The widespread adoption of lithium-ion battery technology is a cornerstone of the modern energy transition. Their high energy density, long cycle life, and declining cost have made lithium-ion batteries indispensable for portable electronics, electric vehicles, and grid-scale energy storage systems. However, this rapid deployment has been accompanied by growing safety concerns, including incidents of thermal runaway, fire, and explosion. These events often originate from complex internal electrochemical side reactions that are not directly detectable by conventional battery management systems (BMS), which typically monitor only external parameters like terminal voltage, current, and surface temperature. To enhance the safety, reliability, and longevity of lithium-ion batteries, there is a critical need for advanced diagnostic tools capable of performing real-time, in-situ monitoring of internal state parameters under real operating conditions.
Among various sensing technologies, optical fiber sensors present a compelling solution for integration into the harsh environment of a lithium-ion battery. Their inherent advantages include immunity to electromagnetic interference, electrical passivity (eliminating spark risks), resistance to corrosion by electrolytes, and the potential for multiplexed, distributed sensing. While fiber Bragg grating (FBG) sensors have been successfully used to monitor temperature and strain, direct and precise measurement of internal gas pressure—a critical indicator of state-of-health and precursor to failure—remains a significant challenge. Pressure changes within a lithium-ion battery arise from a confluence of factors: the reversible “breathing” due to electrode lattice expansion/contraction during lithium intercalation/deintercalation, thermal expansion of materials and electrolyte, and the irreversible generation of gases from parasitic side reactions (e.g., electrolyte decomposition, solid electrolyte interphase (SEI) growth and breakdown). Accurately decoupling and tracking these pressure components requires a sensor with exceptional precision, stability, and minimal cross-sensitivity to temperature.
This article delves into the development and application of a novel, high-fidelity micro-electromechanical system (MEMS) fiber-optic Fabry-Perot (F-P) pressure sensor specifically engineered for long-term implantation and monitoring within commercial lithium-ion battery cells. The core innovation lies in the sensor’s fabrication process, which utilizes a gold-gold (Au-Au) thermocompression bonding technique to create the F-P microcavity. This approach fundamentally addresses key limitations of previous designs, such as thermal stress at material interfaces and residual gas pressure within the cavity, thereby drastically improving pressure measurement accuracy and reducing temperature cross-sensitivity. I will discuss the operational principles, detailed fabrication workflow, rigorous performance characterization, and finally, present compelling results from long-cycle in-situ pressure monitoring experiments conducted on commercial lithium iron phosphate (LFP) cells. The data reveals intricate correlations between electrochemical signals and internal pressure dynamics, enabling the clear separation of reversible and irreversible pressure components and offering a powerful new tool for lithium-ion battery prognostics and health management.
Fundamental Principles of the MEMS Fiber-Optic Pressure Sensor
Sensing Mechanism and Optical Interferometry
The sensor is a non-intrinsic, extrinsic Fabry-Perot interferometer (EFPI). Its core sensing element is a miniaturized MEMS chip attached to the end face of a standard optical fiber. The chip itself comprises a thin, single-crystal silicon diaphragm anodically bonded over an etched micro-cavity on a Pyrex glass substrate. The key advancement is the implementation of an intermediate Au-Au bonding layer between the silicon and the glass. Light from a broadband source is launched through the fiber. At the interface between the fiber end and the glass substrate (Point P1), and at the two surfaces of the silicon diaphragm (inner surface P2 and outer surface P3), partial reflections occur. The two primary reflected beams—one from the glass reference surface (P1) and one from the inner silicon surface (P2)—travel back along the fiber and interfere. The intensity of this interference signal is a function of the optical path difference (OPD) between the two beams, which is directly proportional to the physical length of the air-gap cavity between P1 and P2.
When external pressure is applied to the outer surface of the silicon diaphragm, it deflects inward, reducing the cavity length. This change in cavity length (ΔL) modifies the OPD and consequently shifts the phase of the interference signal. By precisely demodulating this phase shift, the applied pressure can be quantified. The relationship between the pressure (P) and the center deflection of a circular, edge-clamped diaphragm (ω0) is derived from plate theory and can be expressed as:
$$
\omega_0 = \frac{3(1 – \nu^2)}{16E t^3} R^4 P
$$
where:
- \( E \) is the Young’s modulus of silicon,
- \( \nu \) is the Poisson’s ratio of silicon,
- \( t \) is the thickness of the silicon diaphragm,
- \( R \) is the radius of the diaphragm (effective cavity radius).
