not only decreased with the increase of the binder phase by comparing samples PZT-B5, PZT-B10, PZT-B20, and PZT-B30 but also with the increase of the Br concentration in the (PTMA)CdBr 3x Cl 3(1-x) binder by comparing samples PZT-A5, PZT-B5, and PZT-C5 (Figure 4a).The former decrease agreed with the situation for the permittivity (Figure 2), which confirmed the determinate role of the low-permittivity, non-piezoelectric binder phases in the effectiveness of poling for the composites, as discussed above.The latter decrease revealed some new and valuable information that the Br/Cl ratio of the (PTMA)CdBr 3x Cl 3(1-x) binder phase might influence the d 33 values of the composites to a large extent of up to 25%, where the d 33 values of PZT-A5, PZT-B5 and PZT-C5 were roughly 65, 75, and 100 pC N −1 , respectively.The presence of Br seemed to suppress the development of piezoelectric properties in the composites.Thus, it is suggested that future research of other upside-down composites should adopt the (PTMA)CdCl 3 composition as the binder phase and try to eliminate the presence of Br.
Despite the inferior polarization, strain, and d 33 values of the composites compared to those of the PZ29 ceramic counterpart, the g 33 values of all the composites were comparable to, and some of them were even superior to, that of the commercial PZ29 ceramic (Figure 4b). [24]In particular, composite samples containing 5 wt.% of the (PTMA)CdBr 3x Cl 3(1-x) binder phase possessed the largest g 33 values in the range of 29-38 mVm N −1 with the PZT-C5 samples reaching the optimum of 37-38 mVm N −1 .][35] Similar to the trend for d 33 , the g 33 values also decreased with the increase of Br concentration among the PZT-5 wt.% (PTMA)CdBr 3x Cl 3(1-x) composites.Earlier reports have suggested that the incorporation of a suitable low-permittivity secondary phase into the highpermittivity piezoceramic matrix could minimize the permittivity of the composite while preserving the piezoelectric charge coefficient to the best possible extent. [32,36]The relatively large g 33 values obtained in this work underlined the same mechanism where the g 33 was independent of either the individual d 33 or the permittivity.An essential finding here is that the composite sample with 0.1 wt.% MAPbI 3 additive also exhibited a decent g 33 value which was comparable to that of the PZ29 ceramic (Figure 4b).However, for all the recycled composite samples, the FOM values were not yet comparable to those of the PZ29 ceramics (Table S1, Supporting Information).
Table 2 provides a broad perspective by comparing the data collected from various piezoelectric ceramics and composites that have been fabricated by different densification methods.It can be seen that among the ceramic or composite materials produced at ultralow temperatures, this work achieved uncompromised sensing functionality.

Influence of Stability of the Halide Perovskite Binder and Additive
Organometal halide perovskites are generally considered unstable when subject to moisture, elevated temperature, or light over an extended period. [23,46]Their ions may also be mobile when subject to external electric field. [47]Figure S14 (Supporting Information) indicates variations of Cd and Cl potentially caused by the electric field during poling (ion migration) and by aging of the samples being stored in desiccator at room temperature under ambient indoor light for one year.Pb and I from the additive compound could not be detected by EDX because of their extremely low contents.
It can be clearly seen that after poling, Cd + was driven away from the center part of the sample toward the surfaces.This observation is indirectly supported by literature where migration of Pb + and I − under electric field was visualized. [46]Possible variation of ions caused by aging over time seemed to be ambiguous according to the EDX results.However, according to Table S2 (Supporting Information) which compares the d 33 values measured at different time points, piezoelectric property of the PZT-C5 sample decreased by almost 50% after one year, despite staying stable in the course of the first several days.As the sample was repoled before measurement, such a substantial change might be caused by the instability of the halide compounds instead of aging of the poling status in the PZT filler particles.Evolution of the corresponding dielectric properties shown in Figure S13 (Supporting Information) was correlated to the changes caused by poling and aging.
On the other hand, the d 33 values of the PZT-C5-D01 sample seemed to be stable over a period of one year.A hypothesis for reasons could be that, to some extent, the semiconductive but highpermittivity MAPbI 3 additive in PZT-C5-D01 suppressed possible degradation or ion migration supposed to be stimulated by the depoling field existing in the PZT-C5 sample.Although the addition of MAPbI 3 seemed to alter the Cd/Cl ratio and induced changes of dielectric properties (Figures S13 and S14, Supporting Information), further investigation, e.g., by increasing the MAPbI 3 content to a detectable level, is needed to understand the influence of stability of MAPbI 3 on piezoelectric properties as well as on the (PTMA)Cd(Br,Cl) 3 binder phase of the composites.

