Construction of 1D/2D α-Fe2O3/SnO2 Hybrid Nanoarrays for Sub-ppm Acetone Detection

Exhaled acetone is one of the representative biomarkers for the noninvasive diagnosis of type-1 diabetes. In this work, we have applied a facile two-step chemical bath deposition method for acetone sensors based on α-Fe2O3/SnO2 hybrid nanoarrays (HNAs), where one-dimensional (1D) FeOOH nanorods are in situ grown on the prefabricated 2D SnO2 nanosheets for on-chip construction of 1D/2D HNAs. After annealing in air, ultrafine α-Fe2O3 nanorods are homogenously distributed on the surface of SnO2 nanosheet arrays (NSAs). Gas sensing results show that the α-Fe2O3/SnO2 HNAs exhibit a greatly enhanced response to acetone (3.25 at 0.4 ppm) at a sub-ppm level compared with those based on pure SnO2 NSAs (1.16 at 0.4 ppm) and pure α-Fe2O3 nanorods (1.03 at 0.4 ppm), at an operating temperature of 340°C. The enhanced acetone sensing performance may be attributed to the formation of α-Fe2O3–SnO2 n-n heterostructure with 1D/2D hybrid architectures. Moreover, the α-Fe2O3/SnO2 HNAs also possess good reproducibility and selectivity toward acetone vapor, suggesting its potential application in breath acetone analysis.


Introduction
As a potential alternative for the noninvasive diagnosis of disease, exhaled breath analysis has been proposed and developed over the past decades [1][2][3]. The exhaled breath of human beings includes not only nitrogen, oxygen, carbon dioxide, nitric oxide, and water vapor but also a mixture of volatile organic compounds (VOCs) and some other nonvolatile molecules. Encouragingly, a few of them have been regarded as biomarkers to diagnose diseases (Table S1). For example, formaldehyde (lung cancer) [4], toluene (lung cancer) [5], ammonia (hemodialysis) [6], H 2 S (halitosis) [7], isoprene (heart disease) [8], benzene (smoker) [9], and pentane (acute asthma) [10] at few dozens to few thousands of ppb are known as biomarkers for patients. Researchers have also found that exhaled acetone can intuitively correlate with type-1 diabetes, which may exceed 1.8 ppm (only 0.3-0.9 ppm for healthy people) [1,11]. Therefore, an ultrasensitive acetone sensor is of great importance to detect acetone vapor at a sub-ppm level.
Metal oxide semiconductors (MOXs), such as SnO 2 , ZnO, α-Fe 2 O 3 , CuO, and NiO, have been widely explored in the field of gas detection owing to their simple and cost-effective synthesis, high sensitivity, and good stability. Among these MOXs, α-Fe 2 O 3 is a multifunctional n-type semiconductor with a direct bandgap (E g = 2:2 eV at 300 K) that has been intensively investigated in the field of gas sensing [12][13][14][15]. Several effective strategies have been designed to improve the gas sensing properties of these MOXs, such as doping, surface modification, porous/hollow structures, and hierarchical architectures [16][17][18]. Recently, construction of hybrid nanostructures is rapidly emerging as a fascinating strategy that combines different MOXs with precise control of their morphologies, such as hollow ZnO/ZnFe 2 O 4 heterostructures that were synthesized by growing ultrathin ZnFe 2 O 4 nanosheets on the outer surface of ZnO hollow microspheres [19], NiO nanoparticle-decorated SnO 2 nanosheets [20], CuO nanosheets/ZnO nanorods (NRs) [21], and Fe 2 O 3 nanoparticle-decorated CuO NRs [22]. For this purpose, the rational combination of SnO 2 and α-Fe 2 O 3 has been proven to improve their gas sensing performances (Table S2). The results show that the α-Fe 2 O 3 /SnO 2 composites present excellent sensing performances to acetone [23,24], ethanol [25][26][27][28][29][30], toluene [31], and LPG [32,33]. Moreover, their gas sensing properties can be largely affected by the size and shape of nanobuilding blocks (α-Fe 2 O 3 and SnO 2 ). As far as we know, only a few reports about α-Fe 2 O 3 -SnO 2 system have concerned on the detection of ultralow concentrations of acetone.
