CO2Laser Tapering of Intrinsic Fabry–Perot Interferometers for Sensing

We present a highly reproducible method of fabricating a tapered intrinsic Fabry–Perot interferometer (IFPI) device with 5–6 <inline-formula> <tex-math notation="LaTeX">$\mu \text{m}$ </tex-math></inline-formula> diameter at the taper waist. A femtosecond laser was applied to inscribe an IFPI with 3-cm cavity length in a single-mode fiber. A CO2 laser-heated tapering process enabled by digitally controlled mirrors and a precision motorized fiber feed system was used to create a stable heating zone with the desired temperature profile for tapering the fiber IFPI cavity. The well-engineered tapering process produced tapered IFPI devices with insertion loss less than 0.3 dB at 1550 nm. A strong evanescent field exposed by the taper Section was explored for refractive index (RI) sensing. Using swept optical frequency domination reflectometry (OFDR), the tapered IFPI fiber sensor achieved a minimal RI sensing resolution of <inline-formula> <tex-math notation="LaTeX">$2\times 10^{-{5}}$ </tex-math></inline-formula>. This article demonstrates an integrated laser fabrication technique to produce tapered fiber optic devices for sensing applications.


CO 2 Laser Tapering of Intrinsic Fabry-Perot
Interferometers for Sensing Xinruo Yi , Yuqi Li , Kehao Zhao , Zekun Wu , Qirui Wang , Bo Liu , Member, IEEE, Michael Buric, Ruishu Wright, and Kevin P. Chen Abstract-We present a highly reproducible method of fabricating a tapered intrinsic Fabry-Perot interferometer (IFPI) device with 5-6 µm diameter at the taper waist. A femtosecond laser was applied to inscribe an IFPI with 3-cm cavity length in a single-mode fiber. A CO 2 laser-heated tapering process enabled by digitally controlled mirrors and a precision motorized fiber feed system was used to create a stable heating zone with the desired temperature profile for tapering the fiber IFPI cavity. The well-engineered tapering process produced tapered IFPI devices with insertion loss less than 0.3 dB at 1550 nm. A strong evanescent field exposed by the taper section was explored for refractive index (RI) sensing. Using swept optical frequency domination reflectometry (OFDR), the tapered IFPI fiber sensor achieved a minimal RI sensing resolution of 2 × 10 −5 . This article demonstrates an integrated laser fabrication technique to produce tapered fiber optic devices for sensing applications.

I. INTRODUCTION
T APERED optical fiber devices have drawn significant attention in recent years for use as sensors. Through adiabatic tapering, standard telecom fibers with 125 µm diameter can be drawn down to <10 µm waist diameter with low loss. The drastic reduction of diameter in the tapered section facilitates large wavelength dispersion and a very small bending radius. A large fraction of power in the evanescent field surrounding the taper provides strong coupling between the fiber section and the outside ambient surrounding. These unique traits enable tapered fibers to perform various physical and chemical measurements.
The fabrication of high-quality tapered fibers is the key to developing functional sensor devices of this type. Ideally, the tapering process should minimize the propagation loss in the fiber while preserving the desired tapering profile to maintain single-mode propagation. Since the optical properties of fiber tapers are highly sensitive to taper geometry, it is of great importance to develop a reproducible fabrication method to produce devices with consistent characteristics. Several approaches have been used previously to fabricate tapered fibers. These existing methods involved either a flaming torch [1], microfurnace [2], or electrical arcing [3]. Well-engineered and optimized tapering systems using these methods can produce fiber tapers with low loss. However, the use of any of these bulk-heating systems results in a long heated zone on the fiber. Furthermore, they do not provide sufficient spatial resolution to carefully control the heating profile. Although low-loss fiber tapers can be produced, a long piece of thermally altered fiber can be fragile and difficult to handle. Another drawback of using bulk-heating systems for fiber tapering is the difficulty in integrating the tapers produced This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ with other features or devices on the tapered fiber. Since bulkheating systems alter a long section of fiber, this eliminates the possibility of integrating other fiber devices such as fiber Bragg gratings or in-fiber Fabry-Perot interferometers (FPIs) with fiber tapers, which reduces the usefulness of tapers fabricated with bulk-heating methods. As conventionally produced tapered fibers are difficult to handle, it is exceptionally challenging to fabricate fiber devices after the tapering process. Therefore, it is highly desired to fabricate in-fiber devices or features first and then to conduct the tapering process.
Laser heating has long been used for more precise control of thermal processes such as the laser-heated pedestal growth of single-crystal optical fibers [4], cutting of sheet goods and thin materials [5], and micro-welding of leads or metal parts [6]. We note that laser heating is also a viable approach to fabricate optical fiber tapers. A CO 2 laser can precisely control the length and temperature profile of the heated zone in a highly repeatable way by scanning the laser beam along the fiber with a digitally controlled scanning mirror. Laser heating can also avoid the production of air turbulence and combustion byproducts incurred using flame heating [7]. In practice, the length of the thermally altered fiber section produced by heating with a CO 2 laser can be as small as 5 mm, which introduces the possibility of integrating other devices with the taper to perform both transmission and reflection measurements.
In this article, we demonstrate the fabrication of a double fiber taper inside an intrinsic Fabry-Perot fiber interferometer. Using a femtosecond laser direct writing scheme, one pair of Rayleigh-enhanced backscattering points is inscribed inside a single-mode optical fiber to form an intrinsic FPI (IFPI) with a cavity length of 3 cm.
Femtosecond laser fabrication techniques have been used to produce FPI fiber sensors. Most of device fabrication works involve laser micromachining to create an optical cavity on cleaved optical fiber end facets [8], [9] or produce micronotch along the optical fibers [10], [11], [12]. However, laser micromachining techniques on silica fiber can incur high insertion loss and compromise the mechanical integrities of fibers. In this article, we use a femtosecond laser direct writing scheme to induce strong backscattering points to form IFPI sensors. It creates unique possibilities to explore the integration of IFPI sensors with tapered fiber section for sensing applications.
Using the CO 2 laser heating method, a biconical tapered fiber section 16.5 mm long is successfully drawn in between the cavity ends (3 cm apart) to achieve a 4.65-cm-long tapered IFPI cavity with typical transmission loss in the range of 0.1-0.3 dB at 1550 nm. Using swept wavelength optical frequency-domain reflectometry, cavity strain changes as small as 100 nε can be unambiguously determined. When used for refractive index (RI) sensing in a surrounding medium of changing index, the same device can measure index changes as small as ∼2 × 10 −5 .

