Fast-LPG Sensors at Room Temperature by α-Fe 2 O 3 / CNT Nanocomposite Thin Films

We present performance of a room temperature LPG sensor based on α-Fe2O3/CNT (carbon nanotube) nanocomposite films.The nanocomposite film was fabricated via the metallic Fe catalyst particle on CNTs in which both the catalyst particles and the CNT were simultaneously synthesized by chemical vapor deposition (CVD) synthesis and were subsequently annealed in air to create α-Fe2O3. These methods are simple, inexpensive, and suitable for large-scale production.The structure, surface morphologies, and LPG response of nanocomposite films were investigated. Raman spectroscopy and XPS analysis showed the formation of α-Fe2O3 on small CNTs (SWNTs). Morphological analysis using FE-SEM and AFM revealed the formation of the porous surface along with roughness surface. Additionally, the sensing performance of α-Fe2O3/CNTs showed that it could detect LPG concentration at lower value than 25% of LEL with response/recovery time of less than 30 seconds at room temperature. These results suggest that the α-Fe2O3/CNTs films are challenging materials for monitoring LPG operating at room temperature.


Introduction
Liquefied petroleum gas (LPG) is a complex mixture of hydrocarbon compounds, which mostly consist of propane (C 3 H 8 ) and butane (C 4 H 10 ).LPG is widely used as a combustion apparatus in a heater, cooking equipment, and automotive vehicles.Besides, LPG is of most harmful gases due to its flammable, explosive nature, which presents many hazards to the human being as well as environment.To avoid the damage caused by leaks and gas explosions, there is prevailing need to detect LPG leakage at the lower explosive limit (∼2.0 vol.% of LPG) [1].In last few decades, metal oxide semiconducting (MOS) materials have been extensively used as an LPG sensor [2,3].However, the optimal sensing properties of them operated along with high temperature (300-500 ∘ C).In such case, a heater needs to be installed for the sensors to function, causing increased power consumption, complexity, and investment budgets.To overcome these limitations, the fabrication of a LPG sensor operable at room temperature has gained essence importance [4,5].
Many reports are available on the development of the LPG sensors that operate at room temperature, the reduction of the MOS size [6,7], and the mixing of MOS with metal nanoparticles [8] including the fabrication of heterojunctions [4,[9][10][11], which is one of the most strategies used.Due to the mixing of MOS and other materials, it formed a variety of the unique properties for sensing material, such as a change in conductance, improved surface catalytic property, increasing surface reaction sites, and producing a high porosity [12].In addition, a formed contact potential at the interface between MOS and other nanostructured materials has also enhanced the sensing performance of a gas sensor [13,14].Therefore, more studies in fabricating devices based on heterojunctions are necessary, especially in nanostructure systems.
Iron oxide (Fe 2 O 3 ) is n-type semiconductor, of which hematite (-Fe 2 O 3 ) is the most stable of Fe 2 O 3 .It has been widely studied for the various applications, including magnetic devices, pigments, catalysts, sensors, and medical fields [15].In gas sensor applications, Fe 2 O 3 has been continuously researched because it can provide high electron mobility, high chemical/thermal stability, multiple functions, and low cost [16,17].Recently, the nanostructures of Fe 2 O 3 materials have attracted much attention as an effective material for various gases because of their high surface activities, high surface-tovolume ratios, and high carrier mobility [18].However, their electrical conductivity is limited, because it requires high temperatures to work [19].Meanwhile, CNT can be considered as highly conductive material at room temperature.Additionally, CNTs have a hollow structure, nanosized morphology (diameter 1-10 nm), and high surface area (500-1500 m 2 /g) [20][21][22][23][24]. Therefore, CNTs have been used for mixtures enhanced in the conductivity of Fe 2 O 3 at room temperature [19,25].Moreover, Fe 2 O 3 and CNTs nanocomposites have been fabricated through asynchronous methods, and then the mixture of Fe 2 O 3 and CNT has been deposited on the substrate, usually using the usually spinning and screenprinting technique [25,26].However, these techniques are the major performance-limiting factors in the sensor because of its inhomogeneous film thickness.Therefore, the sensor fabrication process also provides the enhancement in sensing performance, which can control the porosity and the roughness, resulting in the improvement of sensing properties of the sensor.Recently, a new strategy to synthesize metal nanoparticle/CNT hybrid materials has been reported.Both of the Fe catalyst particles and the CNT are simultaneously synthesized.Then, the metallic Fe was transformed into maghemite (/-Fe 2 O 3 ) nanoparticles via a potential-cycling method [27].
Here, we report an effective approach to fabricate the -Fe 2 O 3 /CNT nanocomposite films, in which the CNT was uniformly coated with -Fe 2 O 3 particles (-Fe 2 O 3 /CNTs).According to our previous work [28], we found that the Fe catalyst particles adhere to as-grown CNTs.In this work, those particles were annealed in air to create -Fe 2 O 3 into CNTs films.In addition, we also demonstrate the application of the -Fe 2 O 3 /CNTs thin films as a LPG sensor.Obtained results indicated that the response and recovery time of sensor have been significantly improved for detecting LPG at room temperature (28 ∘ C).Moreover, the sensing mechanisms of -Fe 2 O 3 /CNTs-based LPG sensor were also discussed.

