A CH3NH3PbI3 film for a room-temperature NO2 gas sensor with quick response and high selectivity

Xianwei Fu, Shilong Jiao, Ning Dong, Gang Lian, *a Tianyu Zhao, Song Lv, Qilong Wang and Deliang Cui *a A room-temperature NO2 gas sensor with excellent performances is fabricated using an MAPbI3 (MA 1⁄4 CH3NH3 ) thin film. It presents a high response even under extremely low NO2 concentrations. Its average response and recovery times are only 5 s and 25 s at room temperature, respectively, exhibiting its quick-responsive character. In addition, the MAPbI3-based NO2 sensor exhibits good selectivity. Interestingly, the sensitivity of the MAPbI3-based NO2 sensor is strikingly improved under high-pressure gas conditions. This phenomenon can be used for the online monitoring of chemical reaction processes and in situ detection of trace-level gas impurities under high-pressure conditions. Furthermore, based on theoretical calculations, a simple model is proposed to illustrate the corresponding gas-sensing mechanism.


Introduction
In modern industrialized society, the emission of harmful gases has increased substantially. Toxic gases are hazardous to human health and extremely dangerous air pollutants, among which NO 2 is one of the most common harmful gases. The increasing concern for protecting the environment and biological systems has resulted in continuous endeavors in improving gas sensors. [1][2][3][4] To date, many types of inorganic and organic compounds have been utilized as sensing materials. 5,6 Although traditional metal oxides generally present high responses, they have an obvious disadvantage, which is the high working temperatures. For example, although CuO plates showed high sensitivity to NO 2 , the optimum operating temperature was 200 C. 7 A ZnO nanowire sensor exhibited high sensitivity and fast response to ethanol gas at a temperature as high as 300 C. 8 WO 3 was believed to be a promising material for NO 2 gas sensor, but it was typically used in the temperature range from 200 C to 350 C. 9,10 This disadvantage severely limits their further applications in the eld of gas sensors. To achieve the requirement of a low optimum working temperature, some organic compounds have been used as low-temperature gas sensing materials. However, quick response and recovery speed are other challenging issues for organic gas sensors. For instance, the response and recovery time of a lead phthalocyanine (PbPc) lm sensor were about hundreds of seconds for NO 2 . 11 The gas sensing behaviors of tetra-(tert-butyl) phthalocyanine copper (CuPc(t-Bu) 4 ) lms were investigated at room temperature, but their response and recovery speed were very slow. 12 Therefore, based on the advantage of the low working temperature of organic gas sensors, how to greatly improve the response and recovery speed has been a great challenge to date.
Organometallic halide perovskites MAPbX 3 (X ¼ Cl, Br, I) have attracted much attention due to their interesting and intriguing properties, including moderate bandgaps, large optical absorption coefficients, 13,14 extraordinarily long carrier diffusion lengths and high carrier mobilities. 15,16 They present some extensive applications in solar cells, light emitting diodes (LEDs), 17 photodetectors, 18 sensors, 19 eld effect transistors (FETs), 20 lasers 21 and so on. As a typical bipolar charge transport semiconductor, 22 the electrical conductivity of MAPbX 3 is greatly inuenced by some oxidizing and reducing gases. For instance, ammonium gas could induce a quick reversible phase transformation in a perovskite lm, resulting in an obvious decrease in its resistance. 23,24 MAPbI 3 was also sensitive to some polar organic compounds, such as 4-tert-butylpyridine 25 and acetonitrile. 26 MAPbI 3Àx Cl x lms could absorb many types of gases, e.g. NO 2 , SO 2 , alcohol and acetone, which caused obvious changes in their conductivity. 27 In addition, the resistance of an MAPbI 3Àx Cl x thin lm also exhibited a quick reversible decrease in the presence of moisture. These quick room-temperature reversible behaviors of perovskite lms indicate their potential application in the eld of gas sensors. Thus, MAPbX 3 -based gas sensors with low working temperatures, and quick responses and recovery speeds can be expected. However, most research groups have focused on the photoelectric elds of MAPbI 3Àx Cl x and a few investigations have been reported on MAPbX 3 -based gas sensors. Furthermore, because the stability of organic-inorganic hybrid semiconductors is strikingly improved under a high-pressure atmosphere, 28 it would be quite interesting to investigate the response performance of stabilized perovskite lms to target gases under high pressure.
Herein, we prepare MAPbI 3 thin lm gas sensors and investigate their gas-sensing properties under both ambient and high pressure. Interestingly, the gas sensors exhibit rather quick responses and recovery speeds, excellent selectivity and high sensitivity to NO 2 gas even at room temperature. Furthermore, the sensitivity of the prepared sensor is strikingly improved under high-pressure gas. Additionally, theoretical simulation of the adsorption of NO 2 molecules on the surface of perovskite is conducted to illustrate the response mechanism of the sensor.

