Laser-assisted embedding of all-glass optical fiber sensors into bulk ceramics for high-temperature applications

Abstract

We develop a laser-assisted sensor embedding process to embed all-glass optical fiber sensors into bulk ceramics for high-temperature applications. A specially designed two-step microchannel was fabricated on an Al₂O₃ substrate for sensor embedment using a picosecond (ps) laser. An optical fiber Intrinsic Fabry-Perot Interferometer (IFPI) sensor was embedded at the bottom of the microchannel and covered by Al₂O₃ slurry which was subsequently sintered by a CO₂ laser. The sensor spectrum was in-situ monitored during the laser sintering process to ensure the survival of the sensor and optimize the laser sintering parameters. By testing in furnace through high temperature, the embedded optical fiber shows improved stability after CO₂ laser sealing, resulting in the linear temperature response of the embedded optical fiber IFPI sensor. To improve the embedded IFPI sensor for thermal strain measurement, a dummy fiber was co-embedded with the sensing fiber to improve the mechanical bonding between the sensing fiber and the ceramic substrate so that the thermal strain of the ceramic substrate can apply on the sensing fiber. The response sensitivity, measurement repeatability and high-temperature long-term stability of the embedded optical fiber IFPI sensor were evaluated in this work.

1. Introduction

Due to their brilliant mechanical and thermal properties, ceramic materials have been widely applied as the critical components of systems working in high temperature, such as energy production systems, high temperature heating equipment, and aerospace facilities [1–5]. Since these systems normally work under the extremely harsh conditions for a long period, the evaluation of their structural health is necessary for system maintenance and optimization. An embedded sensor is one of the effective ways to accomplish this objective. The real-time information of the part, such as temperature and strain, can be continuously collected through in-situ monitoring of the embedded sensors during system operation [6].

All-glass optical fiber sensors are among the promising candidates for structural status monitoring under harsh environment [7,8]. In addition to the well-known advantages such as compact size, high spatial resolution, fast response and immunity to electromagnetic interference, the all-glass optical fiber sensors are robust to operate under high temperature. For example, the optical fiber Intrinsic Fabry-Perot interferometer (IFPI) has shown great long-term high-temperature stability up to 1100°C for over 1200 hours [9,10]. Since the optical fiber IFPI sensor is highly sensitive to the tensile stress applied to the optical fiber, this sensor is capable of sensing the thermal strain of the components if the fiber is well attached to the parts [11].

In general, attaching the sensor to the part without damaging the optical fiber is crucial in the fabrication of the optical fiber sensor embedded components. One of the common methods is to mount the fiber sensors on the surface of the finished parts with robust protectors. This technique has been proposed for years to monitor the health status of concretes [7,8,12]. However, the surface-mounted method usually results in poor attachment between the sensors and the components, leading to offset between the sensing signals and the real variation of the part [13,14]. In addition, for harsh environment application, the sensors are usually mounted far away from the operating points to avoid damage on the optical fiber. In this way, the sensor only detects the variation of the part indirectly with low spatial and temporal resolution [6].

Additive manufacturing (AM) has been developed to embed the optical fiber sensors into the bulk materials during the part fabrication. In this way, the optical fibers are buried inside the components, which significantly improves the attachment between the sensors and parts and protects the sensors under harsh environment [15,16]. AM methods have been developed to embed the glass optical fiber sensors into the metallic components for high temperature applications [17,18]. The main challenge for this internal sensor embedding process is the thermal expansion mismatch between the glasses and the metals. At rising temperature, large thermal strain applied on the optical fiber from the metallic parts will delaminate the fibers from the components and degrade the performance of the embedded sensors [17]. In addition, the AM methods are only suitable to embed the glass optical fiber sensors into the materials whose melting point is lower than the working temperature of the fused silica glasses. It is still quite difficult to embed the all-glass optical fiber sensors into the high-temperature ceramics, such as Al₂O₃ and yttria-stabilized zirconia (YSZ), using the AM methods, since the glass cannot survive the sintering temperature of most ceramics.

