NO2 sensing characteristics by α-Fe2O3 nanorod arrays with atomic layer deposited amorphous Al2O3 overlayer

We have grown α-Fe2O3 nanorods by solution processing followed by the deposition of Al2O3 overlayer using atomic layer deposition. Al2O3 layer was deposited for two different thicknesses 4 nm and 8 nm and a post-deposition annealing at 550 °C for 2 h in air atmosphere was performed. Crystallinity analysis through x-ray diffraction (XRD) reveals that the α-Fe2O3 nanorods crystallized into rhombohedral structure, whereas the outer Al2O3 layers remained largely amorphous. Interestingly, the interface showed signs of AlFexOy formation as observed through high-resolution transmission electron microscopy images. Gas sensing characteristics were studied using NO2 with 10, 50, and 100 ppm concentrations at operating temperatures of 30 °C, 100 °C, 150 °C and 190 °C. The room temperature sensitivity values obtained in response to 10 ppm NO2 were 31%, which surpassed the previously reported values. A higher concentration of surface adsorbed oxygen on the Al2O3 overlayer, as revealed by the x-ray photoelectron spectroscopy (XPS) analysis, led to enhanced NO2 sensing at room temperature. A lower activation energy (0.29 eV) of barrier to charge transport for Al2O3 coated α-Fe2O3 nanorods compared to that of bare nanorods (0.45 eV), as calculated from the temperature dependent I-V measurements, supported observation of higher sensitivity at room temperature.


Introduction
Gas sensing applications are crucial in modern day environmental research where transition metal oxides play an important role due to their variable valence state, good catalytic properties and higher thermal and temporal stability.Although materials such as SnO 2 , ZnO, and TiO 2 have been extensively studied for gas sensing, a higher band gap (> 3 eV) causes insulating nature and low current leading to a lower signal to noise ratio over a moderate band-gap transition metal oxide (2-3 eV).α-Fe 2 O 3 (hematite), with its low bandgap (2.1 eV), variable oxidation state, good chemical and thermal stability, and high oxygen ion mobility at the surface is a potential candidate for oxidative reactions and is considered as a suitable material for gas sensing applications [1][2][3][4][5].
Based on the surface morphology and structure of α-Fe 2 O 3 the selectivity of gases change as the gas adsorption varies with different morphologies and it is also evident from the studies shown in the literature.For example, thin films or nanoparticles of α-Fe 2 O 3 , have been used for NO 2 detection [6][7][8] and nanowire type structures have shown good sensitivity toward CO sensing [9].A flower-like α-Fe 2 O 3 structure is proven to be a better sensor for reducing gases like ethanol [10].Gas selectivity also depends on nanocrystal facets and surface functional groups, which govern the adsorption of specific molecules at particular conditions.The active surface area of materials plays a crucial role in the gas sensing characteristics since the adsorption of gases is a surfacedriven phenomenon.It is therefore natural that high surface area nanostructures, in particular one-dimensional crystalline forms like nanorods/nanowires, are preferred in gas sensing due to their directional charge flow, offering higher surface-to-volume ratios and reduced charge scattering, leading to superior signal-to-noise ratios.In one of the earliest works, Wu et al [11] fabricated α-Fe 2 O 3 nanorods by chemical techniques and studied the formaldehyde sensing properties.Dai et al reported the growth of α-Fe 2 O 3 nanorods on multiwalled carbon nanotubes and studied the acetone sensing performance [12].In another interesting work, Donerelli et al [13] reported anomalous hydrogen sensing by α-Fe 2 O 3 nanorods that was attributed to unusual band bending at the gas-solid interface.Touba and Kimiagar fabricated ZnO decorated α-Fe 2 O 3 nanorods and found superior H 2 S sensing capabilities.They attributed this to enhanced depletion layer formation at the surface because of the presence of ZnO nanoparticles [14].It is therefore evident that the α-Fe 2 O 3 nanorod morphology has been explored for several gas sensing applications.
NO 2 is known to be extremely harmful to our health and the natural ambiance.It is a strongly oxidizing gas, reasonably corrosive in nature and is generally emitted from automobile exhaust, and various industrial processes and also contributes to the formation of another hazardous gas ozone (O 3 ).It also affects the human nervous system which leads to loss of consciousness in a relatively prolonged exposure.It is therefore important to have sensors for the detection of low concentrations of NO 2 in the atmosphere and at room temperature.Although several morphologies of α-Fe 2 O 3 such as nanoparticles, thin films and composites with other materials have been explored for the sensing of NO 2 [5,6,15], but the detection was only possible at high temperatures like 200 °C and 300 °C.Introduction of co-catalyst and doping helped in bringing down the temperature to 150 °C and 135 °C respectively [16,17].A mixture of flake and rod like structure of α-Fe 2 O 3 as shown by Sangale et al in 2020 shown NO 2 sensing at room temperature but the sample was prepared in powder form and then deposited on a substrate by doctor blade technique [18].However, vertically oriented α-Fe 2 O 3 nanorods, grown directly from an underlying substrate have not been explored so far for NO 2 sensing although it has the advantage of better adhesion in terms of device formation.
One of the fundamental problems with metal oxide gas sensor is lack of room temperature sensitivity due to very low kinetics of catalytic reactions.Enhancement methods can include doping, core-shell designs, and cocatalyst functionalization.When a catalytic layer is deposited on a surface, it is crucial to obtain an ultra-thin and ultra-conformal coating so that the gas adsorption occurs uniformly and also the underlying active layer remains effective.Although several chemical and vacuum-based processes have been used to deposit such layers over a host/base material, atomic layer deposition (ALD) provides the most conformal coating without damaging the underneath layer properties.Xie et al [19] demonstrated ALD coated ZnO layers over CVD grown graphene and both thickness and temperature dependent NO 2 sensing properties were reported.In another work, Lin et al [20] used an ALD coated AZO (Aluminium doped ZnO) layer over the ZnO nanorods and studied ethanol sensing behavior.There are few reports where an ultrathin Al 2 O 3 overlayer has been used as a surface passivation layer to neutralize the effect of trap states which resulted in enhancing the performance of Fe 2 O 3 as the anode in PEC water splitting and in solar cell [21,22].Best of our knowledge Al 2 O 3 coated α-Fe 2 O 3 vertical aligned nanorods have not been explored for gas sensing.
In this work, we report the gas sensing properties of bare and Al 2 O 3 coated α-Fe 2 O 3 nanorods grown on FTO coated glass substrates.The most common and well-studied metal oxide precursors for ALD coating are Al 2 O 3 , TiO 2 , ZnO x , SnO 2 , etc Among these, Al 2 O 3 possess a similar crystal structure to Fe 2 O 3 as both have rhombohedral lattice systems, it shows high corrosion resistance and has exceptional thermal and chemical stability which is suitable for gas sensing.So, the chemically grown nanorods were coated with Al 2 O 3 using the ALD process and the gas sensing characteristics of the bare and core-shell nanorods have been studied for NO 2 .Further, the charge transport mechanism in the presence of target and inert gases (N 2 ) has been explored using current-voltage (I-V ) measurements.At ultrathin film thickness, Al 2 O 3 coating increases the gas sensitivity of α-Fe 2 O 3 nanorods at room temperature.The novelty of this study lies in highlighting the effectiveness of an overlayer coating of amorphous Al 2 O 3 by ALD in enhancing sensor performance.Our results provide an insight into the gas sensing characteristics of vertically aligned α-Fe 2 O 3 nanorods both in terms of ALD coating assisted sensitivity enhancement and charge transport studies.

