The UV-Vis absorption spectrum in Fig. 1a shows the CQDs has an absorption peak at 284 nm, resulting from the π-π* transition of C = C (Pan et al. 2010). As shown by the inset in Fig. 1a, the CQDs ink is brown-yellow and the CQDs ink glows green irradiated with 365 nm of light. As shown in Fig. 1b, the excitation spectrum of CQDs ink was measured at an emission wavelength of 400 nm, which showed that the CQDs ink had the highest absorption at about 450 nm. The fluorescence (FL) emission spectra of the CQDs ink excited at 390, 410, 420, 430, 440, 450 and 460 nm. These FL emission spectra showed an increase in intensity and a red shift in the wavelength of peak emission. When the wavelength of the excited light reached 450 nm, the emission intensity of CQDs ink reached the maximum. These excitation-dependent behaviors could be attributed to the optical selection of different-sized nanoparticles or different surface emissive traps (Liang et al. 2016). The TEM image shows that CQDs has good dispersion and uniformity (Fig. 1c). The HRTEM image shows the relatively clear lattice fringes (Fig. 1d) and the lattice spacing of 0.3 nm (the insert in Fig. 1d), which is consistent with the (002) diffraction facets of graphite diffraction (Dong et al. 2013). By randomly calculating 50 particles, the particle size distribution is obtained. The calculation results show that the particle size distribution of CQDs was in the range of 2 ~ 6 nm, of which 3.5 nm CQDs accounted for the highest proportion (Fig. 1e). The FT-IR spectrum shows the CQDs has rich functional groups (Fig. 1f). The peak appearing at around 3316 cm− 1 corresponds to the OH/NH bond, and the peaks at 1668 and 1745 cm− 1 are attributed to the stretching vibration of C = O in aromatic aldehydes (Hu et al. 2017). The peak at 1158 cm− 1 corresponds to the C = N, C = S and C = O bonds (Hu et al. 2017) and the peaks at 2925 and 2854 cm− 1 correspond to stretching vibration of CH/CH2. The peak at 1460 cm− 1 corresponds to the bending vibration peak of CH/NH. The peak at 776 and 702 cm− 1 correspond to the CH bending vibration of the benzene ring. The abundant functional groups of CQDs can improve the interaction between CQDs and various target gases. The main component of cotton fiber is cellulose, containing a large number of hydroxyl groups (Zhang et al. 2020). At the same time, the N, O and S contained in the CQDs form hydrogen bond with the H atom in hydroxyl groups of cellulose, which is convenient for the CQDs to adsorb on cotton fiber, forming the stable CQD/PCFT film.
As shown in Fig. 2a-d, the cotton fibers of PCFT film were in the state of nonspinning, namely the state of disorderly, which increased its specific surface area and gas permeability. After further magnification (Fig. 2c-d), the diameter of the cotton fibers was about a dozen microns. The optical images of PCFT film were white, while the CQD/PCFT film was yellow, indicating that the CQDs uniformly adhered to the cotton fibers. As shown by SEM images in Fig. 2e, f, it can be clearly seen that the surfaces of cotton fibers in PCFT and CQD/PCFT film were wrinkled. However, no CQDs were found on the cotton fibers of CQD/PCFT film, possibly because the particle size of the CQDs was too small to be easily observed.
As shown in Fig. 3a-f, the element distribution showed that pure cotton fiber of PCFT contained only C and O, but no N, P and S. The results of element analysis were consistent with the fact that cotton fiber was mainly composed of cellulose (Jonoobi et al. 2015; Zhang et al. 2020). It is well known that milk contains proteins, which contain N, P and S, so the CQD derived from milk also contain N, P and S. As shown in Fig. 3g-l, there are not only C and O elements, but also N, P and S elements on the surface of cotton fiber soaked by CQDs ink, and these elements are evenly distributed on the CQD/PCFT fibers, indicating that CQDs are uniformly adsorbed on the surface of cotton fibers through hydrogen bond between CQDs and cellulose (Fig. 3m). This is consistent with the results of optical microscopy (Fig. 2c, d) and FT-IR spectra of CQDs (Fig. 1f). The small size of CQDs, the excellent flexibility of cotton fibers and the uniform adsorption of CQDs lay the foundation for the flexible gas sensor. At the same time, the doping of a variety of heterogeneous atoms will have a significant effect on the gas sensitive properties.