II. FIBER TAPERING SYSTEM
A schematic of the optical fiber taper fabrication rig is shown in Fig. 1(a). A custom-designed LabView control program is used to achieve automated operation for several (a) Schematic of the optical fiber tapering system. (b) Photograph of the tapering system layout. (c) Photograph of blackbody radiation generated by the CO 2 laser beam during the heating process. components in this system, including feedback control to stabilize CO 2 laser output power, scanning galvo mirror positioning, movements of both upper and lower motors, as well as real-time image acquisition through camera monitoring. Through the use of this automation system, excellent reproducibility of the optical fiber tapering process is achieved. A 30-W Synrad v30 CO 2 laser is used as the heat source for tapering. A dichroic mirror is used to guide a He-Ne alignment laser collinear to the CO 2 laser beam. A ZnSe lens with a 10.2-cm focal length is used to focus the CO 2 laser beam onto the fiber. The fiber itself is contained inside a protective ceramic tube to help normalize the circumferential heating zone profile despite the laser impinging on the fiber from one side only. A high-speed scanning galvanometer mirror with a maximum mechanical scan angle of ±20 • is used to steer the focused CO 2 laser beam. The galvanometer can scan along a 6.5-cm length of fiber. To maintain the desired temperature profile along the entire scanned section of the fiber to be tapered, the scanning rate was set at 100 Hz [13]. In addition, the laser power was absorbed by the protective sleeve rather than being directly incident on the fiber to reduce the impact of laser power fluctuation and one-sided heating [14]. Two motorized stages with two sets of rubber drive wheels were used to feed fibers through the laser-heated sleeve, as shown in Fig. 1(b). When the temperature of the fiber is at the softening point, the stages start pulling both ends of the fiber. By adjusting the motor pulling speed, the applied power, and the scan rate, various taper geometries can be induced. As the taper is being drawn, fiber transmission is measured at 1550 nm via a laser and power meter. To observe the change in fiber diameter during the tapering process, a camera combined with a 6× magnification telecentric lens is used to record the process. The camera images can then be used to measure the fiber diameter using a machine-vision algorithm.
During the tapering process, the CO 2 laser power is controlled carefully via an applied control voltage. In order to maintain the fiber at its melting temperature during the tapering process, the CO 2 laser's output power should progressively increase to achieve the required energy flux at the fiber as the diameter of fiber is decreased. The CO 2 laser power output is monitored and adjusted automatically through a PID control algorithm, which relies on feedback from the laser power meter on the backside of the heated fiber, as shown in Fig. 1(a). Based on the conservation of mass during the tapering process, the tapered diameter profile can be predicted by using the constant hot zone model described by Birks and Li [15]. To describe the shape of a complete fiber taper, the taper is assumed to form symmetrically where the ends of the taper are pulled at equal and opposite speeds relative to the center of the heat source. The taper radius closely follows an exponential profile: where r 0 is the initial fiber radius before tapering, L is the scan length of the hot zone, and z is the pull length. In this work, the laser-swept section along the protection sleeve has a length of around 10 mm, while the real illuminated length is about 5-6 mm, which emits strong blackbody radiation, as shown in Fig. 1(c). The length of hot zone above the softening temperature for silica is shorter than the scan range because heat is dissipated toward both unheated ends of the protection sleeve.

