Microcavity strain sensor for high temperature applications

Abstract. A microcavity extrinsic Fabry–Perot interferometric (EFPI) fiber-optic sensor is presented for measurement of strain. The EFPI sensor is fabricated by micromachining a cavity on the tip of a standard single-mode fiber with a femtosecond (fs) laser and is then self-enclosed by fusion splicing another piece of single-mode fiber. The fs-laser-based fabrication makes the sensor thermally stable to sustain temperatures as high as 800°C. The sensor exhibits linear performance for a range up to 3700  με and a low temperature sensitivity of only 0.59  pm/°C through 800°C.


Introduction
Optical fiber-based sensors have gained wide application for strain monitoring due to their compact size, immunity from electromagnetic interference, multiplexing capabilities etc. thus offering an alternative to traditional electrical sensors and conventional pneumatic-based sensors. 1-3 Different types of optical fiber instruments like fiber Bragg grating sensors, extrinsic Fabry-Perot interferometric (EFPI) sensors, intrinsic Fabry-Perot interferometric (IFPI) sensors, long-period fiber grating sensors, and related hybrid combination sensors have been used for monitoring strain, stress, temperature, and pressure. [3][4][5][6][7] Temperature sensitivity and temperature maximums are important practical limitations for many of these sensors. For instance, the temperature sensitivity of Bragg grating and IFPI sensors are well known. 8,9 The EFPI-type sensors are better suited for strain monitoring applications with high ambient temperatures as opposed to Bragg gratings and IFPI sensors. These rugged sensors have excellent noise-free performance and fatigue characteristics. 10 The most widely used method for realizing the EFPI sensor is by epoxying two pieces of fiber, with cleaved ends, inside a hollow tube (glass or ceramic) and controlling the separation distance between the two fiberends. [11][12][13][14][15] In addition to the cumbersome fabrication process and the calibration issues related to controlling the cavity gap, this design has limited thermal performance due to the thermal expansion of the tube and the temperature limitation of the epoxy, e.g., Loctite epoxy extra time pro (slow setting) is effective up to 150°C once cured. Alternative approaches with low temperature sensitivities have been demonstrated by splicing a hollow-core fiber between two sections of single-mode fiber, 16 by forming voids at splices between photonic crystal fiber and conventional singlemode fiber, [17][18][19] and by laser-machining micro-cavities into single-mode fiber. 20 The microcavity sensor in Ref. 20 provides easily reproducible characteristics, but it has an open cavity design that exposes the cavity to the environment. In particular, this open cavity limits embedded applications. An EFPI sensor can also be fabricated using wet chemical etching in which diluted hydrofluoric acid forms a cavity in the tip of a multimode fiber, and this cavity is fused with a single-mode fiber. 21 This latter EFPI alternative has good temperature characteristics, but it suffers from safety concerns during fabrication and from difficulty in controlling the etch, i.e., for calibrating the cavity length.
In this work, a microcavity EFPI strain sensor is fabricated using femtosecond (fs) laser micromachining to form the cavity and is self-enclosed with a fusion splice. This sensor is less bulky than a tube-based EFPI, the fs-laser processing is fast and the resulting cavity length is precisely controlled, and the performance is relatively temperature insensitive and is thermally stable. The sensor is capable of operating in hightemperature applications. Fabrication, strain performance, and thermal effects are discussed.

