Introduction

Portable biomarker detection has been widely demonstrated in various applications such as disease diagnosis and healthy monitoring1,2,3,4,5,6,7,8. The implementation of portable biomarker devices enables deeper profiling of individuals by extending the analytical platform of chemical analytes from restricted sites to expansive environments, including hospitals and homes. Besides, portable biomarker devices allow rapid and real-time monitoring of chemical analytes. The detection of ammonia (NH3) is of great value among various chemical analytes of interest. For instance, kidney diseases-related patients have an abnormal NH3 concentration in their exhaled gas (healthy: <1 ppm; sick: >1.5 ppm; end-stage renal disease (ESRD): >4.88 ppm), which implies a non-invasive and self-operating diagnosis for diseases screening9,10. Under the action of microbial decomposition, the foods stored under natural conditions can release various total volatile basic nitrogen (TVB-N) such as NH3 and triethylamine which can be used for characterizing the food freshness (shrimp: <20 mg/100 g; chicken: <15 mg/100 g; beef: <15 mg/100 g)11,12,13. Implementing sensing devices on the food delivery chain, such as production factories or food packages could assist personalized quantification of food spoilage, preventing foodborne diseases, and eliminating food waste. Meanwhile, hazardous leakage in industrial occasions is of great concern due to its adverse effects on human life (acceptable NH3 exposure limit: 25 ppm within 8 h) and eco-system14. In the industry, NH3 is often transported through sinuous and large-scale pipelines, in which the macroscopical leakage originates from inconspicuous microcracks. Those failures are arduous to be observed with the naked eye and commonly-used fixed alarm systems. In this regard, extensively and actively profiling the imprinting of trace NH3 around the microcracks provides an in-time and accurate detection of NH3 leakage, avoiding potential safety hazards. Unfortunately, the need for real-time and accurate NH3 detection in disease diagnosis, food freshness evaluation, and hazardous leakage warning is unmet. Despite traditional gas chromatography-mass spectrometry instruments providing accurate identification of NH3, they not only have large volume, high costs, and long start-up time, but also require specialized professional personnel, limiting their use in the aforementioned occasions15,16. Fluorescent devices for NH3 sensing have favorable mobility, whereas they only provide qualitative information about NH3 by judging the color change of the device11,17. Furthermore, the chemiresistive devices often rely on specialized external wired equipment for data acquisition, which limits their practical deployment for on-site applications18,19. To make matters worse, these devices are often subject to humidity interference, which severely degrades the detection accuracy20. In spite of some present humidity-compensating methods such as extra humidity sensor compensation, they can increase the complexity and volume of the measurement system21,22. Hence, it highlights the necessity to develop a device that can perform real-time and accurate NH3 biomarker detection in arbitrary occasions and locations, free from behavioral control, and without access to specialized facilities and personnel.

Recently, wireless passive devices have been emerged as a group of promising tools for real-time measurement, which dispense wiring processes for supplying energy and transmitting transduced signals23,24,25. Especially, inductor-capacitor (LC) devices have been extensively studied in chemical sensing as they can synchronously perform multiple functions such as energy supply, signal recognition and transduction, and signal transmission functions26,27. However, due to the intrinsic working mechanism, the operating voltage in the LC device is periodically variable, which fails to maintain the stable electric field for effectively migrating charge transfer in chemical sensing28. In addition, LC sensor has a greater background noise than wired devices, which can easily submerge weak transduced signals in low-concentration NH3 detection29. As a result, the sensitivity of the LC gas device is significantly limited. Previously-reported LC NH3 sensors show a part per million (ppm) level detection limit much lower than the parts per billion (ppb) level of wired devices, which hampers their practical use in real-world settings30,31. The key to achieve ultra-low limit gas detection is to develop NH3 sensing materials with great adsorption-charge transfer dynamics, thereby offsetting the inherent drawback of LC devices. On the other hand, the conductivity of the sensing material plays an important role in the impedance properties of the LC device, further determining radio frequency performance and upper detection limit of LC NH3 sensor. The present sensitization strategies including heterojunction and Schottky junction designs cannot balance the requirements in the adsorption-charge transfer dynamics, conductivity, and semiconductor type of sensing materials, leading to the confinement in signal transmission and upper detection limit (several hundred ppm) of wireless NH3 sensing31,32,33,34.

Humidity plays the main role in various interfering factors of NH3 sensing, demanding for an effective humidity calibration to ensure accurate NH3 sensing. LC technology allows for multi-parameter measurement in such a way that it wirelessly monitors multiple LC sensors with one antenna35,36,37. This operating method may provide a feasible solution to achieve humidity compensation, for example, by adding an extra LC humidity sensor near the LC NH3 sensor. However, this scheme suffered from complex integrated structures and intractable crosstalk, leading to complex manufacturing processes and unsatisfactory measurement accuracy. Interestingly, one LC device can simultaneously output multiple transducing parameters such as frequency (f) and return loss (S11). Making use of these parameters for transducing humidity and NH3 might be a promising method for accurate NH3 sensing. This measurement strategy could largely simplify the device design and fabrication process. However, there have been no reports demonstrating this capability because most sensing materials lack ideal specificity in multi-parameter transducing.