The change in cavity length is approximately equal to this center deflection (\( \Delta L \approx \omega_0 \)). Therefore, the pressure sensitivity (\( S \)) of the sensor, defined as the change in OPD (\( 2 \cdot \Delta L \)) per unit pressure, is:
$$
S = \frac{\partial (2 \Delta L)}{\partial P} \approx \frac{3(1 – \nu^2)}{8E} \frac{R^4}{t^3}
$$
This equation highlights the critical design parameters. Sensitivity scales with the fourth power of the diaphragm radius and the inverse cube of its thickness. Thus, for high sensitivity, a large, thin diaphragm is desired. However, practical limits are imposed by the diaphragm’s burst pressure and the manufacturing process.
The demodulation of the interference signal is achieved using a Polarization Low-Coherence Interferometry (PLCI) system. This technique is highly effective for accurately recovering the absolute OPD of low-finesse F-P cavities. In a PLCI system, the reflected broadband light is passed through a scanning interferometer (often a birefringent crystal wedge). When the OPD introduced by the scanning interferometer matches the OPD of the sensor cavity, a coherence peak (interference fringe) is detected by a linear CCD array. Tracking the position of this peak allows for high-resolution, drift-free measurement of the sensor’s cavity length, which is directly converted to pressure.
The Critical Role of Au-Au Thermocompression Bonding
The performance and long-term stability of the sensor are predominantly determined by the quality and properties of the F-P microcavity. Traditional anodic bonding (glass-to-silicon) creates a hermetic seal but introduces two major sources of error:
- Thermal Stress: The Pyrex glass and silicon have different coefficients of thermal expansion (CTE). When bonded at high temperature (~400°C) and cooled to room temperature, the CTE mismatch generates significant residual thermal stress (\(\sigma_{thermal}\)) at the interface. This stress alters the diaphragm’s stiffness and its deflection characteristics under pressure and temperature, leading to a strong temperature-pressure cross-sensitivity.
- Residual Gas: The anodic bonding process can trap gases or produce them electrochemically within the sealed cavity. This results in a non-vacuum reference pressure (\(P_{residual}\)) that is temperature-dependent (following the ideal gas law), further corrupting the pressure measurement and increasing temperature dependence.
The proposed Au-Au thermocompression bonding strategy elegantly mitigates both issues. In this process, thin gold films are deposited on both the silicon diaphragm and the glass substrate. Under high temperature and pressure in a vacuum chamber, the gold layers diffuse into each other, forming a solid-state, hermetic bond. The advantages are profound:
- CTE Matching: The bonding interface is now between two identical gold layers. Since they have the same CTE, the residual thermal stress from the bonding process is drastically reduced. The primary thermal stress now arises only from the silicon-Pyrex bulk mismatch, which is less critical for the cavity’s zero-point stability.
- High-Vacuum Cavity: The bonding is performed in a high-vacuum environment, and the gold diffusion seal is excellent at preventing outgassing. This ensures the microcavity maintains a high vacuum (\(P_{residual} \approx 0\)), providing a stable and temperature-insensitive pressure reference.
The effective stress influencing the diaphragm deflection, as mentioned in the sensing principle, can be modified to include a thermal stress term. A more general form for center deflection under pressure \(P\) and with interfacial stress \(\sigma\) is:
$$
\omega_0 = \frac{(R^2 – r^2)^2}{64D(1+\xi)} P
$$
where \(D = \frac{Et^3}{12(1-\nu^2)}\) is the flexural rigidity, and \(\xi\) is a correction factor for the tensile stress:
$$
\xi = \frac{12 \sigma R^2 (1-\nu^2)}{14.68 E t^2}
$$
The interfacial stress \(\sigma\) is dominated by \(\sigma_{thermal}\). By using Au-Au bonding, we minimize \(\sigma_{thermal}\), making \(\xi\) very small. This leads to a diaphragm deflection that follows the ideal plate theory more closely and, crucially, makes the deflection (and thus the sensor’s pressure sensitivity) much less sensitive to temperature fluctuations.