Conclusion
This study has demonstrated the possibility of recycling broken oxide perovskite piezoceramics and waste organometal halide perovskite solar cells simultaneously with the method of constructing oxide-halide perovskite upside-down composites.The halide perovskite binder has effectively formed intergranular interfaces between densely packed oxide perovskite filler due to excellent wettability of the halide compound on the oxide surface.The recycled halide perovskite solar cell compound has also been introduced to the intergranular areas as a minor additive.While the fabrication of these oxide-halide perovskite composites has required only a marginal energy budget that is much lower than those in previous studies, the density of these composites has been significantly improved.Without the solar cell additive, the good densification has induced a reasonable piezoelectric charge coefficient (d 33 ) of 90-100 pC N −1 .Meanwhile, the low-permittivity binder phase has resulted in a piezoelectric voltage coefficient (g 33 ) of 37-38 mVm N −1 , which is comparable to many superior piezoelectric ceramics, making these composites potentially useful for sensing applications.This study has also found a threshold of 0.1 wt.%-0.25 wt.% (or volume fraction of 0.2%-0.4%)for the solar cell additive, below which the composites are still been able to show decent g 33 values.
By considering the fabrication temperature which implies the energy to be consumed during either manufacturing new materials or potential recycling production to give a second life to the used materials, the approach introduced in this work is advanced in terms of achieving uncompromised sensing functionality using only marginal energy budget to produce piezoelectric materials via recycling.Although further studies are needed to determine viable approaches to improve the recycling capacity per unit volume for waste perovskite solar cells without compromising the functionality of the composites (e.g., the energy harvesting capability), this study is already conclusive that broken piezoceramics and waste solar cells can be recycled and then be given a second life in applications with piezoelectric sensors.Such promising results lay down the foundation for future works on the simultaneous recycling of piezoceramics and organometal halide perovskite solar cells by improving the recycling rate of the halide composition, i.e., further increasing the additive volume fraction while retaining the good piezoelectric properties, especially for the g 33 .This will enhance energy-and environmental friendliness and thus will result in a cleaner ecological footprint for piezoelectric and photovoltaic research and industry.

Experimental Section
Synthesis of (PTMA)CdBr 3x Cl 3(1-x) Crystals: Figure S15 (Supporting Information) shows a flow chart of the synthesis works in this study.(PTMA)CdBr 3x Cl 3(1-x) crystals with x values of 0, 0.5 and 0.9 were synthesized by slowly precipitating mixed solutions of filtered acetonitrile and deionized water containing PTMA, Cd 2+ , Br + and Cl + .Starting reactants of PTMA-Br (98%, Sigma-Aldrich, USA), PTMA-Cl (≥98%, Sigma-Aldrich, USA), CdBr 2 (98%, Sigma-Aldrich, USA), and CdCl 2 (99.99%,Sigma-Aldrich, USA) were weighed stoichiometrically and were then dissolved in the mixture of acetonitrile and deionized water. [16]The volume ratio of acetonitrile to deionized water was 3:2 when the x value was 0 and 0.5 and 4:1 when the x value was 0.9. [19]The overall concentration of the mixed reactants was 13.3 w/v% of the mixed solution.The dissolved mixture was stirred for 24 h, and the obtained solution went through a 0.2 μm PTFE (polytetrafluoroethylene) filter.Consequently, from the filtered, transparent solution, cm-sized crystals were grown after leaving the solution still for three to four weeks.
Recycling of MAPbI 3 : Waste perovskite solar cells from previous research were collected from the laboratory. [48]The as-fabricated solar cells were the dye-sensitized type with the infiltrated perovskite into mesoporous TiO 2 and porous ZrO 2 layers.The thickness of the perovskite layer was estimated to be a few microns.As the solar cell structure was sandwiched between two glass substrates by epoxy, to access and hence recycle the absorber layer, MAPbI 3 , the sandwich structure was split using a diamond saw.The graphite and silver electrode materials were separated by scratching using a doctor blade, and the remaining absorber material (containing a mixture of perovskite, TiO 2 and ZrO 2 ) was scratched off, gathered, weighed, and mixed as an additive with the crushed (PTMA)CdCl 3 crystal in a mortar using a pestle.It should be noted that the relatively thick perovskite layer in the dye-sensitized solar cells allowed an easy collection of the perovskite compound by scratching.However, in thin film perovskite solar cells where the perovskite layer is normally thinner than 1 μm, proper solvents may be needed to help detach the thin layer, which requires further study.The quantity control of the perovskite additive is explained in Section 4.3 below.

Figure 1 .
Figure 1.a) STEM micrograph and EDX elemental maps for crushed PZT particles coated by binder phase C ((PTMA)CdCl 3 ), and b) FESEM micrograph and EDX elemental maps for the composite sample PZT-C5.

Figure 3 .
Figure 3. Dependence of a) polarization and b) strain on electric field for the binder phase and composite samples measured at 1 Hz.

Table 1 .
Summary of the compositions and densities of the samples studied in this work.

Table 2 .
Comparison of selected piezoelectric ceramics and composites fabricated by various densification methods.