Carbon monoxide (CO) which is colorless, odorless, but toxic, explosive, mainly produced by automotive emissions, coal combustion and forest fires etc.^[1] Actually, any incomplete combustion of carbon could generate CO. However, CO is extremely harmful to human beings causing dizziness, nausea or even death^[2]. According to the data from World Health Organization, one can tolerate 9 μL/L CO within 8 h, 26 μL/L within 1 h, and beyond this the gas could threaten human life^[3]. As a result, there are various CO gas sensors to monitor CO. However, to the best of our knowledge, the detection ranges of these CO gas sensors are limited. Therefore, developing CO gas sensors with a wide detection range becomes a crucially important task.

Many materials could be used to detect CO gas^[4]. Taking flexibility and versatility in practical applications into account, metal oxides are promising sensing materials for CO gas sensors^[5]. Specifically, SnO₂, ZnO, Co₃O₄ and TiO₂etc. have been reported as CO gas sensors, which exhibit a detection range from dozens of μL/L to thousands of μL/L^[6,7]. Among them, cerium oxide (CeO₂), featured of structural stability and high redox capability^[8], has been extensively investigated in gas sensing field^[9]. Good CO sensing performance was obtained by pristine CeO₂ nanospheres with high surface area and the detection range was 10-30 μL/L^[10]. SnO₂ mixed CeO₂ thin films exhibited a response/recovery time of 26 s/30 s towards 500 μL/L CO and the detection concentration was in the range of 5-5000 μL/L^[11]. Although various CeO₂, such as samples with different morphologies, mixture of other metal oxides or precious metals, has been reported, the detection range of these CO sensors are not wide enough as well as dozens of seconds of response time to meet practical requirements in some cases.

Two-dimensional nanosheets exhibit superiority in specific surface area, mobility of ions and the number of active sites, being applied in many aspects^[12]. In the context of gas sensing, two-dimensional nanosheets also show remarkable performance. For instance, Liu et al.^[13] synthesized porous ultrathin ZnO nanosheets with high specific surface areas for acetylacetone gas sensing and presented significant results, demonstrating the important role of oxygen vacancies in gas sensing spontaneously. Self-assembled monolayer CuO nanosheets were used for H₂S gas sensing with improved sensitivity^[14]. Selective alkene gas sensing was achieved by SnO₂ nanosheets with distinct performance, which is affected by the content of oxygen vacancies^[15]. However, to the best of our knowledge, there is almost few reports about CeO₂ NSs, especially for the influence of oxygen vacancies of CeO₂ NSs on the sensing performance.

In this study, we prepared CeO₂ NSs via flame annealing under different time intermediate product CeCO₃OH nanosheets synthesized by simple hydrothermal method. The as-prepared samples were characterized by various methods. These flame-annealed CeO₂ NSs show ultrafast response/recovery time towards CO, wide-range CO detection as well as good reproducibility and selectivity. Moreover, the mechanism of CO gas sensing is discussed.

Cerium (Ⅲ) chloride heptahydrate (CeCl₃·7H₂O) and sodium oleate were purchased from Shanghai Macklin Biochemical Co., Ltd. Ammonium hydroxide (NH₃·H₂O) and cyclohexane were provided by Sinopharm Chemical Reagent Co., Ltd. All of the chemicals and solvents were reagent-grade without further purification. Deionized water was used in all procedure.

Intermediate product CeCO₃OH was synthesized by a simple hydrothermal method according to our previous work^[16] as shown in Fig.1. The specific procedure was listed as follows. CeCl₃·7H₂O solution was dropped into sodium oleate solution with stirring. After that, NH₃·H₂O was dropped into it with stirring. Subsequently, the suspension was proceeded with hydrothermal reaction. The precipitant was collected by centrifugation and washed with cyclohexane. Whereafter, the intermediate product was annealed with liquefied butane gas flame for 0.5, 2, 5 min and the resulted CeO₂ nanosheets were denoted as CeO₂-0.5min NSs, CeO₂-2min NSs, and CeO₂-5min NSs, respectively (The flame temperature is in the range of 1000-1200 ℃). The samples were collected for further characterization.