In this work, a LC wireless device based on platinum-doped partially deprotonated-polypyrrole (Pt-PPy+ and PPy0) for real-time and accurate NH3 measurement is demonstrated. The impedance of the LC device is effectively tuned by directly adjusting the protonation doping of PPy, thus greatly improving the S11 of LC sensor. In order to achieve an ultra-wide detection range of wireless NH3 sensing, we further consider to synchronously improve the adsorption energy and charge transfer of the sensing material toward NH3 and maintain the intrinsic semiconductor type of the sensing material by a transition metal-doping method. As a result, the resultant device shows a detection range of 125 ppb to 2000 ppm. By virtue of the multi-parameter output characteristic of the LC device and the sensing specificity of Pt-PPy+ and PPy0 to NH3 and humidity, the wireless NH3 sensing with self-humidity compensation has been achieved. The portable system combined with our device and a handheld instrument was demonstrated in various applications, indicating a significant tool for individually rationalizing strategies toward those applications.

Results

Materials and device design

Figure 1a is the overall design of the Pt-PPy+ and PPy0-based LC sensor for accurate and real-time NH3 monitoring. This design includes a square inductor (Ls), an interdigital electrodes-based capacitance (Cs), and a Pt-PPy+ and PPy0-based chemiresistor (Rs) on one side of the flexible polyimide substrate (thickness: ~12.5 μm) (See detailed device design in Supplementary Fig. 1 and Supplementary Table 1). The working principle of Pt-PPy+ and PPy0-based LC sensor relies on the electromagnetic coupling effect between the interrogating antenna and the Ls of the device to wirelessly perform the energy supply and transducing signal reception (Supplementary Fig. 2). The wireless NH3 transducing mechanism of Pt-PPy+ and PPy0-based LC sensor can be explained that the conductivity of the Pt-PPy+ and PPy0 varies upon the exposure to NH3, which results in the change of the Rs, further leading to the shift in the S11 of the Pt-PPy+ and PPy0-based LC sensor (Supplementary Note 1)38,39. Once the S11-f curve of the device is wirelessly read by the antenna, the NH3 concentration can be readily quantified. Here we define the S11 of the Pt-PPy+ and PPy0-based LC sensor in the air and in the NH3 as S11air and \(S11_{\tiny{\displaystyle{\rm{NH}}_{3}}}\).

Fig. 1: Materials and device design.
figure 1

a Schematic illustration showing the overall design of the Pt-PPy+ and PPy0-based LC sensor for diagnosis detection, food freshness qualification, and hazardous leakage alarm. b The design route of Pt-PPy+ and PPy0 sensing material for wireless NH3 sensing. c The energy gaps of PPy with different oxidation degrees: the band gaps of PPy significantly broaden as the decrease of the oxidation state. d The comparison of the adsorption energy (ΔEads) and charge transfer (Q) of different sensing materials toward NH3, including PPy-based sensing materials with different oxidation degrees, transition metal-doped PPy+ and PPy0, and previously-reported sensing materials. e The charge densities of different PPy-based sensing materials to NH3 (PPy-based sensing materials with different oxidation degrees and transition metal-doped PPy+ and PPy0). E is the electric field between interdigital electrodes. Source data are provided as a Source Data file.

Full size image

To achieve high-performance wireless NH3 sensing, we designed a Pt-PPy+ and PPy0 (Fig. 1b). As the S11air is the quality factor for wireless transmission signal that determines the wirelessly-identified range between the device and antenna, we consider to directly improve the S11air of the LC sensor by chemically tuning the conductivity of the Pt-PPy+ and PPy0 sensing material. More specifically, we engineered the oxidation state of the PPy to adjust the Rs in LC sensor, thereby improving the S11air. This scheme is theoretically verified by the density functional theory (DFT) calculations (Supplementary Fig. 3 and Fig. 1c). With the hole concentration decreases, the band gaps of PPy exhibit a remarkable increase from 0.04 eV to 2.71 eV, which indicates the widely-tuning characteristic in the conductivity. Thus, the theoretical result supports a prospective method in the S11air regulation of the LC sensor by chemically tuning the conductivity of the Pt-PPy+ and PPy0.

