Introduction

The Environmental Protection Agency’s Clean Air Act was amended in 1990 to include Title VI—Stratospheric Ozone Protection [1]. Under these changes, we have previously seen the phaseout of “Class I” chlorofluorocarbons (CFCs) which is currently being followed by the phasing out of “Class II” hydrochlorofluorocarbons (HCFCs), collectively known as Freon. A 99.5% reduction of HCFC production and consumption was mandated effective in 2020, and therefore, a pressing need exists for greener alternative refrigerants to Freon gases [2]. Compounding this issue, refrigerant gas [e.g., ammonia (NH3)] leaks are one of the most common faults within commercial systems [3], accounting for an annual leakage rate between 11–30% [4, 5]. Management of these leaks not only results in significant economic loss but also raises health and environmental concerns [6]. Naturally derived substitutes such as hydrocarbons or NH3 can be toxic below 50 ppm and/or flammable at concentrations of 15–28% by volume in air [7, 8] Therefore, in our progression toward emerging alternative refrigerants to HCFCs, it is crucial to develop a commercially viable early leak detection method for improved economic and safety outcomes.

With an ozone depletion potential (ODP) = 0 and global warming potential (GWP) < 1, NH3 stands in stark contrast to typical fluorinated refrigerants (e.g., HCFC-142b ODP = 0.07, GWP = 2310) [9]. One of the most promising avenues toward widespread adoption of NH3 as a green refrigerant relies on the advancement of chemiresistive sensors. In particular, polyaniline (PANI) has received much attention owing to its ability for rapid NH3 detection and its relatively easy and low-cost synthesis methods [10,11,12]. In its conductive emeraldine salt (ES-PANI) form, NH3 gas can abstract a proton from PANI in a reversible acid–base reaction that renders the polymer in non-conductive emeraldine base (EB-PANI) form [13]. As protonated ES-PANI retains a free polaron and positive charge, various dopants can serve as the counter-anion to significantly improve processability, as in the case of dodecylbenzenesulfonic acid (DBSA)-doped PANI, which readily forms aqueous dispersions [14]. These dispersions have been developed into inks and are being used in the development of additively manufactured gas sensors [15,16,17]. Accordingly, on the route toward commercialization of printed PANI-based sensors, some of the most exciting work was pioneered by Crowley et al. where a dual-step screen-printed silver (Ag) interdigitated electrode (IDE) followed by inkjet-printed ES-PANI is employed to construct fully functional sensors with detection capabilities as low as 5 ppm [17].

In the past decade, a rapid progress has been made in the design and manufacturing of additive digital printing systems opening up a path toward an entirely new world of electronic components, devices, and systems. Among the emerging printing techniques, aerosol-jet printing enables high-resolution non-contact deposition of both the Ag electrode and PANI materials. Despite their low cost and widespread utilization, inkjet and screen printing are restricted by low spatial resolution typically to a lower limit of 30 μm for linewidth, while aerosol jets can achieve 10 µm linewidths. Previous fabrications typically have employed screen printing for electrode arrays because inkjet printers experience high rates of nozzle clogging and are limited to a narrow viscosity range near that of water (0.01–0.02 Pa-s for inkjet vs. 0.5–5.0 Pa-s for screen) [18] which also presents challenges for reproducible print quality. Because direct-write aerosol techniques allow for a wide range of ink viscosities (0.001–0.16 Pa-s) [18], the manufacturing throughput can be effectively increased as combined printing processes (e.g., screen followed by inkjet printing) are reduced to a single technique requiring minimal preparation, maintenance, and operator input. Additionally, due to the long focal length of the nozzle beam and wide distance between the nozzle and substrate in aerosol jet, prints maintaining nearly constant linewidths can be rendered on curved and textured surfaces [18]. High resolutions, good film uniformities, and streamlined manufacturing processes warrant the adoption of aerosol-jet printing for the simplified fabrication of miniaturized functional NH3 sensing devices across various commercial applications.

