Uncovering the Printability, Morphological, and Functional Properties of Thick Ferroelectric Composites for Next‐Generation Low‐Cost Scalable Sensors

Herein, new insights into the printability, rheological, morphological, dielectric, and piezoelectric characteristics of microscale thick‐film particulate ferroelectric composites are provided. This is achieved by the fabrication and printing of diethylene glycol monoethyl ether acetate (DGMEA)/DI7025/barium titanate (BaTiO3) inks with varying filler content. The effects of DGMEA and BaTiO3 microparticles on the rheological response are investigated using shear, stress frequency dependence, and three‐part recovery tests. The structural and morphological characteristics are analyzed using Fourier‐transform infrared spectroscopy, thermogravimetric analysis, and scanning electron microscopy. The functional sensing behavior is investigated through impedance spectroscopy, piezoelectric, and ferroelectric experiments. The dependences of the dielectric and piezoelectric properties of the composites on BaTiO3 volume fraction are reported and analyzed in terms of Yamada model. The best performance is obtained from the composites with 40 vol% BaTiO3 with parallel‐layered structured connectivity. The results offer important insights for the future development of new and improved thin film sensors.


Introduction
Over the past few years, the demand for flexible wearable sensors has grown rapidly, due to the increasing demand for wearable devices in consumer electronics and healthcare services. [1,2]he most important component of a functional flexible wearable system is its active functional (sensing, actuating, and energy harvesting) element. [3][6] These materials have shown promise in a variety of applications, ranging from energy storage to advanced sensing technologies.Among the ABO 3 piezoelectric materials, barium titanate is a well-studied perovskite oxide material with unique piezoelectric and dielectric properties that have attracted widespread attention for use in flexible electronic systems. [7]These properties make composites fabricated from BaTiO 3 ideal for a variaty of applications such as sensors, actuators, energy harvesters, smart materials and devices, structural health monitoring, and energy conversion. [8]Despite their great potential, traditional methods of producing BaTiO 3 -based composites, such as spin coating, spray printing, hot pressing, and tape casting, are often limited in terms of macrostructure patterning, scalability, and cost-effectiveness. [9]he screen-printing technique offers an alternative and more cost-effective approach for the manufacturing of piezoelectric composites for electronic applications. [10]More specifically, it allows the simple and scalable fabrication of BaTiO 3 -based composites with tailored microstructures and properties, making it suitable for large-scale production.13] In this study, the structural and functional properties of screen-printed BaTiO 3 -based composites were investigated in detail.The focus was on the effect of processing parameters on the microstructure and piezoelectric performance of the composites and the resulting impact on their performance as sensors.The results of this study provide a comprehensive understanding of the relationship between the microstructure and functional properties of the screen-printed BaTiO 3 -based composites and offer important insights for optimizing the performance of these systems as sensors.This work will pave the way for the future implementation of screen-printed ferroelectric composites in various sensing applications and help to drive the development of new and improved sensors based on these materials.