The crystal structure was confirmed by X-ray Diffraction (XRD, Bruker AXS D8 Advanced Diffractometer, German) with Cu Kα radiation (λ=0.15418 nm). The morphology and structures were characterized via transmission electron microscope (TEM, Tian G2 60-300, FEI, USA). The element compositions and chemical states were obtained by X-ray photoelectron spectroscope (XPS, Thermo Scientific Escalab 250Xi, USA) equipped with a monochromatic Al Kα source. The specific surface area and porosity were evaluated using an automated surface area and pore size analyzer (Quantachrome Autosorb iQ3, USA), calculated by Brunauer-Emmett-Teller (BET) and Barrett-Joyner-Halenda (BJH) methods on the basis of nitrogen adsorption-desorption isotherms, respectively. The presence of oxygen vacancies was determined using electron paramagnetic resonance (EPR, Bruker A300, German).

Specialized CGS-1 TP intelligent gas sensing analyzing system (Beijing Elite Tech Co., Ltd, Beijing, China) was adopted in this work and the fabrication detail and testing principal could be found in our previous report^[17]. The gas sensing response (S) was defined as the following formula:

where R_a and R_g represent the resistance of sensing materials in air and target gas, respectively. The response and recovery time were labelled as the time to reach 90% of the total resistance change.

In order to identity the crystal structure, the intermediate product and the as-prepared CeO₂ nanosheets were carefully investigated by XRD. The patterns of intermediate product perfectly matches with CeCO₃OH (JCPDF 32-0189) as shown in Fig. S1. XRD patterns of CeO₂- 0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs show sharp diffraction peaks, which demonstrates high crystallinity as shown in Fig. 2. These diffraction peaks can be well indexed to cubic fluorite cerium oxide (JCPDF 34-0394). The typical peaks located at 2θ=28.6°, 33.0°, 47.5°, 56.3°, 59.1°, 69.4°, 76.7°, 79.0°, correspond to (111), (200), (220), (311), (222), (400), (331), (420) crystal planes of fluorite phase of CeO₂, respectively. However, the width of main diffraction peak of (111) at 2θ=28.6° is quite different. With flame annealing time increasing, the diffraction peak width become narrower, demonstrating the increment of crystal sizes. The crystal sizes based on (111) crystal plane of CeO₂-0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs were calculated to be 12.12, 17.02 and 20.32 nm, respectively, according to the Scherer equation.

The element compositions and surface chemical states could be detected by XPS, and the details are given in Fig. 3. For O element, the O1s core level spectra of CeO₂- 0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs can be deconvolved into three peaks, i.e. lattice oxygen (O_L, ca. 529.0 eV), deficient oxygen (O_V, ca. 529.5 eV) and sorbed oxygen (O_S, ca. 531.2 eV) in Fig. 3(a, b, c), respectively^[18]. As is well-known, O_V and O_S (O₂^-, O^- or O^2-) play absolutely important roles in gas sensing, contributing to improving response^[19]. Table S1 provides the ratio of each oxygen species to the whole oxygen extent. In this study, CeO₂-2min NSs possesses the largest amount of O_V (43.2%), and these for CeO₂-0.5min NSs and CeO₂-5min NSs are relatively smaller. For Ce element, the Ce3d core level spectra of CeO₂-0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs can be deconvolved into ten peaks in Fig. 3(d, e, f), respectively. The peaks at about 916.4, 907.1, 902.4 and 900.5 eV are designated to Ce3d_3/2, and that at about 898.2, 897.8, 888.5, 884.5, 882.1 and 881.4 eV pertain to Ce3d_5/2^[20]. Thereinto, the peaks at about 881.4, 884.5, 897.8 and 902.4 eV are assigned to Ce³⁺, and those at around 882.1, 888.5, 898.2, 900.5, 907.1 and 916.4 eV belong to Ce^4+[21]. Moreover, the ratio of Ce³⁺ to the total extent of Ce is presented in Table S1. It was reported that there is a transformation between Ce⁴⁺ and Ce³⁺ which determines by the oxygen partial pressure. That means that with more Ce³⁺, more O_V exists in the samples^[8]. In this case, CeO₂-2min NSs holds the largest part of Ce³⁺ (28.6%), consistently with the largest part of O_V (43.2%).