Preparation of perovskite thin lm and gas sensor devices
In a typical process, 1 mmol of methylammonium iodide (MAI, AR, Tokyo Chemical Industry Co., Ltd, Tokyo, Japan) and 1 mmol of PbI 2 (AR, Aladdin Industrial Corporation, Shanghai, China) were successively dissolved in 1 mL N,N-dimethyl formamide (DMF, AR, Sinopharm Chemical Reagent Co., Ltd, China) at 65 C under stirring for 12 h. 29,30 Aerwards, the precursor solution was spin-coated on an alumina (Al 2 O 3 ) ceramic substrate (10 Â 14 mm) at 1500 rpm for 9 s, followed by spinning at 2000 rpm for another 10 s. The lm was then dried under high purity argon gas at 100 C for 30 min. Finally, gold interdigital electrodes were sputter-coated on the perovskite lm ( Fig. S1 †). Then, the gas sensor was prepared for further characterization.

Characterization of samples
Fourier transformation infrared (FTIR) absorption spectra were collected on a Thermo-Nicolet NEXUS 670 infrared spectrometer using the attenuated total reection (ATR) mode, and the wavenumber resolution was 4 cm À1 (range 4000-650 cm À1 ). Xray diffraction (XRD) patterns were recorded on a Bruker-AXS D8 Advance X-ray diffractometer with CuKa radiation (l ¼ 1.54186Å) in the 2q range of 10-60 . The morphology was observed via scanning electron microscopy (SEM, Hitachi S-4800) and scanning probe microscopy (SPM, Veeco Instruments Inc, Dimension Icon). The transient gas response characteristics (I-t curves) of the gas sensors were examined using a Keithley 4200-SCS semiconductor parameter analyzer at room temperature (25 AE 3 C) in a specially designed gas cabinet with a volume of 180 mL, and the relative humidity was 25-30%.
The high-pressure and atmosphere pressure gas-sensing performances of the sensors were examined using a specially designed chamber with a capacity of 180 mL (Fig. S2 †), which was connected to the Keithley 4200-SCS semiconductor parameter analyzer. Before characterization, the air in the chamber was completely expelled with high-purity argon gas. Aerwards, a specic volume of target gas was stored in the target gas cylinder which was connected to the chamber. When the pressure inside the chamber was increased to designated pressures (1, 2, 3, 4, 5, 6, 7 and 8 MPa) by introducing highpressure argon gas, the target gas stored in the cylinder was injected into the testing chamber. The variation in the resistance of the sensors was recorded simultaneously. When the measuring process was nished, the high-pressure gas was released from the chamber, which carries away the target gases. Then the resistance of the sensor recovered to its original value. Details of the operation process are presented in ESI Fig. S3. † For oxidizing gases, we dene the response R of a gas sensor as: whereas, for reducing gases, the sensor response R is dened as: For both oxidizing and reducing gases, the sensitivity S is dened as: where, R g and R a are the resistances of the sensor with and without target gases, respectively. c t is the concentration of target gases. Besides, the response (recovery) time of the sensors is dened as the time interval in which the variation in resistance reaches 90% of the total change upon introducing (evacuating) the target gases, respectively.

Theoretical calculation
To explain the gas-sensing experimental results, molecular dynamics simulation was conducted using the FORCITE package in the Materials Studio. The COMPASS (Condensedphase Optimized Molecular Potentials for Atomistic Simulation Studies) 31,32 force-eld was performed to describe the interaction between NO 2 and the surface atoms of perovskite. For the case in which NO 2 interacted with perovskite, the (110) surface was created by periodical replication of an elementary cell in both x and y directions, where the periodical condition was applied. The thickness of the vacuum slab was set to be 15Å in order to avoid unnecessary interactions between adjacent layers. The NVT ensemble with a simulation time of 5 ns was used to run the test, and the step was maintained at 1 fs.