Results & Discussion
Figure 1a shows the flow curves obtained for the ferroelectric inks at different BaTiO 3 volume fractions.In the low-shear rate region, it is evident that viscosity increases with the increase of the filler phase in the inks.The majority of the inks exhibited a shear-thinning thixotropic behavior, indicating the presence of a weakly attractive network.The viscosity of the inks decreased when the shear rate increased and then increased when the shear rate decreased.The viscosity remains higher for the highervolume-fraction inks across the whole shear rate range due to the enhanced interparticle interaction.The results also indicate that the presence of the viscosity adjuster ensures an effective dispersion of the BaTiO 3 even in the high-volume-fraction inks.The viscosity adjuster helps to ensure that the ink will perform optimally in the printing process, leading to high-quality, consistent, and reproducible printed patterns.In contrast, for the ink with 50 vol% of ceramic filler, shear thickening was clearly evidenced at shear rates of about 28-51 s À1 , where the viscosity of the ink increased significantly with the increase in the shear rate.This can be attributed to the higher volume fraction of ceramic filler, which results in stronger particle interactions and a more rigid structure in the ink under shear.The observation of shear thickening in the ink with 50 vol% of ceramic filler highlights the importance of particle interactions in controlling the rheological behavior of suspensions and non-Newtonian fluids.
Frequency-dependent curves can give insights into the network formed by particle-particle and polymer-particles interaction in the inks.More specifically, the response of the particulate polymer-based inks at low frequencies is related to the microstructural details of the particle dispersion (particle-particle interactions), while the response at the high-frequency region reflects the characteristics of the polymer matrix. [14,15]igure 1b displays the frequency-dependent curves of DI7025-BaTiO 3 inks measured at a stress of 10 Pa.The storage modulus of the inks increased with increasing filler concentration, indicating the formation of a stronger network.In addition, with increasing frequency, G 0 exhibits a weaker dependence from the frequency showing that the network formed in the inks is rigid.
Figure 1c shows the shear-stress dependence of the storage modulus (G 0 ) (solid linestyle) and the loss modulus (G 00 ) (dashed linestyle) of the composites inks.For low stresses, G 0 and G 00 remain stable, indicating the formation of a firm network between the BaTiO 3 and the polymer chains of DI7025.It is clear that the higher the volume fraction, the more insusceptible to deformation the network is.Exemption to this behavior is the ink with 50 vol% of ceramic filler.With the increase of the shear stress, breakdown of the network occurs, with the breakdown stress being higher for the inks with higher particle content.
All the inks indicate a solid-like behavior G 0 > G 00 with exemption to the high-stress area where the inks become more liquidlike dominant, G 0 < G 00 .As described in the Experimental Section, the screen printability of the inks can be evaluated by a three-part recovery experiment which combines two types of shearing experiments (oscillatory and rotational shearing).Figure 1d shows the obtained results.In the first 60 s oscillatory, a minimal amount of disturbance was inflicted on the inks by applying a very low shear stress of 10 Pa in order to measure the storage modules in the linear viscoelastic region.During the second part, a high shear rate of 100 s À1 was introduced, resulting in the disruption of the rheological network and a corresponding decrease in the ink viscosity.The viscosities of the inks decrease consistently.The presence of the microparticles and the viscosity adjuster enable the formation of a network that is gradually dismantled under shearing.The third part is similar to the initial part and helps to determine the restoration of the storage modulus between these two parts, as evidenced by the recovery ratio, demonstrated in Equation ( 4).The calculated values of the recovery ratios are presented in Table 2.All the inks exhibit high and relatively high recovery ratios with immediate return to their initial values reported in the first part of the experiments, indicating small changes in their condition after loaded shear.The ink with 50 vol% of BaTiO 3 recovered only 83.7% and based on the aforementioned rheological behavior, its printing might lead to an inconsistent structure.
Fourier-transform infrared spectroscopy (FTIR) was used to identify the composition of the polymer used for the fabrication of screen-printed composites.Matching the peaks of DI7025 with the literature, it is observed that it is composed of poly(ethylene) vinyl-acetate (PEVA) with 40% of vinyl acetate content and DGMEA which is a plasticizer used to control the elasticity and the flexibility of the polymer.Figure 2a shows the FTIR spectra of DI-7025, DGMEA, and PEVA.The peaks at 2970 and 2869 cm À1 correspond to the antisymmetric and symmetric stretching of CH 2 groups, respectively, and they are attributed to both PEVA and DGMEA.Similarly, the sharp absorbance peak at 1736 cm À1 appeared due to the stretching vibration of the carbonyl group (C=O).The very small peak at 1534 cm À1 very likely corresponds to a hindered phenol or a phosphite with nitro compounds used as a stabilizer to prevent the degradation of the polymer due to oxidation or heat.The weak absorption bands around 1450-1350 cm À1 are attributed to the bending vibration of the methyl groups in DGMEA and PEVA.The peak at 1230 cm À1 is a confluence of the 1241 cm À1 peak of PEVA and the 1236 cm À1 peak of DGMEA.The 1241 cm À1 peak corresponds to the stretching vibration