Sensor Fabrication and Integration Process
The manufacturing of the sensor is a sequence of precision MEMS and optical packaging steps. The following table outlines the core fabrication workflow:
| Step | Process | Description and Purpose |
|---|---|---|
| 1. Substrate Preparation | Wafer Cleaning | Pyrex glass and Silicon-On-Insulator (SOI) wafers undergo rigorous RCA cleaning to remove organic and ionic contaminants. |
| 2. Metallization | E-beam Evaporation / Sputtering | A thin adhesion layer (Cr/Ti) followed by a gold layer (0.5-1 µm) is deposited on both the Pyrex wafer and the device layer of the SOI wafer. |
| 3. Cavity Patterning on Glass | Photolithography & Wet Etching | Photoresist is patterned on the Pyrex gold layer to define the cavity area. The exposed gold is etched, followed by wet chemical etching (e.g., HF-based) of the Pyrex to create the micro-cavity (typically 10-30 µm deep). |
| 4. Dielectric Mirror Deposition | PECVD or Sputtering | A thin, partially reflective dielectric coating (e.g., SiO2/TiO2 stack) is deposited on the bottom of the etched cavity to form the first mirror (P1) of the F-P interferometer, optimizing reflectivity (~30-40%). |
| 5. Au-Au Thermocompression Bonding | Wafer Bonding | The SOI wafer (gold side down) is aligned and pressed against the processed Pyrex wafer in a vacuum chamber at elevated temperature (300-400°C) and high pressure. The gold layers diffuse, forming a hermetic seal around the cavity. |
| 6. Membrane Release | Backside Grinding & Etching | The handle layer and buried oxide (BOX) layer of the SOI wafer are removed by grinding and wet/dry etching, leaving only the single-crystal silicon device layer as the free-standing diaphragm over the cavity. |
| 7. Dicing | Dicing Saw / Laser Cutting | The bonded wafer stack is diced into individual MEMS sensor chips, each containing one pressure-sensitive diaphragm. |
| 8. Fiber Attachment | Active Alignment & UV Curing | A cleaved multimode or single-mode fiber is actively aligned to the center of the MEMS chip’s glass substrate using micro-positioners while monitoring the interference signal. UV-curable epoxy is applied and cured to fix the fiber permanently. |
| 9. Packaging | Metal or Ceramic Housing | The fiber-coupled chip is potted into a miniature stainless steel or ceramic ferrule for mechanical protection and ease of handling during battery integration. |
This MEMS-based fabrication process ensures high uniformity and batch-to-batch consistency, which is essential for commercial viability. The final sensor is extremely compact, with an outer diameter typically matching that of the optical fiber ferrule (e.g., 1-2 mm), making it suitable for minimally invasive implantation into a lithium-ion battery cell.
Performance Characterization of the MEMS Fiber-Optic Sensor
Prior to integration into a lithium-ion battery, the sensors undergo comprehensive calibration and performance testing to quantify their pressure sensitivity, temperature cross-sensitivity, accuracy, and long-term stability.
Calibration Setup and Methodology
The sensor is placed inside a hermetically sealed, temperature-controlled chamber equipped with a precision pressure inlet. A high-accuracy pressure controller (e.g., 0.02% FS) regulates the internal pressure. A thermal chamber controls the environmental temperature. The sensor’s optical output is connected to the PLCI demodulation system. The calibration procedure involves:
- Pressure Sweep at Constant Temperature: At a fixed temperature (e.g., 25°C), pressure is varied across the full scale (e.g., 0 to 300 kPa absolute) in defined steps. The corresponding OPD (or phase) output from the PLCI system is recorded.
- Temperature Cycle at Constant Pressure: At a fixed pressure (e.g., atmospheric pressure), the temperature is cycled over the operational range (e.g., -40°C to +80°C). The sensor’s output drift is measured to determine its inherent thermal zero shift.
- Full 2D Calibration: Combining the above, pressure is measured at multiple stabilized temperature points across the full range to create a temperature-compensated calibration matrix.
Key Performance Results
The performance of the Au-Au bonded sensor is summarized and compared against typical anodically bonded sensors in the table below.
| Performance Parameter | Au-Au Bonded MEMS F-P Sensor | Typical Anodically Bonded Sensor |
|---|---|---|
| Pressure Range | 0 – 300 kPa (Absolute) | 0 – 300 kPa (Absolute) |
| Sensitivity | ~0.47 rad/kPa (Example) | ~0.45 – 0.50 rad/kPa |
| Pressure Accuracy (at 25°C) | < 0.02% FS (e.g., <0.06 kPa) | ~0.1% FS (e.g., ~0.3 kPa) |
| Temperature Cross-Sensitivity | 0.091 kPa/°C | 0.5 – 2.0 kPa/°C |
| Temperature Range | -40°C to +125°C | -20°C to +85°C |
| Zero-Point Drift over T-range | < 10 kPa pk-pk | > 50 kPa pk-pk |
| Long-Term Stability (at const. T&P) | < 0.1% FS / year | ~0.5% FS / year |
| Burst Pressure | > 5x FS | > 5x FS |
The data unequivocally demonstrates the superior performance of the Au-Au bonding approach. The temperature cross-sensitivity is reduced by an order of magnitude. This is a direct consequence of minimizing the thermal stress \(\sigma_{thermal}\) and the residual gas pressure \(P_{residual}\). The low cross-sensitivity can be expressed mathematically. The total sensor output \(O\) can be modeled as a function of pressure \(P\) and temperature \(T\):
$$
O(P, T) = S_P \cdot P + S_T \cdot T + O_0
$$
where \(S_P\) is the pressure sensitivity, \(S_T\) is the temperature sensitivity coefficient, and \(O_0\) is the zero offset. For a sensor with significant residual stress and gas, \(S_T\) is large. For the Au-Au bonded sensor, \(S_T\) is minimized, making \(O(P, T) \approx S_P \cdot P + O_0′(T)\), where \(O_0′(T)\) is a small, predictable, and often linear zero shift that can be easily compensated if temperature is known.
