Boosting Selectivity and Sensitivity to Biomarkers of Quantum Resistive Vapour Sensors Used for Volatolomics with Nanoarchitectured Carbon Nanotubes or Graphene Platelets Connected by Fullerene Junctions

Nanocarbon-based vapour sensors are increasingly used to make anticipated diagnosis of diseases by the analysis of volatile organic compound (VOC) biomarkers from the breath, i.e., volatolomics. However, given the tiny number of molecules to detect, usually only tens of parts per billion (ppb), increasing the sensitivity of polymer nanocomposite chemoresistive transducers is still a challenge. As the ability of these nanosensors to convert the interactions with chemical compounds into changes of resistance, depends on the variations of electronic transport through the percolated network of the conducting nanofillers, it is a key parameter to control. Actually, in this conducting architecture, the bottlenecks for electrons’ circulation are the interparticular junctions giving either ohmic conduction in the case of close contacts or quantum tunnelling when jumps though gaps are necessary. This in turn depends on a number of nanometric parameters such as the size and geometry of the nanofillers (spherical, cylindrical, lamellar), the method of structuring of the conductive architecture in the sensory system, etc. The present study focuses on the control of the interparticular junctions in quantum-resistive vapour sensors (vQRS) by nanoassembling pristine CNT or graphene covalently or noncovalently functionalized with spherical Buckminster fullerene (C60) into a percolated network with a hybrid structure. It is found that this strategy allows us to significantly boost, both selectivity and sensitivity of pristine CNT or graphene-based transducers exposed to a set of seven biomarkers, ethanol, methanol, acetone, chloroform, benzene, toluene, cyclohexane and water. This is assumed to result from the spherical fullerene acting on the electronic transport properties at the nanojunctions between the CNT or graphene nanofillers.


Introduction
The demand for the development of miniaturized sensors arrays for fast, low-cost, low-power detection and discrimination of volatile organic compounds (VOC) has increased dramatically over the years. In particular, the detection and identification of biomarkers in the volatolome (volatiles emitted by the human body) should lead to the development of novel, noninvasive anticipated diagnostic methods for a wide range of diseases. The promise of "volatolomics" [1,2] is to detect at their very early stages of development different types of cancer [3][4][5][6][7][8][9], tuberculosis [10,11] liver failure [12], kidney failure [13], Parkinson's disease [14], and many others [15][16][17][18][19]. The spheres of application of VOC sensing beyond medicine are food degradation monitoring [20][21][22], environment monitoring, i.e., the indication of hazardous chemical leakage, the monitoring of organic limiting devices [55]. Song et al. fabricated fullerene decorated CNT for application in flame-retardant polypropylene. Flame retardancy could be enhanced by utilizing the free electron accepting property of C 60 with the barrier property of CNT [56]. However, without any covalent functionalization of their surface, the nanocarbons interact thanks to the van der Waals interactions, which are well described by a Lennard-Jones potential as modelled by Girifalco et al. [57]. Thus, they can naturally stack to create graphitic structures by self-assembly.
By the control of the geometry and the dispersion of the nanofillers, conductive architecture, organic functionality, etc., it is possible to achieve a concerted disconnection of the percolated network on chemical stress, leading to large change in electrical resistance. The sensitivity of the CNT or graphene-based chemo-resistive sensors can, however, be tuned by the control of the tunnel junctions in the percolated network. A first strategy consists in building a hierarchical structure of hard PMMA microbeads bridged by CNT to form a segregated network, in which the development of a high specific surface can enhance the sensitivity of CNT by 240% while keeping the original selectivity of CNT unaltered [58]. A second strategy to improve the sensitivity of the sensors consists in preventing the aggregation between nanocarbons in the conducting network by separating them with "spacers" in order to make the conductive architecture more easily disconnectable. The effectiveness of the different kinds of "spacers" to enhance conductive polymer nanocomposites' (CPC) sensitivity has been evidenced in different configurations: cyclodextrins [41] and POSS [59] with CNT, AgNP [60] and Fe 3 O 4 [61] with graphene.
In this study, the idea was to graft Buckminster fullerenes (C 60 ) onto carbon nanotubes (MWNT) and reduced graphene oxide (rGO) to build a novel conductive architecture easily disconnectable by the presence of nanoscopic C 60 at junctions. The robustness of the hybrid network should be ensured by a combination of two nanostructuring strategies, selfassembly and spray layer-by-layer (sLbL) deposition [60,62,63] to control the 3D conductive architecture of chemo-resistive transducers. The approach of inserting spherical C 60 at CNT or GNP junctions to enhance the sensitivity of single carbon nanofiller based chemoresistive vapour sensors has to be significant enough to justify an implementation into an electronic nose to improve its discrimination ability of a set of disease biomarkers. The final objective is to keep these sensing performances at the ppb level required to make a sharp diagnosis by volatolomics.