of C─OH group in the acetate of PEVA while 1236 cm À1 is typically associated with the C─O─C stretching vibration of the ether functional group in DGMEA.The peaks between 1051 and 1020 cm À1 are associated with the stretching of C─O groups in both PEVA and DGMEA.Finally, the broad absorption band around 3400-3200 cm À1 appeared only in the PEVA spectra, corresponding to O─H stretching vibrations from the residual moisture in it. [16,17]he FTIR spectra of the formulated inks and the cured screenprinted composites are shown in Figure 2b,c, respectively.The presence of DI7025 is confirmed in all cases of inks and printed composites.It is also observed that the stainless steel substrate affects the FTIR spectra of the printed composites.The stainless steel substrate can cause radiation from the instrument and lead to interference with the signal.This interference can cause changes to the baseline and the shape of the spectra, making it difficult to analyze or interpret the data accurately.However, a weak peak around 3300 cm À1 is shown in all printed composites.This peak corresponds to residual moisture after the fabrication procedure.
The chemical stability of the composites and the polymer used for the fabrication of the composites was determined by thermogravimetric and derivative thermogravimetric analysis (TGA) at elevated temperatures, as shown in Figure 3.It can be seen that the curves declinesharply between 100 and 200 °C due to the volatilization of the solvent DGMEA (flash point: 96 °C, boiling point 202 °C). [18,19]The structural decomposition at around 250 °C is attributed to the carbonyl group of DGMEA.The two broad peaks on the TGA first-derivative curves at about 350 and 460 °C are related to PEVA and verify the general theory of the two-stage degradation process in the thermal decomposition of PEVA.More specifically the first peak can be attributed to the deacetylation of the vinyl acetate segments while the second one is related to the decomposition of the residual ethylene components along the polymeric chains. [20,21]) (b) (c) When the concentration of the BaTiO 3 increased, the weight loss of the composites decreased.Moreover, the presence of the inorganic ferroelectric phase and the plasticizer contribute to a small shift in the maximum of the peaks toward lower temperature, accelerating the degradation rate of the material.
Figure 4 shows the cross-sectional scanning electron miscopy (SEM) images of the cured screen-printed composites.In all cases, the shape of the particles is irregular.It is also evident that the BaTiO 3 microparticles are distributed in the polymer matrix for concentrations up to 30 vol% with only very minor BaTiO 3 agglomerations.As a result, the printed composites up to this concentration of the inorganic ceramic filler exhibit a 0-3 type of connectivity since there is a continuous polymer matrix between the filler microparticles (light gray particles in the middle section).However, printing challenges arise at higher BaTiO 3 concentrations, impacting the overall microstructure.These complications are attributed to the increased inhomogeneity and stronger interparticle interactions at higher ceramic phase concentrations.It is assumed that such interactions can lead to the formation of more organized, albeit less homogeneous, structures with varying ceramic phase densities.A comprehensive analysis revealed microstructural anomalies at higher concentrations (40 and 50 vol% BaTiO 3 ).In the 40 vol% BaTiO 3 samples, delamination and interlayer volume fraction variations were observed (see Figure S2, Supporting Information).These irregularities are thought to arise from excess curing times of individual layers, thus suggesting a need for optimized curing conditions.In the 50 vol% BaTiO 3 samples, an increased volume fraction variation within the active layer was observed, attributed to inadequate amounts of plasticizer in the ink formulation.Such variations could be beneficial for sensing applications, as they can result in a reduced overall system permittivity.Finally, the  homogeneity of the ink can be affected by many other factors, for example, the ferroelectric filler, the viscosity of the ink, the mesh size of the screen, and the pressure applied during the printing.Further improvements can be achieved by more advanced particle processing and composite mixing, as well as by printing multiple thick-layer composite ink layers with optimized and fully controlled printing parameters. [7]n the analysis of the SEM cross-sectional images of the cured screen-printed composites using imageJ software, it was noticed that the measured volume fraction (see Table 1) differs from the volume fraction of the formulated inks.Taking into account the TGA analysis of the inks, it is observed that there is a noticeable weight loss in the temperature range where the curing process takes place.The evaporation of DGMEA leads to the reduction of the volume and as a result the increase of the packing density of the BaTiO 3 particles.
Sawyer-Tower technique was used for the study of ferroelectric properties of the composites for polarization switching. [22]he ferroelectric hysteresis loops at room temperature, under 100 kV cm À1 and 10 Hz of the screen-printed composites with varied filler contents, are shown in Figure 5.The hysteresis measurements reveal that the value of remnant polarization (P r ) and the saturation polarization (P s ) of the composites increase with the increasing of BaTiO 3 up to 30 vol%.It can be observed that with increasing the amount of the ferroelectric phase above 30 vol%, the samples start exhibiting a tilted loop which, in general, is an indication of compositional inhomogeneity of the microparticle powder.
