🔬 applied physics

Gradient Residual Stress in Transferred Thin-Film Lithium Niobate and Its Compenstation Using Periodically Poled Piezoelectric Bilayers

This study experimentally characterizes the thickness- and orientation-dependent gradient residual stress in transferred thin-film lithium niobate and demonstrates that engineering periodically poled piezoelectric bilayers can effectively mitigate this stress to enhance the mechanical stability of MEMS devices.

Original authors: Byeongjin Kim, Ian Anderson, Tzu-Hsuan Hsu, Ruochen Lu

Published 2026-04-23

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Byeongjin Kim, Ian Anderson, Tzu-Hsuan Hsu, Ruochen Lu

Original paper licensed under CC BY 4.0 (http://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

Imagine you are baking a very thin, delicate layer of cake (the Lithium Niobate film) on top of a heavy, sturdy cake board (the silicon substrate).

When you pull the cake layer off the board to make a floating bridge or a tiny trampoline (a MEMS device), you expect it to lie perfectly flat. But often, it curls up like a potato chip or sags down like a tired noodle. Why? Because of Residual Stress.

This paper is about figuring out why that cake layer curls in specific, weird ways and how to fix it so we can build better, more reliable tiny machines.

Here is the breakdown of their discovery, using simple analogies:

1. The Problem: The "Curling Chip" Effect

The researchers were working with Thin-Film Lithium Niobate (TFLN), a super-useful material for making tiny sensors, radio filters, and optical devices.

When they made tiny cantilevers (like diving boards) out of this material, they noticed something strange:

Thickness matters: A 100-nanometer-thick film curled differently than a 460-nanometer-thick one.
Direction matters: If they cut the diving board at a 20-degree angle, it curled one way. If they cut it at 55 degrees, it was almost flat.

The Analogy: Imagine a long strip of metal tape. If you heat one side more than the other, it bends. In this case, the "heat" was actually the way the material cooled down after being made. Because the material is "picky" (anisotropic), it shrinks differently depending on which way you look at it. This created a Gradient Stress—meaning the top of the film was being pulled one way, and the bottom was being pushed the other, causing the whole thing to twist or curl.

2. The Discovery: Mapping the "Curl Map"

The team tested films of different thicknesses (100nm, 220nm, 460nm) and rotated them at different angles.

The Result: They found "sweet spots." For thicker films, if you cut them at 55° or 125°, they stayed flat. But for ultra-thin films (100nm), the flat spots shifted to 20° and 160°.
The Takeaway: The thinner the film, the more dramatic the curling effect became. It's like a thin piece of paper curling much more violently than a thick piece of cardboard when exposed to humidity.

They realized that simply knowing the average stress wasn't enough. They had to understand the gradient—the difference in stress from the top surface to the bottom surface.

3. The Solution: The "Opposing Team" Strategy

So, how do you stop a diving board from curling? You can't just make it thicker (that changes how the device works). You need a trick.

The researchers invented a Bilayer Sandwich.

The Setup: Instead of one layer of material, they stacked two layers on top of each other.
The Trick: They rotated the top layer so its "grain" was pointing in the opposite direction of the bottom layer.

The Analogy: Imagine two people pulling on a rope.

Single Layer: One person pulls the rope hard to the left. The rope snaps or bends wildly.
Bilayer: Now, imagine a second person standing on the other side, pulling the rope hard to the right.
The Result: The forces cancel each other out! The rope stays straight.

In their experiment, they stacked a 90nm layer on top of a 110nm layer with opposite orientations. The "curling force" from the top layer fought the "curling force" from the bottom layer.

4. The Outcome: A Flatter, Stronger Device

The result was impressive. By using this "Opposing Team" (Periodically Poled Piezoelectric Bilayer) design:

The crazy curling (gradient stress) was cut in half or more.
The tiny diving boards stayed much flatter and more stable.
Crucially, they didn't lose the material's special electrical powers. The "team" still worked together to do the job (like moving radio waves or sensing motion), but they stopped fighting each other physically.

Why Does This Matter?

Think of TFLN as the "super-material" of the future for tiny electronics. But right now, it's like trying to build a skyscraper out of a material that wants to twist itself into a pretzel.

This paper provides the blueprint to fix that twist. By understanding exactly how the material curls based on its thickness and angle, and by stacking it in a "cancel-out" sandwich, engineers can now build:

More reliable sensors for smartphones.
Better medical devices.
Faster, more efficient communication chips.

In a nutshell: They figured out why their tiny films were curling up like potato chips, mapped out exactly when and why it happens, and then built a "double-layer" sandwich that cancels out the curl, making the films flat and ready for real-world use.

1. Problem Statement

Residual stress is a critical factor affecting the mechanical reliability and performance of thin-film micro- and nano-electromechanical systems (MEMS/NEMS). While previous research on Transferred Thin-Film Lithium Niobate (TFLN) has primarily focused on mean residual stress ( $\sigma_0$ )—often attributed to thermal expansion mismatch between the film and substrate—this work addresses a less understood but equally damaging phenomenon: gradient residual stress ( $\sigma_1$ ).