In this work, we report a two-step chemical bath deposition (CBD) method to construct the α-Fe 2 O 3 /SnO 2 hybrid nanoarrays (HNAs) on-chips with subsequent annealing in air. The one-dimensional (1D) α-Fe 2 O 3 NRs are distributed homogenously on the surface of the 2D SnO 2 nanosheets to construct novel 1D/2D HNAs. In comparison with pure SnO 2 NSAs and α-Fe 2 O 3 NRs, the α-Fe 2 O 3 /SnO 2 HNAs show a dramatically enhanced response to acetone (down to sub-ppm). Moreover, the α-Fe 2 O 3 /SnO 2 HNAs also possess a superior selectivity to acetone against other interfering gases (formaldehyde, toluene, benzene, and ammonia). A possible sensing mechanism based on the formation of α-Fe 2 O 3 -SnO 2 n-n heterostructure is proposed.

Results and Discussion
2.1. Morphological Characteristics. As illustrated in Figure 1(a), a chip with interdigital Au electrodes (200 μm lines separated by 200 μm gaps) was fabricated on a (100) silicon substrate with a 2 μm thermally grown SiO 2 layer, and the SnO 2 NSAs were prepared with an on-chip growth method similar to our previous work [34]. The chip was vertically dipped into the mixed solution (containing Sn 2+ and CO(NH 2 ) 2 ) during this process, and then the prefabricated SnO 2 NSAs were immersed in another aqueous solution (containing Fe 2+ and CO(NH 2 ) 2 ) for depositing FeOOH NRs on SnO 2 NSAs. After annealing in air, the as-prepared sensors were placed on a CGS-4TP gas sensing measurement system. A schematic of the gas sensing measurement systems used in this work is illustrated in Figure S1. Figure 1(b) presents a schematic diagram of the test platform used in this work. A hotplate was used to adjust the operating temperature, and two pins of a sensor were connected with a pair of probes. A digital photograph of the gas sensing chip (3 mm × 6 mm in size) with α-Fe 2 O 3 /SnO 2 HNAs is given in the inset of Figure 1(b). During the test, four sensors were measured simultaneously, as shown Figure 1(c), and the electrical resistance of each sensor was recorded. Figure 2 shows the morphologies of as-prepared pure SnO 2 NSAs and α-Fe 2 O 3 /SnO 2 HNAs. It can be seen from Figure 2(a) that the pure SnO 2 NSAs are composed by oriented growth of nanosheets, where the adjacent SnO 2 nanosheets will interconnect with each other and form a semiopen network. From the cross-sectional scanning electron microscope (SEM) images of SnO 2 NSAs (Figure 2(b)), the flake-like SnO 2 stands vertically on the chip with a uniform film thickness (~100 nm), and the SnO 2 NSAs are robustly adhered to the substrate. After the growth of α-Fe 2 O 3 NRs, it is obvious that the surface morphology of α-Fe 2 O 3 /SnO 2 HNAs is much different from that of the pure SnO 2 NSAs. As shown in Figure 2(c), numerous ultrathin α-Fe 2 O 3 NRs are homogenously distributed among the interconnected SnO 2 NSAs. A SEM close-up image of α-Fe 2 O 3 NRs (inset Figure 2(c)) reveals that the diameter of NRs ranges from 9 nm to 20 nm, and the average diameter is about 12.7 nm. Otherwise, for the second-step (CBD method), the α-Fe 2 O 3 NRs tend to form irregular aggregates without a substrate ( Figure S2). The cross-sectional SEM images of α-Fe 2 O 3 /SnO 2 HNAs in Figure 2(d) further indicate that the 1D α-Fe 2 O 3 NRs are in situ grown on the surface of 2D SnO 2 nanosheets, and novel 1D/2D hybrid nanoarrays can be achieved by a facile two-step CBD method. At the same time, the average film thickness of α-Fe 2 O 3 /SnO 2 HNAs increases up to 220 nm.
To get further insight into the definite morphology of pure SnO 2 NSAs and α-Fe 2 O 3 /SnO 2 HNAs, transmission electron microscope (TEM) images were taken from the   (Figure 3(c)) elements were distributed randomly and uniformly on the substrate, which was in agreement with the fact that the hybrid composites were constructed by hybrid α-Fe 2 O 3 and SnO 2 . The existence of Sn and Fe elements was confirmed by the EDS characterization, and the atomic percentages of Sn and Fe were measured to be 5.38% and 4.17%, respectively.