Recently, sapphire optical fiber has been successfully embedded into the alumina ceramics using the AM method [19]. Since the melting point of single crystal sapphire is over 2000°C, sapphire optical fibers can survive the post-sintering process of the 3D-printed alumina ceramics and have potential to work on temperature over 1500°C. However, due to the lack of cladding layers, the sapphire optical fibers are normally multi-mode fibers with large modal volume, which complicates the interrogation of sapphire optical fiber sensors [20,21]. In addition, the high optical loss of the sapphire fibers also limits it for high-performance sensing applications.

Compared to the AM methods, laser processing technologies are promising to overcome the challenge of embedding the all-glass optical fiber sensors into the high-temperature ceramics. Laser has shown its unique capability for high-resolution processing of ceramic materials [22–24]. Ultrafast laser has been developed for machining micro structures like microchannels on bulk ceramics with a resolution of up to several microns [25,26]. Since the pulse duration is shorter than the typical thermalization time of materials, the ultrafast lasers can machine the materials without thermally degrading the mechanical strength of the parts [27]. In addition, fast, precise and flexible heat treatment on ceramic materials has been realized using the CO₂ laser. The laser heating effective zone can be precisely controlled in three dimensions with ultrahigh heating and cooling rate. The material properties, such as density and cracking propagation, can be flexibly fine-tuned through adjusting the laser processing parameters [28,29]. Both of these laser technologies are promising to accomplish embedding glass optical fiber sensors into finished ceramic products, resulting in the improvement of flexibility and efficiency in the fabrication of sensor-embedded smart ceramic components.

Here we propose a laser-assisted sensor embedding process to embed the all-glass optical fiber sensors into bulk ceramics. A specially designed two-step microchannel was machined on an Al₂O₃ substrate for sensor embedment using a picosecond (ps) laser. An IFPI sensor, which was fabricated on a glass single mode optical fiber by the femtosecond (fs) laser irradiation, was embedded to the bottom of the microchannel and covered by the Al₂O₃ slurry. The filled Al₂O₃ slurry was subsequently sintered by a CO₂ laser to seal the sensor inside the part. The design of the two-step microchannel was based on the shape of the optical fiber and the heating depth of the CO₂ laser. During the laser sealing process, the spectrum of the optical fiber IFPI sensor was in-situ monitored to ensure the survival of the sensor and optimize the laser sintering parameters. The microstructure of the sensor-embedded Al₂O₃ substrate was presented to evaluate the laser sealing quality. By high-temperature measurement in a furnace, the high-temperature response, repeatability and long-term stability of the embedded optical fiber IFPI sensor were investigated.

2. Experiments

2.1. Optical fiber IFPI sensor fabrication

The schematic of the optical fiber IFPI sensor was shown in Fig. 1. The IFPI sensor was formed by two internal partial reflectors created by a femtosecond (fs) laser at the core of a single-mode glass optical fiber (SMF-28, Corning Inc.). Owing to the non-linear effect of the ultrashort laser pulses, the fs laser can locally modify the refractive index of the optical fiber at the focused laser spot. The spot size of the focused fs laser beam is ~1 μm, which is smaller than the diameter of the fiber core (8.2 μm), so that the reflectors can be exactly fabricated within the core area. The detail of the manufacturing technique has been described in our previous publication [30].

Fig. 1.

Schematic of the optical fiber IFPI sensor fabricated by the femtosecond (fs) laser irradiation.