Additionally, it should be noted that the ferroelectric loops exhibit a subtle increase in polarization after the peak field.This can be attributed to the inherent dielectric properties of the polymer matrix and the residual amount of the plasticizer in the composites.These factors contribute to higher dielectric losses, particularly at lower frequencies, as confirmed by Figure 1c, Supporting Information.This frequency-dependent behavior is indicative of some degree of loss or conductivity in the composite material, confirming the observed characteristics in the ferroelectric loops.The values of remnant polarization, saturation polarization, as well as the coercive field of the screenprinted composites are displayed in Table 1.
The dependence of the relative permittivity and the dielectric loss of the screen-printed composites on the volume fraction at 1 kHz are shown in Figure 6a,b respectively.Although a standard sampling frequency of 1 kHz was employed for these measurements, it should be clarified that the dielectric properties exhibit some frequency dependence, particularly at lower frequencies.This is corroborated by the presence of higher dielectric losses at these frequencies, as indicated in Figure S1, Supporting Information.Additionally, the loss tangent data within the same frequency range and temperature, where no relaxation peaks are present, provides further insights into this frequency-dependent behavior (see Figure S1, Supporting Information).
The relative permittivity experimental data reveal a linear increase in relation to the volume fraction of the filler up to 30 vol%.With a further increase of the ceramic phase, a decrease in the relative permittivity is observed.Based on the morphological characterization, this behavior can be attributed to the inconsistent distribution of the BaTiO 3 microparticles within the polymer matrix during the screen-printing process, leading to a parallel layered structure connectivity with different concentrations of ceramic fillers in the individual layers (see Figure S2, Supporting Information).In general, if the individual layers of a composite exhibit a different relative permittivity, the overall effective permittivity of the composite will be lower than the permittivity of the high-permittivity layer since the low-permittivity layers can reduce the average electric field within the composite.The deviation of the total permittivity from the high-permittivity layer also depends on the volume fraction of the individual layers, their thickness, and the arrangement of the layers.The reduction in the overall permittivity of parallel-layered structured systems with layers of different densities of the same ferroelectric material has been also reported by Roscow et al. [23] In addition, the dielectric loss curves (tanδ), Figure 6b, illustrate a decreasing behavior with an increase in the volume fraction of the ferroelectric phase.In general, increasing the volume  fraction of the ferroelectric phase may lead to an increase in dielectric losses due to the movement and the alignment of the ferroelectric material's dipole to the applied electric field.However, this behavior is not observed in this study.Taking into account the TGA analysis (Figure 3), the composite systems with lower volume fraction demonstrated higher remnant amounts of DGMEA after the curing process.DGMEA is a polar solvent and can increase the polarity of the composites.This increased polarity can lead to an increase in the dielectric losses of the system.The dielectric loss (tanδ) of a piezoelectric material is important for the efficient operation of the piezoelectric transducer as it is related to the amount of energy dissipated during the energy conversion process.In a piezoelectric system, a low tanδ is favorable to increase the electromechanical coupling and therefore the efficiency of conversion from mechanical to electrical energy. [24]t is important to address an observed abnormality in the behavior of 40 and 50 vol% BaTiO 3 samples.Despite the similarity in measured volume fractions, variations in their properties are noted.This inconsistency can be attributed to two key factors.First, the TGA indicated that the residual amount of the DGMEA in the 50 vol% samples is less than that in the 40 vol% samples.This can also be confirmed by the lower dielectric losses of the 50 vol% samples.Second, a more pronounced variation in volume fraction within the material for the 50 vol% samples was detected, as confirmed by the morphological characterization.These factors influence the dielectric properties, emphasizing the complex relationship between composition and structure in the performance of these composites.
The piezoelectric charge coefficient which relates the dielectric displacement to stress and strain to the electric field was used for the evaluation of the piezoelectric performance. [10]In this study, the effective piezoelectric coefficient was measured in the thickness mode (perpendicular to the electrode surface).Figure 6c shows the d 33 data as a function of the volume fraction of the BaTiO 3 filler.The piezoelectric coefficient d 33 of BaTiO 3 / DI7025 screen-printed composites increases with the content of BaTiO 3 when contents are equal or below 40 vol%.It is also observed that the piezoelectric performance of 50 vol% composites is similar to the 40 vol% ones and does not further increase.The distributions of the ferroelectric particles, as well as the structural connectivity, are factors that can have a significant impact on the output piezoelectric performance of composite materials, with well-ordered structures and strong connectivity leading to a stronger piezoelectric response.