Moreover, XPS analysis was used to obtain more information about the chemical valences of our samples. Figure 4(a) displays the high-resolution Sn 3d spectra of SnO 2 NSAs and α-Fe 2 O 3 /SnO 2 HNAs. In the pure SnO 2 NSAs, the two peaks centered at 487.4 and 495.8 eV can be ascribed to the peaks of Sn 3d 5/2 and Sn 3d 3/2 , respectively, which are in good agreement with Sn 4+ . With the modification of α-Fe 2 O 3 , a slight negative shift of the binding energies is observed in α-Fe 2 O 3 /SnO 2 HNAs, shifting to 487.3 and 495.7 eV, respectively, as a result of the formation of α-Fe 2 O 3 /SnO 2 heterojunction interface. In the spectrum of Fe 2p in Figure 4(b), interference peaks are detected at 717.6 eV in SnO 2 NSAs and 716.5 eV in α-Fe 2 O 3 /SnO 2 HNAs, which come from the Sn 3p peak. The Fe 2p peaks are not found in α-Fe 2 O 3 /SnO 2 HNAs due to the strong interference peak. In comparison, the two peaks at 711.9 and 725.2 eV detected in pure α-Fe 2 O 3 NRs are attributed to Fe 2p 3/2 and Fe 2p 1/2 , respectively, corresponding to Fe 3+ in α-Fe 2 O 3 .

Gas Sensing Properties.
As is well-known, the gas sensing properties of MOXs are highly dependent on the operating temperature. To confirm it, the as-prepared gas sensors were examined at various temperatures (280-380°C) toward 5 ppm acetone. The sensor response is defined as R a /R g , where R a and R g are the sensor resistance in air and target gas, respectively. As shown in Figure 5(a), the response of α-Fe 2 O 3 /SnO 2 HNA-based sensor increases with the increase in operating temperature and reaches its maximum value (13.63) at 340°C, then decreases with the further increase of operating temperature. Therefore, 340°C can be chosen as the optimum operating temperature of α-Fe 2 O 3 /SnO 2 HNAs. Differently, the pure SnO 2 NSA-based sensor exhibits no obvious variation over the whole temperature range (2.00, at 340°C). The pure α-Fe 2 O 3 NR-based sensor shows a monotonic decrease of the response with an increase in operating temperature, and the highest response value is about 4.28 at 280°C. It is clear that the α-Fe 2 O 3 /SnO 2 HNA-based sensor displays the highest response in the three sensors, revealing that the acetone sensing properties of SnO 2 NSAs can be significantly enhanced by the modification of α-Fe 2 O 3 NRs. Figure 5(b) gives the acetone sensing properties of the above three sensors at the same operating temperature of 340°C. It is clear that the sensor response increases with the acetone concentration ranging from 0.4 to 20 ppm for each sensor. Especially, in the case of α-Fe 2 O 3 /SnO 2 HNAs, the response increases rapidly over the whole concentration range, which is rather different from the other two sensors.      Figure 5(e)) than that of pure SnO 2 NSAs (204.05 kΩ, Figure 5(d)) and α-Fe 2 O 3 NRs (132.298 MΩ, Figure S3(b)). Upon exposure to acetone gas, the sensor resistance quickly decreases as expected and then recovers to its R a after being exposed to air. R g decreases monotonically with the increase of acetone concentration; in other words, the sensor response increases (refer to Figure 5(c) and Figure S3(a)).
The response and recovery times (t res and t rec ) are very important parameters for high-performance gas sensors. The response time t res (or recovery time t rec ) is defined as the time required to reach 90% resistance change when the sensor is exposed to target gas (or air). As shown in Figure S4, the α-Fe 2 O 3 /SnO 2 HNA-based sensor shows a faster t res (14 s, at 1 ppm) at 340°C compared with that of the SnO 2 NSAs (37 s, at 1 ppm). On the contrary, the t rec of the α-Fe 2 O 3 /SnO 2 HNA-based sensor always exceeds one minute (62-159 s, in the range 0.4-20 ppm), which is apparently higher than that of the SnO 2 NSA-based sensor (22-34 s, in the range 0.4-20 ppm). According to the previous studies, the vertically ultrathin SnO 2 NSAs can provide as much surface area as possible to adsorb gas molecules and facilitate the adsorption/desorption of the acetone gas. When the α-Fe 2 O 3 NRs were introduced, the branched α-Fe 2 O 3 NRs on the surface of SnO 2 NSAs will adsorb more acetone molecules, making the α-Fe 2 O 3 /SnO 2 HNAs more sensitive and present faster response toward acetone.   To study the selectivity in the above sensors, some interfering gases (formaldehyde, toluene, benzene, and ammonia) were measured at 340°C with a low concentration of 1 ppm. It can be seen in Figure 5 The reproducibility of the sensors at 340°C has been investigated by continuously testing the sensors to 5 ppm and 1 ppm acetone with 5 cycles for each. As shown in Figures 6(a)-6(c), all the sensors maintain their response   To assess the long-term stability of our sensor, we tested the α-Fe 2 O 3 /SnO 2 HNA-based sensor for 93 days toward 5 ppm acetone at 340°C. The mean response of the α-Fe 2 O 3 /SnO 2 HNA-based sensor is 11.62 with a standard deviation estimated to be 0.93 during the whole period, suggesting its stability for acetone detection over a long period. Furthermore, it can be clearly seen that there is no obvious change between the nanostructures of α-Fe 2 O 3 /SnO 2 HNAs before (Figure 2(c)) and after (inset of Figure 6(e)) a number of gas sensing tests. This observation is consistent with the good long-term stability of the α-Fe 2 O 3 /SnO 2 HNA-based sensor.