2.2. Laser-assisted sensor embedding process

Fig. 1 schematically illustrates the process flow to embed the all-glass optical fiber IFPI sensor into the Al₂O₃ substrate. A microchannel with a two-step structure was machined on a commercial Al₂O₃ ceramic substrate (AO96, MTI Corporation) by a ps Nd:YAG laser (APL-4000, Attodyne Inc.). The ps laser firstly focused on the top surface of the substrate and did raster scanning in predesignated dimensions to fabricate the first-step groove. Upon completion of the first step, the laser beam was moved downward to focus on the surface of the first-step groove to machine a narrower groove inside it, which has the similar shape of the optical fiber for sensor housing (Fig. 2A).

Fig. 2.

Schematics of the process flow to embed the all-glass optical fiber sensors into the Al₂O₃ ceramic substrate assisted by multiple laser processing techniques.

To bury the sensor inside the Al₂O₃ substrate, the Al₂O₃ slurry, which is a high-temperature alumina filled ceramic adhesives (Ceramabond 503, Aremco Inc.), was applied to cover the sensing fiber by the on-demand slurry dispensing process. The flow rate, thickness and width of the dispensing line were controlled by a micro-dispenser (eco-Pen 300, Preeflow), which is capable of filling the microchannel with precise location and slurry quantity control (Fig. 2B).

After the filled Al₂O₃ slurry completely dried in air, a CO₂ laser (Firestar v20, SYNRAD Inc.) was used to scan along the channel to densify the filled Al₂O₃ slurry and seal the sensor inside the Al₂O₃ substrate. Before laser sintering, the Al₂O₃ slurry was preheated by fast CO₂ laser scanning at low laser power density for 1 minute. Subsequently, the CO₂ laser scanned the preheated Al₂O₃ slurry with higher laser power density and slower scanning speed for laser sintering. The spectrum of the sensor was in-situ monitored by an optical spectrum analyzer (OSA, AQ6370D, Yokogawa) during the whole laser heating process (Fig. 2C).

3. Results and discussion

3.1. Ps laser machined two-step microchannel

To firmly attach the optical fiber to the ceramic substrate, the microchannel needs to have a similar shape of the optical fiber to host the fiber at the bottom. In addition, to protect the optical fiber during the laser sealing process, the distance between the top of the optical fiber and the surface of the substrate should be slightly larger than the sintering depth of CO₂ laser on the Al₂O₃ slurry, which is about 100 μm as reported previously [31]. Since the diameter of the standard optical fiber is 125 μm, the depth of the microchannel is designed as 250 μm to ensure that the Al₂O₃ slurry filled on top of the embedded optical fiber can be sintered by the CO₂ laser and the embedded fiber can survive the laser sintering process. However, due to the cone shape of the focused ps laser beam, it is difficult to precisely control the shape of the microchannel when its depth is larger than the opening width.

In this case, a two-step microchannel was adopted to facilitate the microchannel fabrication. As shown in Fig. 3A, the top layer of the microchannel was fabricated with an opening width of 350 μm and a depth of 125 μm, and a narrower channel was machined inside the top layer one with a width of 150 μm and a depth of 125 μm. The laser parameters such as the output power, repetition rate, spot size and scanning speed for the micromachining process were set at 4 W, 10 kHz, 20 μm and 15 mm/s, respectively. The spacing of raster scanning was set at 15 μm to ensure the overlaps between the adjacent laser scanning lines. The depth of the groove was controlled by the times of raster laser scanning on the substrate. For every raster laser scanning, the depth of the groove increases about 20 to 30 μm on the ceramic substrate. After 5 times laser scanning, a groove with a depth of ~125 μm was obtained. Upon the completion of the first-step groove, the laser beam was focused on the surface of the first-step groove and scanned another 5 times to fabricate the narrower inner groove to form the two-step microchannel structure.

Fig. 3.

Optical microscope images of (A): the cross section of the ps laser-machined two-step microchannel; (B): the top view of the microchannel embedded with an optical fiber.