Experimental Section
2.1.Fabrication of -Fe 2 O 3 /CNT Thin Films.An -Fe 2 O 3 / CNTs films-based sensor was fabricated through the following processes.Firstly, as-grown CNTs were synthesized through the vertical floating catalyst chemical vapor deposition method (FC-CVD) using pyrolysing solutions of ethanol-ferrocene.The details of synthesized as-grown CNTs have been previously reported [28].Briefly, a precursor was prepared by dissolving ferrocene powder (Fe (C 5 H 5 ) 2 ; Sigma-Aldrich) in ethanol solution (C 2 H 5 OH; Carlo Erba) at a 0.25 wt.% ferrocene/ethanol ratio.The mist of precursor solution was carried into the 960 ∘ C reactor via argon (Ar) gas with a flow rate of 1000 sccm.The synthesized CNTs were collected onto a membrane filter directly from the FC-CVD reactor at the bottom of the reactor under room temperature (hereinafter called "as-grown CNTs").The density of as-grown CNTs was controlled by adjusting deposition time (study; 15, 30, 45, 60, 90, and 120 min).Secondly, asgrown CNT films were then transferred from the filter to a slide glass or silicon (Si) substrate by pressing and then dissolving the filter in acetone.Finally, the as-grown CNTs on substrate were put into a laboratory oven and annealed in air atmosphere at 350 ∘ C for 8 h to create -Fe 2 O 3 (hereinafter "-Fe 2 O 3 /CNTs").In order to investigate the effect of standalone CNTs on the gas sensing properties, -Fe 2 O 3 /CNTs films were purified by immersing in 3 M hydrochloric acid (HCl) to remove -Fe 2 O 3 , resulting in a purified CNTs film (hereinafter called "purified CNTs").

2.2.
Characterizations.The characteristics of CNTs were analyzed by Raman spectroscopy (Renishaw inVia Reflex) with an Ar laser (514.5 nm) as the excitation source.The surface composition changes of as-grown and -Fe 2 O 3 /CNTs were measured by X-ray photoelectron spectroscopy (XPS; PHI 5700), which were carried out on an AXIS Ultra DLD-X-ray photoelectron spectrometer using a monochromatic Al K  source (1486.6 eV).In order to analyze the chemical functional group of the samples, the Fe 2p, O 1s, and C 1s core level were deconvoluted using the Shirley background function and Vigt fit.The surface morphology of films was performed by field-emission scanning electron microscopy (FE-SEM: JEOL, Hitachi s4700), atomic force microscopy (AFM: Park Systems XE100), and transmission electron microscopy (TEM: JEOL, JSM-2010).The sensor was placed into the chamber; N 2 (or zero air) and LPG were then injected via a mass flow controller unit (MFC).The gas sensing characteristics of the sensor were examined by monitoring changes in resistance with a constant voltage of 5 V, using Keithley source meter (Model 2004).Data acquisitions (DAQ), storage, and plotting in real time were realized using a personal computer with LabView6 software via a GPIB interface control.

