1. Introduction

Graphene, a two-dimensional sheet of sp²-bonded carbon atoms, has many outstanding properties such as extremely high carrier mobility (∼10000 cm²V⁻¹s⁻¹) [1], huge area-mass ratio (theoretically 2630 m²/g) [2], high thermal conductivity (∼5000 Wm⁻¹K⁻¹) [3], low Johnson as well as 1/f noise [4]. Because of its unique atomic structure, graphene is an ideal candidate material for sensing gas molecule. Prior works show that graphene demonstrates ultra high sensitivity to gas molecule and can sense only one gas molecule [4]. Reduced graphene oxide (rGO), a type of low-cost and easy-preparation graphene, has been used as electrochemistry gas sensors frequently [5,6]. Recently, there is a growing interests in graphene-based novel optical devices, particularly with reports about electro-optical [7,8] and all-optical modulators [9–12], polarizers [13–16], and gas sensors [17–26]. The main advantages of graphene-waveguide structures for gas sensing include small size, remote sensing capabilities, and immunity to electromagnetic interference, etc. Nowadays, the reported graphene/waveguide-based gas sensors can be divided into about four types: SPR [17,18], phase-modulation [19–21], power (or intensity)-modulation [22–25], photoacoustic spectroscopy [26].

Toluene is a volatile organic compound (VOC) that may jeopardize health of people. A sensitive toluene sensor is an effective fundamental device to monitor and control the concentration of toluene efficiently. Various sensitive material have been proposed to fabricate toluene sensors, for instance, metal oxide (MOX) [27], intrinsically conductive polymer (ICP) [28], carbon nanotube (CNT) [29] for resistive-based sensors, and fluorosiloxane polymer [30], Zeolite thin film [31], calixarene [32], porous silica xerogel film [33], ZnO nanorod [34] for fiber-optic sensors. However, they all have their pros and cons. Although MOX demonstrates high sensitivity, its functionality operation relies on high temperature (usually hundreds of centigrade degree). ICP works at room temperature but it faces with the problem of low sensitivity or liable to deteriorate. CNT must be specially decorated and its response time is relatively long. Most sensitive materials used in fiber-optic toluene sensors possess porous structure. Pores of improper sizes may deteriorate the performance in terms of response and recovery. Therefore, it is of great importance to find alternative sensitive material for low-consumption, low-cost, fast, sensitive, and highly reversible toluene detection.

In this paper, a graphene-clad side-polished fiber (GCSPF) is fabricated by using rGO and SPF. For the first time, a comprehensive study both in theory and via experiments is carried out on its toluene sensing property, such as sensitivity, reversibility, response time, concentration detection limit, and sensing range of gas concentration. Based on the theoretical model of sensing mechanism, the influence of the doping level of graphene on the sensitivity of sensor is also estimated.

2. Fabrication and characterization of the sensor

The SPF used in the sensor was fabricated with single mode optical fiber (SMF) by using wheel polishing technique [35]. The polished depth is 60 μm (that is into the core with polished depth of ∼1.5 μm) and the polished length is 1 cm. By using two-step chemical method, rGO was fabricated with the size of 2 μm × 4 μm per flake and conductivity of 1.43 × 10² S/m [36]. By using the evaporation-deposition method similar to [24], Xiao et al, rGO film with thickness of 200~1000nm was deposited on the polished surface of the fiber. The sketch maps of GCSPF and its cross section are shown in Figs. 1(a) and 1(b) respectively. Figure 1(c) is the SEM photograph of the rGO film-coated fiber polished surface taken by FESEM (ultra55, ZEISS). It can be seen from Fig. 1(c) that the coated rGO film has nearly uniform thickness but with some folds distributed on the surface. The detachment of a small part of the rGO film from the right side of the fiber polished surface is caused by the destructive broken of the polished segment. The thickness of the rGO film coated on the polished surface can be measured by SEM photograph of the cross section of GCSPF and its zoom-in, as shown in Fig. 1(d).

 

Fig. 1 (a) Sketch maps of GCSPF and (b) its cross section; (c) SEM photograph of rGO film-coated fiber polished surface with a small section of rGO film on the right side detached; (d) zoom-in of the SEM photograph of the cross section of GCSPF.

