Electrospun Cu-doped In2O3 hollow nanofibers with enhanced H2S gas sensing performance

One-dimensional nanofibers can be transformed into hollow structures with larger specific surface area, which contributes to the enhancement of gas adsorption. We firstly fabricated Cu-doped In2O3 (Cu-In2O3) hollow nanofibers by electrospinning and calcination for detecting H2S. The experimental results show that the Cu doping concentration besides the operating temperature, gas concentration, and relative humidity can greatly affect the H2S sensing performance of the In2O3-based sensors. In particular, the responses of 6%Cu-In2O3 hollow nanofibers are 350.7 and 4201.5 to 50 and 100 ppm H2S at 250 °C, which are over 20 and 140 times higher than those of pristine In2O3 hollow nanofibers, respectively. Moreover, the corresponding sensor exhibits excellent selectivity and good reproducibility towards H2S, and the response of 6%Cu-In2O3 is still 1.5 to 1 ppm H2S. Finally, the gas sensing mechanism of Cu-In2O3 hollow nanofibers is thoroughly discussed, along with the assistance of first-principles calculations. Both the formation of hollow structure and Cu doping contribute to provide more active sites, and meanwhile a little CuO can form p–n heterojunctions with In2O3 and react with H2S, resulting in significant improvement of gas sensing performance. The Cu-In2O3 hollow nanofibers can be tailored for practical application to selectively detect H2S at lower concentrations.


Introduction 
Hydrogen sulfide (H 2 S) is a highly toxic reducing gas, which can pose major hazards to the living environment and health of human being. The main sources of H 2 S gas are oil and gas wells, crude oil refining, coal † Yu Zhang, Shuai Han, and Mingyuan Wang contributed equally to this work.
In recent years, metal oxide semiconductors (MOSs), such as ZnO, In 2 O 3 , SnO 2 , TiO 2 , CuO, and WO 3 [8][9][10][11][12][13], have been widely used for detecting various gases (involving NO 2 [14], CO [15], H 2 S [16], Cl 2 [17], ethanol [18], acetone [19], etc.) due to the advantages of high sensitivity, good selectivity, fast response and recovery characteristics, and long-term stability [20]. Among them, In 2 O 3 , as a typical n-type semiconductor material, possesses high sensitivity and selectivity in the detection of H 2 S owing to its wide bandgap of ~3.6 eV, good electrical conductivity, and excellent catalytic activity [1,3,21]. Considering that In 2 O 3 is suitable for gas sensors, there are demands for more accurate response to the gas to be detected and improved selectivity in identifying the target gas. According to the previous reports, there are three main technical means to enhance the gas sensing performance of In 2 O 3 : (i) modulating the material morphology to achieve larger specific surface area [22,23]; (ii) elemental doping to cause changes in the grain size and energy band structure of the material [24,25]; (iii) compounding with other semiconductor materials to build heterogeneous structures [26,27]. Actually, all the three aspects are essentially designed to provide more active sites. For instance, Wei et al. [25] synthesized La-doped In 2 O 3 (La-In 2 O 3 ) hollow microspheres using a hydrothermal process, and obtained the highest response of 17.8 to 10 ppm H 2 S, which is approximately 4.8 times higher than that of pristine In 2 O 3 at 200 ℃. Li et al. [27] fabricated a bamboo-like CuO/In 2 O 3 heterostructure via a solvothermal method, and achieved the response of 229.3 to 5 ppm H 2 S, which can be mainly attributed to the severe sulfation of CuO by H 2 S, as a result of that the resistance value cannot be restored naturally. At present, the related investigations on In 2 O 3 -based H 2 S gas sensors are still not extensive, and some H 2 S sensors present the disadvantages of low sensitivity, difficulty in recovering the resistance, and poor stability. Especially for the H 2 S sensors with high response value, the complete recovery mostly requires a thermal pulse assistance or high temperature environment exceeding 500 ℃ [27,28]. Hence, further explorations are needed to solve these problems.
