Optical Graphene Gas Sensors Based on Microfibers: A Review

Graphene has become a bridge across optoelectronics, mechanics, and bio-chemical sensing due to its unique photoelectric characteristics. Moreover, benefiting from its two-dimensional nature, this atomically thick film with full flexibility has been widely incorporated with optical waveguides such as fibers, realizing novel photonic devices including polarizers, lasers, and sensors. Among the graphene-based optical devices, sensor is one of the most important branch, especially for gas sensing, as rapid progress has been made in both sensing structures and devices in recent years. This article presents a comprehensive and systematic overview of graphene-based microfiber gas sensors regarding many aspects including sensing principles, properties, fabrication, interrogating and implementations.


Introduction
Microfibers with sub-wavelength geometry and large index contrast between the fiber core and surroundings [1,2] has become a widely used technique in fiber optics, benefiting from its outstanding properties: evanescent field excitation, electromagnetic confinement, low transmission loss, and coupling convenience [3,4]. Among diverse microfiber applications, optical sensing is very interesting due to its potential of realizing miniaturized fiber optic sensors with small footprint, high sensitivity, fast response, good flexibility, and low power consumption [5,6]. Especially in recent years, people have found that microfibers can work as an ideal substrate for realizing high-performance chemical sensors incorporated with two-dimensional optoelectronic materials, such as graphene or graphene oxide.
Among them, graphene-based gas detection is one of the most creative and successful applications, showing the potential to achieve ultimate sensitivity: single molecule gas detection [30]. Moreover, benefiting from its atomic thickness with ultrahigh conductivity, graphene can also realize remarkable functionalities, such as wearable sensors for smart systems [31][32][33] and switching-sensing devices for high-precision measurements with thermal compensation [34,35].

Principles of Graphene-Based Gas Sensing on Microfibers
The electronic conductive band and valence band crosses at the Dirac point, as shown in Figure 1a [36]. The Fermi level of graphene can be simply described by using the dispersion relation E ± (κ) = ±hν F |κ|, where κ is the wave vector, ν F ≈ 10 6 m/s is the Fermi velocity, andh is Planck's constant [37]. When the Fermi level is higher than the Dirac point, graphene is N-doped; otherwise it is P-doped. Accordingly, the optical conductivity of graphene can be written as [38]: Specifically, the intraband conductivity and the interband conductivity can be approximately separated as: The theoretically calculated results are shown in Figure 1b. Here E F is the quasi Fermi level, directly determined by the external bias. f is the optical frequency, τ ≈ 10 −13 s is the carrier relaxation lifetime, T is the temperature, f d ( ) = {exp[( -µ)/k B T] + 1} −1 is the Fermi-Dirac distribution, h = 1.05 × 10 −34 J·s, k B = 1.3806505 × 10 −23 J/K is Boltzmann's constant, and e = −1.6 × 10 −19 C is the unit charge. External electric tuning majorly influences the σ g,inter , which is also directly related to the saturable absorption, which is driven by the photoexcited electron kinetics [39]. Considering the fact that graphene has an atomically thick planar waveguide with such a complex sheet conductivity, we write its effective optical permittivity as where Re(σ g ) and Im(σ g ) are the real and imaginary parts of σ g , respectively. By regarding the graphene monolayer thickness ∆ = 0.4 nm, the refractive index of the graphene layer can be derived from [40]: Re n g + iIm(n g ) 2 = Re g + iIm g . For media modes with real permittivity, we can calculate the relationship as Figure 1c maps the numerically calculated results of the complex permittivity and complex refractive index. These results directly indicate the optical dispersion of graphene-based microfiber structures, which was experimentally demonstrated by Yao et al. in 2013 [41].  Once incorporated with a microfiber, the graphene performs as a part of the cladding, and can influence the optical transmission via mode effective refractive index modulation. According to different graphene-microfiber hybrid geometries, the mode field distributions and complex effective index neff can be numerically simulated or calculated via finite element method (FEM) [43,44]. Figure 2a illustrates the idea of modelling the graphene-based microfibers for effective index approximation. Figure 2b provides examples of the spatial distributions of electric field intensity (fundamental mode) for different types of graphene-based microfiber structures. Upper panels: graphene-wrapped microfiber with core diameter ≈ 0.5 µm and 1 µm [43,45], lower panels: graphene-wrapped microfiber with core diameter ≈ 8 µm while cladding thickness ≈ 2 µm [46]. The mode effective index of a hybrid waveguide is also influenced by the optoelectronic dynamics of graphene. Figure 2c plots the simulated "neff vs. |EF|" correlation of a graphene-based microfiber with core diameter ≈ 8 µm. It can be seen that by changing the Fermi level of graphene, both the transmission phase and the transmission loss of a graphene-based microfiber could be modified remarkably. Such a mechanism can not only be applied to sensing, but has also been widely used for fast optical modulation [47,48].