Despite the significant promotion in the S11air of LC sensor, the adsorption energy and charge transfer of PPy to NH3 show distinct attenuation upon exposure to NH3 as the oxidation state of PPy decreases (from PPy+ to PPy0) (Fig. 1d, e), limiting the sensitivity of the wireless NH3 sensing. To enhance the sensitivity of NH3 sensing, we utilized the transition metal doping to modify the partially deprotonated PPy (PPy+ and PPy0) which constructed transition metal-doped PPy+ and PPy0 hybrids, thereby improving the adsorption behavior and charge transfer of the PPy+ and PPy0 toward NH340. In the study, three transition metals of copper (Cu), palladium (Pd), and Pt were selected to study ΔEads and Q of those transition metal-doped PPy+ and PPy0 hybrids to NH3. The DFT calculations show that Pt-PPy+ and PPy0 hybrid has the highest ΔEads of −3.99 eV and the highest Q value of −0.4 e, showing remarkable sensitivity to NH3. Notably, the ΔEads and Q of Pt-PPy+ and PPy0 are not only much higher than those of the PPy+ but also far exceed those of previously-reported NH3 sensing materials, which promises an ultra-low detection limit of Pt-PPy+ and PPy0-based LC NH3 sensor41,42,43,44,45,46,47. Moreover, the Pt-PPy+ and PPy0 have a p-type semiconductor behavior similar to that of PPy+, which could exhibit a decreased conductivity upon exposure to NH3 (p-type response). As a result, the Pt-PPy+ and PPy0-based LC sensor can show a positive S11 shift (|S11air|<|\(S11_{\tiny{\displaystyle{\rm{NH}}_{3}}}\)|) in NH3, thereby achieving high-concentration NH3 detection without signal transmission interruption. To achieve accurate NH3 sensing, the f of Pt-PPy+ and PPy0-based LC sensor was arranged as the humidity-transducing parameter, providing reliable calibration for NH3 sensing. The humidity sensing mechanism of the Pt-PPy+ and PPy0 could be explained that the Pt-PPy+ and PPy0 upon exposure to water molecules could induce proton conduction, further changing the permittivity of the Pt-PPy+ and PPy0. As a result, the Cs of Pt-PPy+ and PPy0-based LC sensor will be changed, which leads to the shift of f (Supplementary Note 1).

Materials characterization

To synthesize the Pt-PPy+ and PPy0 sensing materials, we first obtained PPy+ by chemical oxidative polymerization with Fe3+ which has a high redox potential (Fig. 2a). After that, the BH4- was used to simultaneously reduce Pt4+ and adjust the oxidation state of the PPy+, further yielding the Pt-PPy+ and PPy0 nanohybrids. The Pt-PPy+ and PPy0 nanohybrids with different mass ratios of Pt and PPy+ and PPy0 are denoted as PPy+, 10 Pt-PPy+ and PPy0 and 30 Pt-PPy+ and PPy0. Supplementary Fig. 4 is the scanning electron microscope (SEM) images and energy dispersive spectrometer (EDS) results of PPy+, 10 Pt-PPy+ and PPy0, and 30 Pt-PPy+ and PPy0. In all samples, the PPy emerges tightly-bound nanosheet (NS) morphologies with high surface volume ratios, providing both abundant active sites and fast diffusion pathways for NH3. The size of Pt nanoparticles (NPs) magnifies as the Pt doping increases. Transmission electron microscope (TEM) images of PPy+ and 10 Pt-PPy+ and PPy0 (Fig. 2b, c) and the elemental mappings of 10 Pt-PPy+ and PPy0 nanohybrid (Fig. 2d) indicate that Pt NPs are uniformly loaded on PPy NSs, producing a 0D-2D nanohybrid structure. Such the 0D-2D structure can improve the charge transfer efficiency and make full use of the high adsorption energy and significant charge transfer characteristics of Pt-PPy+ and PPy0. The Pt NPs have a particle size of ~15 nm and the lattice fringe Pt (111) could be clearly identified (Fig. 2e). The selected area electron diffraction pattern indicates the single crystal structure of Pt NPs (Fig. 2f).

Fig. 2: Materials characterization.
figure 2

a Schematic diagram illustrating the synthesis of the Pt-PPy+ and PPy0. b The TEM image of the PPy+ NSs. c TEM image of the 10 Pt-PPy+ and PPy0 nanohybrid, showing the Pt NPs loaded on PPy NSs form 0D-2D nanohybrid structures. d The elemental mappings of 10 Pt-PPy+ and PPy0 nanohybrid. e High-resolution TEM image of 10 Pt-PPy+ and PPy0 nanohybrid. f The selected area electron diffraction pattern of 10 Pt-PPy+ and PPy0 nanohybrid. g XPS spectra of the N 1s region of PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid. h XPS spectra of the C 1s region of the PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid. i XPS spectra of the Pt 4f region of the 10 Pt-PPy+ and PPy0 nanohybrid. j The FTIR spectra of PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid. k Raman spectra of PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid at peak 1053.6 cm−1. l The conductivity result of PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid. Source data are provided as a Source Data file.

