1. Introduction

Piezoelectric MEMS operating at elevated temperatures are increasingly demanded in applications where direct electrical sensing or actuation must be performed under extreme thermal conditions, including energy conversion systems, oil and gas processing, automotive engines, and aerospace propulsion. In such environments, device temperatures frequently exceed 700 °C, a regime that challenges the material limits of conventional silicon-based MEMS technologies.

The operational temperature of traditional MEMS is often restricted by degradation of structural materials, metallization failure, and stress induced by thermal expansion coefficient (CTE) mismatch between functional layers and the supporting substrate. While hybrid MEMS–fiber systems have demonstrated operation beyond 1000 °C, their complexity and lack of scalability limit their suitability for compact, integrated sensor platforms.

Lithium niobate (LN) offers several advantages for high-temperature piezoelectric applications, including a high Curie temperature (~1200 °C), strong piezoelectric coupling, and excellent electro-optic and acousto-optic properties. In particular, stoichiometric lithium niobate (SLN) exhibits superior thermal stability compared to congruent lithium niobate (CLN), which suffers from lithium vacancies and defect-driven degradation above approximately 300 °C. Although high-temperature LN-based surface acoustic wave (SAW) devices on bulk substrates have been widely studied, the thermal survivability of suspended thin-film LN platforms—which enable bulk acoustic wave (BAW) and Lamb-wave devices—remains insufficiently explored.

Suspended MEMS structures offer enhanced electromechanical coupling and acoustic confinement but are inherently more vulnerable to thermomechanical stress, fracture, and collapse under extreme conditions. Understanding their thermal limits is therefore essential for the development of reliable high-temperature MEMS.

2. Device Design and Fabrication

The devices investigated in this work are suspended thin-film LN acoustic resonators designed to support symmetric Lamb wave modes. The resonators are fabricated on a multilayer stack consisting of a high-resistivity silicon substrate, a sacrificial amorphous silicon layer, and a 600 nm thick X-cut stoichiometric LN film. X-cut LN is selected due to its widespread use in MEMS and photonic systems and its favorable electromechanical properties.

Platinum is employed as the electrode material because of its high melting point and chemical stability at elevated temperatures. A thin titanium adhesion layer is introduced between the LN and Pt to improve adhesion and mitigate metal delamination during thermal cycling. The resonator geometries include variations in in-plane rotation angle, anchor configuration, and interdigital electrode layout in order to avoid biasing the thermal endurance results toward a single design.

In addition to functional resonators, serpentine metal resistors are co-fabricated on the same substrate using identical metallization. These structures enable direct monitoring of metal resistivity as a function of annealing temperature, providing insight into metallization degradation and its impact on device performance.

3. Experimental Methodology

Thermal endurance is evaluated using a stepwise annealing and characterization protocol. Annealing is performed under vacuum conditions to minimize oxidation, with controlled heating and cooling rates to suppress pyroelectric effects in LN. The initial annealing temperature is set to 250 °C, followed by successive cycles with temperature increments of 50 °C. Each annealing step is held at the target temperature for 10 hours, except for the highest temperatures, where furnace limitations require shorter dwell times.

After each annealing cycle, devices are characterized using optical microscopy to assess structural integrity, four-point probe measurements to evaluate metal resistivity, radio-frequency (RF) electrical measurements to extract resonance frequency and quality factor (Q), and X-ray diffraction (XRD) to examine crystalline quality and strain evolution.

4. Results and Discussion

4.1 Structural Evolution

Optical inspection reveals minimal visible changes in the suspended LN membranes up to approximately 400 °C. Beyond 500 °C, stress-induced cracking begins to appear within the suspended regions, although most devices remain mechanically intact and functional. Up to 550 °C, cracks generally do not propagate to the anchors or cause catastrophic collapse.

Severe structural degradation occurs between 600 °C and 750 °C. In this temperature range, increased cracking, membrane warping, LN delamination, and anchor fracture are observed. At approximately 700 °C, cracks preferentially form along crystallographic directions associated with high in-plane CTE and low cleavage energy. This behavior is attributed to the large CTE mismatch between LN and the silicon substrate, combined with the intrinsic anisotropy of X-cut LN.

At 800 °C, extensive metallization damage and anchor failure render the resonators non-functional.

4.2 Metallization Degradation

Metal resistivity measurements indicate an initial decrease in resistivity after the first annealing cycle, likely due to grain growth and defect annealing in the Pt film. However, at higher temperatures, resistivity increases significantly, signaling the formation of voids, hillocks, and discontinuities in the metal layer.

Above 650 °C, Pt films exhibit pronounced degradation, including pore formation and partial loss of electrical continuity. This degradation directly contributes to increased electrical losses and eventual device failure, even when the LN membrane remains partially intact.

4.3 Acoustic Performance

RF measurements show that resonant frequencies gradually decrease with increasing annealing temperature, consistent with thermally induced stress relaxation and changes in effective elastic constants. Interestingly, the quality factor of several resonant modes increases after high-temperature annealing, particularly above 700 °C. This improvement is attributed to stress redistribution and reduced acoustic energy leakage in partially cracked or stress-relieved structures.

Despite these localized performance enhancements, overall device operability declines sharply beyond 750 °C due to metallization failure and anchor breakage.

5. Failure Mechanisms

The dominant failure mechanisms identified in this study include:

1. Thermal expansion mismatch between LN, metal electrodes, and the silicon substrate, leading to stress accumulation and cracking.

2. Crystallographic cleavage of LN, particularly along planes with low fracture energy under high thermal stress.

3. Metallization instability, including grain coarsening, void formation, and loss of conductivity in Pt films.

4. Anchor degradation, which compromises mechanical support and electrical continuity.

These mechanisms act synergistically to define the ultimate thermal limit of suspended thin-film LN MEMS.

6. Conclusions

This work demonstrates that suspended thin-film lithium niobate acoustic resonators can withstand annealing temperatures up to 750 °C, representing one of the highest verified thermal endurance limits for purely MEMS-based piezoelectric platforms. Although significant degradation occurs at elevated temperatures, device survival and partial functionality at such extreme conditions highlight the robustness of stoichiometric LN for high-temperature MEMS applications.

The insights gained from this study provide practical guidelines for material selection, metallization design, and structural optimization aimed at extending the operational temperature range of suspended LN devices. These findings open pathways for deploying LN-based MEMS in harsh environments and for advancing high-temperature photonic, electro-optic, and acousto-optic systems.