Flexible Paper-Based Polyacrylic Acid-Coated Silver Nanoparticle Film Sensors Prepared by Inkjet Printing for Ecient Ammonia Detection at Room Temperature

A series of disposable gas-sensing paper-based lm sensors were prepared rapidly by inkjet printing polyacrylic acid-coated silver nanoparticle (PAA/AgNP) ink onto the ordinary copy paper. The surface morphologies and chemical structures of the printed PAA/AgNP based sensing lms with various thicknesses and spacing widths of interdigital electrodes were characterized. The electrical properties and the gas sensing performance of the lm sensors were investigated and the results showed that the PAA/AgNP lm sensors presented excellent selectivity, reproducibility and long-term stability to ammonia (NH 3 ) gas at room temperature. The response of the PAA/AgNP lm sensor to NH 3 with the concentration of 25 ppm is 42.6% at 20 ℃ and 50% relative humidity (RH). The inuences of thickness, spacing of interdigital electrodes and relative humidity on the sensing properties of the PAA/AgNP lm sensors were also discussed and analyzed. Additionally, the PAA/AgNP lm sensor presented perfect exible stability and showed minor change in response value after 100 folding/extending cycles with the angle of 90°. In conclusion, the proposed high-performance paper based PAA/AgNP thin lm sensor holds great promise for exible, low-cost, portable, disposable and recyclable application in detecting NH 3 at room temperature.


Introduction
Ammonia (NH 3 ) has been widely employed in many industries like food processing, medical prognosis, agriculture, and refrigeration systems [1] . However, excessive NH 3 emission not only does harm to the surrounding environment and ecosystem, but also has negative impact on human's health [2] . NH 3 with a concentration higher than 35 ppm will carry a risk to damage the human cells, irritate skin and eyes, and destroy the mucosa of respiratory tract within 15 minutes [3] . Accordingly, it is greatly demanded to develop highly sensitive, disposable, convenient, and economical devices for NH 3 detection at room temperature. There are many methods to detect NH 3 , including mass spectrometry, chemical analysis, infrared spectrophotometry, gas chromatography, and gas sensor [4] up to now, among which, the gas sensor is the most commonly used for NH 3 detection due to its simplicity and convenience [5] .
Printing technology is aiding and revolutionizing the booming eld of exible electronic devices by providing a cost-effective way for processing kinds of electronic materials which are compatible with lowcost and exible substrates [6] . Recently, inkjet printing has received much attention in material and sensor preparation procedures due to its unique advantages, such as the potential for non-vacuum and low temperature processing, low materials waste, easy control, etc. [7] . What is more, inkjet printing technology is environmentally friendly because it does not need any harmful chemicals to wash away the excess materials on the substrate surface [8] . Meanwhile its advantages of fast fabrication and ease of mass production can help lower the cost of the inkjet-printed devices [9] .
Selecting optimal substrates is one of important matters to effectively realize low-cost and exible inkjetprinted devices [10] . Paper, made of plant bers, is one of the cheapest materials in the world. It has the advantages of availability, exibility, disposability and recyclability. In addition, paper allows passive liquid transport and has excellent compatibility with chemicals [9,11] . Moreover, paper has inherent porous structure and rough surface, which can provide a large surface area compared with at glass or nonporous plastic [12] . These advantages make paper a conducive platform for ink-jet printing to develop exible and low-cost functional devices [13] . Up to now a few paper-based NH 3 sensors have been presented in the literatures. Avisek Maity et al. [14] reported a type of paper sensor working at room temperature can be made using perovskite halide CH 3 NH 3 PbI 3 (MAPI) to detect the presence of NH 3 by color change. However, MAPI has strong toxicity, which is not environmentally friendly and is not suitable for daily NH 3 detection. Lianghui Huang et al. [15] fabricated colorimetric NH 3 gas sensors by facile ltration of modi ed Berthelot's reagents on a porous paper substrate for the chemical analysis of wastewater or seawater. However, this detection was based on the separation of NH 3 gas from the aqueous matrix and irreversible chemical reactions at the gas-solid interface, which may be not gainful for NH 3 detection in air.
In this work, we rst report the preparation of high performance polyacrylic acid-coated silver nanoparticle (PAA/AgNP) based NH 3 sensor on paper substrate by inkjet printing at room temperature. PAA/AgNP based gas sensing lms with different thicknesses were fabricated by printing inks onto the paper substrate coated with interdigital silver electrodes of different interdigital spacing. The electrical response of PAA/AgNP thin lms towards NH 3 with low concentration (1-25 ppm) was demonstrated at room temperature and the gas sensing mechanism was proposed.