According to the XRD data in Fig. 4a, the pure PCFT exhibited a strong peak at 23.1o and two weak peaks at 15.1o and 16.5o, respectively. This is consistent with the reported structure of pure cotton (Zhang et al. 2020). Compared with the PCFT film, the structure of the CQD/PCFT film did not change significantly. This might be because the crystallinity of the micron-scale cotton fiber was good, while the crystallinity of the nanoscale CQDs was poor, and only a very thin layer of CQDs was on the cotton fiber, so the characteristic peak of the CQDs was not seen. However, compared with the PCFT film, the Raman spectra and UV-Vis spectra of the CQD/PCFT film changed obviously (Fig. 4b, c). Compared with the PCFT film, there were two weak characteristic peaks (G band at 1580 cm− 1 and D band at 1375 cm− 1 in Fig. 4b) of carbon materials on Raman spectra of CQD/PCFT film (Cao et al. 2021). As shown in Fig. 4c, the light absorption of CQD/PCFT film in the UV region was significantly enhanced, which indicated that the CQD/PCFT film had the potential of making UV shielding clothing. As can be seen from the inserts of Fig. 4c, the PCFT film was white, while the CQD/PCFT film was pale yellow, which was consistent with the light microscope observation (Fig. 2a-d). These results proved that a thin layer of CQDs was uniformly coated on the cotton fibers, forming the CQD/PCFT composite film.
To evaluate the gas-sensitive performance of the flexible CQD/PCFT sensor, six target gases were tested. As shown in Fig. 5, without bending, the flexible CQD/PCFT sensor showed the excellent sensing capability for the 85% RH, NH3, H2O2 and CH2O vapor, and extremely low sensitivity for C2H6O and C3H6O vapor. Interestingly, the shapes of the sensing curves of the CQD/PCFT sensors to the 85% RH, NH3, H2O2 and CH2O were different, which might be able to realize the discriminative detection. This may be related to the fact that the rich functional groups in CQDs, like the various components in an army knife, allow CQDs to react specifically with a variety of target gases and produce sensing curves of different shapes. Previous discriminative detection often relied on a sensor array or electronic nose containing multiple sensors, and then through data processing such as principal component analysis and radar map analysis could be realized (Sun et al. 2019). Nowadays, with the development of big data and image recognition technology, the discriminative detection of target vapor can be realized through single sensor producing different sensing images corresponding to different target vapors.
As shown in Fig. 6, the bending stability of the flexible CQD/PCFT sensor was also tested. The bending angle indicated in the bending test was shown by the insert in Fig. 6a. When the bending angle was 90o, it corresponded to the folded state of the flexible sensor. As shown by Fig. 6a, the current of the flexible sensor was about 18 nA, and there was no significant fluctuation with the increase of bending angles. It can be clearly seen from Fig. 6b-e, at bending angles of 0°, 30° and 60°, the sensing curves of 85% RH, NH3, H2O2 and CH2O did not change significantly, showing good bending stability. As can be seen from the statistical data in Fig. 6f, g, under the three bending states, the responses of the four target analytes changed by less 5%, and the response time changed by less than 20%, also showing good bending stability. At the same time, the flexible CQD/PCFT sensor also demonstrated rapid recovery ability, and the recovery times were less than 2 s for all four analytes.