The options of material morphology and additives are particularly critical for the enhancement of gas sensing performance. The hollow nanofibers or nanotubes are gradually attracting the attention of researchers, since these hollow structures with high aspect ratio show excellent promise for gas sensing application. For instance, Zeng et al. [23] synthesized pearl-necklaceshaped In 2 O 3 nanotubes by electrospinning and calcination, indicating a higher response of 38.3 to 100 ppm formaldehyde, and the shorter response and recovery time (6 and 16 s, respectively) compared with the corresponding non-hollow In 2 O 3 nanofibers. Noted that the p-type CuO-based H 2 S sensors can show the enhanced sensing performance by establishing p-n heterojunctions and sulfation reaction. For instance, Park et al. [29] synthesized highly porous SnO 2 -CuO hollow nanofibers using a combination of electrospinning and annealing, showing an extremely high response of 1395 to 5 ppm H 2 S at 200 ℃ and short response time of about 5.27 s. In short, hollow nanofibers are of great significance for the improvement of gas sensing performance, and especially few investigations on H 2 S sensor based on In 2 O 3 hollow nanofibers were explored.
In this study, pristine In 2 O 3 and Cu-doped In 2 O 3 (Cu-In 2 O 3 ) hollow nanofibers were prepared by electrospinning and calcination. The Cu doping concentration and the operating temperature of the sensor were optimized to improve its gas sensing performance to H 2 S. The as-fabricated pristine In 2 O 3 and Cu-In 2 O 3 hollow nanofibers were eventually used to detect H 2 S gas at concentrations of 1-100 ppm at 150-300 ℃. Meanwhile, the selectivity of 6%Cu-In 2 O 3 to H 2 S was investigated experimentally. Finally, the mechanism of gas sensitivity enhancement of Cu-In 2 O 3 hollow nanofibers was discussed by incorporating the first-principles calculations. The Cu-In 2 O 3 hollow nanofiber sensor can achieve high response to H 2 S, and its excellent sensing performance will increase potential as an efficient candidate for gas sensors. Likewise, our work overcomes the problem that it is difficult for H 2 S gas sensors with high response to recover naturally.

2 Preparation of pristine In 2 O 3 and Cu-In 2 O 3 hollow nanofibers
The pristine In 2 O 3 hollow nanofibers were prepared by a combination of electrospinning and calcination processes. In(NO 3 ) 3 4.5H 2 O (0.4 g) was firstly dissolved in a solution composed of ethanol (5.5 mL) and DMF (4.5 mL) by stirring for 30 min to obtain a homogeneous electrospun solution, and then PVP (0.8 g) was added into the above solution and stirred vigorously at room temperature for 12 h. Subsequently, the as-prepared precursor solution was moved into a 10 mL syringe using a stainless steel needle of Ø0.4 mm. The voltage was maintained at 15 kV during the spinning process, the distance between the needle tip and the collector was ~15 cm, and the injection rate of the syringe was 0.6 mL/h. After electrospinning, the nanofibers on the silicone oil paper were collected and then dried at 60 ℃ for 12 h. Finally, the as-obtained nanofibers were calcined at 500 ℃ for 2 h in air.  Figure 1 shows the schematic of preparation and formation of Cu-In 2 O 3 hollow nanofibers.

3 Material characterization
The crystal structures and phase composition of the as-fabricated nanofibers were characterized by X-ray diffraction (XRD; Bruker D8 Advance, Cu Kα, λ = 1.5418 Å) in the 2θ range of 15°-70°. The structures of pristine In 2 O 3 and x%Cu-In 2 O 3 were refined based on the In 2 O 3 (ICSD Code #14387) model using the Rietveld type method supported by the FullProf software, as a result of that the corresponding lattice constants can be obtained. The morphologies and microstructures of the as-fabricated materials were observed and analyzed by scanning electron microscopy (SEM; FEI, Nova Nano 450) and high-resolution transmission electron microscopy (HRTEM; JEOL JEM 2100F, 200 kV) coupled with selected area electron diffraction (SAED) and energydispersive spectroscopy (EDS). The N 2 adsorptiondesorption isotherms were acquired using an adsorption instrument (Micrometrics, ASAP 2460), and the specific surface area was calculated by the Brunauer-Emmett-Teller (BET) method. X-ray photoelectron spectroscopy (XPS; Thermo Scientific K-Alpha) measurements were performed for surface element analysis.