Materials
Multiwall carbon nanotubes (NC 7000) were kindly provided by Nanocyl S.A. (Sambreville, Belgium). This series of multiwall carbon nanotubes are produced via a catalytic carbon vapour deposition (CVD) process. Nanotubes which exit the reactor are then purified to greater than 90% carbon to produce the NC 7000 grade. These CNT have average mean diameter of 9.5 nm, average length of 1.5 µm. CNT were used in the same manner as received from the production source without purification. Graphene oxide (GO) was synthesized from natural graphite powder by a modified Hummer's method [64] followed by reduction with hydrazine hydrate at 100 • C under reflux condition for 24 h to give reduced graphene oxide (rGO). Fullerene-C 60 of 99.5% purity was purchased from Sigma Aldrich (Saint-Quentin Fallavier, France). All solvents, ethanol, methanol, acetone, chloroform, benzene, toluene, and cyclohexane, obtained from Acros Organics (Antwerp, Belgium) were used as received.

Synthesis of Hybrid Nanomaterials
In the first step of synthesis, lithiated CNT were produced by the dropwise addition of n-butyl lithium (in hexane) into a dispersion of CNT in toluene under the condition of continuous stirring and nitrogen protection. In the next step the lithiated CNT is reacted with a large excess of Buckminster fullerene by a nucleophilic addition reaction, resulting in C 60 -g-CNT as described in Figure 1. C 60 -g-rGO was then produced following a similar with a large excess of Buckminster fullerene by a nucleophilic addition reaction, resulting in C60-g-CNT as described in Figure 1. C60-g-rGO was then produced following a similar procedure as for C60-g-CNT [52].

Fabrication of Sensors
For the fabrication of vapour sensors, first C60-g-CNT or C60-g-rGO hybrids were dispersed in chloroform under ultrasonication for 1 h at 50 °C with a BRANSON 3510 device (100 W, 40 kHz). In the second step, vapour sensors were processed by the spray deposition layer-by-layer (LbL) technique [43,65] onto homemade interdigitated electrodes. The sensors were dried in controlled atmosphere at 30 °C for 8 h. The initial resistance of all the sensors used in this study was adjusted within the R0 = 8 ± 3 kΩ range, so that only the composition and not the variation of initial resistance may be considered to be responsible for any variation of molecular selectivity of the sensors when exposed to a set of VOC cancer biomarkers. The details of composition of the hybrid sensors are provided in Table  1.

Characterization Techniques
Thermo gravimetric analysis was carried out on the PYRIS 6 TGA instrument of PER-KIN ELMER in a nitrogen atmosphere at a heating rate 20 °C·min −1 in the temperature range 50 °C to 750 °C in order to calculate the amount of C60 grafted on to CNT or rGO. The architecture of as-synthesized nanohybrids was observed by morphological analysis such as atomic force microscopy (AFM) and scanning electronic microscopy (SEM). The nanoscale characterizations were done using AFM in ambient conditions using light tapping mode (TM-AFM) on a calibre multimode scanning probe microscope from BRUKER-VEECO, Lille, France. For AFM observation, sample's preparation was done by spin coating onto freshly cleaved silicon substrate at 2500 rpm for 30 s using SPIN150-NPP (SPS-Europe), followed by evaporation of the solvent. The morphology was also observed under the ZEISS EVO series scanning electron microscope model (EVO 50 SEM).

Fabrication of Sensors
For the fabrication of vapour sensors, first C 60 -g-CNT or C 60 -g-rGO hybrids were dispersed in chloroform under ultrasonication for 1 h at 50 • C with a Branson 3510 device (100 W, 40 kHz). In the second step, vapour sensors were processed by the spray deposition layer-by-layer (LbL) technique [43,65] onto homemade interdigitated electrodes. The sensors were dried in controlled atmosphere at 30 • C for 8 h. The initial resistance of all the sensors used in this study was adjusted within the R 0 = 8 ± 3 kΩ range, so that only the composition and not the variation of initial resistance may be considered to be responsible for any variation of molecular selectivity of the sensors when exposed to a set of VOC cancer biomarkers. The details of composition of the hybrid sensors are provided in Table 1.

Characterization Techniques
Thermo gravimetric analysis was carried out on the Pyris 6 TGA instrument of Perkin Elmer in a nitrogen atmosphere at a heating rate 20 • C·min −1 in the temperature range 50 • C to 750 • C in order to calculate the amount of C 60 grafted on to CNT or rGO. The architecture of as-synthesized nanohybrids was observed by morphological analysis such as atomic force microscopy (AFM) and scanning electronic microscopy (SEM). The nanoscale characterizations were done using AFM in ambient conditions using light tapping mode (TM-AFM) on a calibre multimode scanning probe microscope from Bruker-Veeco, Lille, France. For AFM observation, sample's preparation was done by spin coating onto freshly cleaved silicon substrate at 2500 rpm for 30 s using SPIN150-NPP (SPS-Europe), followed by evaporation of the solvent. The morphology was also observed under the Zeiss Evo series scanning electron microscope model (Evo 50 SEM).

Dynamic Vapour Sensing
The chemo-resistive properties of the fabricated sensors were studied by measuring the electrical resistance when exposed alternatively to a 5 min cycles of dry nitrogen and saturated VOC while the sensors were placed in a 100 mm × 10 mm× 3 mm chamber with 10 slots. As one can see, the amplitudes are very similar and thus the error bar would not be clearly visible.
The dynamic vapour sensing device is provided with a mass flow controller, solvent bubbler, electrical valves controlled by LabView software program and Keithley 6517 multimetre. The total flow rate is kept constant at 100 cm 3 ·min −1 . The chemo-resistive response of nanocarbon based sensors can be expressed by calculating the relative amplitude of electrical signals (Ar) against solvent according to Equation (1).
where R 0 and R are the initial resistance of the sensor in pure nitrogen and the resistance of the sensor in the presence of solvent vapour, respectively. The vapour sensing performances of sensors under exposure of only ppm-ppb level concentration of VOC was carried out using a combination of LabView software and OVG-4 device, a versatile chemical vapour generator which consist of two main sections, the oven and the flow control system. The flow control system comprises of sample flow and split flow. The sample flow was kept fixed at 100 cm 3 ·min −1 and split flow was varied to alter the concentration of vapour at a fixed temperature. For each analyte, permeation tubes have been prepared following the providers' recommendations (https://www.owlstoneinc. com/products/build-your-own-permeation-tubes/ accessed on march 2021). For instance, the permeation rate at a given temperature (that depends on the length of the tube and the nature of the solvent) was 1005 ng·min −1 at 35 • C, 796 ng·min −1 at 40 • C, 2335 ng·min −1 at 40 • C and 2360 ng·min −1 at 40 • C for water, methanol, toluene and benzene, respectively. Thus, by acting on the sample flow (100 cm 3 ·min −1 ), the exhaust flow (cm 3 ·min −1 ) for a fixed detector flow of 100 cm 3 ·min −1 , the concentration of benzene was 410 ppb. The calibration of analytes' concentrations lasting one day per volatile organic compound, only eight biomarkers were tested at two concentrations at 1000 and 400 ppb. In this range the evolution of the amplitude of the chemo-resistive response Ar is assumed to be linear with the amount of analyte molecules when Langmuir adsorption is concerned, as already established in previous works [26,66,67]. Then, at higher contents, over several ppm, another linear range corresponding to Henry diffusion will be found and finally, over thousands of ppm, (more than 50% of saturation) clustering will take place. To describe this complete chemo-resistive behaviour on the whole range of concentrations, the LHC model can be used [58,[68][69][70]. Moreover, here, as the targeted concentrations are extremely low in the ppb range, the key point is to evaluate the limit of detection of vQRS, to be able to extrapolate the response Ar down to this value.

Biomarkers
Chemo-resistive vapour sensors able to detect biomarkers in the exhaled breath of human beings have opened up a new era for early cancer diagnosis [71,72]. The discrimination between cancer patients and healthy subjects can be simply done through the comparison of their exhaled breath VOC profile via pattern recognition algorithms [73,74]. Recent studies using gas chromatography/mass spectroscopy (GC-MS) [75] linked with a preconcentrator have shown that the compounds of interest are generally to be found at 1-20 parts per billion (ppb) in healthy human breath [76,77]. However, VOC can also be seen in distinctive mixture compositions at higher levels from 10 to 3000 ppb in the breath of cancer patients; for example, see Table 2 that summarizes some cancer biomarkers found in the breath [78,79]. The VOC biomarkers of cancer and other diseases can be broadly divided into polar and nonpolar compounds. The polar VOC belong to alcohols, carbonyl compounds like aldehydes and ketones, whereas nonpolar VOC are of two types, aliphatic hydrocarbons and aromatic hydrocarbons. However, the concentrations of alkanes and aromatic hydrocarbons in human breath are from 10 ppt to 40 ppb, hence preconcentration becomes a very important step before analysis [77,80,81]. In this study we have selected seven of these biomarkers-ethanol, methanol, acetone, chloroform, benzene, toluene, cyclohexane and water-to illustrate the detection capabilities of our vQRS, being aware that only polar analytes can be detected without preconcentration. The three-dimensional AFM images of pristine CNT, GO and C 60 are exhibited in Figure 2a-c, respectively. The diameters of the C 60 nanoparticle clusters were calculated to be around 230 nm on an average. The comparison of an AFM image of pristine CNT with that of C 60 -g-CNT shown in Figure 2d, distinctly proves that a large number of C 60 particles have been grafted onto CNT.
The pristine CNT appears to have a smooth cylindrical surface without any visible catalyst particles, whereas C 60 -g-CNT consists of many protrusions from the surface of CNT, mostly at the junctions, which indicates that the C 60 nanoparticles are grafted mostly at the junctions of entangled CNT network. On the other hand, the surface of the pristine GO nanosheets appears to be quite smooth, unlike that of C 60 -g-rGO shown in Figure 2e, where dotted clusters are clearly visible on the surface of the rGO nanosheets due to the grafting of C 60 nanoparticles. Figure 3a shows SEM images of C 60 nanoparticles when drop casted onto a silica substrate. Originally, polyhedral pentagonal hexagonal shaped particles of some hundred nanometres in diameter are observed. Figure 3b,c exhibit SEM images of C 60 , after dispersion in toluene abs sprayed onto silica substrate at high pressure at low and high resolution, respectively. Surprisingly, fullerene nanowhiskers (FNW) are formed in this case. It is reported by several researchers that C 60 can be converted into FNW by the application of high temperature, pressure or velocity [84].

Scanning Electronic Microscopy (SEM)
The interfacial polymerization of C 60 nanoparticles may also produce FNW of diameter ranging from 300 nm to 1 µm and length several micrometres [85,86]. Therefore, our case definitely provides an interesting finding where the spray layer-by-layer technique is proved to be a novel tool for production of fullerene nanowhiskers. The surface of C 60 -g-rGO shown in Figure 3d is full of dotted clusters of C 60 nanoparticles unlike that of pristine rGO as in Figure 3e, which supports the finding from AFM and confirms the presence of grafted C 60 on the surface of rGO. By comparison of SEM images of CNT with C 60 -g-CNT as observed in Figure 3f,g, C 60 nanoclusters are found to be fully coated with CNT. The pristine CNT appears to have a smooth cylindrical surface without any visible catalyst particles, whereas C60-g-CNT consists of many protrusions from the surface of CNT, mostly at the junctions, which indicates that the C60 nanoparticles are grafted mostly at the junctions of entangled CNT network. On the other hand, the surface of the pristine GO nanosheets appears to be quite smooth, unlike that of C60-g-rGO shown in Figure 2e, where dotted clusters are clearly visible on the surface of the rGO nanosheets due to the grafting of C60 nanoparticles. The interfacial polymerization of C60 nanoparticles may also produce FNW of diameter ranging from 300 nm to 1 μm and length several micrometres [85,86]. Therefore, our case definitely provides an interesting finding where the spray layer-by-layer technique is proved to be a novel tool for production of fullerene nanowhiskers. The surface of C60g-rGO shown in Figure 3d is full of dotted clusters of C60 nanoparticles unlike that of pristine rGO as in Figure 3e, which supports the finding from AFM and confirms the presence

Thermo Gravimetric Analysis
On Figure 4 C 60 undergoes only 3% weight loss up to 650 • C, whereas around 9% weight loss takes place across the 650 • C to 800 • C temperature range. The slope of the decomposition curve changes clearly at around 650 • C. rGO undergoes around 12% weight loss up to 800 • C with the TGA curve following a continuous slope. C 60 -g-rGO, on the other hand, shows a combined decomposition trend. It decomposes by 30% up to 800 • C, which may be due to the degradation of butyl group used as a linker between rGO and C 60 . Beyond 600 • C the slope of its decomposition curve is similar to that of C 60 , which proves successful grafting of C 60 on to rGO. Similarly, C 60 -g-CNT undergoes around 20% weight loss up to 800 • C although the component nanomaterials like CNT and C 60 degrade by 4% and 12%, respectively.   Figure 5 shows the normalized average maximum relative amplitude of a combination of three sensors CNT (NC 7000), C60-l-CNT, C60-g-CNT, when exposed to a set of eight VOC lung cancer biomarkers (ethanol, methanol, acetone, chloroform, benzene, toluene, cyclohexane and water) at 1000 ppb. The responses of all sensors upon exposure to analytes are recorded five times (see Figure 6a,b for example) and their amplitude is calculated by making the average on the five signals, the first of which can sometimes be slightly different (see inlet of Figure 5), presumably due to a relaxation of the conducting network. As evident from Figure 5 the sensitivity of pristine CNT sensors underwent massive enhancement after hybridization with C60 either by layer after layer deposition in C60l-CNT or by covalent grafting in C60-g-CNT. This gives a very good illustration of the boosting effect of the grafting of C60 on CNT, where it can be seen that the chemo-resistive response of C60-g-CNT based sensor to cyclohexane is twice that of C60-CNT and thirteen times that of the pristine carbon nanotubes. Additionally, it can be noted that the grafting of fullerene does not really change the selectivity of CNT sensors but it enhances the sensitivity of the same, dramatically keeping the order of selectivity same. The sensitivity of the sensors is directly proportional to the structure of the conductive architecture, whereas the difference in sensitivity of CNT based sensors to different VOC most likely results from specific interactions between the carbon species' surface and the analytes [43]. Moreover, as evident from the morphological images presented in Figure 3, C60 is grafted mostly at the junctions of the CNT. Presence of spherical, strong electron acceptor C60 nanoparticles of exceptionally high surface to volume ratio at the CNT junctions can have a strong influence on the electronic mobility in the CNT network. This can induce a faster electronic transport, but also larger disturbances of this flow upon the adsorption of analytes, leading to a dramatic rise of sensitivity to a number of VOC. A strong influence on sensitivity naturally leads to an enhanced selectivity of sensors too. Thus, the grafting of C60 onto CNT leads to a sensor with a high inclination towards nonpolar VOC like cyclo-  Figure 5 shows the normalized average maximum relative amplitude of a combination of three sensors CNT (NC 7000), C 60 -l-CNT, C 60 -g-CNT, when exposed to a set of eight VOC lung cancer biomarkers (ethanol, methanol, acetone, chloroform, benzene, toluene, cyclohexane and water) at 1000 ppb. The responses of all sensors upon exposure to analytes are recorded five times (see Figure 6a,b for example) and their amplitude is calculated by making the average on the five signals, the first of which can sometimes be slightly different (see inlet of Figure 5), presumably due to a relaxation of the conducting network. As evident from Figure 5 the sensitivity of pristine CNT sensors underwent massive enhancement after hybridization with C 60 either by layer after layer deposition in C 60 -l-CNT or by covalent grafting in C 60 -g-CNT. This gives a very good illustration of the boosting effect of the grafting of C 60 on CNT, where it can be seen that the chemo-resistive response of C 60 -g-CNT based sensor to cyclohexane is twice that of C 60 -CNT and thirteen times that of the pristine carbon nanotubes. Additionally, it can be noted that the grafting of fullerene does not really change the selectivity of CNT sensors but it enhances the sensitivity of the same, dramatically keeping the order of selectivity same. The sensitivity of the sensors is directly proportional to the structure of the conductive architecture, whereas the difference in sensitivity of CNT based sensors to different VOC most likely results from specific interactions between the carbon species' surface and the analytes [43]. Moreover, as evident from the morphological images presented in Figure 3, C 60 is grafted mostly at the junctions of the CNT. Presence of spherical, strong electron acceptor C 60 nanoparticles of exceptionally high surface to volume ratio at the CNT junctions can have a strong influence on the electronic mobility in the CNT network. This can induce a faster electronic transport, but also larger disturbances of this flow upon the adsorption of analytes, leading to a dramatic rise of sensitivity to a number of VOC. A strong influence on sensitivity naturally leads to an enhanced selectivity of sensors too. Thus, the grafting of C 60 onto CNT leads to a sensor with a high inclination towards nonpolar VOC like cyclohexane, benzene, toluene, etc., although the principle of the doping of the sensing performances of the CNT network by the specific addition of fullerene is the same for both C 60 -l-CNT (layer after layer deposited separately) and C 60 -g-CNT (fullerene grafted onto CNT). However, the enhancement of sensitivity in C 60 -g-CNT is much higher than that of C 60 -l-CNT. In C 60 -g-CNT, the morphology is expected to be better controlled with C 60 nanoparticles grafted at the junctions of CNT, unlike C 60 -l-CNT where one layer of C 60 solution was deposited after the deposition of one layer of CNT solution onto the electrodes. This finding is further supported by the comparison of the chemo-resistive signals of pristine rGO and C60-g-rGO made in Figure 6a,b, respectively, which shows that the order of selectivity is the same for both. However, the sensitivity, i.e., the relative amplitudes were found to be much higher in case of C60-g-rGO sensors. It is clearly visible that rGObased sensors are about twenty times less sensitive than C60-g-rGO. Moreover, Figure 6c,d confirm the interest of using spacers between nanosheets as already found [41,61], and also evidence the stabilizing effect of C60 at junctions leading to cleaner chemo-resistive responses even at the sub-ppm level.

Dynamic Vapour Sensing
The work presented here corresponds to the preliminary step in the design of an enose for breath analysis in order to diagnose severe diseases such as cancers, i.e., the determination of the selectivity and sensitivity to volatiles organic compounds considered as biomarkers of a considered disease. To go further it will be necessary to assemble more than 10 different vQRS of different selectivities into an array (e-nose) and to classify their responses with an algorithm such as PCA to be able to discriminate the different VOC into a 2D map, which was done there [59,87,88]. Another step consists of checking the influence of water molecules on the detection of biomarkers, as breath contains more than 80% moisture, which was done previously [89]. Therefore, it is believed that if sensors are not very sensitive to water, this drawback can be compensated by including a vQRS with a much higher sensitivity to water, such as chitosan functionalized CNT, for instance [70,87,88], in the e-nose. This finding is further supported by the comparison of the chemo-resistive signals of pristine rGO and C 60 -g-rGO made in Figure 6a,b, respectively, which shows that the order of selectivity is the same for both. However, the sensitivity, i.e., the relative amplitudes were found to be much higher in case of C 60 -g-rGO sensors. It is clearly visible that rGObased sensors are about twenty times less sensitive than C 60 -g-rGO. Moreover, Figure 6c,d confirm the interest of using spacers between nanosheets as already found [41,61], and also evidence the stabilizing effect of C 60 at junctions leading to cleaner chemo-resistive responses even at the sub-ppm level.
The work presented here corresponds to the preliminary step in the design of an e-nose for breath analysis in order to diagnose severe diseases such as cancers, i.e., the determination of the selectivity and sensitivity to volatiles organic compounds considered as biomarkers of a considered disease. To go further it will be necessary to assemble more than 10 different vQRS of different selectivities into an array (e-nose) and to classify their responses with an algorithm such as PCA to be able to discriminate the different VOC into a 2D map, which was done there [59,87,88]. Another step consists of checking the influence of water molecules on the detection of biomarkers, as breath contains more than 80% moisture, which was done previously [89]. Therefore, it is believed that if sensors are not very sensitive to water, this drawback can be compensated by including a vQRS with a much higher sensitivity to water, such as chitosan functionalized CNT, for instance [70,87,88], in the e-nose.