Materials
All chemicals used in this work were analytical grade reagents without further puri cation. where V L is the volume of corresponding organic liquid, P is the gas pressure, C ppm is the gas concentration, V C is the volume of test cavity, R is the gas constant, T is the temperature and ω, ρ and M are the mass fraction, the density and the molar mass of the organic liquid respectively. According to the actual situation of the static test in this experiment, the gas pressure P is 1.01*10 5 Pa, R is 8.314, T is 273.15 K, and the volume of the quartz testing tube is 50 ml. Taking pure ethanol gas as an example, M is 46 g/mol, ρ is 0.791 g/ml and ω is 1.

Fabrication of PAA/AgNP ink and thin lm sensor
Based on our previous work [16] , we made some improvements in preparing AgNPs in order to implement the fabrication of PAA/AgNP lm sensor. The procedure to synthesize precursor ink is shown in Fig. 1.
Firstly, TEA (15 g) and PAA (4 g) were dissolved in deionized water (60 ml). This mixture was stirred at room temperature for 2 h, resulting in a transparent colloidal solution. Secondly, AgNO 3 (9 g) was dissolved in 10 ml of deionized water and underwent ultrasonic dispersion for 20 minutes, which then was dropped into the transparent colloidal solution with a dropper. Thirdly, this mixture of TEA, PAA and AgNO 3 was stirred at room temperature for 18 h and then heated to 65℃ with continuous stirring for 2 h, leading to the PAA/AgNP solution with the color of dark red. The PAA/AgNP ink was prepared by mixing PAA/AgNP aqueous solution and SDS with the mass ratio of 1:0.009. Here, SDS used as surfactant to decrease the surface tension of the ink. Finally, the PAA/AgNP ink was loaded into different cartridges and then printed onto a paper substrate coated with interdigital silver electrodes using our modi ed ink-jet printer. The details are shown in Fig. 1a. After printing, the lms were dried in air for 2 h. As shown in Fig. 1b, the PAA/AgNP lm sensors were divided into four types according to the spacing between the neighboring interdigitated electrode ngers. One had no electrode ngers, and the spacing widths of other three were 5.0 mm, 2.0 mm and 1.0 mm, marked as E-1, E-2, E-3 and E-4, respectively.

Characterization of PAA/AgNP thin lms
The surface morphologies of the samples were examined by a eld emission scanning electron microscope (FESEM, Hitachi S4800). The Energy-dispersive X-ray spectroscopy (EDX) was used in order to analyze the composition of the PAA/AgNP lm sensor. The thermogravimetric differential thermal analysis (TG/DTA) of the PAA/AgNP powder was performed using a Perkin-Elmer Pyris Diamond TG/DTA in air with a heating rate of 10 K min − 1 . The chemical structure and bonding characteristics were investigated by Fourier transform infrared (FTIR) spectroscopy (Model: Perkin Elmer100). X-ray photoelectron spectroscopy (XPS) was performed in a Thermo Scienti c K-Alpha X-ray photoelectron spectrometer, using monochromatic Al Kα radiation (1486.6 eV). The XPS spectra were calibrated with reference to the C 1s speak (284.8 eV).
The detailed gas-sensing experimental process and platform have been shown in Fig. 2. Sensing measurements were performed in a quartz tube with a volume of 50 ml. Standard NH 3 (Zhengzhou Xingdao Chemical Technology Co., Ltd.) was rstly lled into a gas sampling bag (50 ml, Dalian Delin Gas Packing Co., Ltd., China), and then was injected into the quartz testing tube through the syringe to obtain the NH 3 with corresponding diluted concentration. The resistances of the lms were measured by a digital multimeter (VICTOR 8246A), and the values were recorded (every 0.3 s) by the matching LabVIEW program. The experimental relative humidity was about 50% at 20℃. The effect of different humidity on the response of the sensor was examined at 20℃ using a humidity meter purchased from Shanghai Duohe Equipment Co., Ltd, which could control the testing temperature and humidity precisely.
At a certain temperature and humidity, the sensor response (S) was de ned as S= (R-R 0 )/ R 0 × 100 (%), where R 0 and R represent the resistance values of the sensor in air and tested gases, respectively. The response time is de ned as the time required for the resistance variation value to reach 90% of the maximum value. The recovery time is de ned as the time required for the resistance recovery value to reach 10% of the maximum value after the gas was discharged. The TG/DTA result of the PAA/AgNP composite is presented in Fig. 4. The sample shows a continuous weight loss from room temperature to 950℃ with a cumulative weight loss of 13.4%, which is attributed to the desorption of the organic constituents from the surface of the AgNPs [16] . There is a strong endothermic peak at approximately 955℃, which may be attributed to the melting of bulk silver. There is a cumulative loss of 0.3% from 955℃ to 1010℃, which may be attributed to the temperature disturbance. These results demonstrate that the solid AgNPs powder contained more than 86.3 wt% silver.