Materials and methods
Commercially available fluorine-doped Tin Oxide (FTO) (Ants Ceramics, India) coated glass slides were used as the substrates.The sheet resistance of FTO was around 7 Ω/sq and the dimensions of each piece made were around 20 mm × 15 mm area and the glass thickness were 2.3 mm.These substrates were cleaned by ultrasonication in a diluted soap solution and deionized (DI) water for 2 min each to remove all dust particles.Thereafter, these FTO pieces were ultrasonicated in acetone, ethanol and isopropyl alcohol respectively for 2 min each to remove all organic residues without damaging the FTO layer and finally, these cleaned FTO pieces were dried under N 2 jet for 10 min.
FeOOH nanorods were grown on the FTO-glass substrates by a chemical bath deposition method using Urea (CH 4 N 2 O) (Merck, purity > 99.0%) and Ferric Chloride (FeCl 3 ) (Fisher Scientific, purity 96.0%) precursors mixed in a ratio of 3:2 in DI water.This technique is also mentioned in detail in our previous work [23].The mixture was stirred well for at least 1 h to prepare a homogenous solution.The FTO pieces were kept inside the prepared solution in a beaker, supported on the beaker wall by making an angle with the bottom to avoid any direct precipitation from the solution.Then the beaker was sealed properly and kept inside a hot air oven maintained at 95 °C temperature for 8 h.Once the oven cooled down to room temperature, FeOOH coated light yellow FTO glasses were taken out and rinsed thoroughly with DI water.The prepared samples were dried at room temperature under N 2 flow.FeOOH nanorods were converted into α-Fe 2 O 3 by annealing in air at 550 °C for 2 h.