4 Gas-sensor fabrication and performance test
The as-fabricated pristine In 2 O 3 and x%Cu-In 2 O 3 nanofibers (~2 mg) were uniformly dispersed in ethanol (50 μL) by ultrasonic vibration for 30 min, and then a pipette (10 μL) was used to apply the resulting pastes on the surfaces of alumina substrates (i.e., Pt-coated alumina electrodes). Here, the Pt-coated alumina electrodes were purchased from Wuhan Huachuang Ruike Technology Co., Ltd., China. Afterwards, the nanofiber-coated substrates were dried in an oven for 60 min to remove ethanol, and aged in a furnace at 250 ℃ for 24 h to improve the stability of the sensitive materials. The as-obtained substrates were put into the dynamic four-channel gas sensor test system (SD101) to test the performance index of the gas sensors towards the target gases. The gas sensor test system consists of five components: gases source, test electrodes, a four-channel co-mingled gas flow control module, an aluminum alloy test chamber, and a data acquisition system, as shown in Fig. S1 in the Electronic Supplementary Material (ESM). Here, N 2 serves as the base gas for 100 ppm target gases (H 2 S, NO 2 , NH 3 , and SO 2 ) in the measurement. According to the formula ρ = m/v, the stacking density of pristine In 2 O 3 and Cu-In 2 O 3 hollow nanofibers on the electrode is estimated to be ~0.64 g/cm 3 , where  is the stacking density, m is the mass of nanofiber coating on the electrode, and v is the volume of the coating. Typically, for an n-type semiconductor the response (R) of a sensor is specified as R a /R g , and the response/ recovery time ( res / rec ) is defined as the one required for 90% change in sensor resistance in the target gas and air, respectively.

5 Computational method
In this work, all the calculations were performed based on the density functional theory (DFT) as implemented in the VASP code [30]. The Perdew-Burke-Ernzerhof (PBE) of generalized gradient approximation (GGA) was chosen as the exchange-correction functional [31]. The kinetic energy cutoff was selected to be 450 eV and a 3 × 3 × 1 k-points mesh was sampled. A vacuum space of ≥ 30 Å was set in the unit cell to minimize the interactions of the system and its replicas derived from the periodic boundary conditions [32]. The van der Waals interactions were amended by analyzing the interactions of pristine In 2 O 3 or 6%Cu-In 2 O 3 nanofibers with gas molecules using vdW-D3 [33]. All the atomic structures were completely relaxed until the force on each atom was less than 0.01 eV/Å. The energy convergence criteria were set to be 10 -4 eV. After structure optimization, the unit cell parameters of pristine In 2 O 3 were a = 14.60 Å, b = 14.60 Å, c = 30 Å, α = 90.00°, β = 90.0099°, γ = 120.00°. To be basically consistent with the experiment and ensure the accuracy and validity of calculation, the (111) where E gas@slab , E slab , and E gas are the energies of the total system (gas adsorbed on the slab, pristine In 2 O 3 or 6%Cu-In 2 O 3 ), the slab of pristine In 2 O 3 or 6%Cu-In 2 O 3 , and the adsorbed gas molecules, respectively. , and no peaks derived from Cu-related compounds are observed as the amount of Cu doping is below 6%. This is due to the fact that most of the Cu 2+ ions enter the In 2 O 3 lattice to cause lattice distortion, without formation of new products. On the other hand, it is difficult to detect the low concentration of CuO by the X-ray diffraction. However, a visible characteristic peak (111) of CuO except the (002) appears while introducing 8% Cu 2+ . Moreover, the main diffraction peak indexed to the (222) crystal plane of the Cu-In 2 O 3 hollow nanofibers shifts slightly toward the larger diffraction angle compared with the XRD pattern of pristine In 2 O 3 hollow nanofibers ( Fig. 2(b)). This is because Cu 2+ (0.72 Å) with a smaller ion radius enters into the In 2 O 3 lattice instead of In 3+ (0.81 Å) ions with a larger ion radius, which can cause the lattice spacing to shrink, resulting in the lattice contraction. The Rietveld refinement profile for 6%Cu-In 2 O 3 based on the