Herein, to the best of our knowledge, we report the first all-aerosol-printed PANI-based NH3 gas sensor capable of < 5 ppm sensitivity and a sub-ppm detection limit. More importantly, because of the low power consumed by the sensor (typically microwatts or less), the present approach is also promising for the advancement of miniaturized multifunctional self-powered wireless sensor platforms to realize energy efficiency and cost-saving opportunities across widespread applications. Not only are NH3 gas sensors possessing wireless modes of operation highly desirable for residential and commercial refrigerant industries, but the mass production of these miniaturized devices may also find use as smart labels for food spoilage [19], diagnostics in kidney disease [20], and general environmental gas pollution monitors in the agricultural and automotive sectors [21]. Additionally, miniaturization not only aids manufacturing of standalone devices but is also particularly important for enabling lightweight, portable NH3 detection instruments that may find use in tactical military operations, early warning systems for the coal mining industry, and other diverse applications.

Materials and methods

Formulation of PANI ink

The synthesis of inkjet printable DBSA-doped PANI has been previously reported and is followed [22, 23]. Briefly, 0.6 mL of fresh aniline is sonicated in 20 mL of Milli-Q deionized (DI) water for 30 min to evenly disperse the monomer. In a 125-mL round-bottom flask, 3 g of 70% (wt.%) DBSA/isopropanol is added to 20 mL Milli-Q DI water and stirred for 30 min. The aniline dispersion is slowly added to the DBSA solution while stirring, and once solvated, 0.36 g ammonium persulfate is carefully added to initiate the oxidative polymerization. The reaction is carried out at room temperature (RT = 25 °C) for 2.5 h and proceeds with the formation of brown aniline oligomers, followed by undoped blue EB-PANI, and finally, the deep emerald doped ES-PANI product. To isolate PANI nanoparticles, the reaction mixture is centrifuged at 5000 rpm for 30 min and the supernatant is collected. Further nanoparticle separation is achieved via dialysis against DI water to remove unreacted oligomers, aniline, and excess DBSA, followed by 0.1 μm pore syringe filtration. All chemicals were purchased from Sigma-Aldrich and used as received. Spectroscopic characterizations of PANI were carried out by using UV–Vis (Carry 5000) and nuclear magnetic resonance (1H NMR, Bruker Avance III 400) spectroscopy (Figs. S1 and S2, respectively).

Sensor fabrication

Ag IDEs with finger and gap widths of 50 μm and 65 μm, respectively, are prepared on polyimide (PI) substrates in-house using digital aerosol-jet printing (Optomec) with the following parameters (see SV1 and SV2): nozzle print head size = 150 μm; sheath gas flow = 20 ccm; atomizer gas flow = 40 ccm; atomizer current = 0.4 mA; printing speed = 4 mm/s; and line size = 20 μm. Substrates are cured at 120 °C for 10 min on a hot plate (Torrey Pines Scientific). Two coats of PANI ink are then applied via aerosol printing to form a ~ 100-nm-thick circular film on the electrode. Morphological characterization of PANI films is performed using a Zeiss Merlin VP high-resolution field-emission scanning electron microscope (SEM) equipped with an energy-dispersive X-ray spectroscopy (EDS) unit. Figure 1a and b depicts a schematic representation and the as-fabricated NH3 sensor, respectively. Figure 1c shows a typical morphology of PANI bridging a pair of Ag electrode fingers.

Figure 1
figure 1

a Schematic of sensing film on interdigitated electrodes b as-fabricated PANI-based NH3 gas sensor, and c SEM image of the PANI film on the sensing device

Full size image

Gas sensing characterization

A custom flow-dilution apparatus is used to provide a humidity-controlled environment of NH3 gas in air (Fig. S3). Flow controllers (MKS 1479) are used to mix bottled air (Airgas) with NH3 prediluted to 0.2% in nitrogen (Airgas). The air is humidified as required by mixing dry air proportionately with air passed through an evaporative water boiler with a condenser controlled to approximate RT by a water recirculator. The water content of the humidified air is monitored by a chilled-mirror hygrometer (General Eastern). Colorimetric tubes (Gastec) are used to verify that the diluted NH3 concentrations are consistent with flow dilutions. During the experiment, a 20-min presoak of NH3 through the flow controllers is necessary to avoid retention and low concentrations of NH3. A total flow of 400 sccm is maintained by software at all times regardless of the humidity levels and NH3 concentrations. A PTFE hose conducts the mixed gas to a custom-machined plastic enclosure (Delrin) with an internal volume of roughly 1 mL in which the printed sensor is placed. Spring contacts connect the sample to a power supply (Agilent) typically set to 50–100 mV and a data logger (HP 34970A) to read the current (0.05–50 µA) through the sensor. Readings are automatically recorded at 10–20 s intervals. For temperature studies, the sample enclosure is placed on a temperature-controlled hotplate (Torrey Pines Scientific) and insulated with aluminum foil.