Results and discussion
Initially, the phase composition of the lm was identied via XRD. In Fig. 1a Fig. 1b presents the surface morphology of the MAPbI 3 lm, which was constructed by interconnected MAPbI 3 nanoplates. The lm has a large number of pores, which is also supported by the cross-section image (Fig. S4 †). The MAPbI 3 nanoplates were also measured via TEM (Fig. S5 †). Furthermore, the height prole of the lm recorded with AFM (within an area of 20 Â 20 mm 2 , Fig. 1c and d) reveals a rough surface with a root mean square roughness (MRS) of $125 nm. This structural characteristic provides plentiful diffusion channels and active surface sites for the target gas molecules. Thus, a good gas-sensing performance can be anticipated. Furthermore, the gas-sensing performance of the MAPbI 3 lm to NO 2 gas at different concentrations was tested. The sensor response was then calculated using eqn (1), as plotted in Fig. 2. When NO 2 gas was injected into the testing chamber, the resistance of the lm sensor with NO 2 adsorbed obviously decreased. This tendency was more prominent at higher concentrations of NO 2 (Fig. 2a). The inset reveals an almost linear relationship between the sensor response and the concentration of NO 2 , and the average sensitivity was as high as 0.62 ppm À1 (Fig. 2a). Besides, the dynamic response curves and sensor response vs. NO 2 concentration under different bias voltages are shown in Fig. S6. † The signal-to-noise ratio slightly increased with the improvement in bias voltage. More importantly, the detection concentration of NO 2 gas was as low as 1 ppm, and a clear response was still presented (Fig. 2a and S6 †). An extremely low detection limit of the MAPbI 3 lm sensor, which indicates rather high sensitivity at room temperature, could be undoubtedly expected. More interestingly, the room-temperature sensing performance of the MAPbI 3 lm sensor for NO 2 strikingly surpassed the hightemperature performance of most conventional metal oxide sensors (Table 1). 19,[35][36][37][38][39][40] Compared to the room-temperature CuTAP(t-Bu) 4 lm sensors, the gas-sensing performance of the prepared MAPbI 3 lm is also superior ( Table 1). In addition, the response-recovery curve of the lm (Fig. 2b) exhibits a much quicker response and recovery speed than MAPbI 3Àx (SCN) x lm sensors at room temperature. The average response and recovery times were $5 s and $25 s, respectively, which are preponderant for a room-temperature gas sensor.
Besides the sensor response, reproducibility and selectivity are also crucial for gas sensors. In order to investigate the reproducibility of the MAPbI 3 lm gas sensor, NO 2 was repeatedly injected into and evacuated from the testing chamber. The sensor response of the lm remained stable aer more than 12 cycles (Fig. 3a). For comparison, the reproducibility of the MAPbI 3 lm sensor in ambient environment with a relative humidity of 35% was also examined (Fig. S7 †). Compared with the situation in Ar atmosphere, although the current of this   In order to analyze the gas-sensing mechanism, in situ FTIR measurements of the MAPbI 3 lm exposed to NO 2 gas for different durations were conducted. The obtained spectrum (Fig. 4a) is in good accordance with the reported result. 41 The bands at 3091, 2819, 1659, 1487, 1416, 1244 and 902 cm À1 are attributed to the N-H stretch, C-H stretch, N-H bend, C-H bend, C-N stretch, C-H twist and C-H rock vibration modes, respectively. 42 It was worth noting that a new band appeared at 1325 cm À1 (Fig. 4b), which can be assigned to the surface bound NO 2 À ions formed by charge transfer chemisorption. 43,44 When the lm was exposed to NO 2 for a longer time, more gas molecules were adsorbed on the active sites in the porous lm. Therefore, this new band became stronger (Fig. 4c). The electron transfer process is also supported by the red-shi of the N-H stretching vibration band from 3091 cm À1 to 3076 cm À1 . The corresponding increase in hole concentration of the P-type MAPbI 3 lm led to the resistance decreasing in the electrical eld.