It is known that the electromagnetic field distributed along microfibers obeys the Bessel equations in cylindrical coordinates [49], which has a general solution Here (a,b) are reproduced from Refs. [36,42], respectively.
Once incorporated with a microfiber, the graphene performs as a part of the cladding, and can influence the optical transmission via mode effective refractive index modulation. According to different graphene-microfiber hybrid geometries, the mode field distributions and complex effective index n eff can be numerically simulated or calculated via finite element method (FEM) [43,44]. Figure 2a illustrates the idea of modelling the graphene-based microfibers for effective index approximation. Figure 2b provides examples of the spatial distributions of electric field intensity (fundamental mode) for different types of graphene-based microfiber structures. Upper panels: graphene-wrapped microfiber with core diameter ≈ 0.5 µm and 1 µm [43,45], lower panels: graphene-wrapped microfiber with core diameter ≈ 8 µm while cladding thickness ≈ 2 µm [46]. The mode effective index of a hybrid waveguide is also influenced by the optoelectronic dynamics of graphene. Figure 2c plots the simulated "n eff vs. |E F |" correlation of a graphene-based microfiber with core diameter ≈ 8 µm. It can be seen that by changing the Fermi level of graphene, both the transmission phase and the transmission loss of a graphene-based microfiber could be modified remarkably. Such a mechanism can not only be applied to sensing, but has also been widely used for fast optical modulation [47,48].
It is known that the electromagnetic field distributed along microfibers obeys the Bessel equations in cylindrical coordinates [49], which has a general solution where A is the optical amplitude, Re(n eff ) and Im(n eff ) are the real and imaginary parts of the graphene-based microfiber, z is the transmission distance, t is the time delay, c is the light velocity in vacuum, and ω is the optical frequency. As a conclusion, the Fermi-level modification of graphene would modulate both the output phase and output power of the microfiber. In optical nonlinear processes, such an effect also alters the phase-matching and nonlinear threshold [50]. where A is the optical amplitude, Re(neff) and Im(neff) are the real and imaginary parts of the graphenebased microfiber, z is the transmission distance, t is the time delay, c is the light velocity in vacuum, and ω is the optical frequency. As a conclusion, the Fermi-level modification of graphene would modulate both the output phase and output power of the microfiber. In optical nonlinear processes, such an effect also alters the phase-matching and nonlinear threshold [50].  When gas molecules are adsorbed on the surface of graphene, the Fermi level of graphene is tuned [54,55]. Figure 3a shows chemical connections between graphene and gas molecules  where A is the optical amplitude, Re(neff) and Im(neff) are the real and imaginary parts of the graphenebased microfiber, z is the transmission distance, t is the time delay, c is the light velocity in vacuum, and ω is the optical frequency. As a conclusion, the Fermi-level modification of graphene would modulate both the output phase and output power of the microfiber. In optical nonlinear processes, such an effect also alters the phase-matching and nonlinear threshold [50].  When gas molecules are adsorbed on the surface of graphene, the Fermi level of graphene is tuned [54,55]. Figure 3a shows chemical connections between graphene and gas molecules When gas molecules are adsorbed on the surface of graphene, the Fermi level of graphene is tuned [54,55]. Figure 3a shows chemical connections between graphene and gas molecules schematically. For pristine graphene film, the absorption of polar molecules is dominant, hence pristine graphene had been demonstrated with high electronic sensitivity to polar gases such as NH 3 , NO 2 , and H 2 O. It is worth pointing out that for graphene oxide (GO) or other functionalized graphene, non-polar gas such as H 2 is also detectable [56]. Figure 3b plots the measured "gas adsorption vs. conductivity" for graphene film, reported in Ref. [30]. As expounded above, the "gas adsorption" and the "conductivity-index relationship" of graphene form the basis of the optical graphene gas sensing. By using microfibers, the gas adsorption can be detected via optical interference, power metering, or nonlinearity excitation, as shown in Figure 3c.

Design and Fabrication of Graphene-Based Microfiber Structures
Since the first graphene-based microfiber optical gas sensor reported in 2012 [57], a variety of graphene-microfiber hybrid structures have been investigated for gas sensing applications. Their fabrication process can be summarized with the following three steps: (1) microfiber fabrication; (2) graphene growth; (3) graphene-microfiber installation.
Typically, microfibers are fabricated from commercial glass fibers or bulks, by using the fusing & drawing method [1], which can be controlled either manually or automatically. Figure 4a shows the microscopic pictures of typical microfibers. The heater source can be either a flame (usually a hydrogen flame), an electrical heater, or a laser-heated tube [58][59][60]. Figure 4b demonstrates a typical setup for fusing & drawing microfibers from glass fibers. Optical loss of a microfiber is mainly determined by its taper quality. Very recently, taking advantage of the "flame-brushing" technique, high-quality silica micro-nano fibers with diameters in the range of 800 nm to 1.3 µm, unevenness < 5 nm, and waist length larger than 30 cm have been achieved [61]. schematically. For pristine graphene film, the absorption of polar molecules is dominant, hence pristine graphene had been demonstrated with high electronic sensitivity to polar gases such as NH3, NO2, and H2O. It is worth pointing out that for graphene oxide (GO) or other functionalized graphene, non-polar gas such as H2 is also detectable [56]. Figure 3b plots the measured "gas adsorption vs. conductivity" for graphene film, reported in Ref. [30]. As expounded above, the "gas adsorption" and the "conductivity-index relationship" of graphene form the basis of the optical graphene gas sensing. By using microfibers, the gas adsorption can be detected via optical interference, power metering, or nonlinearity excitation, as shown in Figure 3c.

Design and Fabrication of Graphene-Based Microfiber Structures
Since the first graphene-based microfiber optical gas sensor reported in 2012 [57], a variety of graphene-microfiber hybrid structures have been investigated for gas sensing applications. Their fabrication process can be summarized with the following three steps: (1) microfiber fabrication; (2) graphene growth; (3) graphene-microfiber installation.
Typically, microfibers are fabricated from commercial glass fibers or bulks, by using the fusing & drawing method [1], which can be controlled either manually or automatically. Figure 4a shows the microscopic pictures of typical microfibers. The heater source can be either a flame (usually a hydrogen flame), an electrical heater, or a laser-heated tube [58][59][60]. Figure 4b demonstrates a typical setup for fusing & drawing microfibers from glass fibers. Optical loss of a microfiber is mainly determined by its taper quality. Very recently, taking advantage of the "flame-brushing" technique, high-quality silica micro-nano fibers with diameters in the range of 800 nm to 1.3 µm, unevenness < 5 nm, and waist length larger than 30 cm have been achieved [61]. Sometimes one not only wants a microfiber-guiding evanescent field, but also hopes to keep the inner microstructures of the original fiber (such as photonic crystals and Bragg gratings); then, chemical etching becomes a choice [67,68]. For example, for silica-based fiber etching, hydrofluoric acid is commonly used as the corrosive agent. Compared to microfibers fabricated by fusing & drawing, the chemically etched microfibers have larger average diameters, usually in the range of 8-12 µm. Figure 4c shows this method and pictures of the chemically etched microfiber samples. In addition, for further specific purposes, microfibers are not only fabricated from glass materials, but Sometimes one not only wants a microfiber-guiding evanescent field, but also hopes to keep the inner microstructures of the original fiber (such as photonic crystals and Bragg gratings); then, chemical etching becomes a choice [67,68]. For example, for silica-based fiber etching, hydrofluoric acid is commonly used as the corrosive agent. Compared to microfibers fabricated by fusing & drawing, the chemically etched microfibers have larger average diameters, usually in the range of 8-12 µm. Figure 4c shows this method and pictures of the chemically etched microfiber samples. In addition, for further specific purposes, microfibers are not only fabricated from glass materials, but also other materials such as polymers, silicon, and metal-oxides, by using diverse means such as thermal process, chemical growth, and self-assembly [62][63][64][65][66]. Figure 4d shows examples of these special microfibers.
The chemical vapor deposition (CVD) and wet transfer techniques provide inch-level large-scale monolayer graphene for incorporation with microfibers [69][70][71][72]. Figure 5a illustrates a schematic view of a growing graphene film on a copper foil. Graphene on Cu is grown by the decomposition of CH 4 gas in a dilute H 2 environment over the surface at 1000 • C. With the exposure of Cu foil in a CH 4 /H 2 environment, the nucleation of graphene islands starts taking place randomly, eventually aggregating into a continuous graphene film. Figure 5b shows the optical image and pictures of CVD graphene film samples. CVD graphene has good uniformity in inch-size large-scale, which has been widely used in optoelectronic devices, via the wet transfer technique. Graphene films can also be deposited on fiber structures directly by reducing from GO in liquid. Figure 5c shows the chemical structures of GO and graphene; there are many functional groups containing oxygen connecting the carbon atoms [73]. Figure 5d shows pictures of liquid dispersions of GO and graphene reduced from GO [74]. Compared to CVD graphene films, the reduced GO film usually has more defects, but the deposition of the GO film does not need an additional transferring process, which is more convenient for implementation. In recent years, by optimizing the solution-based reduction method, large-area GO films with acceptable uniformity can be obtained [75]. also other materials such as polymers, silicon, and metal-oxides, by using diverse means such as thermal process, chemical growth, and self-assembly [62][63][64][65][66]. Figure 4d shows examples of these special microfibers. The chemical vapor deposition (CVD) and wet transfer techniques provide inch-level large-scale monolayer graphene for incorporation with microfibers [69][70][71][72]. Figure 5a illustrates a schematic view of a growing graphene film on a copper foil. Graphene on Cu is grown by the decomposition of CH4 gas in a dilute H2 environment over the surface at 1000 °C. With the exposure of Cu foil in a CH4/H2 environment, the nucleation of graphene islands starts taking place randomly, eventually aggregating into a continuous graphene film. Figure 5b shows the optical image and pictures of CVD graphene film samples. CVD graphene has good uniformity in inch-size large-scale, which has been widely used in optoelectronic devices, via the wet transfer technique. Graphene films can also be deposited on fiber structures directly by reducing from GO in liquid. Figure 5c shows the chemical structures of GO and graphene; there are many functional groups containing oxygen connecting the carbon atoms [73]. Figure 5d shows pictures of liquid dispersions of GO and graphene reduced from GO [74]. Compared to CVD graphene films, the reduced GO film usually has more defects, but the deposition of the GO film does not need an additional transferring process, which is more convenient for implementation. In recent years, by optimizing the solution-based reduction method, large-area GO films with acceptable uniformity can be obtained [75]. Graphene can be either attached on or wrapped around a microfiber, as schematically shown in Figure 6a. Figure 6b sketches the fabrication flows of a graphene-based fiber structure by using CVD technique [41]. For graphene attached on microfibers, graphene is transferred to a low refractive index substrate; afterwards, microfibers are put on the graphene, kept contacted. Such a van der Waals contact is firm and stable [41]. For graphene-wrapped microfibers, a substrate to carry graphene is unnecessary. The wet transfer of a CVD graphene film is done as follows: spin-coating a layer of polymethyl methacrylate (PMMA) on the surface of the graphene, forming the PMMA/graphene/Cu sandwich-like hybrid; dissolving the Cu under graphene by using FeCl3 solution; covering the PMMA/graphene flexible film on a substrate or wrapping it on a microfiber; removing the PMMA by using acetone, leaving only graphene remaining. Figure 6c shows the process of depositing a reduced GO film on the surface of a microfiber [74]. The process involves oxidizing graphite powder to GO by using strong oxidants such as potassium permanganate; immersing a microfiber in the GO dispersion liquid; reducing the GO to be graphene Graphene can be either attached on or wrapped around a microfiber, as schematically shown in Figure 6a. Figure 6b sketches the fabrication flows of a graphene-based fiber structure by using CVD technique [41]. For graphene attached on microfibers, graphene is transferred to a low refractive index substrate; afterwards, microfibers are put on the graphene, kept contacted. Such a van der Waals contact is firm and stable [41]. For graphene-wrapped microfibers, a substrate to carry graphene is unnecessary. The wet transfer of a CVD graphene film is done as follows: spin-coating a layer of polymethyl methacrylate (PMMA) on the surface of the graphene, forming the PMMA/graphene/Cu sandwich-like hybrid; dissolving the Cu under graphene by using FeCl 3 solution; covering the PMMA/graphene flexible film on a substrate or wrapping it on a microfiber; removing the PMMA by using acetone, leaving only graphene remaining. Figure 6c shows the process of depositing a reduced GO film on the surface of a microfiber [74]. The process involves oxidizing graphite powder to GO by using strong oxidants such as potassium permanganate; immersing a microfiber in the GO dispersion liquid; reducing the GO to be graphene by   To characterize the quality of graphene on the microfibers, Raman spectroscopy [77][78][79], scanning electron microscopy (SEM), scattering measurement, and X-ray photoelectron spectroscopy (XPS) [80] are commonly applied. Figure 7a-d show the pictures of graphene-based microfibers, measured by optical microscopy and SEM. One can check the quality of the microfiber-graphene   To characterize the quality of graphene on the microfibers, Raman spectroscopy [77][78][79], scanning electron microscopy (SEM), scattering measurement, and X-ray photoelectron spectroscopy (XPS) [80] are commonly applied.  To characterize the quality of graphene on the microfibers, Raman spectroscopy [77][78][79], scanning electron microscopy (SEM), scattering measurement, and X-ray photoelectron spectroscopy (XPS) [80] are commonly applied. based on these images. Figure 7e plots a typical Raman spectrum of graphene on a fiber structure. Commonly, high-quality graphene on microfibers should have an ignorable D peak, and a G/2D ratio 0.3~0.5; the locations of the G peak and the 2D peak are influenced by the graphene doping. Commonly, CVD graphene is pure in chemistry, while graphene film reduced from GO may contain more functional groups, such as -OH and -COOH. These functional groups can be extremely useful for specific sensing applications; XPS is helpful to check them. Figure 7f plots the XPS of CVD graphene and reduced GO. For better deposition on microfibers, the reductions are usually controlled to make a C:O ratio > 4 [74].

Graphene Gas Sensors with Microfibers
In recent years, graphene gas sensors with microfibers develop rapidly, some of them are also reviewed in Refs. [81][82][83][84]. The roadmap demonstrates that the sensitivity of these graphene-based microfiber gas sensors increases from parts per kilo (ppk) to part per billion (ppb) by gradually optimizing the sensing structure.
Based on the graphene-attached microfiber scheme, we reported gas sensors based on both optical intensity detection and interferometric demodulation by using mode field analysis or Mach-Zehnder Interferometer (MZI) [52,84]. Figure 8a,b show their implementations. A microfiber with a length of several centimeters and ≈1 µm diameter was used to couple light interacting with graphene. Such a scheme is similar to a graphene-based D-shaped fiber [85,86], but is easier to manipulate. In Ref. [84], we illustrated that the polarization-dependent transmission of the graphene-attached microfiber was sensitive to gas adsorptions, especially large molecules. As Figure 8c shows, the adsorption of acetone gas molecules would dramatically dampen the optical transmission; for example, for acetone gas with a concentration of 1150 ppm, the transmission loss increased over 3 dB, the maximum sensitivity of this structure was about 0.3 dB/ppk.
The regeneration time for the acetone gas detection was in the rage of several minutes. The interferometer scheme was a big step forward to achieve a much higher sensitivity. Figure 8d shows the performance of the graphene-microfiber-based MZI for NH 3 gas detection [52]. It illustrated both high sensitivity and fast response. For NH 3 trace detection, 0.3 ppm resolution and 0.5 s response delay was achieved. incorporation based on these images. Figure 7e plots a typical Raman spectrum of graphene on a fiber structure. Commonly, high-quality graphene on microfibers should have an ignorable D peak, and a G/2D ratio 0.3~0.5; the locations of the G peak and the 2D peak are influenced by the graphene doping. Commonly, CVD graphene is pure in chemistry, while graphene film reduced from GO may contain more functional groups, such as -OH and -COOH. These functional groups can be extremely useful for specific sensing applications; XPS is helpful to check them. Figure 7f plots the XPS of CVD graphene and reduced GO. For better deposition on microfibers, the reductions are usually controlled to make a C:O ratio >4 [74].

Graphene Gas Sensors with Microfibers
In recent years, graphene gas sensors with microfibers develop rapidly, some of them are also reviewed in Refs. [81][82][83][84]. The roadmap demonstrates that the sensitivity of these graphene-based microfiber gas sensors increases from parts per kilo (ppk) to part per billion (ppb) by gradually optimizing the sensing structure.
Based on the graphene-attached microfiber scheme, we reported gas sensors based on both optical intensity detection and interferometric demodulation by using mode field analysis or Mach-Zehnder Interferometer (MZI) [52,84]. Figures 8a,b show their implementations. A microfiber with a length of several centimeters and ≈1 µm diameter was used to couple light interacting with graphene. Such a scheme is similar to a graphene-based D-shaped fiber [85,86], but is easier to manipulate. In Ref. [84], we illustrated that the polarization-dependent transmission of the graphene-attached microfiber was sensitive to gas adsorptions, especially large molecules. As Figure 8c shows, the adsorption of acetone gas molecules would dramatically dampen the optical transmission; for example, for acetone gas with a concentration of 1150 ppm, the transmission loss increased over 3 dB, the maximum sensitivity of this structure was about 0.3 dB/ppk. The regeneration time for the acetone gas detection was in the rage of several minutes. The interferometer scheme was a big step forward to achieve a much higher sensitivity. Figure 8d shows the performance of the graphene-microfiber-based MZI for NH3 gas detection [52]. It illustrated both high sensitivity and fast response. For NH3 trace detection, 0.3 ppm resolution and 0.5 second response delay was achieved.  Here the results are reproduced from Ref. [52,84]. Similarly, Figure 9b shows the experimental results [87]. Sridevi et al. chose reduced GO rather than CVD graphene, getting better selectivity for NO 2 gas. A sensitivity of 0.5 ppm was achieved in this work. As the reduced GO film was thicker and fluffier than the CVD graphene, its response time was limited to minute-level. In this type of sensor, it is also important to enhance the mode distribution ratio out of the core and optimize the microfiber diameter (or the thickness of the remained cladding), as discussed by Zhang et al. in Ref. [88]. Figure 9c replots the experimental results: a larger microfiber diameter brought lower sensitivity, but also lower attenuation and larger dynamic range. Considering a composite index H~ln(S)ln(D)/A, the best diameter could be around 10 µm for graphene-wrapped silica fiber-Bragg-gratings (FBGs) with standard 8 µm core.
Since the scheme of graphene-microfiber attachment spatially limits the light-graphene interaction, since 2014, graphene-wrapped microfiber structures have become a trend. For instance, based on graphene-wrapped micro fiber Bragg gratings (MFBGs), the footprint of the sensors can be dramatically miniaturized. Wu et al. [89] and Sridevi et al. [87] reported ultrasensitive sensors based on MFBGs for NH 3 gas sensing and NO 2 gas sensing, respectively. As shown in Figure 9a Similarly, Figure 9b shows the experimental results [87]. Sridevi et al. chose reduced GO rather than CVD graphene, getting better selectivity for NO2 gas. A sensitivity of 0.5 ppm was achieved in this work. As the reduced GO film was thicker and fluffier than the CVD graphene, its response time was limited to minute-level. In this type of sensor, it is also important to enhance the mode distribution ratio out of the core and optimize the microfiber diameter (or the thickness of the remained cladding), as discussed by Zhang et al. in Ref. [88]. Figure 9c replots the experimental results: a larger microfiber diameter brought lower sensitivity, but also lower attenuation and larger dynamic range. Considering a composite index H ~ ln(S)ln(D)/A, the best diameter could be around 10 µm for graphene-wrapped silica fiber-Bragg-gratings (FBGs) with standard 8 µm core.