Results and Discussion
3.1.Raman Spectroscopy Analyses.Figure 1(a) presents the Raman spectra of the as-grown CNT and -Fe 2 O 3 /CNT nanocomposite films in the low-frequency (∼100-500 cm −1 ) region.The as-grown CNT films show several characteristic peaks at 162 and 267 cm −1 .These peaks are the Raman spectra in radial breathing mode (RBM: ∼100-350 cm −1 ) of the small-diameter CNTs (SWNTs: single-walled CNTs), which corresponded to the diameters (Φ SWNTs ) of the CNTs films varying in the range of 0.9-1.5 nm [29].It is clearly seen from this figure that CNTs with 1.5 mm diameter remain after the anneal, while the small tube (0.9 nm) is easily burned at 350 ∘ C due to curvature strain [30].In addition, Raman spectra of -Fe 2 O 3 /CNT nanocomposite films also reveal additional peaks at 227, 244, 298, and 414 cm −1 but these peaks are small signals (see inset of Figure 1(a)).These peaks remained after removing CNTs by annealing at 550 ∘ C, which can clearly observe the main peak at 298 cm −1 , 414 cm −1 , and 1320 cm −1 and minor peaks at 227 cm −1 and 244 cm −1 .These peaks are identified to the hematite (-Fe 2 O 3 ) nanostructure [31,32].
Figure 1(b) presents the Raman spectra of the as-grown CNT and -Fe 2 O 3 /CNT nanocomposite films in the highfrequency (∼1200-1800 cm −1 ).In the high-frequency regions,  two apparent peaks at 1570 cm −1 and 1590 cm −1 are observed and correspond to the D and G band of CNTs [29].The intensity ratio of G and D mode ( G / D ): a higher ratio indicates a lower defect, therefore, better structural quality [29]. G / D of as-grown CNT and -Fe 2 O 3 /CNTs are 11.9 and 12.3, respectively.It can be seen that  G / D of -Fe 2 O 3 /CNT films had higher than that of as-grown CNT films. G / D results indicated that the purity of CNTs increased and the structure of CNTs was not damaged even when annealed in air.

XPS Analyses.
The XPS technique is used to analyze chemical composition on the top surface of the materials (depth < 10 nm).Compositional elements can be identified by the peak position in terms of binding energy and the peak intensity (peak area) can be related to the amount of elements in the material.Figure 2(a) demonstrates that the XPS survey spectrum of both as-grown CNTs and -Fe 2 O 3 /CNT films was carbon (C 1s; ∼284 eV), oxygen (O 1s; ∼530 eV), silicon (Si 2p; ∼100 eV), silver (Ag 3d; ∼370 eV), and iron (Fe 2p; ∼700 eV) [33][34][35].In -Fe 2 O 3 /CNT thin films, peak at 100 eV is observed and corresponds to silicon of silicon substrate; this is because of the ultra-thin films of the Fe 2 O 3 /CNTs.In addition, the atomic percent (At%) table (see the inset table) demonstrated that the O : Fe atomic ratio on the -Fe 2 O 3 /CNT films was close to 2 : 3, which is probably Fe 2 O 3 [25].
The phase analyses in -Fe 2 O 3 have been achieved via the Fe 2p deconvolution, as shown in Figure 2(b).The peaks in the Fe 2p spectrum were considered as species of Fe in films, which is referenced in the previous reports of Fe 2 O 3 [36][37][38].Fe 2p 1/2 and Fe 2p 3/2 are doublet spin orbit component of Fe 2p.According to Figure 2(b), the as-grown CNT films were composed mostly of Fe metal, followed by Fe 3 C (Fe-C), FeO, Fe 3 O 4 , and Fe 2 O 3 , respectively.After annealing, the dominant component in -Fe 2 O 3 /CNT films was Fe 2 O 3 .In addition, two distinct peaks are observed at binding energies of 711.4 (Fe 2p 3/2 ) and 724.4 (Fe 2p 1/2 ) eV with a broad shake-up-satellite peaking at 719.3 eV.These peaks suggested that the phase formation of -Fe 2 O 3 was as good as those reported in the previous -Fe 2 O 3 studies [18,36].The occurrence of -Fe 2 O 3 is similar to that obtained by our Raman results.These results indicated that the most of Fe elements in -Fe 2 O 3 /CNT films were -Fe 2 O 3 , followed by the Fe 3 O 4 , Fe (OH)  , and Fe-Si, respectively.The increase in intensity of the O 1s deconvoluted peak at 530.4 eV indicated that the surface of -Fe 2 O 3 had mostly adsorbed oxygen ions [39], as shown in Figure 2(c Furthermore, more structure and characterization of the CNTs were carried out by the C 1s deconvolution [39], as shown in Figure 2(d).Both of as-grown CNT and -Fe 2 O 3 /CNT films presented the main peak at 284.9 eV, which can be assigned the C 1s (Sp 2 ) binding energy of CNTs.The It is widely recognized that the mixing of oxide NPs by the formation of nucleus Fe + on CNTs induces damage in CNTs [25].However, our method did not further significantly affect the CNTs due to the Fe spontaneously grow and uniformly assemble on the entire surface of each nanotube by CVD.
In addition, the sp 2 /sp 3 ratio of the -Fe 2 O 3 /CNT films was higher than that of the as-grown CNT films.This is consistent with  G / D ratios in the Raman results, implying that the structure of CNTs was not damaged even when annealed in air.