cubic crystal structure (space group, la 3 ) reveals that the final agreement factors converge to R p 10.5%, R wp 11.8%, and R exp 4.55% (Fig. 2(c)). The lattice constant decreases monotonously with the Cu doping concentration increasing ( Fig. 2(d)), and the similar results were reported [35,36]. The above results indicate that a large portion of Cu 2+ ions is doped into the In 2 O 3 lattice and a small portion is applied to generate CuO. The SEM images of pristine In 2 O 3 and x%Cu-In 2 O 3 hollow nanofibers indicate the hollow structure, as shown in Fig. 3. The surface rough of In 2 O 3 hollow nanofibers is gradually decreased with the Cu doping concentration increasing from 0 to 8%, and the 8%Cu-In 2 O 3 hollow nanofibers are broken into the shorter ones with visible fracture (Fig. 3(f)). Meanwhile, the diameter and wall thickness of In 2 O 3 hollow nanofibers decrease in varying degrees with the Cu doping concentration increasing. In particular, the diameters and average wall thickness of pristine In 2 O 3 hollow nanofibers are of ~200-400 nm and 20 nm, respectively, and the diameter of 6%Cu-In 2 O 3 hollow nanofibers is ~150-180 nm. Hence, the doping of Cu can reduce the diameter of In 2 O 3 nanofibers. Actually, the finer hollow nanofibers can provide more adsorption sites for gas, which is more conducive to the improvement of gas sensitivity. The formation of the hollow structure is due to the Kirkendall effect that occurs during the heat treatment, where atoms are exchanged with vacancies  after the removal of PVP, with In and Cu atoms forming In 2 O 3 and CuO, and diffusing rapidly toward the surface [29,37], as shown in Fig. 1. On the other hand, the formation of the hollow structure is also related to the rate of temperature increase. Actually, the high rate of temperature increase can cause the high decomposition rate of PVP. The rate of gas release is greater than that of gas diffusion through the fiber surface as the decomposition rate of PVP is higher, so that the increase of internal pressure of the fiber is higher than that of the external pressure, resulting in the formation of hollow nanofibers [38].

1 Structure and morphology
The 6%Cu-In 2 O 3 hollow nanofibers were further characterized by using HRTEM, SAED, and EDS mapping, as shown in Fig. 4. The HRTEM image ( Fig.  4(b)) reveals that the lattice fringe spacings of 0.29 and 0.41 nm correspond to the (222) and (211) crystal planes of In 2 O 3 , respectively. Based on the previous reports [27,36], the crystal plane corresponding to a lattice fringe spacing of 0.25 nm is the CuO (002). The SADE pattern (Fig. 4(c)) corresponds to the (211) Figure 5 shows the nitrogen adsorption-desorption isotherms and BJH pore size distributions of pristine In 2 O 3 and 6%Cu-In 2 O 3 hollow nanofibers. The calculated specific surface area of 6%Cu-In 2 O 3 nanofibers is 71.6 m 2 /g, which is higher than that of pristine In 2 O 3 nanofibers (60.7 m 2 /g). Note that this specific surface area value is much larger than those (< 40 m 2 /g) of other In 2 O 3 -based hollow nanofibers or nanotubes [39][40][41]. The pore sizes (calculated by BJH equation) on the walls of pristine In 2 O 3 and 6%Cu-In 2 O 3 nanofibers are distributed in ~2-55 and 2-70 nm, respectively (insets of Fig. 5), indicating that the Cu doping can cause the slight enlargement of pore on the wall. To examine the surface elemental compositions and chemical states of the pristine In 2 O 3 and 6%Cu-In 2 O 3 hollow nanofibers, XPS analysis was performed, as shown in Fig. 6. The two survey spectra show that the pristine In 2 O 3 nanofibers are only composed of In, O, and C elements, while there is also Cu element in 6%Cu-In 2 O 3 nanofibers, which further proves the presence of Cu element (Fig. 6(a)). In the spectra of In 3d (Fig. 6(b)), the peak positions of In 3d 5/2 and In 3d 3/2 move from 444.38 to 444.48 eV and from 451.98 to 452.08 eV after the Cu doping, respectively, due to the electron transfer from In 2 O 3 to CuO [26]. For the O 1s spectra (Fig. 6(c)), three fitting peaks can be obtained by Gauss fitting, and these binding energies are located at 530.09, 531.21, and 532. 