⚡ electrical engineering

Nonlinear Characterization of Thin-Film LiNbO3 Acoustic Filters

This paper introduces a new nonlinear characterization methodology for high-frequency LiNbO3 acoustic filters and demonstrates that devices fabricated on sapphire substrates outperform those on silicon at mmWave frequencies by achieving lower insertion loss and superior thermal stability with reduced passband distortion.

Original authors: Omar Barrera, Bryan T. Bosworth, Taran Anusorn, Kenny Huynh, Ian Anderson, Nicholas R. Jungwirth, Michael Liao, Sinwoo Cho, Jack Kramer, Lezli Matto, Mark S. Goorsky, Nathan D. Orloff, Ruochen Lu

Published 2026-04-14

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Omar Barrera, Bryan T. Bosworth, Taran Anusorn, Kenny Huynh, Ian Anderson, Nicholas R. Jungwirth, Michael Liao, Sinwoo Cho, Jack Kramer, Lezli Matto, Mark S. Goorsky, Nathan D. Orloff, 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 trying to listen to a very faint radio station while standing next to a roaring jet engine. To hear the music clearly, you need a filter that blocks the jet noise but lets the music through. In the world of smartphones and 5G/6G networks, these "filters" are tiny acoustic devices that sort out signals so your phone doesn't get confused by the chaos of the airwaves.

This paper is about building a super-fine, high-speed filter for the next generation of wireless technology (millimeter waves), and figuring out why some materials work better than others when the device gets hot and stressed.

Here is the breakdown of their work using simple analogies:

1. The Problem: The "Hot Mess" of High Frequencies

As we move to faster internet (6G), the signals get higher in frequency. To catch these signals, engineers use tiny acoustic resonators (think of them as microscopic tuning forks that vibrate to specific notes).

The Challenge: When you push a lot of power through these tiny tuning forks, they get hot.
The Analogy: Imagine a rubber band. If you stretch it gently, it snaps back perfectly. But if you stretch it too hard and fast, it heats up, gets floppy, and changes shape. In our filters, this "heating up" causes the signal to drift off-key (frequency shift) and get distorted. This is called nonlinearity.

2. The Solution: Two Different "Beds" for the Tuning Fork

The researchers built these filters using a special material called Lithium Niobate (a piezoelectric crystal that turns electricity into vibration). They placed this crystal on top of two different "beds" (substrates) to see which one kept the filter cool and stable:

Bed A: Sapphire (Al2O3) – Think of this as a high-tech cooling mat. It conducts heat away well.
Bed B: Silicon (Si) – Think of this as a thick, insulating foam pad. It traps heat.

They also used a "sacrificial layer" (like a temporary glue) to lift the filter slightly off the bed, creating an air gap. This is like suspending the tuning fork in a vacuum so it can vibrate freely without touching the table.

3. The Experiment: The "Stress Test"

The team built two sets of filters: one on the Sapphire bed and one on the Silicon bed. They then cranked up the power to see what happened.

The Heat Test: When they turned up the volume (power), the filter on the Silicon bed got very hot very fast. Because the air gap was deeper and the silicon trapped heat, the filter's "note" drifted significantly. It was like the tuning fork getting so hot it started singing a different song.
The Winner: The filter on the Sapphire bed stayed cool. The heat escaped efficiently, so the filter stayed on pitch even at high power.

The Result: The Sapphire filter was much more stable. It didn't lose signal strength, and its frequency didn't wander.

4. The "Traffic Jam" Test (Intermodulation)

In radio, if you mix two signals together, they can create unwanted "ghost" signals (intermodulation) that jam the channel.

The Analogy: Imagine two people talking at once. If the room is quiet, you hear them clearly. If the room is loud and chaotic (nonlinear), their voices might blend into a weird, unintelligible noise.
The Finding: The Sapphire filter was excellent at keeping the voices clear. It had a "linearity score" (IIP3) of 50.8 dBm, which is incredibly high. The Silicon filter was good too (46.5 dBm), but the Sapphire one was the clear champion.

5. Why This Matters

Speed: These filters work at 21.8 GHz, which is in the "millimeter-wave" range needed for super-fast 6G internet.
Size: They are tiny (smaller than a grain of sand) but handle huge amounts of data.
Reliability: By choosing the right "bed" (Sapphire), engineers can make sure that when your phone is transmitting at full power (like when you are uploading a video), the filter doesn't overheat and ruin your connection.

The Bottom Line

The researchers proved that to build the next generation of wireless filters, you can't just focus on the material itself; you have to look at how the heat escapes.

By placing their high-performance Lithium Niobate filter on a Sapphire substrate instead of Silicon, they created a device that stays cool, stays on pitch, and handles high power without distortion. It's like upgrading from a flimsy tent in a storm to a sturdy, well-ventilated cabin—the signal gets through loud and clear, no matter how hard the wind blows.

1. Problem Statement

As wireless communication systems evolve toward 5G and future 6G standards, there is an urgent demand for compact, high-performance acoustic filters operating in the millimeter-wave (mmWave) spectrum (above 20 GHz). While Lithium Niobate (LiNbO3) based on first-order antisymmetric (A1) mode laterally excited bulk acoustic resonators (XBARs) has demonstrated superior linear performance at these frequencies, nonlinear characterization remains a critical gap.