FE-SEM/AFM Analyses.
From the obtained results from XPS and Raman spectroscopy, it is confirmed that the asgrown CNTs were decorated with metallic Fe.However, these metallic Fe could be easily transformed to -Fe 2 O 3 by air annealing.In this section, we investigated that the  As-grown CNT films consist of CNTs and metallic Fe particles.The decrease in the distribution of CNTs and Fe particles suggest that the density of CNTs and Fe particles increases with increasing the deposition time.In addition, the hydroxide (OH-) part was observed onto CNTs and Fe particles in XPS result.These components may be attributed to the remained filters.The CNTs were the loosely packed and were consisted of the randomly oriented entangled CNT bundles, which were decorated with Fe NPs (size averaged 7.33 nm in diameter; Supplement Figure S1).

Sensing Properties of LPG Sensors.
The sensor response to LPG was defined as  (%) = (|Δ|/  ) × 100%, Δ is   −  , where   and   denote the resistance of sensor when exposed to LPG gas and that when exposed to baseline gas, respectively [48].Response/recovery time [49] is an important sensing performance indicator for a sensor to detect a flammable gas.A sensor's response time ( resp .) is defined as 90% change in resistance from its baseline value to the maximum value in the presence of LPG.Recovery time ( rec .) is defined as the time required for recovering of the original resistance.In the examination of the sensor, carbon conductive electrodes were added onto the top of sensing films.
Figures 5(a), 5(b), and 5(c) display the change in the resistance of as-grown CNTs, -Fe 2 O 3 /CNTs, and purified CNTs sensors in the presence of LPG at various concentrations.The baseline resistances in N 2 for as-grown CNTs, -Fe 2 O 3 /CNTs, and purified CNTs were approximately 2.5 kΩ, 1.9 MΩ, and 360 kΩ, respectively.Figure 5  the p-type LPG sensing behavior is observed for as-grown CNTs, which resulted from the CNTs sensing while the Fe catalyst was encapsulated with carbon [27].Meanwhile, -Fe 2 O 3 /CNT films display an n-type sensing behavior due to the formations of -Fe 2 O 3 that act as LPG sensing.It is well known that the sensing behavior of gas sensor has been related to the properties of sensing materials such as p-type material, which had hole majority carriers.The amount of hole will be decreased dramatically when reducing gas (LPG) exposure, due to the fact that an electron of reducing gas was injected to p-type material resulting in increased resistance of the sensor [50].This could explain the sensing behavior of as-grown CNTs and -Fe 2 O 3 /CNTs composite based on LPG sensor.Figure 5(c) clearly depicts that the sensing behavior of CNTs without -Fe 2 O 3 is of p-type.
Figure 5(d) presents the sensor response, and -Fe 2 O 3 / CNT films showed the maximum in response along with fast response/recovery time, while as-grown CNT and purified CNT films had a low response.This was due to the fact that high bonding energy among LPG atoms allows limited electron transfer from LPG molecules to the CNTs [51].However, the purified CNTs had a quick response/recovery time because of the highly purified CNTs than that of CNT films.These results indicated that the -Fe 2 O 3 /CNT films had an excellent responsive to detect the LPG.Moreover, we observed that the time response of -Fe 2 O 3 /CNTs was faster than that of as-grown CNTs.This observation can be explained by the different sensing mechanism.For as-grown CNT films, the response is attributed to the adsorption between the LPG and the surface of the carbon nanotube.On the other hand, the fast response of -Fe 2 O 3 /CNT films might be due to the sensing process of the interface between the CNTs and the -Fe 2 O 3 .According to this mechanism, the quick response of Fe 2 O 3 /CNT-based sensors could be attributed the Schottky barrier formed at -Fe 2 O 3 and CNTs junction [52].Moreover, the annealing in air could improve the response time of -Fe 2 O 3 /CNT films [53].In next section, we investigated the effect of the deposition time on LPG sensing, as showed in Figure 6."S1" and "S2" are defined as the sensor response of n-and p-type, respectively.The -Fe 2 O 3 /CNT films were prepared under varying deposition time of 15, 30, 45, 60, and 90 min.For the sensing behavior of -Fe 2 O 3 /CNT films, we found that the -Fe 2 O 3 /CNT films present the both of n-and p-type behavior.However, notable behaviors of the sensor were attributed to the deposition time.