ALD deposition of Al 2 O 3 layers
Al 2 O 3 films were deposited at temperatures ranging from 100 °C to 250 °C at the intervals of 50 °C using a commercial ALD reactor manufactured by Oxford Instruments, UK (Model: FlexAL).An electronic grade trimethylaluminum (TMA) (5 N purity) (EpiValance, UK) and ultra-pure deionized H 2 O were used as a precursor and co-reactant respectively for Al 2 O 3 deposition from thermal atomic layer deposition (TH ALD).One cycle of ALD deposition gets completed in four steps.These steps include (i): Dose; in this step, a precursor is put in the deposition chamber to react with substrate; (ii): Purge; this step is required to remove unreacted precursors from the chamber after the completion of reaction with the substrate; (iii): Co-reactant dose; here H 2 O vapor is sent into the chamber to complete the reaction; and (iv): Purge; final purge of the chamber is required to remove unreacted co-reactant and byproducts.To optimize the thickness of the film and to determine an ALD deposition window [19,20] for TMA precursors, the growth per cycle (GPC) was monitored and quantified at various temperatures using spectral ellipsometry (SE) (M2000, JA Woollam) and x-ray reflectivity (XRR) (Rigaku SmartLab).The optimized data reveals 200 °C to 250 °C deposition temperature as the ALD parameter for TMA precursors.After optimization of Al 2 O 3 deposition, we selected 200 °C temperature for the deposition of different thicknesses of Al 2 O 3 on α-Fe 2 O 3 nanorods deposited on FTO-coated glass.The GPC of Al 2 O 3 obtained using TH ALD is approximately 1.5 Å-cycle −1 .We have deposited Al 2 O 3 multiple layers from 4 nm to 20 nm on FeOOH nanorods as per the need for varying thicknesses and observed that the optimal film thickness range was between 4 and 8 nm.After coating with Al 2 O 3 , FeOOH nanorods were annealed at 550 °C with a ramp rate of 1 °C min −1 for two hours in the air.Hereafter, the 4 nm of Al 2 O 3 coated and annealed samples will be identified as 'AFO-4', and similarly, 8 nm samples as 'AFO-8'.The same notations are used throughout the article.

Characterization and measurements
The crystallinity and phase formation were investigated by x-ray diffraction (XRD) using Aeris Benchtop x-ray diffractometer, operating at 40 kV and 15 mA anode current using the radiation Cu Kα (λ = 0.154 nm) as source.The surface morphology was analyzed using field emission scanning electron microscope (FESEM) (FEI Inspect F50) operating at 30 KeV.The microstructure and crystalline phase were further confirmed by high resolution transmission electron microscope (HR-TEM) using JEOL Japan, JEM -2100 Plus.To perform the HR-TEM measurements the nanorods were scratched out of the FTO substrate and then dispersed in ethanol by ultrasonication.Few microliters of this suspension were drop-casted on a TEM grid and then put under the microscope.The elemental composition, chemical and electronic state analysis were performed using an x-ray photoelectron spectrometer by ULVAC, Model : PHI5000 Version Probe III.The optical absorption spectra were recorded using a Perkin Elmer LAMBDA 950 UV-VIS-NIR spectrophotometer in diffuse reflectance spectroscopy mode equipped with an integrating sphere accessory.

Gas sensor unit and measurement
Figure 1 shows a schematic depiction of the gas sensing set-up used in this work.All the experiments were performed in a 100 ml sealed glass chamber containing a sample heating stage, a thermocouple and two electrical probes.The electrical probes were connected outside with a source meter (Keithley 2450 SMU) for measuring the change in resistance with exposure to gas.The chamber was equipped with gas injection and ejection ports and a sampling port.The heater could provide the temperature variation of the sample stage from ambient to 200 °C.The amount of injected gas was controlled by two mass flow controllers (MFC), one for the test gas (NO 2 ) and the other for the carrier gas (N 2 ).The concentration of the test gas was measured by taking a ratio of the gases flowing through the MFCs.Gas sensing performance was measured in terms of change in resistance with time at different temperatures and for different concentrations of gas.The sensitivity was calculated using the following equation [6, 24] Where, S is sensitivity factor, R NO2 is the resistance measured in the presence of NO 2 and R N2 is the resistance measured in the presence of N 2 which is the carrier gas.

Crystallinity and phase formation
Figure 2 shows the crystallinity of different samples, namely, α-Fe 2 O 3 AFO-4 and AFO-8.The pattern (i) representing FTO coated glass substrate, (ii) α-Fe 2 O 3 , peaks appearing at positions of 24.13°, 33.15°, 35.59°c orrespond to (012), ( 104), (110) planes respectively.This XRD result obtained from annealed FeOOH which is matching with the rhombohedral α-Fe 2 O 3 (Hematite) phase (ICSD reference no # 01-089-2810).The high intensity peaks are originating from the FTO substrate, which has a strong crystallinity.Patterns (iii) and (iv) correspond to AFO-4 and AFO-8 samples.However, the presence of Al 2 O 3 coating was not detected, which may be attributed to the layer of Al 2 O 3 thinner than the detection limit of XRD.Also, annealing at 550 °C for 2 h may not be sufficient to induce crystallinity in the Al 2 O 3 layers, which crystallizes at 900 °C or higher [25,26].The comparison of XRD patterns of higher coating thickness of 12 nm, 16 nm and 20 nm are also presented in figure S1 in supplementary information.The presence of Al 2 O 3 couldn't get detected by XRD even at 20 nm coating thickness also.To further confirm the crystallinity of the ALD coated Al 2 O 3 layer, high resolution TEM studies were carried out and the results are discussed in a subsequent section.