Finally, the piezoelectric voltage constants (g 33 ) were calculated for each composite using Equation ( 5) and are presented in Figure 6d.The g 33 values generally increase from 8.76 to 33.16 m Vm N À1 as the volume fraction of the BaTiO 3 in the screen-printed composites increases.However, the data does not follow a linear trend with the volume fraction of the filler.Since the piezoelectric voltage constant is strongly dependent on relative permittivity, the composites with different structural connectivity exhibit different performances.More specifically, the composites that demonstrated a 0-3 connectivity exhibit lower g 33 values than the composites of a parallel-layered structure attributed to their lower relative permittivity values.When piezoelectric systems are used as sensors, a high g 33 value is desirable to increase the sensitivity of the sensor as it describes how much electrical signal can be generated in response to a given mechanical stress.
The dielectric and piezoelectric performances of the screenprinted composites obtained from the experimental measurements are compared in Figure 7 with the theoretical model of Yamada [25] .The Yamada model is a mathematical model based on the effective medium approximation (EMA) and usually shows close matching with the experimental behavior of piezoelectric composites.The equation that can describe the effective relative permittivity is given below.
where ε f is the relative permittivity of the BaTiO 3 microparticles, ε m the relative permittivity of the polymeric matrix, V f the volume fraction of the filler, and n = 4π/m a dimensionless parameter depending on the shape and the orientation of the ceramic filler.
On the other hand, the effective piezoelectric coefficient d 33(ff ) can be calculated by the equation below.
where V f the volume fraction, α the poling ratio, G the local field coefficient, and d 33(f ) the piezoelectric charge coefficient of the filler particles.The local field coefficient depends upon the relative permittivities of the polymer matrix and the ceramic filler and can be given as where n is the parameter attributed to the shape factor as previously described and ε m and ε f the dielectric constant of the polymer matrix and the ceramic filler accordingly. [26]he image processing software ImageJ was used for further analysis of the SEM images and the evaluation of the microparticle shape characteristics.More specifically, the aspect ratio was calculated by fitting an ellipse on each well-defined particle on the images (AR ¼ MajorAxis ½ = MinorAxis ½ ).Similarly, circles were fit on the particles for the estimation of their circularity (Circularity ¼ 4π Â Area ½ = Perimeter ½ 2 ) with the circularity approaching 1, indicating a perfect circle.On the other side, if the circularity approaches 0, it indicates an increasingly elongated shape.
The histograms and the fitted normal distributions of these two standard shape descriptors are demonstrated in Figure 7a, b.In Figure 7c the shape parameter n has been calculated for a range of aspect ratios.Taking into account the limits of the aspect ratio histogram, it was estimated that the shape factor must be in the range of 3-9.2.The effective relative permittivity was initially estimated using the relative permittivity of the polymer (DI7025) as it was measured using the impedance analyzer (ε m = 5.36) (Figure S3, Supporting Information).The shape parameter n was calculated by fitting the experimental data using the Yamada model.However, the n (n = 14) estimated utilizing the relative permittivity of the pure polymer as an input cannot be correct since it is out of the range given in the study shown in Figure 7c.When a solvent is added to a polymer or a polymeric system to control the viscosity, it can modify the dielectric properties and potentially increase the overall permittivity.This effect occurs because the solvent molecules can separate the charges in the polymer, creating a more conducive environment for electrical energy to be stored.The change in permittivity will depend on the specific solvent used, its concentration in the solution, and the dielectric properties of both the solvent and polymer.Furthermore, the effect on the permittivity may also be influenced by any chemical or physical interactions between the solvent and polymer molecules.In the case of the screen-printed DI7025-BaTiO 3 composites, from the previously described physiochemical characterization, it is evident that the remaining plasticizer (DGMEA used to control the viscosity) and water are present in the different composites systems, thus, causing an increase in the measured relative permittivity of the polymer matrix.
Given the fact that the aspect ratio and the shape parameter of the microparticles are known, the permittivity of the polymer matrix can be estimated using the equation of Yamada for the effective permittivity of a composite system.Consequently, the shape parameter n was re-estimated by fitting the experimental data to the theoretical one and using increments of the relative permittivity of the pure polymer (min = 5.36, max = 18.36, step = 1) as an input (Figure 7d).As shown in Figure 7d, if n % 5, then the relative permittivity of the matrix is estimated to be ε m = 13.36.In Figure 7e, the fitted curve appears to match the experimental permittivity of the composites with 0-3 connectivity (10-30 vol% filler content before printing and curing).A similar behavior is observed for the experimental d 33 values (Figure 7f ), where the data follows the overall trend of the estimated data curve.The poling ratio was found to be around 0.8; thus, it is a parameter depending on the poling conditions such us poling time, poling temperature, and poling field strength.The poling ratio is defined as the ratio of the actual polarization achieved during the poling process to the maximum polarization that can be achieved theoretically and is an important parameter that characterizes the effectiveness of the poling process.A high poling ratio indicates that the material has a high degree of polarization and thus a strong piezoelectric response.

Conclusions
This study has shown the feasibility of reproducibly manufacturing multilayer microscale thick-film ferroelectric composites using screen printing as a deposition process.Initial investigations focused on the rheological behavior of ferroelectric inks, revealing that both viscosity and storage modulus are dependent on the filler phase concentration.Most inks displayed shearthinning and thixotropic behavior, indicating a weakly attractive network among the particles.Screen-printing simulation experiments showed high recovery ratios across all formulations.However, the ink containing 50 vol% BaTiO 3 is notable for its lower recovery ratio, suggesting that it could result in inconsistent printed structures.This makes the manufacturing and characterization methodology suitable for semiautomatic or automatic pilot-line production processes.
FTIR spectroscopy confirmed that the polymer matrix comprises poly(ethylene) vinyl-acetate and DGMEA, which is crucial for understanding composite behavior.Microstructural analysis indicated that up to 30 vol% BaTiO 3 , the particles were evenly distributed within the polymer matrix.However, concentrations beyond this led to challenges in printing and resulted in a layered structure.TGA showed that the presence of the inorganic ferroelectric and plasticizer accelerated the degradation rate of the composites.The evaporation of DGMEA in the temperature range of the curing process leads to composites with higher volume fractions than the targeted ones.The screen-printed composites for concentrations higher than 30 vol% exhibited flattened hysteresis loops and a decrease in relative permittivity.The dielectric loss decreased with increasing volume fraction of the ferroelectric phase due to the less-conductive polymeric phase in the composites The screen-printed composites with parallel layered structure connectivity showed enhanced piezoelectric performance due to their lower relative permittivity.Finally, the dielectric and piezoelectric performance of the composites with 0-3 connectivity showed good agreement with the Yamada model.Overall, the study provides comprehensive insights that are anticipated to facilitate the development and integration of screen-printed ferroelectric composites in various sensing applications, thus promoting the advancement of cost-effective sensor technologies based on these materials.

Experimental Section
Ink Preparation: Barium titanate (BaTiO 3 ) micropowder (<3 μm) (Sigma-Aldrich) with a density of 6.08 g cm À3 was mixed with a solvent-based dielectric ink with excellent stretch properties and density of 0.93 g cm À3 called DI-7025 (Nagase ChemteX, USA).Diethylene glycol monoethyl ether acetate (DGMEA) (99%, Thermo Scientific) was used as a viscosity adjuster.The composition of the different inks prepared for this study is listed in Table 2.
Screen Printing: A typical screen-printing setup consists of a screenprinter bed, a substrate, a screen with a mask, and the squeegee.The squeegee was moved across the screen and forced the ink to the open mesh apertures and onto the substrate.The mask defined the desired pattern (see Figure 8). [3]The manufacturing of the DI7025-BaTiO 3 composites was performed using a manual printed circuit board (PCB) stencil screen-printer machine.A polyester screen mesh (140 holes inch À1 , 22°a ngle) was used for the printing.A 15 cm-long polymer squeegee forming an angle of around 60°with the screen was used to push the ink through the mesh.The "off-contact" (the distance between the screen mesh and the printing surface) was fixed at 1.5 mm.All composites were printed on stainless steel shim (t = 100 μm).Three layers of ferroelectric ink were printed in each sample.Each layer was cured separately in an off-vacuum oven at 120 °C for 5 min.After the printing of the final composite layer, the samples were kept in the oven for 10 min in order to fully cure the composites and remove the excess solvent without damaging the samples.Finally, a low-temperature curable silver ink (1901-S) provided by Ferro was printed as a top electrode.After manufacturing, the excess stainless-steel shim was cut in the dimensions defined by the printed composites, allowing the individual characterization of the samples (see Figure 1).Inkscape was used for the design of the artwork.A solvent and water-resistant Diazo-UV-polymer photoemulsion (AZOCOL Z1) purchased by Screen Stretch Ltd. was used for the production of the stencils.
Scanning Electron Microscopy (SEM): The microstructure and the thickness of the screen-printed composites were evaluated using cross-section images acquired by SEM (Hitachi, SU3900) at a voltage of 15 kV and a working distance between 9 and 10.5 mm.
Thermogravimetric Analysis (TGA): TGA was used in order to evaluate the behavior of the composite inks when heated, as well as the amount of ceramic material present in each individual system.The analyses were conducted in nitrogen atmosphere using a Setsys Evolution TGA 16/18.
Fourier Transform Infrared Spectroscopy (FTIR): The polymers used for the formulation of the ferroelectric inks, the inks, as well as the printed composites were analyzed via Fourier transform infrared spectroscopy (FTIR) using a Frontier FT-IR (PIKE Technologies Inc.) system.The FTIR spectra were recorded in the range of 800-4000 cm À1 .
Rheological Behavior: The rheological behavior of the composite inks was evaluated using a Discovery Series Hybrid Rheometer from TA Instruments and a parallel-plate geometry with a diameter of 25 mm.All the experiments were performed at room temperature (%20 °C).After approaching the top plate to the geometry gap, the excess ink was wiped off in all cases.The gap was set to 500 μm.The flow curves of ferroelectric inks were analyzed, varying the shear rate from 0.1 to 100 s À1 .The viscoelastic behavior of the inks was determined using oscillation experiments.The storage modulus (G 0 ) and the loss modulus (G 00 ) were quantified using stress sweep experiments, increasing the oscillation stress from 0.1 to 1000 Pa at a frequency of 1 Hz.The frequency dependence of the storage modulus was also evaluated at a constant stress amplitude of 10 Pa in a range of frequencies from 0.01 to 100 Hz.Finally, the rheological behavior of the inks during screen printing was simulated following the procedure by Inukai et al. [15] In order to include all the stages of screen printing, the experiments were divided into three parts.In the first part, an oscillatory shear stress of 10 Pa was applied at 1 Hz for 60 s.In the second part, a high-rate rational shear of 100 s À1 was applied to the inks for 30 s. Finally, the initial test parameters were applied again.The recovery ratio of the inks was calculated by the measured storage modulus at 60 and 150 s using Equation (4) and is shown in Table 2.
Dielectric, Ferroelectric, and Piezoelectric Properties: The dielectric behavior of the screen-printed composites was examined through impedance spectroscopy using an Agilent 4194A impedance analyzer.The printed barium titanate-based composites were tested at the frequency of 1 kHz at room temperature under an applied voltage of 0.5 V using a parallel-plate capacitor geometry.The maximum polarization, the remnant polarization, and the coercive field were measured on unpoled screen-printed composites using a Radiant RT66B-HVi ferroelectric test system with a frequency of 100 Hz and a double-bipolar input as the electrical signal.A custommade corona discharge single-tip electrode setup was utilized to perform the poling of screen-printed DI7025-BaTiO 3 composites at 115 °C with an applied field of 15 kV for 1 h from a 50 mm point source.Given the fact that the samples had cubic geometry (see Figure 8) with an upper active surface area of %1.2 cm 2 , the needle base distance can was set at 5 cm, giving an active area of 78.5 cm 2 .The samples were placed as close to the center of the electric field as possible.The longitudinal piezoelectric strain coefficients (d 33 ) were measured using a Berlincourt piezometer (Piezotest, PM300, Singapore) at 97 Hz, while the longitudinal piezoelectric voltage constants (g 33 ) were calculated using Equation (5).(5)   d 33 is the piezoelectric charge constant in pC/N, ε 0 the permittivity of free space, and ε T 33 the relative permittivity of the sample as poled through thickness and at constant stress.For consistency, the d 33 readings were taken 24 h after poling and after 10 s of the sample insertion in the piezometer's clamp.

Figure 1 .
Figure 1.a) Viscosity as a function of shear rate for the DI7025-BaTiO 3 composites, b) frequency dependence of the composite inks at 10 Pa, c) shear stress dependence of the storage (solid linestyle) and loss modules (dashed linestyle) and d) time-dependent three-part recovery test for the composites inks.

Figure 2 .
Figure 2. a) FTIR spectra of DI-7025, DGMEA, and PEVA.b) FTIR spectra of the uncured inks and c) FTIR spectra of the screen-printed composites.

Figure 3 .
Figure 3. TGA of the polymer and the composites inks after formulation: a) TGA thermographs indicating the remaining mass for each tested system and b) TGA first-derivative curves of polymer and the inks.

Figure 4 .
Figure 4. Backscattered SEM micrographs of the screen-printed composites' cross section.

Figure 5 .
Figure 5. Ferroelectric hysteresis loops of the screen-printed DI7025-BaTiO 3 composites at room temperature and 10 Hz.

Figure 6 .
Figure 6.a) Relative permittivity before poling, b) loss tangent before poling, c) d 33 , and d) g 33 data for the DI7025-BaTiO 3 composites as a function of volume fraction.

Figure 7 .
Figure 7. Histogram and normal distributions of a) particle aspect ratio and b) circularity.c) Estimation of n Yamada according to the aspect ratio of particles.d) Estimation of the relative permittivity of the polymer matrix according to n Yamada .e) Relative permittivity ε T 33 =ε 33 and f ) piezoelectric charge coefficient d 33 as a function of volume fraction.

Figure 8 .
Figure 8. Schematic representation of the screen-printing process for the manufacturing of micrometer-scale thick-film ferroelectric particulate composites.

Table 1 .
Characteristics of the composites.

Table 2 .
Composition and characteristics of the ferroelectric inks.