Download Full Size | PDF

The Raman spectrum of rGO flakes on the polished surface of the fiber is measured at room temperature with a RENISHAW inVia Raman Microscope and laser excitation at 514 nm, and the result is shown in Fig. 2(a). The D and G peak positions are at 1348 cm⁻¹ and 1600 cm⁻¹, respectively. The prominent D peak (absent in mechanically cleaved graphene) indicates the presence of structural imperfections induced by the boundary effect [37] and the attachment of residual hydroxyl and epoxide groups on the carbon basal plane [36]. The smaller intensity of 2D peak with respect to the D and G peaks is due to disorder [6]. Nevertheless, the number of rGO flake layers is estimated in a non-destructive manner by using the shift of this peak. By comparing the 2D band center (2710 cm⁻¹) in Fig. 3(a) with [37], Ferrari et al, the rGO flakes, which constitute the “thick” rGO film coated on the polished surface of the fiber, is estimated to have less than 10 atom layers. The X-ray diffraction (XRD) pattern of rGO flakes on the polished surface of the fiber is also measured at room temperature with a Rigaku MiniFlex 600 X-ray diffractometer, and the result is shown in Fig. 2(b). The broad peak at 2θ = 24.8° as rGO mark is similar to [36], Cai et al. The low XRD signal-to-noise ratio is caused by the poor crystallinity of the sample. The rGO flakes in the outside surface layer of the film will play a key role in toluene sensing.

 

Fig. 2 (a) Raman spectrum and (b) XRD pattern of rGO on the SPF.

Download Full Size | PDF

 

Fig. 3 Experimental set-up.

Download Full Size | PDF

3. Experimental result

The experimental set-up is constituted with a two-channel flow meter, 2 mass flow controllers (MFC), a gas mixing unit, a chamber (6 litres), a highly steady laser source (1550 nm), an 1 × 2 coupler, a two-channel optical power meter, and a computer, as shown in Fig. 3. There are two kinds of standard gases (with the output pressure of 0.2 MPa): one is dry air, and the other is mixture gas of 1% toluene and 99% dry air. Their flows can be adjusted through the two MFCs. After being further mixed in the gas mixing unit, the final mixture gas with desired toluene concentration can be fed into the chamber. The two output terminals of the coupler are connected to two fiber samples: one is GCSPF, which is placed inside the chamber to face with the gas inlet with a distance of about 3 cm, the other is an unpolished SMF. The output optical powers of GCSPF and SMF can be measured by the optical power meter and be recorded continuously by a computer.

The experiment was conducted at room temperature (about 25°C). The reference gas was dry air, so the concentrations of nitrogen, oxygen, and other components in dry air were nearly unchanged and did not interfere with the sensing experiment. The main interfering gas was the residual water vapor in the chamber. In order to evaluate the influence of the residual water vapor, a humidity meter was also placed in the chamber to monitor the change of humidity. The flow of the standard mixture gas of toluene and dry air remained at 0 and the flow of the standard dry air was adjusted to 500 ml/min and remains unchanged for more than 2 hours, during which the humidity inside the chamber decreased from 62.2%RH to about 5.2%RH with a slower and slower decreasing speed. The output optical power of GCSPF increased gradually due to the desorption of water molecules from graphene, which is similar to the experimental results in [24], Xiao et al. When the output optical power of GCSPF was relatively stable, the flow of the standard mixture gas of toluene and dry air was adjusted to 2, 0, 4, 0, 6, 0, 8, 0, 10, 0 ml/min in sequence, each value maintains 10 minutes, and at the meantime the flow of standard dry air was remained at 500 ml/min. The relative output optical power of GCSPF as a function of time after the relative humidity down to 32% was depicted in Fig. 4(a). A long time output optical power record of SMF during the experiment indicated that the laser source was stable with a power fluctuation of 0.01 dB, as shown in Fig. 4(b). It is observed that the injection of standard mixture gas of toluene and dry air into chamber leads to a decrease of the output optical power of GCSPF. This decrease was not caused by either water molecules which still leaving graphene, or a little increase of gas flow which results in a faster desorption of water molecules. Instead, it was caused by the adsorption of toluene molecules on the graphene. When the injection of standard mixture gas of toluene and dry air stopped, the output optical power of GCSPF increase to an even higher level (with respect to before the standard toluene mixture gas was injected). This was caused by the desorption of toluene molecules from the graphene, and the even higher recovery power level was caused by water molecules leaving graphene as before. The power variation caused by exposure of sensor to toluene with respect to exposure to dry air, ΔP, as shown in Fig. 4(a), is a function of toluene concentration C. The relationship between ΔP and C is depicted as black rectangular markers in Fig. 5. The straight fitting line equation between ΔP and C can be obtained as$\Delta P=-0.0004622C+0.01463$, with a linear correlation coefficient of 96%. The power fluctuation of the laser source was 0.01 dB, and the resolution of the optical power meter was also 0.01 dB, i.e. the minimum power variation caused by exposure of sensor to toluene that can be resolved by the sensor is 0.02 dB. So the concentration detection limit of GCSPF to toluene is less than or equal to 79 ppm, at which the optical power variation is 0.02 dB

 

Fig. 4 (a) Relative output optical power of GCSPF (α) as well as the relative humidity in the chamber (β) as a function of time when dry air (DA) and toluene (T) with concentration of 40, 79, 119, 157, 196 ppm were fed into the chamber alternately; (b) a long time output optical power record of SMF during the experiment period.

