⚡ electrical engineering

Wafer-bonded symmetric four-terminal PZT/insulator/PZT MEMS bimorph actuator with near-ideal displacement superposition

This paper presents a wafer-bonded, silicon-shim-free four-terminal PZT/insulator/PZT MEMS bimorph actuator that achieves near-ideal displacement superposition and significantly enhanced tip displacement through symmetric active-active actuation without shared internal conductive nodes.

Original authors: Yukio Suzuki, Xuchen Wang, Shuji Tanaka

Published 2026-08-26

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Yukio Suzuki, Xuchen Wang, Shuji Tanaka

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Tiny machines that move with electricity are becoming essential tools for modern technology, from the lenses in smartphone cameras to the sensors that guide medical instruments. At the heart of many of these devices are actuators, components that convert electrical signals into physical motion. For decades, engineers have relied on a specific type of material called lead zirconate titanate, or PZT, to create this movement. When electricity is applied to a thin layer of this material, it stretches or shrinks, and if that layer is attached to a passive backing, the whole structure bends. This bending action is the engine behind countless micro-machines. However, there is a limit to how much force these single-layer designs can generate before they become too stiff or unstable, prompting researchers to look for ways to make them stronger and more efficient without making them larger.

A team of researchers at Tohoku University in Japan has developed a new way to build these tiny motors that doubles the power of the movement while keeping the device perfectly balanced. Instead of using a single active layer glued to a passive one, they created a device where two active layers work together, sandwiching a thin insulating sheet in the middle. They achieved this by taking two separate wafers, each already containing a working motor, and bonding them face-to-face. This process creates a symmetrical structure where both the top and bottom layers are active and can be controlled independently. The result is a device that bends with nearly double the force of a standard design, offering a significant leap in performance for microscopic machinery.

The researchers began by designing a structure that looked like a tiny diving board, but instead of a single layer of material, it consisted of two layers of the active PZT material separated by a very thin insulating barrier made of silicon nitride and silicon dioxide. This barrier is only 1.4 micrometers thick, yet it is strong enough to hold the two layers together while keeping their electrical connections completely separate. By bonding two pre-made wafers together, the team ensured that the top and bottom layers were perfectly mirrored, a feat that is difficult to achieve when building layers one by one on top of each other. This symmetry is crucial because it balances the internal stresses that often cause these tiny devices to warp or break before they are even turned on.

To test how well this new design worked, the team simulated the behavior of the device on a computer before building it. The simulations predicted that when they applied voltage to both layers in opposite directions, the tip of the tiny beam would move 92.2 micrometers. This is exactly twice the movement produced by activating only the top layer, and it is about 80 percent more movement than the best single-layer designs currently available. The computer models suggested that by using two active layers instead of one active and one passive, the device could generate much more bending force without needing a thicker or heavier structure.

The team then moved from the computer screen to the laboratory to build the actual device. They started with two silicon wafers, each coated with the necessary layers of metal and PZT. Using a specialized process, they cleaned the surfaces and brought the two wafers together, allowing them to bond directly without any glue. This face-to-face bonding created a single, solid unit where the two active layers were mechanically linked but electrically isolated. After the bonding was complete, they carefully removed the bulk silicon backing from the wafers, leaving behind a free-standing cantilever that was only 1.5 millimeters long. When they examined the finished device under a microscope, they found that the bonding was nearly perfect, with only a few tiny, isolated defects and no large areas where the layers had separated.

The most striking result came when they tested how the device moved. They applied a small electrical signal to the top layer and then to the bottom layer, measuring how far the tip of the beam moved in each case. The two layers performed almost identically, confirming that the bonding process had created a perfectly balanced structure. When they activated both layers at the same time with opposing signals, the movement added up almost perfectly. At a voltage of 5 volts, the device moved 23.1 micrometers, which was 96 to 99 percent of the sum of the movements from the individual layers. This near-perfect addition of movement proves that the two active layers are working together in harmony, pushing the beam in the same direction without canceling each other out or getting in the way.

The researchers also compared their new device to other types of piezoelectric actuators to see how efficient it was at converting electrical strain into physical bending. They calculated a specific factor that measures how much curvature is generated for a given amount of strain. Their new design achieved a value that was 2.3 times higher than the best single-layer devices they compared it against. While some older designs with thinner layers showed slightly higher efficiency numbers, those designs often lacked the symmetry and independent control of the new device. The new actuator offers a unique combination of high movement, structural balance, and the ability to control each side independently, which allows for more complex and precise movements.

This work demonstrates that it is possible to build microscopic motors that are stronger and more versatile than previous generations by simply changing how the layers are assembled. By bonding two complete motor units together, the researchers created a device that is not only more powerful but also more stable and easier to control. The ability to move the tip of the beam by nearly 23 micrometers with a small voltage suggests that these devices could be used in applications requiring precise positioning, such as adjusting the focus of tiny optical lenses or moving micro-sensors in medical tools. The success of this face-to-face bonding technique opens the door to creating even more complex and capable micro-machines in the future, proving that sometimes the best way to make something stronger is to build it in two perfectly matched halves.

