WorksEngineering
Bilayer Periodically Poled Piezoelectric Thin-Film Lithium Tantalate Platform for Passive-Layer-Free Flexural and Torsional Actuation
Abstract
This work demonstrates, to the best of our knowledge, the first passive-layer-free thin-film lithium tantalate (LT) actuator platform, supporting both out-of-plane flexural and in-plane torsional actuation from a single bilayer periodically poled piezoelectric film (P3F) X-cut LT stack. The polarity reversal across the bond interface flips the piezoelectric stress below the neutral axis, generating a bending moment without any passive elastic layer. Moreover, the two layers share identical elastic and thermal-expansion constants, making the stack thermally matched by construction. Suspended cantilever devices were fabricated on a single 21.5 m X-cut LT stack and verified by laser Doppler vibrometry (LDV). The tapered flexural resonator exhibits a 5.5 tip velocity amplification and a measured peak tip velocity of 10.3 m/s at , and a 0.49 MHz torsional cantilever resonator with is demonstrated on the same stack. These results position bilayer P3F LT as a thermally matched, passive-layer-free actuation platform for vacuum-operated and optically interrogated MEMS such as scanners, gyroscopes, and ultrasonic transducers.
Introduction
Out-of-plane flexural and in-plane torsional modes are key building blocks of modern transducers. Flexural devices turn in-plane piezoelectric strain into out-of-plane displacement for ultrasonic transducers [1] or inertial sensors [2]. Torsional devices can provide angular motion for optical scanning mirrors [3] or rate gyroscopes [4]. In thin-film piezoelectric devices, the dominant structure of choice is the unimorph, in which an active piezoelectric film is paired with a passive elastic layer that offsets the neutral axis so that the piezoelectric stress produces a net bending moment [5].
The passive layer, however, has two main drawbacks. First, it adds mass without contributing to transduction. Second, its coefficient of thermal expansion (CTE) generally differs from that of the piezoelectric film [6], so any temperature change bends the beam. This is particularly problematic in vacuum operation, where the absence of convective cooling turns even sub-µW of absorbed heat from sources such as an optical probe beam into static bending.
Recently, P3F stacks with films of opposite polarity bonded together have replaced the passive layer in a unimorph to form a bimorph structure, as in lithium niobate (LN) piezoelectric micromachined ultrasonic transducers (PMUTs) [7, 8]. In this work, a passive-layer-free flexural architecture is proposed in P3F LT. Compared with LN, LT offers a smaller temperature coefficient of frequency (TCF) in RF acoustics [9] and better optical properties [10]. Here, we present the transduction principle of the bimorph P3F X-cut LT (X-LT) stack, paired with its CTE-matched-by-construction design validation, and transduction verification using five released devices by LDV, including four flexural resonators of different geometries and one torsional resonator.
Design and Principle
Figure 1(a) shows the side view of the stack, which consists of two 1.5 µm X-LT layers bonded with a 180° in-plane flip on high-resistivity Si, with 200 nm Au electrodes on top. The devices are driven through electrodes ( and GND) of width µm running along the beam edges over from the anchor with the device motion measured by LDV. Figure 1(b) shows three cantilever geometries covering five implemented designs, including a straight cantilever (Device A), a tapered cantilever (Device B), and tapered cantilevers on T-shaped supports (Devices C–E). The stack and geometric parameters are listed in Figure 1(c).
Figure 1. (a) Side view of the bilayer P3F stack. (b) Top view of the three geometries covering five devices (A–E). (c) Stack and geometric information. (d) P3F principle: the polarity-inverted Rot-LT layer flips the stress below the neutral axis, producing a net bending moment.
In this work, two accessible piezoelectric constants, and , were selected under lateral-field excitation in X-LT to generate in-plane normal and shear stress, respectively.