The sensor’s high precision is quantified by its non-linearity, hysteresis, and repeatability errors calculated from the calibration data. A typical precision better than 0.05% FS is achievable, which translates to a pressure resolution in the range of 10-50 Pa. This level of precision is essential for detecting the subtle pressure changes associated with early-stage side reactions in a lithium-ion battery.
In-Situ Monitoring of Lithium-Ion Battery Internal Pressure
Experimental Integration into a Commercial Cell
To validate the sensor’s operational capability, it was integrated into a commercial 18650-format lithium-ion battery with a Lithium Iron Phosphate (LFP) cathode and graphite anode. The integration was performed in an argon-filled glovebox to prevent contamination. A small, laser-drilled hole was made in the cell’s negative terminal (case). The sensor’s packaged ferrule was inserted and hermetically sealed using a high-temperature, chemically resistant epoxy. This minimally invasive approach ensures the cell’s internal chemistry and mechanics are disturbed as little as possible. The fiber-optic cable was routed out of the glovebox and connected to the PLCI demodulator. The modified cell was then connected to a standard battery cycler for electrochemical testing. The specifications of the test cell are below:
| Parameter | Specification |
|---|---|
| Cathode Chemistry | LiFePO4 (LFP) |
| Anode Chemistry | Graphite |
| Nominal Capacity | 1500 mAh |
| Nominal Voltage | 3.2 V |
| Charge Cut-off Voltage | 3.65 V |
| Discharge Cut-off Voltage | 2.0 V |
Pressure Dynamics During Galvanostatic Cycling
The cell was subjected to repeated charge-discharge cycles at a 1C rate (1500 mA). The standard cycle protocol was: Constant Current (CC) charge to 3.65V, Constant Voltage (CV) hold until current drops to C/20, rest, CC discharge to 2.0V, rest. The internal pressure, cell voltage, and current were recorded simultaneously.
The results from the initial cycles revealed a stable and repeatable correlation. The internal pressure signal exhibited a distinct “breathing” pattern synchronized with the charge-discharge cycles. During CC charge, pressure increased monotonically. This is attributed to the expansion of the graphite anode as lithium ions intercalate, causing a net increase in the volume of the electrode particles, which translates to a reduction in pore volume and an increase in internal gas pressure. The rate of pressure increase was not perfectly linear, reflecting the different stages of lithium intercalation (staging phenomena) in graphite.
Upon switching to the CV charge phase, the pressure promptly began to decrease. This is because the intercalation reaction slows dramatically as the cell approaches full charge, while the temperature of the cell—which had risen due to Joule heating and reaction entropy during fast CC charge—starts to equilibrate or even drop, causing a contraction of gases and materials. During the rest period, pressure continued to relax towards a baseline as the cell temperature fully equilibrated to ambient.
The discharge process showed the inverse behavior: pressure decreased during CC discharge (as the graphite anode contracts with lithium deintercalation) and then recovered during the subsequent rest period. This reversible pressure swing (\(\Delta P_{rev}\)) is a direct mechanical signature of the cell’s main electrochemical activity.
Long-Term Cycling and Irreversible Pressure Evolution
The experiment was extended over 40+ cycles to observe long-term trends. A critical finding was the evolution of the pressure “baseline.” By plotting the pressure value at the end of each full cycle’s rest period (when the cell is electrochemically idle and at near-ambient temperature), a clear upward drift was observed. This increasing baseline pressure (\(P_{irr}\)) represents the irreversible component of pressure change.
The total measured pressure \(P_{total}(t)\) at any point in cycle \(n\) can be conceptually decomposed as:
$$
P_{total}(n, t) = P_{ambient} + P_{irr}(n) + \Delta P_{rev}(n, t) + \Delta P_{thermal}(t)
$$
where:
- \(P_{ambient}\): External atmospheric pressure.
- \(P_{irr}(n)\): Cumulative irreversible pressure from gas-generating side reactions up to cycle \(n\). This component is monotonic and increases with cycle number and aging.