The interaction between the adsorbed NO 2 molecules and MAPbI 3 lm was further investigated via computational simulations. Fig. 5a-c show the radial distribution functions (RDFs) of NO 2 -H MA + on the typical (110) facet of perovskite, where, H MA + includes two types of -CH 3 and -NH 3 . The peaks at 2.50Å and 2.63Å (Fig. 5a) reveal a stronger hydrogen-bond interaction between -CH 3 and NO 2 , compared to that between -NH 3 and NO 2 (the peaks at 3.53Å and 2.94Å, respectively (Fig. 5b)). This result is consistent with that of FTIR spectra quite well. Subsequently, a schematic illustrating the electron transfer process from perovskite to the NO 2 molecule via the hydrogenbond channel was proposed (Fig. 5c). According to the FTIR spectra and theoretical simulation results, the adsorbed NO 2 molecules rst interact with the -CH 3 groups and then attract some electrons from the organic components. In fact, the inorganic Pb-I skeleton in the perovskite semiconductor serves as the major transportation path for the charge carriers. 45 Thus some electrons on the Pb-I skeleton subsequently transfer to MA + ions via the NH/I bonds (Fig. 5c). As a result, the hole concentration of the perovskite lm is enhanced, leading to a striking decrease in the resistance of the lm.
In addition, based on the theoretical calculation ( Fig. S9 †), the interaction between the NO 2 molecules and perovskite lm is enhanced with the improvement in gas pressure due to the increase in adsorption energy. Consequently, more effective electron transfer can be expected between them under a highpressure circumstance, which plays an important role in the excellent performance of this sensor. Therefore, the effect of pressure on the gas-sensing performance was investigated. In our experiments, V, R and T are all constants, thus the molar number of argon gas n is simply proportional to the pressure p. According to the ideal gas equation, n ¼ p/p 0 Â n 0 (where, n 0 is the molar number of argon gas at atmospheric pressure and p 0 Fig. 3 (a) Reproducibility of the MAPbI 3 sensor exposed to 30 ppm NO 2 and (b) sensitivity to a series reference gases. Fig. 4 FTIR spectra of the MAPbI3 film sensor exposed to 1% NO 2 for different durations. is atmospheric pressure). Besides, the molar number of target gases n t is much less than n (Fig. S10 †), thus the molecule number concentration of target gases at high pressure c t can be expressed as: According to the way that we ignore the molar number of target gases, the error was determined to be only 0.005& (Fig. S10 †), thus this approach is feasible. Fig. 6a shows the dynamic response of the MAPbI 3 lm sensor to NO 2 gas at high pressures, where the molecule number concentration of NO 2 remains constant for all the pressures, namely, 5 ppm. Obviously, the sensor response strikingly increased in a near-linear way with an increase in pressure, and the MAPbI 3 lm sensor still exhibited rather quick response and recovery times of 11-25 s and 20-60 s under high pressure. According to eqn (3) and (4), we calculated the sensitivity of the MAPbI 3 lm sensor under high pressure, and the results are presented in the inset in Fig. 6a. Interestingly, the sensitivity of the MAPbI 3 lm sensor strikingly increased under high pressure, which should be closely related with the fact that the adsorption energy of the NO 2 molecules on the surface of the sensor increased (Fig. S9 †) and more NO 2 molecules were strongly absorbed the surface of MAPbI 3 lm under higher pressure, 46 resulting in more effective electron transfer and an increase in hole concentration. Actually, in the ideal gas model, the adsorption of gas molecules is not only determined by the adsorption energy, but also depends on the partial pressure of the target gas. Therefore, the positive effect of the partial pressure of NO 2 cannot be ignored for the enhanced sensitivity. Besides, the MAPbI 3 lm sensor still exhibited quite high selectivity towards NO 2 even at high pressure (Fig. S11 †). This phenomenon has potential applications in improving the performance of gas sensors when detecting ultra-low concentrations of target gases.

Conclusions
By utilizing an organic-inorganic hybrid semiconductor, a gas sensor with strikingly improved sensitivity and rather quick response and recovery speed was prepared at room temperature. Furthermore, the sensitivity of the gas sensor was greatly increased under high-pressure argon gas, providing the possibility of enabling the sensor to detect target gases with ultra-low concentrations by applying a high-pressure. These ndings may provide a new route for the fabrication of ultra-sensitive gas sensors for the on-line monitoring of chemical reaction processes and in situ analyzing impurity gases under highpressure circumstances.

Conflicts of interest
There are no conicts to declare. Transient curve when high-pressure argon gas and NO 2 are successively introduced into the testing chamber at 1 and 7 MPa. All the curves were recorded at room-temperature under a bias-voltage of 10 V.