49   indicating that the number of oxygen vacancies in the 6%Cu-In 2 O 3 nanofibers is higher than that in the pristine In 2 O 3 nanofibers. The increased oxygen vacancies can provide much more adsorption sites, thus enhancing the sensitivity of the sensor [41,42]. The binding energies of Cu 2p peak (Fig. 6(d)) are around 933.35 and 952.96 eV, corresponding to Cu 2p 3/2 and Cu 2p 1/2 , respectively, with the binding energy difference of ~19.61 eV. Furthermore, there are strong satellite peaks of Cu 2+ , demonstrating that the Cu element exists in the form of Cu 2+ [27].

2 Gas sensing properties
As we know, the operating temperature is one of the crucial indexes to evaluate the performance for MOSbased gas sensors, and it can highly affect the adsorption and desorption of gases and surface interactions. To determine the optimum operating temperature, the pristine In 2 O 3 and Cu-In 2 O 3 hollow nanofibers were tested for response to 100 ppm H 2 S at 150-300 ℃, as shown in Fig. 7(a). Obviously, the optimum operating temperature is 250 ℃, and the responses of Cu-In 2 O 3 nanofiber based sensors to H 2 S are far higher than that of pristine In 2 O 3 nanofibers within the test temperature range. In particular, the response of 6%Cu-In 2 O 3 nanofiber based sensor is the highest (4201.5) while employing the optimum operating temperature, which is over 100 times higher than that of pristine In 2 O 3 (29.3). The responses of all the sensors to H 2 S gradually increase with the temperature increasing from 150 to 250 ℃, and then decrease with the temperature increasing from 250 to 300 ℃. This is because the gas adsorption capacity becomes weaker at higher temperatures, allowing for easy desorption before surface reactions occur, as a result of that the response decreases in varying degrees with the further increase of temperature [43].
The responses of pristine In 2 O 3 and 6%Cu-In 2 O 3 nanofiber based sensors to H 2 S at different concentrations (1-100 ppm) were also investigated at 250 ℃, and the dynamic response/recovery curves are shown in Figs. 7(b) and 7(c). As expected, the response increases with the gas concentration increasing.  (Fig. S2 in the ESM). The resistance of 6%Cu-In 2 O 3 is significantly higher than that of the pristine In 2 O 3 , which is conducive to the increase of response. It is difficult for the resistance to recover after the pristine In 2 O 3 interacts with H 2 S, leading to the decreasing resistance with the increase of gas concentration. Although the resistance of 6%Cu-In 2 O 3 nanofiber based sensor can be restored, it takes a long time, namely that the recovery time is over 400 s at 100 ppm. Figure 7(d) presents the response versus gas concentration for the pristine In 2 O 3 and 6%Cu-In 2 O 3 hollow nanofiber based sensors. For the pristine In 2 O 3 based sensors, the response shows a linear relationship with the concentration of H 2 S. However, for 6%Cu-In 2 O 3 based sensors, the response is increased dramatically after 20 ppm, which is mainly related to the reversible chemical reaction of CuO with H 2 S [44], the role of oxygen vacancy, and the reaction of surface oxygen ions with H 2 S, as discussed below. Figure 7(e) shows their variations of response time with the H 2 S concentration, where the response time is calculated from the two dynamic curves (Figs. 7(b) and 7(c)). As expected, the response time decreases with the gas concentration increasing, and the response time of the 6%Cu-In 2 O 3 based sensor is significantly less than that of the pristine In 2 O 3 at each concentration of H 2 S. The response time of the 6%Cu-In 2 O 3 is 24 and 18 s while employing 50 and 100 ppm H 2 S, which is 20 and 12 s less than that of the pristine In 2 O 3 based sensor, respectively (Fig. S2 in the ESM). The response and recovery time of 6%Cu-In 2 O 3 based sensor for 1 ppm H 2 S is ~125 and 175 s, respectively. Moreover, the response time is shortened rapidly as the gas concentration exceeds 5 ppm.
