Temperature Gradient Method for Alleviating Bonding-Induced Warpage in a High-Precision Capacitive MEMS Accelerometer

Dandan Liu, Huafeng Liu, Jin Quan Liu, Fangjing Hu, Ji Fan, Wen-Jie Wu, Liang Cheng TuSensors 2020 · DOI

TL;DR

This paper introduces a novel bonding technique for high-precision capacitive MEMS accelerometers that significantly reduces warpage caused by thermal expansion mismatches. By applying a temperature gradient during bonding, they achieve near-zero warpage, leading to a 1.5x improvement in interelectrode spacing control and a 2x reduction in self-noise, making these accelerometers among the most sensitive globally.

What Problem Does This Paper Address?

High-precision MEMS accelerometers are crucial for applications like disaster monitoring and inertial navigation. A key component in these devices is the capacitive transducer, which relies on precise control of the spacing between electrodes. However, a common manufacturing step, high-temperature bonding of different materials (like glass and silicon), introduces stress and warpage due to their differing coefficients of thermal expansion (CTE). This warpage directly hinders the precise control of interelectrode spacing, negatively impacting the accelerometer's sensitivity and potentially damaging the device. Existing methods to mitigate this issue are often complex, enlarge the device size, or limit material choices.

Methods

The authors propose a new bonding strategy that utilizes a temperature gradient during the bonding process. Instead of applying a uniform high temperature to both the glass top cap and the silicon spring-mass system, they apply different temperatures to each. The specific temperature difference ($\Delta T$) is calculated based on the CTE difference between glass and silicon to achieve a balanced shrinkage, theoretically eliminating warpage and stress.

The process involves:

  • Fabrication: Standard microfabrication techniques for silicon and glass slices, including UBM (Under-Bump Metal) deposition and electroplating of Tin for bonding.
  • Bonding: Reflow soldering of the glass and silicon slices using a FINEPLACER® sigma, which allows for real-time temperature monitoring of both components.
  • Simulation: Finite Element Analysis (FEA) using COMSOL Multiphysics to model the thermal stress and warpage. This involved a two-step process: thermal analysis followed by static analysis.
  • Experimental Validation: Warpage measurement using a white light scanning profiler. Bonding strength was tested using a multi-function push-pull tester. Self-noise calibration was performed by comparing the device's output to a commercial seismometer in a controlled environment.

Datasets

Not discussed in the paper. The study relies on experimental measurements and simulations of fabricated MEMS devices.

Key Insights and Ideas

  • CTE Mismatch is the Culprit: The fundamental issue is the difference in how glass and silicon expand and contract with temperature changes, leading to internal stress and deformation.
  • Temperature Gradient as a Solution: Applying a controlled temperature difference during bonding can counteract the differential shrinkage, effectively "flattening" the bonded structure. This is a clever way to leverage the material properties rather than fighting them.
  • Interelectrode Spacing is Paramount: The sensitivity of capacitive accelerometers is directly proportional to how closely the electrodes can be placed. Reducing warpage is critical for achieving this precision.
  • Low-Cost and Simple: The proposed method doesn't require exotic materials or complex additional fabrication steps, making it highly attractive for industrial adoption.
  • Noise Reduction is a Direct Benefit: By improving interelectrode spacing control, the signal-to-noise ratio of the accelerometer is significantly enhanced.

Key Results

  • Quasi-Zero Warpage: Experimentally, the proposed bonding method achieved a warpage close to zero, as demonstrated by the topography measurements.
  • Improved Interelectrode Spacing: The interelectrode spacing was controlled to 8 µm, which is approximately two-thirds of that achieved with uniform-temperature bonding.
  • Ultra-Low Self-Noise: The MEMS accelerometer fabricated using this method achieved an ultra-low self-noise of 6 ng/√Hz @ 0.07 Hz, which is two times lower than accelerometers bonded using a traditional uniform-temperature process.
  • Bonding Strength: The bonding strength met MIL-STD-883E standards, indicating the reliability of the process.
  • Simulation Accuracy: Experimental, calculation, and simulation results showed a consistent trend in warpage reduction with temperature difference, with errors within 23%.

Limitations and Future Work

  • Plastic Deformation: The authors note that plastic deformation, which was not considered in their simplified elastic simulation, might explain discrepancies between experimental and simulated results when warpage is large. This suggests that for extreme cases, a more advanced simulation might be needed.
  • Future Applications: The authors highlight the broad applicability of this bonding technology beyond accelerometers, including forming multi-layer structures, providing mechanical support, and isolating ambient disturbances in various MEMS devices and integrated circuits. This suggests future work could explore these diverse applications.

Summary updated

BibTeX
@article{liu2020temperature,
  author = {Dandan Liu and Huafeng Liu and Jin Quan Liu and Fangjing Hu and Ji Fan and Wen-Jie Wu and Liang Cheng Tu},
  title = {{Temperature Gradient Method for Alleviating Bonding-Induced Warpage in a High-Precision Capacitive MEMS Accelerometer}},
  journal = {Sensors},
  year = {2020},
  doi = {10.3390/s20041186},
}
Figure