Results
The FTIR absorption spectrum of the PAA/AgNP lm is illustrated in Fig. 5. The peak at 1700 cm − 1 is assigned to C = O stretching vibration mode, and the wide peak at 3300 cm − 1 corresponds to O-H stretching mode [17] . This result suggests that the lm contains carboxyl group which origins from PAA. Figure 6 shows the XPS spectra for Ag 3d peaks of pure AgNPs and PAA/AgNP sensing lm, respectively.
Based on the results of deconvolution, the AgNPs exhibit two characteristic peaks at binding energies of 374.2 eV (Ag 3d 3/2 ) and 368.2 eV (Ag 3d 5/2 ), corresponding to metallic Ag. However, the PAA/AgNP sensing lm shows the presence of two Ag 3d chemical states as one doublet at 374.0 eV (Ag 3d 3/2 ) and 368.0 eV (Ag 3d 5/2 ), and the other one at 376.0 eV (Ag 3d 3/2 ) and 370.0 eV (Ag 3d 5/2 ). The shifting of the doublet to higher binding energy may be attributed to the interfacial interaction and charge transferring between the AgNP and the polymer (PAA) [18] .

Electrical property of PAA/AgNP thin lm
The thickness of the PAA/AgNP lm could be controlled by the number of printing cycles. The resistance curves of the lms with different printing cycles and different shapes of interdigital electrodes are shown in Fig. 7. It illustrates the changes in resistances of the PAA/AgNP thin lms associated with different spacing of interdigital silver electrodes and number of printing cycles. At the beginning, the resistances of E-1, E-2 and E-3 were all larger than 50 MΩ with the printing cycles less than 5, which was beyond the range of the multimeter. Increasing the number of printing cycles leads to an increase in the lm's thickness and a decrease in the lm's resistance. The high resistance of the lms printed for 5 cycles may be due to the discontinuity of the lm surface. The lms become more continuous and their resistances decrease obviously after printed for 9 cycles. Afterwards the resistances reduce slightly and then tend to be stable. For the same printing cycle, the resistances of PAA/AgNP thin lms reduce as the spacing between interdigital electrode ngers decreases. This may be determined by the length of the conductive path in PAA/AgNP layer between two interdigital electrode ngers [19] . So in the next study, we chose the sensor of E-4 printed for 9 cycles with low initial resistance for the performance investigation.

The selectivity of PAA/AgNP lm sensor
Selectivity is the most critical parameter for gas sensor, which determines whether the sensor could be used in complex atmospheric environment [20] . Herein, three kinds of interfering gases including 1000 ppm formaldehyde, 100 ppm methanol and ethanol were chosen to evaluate the selectivity of the PAA/AgNP lm sensor (E-4, printed for 9 cycles) to NH 3 at 20℃. As shown in Fig. 8, the PAA/AgNP lm sensor exhibits a signi cant response of 42.6% to 25 ppm NH 3 , while it shows low responses of -10.2% and − 5.8% to methanol and ethanol with concentrations 4 times higher than NH 3 , and little response of -1.7% to formaldehyde with a concentration 40 times higher than NH 3 . What is more, the PAA/AgNP lm sensor responds with a decrease in resistance when exposed in formaldehyde, methanol and ethanol, which is contrary to its response to NH 3 . It is con rmed that the PAA/AgNP lm sensor possesses a better selectivity to NH 3 and is a promising candidate for future applications to detect NH 3 at room temperature.