Good selectivity is one of the basic requirements for a gas sensor to avoid the interference of other gases during the sensing process. To investigate their selectivity, the responses of pristine In 2 O 3 and 6%Cu-In 2 O 3 based sensors to 100 ppm H 2 S and other gases, involving NO 2 , NH 3 , SO 2 , ethanol, and acetone, were recorded at 250 ℃. It reveals that the response (4201.5) of 6%Cu-In 2 O 3 based sensor to H 2 S is remarkably superior to other gases, exhibiting excellent selectivity (Fig. 7(f)).
We also checked the current-voltage (I-V) characteristics of 6%Cu-In 2 O 3 based sensor at various temperatures in air. As shown in Fig. S3 in the ESM, the I-V curves of 6%Cu-In 2 O 3 are linear, indicating the formation of ohmic contact between the 6%Cu-In 2 O 3 and Pt electrode [44]. Moreover, the reciprocal of slope (i.e., resistance) decreases with the temperature increasing, which is in good agreement with the characteristic of n-type semiconductor.
Humidity is also an essential aspect to be considered for gas sensors in practical applications. The effect of humidity on the sensing performance of 6%Cu-In 2 O 3 based sensor was investigated, as shown in Figs. 8(a) and 8(b). The response to 20 ppm H 2 S shows a decreasing trend as the relative humidity is increased from 0 to 65%. This is due to the reduced interaction between the sensing material and H 2 S gas, since more water molecules are adsorbed on the nanofiber surface under higher humidity, weakening the adsorption of H 2 S on the sensing material [45,46]. In particular, the response still remains 29.3 (~60% of the initial value) as the relative humidity arrives at 65%, indicating that the 6%Cu-In 2 O 3 based sensor can be used to detect H 2 S under higher humidity.
Whether in scientific research or industrial production, the reproducibility and stability at the optimum operating temperature of sensors are essential for accurate monitoring of the target gas. Figure 8(c) illustrates the reproducibility of the 6%Cu-In 2 O 3 based sensor for consecutive exposure and removal of 50 ppm H 2 S (5 cycles), indicating that the sensor performs well in 5cycle test with response of 350.7 ± 38.6 ( Fig. 8(d)). The sensor also has a good stability in sensing performance due to no obvious fluctuation in the variations of resistance and response/recovery rate during sensing. Moreover, the response of 6%Cu-In 2 O 3 based sensor to 50 ppm H 2 S for continuous 33 days presents a non-significant decrease (Fig. 8(e)), showing good long-term stability. Table 1 lists the responses and response time of In 2 O 3 -based sensors to 50 and 100 ppm H 2 S at ≥ 150 ℃ for sensing performance comparison. As shown in Table 1, the response of In 2 O 3 -based sensors to 50 ppm H 2 S at 200-300 ℃ varies from 1.08 to 350.7, with the response time of ~2-60 s, indicating that the asfabricated 6%Cu-In 2 O 3 based sensor has the highest response to 50 ppm H 2 S. Similarly, the response of 6%Cu-In 2 O 3 based sensor to 100 ppm H 2 S is far higher than those of other In 2 O 3 -based sensors except the comparable value of Mg-In 2 O 3 nanotubes. In short, the as-fabricated 6%Cu-In 2 O 3 based sensor has the highest response to H 2 S under the comparable conditions compared with the reported In 2 O 3 -based sensors. Thus, the Cu-In 2 O 3 hollow nanofibers fabricated by a combination of electrospinning and subsequent calcination have a promising application prospect in H 2 S gas sensing.