Morphology
The surface morphology of the samples investigated through FESEM are presented in figure 3. Figure 3(a) shows the as-prepared FeOOH nanorods with square cross section grown directly on the FTO coated glass substrate.The average dimension of nanorods is approximately 650 nm * 50 nm * 50 nm.The cross-section view of FeOOH nanorod arrays grown on the FTO coated glass substrate is shown in supplementary information figure S2(a).Vertical growth from the underlying substrate is observed.Coating at higher thicknesses of 12 nm, 16 nm and 20 nm is also presented in supplementary information in figure S3 for a better understanding of the conformal coating, where the decrease in inter nanorod spacing confirms higher thickness coating.A careful observation of the FESEM images of α-Fe 2 O 3 nanorods (annealed FeOOH nanorods) without and with Al 2 O 3 coating also reveals that, the samples without ALD coating are not able to sustain annealing and the sharp edges of FeOOH rods disappear after the high temperature processing.On the other hand, there is no change in shape of ALD coated nanorods even after high temperature annealing.Thus, we infer that the amorphous Al 2 O 3 also acts as a protective layer for the nanorods.

TEM and EDS
Further detailed study of the structure and crystallinity of the prepared nanorods were done under the highresolution transmission electron microscope (HR TEM).Figures 4(a

X-ray photoelectron spectroscopy
The chemical composition was analyzed using XPS on annealed samples with and without the Al 2 O 3 coating (α-Fe 2 O 3 and AFO-4).Figure 6(a) shows the full survey scan obtained from both types of samples.Comparison of the spectra shows that Fe, O, C elements are present in both survey spectra (i) and (ii); however; Al is only present in spectrum (ii).In figure 6(c), the high resolution Fe core level spectrum of pure α-Fe 2 O 3 shows two main peaks at 710 eV and 724 eV, which are assigned as Fe 2p 3/2 and Fe 2p 1/2 respectively.In addition, there are satellite peaks at 719.3 eV and 732.9 eV respectively.The main peaks for 2p 3/2 and 2p 1/2 are further fitted into two peaks each, one at higher energy for Fe 3+ (octahedral site) state and the other at lower energy for Fe 2+ (tetrahedral site) state [12,28].But there are no Fe 2+ satellite peaks observed at 715 eV, which implies the formation of pure α-Fe 2 O 3 phase [29].The peak positions are almost the same in case of AFO-4 presented in figure 6(d).Yet, considering the integral value of Fe 3+ state and Fe 2+ state, the total peak area of Fe 3+ state becomes half in case of Al 2 O 3 coated α-Fe 2 O 3 sample.This phenomenon occurs because of aluminum diffusion  Unlike the Fe peaks, there is a significant difference between the O1s peaks as obtained from pure α-Fe 2 O 3 and AFO-4 samples presented in figures 6(e) and (f) respectively.The asymmetric peaks obtained from O1s are fitted into three peaks and assigned as lattice oxygen (O L ), oxygen deficiency (O V ) and chemisorbed oxygen (O C ), starting from lower energy to higher energy in both the samples [30,31].In bare Fe 2 O 3 the position of the three peaks are at 529.8 eV, 531.3 eV and 533.3 eV for O L , O V and O C respectively and for AFO-4 these are at 529.9 eV, 531.1 eV and 532.2 eV.The percentage of chemisorbed oxygen is increased from 0.39% for bare Fe 2 O 3 to 52.25% in Al 2 O 3 coated Fe 2 O 3 which is attributed to the amorphous nature of the Al 2 O 3 layer that promotes vacancy induced adsorption of the oxygen ions.The high concentration of chemisorbed oxygen helps in achieving enhanced sensitivity, in particular for thin (4 nm) coating of Al 2 O 3 .