Full size image

X-ray photoelectron spectroscopy (XPS) provides deep analysis for the surface atomic states of PPy and PPy-based nanohybrids. In Supplementary Fig. 5, the C and N peaks attributed to PPy rings of PPy+ NSs, 10 Pt-PPy+ and PPy0 nanohybrid and 30 Pt-PPy+ and PPy0 nanohybrid can be clearly identified. Moreover, the characteristic peaks of Pt NPs can be recognized in the total spectrum of 10 Pt-PPy+ and PPy0 nanohybrid and 30 Pt-PPy+ and PPy0 nanohybrid. In Fig. 2g, we deconvoluted the N 1s spectrum of PPy+ and 10 Pt-PPy+ and PPy0 into four peaks (C=N+, C-N+, N-H, and C=N). Compared with PPy+, The binding energy of N 1s of 10 Pt-PPy+ and PPy0 exhibit a negative shift of 0.27 eV, proofing the possible charge transfer between the Pt atom and the N atom on PPy NSs. The possible charge transfer between Pt NPs and PPy NSs changes electron conjugation in the PPy ring as verified by the increased [=N−]/[−N−] ratio of 10 Pt-PPy+ and PPy048. Moreover, it can be observed that PPy+ has a quite weak peak of C=N, which indicates the high protonation degree of PPy+. However, due to strong interactions between the electrons in BH4- and the polarons and bipolarons in PPy+, the C=N peak in Pt-PPy+ and PPy0 is significantly enhanced, implying the obvious decrease in the carrier concentration. Further, the C 1s fine spectrum of PPy+ NSs and 10 Pt-PPy+ and PPy0 nanohybrid can be divided into five peaks, including 288 eV (N-C=O/O-C=O), 286.7 eV (C=N/C=O), 285.9 eV (C-N/C-O), 284. 8 eV (C-sp3) and 284.1 eV (C-sp2), which are consistent with previous report (Fig. 2h)49. The 10 Pt-PPy+ and PPy0 samples show a positive shift in the binding energy of C 1s compared with PPy+, indicating the strong charge transfer between Pt NPs and C atoms. It is noteworthy that the charge transfer between Pt NPs and the C atom increases the C-sp2/C-sp3 ratio, indicating the high catalytic activity of 10 Pt-PPy+ and PPy0 nanohybrid50. Further, the Pt 4f spectrum demonstrates the co-existence of Pt0 and Pt2+ (Fig. 2i). Specifically, peaks at 71.2 eV and 74.5 eV are attributed to Pt0, while 73.9 eV and 76.4 eV are attributed to Pt2+51. The presence of Pt2+ is attributed to the possible charge transfer between Pt NPs and PPy NSs. The co-existence of Pt0 and Pt2+ can further enhance the charge transfer of Pt052. The Fourier transform infrared spectrometer (FTIR) results indicate all the characteristic peaks of PPy could be observed in PPy+ and 10 Pt-PPy+ and PPy0 samples (Fig. 2j, Supplementary Table 2)53. In comparison with PPy+, 10 Pt-PPy+ and PPy0 exhibit a red shift at the peaks of 904.5 cm−1 and 1546.6 cm−1, indicating reduced polaron concentrations. In addition, 10 Pt-PPy+ and PPy0 show significant blue shifts at the peaks of 794.5 cm−1 and 3442.3 cm−1, demonstrating the strong chemical interaction formed between Pt NPs and PPy NSs. The Raman spectra of PPy+ and Pt-PPy+ and PPy0 are surveyed (Supplementary Fig. 6 and Table 3). As shown in Fig. 2k, the peak of Pt-PPy+ and PPy0 at 1053 cm−1 exhibits a significant blue shift, proving an obvious decrease in charge carrier concentration, which is consistent with the observation of XPS and FTIR. The electrical conductivity of PPy+ is 5.1 times higher than that of 10 Pt-PPy+ and PPy0, demonstrating a significant decrease in the conductivity of Pt-PPy+ and PPy0 (Fig. 2l). Other necessary characterizations are included in Supplementary Fig. 7 and Note 2.

Wireless transmission capability of the Pt-PPy+ and PPy0 LC sensor

Figure 3a is the optical image of the Pt-PPy+ and PPy0 LC sensor fabricated with flexible printed circuit board process and spraying technique (Supplementary Fig. 8). The enlarged optical image clearly shows physical contact between the interdigital electrode and Pt-PPy+ and PPy0, demonstrating the success insertion of Pt-PPy+ and PPy0-based chemiresistor into the flexible LC resonator. The sensing film has a thickness of 56.1 μm (Supplementary Fig. 9).

Fig. 3: Wireless transmission capability of the Pt-PPy+ and PPy0 LC sensor.
figure 3

a Optical image of the Pt-PPy+ and PPy0-based LC sensor. The enlarged optical image of the IDEs on the device exhibits the successful insertion of Pt-PPy+ and PPy0-based chemiresistors into the device. b The S11air values of the different Pt-PPy+ and PPy0-based LC sensors. c The f values of different Pt-PPy+ and PPy0-based LC sensors. d Comparison of this work and those previous reports in the S11 enhancement capability (Methods 1, 2, and 3 are substrate substitution, geometrical structure optimization, and bias voltage regulation, respectively). e Comparison of PPy+- and 10 Pt-PPy+ and PPy0-based LC sensors in RL. The inset is the schematic diagram showing the difference in RL. f Comparison of PPy+- and 10 Pt-PPy+ and PPy0-based LC sensors in dL. The inset is the schematic diagram showing the difference in dL. IDE is the abbreviation of the interdigital electrode. Error bar: the standard deviation calculated from five independent tests. Method 1: substrate substitution; Method 2: structure optimization; Method 3: bias voltage regulation. Source data are provided as a Source Data file.