3 Sensing mechanism
There are many factors affecting the gas sensing behavior of 6%Cu-In 2 O 3 based sensor, and the reasons for the improved sensing performance are mainly reflected in the following aspects.
(1) The increase of oxygen vacancy concentration derived from the Cu doping. The substitution of Cu 2+ ions for In 3+ ions can generate additional oxygen vacancies, and meanwhile defects involving oxygen vacancies can dominate the physical and chemical properties of oxide materials and serve as the preferred adsorption www.springer.com/journal/40145 sites for gas molecules [24,37]. So, the as-prepared Cu-In 2 O 3 has great advantages in gas adsorption. Moreover, the doped hollow nanofibers have larger specific surface area, which can significantly promote surface interactions by providing more active adsorption sites, resulting in the enhancement of gas-sensitive response.
(2) The redox reactions occurred on the material surface.
(i) The reaction of adsorbed oxygen with H 2 S. When the n-type In 2 O 3 is exposed to air, O 2 in the air can capture electrons in the conduction band of In 2 O 3 with certain energy, forming electron depletion layer on the surface. Under the operating temperature (250 ℃), there are mainly oxygen ions in the forms of O − and O 2 − adsorbed on the 6%Cu-In 2 O 3 nanofibers [9,56], so that the H 2 S is oxidized to H 2 O and SO 2 by the oxygen ions when exposed to H 2 S ( Fig. 9(a)), and meanwhile electrons are released. The sensing reactions are shown in Eqs. (2) and (3): The electrons are released into the conduction band of In 2 O 3 in this process, so the resistance tends to decrease.
(ii) The reaction of CuO with H 2 S. According to previous reports [29,57], CuO can be converted to Cu 2 S when the 6%Cu-In 2 O 3 based sensor is exposed to H 2 S at a temperature of over 200 ℃. The resistance of the material can decrease rapidly because Cu 2 S has high electrical conductivity, and the chemical reaction is shown in Eq. (4): Once air is passed through to remove H 2 S, Cu 2 S will be oxidized slowly by the atmospheric oxygen and the resistance will begin to recover, the following reaction (Eq. (5)) will occur.
However, the generation of too much Cu 2 S can make the resistance difficult to recover to the original state in air, as a result of that the response tends to decrease when the Cu doping concentration exceeds 6%.
(3) Formation of p-n heterojunction. The electrons in In 2 O 3 and holes in CuO flow in opposite directions due to the inhomogeneous carrier concentration as the p-type CuO is in contact with n-type In 2 O 3 [58,59]. During this process, the energy band bends until the Fermi energy level (E f ) of the two materials reaches equilibrium, forming p-n heterojunctions with the electron depletion layer at the CuO/In 2 O 3 interface ( Fig. 9(b)). The p-n heterojunction can exhibit high resistance state, and offer a greater range of resistance drop for 6%Cu-In 2 O 3 based sensor in H 2 S gas, thus enhancing the gas sensing performance. The first-principles calculations based on VASP were conducted with DFT to explain the gas adsorption mechanism of 6%Cu-In 2 O 3 (111) from the atomic scale. Although CuO can react with H 2 S, here we ignore the role of CuO in the chemical reaction and only investigate the doping effect of Cu. We calculated the adsorption energies of the six gases with different atoms adsorbed on the In and Cu sites on the surface of the constructed model. Figures 10(a)-10(d) show the structural models with the adsorption energy for H 2 S gas, and the other gas adsorption structural models are shown in Fig. S4 in the ESM. The calculated adsorption energies and charge transfer (q) of the six gases (H 2 S, NO 2 , SO 2 , NH 3 , ethanol, and acetone) on the surface of pristine In 2 O 3 (111) and 6%Cu-In 2 O 3 (111) nanofibers are listed in Fig. 10(e) and Table S1 in the ESM. Compared with the pristine In 2 O 3 nanofibers, the adsorption energy of the four gas molecules (H 2 S, NO 2 , NH 3 , and acetone) on the 6%Cu-In 2 O 3 (111) increase in varying degrees except the two comparable values of SO 2 and ethanol, and especially that of H 2 S increases sharply from 2.22 to 3.88 eV, indicating that the gas adsorption capacity of H 2 S gas is greatly enhanced by the Cu doping, which can contribute to the high response and selectivity. Moreover, from the perspective of charge transfer, H 2 S loses the maximum electron of 0.73e as donor to the 6%Cu-In 2 O 3 when the target gas is adsorbed on the 6%Cu-In 2 O 3 (111) ( Table S1 in the ESM), showing the highest response and good selectivity of 6%Cu-In 2 O 3 to H 2 S (Fig. 7(f)). However, the H 2 S molecules are dissociated during the adsorption of H 2 S on 6%Cu-In 2 O 3 (Figs. 10(c) and 10(d)). This is because H 2 S is strongly reductive and H 2 S undergoes deprotonation reaction through the action of S atoms and Cu 2+ to form HS and H [60], leading to the increase in the adsorption energy as well, which contributes to the high selectivity of 6%Cu-In 2 O 3 to H 2 S.