Since the scheme of graphene-microfiber attachment spatially limits the light-graphene interaction, since 2014, graphene-wrapped microfiber structures have become a trend. For instance, based on graphene-wrapped micro fiber Bragg gratings (MFBGs), the footprint of the sensors can be dramatically miniaturized. Wu et al. [89] and Sridevi et al. [87] reported ultrasensitive sensors based on MFBGs for NH3 gas sensing and NO2 gas sensing, respectively. As shown in Figure 9a  Another method to enhance light-graphene interaction in graphene-based microfiber structures is to excite high-order mode propagation with larger mode-field area or to excite plasmons. In 2014, Yao et al. demonstrated a graphene-based microfiber multimode interferometer, as shown in Figure  10a [46]. In this structure, the HE21 mode is more sensitive to local refractive index alteration than the in-core HE11 [90], hence its interference Free Spectrum Range (FSR) could be tuned by gas adsorption, resulting in a spectral resonance dip shift. In a sensing experiment, ~0.1 ppm for NH3 gas detection Another method to enhance light-graphene interaction in graphene-based microfiber structures is to excite high-order mode propagation with larger mode-field area or to excite plasmons. In 2014, Yao et al. demonstrated a graphene-based microfiber multimode interferometer, as shown in Figure 10a [46]. In this structure, the HE 21 mode is more sensitive to local refractive index alteration than the in-core HE 11 [90], hence its interference Free Spectrum Range (FSR) could be tuned by gas adsorption, resulting in a spectral resonance dip shift. In a sensing experiment,~0.1 ppm for NH 3 gas detection and~0.2 ppm for H 2 O vapor detection were achieved. By using GO-ZnO film, Hu et al. also recently realized a NH 3 gas sensor, demonstrating sub-ppm sensitivity, as shown in Figure 10b [91]. This GO-ZnO incorporated interferometric optical microfiber illustrated very high selectivity to NH 3    Limited by the spectral resolution of the optical interference and the linear loss of the passive devices, sub-ppm seemed to be the limitation for a graphene-based microfiber gas sensor. Taking    Limited by the spectral resolution of the optical interference and the linear loss of the passive devices, sub-ppm seemed to be the limitation for a graphene-based microfiber gas sensor. Taking Limited by the spectral resolution of the optical interference and the linear loss of the passive devices, sub-ppm seemed to be the limitation for a graphene-based microfiber gas sensor. Taking advantage of high Q resonance, the interferometric resolution can be effectively improved. Yu et al. made such an attempt by building a GO-deposited microfiber knot resonator [93]. Figure 11a shows the structural diagram. In this study, the GO film covered the whole microfiber resonator, limiting the Q factor. In future investigations, by optimizing the graphene coverage region [94], the resolution and detection limit can potentially be further improved.
Moreover, in 2017, based on graphene-enhanced Brillouin scattering, a microfiber interrogated Whispering Gallery Mode ptomechanical gas sensor was realized. Figure 11b shows the design and the sensing performance. It reported an unprecedented high sensitivity (1 ppb) for NH 3 gas detection, which is a globally leading number comparable to other advanced techniques [95,96]. Such a remarkable breakthrough revealed the "electron-phonon-photon" interaction in the graphene-based optomechanical resonator, going beyond all the conventional graphene-based optical or solid-state sensors. Optomechanical resonance was generated via Brillouin phase matching and nonlinear gain: where v A is the acoustic velocity, c is the light velocity in vacuum, n p and n s are the effective indexes of the pump mode and the generated Stokes mode, f p and f s are the pump frequency and the Stokes frequency, γ e is the electro-strictive coefficient, and Γ B is the lifespan of the phonons. Gas adsorption on the reduced GO film enables an RF spectral shift 200 kHz/ppm, while the uncertainty of the Brillouin optomechanical generation is only 200 Hz, due to the extremely high Q factor (10 6 ). Moreover, such an optomechanical micro-resonator kept an exceptional dynamic range from 1 ppb to 400 ppm, crossing over five orders. The above review summarizes the progress in optical graphene gas sensors based on microfibers. By optimizing the optical sensing structures, this type of sensor has made significant advancements in sensitivity, which is summarized in Table 1. Benefitting from the advancements in novel optical detection methods, the current performance of microfiber-based optical graphene gas sensors has been comparable to the state-of-the-art gas detection techniques, as displayed in Table 2. Table 1. Major progress in the optical graphene gas sensors based on microfibers.

Year
Sensor Structure Target Gas Performance Reference

Conclusions and Outlook
In this article, we review the principles, fabrications, implementations, and performances of optical graphene gas sensors with microfibers, which have attracted intense interest in research and development and play an important role in industry. The mode of "graphene and microfiber" has become a widely used platform for-but not limited to-gas detections. In this way, higher sensitivity and better selectivity are constant pursuits, while still remaining challenges. On one hand, to increase the sensitivity, more and more new optical mechanisms and techniques are being reported, such as graphene-based laser sensing [98], high-order nonlinearity-based enhancement [42], and plasmonic sensing in the Mid-Infrared Range to THz region [20]. On the other hand, determined by the graphene's nature that it can interact with any gas molecule, most of the above optical gas sensors are focused on polar gas sensing, such as NH 3 , H 2 O, or NO 2 , and these sensors lack selectivity. Towards the realization of graphene-based microfiber sensors for other gas detection, the graphene material itself would be further functionalized, such as by using element-doped films [99], graphene grains [100], or fluorescent resonance energy transformation technology [74]. With the progress in both the microfibers and graphene materials, there is no doubt that more graphene-based fiber-optic gas sensing structures and devices can be foreseen to meet the practical application requirements.