The response of PAA/AgNP lm sensor to NH 3 with various concentrations
The real-time response-recovery curve of the PAA/AgNP lm sensor (E-4, printed for 9 cycles) for 1-25 ppm NH 3 at 20℃ is shown in Fig. 9, which exhibits excellent dynamic response/recovery characteristics to NH 3 gas. The response value of the PAA/AgNP lm sensor is 1.6% to NH 3 with the minimum concentration of 1 ppm. The response value increases with the concentration of NH 3 rising, which reaches to 42.6% when the NH 3 concentration is 25 ppm. It is inferred that there is a linear relationship between the response value and the NH 3 concentration in the inset of Fig. 9. As presented in Fig. 10, the PAA/AgNP lm sensor (E-4, printed for 9 cycles) responds quickly and the response time is about 52 s with an increase in resistance when exposed to NH 3 . The resistance falls back to the initial state within a short time and the recovery time is 231 s, without baseline drift after it is placed in air, which reveals a good reversibility of the PAA/AgNP lm sensor.

The stability of PAA/AgNP lm sensor
Stability is one of the main evaluation criteria for gas sensors and also a comprehensive performance of sensor reliability in applications. In terms of the length of time, it is usually divided into short-term repeatability and long-term stability [21] .
In order to investigate the stability of PAA/AgNP lm sensor, we have tested short-term repeatability and long-term stability of the E-4 sensor printed for 9 cycles. Figure 11a illustrates a series of real-time response of PAA/AgNP lm sensor to 25 ppm NH 3 at 20℃. It is observed that the sensor could be continuously operated for ve cycles with similar response values, and the response could fully return to the initial state at the end of each response-recovery cycle, which indicates the PAA/AgNP lm sensor can be reproduced. The long-term stability of the PAA/AgNP lm sensor aged for a month was tested to 25 ppm NH 3 at 20℃ and the result has been shown in Fig. 11b. It is found that the responses of the lm keep similar and stable as the values a month before. These results indicate a remarkable short-term repeatability and long-term stability of the PAA/AgNP thin lm sensor.

The in uence of printing cycles and different spacing of interdigital electrodes on the response of PAA/AgNP lm sensor
The thickness of the lms is an important factor that in uences the sensor response. Via utilizing the inkjet printing method, the thickness of the lms could be easily changed by varying the number of printing cycles [22] . Figure 12a illustrates real-time response curves of PAA/AgNP sensor to 25 ppm NH 3 at different printing cycles. As shown in Fig. 12b, it is evident that the responses of the E-4 lm sensors increase initially with the increasing of printing cycles. When the number of printing cycles are 5, 9 and 11, the responses of the lms to 25 ppm NH 3 are 39.5%, 42.6% and 51.1%, respectively. The optimal response is obtained after printed for 13 cycles with the response value of 66.5%. However, further increasing the printing cycles leads to a decrease of the sensor response. When the lms are printed for 15 and 20 cycles, the responses are 58.5% and 49.2%, respectively. The above results indicate that the sensor response is largely dependent on the lm's thickness. Both too thin (printing for 5 cycles) and too thick (printing for 20 cycles) lms will lead to limit the NH 3 adsorption capacity.
In order to investigate the NH 3 sensing properties of the PAA/AgNP lm sensor with different spacing of ngers to 25 ppm NH 3 at 20℃, the responses of four type sensors printed for 9 cycles are shown in Fig. 13. The inset gure of Fig. 13 are the dynamic response curves of the sensors. The responses of E-1, E-2, E-3 and E-4 are 127.6%, 85.3%, 68.1% and 42.6%, respectively. It is clear to nd that the greater spacing between the ngers of the PAA/AgNP lm sensors, the higher response to NH 3 . According to C. M. Yang [19] , it is concluded that the sensor with a larger spacing of interdigital electrodes has a higher surface area ratio (the ratio of the paper surface area to the unit area including the paper and Ag electrode, i.e. paper area / (paper area + Ag area)), which insures more NH 3 could be adsorbed for a higher response.