The wider top layer decreases the general aspect ratio of the microchannel, resulting in more precise control on the shape of the bottom channel which hosts the optical fiber. Moreover, due to the cone shape of the ps laser beam, the laser-machined microchannel becomes a bell shape as the channel depth increases during laser machining. As shown in Fig. 3B, the homogeneous bright part under the optical fiber after the fiber was placed on the bottom of the narrower channel indicates that the curved bottom of the bell-shape channel well fits the optical fiber without large gaps between the fiber and the substrate.

3.2. CO₂ laser sealing

To seal the sensor inside the substrate, CO₂ laser sintering process is applied to densify the filled Al₂O₃ slurry. The sensor spectrum was in-situ monitored during laser sealing to ensure the embedded optical fiber IFPI sensor survive the process. To optimize the laser parameters for the sealing process, the Al₂O₃ slurry was processed by different laser power with a fixed scanning speed of 1 mm/s and a fixed spot size of 1 mm. The optimization process was started from a laser power of 1 W, and the power value kept increasing for every next scanning until the interference spectrum was distorted. The sample was cooled down to room temperature before scanned by a different power value.

Fig. 4A shows the spectra of the embedded IFPI sensor after scanned by different laser power in the optimization process. As the laser power increased from 1 W to 8 W, no apparent variation was observed on the interference spectrum after every laser scanning. When the laser power reached 9 W, the interference spectrum shifted ~1 nm to the shorter wavelength region after laser scanning, while the interference pattern kept clean with neglectable loss. The Finesse factors of the interference spectra before and after laser sealing at 9 W laser power were calculated to evaluate the integrity of the IFPI. As shown in Fig. 4A, the small difference of the Finesse factors indicates that the integrity of the IFPI maintained without distortion though the spectrum shifted. After the laser power increased to 10 W, the interference pattern was completely distorted with a huge loss in the fringe visibility, indicating that the reflectors of the IFPI was damaged by this power value. Therefore, the laser power for the sealing process in this case was optimized to 9 W. The spectrum shift after laser scanning at 9 W laser power is attributed to the sintering and shrinkage of the Al₂O₃ slurry around the embedded optical fiber, resulting in the slight compression effect on the IFPI cavity.

Fig. 4.

(A) Spectra of the embedded optical fiber IFPI sensor before and after CO₂ laser sealing at the laser power of 8 W, 9 W and 10 W; (B) The cavity length variation of the embedded IFPI sensor during the laser sealing process at 8 W and 9 W.

The cavity length variation of the embedded IFPI sensor during the laser sealing process was recorded to further investigate the effect of laser on the embedded sensors. As shown in Fig. 4B, the cavity length increased as the laser beam passed the embedded sensor position, and quickly recovered to the original value when the sensor cooled down. Larger variation in the cavity length was observed at higher laser power, indicating that the higher temperature was induced as the laser power increased.

To evaluate the laser sealing quality, the scanning electron microscope (SEM) images of the sensor-embedded ceramic substrate are presented in Fig. 5. As shown in the top-view images (Fig. 5A and B), the Al₂O₃ slurry was fully densified by the CO₂ laser without any cracks. The densified track showed exactly the same width as the opening of the microchannel, indicating that the sensor was firmly sealed at the surface of the part. Observed from the cross-section images in Fig. 5C and D, no gaps can be found between the optical fiber and the filled Al₂O₃ slurry.

Fig. 5.

SEM images of (A): top view of the sensor-embedded microchannel after CO₂ laser sealing; (B): the zoom-in image of (A); (C): the cross-section of the sensor-embedded part after CO₂ laser sealing; and (D): Zoom-in image of (C).

To demonstrate the effect of CO₂ laser sealing on improving the high-temperature stability of the embedded optical fiber, standard-optical-fibers-embedded ceramic substrates with and without laser sealing were both heated to 800°C to evaluate the transmission of the embedded standard optical fibers. As shown in Fig. 6A, when the temperature increased to 400°C, huge fiber loss was observed at the part without laser sealing. As the temperature kept increasing, ripples occurred on the spectrum until the temperature reached 800°C. On the contrary, the transmission spectrum of the embedded optical fiber in the laser-sealed ceramic part was much more stable with 5 dB loss and smaller ripples at 800°C (Fig. 6B).