Both stress fields are uniform through the film thickness as illustrated in Figure 1(d). Figure 2(a) shows the calculated lateral-field coupling, defined as

Figure 2. (a) Lateral-field coupling vs in-plane rotation of X-LT. (b) Thermal simulation of bimorph P3F film shows CTE-matched behavior. (c) Displacement profiles of the top surface and neutral axis of the bimorph P3F film, showing linear expansion following .
with all constants rotated to the in-plane device orientation. They peak at 139° () and 172° (). However, a uniform stress carries no moment about the neutral axis, so a uniformly poled film extends rather than bends even though its full thickness is active. A unimorph restores the moment by shifting the neutral axis into a passive layer, at the cost of a partially active stack and a CTE mismatch. The bimorph instead inverts the polarity below the neutral axis. Considering at the LT-LT bonding interface, which is the neutral axis of two elastically identical layers of thickness , and neglecting the electrodes, the piezoelectric moment per unit width under a lateral field is
whereas the same integral vanishes for a uniformly poled film. An idealized unimorph of the same total thickness with an elastically identical passive layer yields only half of this moment at because only half of the stack is active. The same expression can also be applied to to give an antisymmetric in-plane shear stress that generates the twisting moment for torsional motion. Importantly, a 180° in-plane flip of X-LT about the X axis reverses the sign of the piezoelectric constants but leaves the elastic, dielectric, and thermal expansion constants unchanged. For a uniform temperature rise , the thermal moment of the bilayer thus vanishes,
where and are the in-plane Young’s modulus and CTE, identical in both layers. The stack is thus CTE-matched by construction while retaining full transduction efficiency.
PHOTOTHERMAL SIMULATION
Figure 2(b) presents the thermal FEM simulation of a 450 m long, 70 m wide cantilever based on Device A footprint with 1.5 m per layer under 1 W of absorbed power applied near the tip of the beam (0.9L). The X-LT/SiO unimorph deflects by 80 nm, while the bimorph shows no obvious bending. Figure 2(c) zooms into the residual displacement of the bimorph; the picometer-level displacement is not bending. The neutral axis remains stationary while the top surface follows the temperature-rise profile, corresponding to through-thickness thermal expansion. The linear temperature rise from the anchor to the heat spot reflects conduction along the beam as the only heat path in vacuum. Consequently, the absorbed power required to exceed a typical 2.5 m LDV depth of field rises from 32 W for the unimorph to 245 mW for the bimorph, an increase of 7,700.
FABRICATION
The LT is patterned by ICP-RIE to define the etch windows, followed by 200 nm Au electrode deposition, and the devices are released by isotropic XeF etching of the Si substrate. Figure 3 shows the laser confocal 3D map and height profile of a released device. The cantilever is flat within the confocal resolution over a 60 m deep release cavity, with no residual-stress curl, consistent with the symmetric bilayer stack observed in [11].

Figure 3. Laser confocal 3D map and height profile of the released device.
EXPERIMENTAL RESULTS
The small-signal LDV spectra of Devices A/B and C/D were shown in Figure 4 and Figure 5, respectively. The devices were driven under broadband chirp excitation in air, and the measured mode shapes correspond to the labeled resonances. The extracted quality factors () of the modes used for largesignal testing range from 443 to 634 and are summarized in Table 1. As all four devices share the same stack and electrode width, their difference in response is set by geometry. At identical drive, the tapered cantilever (Device B) raises the peak velocity by ∼1.6× over the straight beam design (Device A) whereas the T-support tapered design (Device C) raises the peak velocity by ∼2.7×. Further narrowing of the T-support base from 110 to 70 µm (Device D) boosts the peak velocity by ∼5.5×.
Figure 4. Small-signal LDV spectra and mode shapes of Devices A and B.
Figure 5. Small-signal LDV spectra and mode shapes of Devices C and D.
| Device | Freq. [MHz] | Velocity factor | [m/s] | Limited by | ||
| A | 0.42 | 569 | 1× (ref) | 2.1 | PA | |
| B | 0.38 | 484 | ∼1.6× | 4.7 | PA | |
| C | 0.73 | 443 | ∼2.7× | 6.8 | PA | |
| D | 0.43 | 634 | ∼5.5× | 10.3 | PA | |
| E | 0.49 | 514 | — | 3.3 | Laser |
Table 1. Summary of measured devices. Velocity factor is referenced to Device A at identical small-signal drive.
Figure 6 and Figure 7 show the corresponding large-signal time-domain velocity as the LDV drive signal is increased from 0.05 V to 0.6–0.8 V, amplified by a power amplifier (PA) with ∼640× gain. The maximum tested tip velocity rises from 2.1 m/s for Device A to 4.7, 6.8, and 10.3 m/s for Devices B, C, and D, respectively, with all devices surviving. The maximum velocity ceiling is currently limited by amplifier compression rather than by device failure.