Results and discussion

The surface morphologies of the PANI films in between and on the IDE conductive traces, along with the corresponding EDS carbon and Ag elemental maps, are displayed in Fig. 2a, c, and e and Fig. 2b, d, and f, respectively. A relatively uniform coating is evident in the space between the electrodes, where the carbon yield is highly homogenous (Fig. 2a,c). On top of the electrodes the film is comparably rougher, revealing the granular structure of Ag and a relatively less intense and less uniform carbon distribution.

Figure 2
figure 2

Surface morphologies of PANI films and corresponding EDS elemental carbon and silver maps in between a, c, e and on b, d, f the IDEs. Elemental mappings are acquired from the boxed regions marked in parts (a) and (b)

Full size image

Sensor response, selectivity and repeatability

As defined by the National Institute for Occupational Safety and Health, NH3 is immediately dangerous to life or health at 300 ppm level, which formed our sensor development and characterization focus. The change in resistance of the aerosol-printed PANI sensor at RT upon exposure to dry NH3 in a logarithmically ascending series of concentrations ranging from 5 to 1000 ppm is displayed in Fig. 3. Initially, the resistance of the device is roughly 103 Ω that increases by over two orders of magnitude as NH3 levels increase to 1000 ppm. At 100 mV, this resistance results in very low power consumption of 10 µW or less. For each pulse of NH3, the resistance of the sensor first increases rapidly and then asymptotically climbs. When the gas flow is stopped, an initial fast drop in resistance is followed by a slow exponential decrease back to baseline. The observed increase in resistance after exposure to NH3 is ascribed to the adsorption of the gas which triggers deprotonation of ES-PANI, progressively rendering the polymer non-conductive. This process is reversible: as NH4+ forms and DBSA− is freed, the conjugate base of DBSA quickly abstracts a proton from ammonium (PANI-DBSA + NH3 ⇄ PANI + DBSA− + NH4+) and in turn the acid serves to re-dope EB-PANI back to its original conductive DBSA-PANI form [24]. Note that despite a sudden response upon exposure to NH3, the longer time intervals toward saturation and recovery may be attributed to the relatively compact (i.e., low surface-to-volume ratio) surface morphology of PANI films on PI substrates. Nevertheless, as can be seen from the inset of Fig. 3, the sensor detects NH3 gas in ambient conditions at concentrations as low as 5 ppm, which is comparable or better than the PANI-based or PANI-hybrid sensors reported in the literature [15, 25,26,27,28].

Figure 3
figure 3

Dynamic resistance transients of PANI-based sensor. Inset shows the response at 5 and 10 ppm NH3 exposure

Full size image

In addition, although under dry conditions the response does not reach equilibrium with each 30 min NH3 exposure, the PANI sensor demonstrates a roughly linear (correlation coefficient, r2 = 0.99) response to higher NH3 concentrations tested (Fig. 4). Here, the response of the sensor is defined as the normalized change in resistance, ΔRS = (RG − RA)/RA, and its sensitivity as S = ΔRS/ΔCG, where RG and RA represent the resistances of the PANI film in the presence of NH3 and air, respectively, and CG denotes the gas concentration. Accordingly, the sensor sensitivity (albeit under non-equilibrium conditions) is expressed by the slope of the line of best fit (18.4 × 10–2/ppm) to the sensing response versus gas concentration. To derive the detection limit (DL) of the sensor, we follow the procedure outlined in previous reports [29] and use the IUPAC definition of DL (ppb) as 3× sensor noise/sensitivity [30]. The sensor noise (1.5 × 10–3) is taken as the variation in the change of normalized resistance at the baseline using the root-mean-square (rms) deviation. Using the definition established above, the DL value is estimated as 24 ppb. The observed linear relationship between the sensor response and gas concentration, coupled with low detection limit and low baseline drift, is all highly valuable for the integration of low-cost, low-power sensors on thin polymer platforms. Since selectivity plays a key factor in defining a sensor, the sensitivities of the PANI-based sensor toward other potentially interfering non-Freon refrigerant gases including isobutane (C4H10) and propane (C3H8) at RT are shown in Fig. 5. Corresponding real-time responses are presented in Fig. S4. The sensor demonstrates a specific response to NH3 gas, exhibiting at least three orders of magnitude higher sensitivity relative to the other refrigerant gases. When corrected for the concentrations tested, the sensitivity ratios toward C4H10 and C3H8 to that of 20 ppm NH3 are 0.09 and 0.06%, respectively.