Full size image

To begin with, we surveyed the S11-f curves of LC sensors based on different Pt-PPy+ and PPy0 in air (Supplementary Fig. 10). Figure 3b shows the S11air of the 10 Pt-PPy+ and PPy0-based LC sensor has the highest S11air of −5.6 dB. Since the LC sensors based on different Pt-PPy+ and PPy0 exhibit negligible variations (~2.9%) in f (Fig. 3c), the Cs of LC sensors based on different Pt-PPy+ and PPy0 are close to each other. Thus, it can be concluded that the significant S11 increase of Pt-PPy+ and PPy0-based LC sensor is radically attributed to the improved Rs of the LC sensor, which is realized by directly adjusting the conductivity of Pt-PPy+ and PPy0. Moreover, the stagnated growth of the S11air can be explained that the continuous addition of BH4- could improve the mass ratio of Pt NPs in Pt-PPy+ and PPy0, thus reducing the conductivity. Notably, the S11air of 10 Pt-PPy+ and PPy0-based device shows a 180% increase compared with that of PPy+-based device. Our strategy has an unprecedented record compared with the conventional methods, which not only have limited efficiency (10–50%) but also require trade-offs in fabrication, cost, power consumption, and volume (Fig. 3d)54,55,56,57,58,59,60,61,62,63.

The 10 Pt-PPy+ and PPy0-based LC sensor with a high S11air indicates the effective energy supply and transducing signal transmission over a wide operating range. In order to study the operating limit of the device, we evaluated the S11air at different bending radius (R) and different working distances (d) (Supplementary Figs. 11, 12 and 13). We show that the 10 Pt-PPy+ and PPy0-based LC sensor has a detectable bending limit (RL) of 5 mm, which is almost 5 times lower than that of PPy+-based device (Fig. 3e). Not only that, the 10 Pt-PPy+ and PPy0-based LC sensor shows a significantly-improved detectable distance limit (dL) of 21.5 mm, while the dL of PPy+-based device is only 5.5 mm (Fig. 3f). In addition, it is noteworthy that the 10 Pt-PPy+ and PPy0-based LC sensor has a f below 500 MHz, which allows portable measurement with a mini network analyzer.

Sensing performances of the Pt-PPy+ and PPy0 LC sensor

To optimize the gas sensitivity, we initially surveyed the gas responses of LC sensors with different Pt-PPy+ and PPy0 towards 100 ppm NH3 in a custom-made gas sensing system (Supplementary Figs. 14 and 15). The gas response characterized with S11 is defined as: \({{{\rm{Response}}}}=({S11}_{{{{\rm{gas}}}}}-{S11}_{{{{\rm{air}}}}})/{S11}_{{{{\rm{air}}}}}\). In Fig. 4a, we observed that 10 Pt-PPy+ and PPy0-based device has the highest response (0.195), after which the responses declined. In order to clarify the mechanism of the sensitivity evolution, the f responses of LC sensors with different Pt-PPy+ and PPy0 upon exposure to 100 ppm NH3 were studied (Supplementary Fig. 16). As shown in Fig. 4b, all devices show negligible changes in f, implying the Cs of the devices are insensitive to NH3. In other words, the NH3-S11 transducing mechanism of Pt-PPy+ and PPy0-based LC sensor is mainly dependent on chemically-resistive variation induced by the adsorption-charge transfer dynamics between NH3 and Pt-PPy+ and PPy0. To further demonstrate this speculation, we surveyed the Raman spectra of 10 Pt-PPy+ and PPy0 before and after exposure to NH3 (Fig. 4c). The C-C peak at 931.5 cm−1 which is related to the bipolaron state in PPy ring disappeared in the presence of NH3, suggesting the strong interactions between NH3 and Pt-PPy+ and PPy0. That is, the electrons were injected into Pt-PPy+ and PPy0 from NH3, leading to a decrease in the hole concentration, which is consistent with the DFT results64. The highest response of 10 Pt-PPy+ and PPy0-based device toward 100 ppm NH3 is ~16.2 times that of PPy+-based device, experimentally indicating that the Pt NPs play an important role in boosting the adsorption-charge transfer dynamics, thereby improving the sensitivity of wireless NH3 sensing. Moreover, the response attenuation in the latter stage of optimization can be explained that both the number of Pt-PPy contacts (Supplementary Fig. 4e) and the hole concentration in PPy significantly decreases as the mass ratio of BH4- improves, synergistically limiting the adsorption efficiency and charge transfer of gas sensing materials and NH3.