It is well-known that human exhaled breath is highly humid (RH ≥ 80%) and the existence of water vapor has a significant influence on the gas sensing performance for MOX-based gas sensors. As shown in Figure 6(f), the response of α-Fe 2 O 3 NRs to 5 ppm acetone was measured as a function of relative humidity (20%-90% RH). The responses of α-Fe 2 O 3 /SnO 2 HNAs under 20%, 40%, 60%, 80%, and 90% RH were 12.34, 6.02, 4.69, 3.86, and 3.62, respectively. Obviously, the response of α-Fe 2 O 3 /SnO 2 HNAs is highly dependent on relative humidity, and some available approaches (such as employing water filtering membranes) are needed to eliminate the influence of water vapor.
Considering the previous reports in Table 1 [11,[35][36][37][38], the α-Fe 2 O 3 /SnO 2 HNA-based sensor in this work possesses relatively medium sensitivity (or operating temperature). We can conclude that the acetone sensing properties of MOXs can be further enhanced by constructing heterostructures or modifying with noble metals. As mentioned before, the acetone detection capability (or resolution) for the diagnosis of diabetes mellitus should be as low as sub-ppm, all of which need sufficient and reliable sensors for acetone. In this sense, the high sensitivity, good selectivity, and excellent reproducibility of the α-Fe 2 O 3 /SnO 2 HNA-based sensor imply that it can potentially be used for breath acetone analysis.

Sensing Mechanism.
For n-type MOXs (SnO 2 and α-Fe 2 O 3 ), their acetone sensing mechanisms can be briefly understood as the reaction between the adsorbed oxygen species and acetone molecules on the active sites of sensitive materials. [24] In general, an electron depletion layer (EDL) can be formed on the near surface of SnO 2 nanosheets (Figure 7(a)) owing to the adsorbed oxygen species (O -, O 2 -, and O 2-) after exposing to air and makes the absorbed oxygen species capture free electrons from the conduction band of SnO 2 . This results in a decrease of electron concentration (or width of EDL) and thus a relatively high resistance in air atmosphere. The generation and transformation processes of the oxygen species at different operating temperatures have the following expressions [39]: On the contrary, upon exposure to reducing gases such as acetone, acetone molecules will react with the absorbed oxygen species, as expressed by Equation (2) and release free electrons back to the SnO 2 nanosheets. Hence, the electron concentration increase will cause an increase in conductivity (or a decrease in sensor resistance), and the width of EDL also becomes broader. According to the SEM and TEM observations, the vertically distributed SnO 2 nanosheets connect with each other to construct an interlaced electron transport network on the substrate. However, the modulation mechanism of this type of transport network is not efficient because of the same energy band structure in the pure SnO 2 NSAs. In other words, the R a of the pure SnO 2 NSA-based sensor is rather low (Figure 5(d)), which is too difficult to obtain dramatic change, especially at low acetone concentrations.
In the case of α-Fe 2 O 3 /SnO 2 HNAs, the sensor exhibits enhanced sensitivity toward acetone, this may be attributed to the following reasons: (1) In the formation of α-Fe 2 O 3 -SnO 2 n-n heterostructures, by combining these two MOXs with different work functions (SnO 2 : Φ = 4:9 eV; α-Fe 2 O 3 : Φ = 5:88 eV) [40,41], the free electrons tend to transfer from the higher side to the lower side, until the equilibrium Fermi level is reached (Figure 7(c)) [26,42]. In this process, the SnO 2 nanosheet near the heterostructure interface will lose more electrons, which leads to a broader conduction region in air (Figure 7(b)) [43]. Similar to other reports, the R a of the α-Fe 2 O 3 /SnO 2 HNAs in this work is much higher than that of the pure SnO 2 NSAs.