-Fe 2 O 3 /CNTs prepared for 15 min showed the p-type sensing behavior while -Fe 2 O 3 /CNTs prepared under deposition time of 30, 45, 60, and 90 min presented n-type sensing behavior.Note that the resistance of the -Fe 2 O 3 /CNT composites prepared under the deposition time of 120 min was also investigated, because of its extremely high resistance at room temperatures, which is out of the range of our instrument.Moreover, SEM and AFM images of -Fe 2 O 3 /CNT films indicated that the choosing a suitable deposition time could be generated the -Fe 2 O 3 /CNT films along with a highly porous structure and high roughness.This could explain the -Fe 2 O 3 /CNT films prepared under deposition time of 45 min, which is the maximum response of ∼6% to 5 vol.% of LPG.
To examine the sensor's multiple-cycle sensing performance under air environment, the -Fe 2 O 3 /CNT films were measured under at various concentrations of 0.1, 0.4, and 0.7 vol.% of LPG diluted in zero air, as shown in Figure 7.The electrical resistance of sensors decreased upon LPG exposure and increased after replacing LPG with zero air.The sensor response, response, and recovery times of -Fe 2 O 3 /CNTs on Figure 7 are summarized in Figure 8.It is observed that the sensor response and response/recovery times were stable and nearly equal at each concentration, indicating a good reproducibility of the sensing performance.Moreover, the -Fe 2 O 3 /CNT films could detect LPG at concentration levels of less than 0.5 vol.% of LPG, which corresponds to 25% LEL of LPG.Note that LEL (lower explosive limit) is defined as the minimum level of concentration of LPG contained in the  air sufficient enough to propagate a flame when exposed to a source of ignition.
Moreover, all -Fe 2 O 3 /CNT films exhibited excessive recovery (point "i" in Figure 6), which is similar to that reported by Dai et al. in a gas sensor made of monolayer -Fe 2 O 3 [55].They explained the excessive recovery due to the one-electron response, one initial electron recovery, and two-electron eventual recovery.In our work, the oneelectron response has received from the LPG molecules; the two-electron recovery might be received from the oxygen/ substance intermediates gas (C  H 2 O) which may be due to the high active site on -Fe 2 O 3 .
The advantageous properties of the CNTs-based gas sensor are high sensitivity and fast response; however, the CNTsbased LPG sensor is slow to respond and slow to recover.This was due to the fact that high bonding energy among LPG atoms allows limited electron transfer from LPG molecules to the CNTs [51].In our experiment, the -Fe 2 O 3 /CNTs-based LPG sensor's improved response was achieved by converting Fe, which served as catalyst for the growth of CNTs, to -Fe 2 O 3 , causing an increase in LPG active sites in the CNTs.In addition, the porous morphology and the new characteristics of the nanocomposite materials such as n-Fe 2 O 3 /p-CNT nanocomposites make it possible to improve LPG sensing performance.Further investigations are required to clarify this kind of sensing mechanism, which will be undertaken in our future work.Moreover, our literature review shows that various materials have been used in LPG sensor applications intended for room temperature use, as shown in Table 1.Our -Fe 2 O 3 /CNT films-based LPG sensors exhibited a fast response/recovery time, indicating the promise of possible applications in LPG leak detection at room temperature.