The morphology and structure were characterized by TEM. The spectra of CeO₂-0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs are given in Fig. 4(a, b, c), respectively, which are transparent nanosheets with pores (labelled by red dotted circles). Nevertheless, there is a difference in pore diameters that with the increase of flame time, the pore diameter becomes larger. This result is consistent with the tendency of nitrogen adsorption-desorption discussed later, which may influence the gas sensing performance. More importantly, with the increment of flame annealing time, the grain sizes become larger, which is consistent with the results of XRD. The crystal structure and crystal lattice spacing of the samples can be observed by HRTEM, as shown in Fig. 4(d,e,f), respectively. The distance of two adjacent crystal planes is 0.31 nm, corresponding to (111) planes of the CeO₂ fluorite phase. Obviously, crystallinity increases along with a longer flame annealing time.

Nitrogen adsorption-desorption isothermal curves were performed to characterize the aspect surface area and porosity of the samples, as shown in Fig. 5. It is observed from Fig. 5(a) that all samples exhibit the isotherm feature of the IV-type with a hysteresis loop according to the IUPAC classification, which indicates the existence of a mesoporous structure in nanosheets^[22]. The surface area of CeO₂-0.5min NSs, CeO₂-2min NSs and CeO₂-5min NSs are 73.344, 46.804 and 40.836 m²∙g^-1, respectively, summarized in Table S2. It is noticed that with the flame annealing time increasing, the surface area becomes smaller, which may be contributed to congregation and growth of small grains during the annealing process. Nevertheless, there is a variance in pore size distribution.As shown in Fig. 5(b), CeO₂-2min NSs (15.5 nm) and CeO₂-5min NSs (15.8 nm) possess larger pore diameter than CeO₂-0.5min NSs (10.1 nm), which indicates gas molecules easily adsorb and desorb on the surface of CeO₂-2min NSs and CeO₂-5min NSs as well as a fast response/recovery time.

To further verify the presence of oxygen vacancies, EPR were performed to ascertain relative amount of oxygen vacancies in each sample, as shown in Fig. 6. The value of g represents different species which contain unpaired electron. In this case, all samples have the same value of g(2.0038), which typically demonstrates the presence of oxygen vacancies in the samples^[19]. The intensity of CeO₂-2min NSs and CeO₂-5min NSs is almost identical and far larger than that for CeO₂-0.5min NSs, which confirms that oxygen vacancies in CeO₂-2min NSs and CeO₂-5min NSs are abundant and much higher than that in CeO₂-0.5min NSs, consistent with the results of XPS.

We systematically investigated the gas sensing performances of the samples, as shown in Fig. 7. First of all, the response of the sensors towards 500 μL/L CO were measured from 300 ℃ to 500 ℃ with the interval of 50 ℃ to determine the optimal temperature. For CeO₂-2min NSs and CeO₂-5min NSs, the optimal temperature is 450 ℃, and that for CeO₂-0.5min NSs is 400 ℃ in Fig. 7(a). However, the responses of CeO₂-2min NSs (10%) and CeO₂-5min NSs (9%) are far higher than that for CeO₂- 0.5min NSs (4%). The responses of CeO₂-2min NSs and CeO₂-5min NSs are considerable, which may be ascribed to the comparable O_V content demonstrated by XPS (Table S1, 43.2% and 42.7%, respectively) and EPR (Fig. 6). For a little higher response of CeO₂-2min NSs than CeO₂-5min NSs, the difference between surface area (Table S2) and the content of Ce³⁺ (Table S1) may account for it. However, there is a large gap between the response of CeO₂-2min NSs and CeO₂-0.5min NSs, which may be resulted from the discrepancy of the amount of O_V (Table S1 and Fig. 6) and pore diameters (Fig. 5(b)). Therefore, CeO₂-2min NSs was chosen to proceed further testing at its optimal temperature(450 ℃).

Reproducibility is a basic requirement for gas sensors to practical application. 500 μL/L CO at 450 ℃ was adopted to test reproducibility. In Fig. 7(b), it is found that the response of 7-time testing is stable without any decrease. It demonstrates that CeO₂-2min NSs possesses well reproducible ability in CO gas sensing.