Fig. 4: Sensing performances of the Pt-PPy+ and PPy0 LC sensor.
figure 4

a The sensing responses of LC sensors with different sensing materials in the presence of 100 ppm NH3. b The variation percents of the f of the LC sensors with different sensing materials in the presence of 100 ppm NH3. c In situ Raman spectra of 10 Pt-PPy+ and PPy0 nanohybrid in a cycle of air-NH3-air exposure. The arrow illustrates the electron doping–dedoping process. d The resistance response of PPy+- and 10 Pt-PPy+ and PPy0-based resistors in the presence of NH3. e The S11-f curves of 10 Pt-PPy+ and PPy0-based LC sensors in air and in 100 ppm NH3. f The dynamic response curves of 10 Pt-PPy+ and PPy0-based LC sensors in 125 ppb−100 ppm. g The dynamic response curves of 10 Pt-PPy+ and PPy0-based LC sensors in 200 ppm−2000 ppm. h Sensing range comparison of our work and those reported wireless passive NH3 sensors. i The tres of PPy+- and 10 Pt-PPy+ and PPy0-based LC sensors in the presence of 2000 ppm NH3. j The tres comparison of our work and previously-reported NH3 sensors. Error bar: the standard deviation calculated with five independent tests. Source data are provided as a Source Data file.

Full size image

To clarify the semiconductor behavior of 10 Pt-PPy+ and PPy0, we then compared the resistance variation curves of 10 Pt-PPy+ and PPy0- and PPy+-based resistors in NH3 environment (Fig. 4d). The 10 Pt-PPy+ and PPy0-based resistor show similar response trend with PPy+-based resistor, demonstrating the p-type semiconductor of 10 Pt-PPy+ and PPy0. The S11 of 10 Pt-PPy+ and PPy0-based LC sensor exhibits a positive shift in the presence of NH3 as shown in Fig. 4e, which proves that 10 Pt-PPy+ and PPy0-based LC sensor could be used for high-concentration NH3 detection.

To eventually determine the sensing range of 10 Pt-PPy+ and PPy0-based sensors, we surveyed the dynamic responses of the device towards different NH3 concentrations (Fig. 4f, g). The device shows monotonically increasing responses as the NH3 concentration improves, indicating the dosimetric behavior of the device. Owing to the enhanced sensitivity, the 10 Pt-PPy+ and PPy0-based device has a much lower detection limit of 125 ppb than that of PPy+-based device (10 ppm) (Supplementary Fig. 17). In addition to the significant charge transfer and the high adsorption energy of the sensing materials toward NH3, this significant detection limit can also be attributed to the lower background noise level than that of PPy+-based device (Supplementary Fig. 18). The comparison of the previously reported reports and our work indicates 10 Pt-PPy+ and PPy0-based device has unprecedented NH3 detection range of 125 ppb to 2000 ppm (Fig. 4h)27,30,31,65,66,67,68,69,70,71,72,73, highlighting the great potential in real-world applications.

Benefited from the remarkable properties in adsorption-charge transfer dynamics, 10 Pt-PPy+ and PPy0-based LC sensor can also achieve a quick response equilibrium with a rapid response time (tres) of 19 s at an ultra-high concentration of NH3, which is much lower than that of the PPy+-based device (Fig. 4i). Our sensor shows a significant advance in tres over reported ammonia sensors so far (Fig. 4j)46,74,75,76,77,78,79,80,81,82. In addition, 10 Pt-PPy+ and PPy0-based LC sensor has fully-recovery characteristics at 125 ppb–2000 ppm NH3, indicating the reversible adsorption-charge transfer dynamics of the sensing material. This is demonstrated by the Raman result in Fig. 4c, in which the 931.5 cm−1 peak of 10 Pt-PPy+ and PPy0 reoccurs in the absence of NH3, suggesting that the holes underwent a reversible transformation before and after NH3 exposure.

Self-humidity compensation of the Pt-PPy+ and PPy0 LC sensor

In order to explore the influence of humidity on the wireless NH3 sensing, the response curves of the 10 Pt-PPy+ and PPy0 LC sensor at 100 ppm, 30–90% RH were measured (Supplementary Fig. 19). In Fig. 5a, b, we can see that the S11air and the responses of the sensor decrease as humidity increases, indicating the dampening effect of humidity. In Fig. 5c, the dampening effect of humidity can be explained that the H2O absorbed by 10 Pt-PPy+ and PPy0 could build an additional conductive path (Path 2) besides the intrinsic conductive path of 10 Pt-PPy+ and PPy0 (Path 1). This humidity-induced path enables effective charge transfer between non-adjacent PPy NSs, decreasing the conductivity of 10 Pt-PPy+ and PPy0. In the meanwhile, the H2O molecules could also cause significant changes in the dielectric constant of 10 Pt-PPy+ and PPy0 owing to the high dielectric constant (80) of the H2O. In other words, it is equivalent to parallel a resistance (Rs(humidity)) and a capacitance (Cs(humidity)) on the basis of the original LC circuit, which synergistically decreases the Rs and increases the Cs of the LC sensor, thereafter decreasing the S11air and the responses of the sensor.