Download Full Size | PDF

 

Fig. 5 Optical power variation of GCSPF as a function of the concentration of toluene.

Download Full Size | PDF

The flow of standard dry air was kept unchanged at 500 ml/min for about 5 hours (the humidity down to about 2.5%RH), before the flow of the standard mixture gas of toluene and dry air was tuned between 0 and 10 ml/min back and forth for several times. The resulting relative output optical power as a function of time is depicted in Fig. 6. It is noticed that both the response and the recovery time of GCSPF are 256 s, and the sensor possesses good reversibility and good repeatability. The step-like feature of sensing curves in Fig. 4 and Fig. 6 are caused by the resolution of the optical power meter (0.01dB), i.e. the power variations that smaller than 0.01dB cannot be shown in curves.

 

Fig. 6 Relative optical power of GCSPF as a function of time when dry air and toluene with concentration of 196 ppm were fed into the chamber alternately.

Download Full Size | PDF

4. Theoretical analysis and discussion

The adsorption of gas molecules on graphene will cause the change of charge carrier density of graphene $n_s$, and consequently cause the change of chemical potential ${\mu }_c$because of the following formula [38]:

Under the conditions of T = 298K, $\omega /2\pi =1.94\times {10}^{14}$Hz, $\Gamma ={10}^{12}$Hz, the theoretical relation between $n_s$ and ${\mu }_c$, and that between $\sigma$and${\mu }_c$, can be obtained by calculating Eq. (1) and (2) numerically, and the results are depicted in Figs. 7(a) and 7(b) respectively. For p-type semi-conductor, the value of chemical potential is negative, the value of corresponding charge carrier density determined by Eq. (1) is also negative. Through simulation with COMSOL, $n_{eff}^{TM}$and$n_{eff}^{TE}$ can be obtained and their real and imaginary part are depicted respectively in Figs. 7(c) and 7(d). The output optical power of GCSPF is calculated approximately through Eq. (4), and depicted in Fig. 7(e). In the calculation the refractive indexes of the core and cladding of the fiber are 1.468 and 1.463 respectively, the polished length is 1 cm, the polished depth is 60 μm, a 15-layer graphene is considered, and the conductivity of N-layer graphene is assumed to be ${\sigma }_{N-layer}=N{\sigma }_{monolayer}$ [39]. With the decrease of chemical potential, there exists a power-descending interval [0, −0.42] ћω, a power-ascending interval [-0.42, −0.6] ћω, a power-sunken interval [-0.6, −0.64] ћω, and a power-unchanged interval [-0.64, −1] ћω, as shown in Fig. 7(e). The existence of the power-descending interval is attributed to the loss of TM-mode wave energy increases with the absolute value of chemical potential in this interval. The existence of the power-ascending interval is attributed to the losses of both TM- and TE-mode waves energy decrease with the absolute value of chemical potential in this interval. The existence of the power-sunken interval is attributed to TE-mode wave can propagate along the graphene sheet in the vicinity of so-called epsilon-near-zero (ENZ) point, therefore the fractional energy of TE-mode wave is confined in the graphene sheet and rapidly damped [16]. The power descending range in the power-descending interval and the power ascending range in the power-ascending interval are all increase with graphene layer number N. The calculated intensity distributions of the TM mode and the TE mode in the cross section of GCSPF are shown in Figs. 7(f) and 7(g) respectively

 

Fig. 7 Under the conditions of T = 298K, ω/2π = 1.94 × 10¹⁴Hz, Γ = 10¹²Hz, the calculated relationship between: (a) μ_c and n_s; (b) σ and μ_c; (c) Re(n_eff) and μ_c; (d) Im(n_eff) and μ_c; (e) relative optical power of GCSPF and μ_c; numerically calculated fundamental TM mode (f) and fundamental TE mode (g) of GCSPF.