Figure 6. Time-domain velocity of Devices A and B vs LDV drive signal. * Denotes amplifier compression.
Figure 7. Time-domain velocity of Devices C and D and of the torsional Device E ( excitation; inset: measured mode shape) vs LDV drive signal. * Denotes amplifier compression.
Aside from the flexural actuation demonstrated in this work, the torsional device (Device E) using rotated in-plane excitation is also presented in Figure 7. Here, the same stack supports torsional actuation at 0.49 MHz with . The velocity reaches 3.3 m/s at a drive of only 0.15 V. Unlike that of the flexural devices, the velocity of the torsional device is capped by optical dropout of the LDV signal under tilt at the edge of the cantilever.
Conclusion
A passive-layer-free actuator platform in bilayer P3F X-cut thin-film LT has been demonstrated. Reversing the polarity across the bond interface restores the bending moment that enables the bimorph actuation. We presented four flexural devices that reach a 5.5× geometric velocity amplification and up to 10.3 m/s at , and a torsional device that operates at 0.49 MHz with within the same stack. P3F LT is thus a promising multifunctional actuator platform for vacuum-operated sensing and optically interrogated MEMS.
Acknowledgements
The authors thank the DARPA NIMBUS program for funding support.
Contact
*T.-H. Hsu; [email protected]
References
- [1]R. Lu, M. Breen, A. E. Hassanien, Y. Yang, S. Gong, “A Piezoelectric Micromachined Ultrasonic Transducer Using Thin-Film Lithium Niobate”, J. Microelectromech. Syst., vol. 29, no. 6, pp. 1412–1414, 2020.DOI
- [2]N. Yazdi, F. Ayazi, K. Najafi, “Micromachined Inertial Sensors”, Proc. IEEE, vol. 86, no. 8, pp. 1640–1659, 1998.DOI
- [3]Y. Lu, K. Liu, Y. Wang, R. Nie, T. Wu, “LNOI Thin-Film Dual-Axis Resonant Micro-Mirror with e₁₆ Torsional Actuation”, in Proc. IEEE MEMS 2023, Munich, 2023, pp. 669–672.DOI
- [4]M. Hodjat-Shamami, F. Ayazi, “Eigenmode Operation of Piezoelectric Resonant Gyroscopes”, Microsyst. Nanoeng., vol. 6, no. 1, 108, 2020.
- [5]G.-L. Luo, Y. Kusano, D. A. Horsley, “Airborne Piezoelectric Micromachined Ultrasonic Transducers for Long-Range Detection”, J. Microelectromech. Syst., vol. 30, no. 1, pp. 81–89, 2021.
- [6]A. Bartatsas et al., “Residual Stresses and Clamped Thermal Expansion in LiNbO₃ and LiTaO₃ Thin Films”, Appl. Phys. Lett., vol. 101, no. 12, 122902, 2012.
- [7]V. Chulukhadze et al., “Bimorph Lithium Niobate Piezoelectric Micromachined Ultrasonic Transducers”, IEEE Trans. Ultrason., vol. 73, no. 5, pp. 574–585, 2026.DOI
- [8]X. Zhao, M. Pertijs, T. Manzaneque, “Piezoelectric Micromachined Ultrasonic Transducer (PMUT) Based on Bilayer X-Cut Lithium Niobate”, in Digest Tech. Papers Transducers’25 Conference, Orlando, June 29–July 3, 2025, pp. 530–533.DOI
- [9]T. Kimura, M. Omura, Y. Kishimoto, K. Hashimoto, “Comparative Study of Acoustic Wave Devices Using Thin Piezoelectric Plates in the 3–5-GHz Range”, IEEE Trans. Microw. Theory Techn., vol. 67, no. 3, pp. 915–921, 2019.DOI
- [10]J. Yu et al., “Tunable and Stable Micro-Ring Resonator Based on Thin-Film Lithium Tantalate”, APL Photonics, vol. 9, no. 3, 036115, 2024.DOI
- [11]B. Kim, I. Anderson, T.-H. Hsu, R. Lu, “Gradient Residual Stress in Transferred Thin-Film Lithium Niobate and Its Compensation Using Periodically Poled Piezoelectric Bilayers”, J. Microelectromech. Syst., in press, 2026.DOI