The in uence of humidity on the response of PAA/AgNP lm sensor
Relative humidity is one of the major matters which could affect the gas sensing performance of PAA/AgNP lm sensor. The in uence of relative humidity on the NH 3 -sensing performance of PAA/AgNP lm sensor (E-4, printed for 9 cycles) was examined by detecting the response to 25 ppm NH 3 under various relative humidity as shown in Fig. 14. The responses of PAA/AgNP lm sensor are 124.5%, 76.4%, 42.6%, 35.8% and 26.3% under relative humidity of 30%, 40%, 50%, 60% and 70%, respectively. It is observed that the sensor response decreases with increasing the relative humidity. It is suggested that water molecules occupy some of the active sites on the surface of exible sensor and hinder the adsorption of target gas molecules at a higher humidity, which is the probable cause to the decreasing of the response at higher humidity conditions [23,24] . From Fig. 14, it is shown that the sensor's response is greatly affected by the relative humidity. Based on this, improving the sensor's anti-interference to humidity is the direction of our future work. Adding a humidi er to the test instrument to x the humidity in the quartz testing tube at a constant value is also an optional proposal to eliminate the in uence of humidity on the sensor's performance.

Flexibility of PAA/AgNP lm sensor
The sensing properties of PAA/AgNP lm sensor (E-4, printed for 9 cycles) to different concentrations of NH 3 was evaluated after 100 folding/extending cycles with the folding angle of 90°, and the results are shown in Fig. 15. After 100 folding/extending cycles, no obvious response change of the sensor is observed for 1 ppm and 25 ppm NH 3 . A slight response increases with a standard deviation of less than 2% could be seen for 5 ppm, 10 ppm and 15 ppm NH 3 . The inset of Fig. 15 illustrates the responserecovery curves of PAA/AgNP lm sensor to 25 ppm NH 3 , which reveals no obvious response change for the sensor after 100 folding/extending cycles. All the results prove that the PAA/AgNP lm sensor possesses a great exibility for sensing properties, which is promising to be integrated into hand-held or wearable device.

Sensing mechanism
The sensing mechanism of the excellent response to NH 3 by PAA/AgNP lm sensor is proposed and illustrated in Fig. 16. As the XPS results inferred, there exists a charge transferring between the silver particles and the PAA coating [18] . When the PAA/AgNP lm sensor is placed in air (Fig. 16a), the electrons of AgNPs normally transport through the AgNPs to the interface between AgNPs and PAA and attract H + of the PAA. Due to the fact that the AgNPs are coated by PAA, the transport of H + between PAA and AgNPs becomes weak, which results in the resistance of the sensor being relatively large. When the PAA/AgNP lm sensor is exposed in NH 3 (Fig. 16b), the NH 3 molecule attracts the H + of PAA to form NH 4 + , which leads to the interfacial electrons between AgNPs and PAA owing back into the AgNPs. As a result, the conductive paths reduced resulting in the increase in the resistance of the PAA/AgNP lm sensor. On the contrary, when the sensor is exposed in ethanol, methanol and formaldehyde, their molecules will give H + to PAA, which promotes electrons owing to the interfacial between AgNPs and PAA. As a result, the conductive paths increased resulting in the decrease in the resistance of the PAA/AgNP lm sensor. However, compared with methanol and ethanol, formaldehyde has a poor ability to give H + , so the sensor shows almost no response to 1000 ppm formaldehyde.

Conclusions
In this work, a exible room temperature NH 3 sensor based on PAA/AgNP lms with different thickness and spacing of interdigital electrodes on paper substrates were successfully developed by inkjet printing method. The NH 3 -sensing performance was evaluated and the results demonstrated that the PAA/AgNP lm sensor possessed remarkable selectivity, response, reproducibility, long-term stability, low detection limit (1 ppm) and outstanding exibility. Meanwhile the inkjet printing technology using paper substrate help to improve the processing of low-cost, portable, disposable, recyclable lm sensors.

Declaration of Competing Interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.