Download Full Size | PDF

It is known that rGO possesses p-type semi-conductivity [5] and water molecules act as electron acceptors [4]. The similar responses of GCSPF to toluene and to water molecules (the power decrease when attaching and increase when detaching, see Fig. 4(a)) indicates that toluene molecules also act as electron acceptors (like water molecules), i.e. their attachment cause the increase of charge carrier (hole) density of rGO. In order to compare the experimental results with theoretical calculated curve, the values of chemical potential induced by toluene with different concentrations should be estimated. When gas molecules are adsorbed on graphene, charge transfer between graphene and adsorbate occurs. The relationship between gas concentration and the chemical potential of graphene can be determined through a definite relationship existed between the number density of gas molecules $n$, the surface density of adsorbed gas molecules on graphene $n_a$, and the induced charge carrier density of graphene $n_s$. As an example, we can first obtain the relationship between water vapor concentration and induced chemical potential of graphene according to the experimental results in [24], Xiao et al. There is a minimum output power of GCSPF when RH = 70% at 25°C [24], and it corresponds to${\mu }_c=-0.42\hslash \omega$, as shown in Fig. 7(e). By using Eq. (1), the $n_s$ corresponding to ${\mu }_c=-0.42\hslash \omega$ can be determined, and $n_a$ can also be determined by the ratio $n_s/n_a=0.025$ (the charge transfer between graphene and one adsorbed water molecule is 0.025e [40]). The water molecule number density $n$ corresponding to the condition of RH = 70% at 25°C can be determined through gas state equation$P=n{k}_{B}T$, where P is the gas (water vapor) pressure. It can be found that the ratio $n_a/n^{2/3}\approx 566$. By assuming $n_a/n^{2/3}$ is a constant for other relative humidity value at 25°C, we can then obtain the relationship between water vapor concentration $C_W$ and induced graphene chemical potential ${\mu }_c$, as depicted in Fig. 8 (Here the contributions of nitrogen, oxygen, and other components except water vapor in air to the chemical potential of graphene are ignored, because the variation of graphene chemical potential in experiment in [24], Xiao et al is mainly caused by the variation of humidity).

 

Fig. 8 The calculated relationship between μ_c and water vapor concentration C_W under the condition of T = 298K, ω/2π = 1.94 × 10¹⁴Hz, Γ = 10¹²Hz, with the air pressure of 101.325 KPa in chamber.

Download Full Size | PDF

The chemical potential induced by toluene with each experimental concentration can also be determined in the order of C→$n$→$n_a$→$n_s$→${\mu }_c$. Due to a lack of data of toluene molecule, the ratios $n_a/n^{2/3}$and $n_s/n_a$are assumed to be same with the case of water molecules (although the existence of certain error is therefore unavoidable in the estimation of the chemical potential induced by toluene, it is unlikely that the absolute value of chemical potential would exceed 0.25$\hslash \omega$, which corresponds to water vapor concentration of 4858 ppm as shown in Fig. 8). Since the optical conductivity of graphene is symmetric for positive and negative chemical potentials due to symmetric band structure in graphene [8], the power response of GCSPF is also symmetric for positive and negative chemical potentials. The chemical potential mentioned below all means its absolute value. The relationship between the experimental relative power of GCSPF and chemical potential of graphene induced by toluene is depicted as black rectangular markers in Fig. 9. The calculated theoretical curves between relative power and chemical potential of graphene are depicted as blue, green, and red solid lines (represent the case of N = 15, 10, and 8 respectively). It is worth noting that there is a match between the experimental results and the theoretical curve of the 15-layer graphene case if the markers are translated to right side for a certain distance. The contribution of retained water molecules to induced chemical potential of graphene should be taken into account.

 

Fig. 9 Solid lines represent calculated theoretical relationship between relative power of GCSPF (graphene layer number N = 15, 10, 8 respectively) and |μ_c| under the condition of T = 298K, ω/2π = 1.94 × 10¹⁴Hz, Γ = 10¹²Hz, the black rectangular markers represent relationship between the experimental relative power and induced |μ_c| of graphene, the red circle markers represent relationship between calculated relative power and induced |μ_c| of highly-doping graphene.

Download Full Size | PDF

It is worth noting that the sensing range of GCSPF to toluene gas concentration is very wide. The power-descending chemical potential interval of [0, 0.42]$\hslash \omega$ corresponds to a concentration range of [0, 21898] ppm, and the power-ascending interval of [0.42, 0.6] $\hslash \omega$corresponds to a concentration range of [21898, 62937] ppm, when the ratios $n_a/n^{2/3}$and $n_s/n_a$ of toluene are assumed to be same with the case of water molecules. It is difficult to verify experimentally the calculated theoretical curve in a much wider concentration range because the maximum concentration of standard toluene mixture gas that can be prepared is about 1%. To verify the response of GCSPF to toluene with much higher concentration, a new experimental set-up should be designed where liquid-phase toluene can be injected into chamber to form high concentration toluene mixture gas, which is beyond the scope of this paper