Gas sensing properties
The gas sensing characteristics of the samples were tested using NO 2 with different concentrations and at different measurement temperatures for bare and AFO-4 nanorod arrays.Before testing, each sample was kept under 1000 sccm (standard cubic centimeters per minute) N 2 flow for 2 h to get a stabilized resistance.The concentrations of NO 2 used for the sensing experiments were 10, 50 and 100 ppm (parts per million) obtained through calculated mixing of the test gas with the carrier gas (N 2 ) flown through respective mass flow controllers (MFCs).Here, it is important to notice that the time of gas flow was kept constant (5 min) for all samples and the sensitivity was calculated by measuring the difference of resistance (of the samples) at the initial (before gas flow) and final (end of 5 min flow) times.However, the change in resistance in response to a fixed amount of test gas for all samples was expected to show the sensitivity for ALD coated samples with respect to the bare α-Fe 2 O 3 samples.
Figure 7 shows the values of sensitivity factor in bare and Al 2 O 3 coated samples with a coating thickness of 4 nm (AFO-4) and 8 nm (AFO-8).The sensitivity factor S is calculated according to the equation (1) mentioned in the experimental section.Four different operating temperatures, namely, 30 (room temp.), 100 °C, 150 °C and 190 °C were used and the sensitivity factor S for different samples with exposure to 10, 50 and 100 ppm of NO 2 are represented in the bar diagram shown in figures 7(a), (b) and (c) respectively.The bare Fe 2 O 3 nanorods (black bar) show a negligible sensitivity at room temperature; but the value increases at higher temperature.On the other hand, the AFO-4 sample (red bar) shows high room temperature sensitivity for all tested concentrations of NO 2 , which also increase with increasing temperature.However, the sensitivity factor decreases on increasing the ALD coating thickness to 8 nm (blue bar).We also observed that sensitivity increases almost linearly as NO 2 concentration increases within our experimental range of 10 to 100 ppm.Same trend followed at different temperatures for AFO-4 sample which is presented in figure 7(d).The most interesting features are the room temperature sensitivity obtained from AFO-4 samples showing values of 0.31, 0.56 and 1.50 respectively for 10, 50 and 100 ppm NO 2 .The variations of sample resistance as a function of time on exposure to 100 ppm NO 2 at room temperature and at 190 °C are shown in figures 8(a) and (b) respectively.The room temperature sensitivity data have slightly poor recovery; however, the samples show better recovery at higher temperature.A comparison of the reported sensitivity values for NO 2 gas obtained at different temperatures and concentrations for α-Fe 2 O 3 along with the results obtained in our work are presented in table 1.The results obtained from our work are better than the reported data for bare α-Fe 2 O 3 in terms of sensitivity at high temperature and AFO-4 shows good room temperature sensitivity.However, sensor recovery is a feature that requires further investigation and improvement.The response and recovery times for the AFO-4 sample at RT and 190 °C is presented in figures 8(c) and (d) and the estimated comparison for α-Fe 2 O 3 and AFO-4 samples are presented in supplementary information figure S5(e).We observed that although AFO-4 samples don't recover quickly at room temperature due to its amorphous outer layer but has a trend of decreasing resistance with time without the application of any heat.However, at a higher temperature of 190 °C, almost complete recovery happens.Further discussion on room temperature and high temperature recovery is presented in supplementary information corresponding to figure S5.
Figure 9(a) shows the selectivity performance and a comparison of AFO-4 sample signals for NO 2 gas and other common gases present in a normal atmosphere for example H 2 , O 2 , CO 2 , etc We have deliberately performed the selectivity experiments for these gases at higher ppm levels (than the NO 2 ) to be sure about the selective response and all the sensitivity responses are presented in supplementary information figure S6.This confirms that Fe 2 O 3 nanorods with ALD coated Al 2 O 3 overlayer are selective to NO 2 in the presence of interfering gases such as hydrogen, oxygen and CO 2 .
We have performed humidity response experiments for pure Fe 2 O 3 and AFO-4.Three different humid environments were chosen-10% (under dry N 2 ), 40% (at normal ambiance), and 75% by saturated KCl solution which were measured at the time of experiment.The change in resistance for Fe 2 O 3 in response to a change in humidity is higher than that for AFO-4.This demonstrates humidity resistance behavior of the Al 2 O 3 coated Fe 2 O 3 .These data are presented in figure 9(b).In general, we find an increase in resistance with increase in humidity level, which was also reported in literature for other materials [36,37].
Figure 9(c) gives us an idea about the long-term stability of the sample.This test is performed on AFO-4 samples and the initial resistance is collected for 25 days at RT. Almost a similar resistance is observed for this long period of time.Thus, we can conclude that Al 2 O 3 -coated α-Fe 2 O 3 nanorod samples are highly stable and a sample can be used multiple times at a stretch.In addition to this, corresponding FESEM images were obtained for both samples after gas sensing after more than 10 times.These are presented in the figures 9(d) and (e).In general, there are no significant morphological changes observed in the samples after sensing.3.6.Gas sensing mechanism Gas sensing is a surface phenomenon.The change in electrical resistance of sensors upon exposure to the target gas is governed by the dominant charge carriers in the material and type of the gas.In the case of an n-type semiconductor, the electrical resistance increases for oxidizing (electron withdrawing) gas and decreases for reducing (electron donating) gas.The situation reverses in case of a p-type material.The gas sensing is governed by surface adsorbed oxygen, which further reacts with the target gas molecules to either extract or donate electrons to the sensing surface.Accordingly, the resistance change occurs, which manifests in sensitivity [38,39].
In n-type materials, electrons are the majority carriers in the conduction band and the adsorption of oxygen leads to the formation of ionized species such as O 2 − , O − and O 2− .This results in electron depletion from a depth near the surface of the semiconductor, which is equivalent to the Debye length [39].Typically, the Debye length for most semiconducting materials remains in the range of 2-100 nm.When an oxidizing gas such as NO 2 comes in contact with the material surface, it leads to further removal of electrons, resulting in further increase in resistance.In case of a reducing gas, the interaction with the surface-adsorbed oxygen leads to the formation of oxygen molecules.Thus, the conductivity increases as the electrons released back into the surface and we see a decrease in the resistance.
In the present case, NO 2 interacts with the Fe 2 O 3 surface through the adsorbed oxygen ions.In the first stage, oxygen adsorption occurs through the steps below [40,41] In case of bare α-Fe 2 O 3 nanorods XPS data shown in figures 6(e) and (f), reveal a lower concentration surface adsorbed oxygen, which leads to weaker interaction of the target gas molecules with the sensor surface.According to the oxygen adsorption model, gas sensing takes place through a two-step process with ambient oxygen creating a depletion layer followed by the adsorption of the target gas.The interaction of the target gas with adsorbed oxygen ions leads to further charge transfer and the depletion layer width gets modified resulting in the change in surface resistance.The interaction of NO 2 with bare Fe 2 O 3 nanorods is minimal at room temperature, and hence a negligible sensitivity factor is obtained.This is schematically explained in figure 10, which shows a lesser band bending for bare α-Fe 2 O 3 nanorods.However, at higher temperatures, adsorption/ desorption on the metal oxide surface increases and hence reasonable sensitivity is obtained even for the bare Fe 2 O 3 nanorods.It is also evident from literature that α-Fe 2 O 3 is most sensitive towards NO 2 gas at high temperature in the range 200 °C-350 °C [6,31,42].On the other hand, a 4 nm coating of Al 2 O 3 results in an appreciable sensitivity factor even at room temperature.It is evident from the XPS data that the Al 2 O 3 overlayer leads to a higher concentration of oxygen adsorption.This may originate from the amorphous nature of the Al 2 O 3 layer, which promotes oxygen adsorption through the surface vacancies [20,43,44].An initial depletion layer formed by the oxygen adsorption is further enhanced by the electron withdrawing nature of the NO 2 gas (oxidizing gas) and for a thinner layer of Al 2 O 3 , the requirement of electrons may come from Fe 2 O 3 underlayer.This is schematically shown in figures 10(c) and (d).However, for thicker Al 2 O 3 layer, the access to Fe 2 O 3 underlayer may be inhibited by the insulating nature of the Al 2 O 3 and hence the sensitivity decreases with 8 nm ALD coating.Nevertheless, the sensitivity factor for 4 nm Al 2 O 3 coated Fe 2 O 3 nanorods (AFO-4) remains higher than that of the other samples.Such a value of room temperature NO 2 sensitivity (31% at 10 ppm) is significantly higher than the previously reported value on the Fe 2 O 3 system [18].