- \(\Delta P_{rev}(n, t)\): The reversible “breathing” pressure change during cycle \(n\). Its amplitude is related to the State of Charge (SoC).
- \(\Delta P_{thermal}(t)\): Fast pressure fluctuations due to rapid temperature changes during operation. (Our sensor’s low cross-sensitivity ensures this is a minor correction).
The growth of \(P_{irr}(n)\) is a direct indicator of cell degradation. Sources of irreversible gas generation include:
- Electrolyte reduction at the anode forming and reforming the SEI, especially in early cycles.
- Electrolyte oxidation at the cathode at high voltage.
- Transition metal dissolution from the cathode.
- Lithium plating on the anode surface at high charge rates or low temperatures.
The high-precision sensor allows for the quantification of this gas generation rate. The data showed that \(P_{irr}\) increased at a gradually accelerating rate over 40 cycles, which correlated with a measurable but small capacity fade. This demonstrates the sensor’s ability to detect degradation mechanisms long before they lead to catastrophic failure or significant capacity loss detectable by the BMS.
The following table summarizes the key pressure-related observations and their electrochemical correlations:
| Cycle Phase | Observed Pressure Trend | Primary Physical/Chemical Origin | Implication for Battery State |
|---|---|---|---|
| CC Charge | Steady Increase | Graphite anode expansion (Li intercalation). Minor heating. | Direct correlate with SoC increase. Slope changes may indicate staging. |
| CV Charge / Rest | Decrease & Relaxation | Cessation of main reaction. Cell cooling. Gas contraction. | Reveals thermal mass and heat dissipation properties. |
| CC Discharge | Steady Decrease | Graphite anode contraction (Li deintercalation). | Direct correlate with SoC decrease. |
| End-of-Cycle Baseline | Gradual Increase over Cycles | Cumulative irreversible gas from side reactions (SEI growth, electrolyte decomposition). | Key Health Indicator (SOH). Early warning for performance fade and safety risk. |
| High C-rate Cycling | Larger \(\Delta P_{rev}\) amplitude; Faster \(P_{irr}\) growth | Enhanced polarization, heating, and accelerated degradation kinetics. | Quantifies stress induced by fast charging/discharging. |
Discussion and Future Perspectives
The successful implementation of this Au-Au bonded MEMS fiber-optic pressure sensor opens a new window into the internal dynamics of lithium-ion battery cells. The ability to accurately measure sub-kPa pressure changes over long periods and wide temperature ranges provides a data stream that is orthogonal and complementary to traditional voltage and current measurements. This work substantiates several important conclusions:
- Technical Feasibility: Optical fiber sensors, particularly those based on robust MEMS Fabry-Perot structures, can survive and operate reliably within the strongly reducing and oxidizing environment of a working lithium-ion battery.
- Superior Performance: The Au-Au thermocompression bonding technique is a key enabler, delivering the high precision and low temperature cross-sensitivity required to resolve the subtle pressure signatures of battery electrochemistry.
- Rich Information Content: The internal pressure signal is information-dense, encoding details about reversible electrochemomechanical coupling (“breathing”), thermal transients, and, most importantly, the irreversible accumulation of gases from degradation side reactions.
The implications for battery research, management, and safety are significant:
- Advanced BMS: Integrating such a sensor into a next-generation BMS would provide a direct internal health parameter (\(P_{irr}\)). This could enable more accurate State-of-Health (SOH) estimation, early failure prediction, and adaptive charging algorithms that limit gas-generating side reactions.
- Materials and Cell Design: Researchers can use this tool for rapid, in-situ evaluation of new electrode materials, electrolytes, and cell designs by directly monitoring their gas generation propensity during cycling and aging.
- Safety Prognostics: A sudden acceleration in the irreversible pressure rise rate could serve as an early warning signal for conditions leading to venting or thermal runaway, potentially triggering safety countermeasures.
- Model Validation: The high-fidelity pressure data provides a critical benchmark for validating and refining multi-physics models of lithium-ion battery cells that couple electrochemical, thermal, and mechanical phenomena.
Future work will focus on several avenues: further miniaturization of the sensor package for integration into different cell formats (e.g., pouch cells); developing multi-parameter sensors that combine pressure, temperature, and perhaps strain sensing in a single fiber; creating distributed sensing networks within large-format battery modules; and employing machine learning algorithms on the pressure, voltage, and current data streams to build predictive models for battery lifespan and failure. The journey towards safer, longer-lasting, and higher-performance lithium-ion batteries will undoubtedly be aided by such advanced, in-situ diagnostic capabilities, with high-precision fiber-optic pressure sensing playing a pivotal role.