III. SENSOR STRUCTURE AND FABRICATION A. Fabrication of IFPI
A femtosecond (fs) Ti:sapphire laser system (Coherent MIRA-D and RegA 9000) operating at 800-nm, 270-fs, and 250-kHz repetition rates is used for IFPI sensor fabrication. The linearly polarized laser beam was focused in the center of fiber core through a 100× oil-immersion objective (Olympus 1-U2B235, NA 1.25). A pair of laser-induced scattering points were used to form an IFPI sensor using an on-targeted energy of 160 nJ. The detailed fabrication process can be found in [16].
The IFPI sensor used in this work consists of one pair of laser-induced Rayleigh backscattering reflectors spaced 3 cm apart to form a 3-cm-long cavity. The Rayleigh backscattering profile of this IFPI device is characterized in real time during writing through an optical backscattering reflectometer (Luna OBR 4600). This OBR system can verify the location of the Rayleigh backscattering points to ensure exact cavity length. As shown in Fig. 2, the laser-induced scattering points produce ∼50-dB enhancement of backscattered power above the Rayleigh backscattering level, which is intrinsic to a standard single-mode optical fiber (Corning SMF-28e+). This increase in backscattered power from the laser-induced scattering points is sufficient to produce interference fringes for the subsequent real-time demodulation of cavity length changes and RIinduced optical path differences (OPDs) [17], [18], [19].

B. Fabrication of Fiber Taper
After the IFPI device was successfully inscribed in the single-mode fiber, the plastic jacket on the fiber was mechanically stripped and thoroughly cleaned using methanol. The cleaned IFPI fiber was then mounted onto the fiber tapering system. The two Rayleigh-enhanced scattering points were centered equidistant above and below the hot zone by observing the He-Ne alignment beam and adjusting the fiber position. During the heating process, the CO 2 laser power meter behind the protection sleeve was used to predict the approximate temperature of the ceramic tube. The incident laser power was linearly ramped up at ∼9.6 mW/s, as shown in Fig. 3(a). After 240 s, the CO 2 laser power reached 2.3 W. At this point, the upper and lower motorized stages pull in opposite directions with a speed of 0.033 mm/s. A tapered section with 16.5 mm length was produced after ∼260 s of pulling. These tapering parameters produce a fiber taper with a minimal waist diameter of 5.7 µm, as shown in Fig. 3(b) (inset). Images of the fiber are recorded by the camera during the tapering process and the diameter is calculated in real time by a machine-vision algorithm [ Fig. 3(b)].  The diameter difference before and after tapering can be clearly seen under an optical microscope, as shown in Fig. 3(c). Transmission loss fluctuates from 0.1 to 0.7 dB during the fabrication process and stabilizes at 0.3 dB once the fabrication was completed and the fiber was cooled to room temperature after 800 s, which is shown in Fig. 3(d).
Fabricating multiple devices using the same tapering algorithm showed that the fabrication process is highly consistent and repeatable, with loss from ten different devices ranging from 0.1 to 0.3 dB.

C. Tapered Fiber Optical Mode Simulation
To estimate the fraction of guided optical waves that propagate through the device as evanescent waves, a beam propagation simulation (Lumerical) was carried out. The simulation studied taper waist diameter of 5.71 µm, along with surrounding refractive indices of 1.0, 1.33, and 1.34. Fig. 4 presents the results of the simulation.
A taper with a 5.71 µm waist diameter creates an evanescent wave with 10.74% of the total transmitted optical power. This fraction increases to 12.08% when the surrounding RI was changed to 1.33, simulating submergence in water (RI = 1.33). A further 0.01 change in RI (to 1.34) introduced a 0.59% change in evanescent field power.