Charge conduction and activation energy
The better sensing characteristics of ALD coated sample AFO-4 compared to that of bare α-Fe 2 O 3 can further be explained and understood through the activation energy calculation.There is a change in resistance upon interaction with gas, this is reflected with a change in the current versus applied voltage (I -V ) curves and has a Schottky-type nature which is attributed to the difference in work function of metal-insulator contact.The Where, J is current density, J 0 is a pre factor, f is the Schottky barrier height, T is the temperature and k is the Boltzmann constant.Equation (8) can be rewritten as Where, m is the slope of the straight line and c is a constant.Thus, the slope of the curve which is plotted in a graph of ln (J) versus 1/T gives the information about barrier height [41,45,46].Now, in an ideal condition where only thermionic conduction happens and no gas adsorption/desorption occurs during heating or cooling, we can consider Schottky barrier height (f) = activation energy (E a ) [45,47].
The activation energy values are calculated from the graph in figure 11(c) and equation (9) at 5 different applied voltages starting from 1 V to 5 V.The obtained curves are not completely linear but to calculate the activation energy a linear fitting has been done.The average value of E a is 0.45 eV for bare α-Fe 2 O 3 whereas it drops down to 0.29 eV for the ALD coated AFO-4 sample.A lower value of the activation energy indicates better sensing performance.Thus, the activation energy also explains why we have enhanced sensing performance with Al 2 O 3 coated Fe 2 O 3 nanorods.
The surface states in bare α-Fe 2 O 3 , in the form of oxygen vacancies and other defects, are passivated by the ultrathin Al 2 O 3 coating.Ultrathin ALD coatings are well proven method of passivation of the surface states as reported in literature [21].However, the Al 2 O 3 layer is amorphous in nature and has a higher density of oxygen vacancies as shown in our XPS data (figure 6(f)) and also reported in the literature [43,44].The higher density of oxygen vacancies promotes enhanced oxygen adsorption on the Al 2 O 3 surface.This effect is consistent with