Limit of Detection at ppb Level
Since breath extract consists of VOC mostly present at tiny concentrations, it is important to analyse the performance of our sensors under real conditions of ppb level exposure to VOC. In order to assess the efficiency of the sensors during sensing of trace level of vapour molecules in the surroundings, the signal-to-noise ratio (SNR) was measured according to LEWIS et al. [90] using Equation (2). baseline R SNR σ max Δ = (2) where ΔRmax = the steady-state resistance change upon exposing the sensor to solvent vapour molecules (analyte), i.e., the difference between the maximum resistance of a sensor obtained after exposure to a solvent and the baseline resistance of a sensor; σbaseline = the standard deviation in baseline resistance of the sensor before analyte delivery, calculated using 10 data points.
It is accepted that the chemo-resistive response can processed if its SNR value is higher than or equal to 2.0 (SNR ≥ 2.0) at a defined concentration of solvent. Table 3 clearly demonstrates that the noise density of the rGO based sensors has been reduced considerably after the grafting of C60.

Limit of Detection at ppb Level
Since breath extract consists of VOC mostly present at tiny concentrations, it is important to analyse the performance of our sensors under real conditions of ppb level exposure to VOC. In order to assess the efficiency of the sensors during sensing of trace level of vapour molecules in the surroundings, the signal-to-noise ratio (SNR) was measured according to Lewis et al. [90] using Equation (2).
where ∆R max = the steady-state resistance change upon exposing the sensor to solvent vapour molecules (analyte), i.e., the difference between the maximum resistance of a sensor obtained after exposure to a solvent and the baseline resistance of a sensor; σ baseline = the standard deviation in baseline resistance of the sensor before analyte delivery, calculated using 10 data points. It is accepted that the chemo-resistive response can processed if its SNR value is higher than or equal to 2.0 (SNR ≥ 2.0) at a defined concentration of solvent. Table 3 clearly demonstrates that the noise density of the rGO based sensors has been reduced considerably after the grafting of C 60 . Table 3. Signal-to-noise (S/N) ratio of rGO and C 60 -g-rGO sensors towards four selected VOC at 400·ppb concentration. As also evident from Figure 6c,d, the enhancement of the sensitivity the of rGO based sensors by around four times, on grafting of the C 60 even at 400 ppb VOC concentration validates the reliability of our sensors even in sub-ppm level exposure of VOC.

Conclusions
It has been demonstrated in this paper that the grafting of Buckminster fullerenes at the junctions of CNT or graphene networks can strongly boost the sensitivity and the selectivity of the conducting architecture of chemo-resistive vapour sensors. It has been assumed that the enhancement of the electronic mobility through the CNT or graphene network, due to the grafting at junctions of spherical Buckminster fullerene with high surface area, is the most likely reason for the important gain of sensing properties obtained. Given that the major limitation of crude CNT or graphene-based sensors results from their too low and nonadjustable selectivity to volatile organic compounds, this work shows that such a lock can be lifted by synthesizing their nanohybrids based on Buckminster fullerene, which improves by tenfold the performances of crude carbon nanotubes or graphene-based sensors. Although the ranking of selectivity towards the different selected VOC remained the same, the increase in sensitivity automatically led to the enhancement of the sensors' discrimination ability. Thus, it can be concluded that grafting C 60 at the nanojunctions of a carbon nanoarchitecture by covalent or noncovalent functionalization is a powerful tool to tune both sensitivity and selectivity of carbon-based nanosensors without functionalization by polymers as usually done. The present study unveils an innovative approach for boosting chemo-resistive transducers by improving the electronic mobility of their junctions through the grafting of Buckminster fullerenes thanks to their high surface area. These promising results are expected to bring an interesting input in the design of sharper sensor arrays for volatolomics in order to make the anticipated diagnostics of severe diseases like cancers more reliable. However, the study also points out that among all selected biomarkers, only ethanol, methanol, acetone, chloroform, and water can be detected directly by our vQRS when they present in breath at a concentration close to 400 ppb. Nonpolar analytes such as benzene, toluene, and cyclohexane, being under the limit of detection in breath, will require a preconcentration step before analysis as expected.