Fig. 6.

Transmission spectra of the embedded optical fibers heated from room temperature to 800°C, (A): without CO₂ laser sealing; (B): with CO₂ laser sealing; Inset of (A): cross-section optical microscope image of the part without laser sealing after testing to 800°C (scale bar: 125 μm).

The ripples and fiber loss on the transmission spectra are caused by the microbending effect of the embedded optical fiber during the high-temperature measurement. As the temperature increased, the ceramic substrate thermally expanded and perturbed the embedded optical fiber due to thermal expansion mismatch. Since the surface of the microchannel is rough after laser machining, the perturbation to the optical fiber at the contact points between the optical fiber and the substrate resulted in the microbending effect.

In the part without laser sealing, since the Al₂O₃ slurry was not pre-sintered by the laser, some additional reactions such as polymer burnout and ceramic sintering occurred as the temperature increased, causing additional perturbation on the embedded optical fiber to induce huge optical loss. In addition, as shown in the inset of Fig. 6A, the filled Al₂O₃ slurry in the part without laser sealing was delaminated with the ceramic substrate after heating through high temperature. Therefore, the laser sealing process can effectively improve the high-temperature stability of the embedded optical fiber through pre-sintering the filled Al₂O₃ slurry and fusing it to the ceramic substrate.

3.3. High-temperature response of the embedded optical fiber IFPI sensor

The sensor-embedded ceramic substrate was tested in an electric furnace from room temperature to 800°C to investigate the high-temperature response of the embedded optical fiber IFPI sensor. As the temperature increased, the interference spectrum of the IFPI sensor shifted to longer wavelength region. The spectral shift as a function of temperature change is plotted in Fig. 7A. Linear regression was used to fit the response curve and the slope was estimated as 14.95 pm/°C, which is considered as the sensor sensitivity in terms of wavelength shift versus temperature. The obtained temperature sensitivity agrees with the typical temperature response of the bare optical fiber IFPI sensor as reported in [30]. When the part was cooled down to room temperature, a similar temperature sensitivity of 14.94 pm/°C was obtained, indicating that only the temperature variation of the ceramic substrate was responded by the embedded IFPI sensor. Due to the large thermal expansion mismatch and weak mechanical bonding between the optical fiber and ceramic substrate after laser sealing, the embedded optical fiber was de-bonded with the ceramic substrate at the rising temperature so that the thermal strain effect of the substrate was not able to apply on the optical fiber.

Fig. 7.

(A): The temperature response of the embedded optical fiber IFPI sensors in two heating-cooling cycles; (B) The spectra of the embedded IFPI sensor at room temperature and 800°C.

To estimate its repeatability and stability in the temperature response, two cycles of heating and cooling processes were performed to the IFPI sensor-embedded ceramic substrate. As shown in Fig. 7A, similar temperature sensitivity was obtained in the second heating-cooling cycle, indicating that the embedded sensor shows good repeatability for high-temperature monitoring of the ceramic components. Fig. 7B shows the spectra of the embedded IFPI sensor at room temperature and 800°C. The calculated Finesse factor of the spectrum kept almost intact after heating to high temperature, indicating that the embedded IFPI sensor is stable at high temperature without apparent optical loss and distortion on the interference spectrum.

To improve the embedded IFPI sensor for thermal strain measurement, a dummy fiber was co-embedded with the sensing fiber to strengthen the mechanical bonding between the sensing fiber and the ceramic substrate (inset of Fig. 8A). Fig. 8A plots the temperature response of the optical fiber IFPI sensor after the dummy fiber was co-embedded. In the first heating process, the sensor sensitivity still shows the typical temperature response of the bare optical fiber IFPI sensor, suggesting that the firm mechanical bonding between the fiber and the substrate was not accomplished during the first heating process. However, in the subsequent cooling process, the interference spectrum moved back toward the shorter wavelength direction with a larger slope of 24.52 pm/°C.