Moreover, the reason why the Cu doping can promote the sensing performance of In 2 O 3 is explained. Actually, the gas adsorption can alter the work function (W) of materials. So, the work function induced by gas adsorption has become a promising way to describe the change of current densities to evaluate the sensor properties [61], and the current density in vacuum is described by [62]: where A, T, and W represent the Richardson constant (A/m 2 ), temperature (K), and work function (eV), respectively. There is a negative correlation between www.springer.com/journal/40145 the current density and work function. The calculated work functions are shown in Fig. 10 (111)). Furthermore, the calculated DOSs indicate that the In 2 O 3 (111) shows the semiconductor properties, with the bandgap of 0.76 eV ( Fig. 11(a)). After the Cu doping, the bandgap of Cu-In 2 O 3 (111) decreases to 0.11 eV (Fig. 11(b)), and much impurity levels are introduced around Fermi level (0 eV  Fig. 11(c)). Specially, when H 2 S is adsorbed on Cu-In 2 O 3 (111), a sharp peak appears at Fermi level due to the orbitals' hybridization of d orbital of Cu and p orbital of S. Therefore, the Cu doping can promote the sensing performance of In 2 O 3 to H 2 S (Fig. 11(d)).

Conclusions
In this study, the pristine and Cu-In 2 O 3 hollow nanofibers were prepared by a two-step method involving electrostatic spinning and annealing. A large portion of Cu 2+ ions are doped into the In 2 O 3 lattice and a small portion are applied to generate CuO while employing 6% Cu doping. The 6% Cu doping can make the surface of nanofibers smooth, homogenize and refine the diameter of nanofibers, increase the specific surface area to a certain extent, and especially enhance the formation of oxygen vacancies. Gas sensing experiments and DFT calculations (involving the adsorption energy, work function, and DOS) demonstrate the excellent selectivity and sensing performance of 6% Cu-doped In 2 O 3 to H 2 S. The response of 6%Cu-In 2 O 3 sensor can reach 350.7 and 4201.5 to 50 and 100 ppm H 2 S at the optimum operation temperature of 250 ℃, and the response time is only 24 and 19 s, respectively. In particular, the response of 6%Cu-In 2 O 3 sensor to 1 ppm H 2 S reaches 1.5. Moreover, its response to 20 ppm H 2 S gradually decreases from ~50 to 30 with the relative humidity increasing from 0 to 65%. The 6%Cu-In 2 O 3 hollow nanofiber based sensor presents excellent sensing performance to H 2 S gas owing to the following aspects: (1) the increase of oxygen vacancy concentration derived from the Cu doping; (2) the redox reactions of adsorbed oxygen and CuO with H 2 S on the material surface; and (3) the formation of p-n heterojunction. This work can provide a promising approach to achieve high-performance In 2 O 3 -based gas sensor for H 2 S detection.