Discriminative detection has always been a bottleneck for gas sensor and sensor arrays and electronic noses have been used to break through this bottleneck (Gao et al. 2016; Wu et al. 2016). However, sensor arrays required higher cost and power consumption, because of the built-in multiple sensors. Therefore, the average response and response time of each analyte under three bending angles (Fig. 6h) are used as a binary parameter (Response time, Response) to evaluate the identification capability of the flexible CQD/PCFT sensor. As shown in Fig. 6h, except for the 85% RH and CH2O, the binary parameters of other analytes are very different. This means that with the help of binary parameters (Response time, Response), a single sensor also shows some identification capability, but it still needs to be improved. It is worth noting that the flexible CQD/PCFT sensor maintains a high resolution of the shapes of the sensing curves for the four analytes despite the bending states. Additionally, the technologies of image recognition and cloud storage have developed rapidly in recent years. Therefore, we assume that the sensing curves from the flexible CQD/PCFT sensor are stored as a graph in the cloud database to form a database of sensing graphs (Fig. 6i). Then, with the aid of image recognition technology, the sensing graph of the unknown vapor is compared with the cloud database, and finally the unknown gas could be identified (Fig. 6i). In this way, it is possible to identify and detect a variety of analytes with single sensor and image recognition technology.
The classical theory holds that the sensitivity of the chemiresistive gas sensor is related to the particle size (D) of the sensing material and the charge depletion layer depth (L) (Schierbaum et al. 1991; Göpel 1995; Franke et al. 2006). As shown by Fig. 7a, c, the charge depletion layer was formed because oxygen molecules attached to the sensing particle took electrons from the sensing particle, forming oxygen anions (O2−, O− or O2−) (Schierbaum et al. 1991; Göpel 1995; Franke et al. 2006). Figure 7a, c show that small sensing particles formed a deeper charge depletion layer (L) in the air than large sensing particles, which also generated a higher surface potential barrier (eVsurface) in the small sensing particles. Take CH2O vapor for example, during the sensing process, when the CH2O molecules were attached to the surfaces of CQDs, the following reaction occurred at room temperature:
CH2O (g) + O2- (s) → CO2 + H2O + e- (1)
When the CH2O molecule adsorbed on the sensing particle, it released electrons to the N-type sensing material, which made the charge depletion layer thin (Fig. 7b, d). However, for small sensing particles, the change of charge depletion layer was more significant than that for large sensing particles (Schierbaum et al. 1991; Göpel 1995; Franke et al. 2006). Accordingly, the change of the surface barrier (ΔeVsurface) of the small sensing particle was much more obvious than that of the large sensing particle. It has been reported that the conductance G of the sensing material is exponentially proportional to the effective barrier height (Franke et al. 2006; Li et al. 2015; Zhang et al. 2018).
G ≈ exp (-eVsurface/kT) (2)
Where eVsurface is the effective potential barrier, k is Boltzmann’s constant and T is absolute temperature. It can be seen that a small change in the effective barrier will cause a significant change in the conductance G of sensor. The conductance of the sensing material in the air and the target vapor corresponds to IA and IG, respectively. According to the definition of Response (Response = IG/IA), small sensing particles will be more sensitive to target vapor than large sensing particles (Schierbaum et al. 1991; Göpel 1995; Franke et al. 2006).
In the case of CQDs/PCFT film, the excellent gas sensing performance should be attributed to the synergistic effect between CQDs and f PCFT film (Fig. 7e). First, the particle size of CQDs was very small, basically less than 10 nm. Second, the CQDs had a highly defective graphene structure, especially those with heteroatomic doping (Wang et al. 2016; Hu et al. 2017; Qi et al. 2019). Due to the doping of N, P, S, and the rich functional groups, CQDs had more adsorption and reaction sites (Hu et al. 2017; Qi et al. 2019), which was beneficial to increase the charge depletion layer depth L (Fig. 7e). Third, the particle size of nanoparticles was small, their surface energy was high, and nanoparticles tended to stick together to reduce surface energy (Alivisatos 1996; Klimov et al. 2000). As a result, the advantage of large specific surface area of nanoparticles was not exploited. The CQDs were attached to the cotton fibers of the CQDs/PCFT film, which avoided agglomeration of CQDs and allowed the CQDs to take advantage of the large specific surface area and the rich functional groups. Fourth, the CQDs/PCFT film with the high permeability facilitated the adsorption and desorption of gas molecules, which was also very important to improve the performance of the gas sensor (Liu et al. 2014; Zhang et al. 2020). The above aspects contributed together to the excellent gas sensing performance of CQDs/PCFT film.