Fig. 8.

(A): The temperature response of the dummy fiber co-embedded component in the 1^st heating (red) and cooling (blue) process; Inset of (A): Cross-section optical microscope image of the dummy fiber co-embedded component; (B) The temperature response of the dummy fiber co-embedded component in the 2^nd heating (red) and cooling (blue) process; (C) High temperature stability of the embedded IFPI sensor in the dummy fiber co-embedded component heated at 800°C for 17 hours.

The increase in sensitivity during the cooling stage is probably contributed to by the thermal shrinkage of the ceramic substrate. Typically, the spectral shift of the IFPI interference signal as a function of temperature can be calculated as:

$$\Delta {\lambda }_0/°\text{C}=\left({\alpha }_{CTO}+{\alpha }_{CTE}\right){\lambda }_0$$ (1)

where α_CTO (8.3 × 10⁻⁶ °C⁻¹) is the thermal-optic coefficient of fused silica glass, and α_CTE is the thermal expansion coefficient (CTE), which should be affected by both the optical fiber and the embedded environment for the cases of embedded sensors. λ₀ is the characterized spectrum position which is used to monitor the spectral shift [30]. In this case, when Δλ₀ is 24.5 pm/°C at λ₀ = 1551 nm, the calculated α_CTE is ~7.35 × 10⁻⁶ °C⁻¹. Since the optical fiber will also thermally expand as the temperature increases, part of the α_CTE should be contributed to by the CTE of fused silica glass (0.55 × 10⁻⁶ °C⁻¹). Therefore, the CTE of the ceramic substrate is calculated as ~6.8 × 10⁻⁶ °C⁻¹, which is close to the typical CTE of pure alumina (8.0 × 10⁻⁶ °C⁻¹). This result evidences that after the first heating, the mechanical bonding between the optical fiber and the ceramic substrate was effectively improved, so that the thermal strain of the ceramic substrate was able to apply on the optical fiber and detected by the embedded IFPI sensor. The improvement in the mechanical bonding is attributed to the firm contact between the sensing fiber and the dummy fiber, which effectively stabilized the sensing fiber after the first heating process.

As shown in Fig. 8B, similar temperature response as the first cooling process was obtained at the second heating-cooling cycle of the dummy fiber co-embedded component, which indicates that the embedded IFPI sensor shows good repeatability in temperature and thermal strain measurement for the ceramic components. Furthermore, the high-temperature long-term stability of the dummy fiber co-embedded component was also evaluated. As shown in Fig. 8C, the embedded IFPI sensor shows good high-temperature long-term stability with < 10°C deviation when heated at 800°C for 17 hours.

4. Conclusion

In summary, an all-glass optical fiber IFPI sensor was successfully embedded into a commercial Al₂O₃ ceramic substrate using the laser-assisted sensor embedding process. The two-step structure effectively reduces the general aspect ratio of the laser-machined microchannel to provide precise control on the channel shape, which is essential to fit the optical fiber inside the ceramic substrate. After sealing by the CO₂ laser sintering, the high-temperature stability of the embedded optical fiber is dramatically improved, while the part without laser sealing shows huge fiber loss at high temperature. By testing from room temperature to 800°C, the embedded optical fiber IFPI sensor shows a linear temperature response, which agrees with the bare optical fiber IFPI sensor. By co-embedding a dummy fiber between the Al₂O₃ slurry and the sensing fiber, the mechanical bonding between the sensing fiber and the ceramic substrate was effectively improved after the first heating process, resulting in the detection of the thermal strain of the ceramic component through the embedded IFPI sensor. The embedded optical fiber IFPI sensor show good repeatability in both temperature and thermal strain monitoring, and long-term stability with < 10°C deviation at 800°C for 17 hours.