Figure 4
figure 4

Sensitivity of PANI-based sensor under varying concentrations of NH3

Full size image
Figure 5
figure 5

Selectivity of the PANI-based NH3 sensor toward selected hydrocarbon refrigerants. Note the concentration of both isobutane and propane was kept at a high level of 1% (i.e., 10000 ppm) to enable detection. Sensor response to lower NH3 concentrations are included for comparison

Full size image

In terms of repeatability and stability, the printed sensors not only display a low detection limit, but also retain a steady performance over time. Figure 6 illustrates the detection–recovery curves at 20 ppm NH3 exposures for 8-h followed by 16-h recoveries over an approximate six-day period. As shown, the PANI film is capable of stable gas sensing without a notable loss of functionality. It is also important to note that these data were collected approximately three months after the initial film deposition to elucidate the environmental stability of the fabricated sensors. Our results suggest that exposure at least to ambient laboratory storage conditions does not affect the subsequent performance and functionality of the sensor over an extended time period. The curves demonstrate a complete recovery to baseline resistance with no obvious drift due to a highly reversible reaction between the analyte gas and sensing material. Taken together, these observations point to the excellent stability of all-aerosol-printed PANI-based sensors for at least several months with preservation of full functionality after repeated gas exposures. After a successful laboratory-scale sensor development and performance characterization, long-term reliability evaluation is critical for commercial applications. Interplay of sensor calibration, power consumption, data processing, and communication requirements must be established for a practical manufacturing technology. For refrigeration technology integration, the key sensor requirements of long lifetime (5–10 years), battery powered operation, and wireless communication must be met to address preventive maintenance and deployment requirements which are cost intensive. At the first step, among the sensor technologies of interest for refrigeration application (infrared, electrochemical cell, metal–oxide–semiconductor, catalytic type and heated diode), the printed PANI sensors show promise to meet the cost, sensitivity, size, and low-power requirements. While it remains to be seen whether these sensors will satisfy the common commercial requirement that sensors have a “10-year operational lifetime”, the present results are highly promising to further focus on packaging pathways to achieve long-term performance and reliability in the environment of interest.

Figure 6
figure 6

Repeatability of PANI-based sensor following repeated exposures (denoted by the dotted lines) to 20 ppm of NH3 in dry conditions over 140 h

Full size image

Response under the influence of humidity and temperature

Next, we test the effect of relative humidity (RH) by subjecting the same sensor to different RH levels under varying concentrations of NH3 (Fig. 7). RH levels of 30 and 50% are chosen as they represent the typical conditions for indoor and air-conditioned spaces. Comparison of the present sensor responses with earlier studies reveals interesting differences. While Crowley et al. [15] report little interference from humidity, our results are in accordance with responses to humidity previously seen in other PANI-based sensors [31]. Under dry conditions, a fast-initial response to NH3 is followed by a continual upward drift, resulting in a sluggish stabilization requiring hours. At 30% RH, a similar steep rise is followed by a smaller downward drift. However, at 50% RH, stable but notably lower responses are observed at higher NH3 concentrations as compared to 30% RH and dry conditions. Most remarkable are the 10%/90% response and recovery to baseline times of roughly 1 min (rise time) and 5 min (recovery time) for both the 30% and 50% RH values. We speculate that atmospheric water vapor may act as a dispersant for PANI and free DBSA to re-dope and restore the film to its conductive ES-PANI form and thus to stimulate more rapid uptake of NH3 when humidity is introduced. In fact, the present results corroborate well with observations of increased conductivity in PANI following moisture absorption [32].