Conclusion
We investigated the NO 2 sensing behavior of bare, 4 nm and 8 nm ALD-coated Al 2 O 3 overlayer on α-Fe 2 O 3 nanorods chemically grown on a conductive glass substrate.The FESEM images show near-vertical alignment of the α-Fe 2 O 3 nanorods and a highly conformal overlayer of Al 2 O 3 , as confirmed by EDS analysis.The crystallinity study by XRD and microstructure analysis by HR-TEM reveals crystalline α-Fe 2 O 3 and an amorphous Al 2 O 3 layer; however, a distinctly crystalline layer corresponding to a d value of 5.9 Å appears at the interface of Fe 2 O 3 and Al 2 O 3 .This indicates the formation of AlFe x O y or a similar phase at the interface possibly by the diffusion of Al and Fe ions during the annealing process.The gas sensing experiments were performed for 10 ppm, 50 ppm and 100 ppm of NO 2 gas at temperatures of 30 °C, 100 °C, 150 °C and 190 °C.A higher concentration of the adsorbed oxygen on AFO-4 nanorods as evident from XPS leads to a better sensitivity of the NO 2 gas compared to that of bare α-Fe 2 O 3 .Most interestingly, room temperature NO 2 sensing has been achieved with a sensitivity factor significantly higher than the previously reported values.This is attributed to the wider depletion layer induced by the high chemisorbed oxygen that completely takes over the 4 nm Al 2 O 3 film.However, for 8 nm thickness of the Al 2 O 3 coating, the insulating nature dominates and the gas sensitivity decreases.The activation energy calculated from temperature dependent I-V characteristics in the presence of NO 2 shows 0.45 eV for bare and 0.29 eV for 4 nm Al 2 O 3 coated α-Fe 2 O 3 nanorods, which indicates an easier charge transfer process between the sensing surface and the gas molecules.This ultimately leads to better sensing performance.

Figure 3 (
b) represents the α-Fe 2 O 3 nanorods which were obtained by annealing the FeOOH nanorods at 550 °C for 2 h.The surface morphology of FeOOH nanorods coated with a 4 nm amorphous Al 2 O 3 layer by ALD shown in figure 3(c).A thin and highly conformal layer wrap around the FeOOH nanorods with high uniformity as observed in figure 3(c).The uniformity and conformality of the ALD coating are maintained even after annealing.Figure 3(d) presents here the 4 nm ALD coated nanorod samples after annealing at 550 °C for 2 h (AFO-4).With the increasing thickness from 4 to 8 nm, there is not much change visible in FESEM top view image as shown in figure S2(c).But the cross-section image shows a little more dense but still discrete nanorods in AFO-8 samples which are presented in figure S2(b) and (d).

Figure 1 .
Figure 1.Schematic of the gas sensing chamber.
) and (d) show the TEM images from a single α-Fe 2 O 3 nanorod before and after coating the Al 2 O 3 layer.The compositional analysis by EDS (Energy Dispersive x-ray Spectroscopy) was done over an array of nanorods.The results give an idea about the elements present in the nanorods.The graph and table presented in supplementary information figures S4(a) and (b) confirms the presence of Al, Fe and O in figure S4(b), whereas only Fe and O are present in S4(a).The d-spacing as estimated from the HR TEM (lattice fringes) image is 0.369 nm which corresponds to the (012) plane of α-Fe 2 O 3 , as confirmed with ICSD database no.#01-086-2810.Further we have obtained the selected area electron diffraction (SAED) pattern, which is presented in figure 4(f).The bright spot pattern represents the single crystalline nature of the α-Fe 2 O 3 nanorods.The probable planes indexed based on position and angles are (104), (018), and (208), which exactly match with the mentioned XRD database confirming the formation of rhombohedral α-Fe 2 O 3 .The high resolution image for Al 2 O 3 coated α-Fe 2 O 3 nanorods (figures 4(d) and (e)) clearly shows the presence of two distinct layers around each nanorod.A careful observation of figure 4(e) reveals that the outermost layer (2-3 nm) is in amorphous form as evident from the absence of lattice fringes.For a better understanding of the distribution of this outermost layer a TEM image of 20 nm thickness coated on α-Fe 2 O 3 nanorod is presented in figure 5.This amorphous region is Al 2 O 3 and the film thickness has decreased after annealing.Interestingly, at the interface of the Al 2 O 3 outer layer and the α-Fe 2 O 3 core, the presence of another layer is clearly visible.The origin of this layer might be due to diffusion of aluminium atoms into the α-Fe 2 O 3 nanorods [27].The calculated d-spacing value of 0.59 nm is higher than the d-spacing of both α-Fe 2 O 3 (d = 0.37 nm for (012)) and Al 2 O 3 (d = 0.34 nm for (012)) (ICSD 01-085-1337).This suggests the formation of a mixed phase Al and Fe oxides.It is known from the literature that AlFeO 3 has a d value of 0.62 nm for (110) (ICSD code #00-030-0024).It is therefore evident that a phase similar to AlFeO 3 may have formed at the interface of amorphous Al 2 O 3 and crystalline α-Fe 2 O 3 .The SAED pattern captured on a single nanorod also shows both bright spotted patterns as well as rings made up of smaller spots, which points to the presence of crystalline α-Fe 2 O 3 and polycrystalline material -possibly the AlFe x O y phase.