Fig. 5: Self-humidity compensation of the device.
figure 5

a The sensing responses of 10 Pt-PPy+ and PPy0-based LC sensors at different RHs. b The S11air of 10 Pt-PPy+ and PPy0-based LC sensor at different RHs. c The transducing mechanism of 10 Pt-PPy+ and PPy0-based LC sensor in NH3 atmosphere and in NH3+humidity atmosphere. d The f of 10 Pt-PPy+ and PPy0-based LC sensor at 30–90% RH: the inset indicates the linear f response of the device toward humidity: f = 517.7 − 0.5 × humidity) e The measuring results of the 10 Pt-PPy+ and PPy0-based LC sensors before and after humidity calibration at 100 ppm NH3. f Selectivity of the 10 Pt-PPy+ and PPy0-based LC sensor. The response scale on the right corresponds 1:1 with the radial axis of the plot. g The response curves of the 10 Pt-PPy+ and PPy0-based LC sensor before and after 200 bending times (R = 5 mm) at 100 ppm. Error bar: the standard deviation calculated with five independent tests. Source data are provided as a Source Data file.

Full size image

Because the H2O molecules can change the dielectric constant of 10 Pt-PPy+ and PPy0, the f of the 10 Pt-PPy+ and PPy0-based LC sensor could exhibit responsive sensitivity to humidity. We investigated the f responses of the device in the humidity range of 30–90% RH (Fig. 5d). The device has a highly-linear f response curve which yields a high sensitivity of 0.5 MHz/% RH (f = 517.7 − 0.5 × humidity). Since NH3 does not affect the f of the device (Supplementary Fig. 16e), such a high sensitivity is fully induced by the humidity transducing. To conclude, the different S11 and f response behaviors of the 10 Pt-PPy+ and PPy0-based LC sensor could be used for realizing a discriminative detection of NH3 and humidity. More importantly, it presents a promising strategy to address the concern of humidity compensation in NH3 measurement. To verify this possibility, we assessed the differences in concentration before and after self-compensation (90% RH, 100 ppm). The after-compensation result shows a weak difference (only a few ppm) compared with the before-compensation concentration (Fig. 5e). This result suggests that our device can perform accurate NH3 measurement without the additional configuration of humidity sensor, which greatly simplifies the device design and improves the mobility of the measurement system. Further tests suggest the bending and detection limits show negligible changes at different humidities as the humidity has little effect on the S11air (Supplementary Fig. 20).

We analyzed the responses of the 10 Pt-PPy+ and PPy0-based LC sensor toward 15 interfering gases at a concentration of 100 ppm (Supplementary Fig. 21). The device shows obvious responses toward biogenic amines (triethylamine, diethanolamine and trimethylamine) whereas it shows negligible responses to the rest of gases (Fig. 5f). Nevertheless, the response of the device toward NH3 is much higher than those of tested biogenic amines. NH3 and biogenic amines are known as TVB-N chemicals, the detection of which is a gold standard for food spoilage evaluation11,12,13. Moreover, the S11-f curves and response curves of the device show weak variations before and after 200 bending times (Fig. 5g and Supplementary Fig. 22), demonstrating the favorable robustness of the device. The bending effect on the response may be illustrated that the bending process could lead to mechanical deformation of the gas sensing, decreasing the conductive efficiency of the sensing film. By exposing the sensor to 100 ppm and 2000 ppm NH3 five times, we show slight fluctuations in sensor responses, demonstrating high repeatability (Supplementary Fig. 23). Moreover, the device has good stability as it exhibits weak shifts in the responses of 100 ppm and 2000 ppm NH3 after 40 days in the atmosphere (Supplementary Fig. 24).

Demonstrations of the sensor for real-time chemical sensing

As shown in Fig. 6a, we envision our device could be integrated with the food package as a smart package to perform real-time monitoring of meat freshness at less limited occasions, avoiding foodborne illness and reducing food waste. This portable measurement scheme bridges the gap between tested and real TVB-N concentrations released during meat spoilage, allowing consumers to accurately determine the edible state of meat instead of unrealistic shelf life on the label. To demonstrate the use of our device in instant meat freshness evaluation, we built a wireless device-based smart package, which is combined with a mini network analyzer for measurement (125 × 75 mm2) (Supplementary Fig. 25). We used this portable system to continuously track the responses of the device to TVB-N in the smart package which contains different meat (shrimp, beef, and chicken) (Supplementary Figs. 26 and 27). Meanwhile, we used laboratory instruments (LD-SP01, LANENDE, China) to analyze the real TVB-N concentrations during the spoilage process of these meat (Supplementary Table 4). Then, we established the linear relationships between the real TVB-N and the responses of our device toward different meat (Fig. 6b–d). To evaluate the sensing performance of our device in actual food freshness, we tested the TVB-N emission of shrimp, beef, and chicken at 7 and 20 °C (storage time: 12 h) with our device and the laboratory instrument (Fig. 6e). The results show that the TVB-N concentrations detected by our device over four types of meat were very close to those detected by the laboratory instrument. Due to the high protein content, shrimp releases a higher concentration of TVB-N than those of beef and chicken. Moreover, meat stored at 7 °C produces less TVB-N concentrations than those stored at 20 °C. This may be illustrated that bacterial proliferation is inhibited at lower temperature, which further limits protein degradation. Thus, it can be inferred that our portable system can be universally applied to meat freshness evaluation of the various types of meat in different storage environments, indicating a promising platform for real-time monitoring of meat freshness.