(2) In the novel 1D/2D α-Fe 2 O 3 /SnO 2 hybrid architectures, when the sensor is exposed to acetone, the stretchedout α-Fe 2 O 3 NRs provide an extra surface area and active sites for the gas adsorption. Thus, more oxygen species and acetone molecules can be adsorbed on the surface of α-Fe 2 O 3 /SnO 2 HNAs (Figure 7(b)), which provides more opportunities for Equation (2). The conduction region in SnO 2 nanosheets will be broadened as well as a decrease in R g . On the other hand, the free electrons generated on the surface of α-Fe 2 O 3 NRs will flow to SnO 2 NSAs and allow a dramatic decrease in the width of the electron depletion region at the interface of the α-Fe 2 O 3 /SnO 2 heterostructure. It may further result in an increase of sensor response toward acetone. So the modulation mechanism of α-Fe 2 O 3 /SnO 2 HNAs becomes more efficient than that of pure SnO 2 NSAs. Additionally, much hard work is still needed to study the influence of ambient humidity, filter units, and clinical tests, making it more suitable for breath acetone analysis.

Conclusion
The α-Fe 2 O 3 /SnO 2 HNA-based acetone sensor has been fabricated via a facile two-step on-chip growth (or CBD method) process. The results indicate that the as-prepared sensor presents a well-defined 1D/2D hybrid architecture, where the ultrathin α-Fe 2 O 3 NRs (an average diameter~12.7 nm) are distributed among the interconnected SnO 2 NSAs. Gas sensing measurements show that the α-Fe 2 O 3 /S-nO 2 HNA-based sensor exhibits superior acetone sensing properties (high sensitivity, good reproducibility, and selectivity), even at a sub-ppm level, compared with those of the pure SnO 2 NSA-and α-Fe 2 O 3 NR-based sensors. The improved acetone sensing performance may be due to the formed α-Fe 2 O 3 -SnO 2 heterostructures and their  unique hybrid nanostructures. Our work suggests that the α-Fe 2 O 3 /SnO 2 HNAs can be a promising candidate for sub-ppm acetone detection in breath analysis.

Materials and Methods
4.1. Preparation of α-Fe 2 O 3 /SnO 2 HNAs. In brief, 0.6 mmol SnCl 2 ·2H 2 O and 0.8 mmol CO(NH 2 ) 2 were dissolved into 20 mL deionized water and stirred for 15 min at room temperature. Then a piece of chip with several Au electrodes was washed with acetone and ethanol and deionized water for several times, which was afterwards vertically dipped into the above solution and maintained at 95°C for 8 h. After washing and drying at 60°C in an oven, the prefabricated SnO 2 NSAs were immersed in an aqueous solution (containing 0.1 M FeSO 4 ·7H 2 O and 1.0 M CO(NH 2 ) 2 ) and kept at 80°C for 1 h. Similarly, the chip was washed and dried again at 60°C. The final chip was annealed at 400°C for 3 h in air to achieve α-Fe 2 O 3 /SnO 2 HNAs. In addition, the pure SnO 2 NSAs were also annealed under the same conditions. For a pure α-Fe 2 O 3 NR-based sensor, 0.1 M FeSO 4 ·7H 2 O and 1.0 M CO(NH 2 ) 2 were mixed and maintained at 80°C for 1 h, and the collected precipitate was dip-coated on the Au electrodes and then annealed at 400°C in air for 3 h.

Characterization and Gas Sensing Measurements.
The morphologies and compositions of as-prepared products were investigated by a scanning electron microscope (SEM, Zeiss Gemini 300) equipped with energy dispersive X-ray (EDX) spectroscope and a high-resolution transmission electron microscope (HRTEM, FEI Tecnai G2 F30). The chemical states of the surface species were determined by using X-ray photoelectron spectroscopy (XPS, ESCALB 250Xi). The gas sensing properties of sensors were performed on a commercial CGS-4TPs system (Beijing Elite Tech Co., Ltd., China). Gaseous acetone diluted with dry air was injected by a syringe. The operating temperature ranges from 280 to 380°C with a relative humidity around 20%. Figure S1: schematic drawing of the gas sensing measurement systems. Figure S2: SEM images of pure α-Fe 2 O 3 NRs. Figure S3: transient response curve of α-Fe 2 O 3 NRs toward different acetone concentrations at 340°C and the corresponding resistance curve. Figure S4: response/recovery time vs. acetone concentration (0.4-20 ppm) with SnO 2 NSAs and α-Fe 2 O 3 /SnO 2 HNAs at 340°C. Table S1: brief summary of results reported on exhaled breath analysis.