The Issue: $\sigma_1$ represents a linear stress gradient through the film thickness. Unlike uniform stress, which causes rigid-body rotation, gradient stress induces beam curvature (bending), leading to structural deformation, buckling, or collapse in suspended TFLN structures.
The Gap: Practical strategies to actively manage or cancel this gradient stress in TFLN MEMS had not been demonstrated, and experimental observations of its orientation and thickness dependence lacked systematic validation through physics-based simulations.

2. Methodology

The authors employed a combined experimental and simulation approach to characterize and mitigate $\sigma_1$ in 128° Y-cut TFLN films.

Fabrication:
- Single-Layer Films: Cantilever arrays were fabricated from TFLN films with thicknesses of 100 nm, 220 nm, and 460 nm.
- Bilayer Films: A Periodically Poled Piezoelectric Film (P3F) bilayer structure was created using two LN layers with opposite crystallographic orientations (90 nm top / 110 nm bottom).
- Orientation Mapping: Cantilevers were patterned with in-plane orientations ranging from 0° to 175° (in 5° increments) to map the anisotropic stress behavior.
- Process: The devices were released using a dry etching process (XeF₂) to remove the underlying silicon substrate without damaging the TFLN.
Measurement & Extraction:
- Curvature Analysis: Released cantilevers were measured using a laser confocal optical profilometer.
- Stress Decomposition: The out-of-plane deflection profiles were fitted to a quadratic model to separate the mean stress ( $\sigma_0$ ) and the gradient stress ( $\sigma_1$ ).
- Anisotropic Modeling: Because Lithium Niobate is anisotropic, the Young's modulus ( $E$ ) was treated as a directional modulus ( $E(\theta) = 1/S'_{11}(\theta)$ ) to accurately calculate stress from curvature ( $R$ ) using the relation $\sigma_1 = E h^2 / (2R)$ .
- Normalization: To compare films of different thicknesses, $\sigma_1$ was normalized by film thickness (units: MPa/nm).
Simulation:
- Finite Element Analysis (FEA) using COMSOL was performed to validate the experimental extraction method. The simulations applied the extracted $\sigma_1$ as an initial stress condition to predict beam deformation.

3. Key Contributions

Characterization of Gradient Stress Anisotropy: The paper establishes that $\sigma_1$ in TFLN is not uniform but exhibits a strong dependence on both crystallographic orientation and film thickness.
Validation of Extraction Method: The authors demonstrate that curvature-based extraction is a robust method for measuring $\sigma_1$ in anisotropic TFLN, showing excellent agreement between experimental data and FEA simulations.
Stress Compensation Strategy: The paper introduces and validates a bilayer P3F design (opposite crystallographic orientations) as an effective mechanism to actively cancel gradient stress, significantly reducing structural deformation without compromising piezoelectric performance.

4. Key Results

Thickness and Orientation Dependence:
- Stress-Free Orientations: For thicker films (220–460 nm), stress-free orientations (where $\sigma_1 \approx 0$ ) occur at ~55° and ~125°. For the ultra-thin 100 nm film, these shift to ~20° and ~160°.
- Magnitude of $\sigma_1$ : The normalized gradient stress increases significantly as the film becomes thinner:
  - 460 nm: −0.12 to 0.08 MPa/nm
  - 220 nm: −0.8 to 0.34 MPa/nm
  - 100 nm: −0.1 to 3.4 MPa/nm (indicating a much stronger through-thickness asymmetry).
- Physical Origin: The gradient stress is attributed to anisotropic thermal expansion during cooldown and asymmetric stress relaxation between the top and bottom of the film, rather than purely elastic effects.
Bilayer Compensation (P3F):
- A 90/110 nm bilayer with opposite orientations successfully reduced the equivalent normalized $\sigma_1$ from the single-layer range of −0.8 to 0.34 MPa/nm down to −0.4 to −0.04 MPa/nm.
- This resulted in significantly reduced cantilever curvature.
- Residual Asymmetry: The bilayer did not achieve perfect cancellation (remaining slightly concave-up) due to the thickness imbalance (110 nm bottom vs. 90 nm top), where the thicker bottom layer contributed a dominant compressive component.
Simulation Validation:
- FEA simulations matched experimental beam tip heights and curvature profiles with high fidelity.
- Simulations confirmed that the observed deformation is driven by $\sigma_1$ , while $\sigma_0$ (mean stress) primarily causes rigid-body rotation and is highly geometry-dependent, making it difficult to extract reliably via curvature alone.

5. Significance and Impact

Mechanical Stability: This work provides a critical pathway to achieving mechanically stable, suspended TFLN MEMS devices. By understanding and engineering $\sigma_1$ , designers can prevent failure modes like buckling and collapse.
Scalability: The demonstrated bilayer compensation strategy allows for the fabrication of high-performance devices across various thicknesses, which is essential for scaling TFLN technology.
Dual Benefit: The P3F bilayer approach offers a unique advantage: it mitigates mechanical stress without degrading piezoelectric performance. The alternating orientations preserve constructive electromechanical coupling (maintaining $k^2$ ) while suppressing stress-induced deformation.
Design Framework: The study establishes "gradient stress engineering" through orientation, thickness, and layer configuration as a fundamental design parameter for next-generation RF resonators, filters, and photonic circuits based on TFLN.