Figure 7
figure 7

Comparison of the sensing response of PANI-based sensor to various concentrations of NH3 under different relative humidities

Full size image

On the other hand, for higher concentrations and higher humidity levels (i.e., 50% RH), the sensor response rapidly stabilizes to a lower sensitivity compared to moisture-free conditions. The observed decrease in resistance with increased RH is suggested to be related to the increased formation of intermolecular H-bonds between PANI and H2O, leading to increased charge mobility in the polymer chain [33] (increased hole concentration through donation of the lone pair from the conducting complex toward water molecules) [33,34,35]. The observed PANI response under the influence of humidity is anticipated considering that humidity has been demonstrated in the literature to influence PANI’s conductance and response as previously mentioned. While these observations help elucidate the fundamental facets of PANI’s conductivity mechanisms, the exact relationship between humidity interference and PANI-based NH3 sensing is not entirely clear. As NH3 readily forms the hydroxide in the presence of moisture, water vapor may generate side reactions with NH4+ ions to form NH4OH. These side reactions with the analyte gas may impact device sensitivity as other basic species (e.g., gaseous amines, triethylamine) can cross-react with PANI and interfere with NH3 detection [35, 36]. Additionally, the backbone of the polymer may swell upon exposure to moisture, changing the morphological configuration of PANI, thus affecting the film’s conductivity and its response to the analyte [35,36,37,38]. Although further work is required to fully understand and clarify the underlying interaction mechanisms between PANI and H2O with respect to NH3 detection, the sensor remains responsive across the entire concentration range of NH3 exposures. Future work will explore the long-term stability of devices under varying environmental conditions (e.g., temperature and humidity).

In addition to humidity, the effect of operating temperature on sensor performance is also evaluated as shown in Fig. 8. It is well known that ES-PANI is less thermally stable with respect to molecular changes than its undoped EB-PANI counterpart. ES-PANI reverts back to EB-PANI upon heating as the doping fraction decreases with increase in temperature [39]. While our experimental conditions likely do not induce degradation of the polymer backbone, conductance of sulfonic acid-doped PANI is known to decrease as temperature increases, sharply dropping off > 160 °C due to loss of the counter-anions and unbound H2O molecules [39, 40]. While the sensor demonstrates operational functionality from ambient temperatures to 80 °C, as expected, the device’s sensitivity toward NH3 decreases. At 80 °C, the sensor is approximately half as sensitive as operation at RT. In fact, thermal stability studies of PANI-based and PANI-hybrid gas sensors have also reflected our findings of decreasing sensitivity to the analyte gas in the range of 22–48 °C [10, 16, 41, 42]. At elevated temperatures, the conductivity changes in ES-PANI are hypothesized to originate from cross-linkages that drive oxidative condensation reactions between abstracted H+ units and atmospheric oxygen. Because ES-PANI units can react with oxidants in either physical or electrochemical contact, the degradation of ES-PANI more readily occurs compared to its non-conductive counterpart [43]. In addition to the thermal limitations of unadulterated PANI-based sensors, the binding interactions between PANI and NH3 may change with respect to temperature, which may result in alterations in the association and dissociation rate constants. Further studies beyond the intended scope of this paper are needed to analyze the reaction kinetics and degradation mechanisms that may interfere with detection responses at elevated temperatures.

Figure 8
figure 8

The effect of operating temperature on sensor response to NH3 gas at a fixed concentration of 50 ppm. The time baseline for each temperature has been offset for clarity

Full size image

Conclusion

We have demonstrated a low-cost, low-power NH3 gas sensor on PANI-modified IDE arrays accomplished through an additive manufacturing-based technique of aerosol-jet printing. The fabricated sensors display rapid response, stable repeatability, and excellent selectivity to NH3 within the desired analytical range of 5–2000 ppm that are all crucial attributes for early leak detection. Our results also reveal the long-term operational potential of these sensors as no notable loss of functionality was observed following three months of exposure to ambient conditions. Additionally, temperature and humidity studies illustrate responsiveness of the sensors at elevated temperatures (80 °C) and/or high humidity levels (50% RH). When coupled with the potential scalability of the present approach, these results hold great promise toward fabrication of self-powered wireless devices such as facile peel-and-stick toxic gas sensors that can be used in hazardous environments. The direct-write printing approach offers promise for hybrid integration of sensors to achieve the desired sensitivity and selectivity in the application environment while maintaining low-cost and high-volume manufacturability attributes. The additive integration of multiple sensors on the same substrate can lead to an efficient electronic nose platform gaining from the advances in the nanomaterials technology and emerging Internet of Things solutions. Advances in functional inks and printed hybrid electronics are key to the continued development toward a greener future where printed devices may be increasingly deployed and perhaps even become the dominant sensor technology.