Conclusions
A simple fabrication of liquid petroleum gas sensor based on -Fe 2 O 3 /CNT nanocomposite films is reported.-Fe 2 O 3 /CNT films were successfully synthesized through the vertical floating catalyst chemical vapor deposition method (FC-CVD) using ferrocene-ethanol mist.To fabricate a sensor, the metallic Fe particles on CNTs were changed into -Fe 2 O 3 by annealing in air at 350 ∘ C. Raman spectroscopy, XPS analysis, SEM, and AFM images reveal the Fe 2 O 3 /CNT structure with porous and roughness structure.The sensing performance of the sensor was tested at room temperature.The sensing performance of -Fe 2 O 3 /CNTs showed that it could detect LPG concentration at lower value than 25% of LEL with response/recovery time of less than 30 seconds at room temperature.These results suggest that the -Fe 2 O 3 /CNTs films are a challenging material for monitoring LPG operating at room temperature.
Journal of Nanomaterials individual NPs on the TEM images.Figure S1(a) shows the TEM image of the as-grown CNT films, which consisted of the randomly oriented entangled CNTs.The small-diameter CNTs were coated with Fe NPs, and some of these CNTs were coated with Fe NP clusters.These NPs were metallic iron (using XPS).An estimation of the distribution of Fe NPs' sizes performed by measuring each NP obtained from TEM images yielded an average particle size of ∼7.33 nm, as shown in Figure 1S(b).(Supplementary Materials)

Figure 1 :
Figure 1: The Raman spectrum of the as-grown CNT and -Fe 2 O 3 /CNT nanocomposite films measured in (a) low-frequency and (b) highfrequency regions.The inset in (a) shows a comparison between Raman spectra of as-grown CNTs, -Fe 2 O 3 /CNTs, and -Fe 2 O 3 reference.
).The peaks at 532.4, 533.4,and 534.2 eV corresponded to carboxylate/carbonyl (O-C=O/C=O), epoxy/hydroxyl (C-O-C/C-OH), and silicon oxide (Si-O) respectively.It should be noted that the -Fe 2 O 3 /CNTs showed the occurrence of a new peak at 531.7 eV (in O 1s deconvolution; Figure 2(c)) and that Fe-C (in Fe 2p deconvolution; Figure 2(b)) disappeared.This may be that bonding of metallic Fe and CNTs (Fe-C) was changed to -Fe 2 O 3 and CNTs (Fe-O-C), of which Fe atom and CNTs were contacted via carbonyl groups (-O-C).The peak at 531.7 eV, therefore, possibly is assigned to the binding of -Fe 2 O 3 and oxygen functional group on CNTs through Fe-O-C bonds.

Figure 3 :
Figure 3: Morphology of (a-c) as-grown CNT films and (d-f) -Fe 2 O 3 /CNT composite films on glass substrate.

Figures 3 (
d), 3(e), and 3(f) present the SEM images of -Fe 2 O 3 /CNT films prepared with different deposition time of 15, 60, and 90 min, respectively.The hydroxide part was removed during air annealing leading to the formation of -Fe 2 O 3 nanoparticles.It is seen that the CNTs network with -Fe 2 O 3 /CNTs prepared under 15 min -Fe 2 O 3 /CNTs prepared under 90 min

Figure 4 :
Figure 4: Morphology of surface measured using AFM on samples of the -Fe 2 O 3 /CNT films with different deposition times.The 3D images recorded at 5 m × 5 m planar in contact mode.

Figure 7 :Figure 8 :
Figure 7: Dynamic responses of -Fe 2 O 3 /CNT composite films to LPG concentrations varying from 0.1 to 0.7 vol.% in a different mix environment of air.

Table 1 :
Nanocomposite materials-based LPG sensors operating at room temperature.