Microcavity Sensor Design and Fabrication
The overall optical response for a Fabry-Perot cavity depends on multiple-beam interference in light transmitted and reflected from the two ends of the cavity. This periodic response is modulated by the wavelength and optical path (gap) length. 22 Figure 1(a) shows a traditional EFPI design in which the cavity is formed between the end faces of optical fiber that are aligned with an epoxied capillary tube. [11][12][13][14] In a strain sensor with an air gap of length d, the gauge length, i.e., the length of the sensing element, is approximately the tube length L and the measured strain is ΔL∕L ¼ Δd∕L. Figure 1(b) shows the microcavity EFPI *Address all correspondence to: Steve E. Watkins, E-mail: steve.e.watkins@ ieee.org in which the cavity is formed in the fiber itself and a second fiber is fusion spliced to self-enclose the cavity. As a strain sensor with an air gap of length d, the gauge length is the cavity gap length d and the measured strain is Δd∕d. The smaller gauge length allows the latter sensor to more closely approximate a point sensor. Also, the tube component causes the former design to be bulkier and to have a more complex fabrication than the microcavity design. Note that the exact gauge length and the initial gap length are more difficult to determine for the traditional design, hence calibration is an issue.
The EFPI response is dependent on any parameter changing the cavity optical path length. For the bare sensor with no applied strain, e.g., a sensor not attached to a structure to be measured, changes in ambient temperature T induces a Δd due to the thermal expansion of the silica fiber. Since the coefficient of thermal expansion (CTE) for silica (0.55 × 10 −6 ∕°C) is small, this temperature dependence is minimal. Hence, the single-mode silica fiber EFPI is an ideal candidate for hightemperature applications. The smaller gauge length and the absence of epoxy reduce the influence of temperature on the microcavity EFPI performance.
The microcavity EFPI has two glass-air interfaces with low reflectivity, which produces a sensor with low finesse F. The reflectance (ratio of the output signal irradiance I R to the input signal irradiance I i ) is 22 where n ¼ 1 is the refractive index of the cavity and λ is the wavelength. The condition for destructive interference is where m is an integer. Note that d can be calculated from adjacent minima at λ 1 and λ 2 as Demodulation methods for the microcavity EFPI are the same as for the traditional EFPI types. For this work, the phase tracking method is used, cf. Ref. 15, and wavelength shifts in the interference spectrum were measured. By Eq. (2), a change in cavity length Δd is proportional to the associated wavelength change for destructive interference Δλ. Hence, the measured strain is Δd∕d ¼ Δλ∕λ: (4) Figure 2 shows the micromachining system with a fs laser. The single-mode fiber is cleaved and the fiber tip is aligned with a five-axis translation stage (resolution 1 μm). The fs laser is focused on the fiber tip and a cavity is precisely ablated as shown in Fig. 3. The fs-laser system (maximum output of 1 W) operates at a center wavelength of 800 nm with the repetition rate and pulse width of 250 kHz and 200 fs, respectively. The laser power used for fabrication was 0.4 μJ per pulse. The sensor fabrication is completed by fusion splicing another single-mode fiber. The resulting cavity is 65 × 65 × 35 μm 3 (see Fig. 3). Figure 4(a) shows the instrumentation used for sensor testing. A 100-nm broadband source is the input, a 3-dB coupler sends the signal to the sensor and receives the reflected signal back, and an optical spectrum analyzer (OSA) then records the wavelength spectra. Figure 4(b) shows the spectra shift for an applied 500-με strain. Several microcavity EFPI sensors were fabricated with similar cavity lengths (∼35 μm),   fringe visibility (10 to 12 dB), and excess loss (4 to 6 dB) (The excess loss is the nonideal power drop, e.g., for destructive interference the return signal is −4 to −6 dB down from the ideal value of 0 dB.). Figure 5 shows the strain-induced response of the microcavity sensor at room temperature. The wavelength spectra shift is plotted with respect to the applied strain. The sensor was fixed between two translational stages and axial strain was applied in steps of 100 με. The sensor response was linear with response slope of 1.5 pm∕με. The strain was applied until the sensor broke at the fusion joint as verified under a microscope. The breaking point for the EFPI sensor was 3700 με approximately. The figure inset shows detailed performance for strains applied from 1000 through 1500 με. Figure 6 shows the temperature-induced wavelength shift with no applied strain. For the temperature testing, the sensor was placed inside a box furnace (Lindberg/Blue M). The temperature of the furnace was raised from 50°C to 800°C in steps of 50°C, and the resultant wavelength shift in the spectrum was recorded using the OSA. The results are plotted in terms of wavelength shift as well as the apparent strain. The slopes of the experimental response for the EFPI sensor were 0.59 pm∕°C and 0.37 με∕°C for wavelength shift and apparent strain, respectively. The CTE of silica was calculated to be

EFPI Sensor Testing
that is 1.3 times larger than that of the reference silica CTE of 0.55 × 10 −6 ∕°C.
The design's capability for handling high temperatures was also tested by keeping a second identical sensor at 650°C for 3 h. No change in the reflection spectrum was observed during this 3-h test. The sensor was then returned to room temperature and temperature-induced wavelength shift with no applied strain was again determined. The sensor survived the whole process without any deterioration in the performance, i.e., it had the same responses as given in Fig. 6 for the sensor with no prior high temperature history.

Conclusions
A robust, compact EFPI strain sensor is demonstrated that is easy to fabricate and calibrate and that has a high operating temperature. A single fusion joint allows for better structural integrity and is less complicated compared to traditional designs. The fs-laser fabrication results in well-controlled cavity for calibration. The wavelength shift with applied strain is linear up to a breaking point of about 3700 με. The sensor has a low cross sensitivity due to low thermal expansion of the silica glass, i.e., the wavelength shift with temperature is small up to at least 800°C. The experimental CTE value was slightly higher than the reference value for silica. Ongoing work is examining the sensor performance for embedded applications in which the sensor must survive high-temperature fabrication processes and must monitor strain at elevated temperatures. Preliminary results show successful sensor operation and strain transfer while embedded in carbon fiber composite laminate plate. 23 Overall, the microcavity EFPI is a good candidate for strain monitoring applications in high ambient temperatures.