Relationship between CO concentration and its response is important to determine real CO concentration in CO gas sensor applications. Different concentration of CO was subject to test at 450 ℃ as shown in Fig. 7(c). The highest concentration of the test is 10000 μL/L, and with the decrease of the concentration, the response also declines. Meanwhile, the lowest concentration tested is 10 μL/L with the response of 1.5%. Furthermore, the response and the concentration of CO (10-10000 μL/L) is fitted as shown in Fig. 7(d) and it keeps functional relationship of y=1.09x^0.45 with R²=0.9908. It proves that CeO₂-2min NSs is a promising material in CO gas sensing with a wide range. In addition, it is impressive that the response/recovery time are ultrafast in the sensing procedure, which is expected for a gas sensor.

To investigate response/recovery time carefully, we chose a cycle of transient response in reproducibility test for further research. As can be seen in Fig. 7(e), the response and recovery time are both 2 s, which is very short in gas sensing. As discussed above, carriers on nanosheets move quickly because the morphology constraints its pathway in planes. In this case, electrons move in planes of CeO₂ nanosheets, which promises ultrafast response/recovery time. Therefore, it is reasonable to deem that CeO₂-2min NSs is advantageous in CO gas sensing. Moreover, recent progresses in CO gas sensing are summarized in Table 1, and it is concluded that CeO₂-2min NSs outperforms in response/recovery time.

Selectivity is a valuable index to evaluate gas sensor. We also studied the selectivity of CeO₂-2min NSs in gas sensing was studied as presented in Fig. 7(f). 500 μL/L of CH₄, H₂, NH₃, NO₂ and CO are subject to test, and it is found that the response of CO is much higher than that to other gases. Therefore, CeO₂-2min NSs is very promising in CO gas sensing.

In addition, it is noticed that humidity has obvious impact on CeO₂-2min NSs response (Fig. S2), namely high humidity for low response, which may be arisen from the water molecules occupy the active sites on CeO₂ NSs at high humidity. Besides, long-term stability of CeO₂-2min was also performed as shown in Fig. S3. Although the response has a small decrease in two weeks, long-term stability is improved by doping or coating in followup work.

The gas sensing mechanism of CeO₂ NSs follows a surface charge model, which can be explained by the change of resistance in different target gases, as shown in Fig. 8^[5,28]. In oxygen atmosphere (ambient air), oxygen molecules are adsorbed on the surface of the sensing materials by capturing free electrons, forming adsorbed oxygen anion species depended on the temperature^[29,30]. In this study, when CeO₂ NSs are exposed to air, oxygen molecules are adsorbed on the surface trapping electrons from them, which forms depletion layer and an increased resistance. Subsequently, oxygen anions change as the sequence of O₂^-, O^- and O^2- along with the temperature rising. At last, when reducing gas CO is introduced, the CO gas molecules adsorb on the surface of CeO₂ NSs and then react with O^2- (the operating temperature is 450 ℃). The whole procedure may proceed as following equations:

During this reaction, the release of electrons into the surface of CeO₂ NSs increases the concentration of carriers, leading to a thinner depletion layer as well as a decrease of resistance.

In this case, CeO₂-2min NSs possesses the largest part of O_V, which is the active site for facilitating oxygen adsorption and dissociation by providing unpaired electrons^[13]. Besides, CeO₂-2min NSs holds largest pore diameter, which guarantees that CO molecules penetrate into the inside of sensing materials to touch more active sites^[31]. Therefore, CeO₂-2min NSs deserves the highest response as discussed above. On the contrary, the small amount of as well as small pore diameters of O_V of CeO₂-0.5min NSs, which pays for a low response. In addition, nanosheets can constrain carriers to transport in planes, which make carriers possess ultrafast mobility. On the other hand, oxygen vacancies can also increase mobility of carriers with easy movement between vacancies. These factors all contribute to ultrafast response/recovery time in this study.

In summary, CeO₂ NSs via flame annealing intermediate product CeCO₃OH nanosheets was synthesized by hydrothermal method with different flame time. Wherein, CeO₂-2min NSs showed superb gas sensing performances. The response/recovery time were only 2 s/2 s, which was ultrafast in gas sensing field. In addition, CO concentration and its response kept well functional relationship at a wide detection range. Moreover, reproducibility and selectivity were also decent. The good performances are attributed to more oxygen vacancies and porous nanosheets morphology. As a result, this CeO₂ NSs is reliable to apply in CO gas sensing. Meanwhile, flame annealing provides a novel method to prepare nanomaterials with simplicity and low cost.