Figure 4 .
Figure 4. (a) TEM images of a single nanorod, (b) high-resolution of an edge of the nanorod, and (c) SAED pattern of bare α-Fe 2 O 3 .Image (d) to (f) are presented for AFO-4 nanorod TEM image, high resolution edge and SAED pattern respectively.

Figure 5 .
Figure 5. TEM image of a Fe 2 O 3 nanorod with Al 2 O 3 coating of thickness 20 nm.

Figure 6 .
Figure 6.All the XPS results for Fe 2 O 3 and AFO -4 are put together here.Image (a) is the survey spectrum of both samples, (b) is Al2p peak coming from Aluminium present in AFO -4 sample.Image (c) and (d) are Fe2p and image (e) and (f) are O1s peak of Fe 2 O 3 and AFO samples respectively.

Figure 7 .
Figure 7.The bar diagram in (a), (b) and (c) presents the summary of gas sensing sensitivity at different temperature and NO 2 gas concentration.(d) presents the sensitivity versus NO 2 concentration for AFO-4 at different temperature.

Figure 8 .
Figure 8. Graph (a) and (b) are the change in resistance versus time graph at room temperature and 190 °C at 100 ppm NO 2 concentration for AFO-4 sample.Graph (c) and (d) are the tests for recovery time with respect to its response of 5 min for NO 2 gas by AFO-4 sample.

Figure 9 .
Figure 9. (a) Presents the selectivity of NO 2 gas with respect to the response of other common gases presented in the atmosphere for AFO-4 sample.(b) Presents the change in resistance with respect to relative humidity for α-Fe 2 O 3 and AFO-4 samples.(c) Presents the longevity of AFO-4 sample with respect to number of days collected in presence of N 2 flow, (d) and (e) present the FESEM images of the used α-Fe 2 O 3 and AFO-4 respectively after gas sensing.
contact material (Au) has a work function of ∼5.1 eV whereas Al 2 O 3 has ∼4.7 eV; such a difference in work function gives rise to Schottky barrier at the metal-insulator interface.The I-V curves for both samples are given in figures 11(a) and (b) for 8 different temperatures.A comparison of I-V nature at room temperature for Fe 2 O 3 and AFO-4 samples are presented in the supplementary information in figure S7 to show poor response of Fe 2 O 3nanorods at room temperature towards NO 2 gas.The Arrhenius equation yields the activation energy, which is the minimum energy required for the charge carriers to overcome the potential barrier upon contact with gas and take part into a reaction[8,44].The Arrhenius equation can be written as[8]

Figure 10 .
Figure 10.Schematic of gas sensing mechanism and change in depletion layer width of bare Fe 2 O 3 versus thin layer Al 2 O 3 coated Fe 2 O 3 nanorods (AFO -4).
findings in the literature, which demonstrate that amorphous layers contribute to higher oxygen adsorption, a correlation supported by our XPS results.When NO 2 adsorption takes place in the presence of the Al 2 O 3 layer, electrons from Fe 2 O 3 can easily participate in the gas-solid interaction process because of the passivation of the states at the Fe 2 O 3 -Al 2 O 3 interface and also due to a very low thickness of Al 2 O 3 layer.Thus, charge transfer kinetics improves compared to the bare Fe 2 O 3 nanorods, which leads to a reduction in the activation energy.The contribution of charges from the underlying Fe 2 O 3 in the sensing behavior is clear evident as the sensitivity response decreases with increasing Al 2 O 3 layer thickness.

Figure 11 .
Figure 11.Graph (a) and (b) are I-V plots of bare Fe 2 O 3 and AFO -4 samples collected for 8 different temperatures.Graph (c) is the Arrhenius plot.The calculation of slope gives rise to the activation energy 0.45 eV for Fe 2 O 3 and 0.29 eV for AFO -4 nanorods.

Table 1 .
Comparison of different α-Fe 2 O 3 structures for NO 2 sensing based on literature.
Subsequently, NO 2 reacts with the electrons as well as the adsorbed oxygen ions.The reactions are as follows: