High-temperature crystallized thin-film PZT on thin polyimide substrates

Flexible thin-film piezoelectric devices are potentially useful in health monitoring (for both infrastructure and the human body) as conformal sensors, actuators, and energy harvesters on structures and objects with complex geometries. For example, there is interest in permanently emplaced ultrasound sensing systems for monitoring corrosion in natural gas pipelines and nuclear power plants; a conformal ultrasonic sensor array in close contact with a pipe would provide timely detection of damage that often occurs at joints, elbows, steps, and welds.¹ Kato et al. have demonstrated a large-area polyvinylidene fluoride (PVDF)-based ultrasonic imaging system that is able to detect targets behind thin objects and to obtain three-dimensional images.² Moving from mechanical systems to humans, a flexible imaging system placed on the body would allow a quick assessment of injury or bleeding of a trauma patient under critical conditions. In addition, a conformal Doppler transducer array positioned on the neck area could report carotid pulse.¹ 

To pursue the aforementioned goals, metallic foils have been employed as flexible substrates for Pb(Zr_x,Ti_1-x)O₃ (PZT) based piezoelectric microelectromechanical systems (MEMS) due to the capability of direct deposition,^3–8 as well as the high mechanical and thermal robustness of the metal. Replacing metallic foils with polymeric layers can provide additional advantages in MEMS applications due to the much lower elastic stiffness of the polymer. Piezoelectric actuators on a compliant elastic layer undergo larger displacement for the same voltage level applied. In the energy harvesting arena, lower elastic stiffness produces lower resonance frequencies that could provide a better frequency match to ambient vibrations. Besides the incentives of realizing conformal piezoelectric devices with enhanced properties, thin-film PZT on polymeric substrates should experience less clamping from the substrate. Both clamping and its combination with the large stress associated with film growth contribute to reduced domain wall motion and inferior piezoelectric and dielectric properties with respect to their bulk counterparts.^9–11 

A robust methodology for integrating high quality piezoelectric films on polymers is needed. However, a challenge lies in the high crystallization temperature required for the perovskite and the low decomposition temperature of plastics; this complicates direct-deposition on polymers. As an alternative, high strain PZT thin films can be first grown on a thermally robust substrate and then released and transferred onto a polymeric layer. Various methods have been employed to carry out this idea. The McAlpine group has demonstrated a transfer process for 5 μm wide, half micron thick PZT nanoribbons from a MgO substrate to polydimethylsiloxane (PDMS) for the flexible energy conversion purpose, in which PZT was first grown and patterned on MgO and then transferred to polydimethylsiloxane (PDMS) through wet-etching MgO in hot phosphoric acid.¹² Tue et al. etched patterned indium zinc oxide layers underneath a PZT layer in phosphoric acid to perform sub 5 μm fine-patterning of the PZT film.¹³ The relatively small structures utilized required short times to undercut the PZT, limiting damage to the piezoelectric. However, etch-induced damage to the PZT becomes more problematic for the release of larger areas.¹⁴ Besides etching, PZT release has also been achieved by laser lift-off,¹⁵ where a KrF excimer laser is pulsed through a sapphire substrate, in order to either melt or thermal shock an interfacial layer between the PZT and the substrate to enable separation.

In this work, PZT thin films were transferred from silicon substrates to few micron thick solution-cast polyimide by etching away an underlying ZnO release layer. Electrical characterization of the released PZT films on the free-standing polymer provides a useful model for studying the fundamental domain wall behavior in unclamped PZT thin films.

High-quality, dense PZT films require a crystallization temperature exceeding 550 ºC, substantially higher than the decomposition temperature of most polymeric materials. While it is in principle possible to utilize localized heating to achieve PZT crystallization on polymers, in practice, the resulting films are generally limited in size and/or functional properties.^16,17 PZT films therefore need to be transferred from a thermally robust substrate (i.e., silicon and magnesium oxide) to a polymer layer instead of being grown directly on the polymer. The basic approach of the transfer technique reported in this work is to etch away a release layer between the thermally robust substrate (here silicon) and the PZT-polymer combination. Throughout this article, denotations of “released” and “clamped” films refer to the PZT films constrained on the original silicon substrate prior to release and the films after release onto polyimide, respectively. The preparation process for both samples is illustrated as a flow diagram in Fig. 1.

In this work, ZnO was chosen as the release layer. Co-processing of PZT thin films and electronics based on ZnO thin film transistors (TFTs) have been accomplished for applications such as piezoelectric MEMS energy harvesters¹⁸ and adjustable x-ray optics,¹⁹ where no degradation of materials or electronics was found. In addition, ZnO TFTs have also been directly fabricated on polyimide.²⁰ This not only suggests that ZnO is compatible with PZT but also implies that the etchant (typically weak acids) used to pattern the ZnO does no harm to the PZT, the metal contact, or the polymer layer.

To assess the use of a ZnO release layer, a 4 cm × 4 cm piece of silicon was first sonicated in acetone followed by isopropanol (IPA), blow-dried with clean air, and ashed in an oxygen plasma to remove any organic residue on the substrate surface. Immediately after the cleaning process, a 130 nm ZnO layer was deposited by weak oxidant plasma enhanced atomic layer deposition (PEALD) at 200 ºC using diethylzinc (DEZ) as the precursor.²¹ Without breaking vacuum, a 10 nm thick Al₂O₃ layer was deposited by PEALD using trimethylaluminum (TMA) as the precursor to serve as a diffusion barrier between ZnO and PZT. Omitting the Al₂O₃ barrier resulted in reduced solubility for the release layer for PZT deposited directly on ZnO and correspondingly poor film release.

After deposition of the release layer, platinum was deposited to serve as an electrode for the PZT film. To allow characterization of clamped and released films from the same PZT deposition, a platinum electrode layer was deposited by magnetron sputtering at room temperature and patterned with circles on half of the substrate and a blanket layer on the other half, as shown in Fig. 2. The blanket Pt served as the bottom electrode for the clamped PZT, while the circular Pt areas served as the top electrodes of the PZT film after transfer to polyimide.

Thin-film Pb(Zr_0.52Ti_0.48)O₃ doped with either 1% Mn or 2% Nb was deposited using chemical solution deposition. 1-butanol and propylene glycol based Mitsubishi E1 (Mitsubishi Materials Corporation, Hyogo, Japan) PZT solution was first spin-cast onto the substrate at 4000 rpm, then hotplate baked at 100 ºC for 1 min to remove solvent, hotplate pyrolyzed at 300 ºC for 2 min, and finally crystallized at 700 ºC by rapid thermal annealing (RTA) for 1 min. These steps were repeated ten to fifteen times to reach the desired final film thickness, which ranged between 0.8 μm and 1.1 μm. A PbO capping layer was added at the end to convert any surface pyrochlore to the perovskite phase.²² 

Figure 3 shows the X-ray diffraction (XRD) pattern of a 0.8 μm sol-gel PZT film with 2% Nb doping deposited on a Pt/ZnO/Si stack. The film is free of secondary phases and has some {001} preferential orientation with a Lotgering factor²³ of 0.76. Field-emission scanning electron microscopy (FESEM) images (Fig. 4) show the PZT films to be dense and crack-free. The plan-view micrograph demonstrates that no excess PbO or pyrochlore is present at the film surface. The cross-section image shows the PZT film coverage over patterned platinum. The lateral lines seen in the PZT cross-section correspond to individual crystallization steps. As expected, the multiple sol-gel layers smooth the film and the initial Pt step is eliminated at the surface for films in this thickness range.

After characterization by XRD and FESEM, the samples were cleaved to separate the PZT on blanket Pt and PZT on patterned Pt sections. To complete the clamped PZT sample, a second Pt layer was then sputtered and patterned into circular top electrodes. For the sample to be released, blanket Pt was deposited and a 10–20 nm thick Al₂O₃ layer was deposited on Pt by PEALD. The Al₂O₃ layer blocks the migration of hydrogen evolved from the polyimide during the imidization which would otherwise degrade the PZT by reducing the switchable polarization.^24,25

For the samples to be released, the PI-2611 (MicroSystems, Parlin, New Jersey) polyimide precursor was spin-coated on the Al₂O₃ protected Pt layer. The precursor was spin-cast at 3000 rpm for 30 s. The film was hotplate baked at 100 ºC for 5 min to remove the solvent and hotplate cured at 330 ºC in air for 6 h. When fully cured, the polyimide had a thickness of approximately 5 μm.

The release process was conducted by immersing the sample stack in 33% acetic acid heated to 50–60 °C. Diluting and heating the etchant bath assist the mass transport of the etchant. Figure 5(a) shows the PZT-polyimide combination in the etchant bath near the completion of the release process.

After the PZT was fully separated from the original substrate, it was rinsed with deionized (DI) water and transferred to a 0.26 N tetramethylammonium hydroxide (TMAH) bath at room temperature to remove the Al₂O₃ diffusion buffer layer between ZnO and the circular Pt contacts.²⁶ The released film was then rinsed with DI water and blown dry. Figure 5(b) shows that the released film is flexible, and Fig. 5(c) shows the film taped onto a rigid carrier wafer with the patterned Pt side up. After release and some post-release handling, both the PZT and Pt are crack-free, and there is no sign of delamination of the electrodes, as can be seen in an optical microscopy image in Fig. 5(d). For a sample and a release layer of the aforementioned lateral dimension and thickness, the release process took approximately one day to complete.

Films were characterized by the measurement of the relative dielectric permittivity, loss tangent, and polarization-electric field (P-E) hysteresis. In addition, the ac electric field dependence of the permittivity was also measured at 1 kHz to assess the Rayleigh characteristics. Some electrodes were poled at three times the coercive field (160 kV/cm–180 kV/cm) at an elevated temperature of 125 °C for 15 min and cooled under the field.

Polarization-electric field hysteresis loop (P-E) measurements were made using a Radiant Technologies Precision LC ferroelectric tester with a 500 kV/cm excursion at 100 Hz and room temperature to assess the switchable polarization.

As shown in Fig. 6, for a released Pb(Zr_0.52,Ti_0.48)_0.98Nb_0.02O₃ film of 1 μm thickness, the remanent polarization $Pr$ increased to 26.0 μC/cm² compared to 17.5 μC/cm² for the clamped films on Si. The maximum polarization $Pmax$ also increased. This is believed to be a result of the improved domain reorientation due to reduced mechanical clamping of the PZT film.^10,27 The de-clamping of the films also appeared to reduce back-switching, which explained why the difference between $Pr$ and $Pmax$ was smaller on the released film. This is consistent with the released film having a smaller residual stress than at growth, when crystallized on the Si substrate. Because of the lower thermal expansion coefficient of the silicon compared to PZT, clamped PZT films (PZT on rigid substrates) are under a tensile stress upon cooling from the crystallization temperature to ambient, which tends to favor in-plane polarization.²⁸ The transfer of the PZT film to the polyimide likely relaxed some of this tensile stress. Shepard et al. demonstrated that tensile stress tends to rotate the P-E loop of PZT thin films clockwise.²⁹ As the released PZT film exhibited a counterclockwise rotation of the loop, the release presumably reduced the as-grown tensile stress.³⁰ 

Capacitance and dielectric loss (⁠$tan δ$⁠) were measured using a Hewlett-Packard 4284A LCR meter with an applied excitation signal of 30 mV in the frequency range from 100 Hz to 100 kHz. Measurements were made at room temperature, and each data point represents the average of five capacitors (600 μm diameter circular electrodes) across the sample. At 1 kHz, the relative permittivity (ε_r) was slightly above 1100 in the clamped PZT films and slightly below 1200 for the released films on polyimide, both with $tan δ$ values around 3%. It was found that in both clamped and released states, the permittivity dropped after poling (Fig. 7), presumably either because poling switched some of the a-domains to c-domains, with the latter possessing lower permittivity, or because of a reduction in the number or the mobility of domain walls. The poling-induced decrease in $εr$ for the released film was six times larger than that of the clamped film. This is possibly because in the released state, there is not as much restoring force from the substrate to suppress the dipole alignment with the electric field, thus resulting in more effective poling.¹⁴ The loss tangent was below 4% in all cases.

The dielectric response in ferroelectrics consists of an intrinsic and an extrinsic component. The former is due to the single domain lattice response, while the latter is due to the displacement of phase boundaries and, most importantly, domain wall motion.³¹ Some insight into extrinsic contribution can be obtained by performing Rayleigh analysis, that is, studying the ac electric field amplitude dependence of the response.^32,33 The Rayleigh law describes the field dependence of $εr$ (in the Rayleigh region) using the following form:

where the first term is the reversible Rayleigh parameter, representing the sum of intrinsic lattice response and reversible motion of domain walls. The second term measures the irreversible domain wall or phase boundary displacements, in which $α′$ is the irreversible Rayleigh parameter and $E0$ is the amplitude of the ac electric field. The Rayleigh parameters and the ratio of $α′$ to $εr, init′$ are excellent tools for quantifying the extrinsic contributions to the dielectric properties of the ferroelectric.

Figure 8 plots the dielectric permittivity measured at 1 kHz with respect to ac electric field amplitude. The films remained in the Rayleigh range to approximately 1/2 $Ec$⁠. The Rayleigh parameters were obtained by linearly fitting the permittivity in the Rayleigh region and extrapolating to zero field, where the intercept and the slope gave $εr, init′$ and $α′,$ respectively. The extracted parameters are summarized in Table I.

The data in Table I show that the clamped and released films have a similar reversible Rayleigh coefficient in the unpoled state, but upon poling, $εr,init′$ becomes much higher in clamped films than in released ones. The ratio of the irreversible to the reversible Rayleigh coefficient is higher in the released films regardless of the poling state. Both reversible and irreversible domain wall motion contribute to the dielectric response of the PZT. The Rayleigh results suggest similar reversible domain wall motion in the unpoled released and clamped films, which is consistent with the dielectric permittivity plots in Fig. 7. For the released PZT, greater irreversible domain wall motion was observed, which explains the increased $α′$ and $α′/ε′init$ in the released film.

In this work, ∼1 μm thick, ∼$1″×1″$ PZT thin films were transferred from thermally robust Si substrates to polyimide by etching away a ZnO release layer. Al₂O₃ buffer layers were used to prevent ZnO-PZT diffusion and to block imidization related hydrogen migrating to the PZT. The released films showed superior material properties compared to the same films on Si due to reduced substrate clamping. Polarization-electric field hysteresis measurements showed nearly a 45% increase in the remanent polarization (from 17.5 μC/cm² to 26 μC/cm²). Poling at 3 $Ec$ for 15 min at 125 ºC either induced more ferroelectric and ferroelastic realignment or significantly decrease the density of domain walls in released films, which reflected in a reduction in the relative permittivity of 17% compared to only 3% for clamped films. These measurement results, along with Rayleigh analysis, strongly confirmed the correlation between the reduced substrate clamping and improved domain wall mobility in the films on polymeric substrates. The released films and the transfer process developed in this work are attractive for flexible sensor, transducer, and energy harvester applications.

The authors would like to convey their appreciation for the financial support provided by the National Science Foundation through Grant No. IIP-1547877 as part of I/UCRC Clusters for Grand Challenges: Center for Tire Research.