At high power levels, acoustic filters suffer from:

Self-heating: Elevated input power raises the resonator temperature. Due to LiNbO3's low thermal conductivity (5.8 W/m·K) and high temperature coefficient of frequency (TCF ≈ -80 ppm/K), this causes significant frequency drift and insertion loss (IL) degradation.
Intermodulation Distortion: Nonlinear mixing generates third-order intermodulation products (IMD3), which can desensitize RF front-ends and compromise signal integrity.
Substrate Dependency: The thermal management of suspended thin-film structures is heavily dependent on the underlying substrate, yet few studies have systematically compared how different substrates affect nonlinear performance at mmWave frequencies.

2. Methodology

The authors developed a comprehensive experimental framework to characterize both linear and nonlinear behaviors of LiNbO3 XBAR filters on two distinct substrate stacks:

LiNbO3-aSi-Al2O3: LiNbO3 film on a sapphire substrate with an amorphous silicon (aSi) sacrificial layer.
LiNbO3-aSi-Si: LiNbO3 film on a silicon substrate with an aSi sacrificial layer.

Key Methodological Steps:

Device Design: Third-order ladder filters were designed using series and shunt resonators with local thickness trimming (75 nm for series, 90 nm for shunt) to achieve a center frequency near 22 GHz.
Thermal Simulation: Finite Element Analysis (FEA) using COMSOL was performed to model heat dissipation. The study compared thermal conductance ( $G$ ) and thermal time constants ( $\tau$ ) between the two stacks, noting that the Si stack creates a larger air gap (35 µm vs. 1 µm) due to isotropic etching, leading to higher thermal resistance.
Material Characterization: High-Resolution X-ray Diffraction (HRXRD) and Transmission Electron Microscopy (TEM) were used to assess crystal quality (mosaicity) and Coefficient of Thermal Expansion (CTE) matching.
Custom Nonlinear Measurement System: A specialized setup was built using a 67 GHz Vector Network Analyzer (VNA) with four internal sources. It utilized low-noise amplifiers, power combiners, and directional couplers to perform:
- Power-dependent S-parameters: Sweeping input power from -9 dBm to +17 dBm to measure IL and frequency drift.
- IMD3 Measurements: Two-tone testing to calculate the Third-Order Input Intercept Point (IIP3).
Calibration: A rigorous two-tier calibration (Coaxial and on-wafer mTRL) was employed to ensure accurate power measurements at the probe tip reference plane.

3. Key Contributions

Novel Nonlinear Methodology: The paper introduces a robust methodology for performing power-dependent S-parameter and IMD3 measurements on mmWave acoustic filters, addressing a significant lack of data in the field.
Substrate Comparison: It provides the first direct comparative analysis of LiNbO3 XBAR filters on Sapphire (Al2O3) vs. Silicon (Si) substrates, isolating the impact of thermal management on nonlinearity.
Thermal-Nonlinearity Correlation: The study quantitatively links thermal resistance (derived from FEA) to observed nonlinear metrics (frequency drift and IIP3), demonstrating that substrate selection is a pivotal factor in mitigating self-heating.
High-Frequency Performance: It demonstrates the feasibility of LiNbO3 XBAR filters operating well above 20 GHz with wide fractional bandwidths.

4. Key Results

Linear Performance:

LiNbO3-aSi-Al2O3: Center frequency ( $f_c$ ) = 21.8 GHz, Insertion Loss (IL) = 1.48 dB, 3 dB Fractional Bandwidth (FBW) = 17.7%.
LiNbO3-aSi-Si: $f_c$ = 21.6 GHz, IL = 2.47 dB, FBW = 18.6%.
Both platforms showed excellent linear performance, with the Al2O3 platform exhibiting slightly lower loss, likely due to better crystal quality (narrower HRXRD rocking curves) and CTE matching.

Nonlinear Performance (Thermal Stability):

Frequency Drift ( $\Delta f$ ): Under high power, the Si-based filter showed a massive drift of 125 MHz (5,680 ppm), corresponding to a temperature rise of ~70 K. In contrast, the Al2O3-based filter drifted only 29 MHz (1,320 ppm), corresponding to a ~17 K rise.
Insertion Loss Degradation ( $\Delta IL$ ): The Si filter degraded significantly faster than the Al2O3 filter as power increased beyond +1 dBm.
Thermal Analysis: The Al2O3 stack had a thermal conductance ( $G$ ) of 270 µW/K, significantly higher than the Si stack's 109 µW/K, confirming superior heat dissipation.

Linearity (IMD3):

IIP3 (In-band): The Al2O3 filter achieved an exceptional IIP3 of 50.8 dBm. The Si filter also performed well at 46.5 dBm.
The Al2O3 platform consistently outperformed the Si platform across the frequency range, confirming that better thermal management directly translates to higher linearity.

5. Significance

This work establishes thin-film LiNbO3 on Sapphire as a superior platform for next-generation mmWave acoustic filters. The findings highlight that:

Thermal Management is Critical: For high-power mmWave applications, the choice of substrate is not just about mechanical support but is the primary determinant of nonlinear performance. The Al2O3 substrate's superior thermal conductivity and CTE matching minimize self-heating, preserving filter linearity.
Scalability: The technology successfully scales acoustic filters to >20 GHz with wide bandwidths (~18%) and low loss, making it a viable candidate for 5G/6G front-end modules.
High Linearity: Achieving IIP3 values near 50 dBm at mmWave frequencies is a significant milestone, suggesting these filters can handle the high dynamic range requirements of modern communication systems without signal distortion.

In conclusion, the paper validates that with proper substrate engineering (specifically using Sapphire), LiNbO3 XBARs can overcome thermal limitations to become the leading technology for compact, high-linearity mmWave filtering.