D. Tapered Fiber Interrogation Setup
To monitor the tapered IFPI fabrication process and to gauge RI sensing performance, a swept optical frequency interrogation system was used, as shown in Fig. 5. This interrogation system calculates the frequency and phase of the interference signal resulting from the two laser-enhanced Rayleigh scattering points to directly demodulate the optical path length of the IFPI sensor.
A LUNA Phoenix 1200 tunable laser is used as the interrogator light source. The laser can produce a 10-mW single wavelength output around 1550 nm with a linewidth of 1.5 MHz [link]. The wavelength of the laser source is set to sweep from 1545 to 1550 nm. A 1-m long "extended optical fiber" is used to calibrate the linearity of the laser wavelength sweep, and an additional gas cell is used to provide an absolute reference wavelength [20]. The optical interference intensity signal in a sweep period detected by the photodetector can be expressed as where k s is the linear sweeping rate, d 0 is the cavity length of the IFPI sensor, n is RI, c is the speed of light, λ 0 is the starting wavelength of the sweeping, and T p is the duration of a sweeping period. After Fourier transform, the demodulation result of interference signal is where sinc(·) is the sinc function. The cavity length of IFPI sensor can be derived from the phase term by where φ is the phase term of (3). OPD changes can be demodulated in real time by determining the optical phase change of (4) using the Buneman frequency estimation. The detailed principles of the demodulation algorithm for the IFPI sensor can be found in [21].

E. Detection of Optical Cavity Length Changes by IFPI
Simulation results presented in Fig. 4 can provide a rough estimation on performance of an IFPI sensor with a tapered section in the optical cavity for RI change measurements. A tapered IFPI sensor has an optical cavity n 1 L 1 + n 2 L 2 , where n 1 L 1 is the effective index and physical cavity length of untapered fiber section and n 2 L 2 is the effective index and physical cavity length of the tapered fiber section. The tapered section has 12.08% of guided light as evanescent waves, as shown in Fig. 4. An RI change n of the surrounding medium will result in a strain change of ∼ 0.12 n L 2 /(n 1 L 1 + n 2 L 2 ).
Therefore, it is important to gauge the minimal static strain that can be detected by the IFPI sensor. A commercial optical frequency domain reflectometry (OFDR) interrogation system was first used to characterize the sensitivity and stability of the IFPI sensor array. The experimental setup is shown in Fig. 6(a). One IFPI sensor with 45-cm cavity length was mounted between two fiber holders separated by 50 cm. A linear motion stage was used to exert tensile strain on the IFPI sensor. The fiber sensor was also monitored by a commercial OFDR interrogator (LUNA OBR4600), which has a minimal strain resolution of 1 µε. Fig. 6(b) shows the measurement results when 1.5 and 150 nεwere exerted on the fiber sensors. Our interrogation system detected 1.50-and 145nε strain changes, while the OBR system reached its sensitivity limit and was unable to resolve any strain change. To test the measurement stability, the IFPI was left still for 20 min. Fig. 6(c) shows the 20-min continuous measurement results, where the standard deviation of the strain change is 12.34 nε, which can be regarded as the minimal strain measurement detectable by our system. Based on the experimental results presented in Fig. 6, we could roughly estimate the minimal RI change detectable by this sensor. If one assumes that 150 nε is the minimal static strain detectable by an IFPI sensor, then a 4.6-cm-long IFPI sensor with 6-mm-long tapered section could theoretically detect the RI change of n ∼ 1.3 × 10 −6 . This estimate is based on the assumption that the entire 6-mmlong tapered fiber has a waist diameter of 5.71 µm. In reality, most of the tapered section has a larger waist diameter, which results in far less portion of light propagating as an evanescent wave. Therefore, the minimal detectable RI change must be significantly larger than n ∼ 1.3 × 10 −6 .

IV. EXPERIMENTAL RESULTS AND DISCUSSION
In addition to camera imaging and monitoring power throughput during tapering, it was also possible to monitor the tapering process by connecting the IFPI device to the interrogator setup described above before and after tapering.
A 3-cm-long IFPI device (before tapering) was used to illustrate this method. Physical separation between two laserinduced scattering points before tapering was measured by the OBR to be 29.72 mm, as shown in Fig. 7(d). The interference spectrum was measured using the custom-IFPI interrogator before and after tapering [ Fig. 7(a) and (b)]. The interference fringes therein are clearly visible. The demodulation of interference fringes before tapering yields a 29.39-mm cavity length [blue line of Fig. 7(c)], which is consistent with the OBR measurement given the resolution of the OBR system. Similar measurements performed after tapering yielded a 45.92-mm demodulated cavity length [red line of Fig. 7(c)]  and a 46.5-mm OBR-measured length, which is still consistent within the resolution of the OBR. We note that there is some increase in Rayleigh scattering along the tapered section, consistent with changing mode profiles as light traverses the taper. Using the OBR, we observed the rise of backscattering signal in the tapered section [ Fig. 7(e)]. This is likely due to physical nonuniformity produced by the tapering process as Rayleigh backscattering is extremely sensitive to fiber uniformity. Rayleigh's backscattering loss in the taper is 20-30 dB/mm above the intrinsic backscattering loss of the pristine fiber (−120 dB/mm) to reach up to −90 dB/mm, which is very weak.
In order to characterize the thermal response of the tapered IFPI device prior to RI sensing, the tapered IFPI sensor is placed into a furnace where the temperature is gradually increased from 25.2 • C to 52.3 • C and then cooled down to 27.2 • C in air. The demodulated OPD is plotted during heating and cooling in Fig. 8. The sensitivity of thermal influence is measured by temperature-induced OPD change of the IFPI device, which is 520 nm/ • C. This is consistent with IFPI sensors without tapered sections. Because the IFPI interrogator is capable of making quick measurements, it is possible to record RI data much faster than slow temperature change. An untapered IFPI sensor can also be used to measure temperatures to mitigate temperature sensor drift [19].
To perform RI measurement using the tapered IFPI sensors, a series of glucose solutions with different concentrations was prepared. The RI of each solution was measured using a refractometer (Cole-Parmer). The tapered IFPI device was mounted to a glass slide, and each solution was pipetted dropwise onto the device. Fig. 9 shows the operation of the sensor demodulation algorithm. Instead of "eyeballing" interference fringe shift as a result of RI change [22], [23], [24], [25], the reflection spectra were digitally acquired. A fast Fourier Transform (FFT) was used to convert reflection spectra from the wavelength domain to optical cavity length, as shown in Fig. 9(b). The imaginary part of the FFT reveals phase changes as a result of the RI change. The phase changes induced by RI changes were then converted into OPD. Fig. 10 compares the measurement repeatability of a single sensor tested in solutions with different refractive indices and measurement repeatability of multiples sensors tested in identical solutions. Three sensors (sensors 1, 2, and 3) were used for the test. Their optical cavity lengths are 45.92, 46.12, and 46.09 mm. Fig. 10(a) plots the OPD with respect to ambient RIs for sensor_1. Between each measurement, the tapered IFPI sensor is rinsed thoroughly with Deionized water (DI water) and air dried. A total of 38 solutions were used for the test; each solution has distinct RI values determined by the dilution of refractometer-measured glucose solutions. The smallest RI difference among solution is ∼2 × 10 −5 . One example, sensor_1 test result around RI 1.337 involved three solutions with refractive indices of 1.33670, 1.33675, and 1.337. Three devices are tested three times under certain RI liquids, and the variation of each device is around 5-50 nm, as shown in Fig. 10(b). The measurement variation produced by different IFPI sensors revealed in Fig. 9 could come from the limit of sensor demodulation algorithm or small temperature variation during the experiments. Fig. 10(b) shows that the variation of OPD measured by different IFPI sensors could be up to 50 nm. This corresponds to ∼1.074-µε strain changes. An RI resolution of ∼2 × 10 −5 was measured, as shown in Fig. 10. Furthermore, this tapered IFPI sensor only requires a microliter of liquid to perform the measurement.

V. CONCLUSION
This article presents a highly reproducible method to fabricate multiplexable tapered IFPI fiber sensors. Using a CO 2 laser heating process controlled by digital mirrors and high-precision motorized systems, a stable heating zone with carefully tailored temperature profile was created to produce tapered sections in silica fibers with desired profiles. Enabled by the well-engineered laser-heated tapering process, a tapered section with a minimal waist diameter of 5.71 µm was created inside an IFPI sensor with less than 0.3-dB insertion loss at 1550 nm. A swept optical frequency-domain reflectometer was used to characterize the tapered IFPI sensors, which achieved a minimal RI sensing resolution of ∼2 × 10 −5 . This article shows that highly multiplexable tapered fiber sensors can be achieved through an integrated laser fabrication process.

ACKNOWLEDGMENT
This work was funded by the Department of Energy, National Energy Technology Laboratory, an agency of the United States Government, through a support contract with Leidos Research Support Team (LRST). Neither the United States Government nor any agency thereof, nor any of their employees, nor LRST, nor any of their employees, makes any warrant, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.