If the initial chemical potential of graphene is adjusted by using gate voltage (or other method) to a high level (for example, μ_c = 0.5 ћω), and a GCSPF with a 15-layer graphene clad is exposed to toluene with each experimental concentrations in this paper, the corresponding relative optical power can be estimated via the following steps: 1) calculate the carrier density of graphene corresponding to μ_c = 0.5 ћω by using Eq. (1), which can be taken as a basis carrier density; 2) calculate the carrier densities induced by toluene with each experimental concentration and then add each of them with the basis carrier density respectively; 3) calculate the corresponding chemical potentials by using Eq. (1), which are 0.5027 ћω, 0.5042 ћω, 0.5055 ћω, 0.5066 ћω, and 0.5077 ћω respectively; 4) calculate $n_{eff}^{TM}$and$n_{eff}^{TE}$ for each chemical potential value through simulation, and the corresponding optical power by using Eq. (4). The relationship between calculated relative optical power and toluene-induced chemical potential of highly-doping graphene is depicted as red circle markers in Fig. 9. The sensitivity of GCSPF in toluene detection can be estimated to reach 0.003667 dB/ppm, which is nearly 8 times higher than the experimental sensitivity (0.0004622 dB/ppm). Under the same condition (both the power fluctuation of the laser source and the resolution of the optical power meter are 0.01 dB), the sensor’s capability for toluene concentration detection reaches its lowest at 6 ppm.

Graphene-based gas sensors show poor reversibility in low concentration detection of polar gas such as H₂O, NH₃, NO₂, CO [4, 24], whereas GCSPF presents good reversibility in low concentration detection of toluene. This is probably because toluene molecules are nearly non-polar [27] and therefore their desorption from graphene is relatively easy. It is worth noting that the retained water molecules have a significant impact on the sensitivity of GCSPF in toluene detection. The continual increase of the reference standard power (corresponding to GCSPF being exposed to dry air) is caused by the continual and slow loss of water molecules from graphene. Theoretically saying, GCSPF can respond to the attachment of all kinds of gases that can change the carrier density (hence the chemical potential) of graphene. In practical application, if we keep humidity constant, the toluene response by the sensor can be distinguished from water response. Furthermore, using the decorated or functionalized graphene to improve the selectivity of GCSPF is also a potential research direction.

There are two main differences between rGO and perfect graphene: the first is that rGO sheet is composed of intact graphene areas interspersed with defect clusters [41], also with residual hydroxyl and epoxide groups on the sheet; the second is that rGO flake has a much smaller area than the perfect graphene. It can be considered that the defects and residual chemical groups only give rise to slight difference in the band structure of the perfect graphene. Their contribution to optical conductivity can be considered as the increase of the electronic scattering. So we can use optical conductivity σ(ω,μ_c,T,Γ) of the perfect graphene to approximate that of rGO. The increase of the electronic scattering induced by the defects and residual chemical groups in rGO can be phenomenologically considered by the scattering rate Γ, whose smaller value corresponds to a larger charge carrier mobility [8]. On the other hand, for 4 μm large rGO flake, the influence of size effect on the band structure can also be neglected. Therefore the theoretical model based on the perfect graphene optical conductivity can be used to analyze the mechanism of rGO-based sensor in this paper. Using rGO solution makes the sensor fabrication simpler and easier, which can help to lower the cost of the sensor, since the CVD graphene is much more expensive than rGO. However, the obvious shortcoming is that the sensing performance of rGO may be worse than the CVD graphene due to the residual chemical groups and defects in rGO. So as far as the sensitivity is concerned, the CVD graphene may have better performance than rGO. However, using CVD graphene requires more complex processes to transfer the CVD graphene film to cover on the fiber, and other technique to control the atom layer number of the CVD graphene for higher sensitivity. Therefore, the rGO was chosen in the paper.

5. Conclusion

This work presents our work on both theoretical analysis and experiments about the toluene gas sensing characteristics of SPF coated with rGO film. The sensor works at room temperature, both the response and recovery time are about 256 s. Its low limit of detection to toluene concentration is less than or equal to 79 ppm. In addition, it shows good linearity and good reversibility. Theoretical analysis suggests that its sensing range for toluene gas concentration is very wide, and that the doping level of graphene has a significant impact on its sensitivity. If the chemical potential of graphene is adjusted to a proper high level, a significant rise in its sensitivity can be expected. Once the problem of poor selectivity is overcome, GCSPF has a great potential to be used in low-consumption, low-cost, fast, sensitive, and highly reversible toluene sensors.