Technical Summary: Wafer-Bonded Symmetric Four-Terminal PZT/Insulator/PZT MEMS Bimorph Actuator

Problem Statement
Conventional thin-film lead zirconate titanate (PZT) MEMS actuators typically utilize a unimorph configuration, where a single active PZT layer is bonded to a passive elastic layer (e.g., silicon). While this converts in-plane piezoelectric strain into out-of-plane bending, it suffers from two primary limitations: (1) sensitivity to residual-stress-induced initial curvature due to the asymmetric stack, and (2) limited displacement efficiency because the passive layer adds mechanical stiffness without contributing active strain. Although bimorph configurations using two active layers can balance residual stress and maximize displacement, existing thin-film PZT bimorphs often employ a three-terminal architecture with a shared internal electrode. This shared node prevents galvanic isolation between the two actuation units, limiting control flexibility. Furthermore, sequentially deposited multilayer stacks often lack mechanical symmetry due to differing thermal histories and polarization orientations of the films.

Methodology
The authors propose and fabricate a "true-bimorph" actuator: a silicon-shim-free, mechanically symmetric four-terminal PZT/insulator/PZT laminate. The core methodology involves the face-to-face direct wafer bonding of two independently fabricated PZT-stacked substrates.

Device Architecture: The device consists of two 5-µm-thick PZT layers, each sandwiched between Pt electrodes, bonded via a 1.4-µm-thick insulating interlayer (SiN/SiO₂). This creates a nominally symmetric active–active laminate with separate electrode pairs for the top and bottom units, ensuring no shared internal conductive node.
Fabrication Process: Two SOI substrates were processed to form PZT stacks. A radical-assisted sputtered (RAS) SiN/SiO₂ interlayer was deposited and planarized via chemical mechanical polishing (CMP) to achieve a surface roughness ( $R_a$ ) of <0.2 nm. The two wafers were plasma-activated and directly bonded, followed by annealing at 400°C. The handle silicon layers were subsequently removed via dry etching to release the silicon-shim-free cantilever.
Actuation Scheme: The device utilizes complementary differential driving. A DC bias ( $V_{DC}$ ) is applied to both layers, with a time-dependent signal ( $v_{sig}$ ) added to the top layer and subtracted from the bottom layer ( $V_{top} = V_{DC} + v_{sig}$ , $V_{bottom} = V_{DC} - v_{sig}$ ). This ensures both layers remain in their polarization direction while generating additive bending moments.
Analysis: The performance was evaluated using finite-element analysis (FEM) and experimental measurements (laser Doppler vibrometry) on a 1.5-mm-long cantilever. A structural strain-to-curvature factor ( $G_{str}$ ) was defined to benchmark performance against other piezoelectric MEMS actuators on a normalized basis.

Key Results

Displacement Superposition: Under complementary actuation at 5 Vpp per layer (off-resonant), the fabricated device achieved a zero-to-peak tip displacement of 23.1 µm. Across the 1–5 Vpp range, the bimorph displacement reached 96.0%–99.4% of the sum of the independently measured top-only and bottom-only responses, demonstrating near-ideal linear superposition.
Comparison with Unimorphs: The true-bimorph provided a displacement approximately 80% larger than the maximum-displacement unimorph (with a 3-µm Si passive layer) and 2.42 times that of a total-thickness-matched unimorph (with a 6.4-µm Si layer) under the same electric field conditions.
Structural Efficiency: The extracted structural strain-to-curvature factor ( $G_{str}$ ) was $18.6 \times 10^4 \text{ m}^{-1}$ . This value is approximately 2.3 times higher than the highest value reported among the compared unimorphs, though it remains lower than some thinner three-terminal bimorphs due to the increased flexural rigidity of the thicker 5-µm PZT layers.
Initial Deformation: The released cantilever exhibited a minimal initial tip deformation of approximately 2 µm (0.13% of its length), indicating effective residual-stress balancing achieved through the symmetric bonding architecture.
Bonding Quality: Scanning acoustic microscopy confirmed uniform bonding over the main device region with only isolated defects and no large-area delamination.

Significance and Claims
The paper claims the first experimental demonstration of a mechanically symmetric, four-terminal thin-film PZT MEMS bimorph formed by face-to-face direct wafer bonding of two independently fabricated stacks. The significance of this work lies in:

Electrical Isolation: The four-terminal design provides galvanic isolation between the two active layers, enabling flexible biasing and feedback control strategies (e.g., bidirectional actuation from a zero position) that are difficult to achieve with shared-node three-terminal devices.
Additive Actuation: The results confirm that the bending contributions of the two active layers add in the same rotational direction under complementary drive, achieving near-ideal displacement superposition without the need for a passive silicon shim.
Residual Stress Management: The symmetric face-to-face bonding approach effectively balances residual stress, resulting in a flat initial state, which is critical for precision micro-positioning applications.

The authors conclude that while the absolute $G_{str}$ is limited by the device's total thickness and flexural rigidity, the architecture successfully combines large absolute displacement, symmetric dual-active-layer actuation, and separate electrical access, offering a robust platform for high-performance MEMS actuators.

More like this