Fig. 6: Demonstrations of the device for various scenarios.
figure 6

a Schematic diagram of instant detection of food freshness, point-of-care (POC) diagnosis, and early warning of chemical leakages with Pt-PPy+ and PPy0-based LC sensors. b TVB-N concentration calibration for shrimp stored at 7 °C and 20 °C within 72 h. c TVB-N concentration calibration for beef stored at 7 °C and 20 °C within 72 h. d TVB-N concentration calibration for chicken stored at 7 °C and 20 °C within 72 h. e Comparison of TVB-N analysis with our device and laboratory instrument. f Detection of exhaled NH3 concentrations in human breath: gray histograms represent the real concentrations of NH3 in human breath, while green, red and purple histograms represent the tested concentrations of NH3 in human breath. g The analysis duration of exhaled gases with different NH3 concentrations by our sensor. h The measurement principle of NH3 leakage detection by our sensor. i The leakage orientation capability of our device at an ultra-low concentration NH3 leakage. j The leak frequency detection of our device at an ultra-low concentration NH3 leakage. Error bar: the standard deviation calculated with five independent tests. Re is the abbreviation of the response of the device, r is the radiation radius and c is the concentration of NH3. Source data are provided as a Source Data file.

Full size image

Our device could also be used for portably examining the exhaled gases with a handheld vector network analyzer, aiding the risk and treatment evaluation of related diseases (Fig. 6a). We developed a POC tester that integrates our device and a 0.25 L chamber (Supplementary Fig. 28). The POC tester has excellent mobility and permits the individual operation, implying a promising tool in the POC diagnosis. To demonstrate the detection capability in the exhaled gas detection, we studied the response curves of the POC tester to several simulated exhaled gases with different NH3 concentrations (1000–5000 ppb) (Supplementary Fig. 29). The results reveal that the NH3 concentrations detected by the device were almost close to the standard concentrations, in which the maximum deviation (500 ppb) occurs at 3000 ppb (Fig. 6f). Moreover, our device has a detection limit of 1000 ppb. This evidence demonstrates that our device can accurately detect the NH3 concentrations in the exhaled gas and distinguish sick and healthy individuals. Interestingly, the detection time of the device is shorter than 76 s, indicating the ability for rapid diagnosis (Fig. 6g). Therefore, POC tester is completely different from the diagnostic methods currently used by medical departments, which often rely on large medical equipments.

NH3 being as an important industrial product is commonly transported through widely-distributed pipelines. To achieve the early leakage monitoring of NH3, the detection system is required for wide-range detection, fast response, and favorable mobility. Meanwhile, the leak frequency which acts as a significant index for evaluating crack expansion is also demanded. We developed a leak guard based on our device which detects hazardous leakage with such a way that monitors the featured response curves around the leakage point, providing timely reference for security measures (Fig. 6a). Specifically, the concentration of NH3 around the leak point changes as the radiation radius creeps, which results in the difference response curves of the device (Fig. 6h). Therefore, the leak point of NH3 can be determined through exploring the maximum response curve of the device. We simulated early pipeline leakage with a custom-made system and evaluated the performances of our device in early NH3 leak warning (Supplementary Fig. 30). By recording the featured response curves of the device along one axis in the coordinate system, we show that our device can successfully track the location of NH3 water down to 0.1 µL, highlighting the ultra-low concentration detection capability of NH3 leakage (Fig. 6i). Moreover, our experimental results show that the device can detect the leak frequency of a 0.1 µL drip in a wide range (0.3–1.5 Hz) (Fig. 6j). That evidence proves that our device can trace extraordinarily-early NH3 leakage, providing very timely and valuable intelligence to guide effective security measures. We envision the device can enable a wireless sensing network for wide-range leakage detection in the pipeline facility.

Discussion

In summary, we have proposed a wireless LC chemical sensor based on Pt-PPy+ and PPy0 for real-time and accurate NH3 detection. We directly adjust the impedance of the LC sensor through chemically tuning the conductivity of Pt-PPy+ and PPy0, which results in an up-to-180% increase in the S11. The transition metal doping-enabled Pt-PPy+ and PPy0 sensing material has remarkable adsorption-charge transfer dynamics and p-type semiconductor properties, leading to an ultra-wide detection range of wireless NH3 sensing. By incorporating the multi-parameter output characteristic of a single LC sensor and the sensing specificity of Pt-PPy+ and PPy0 to discriminatively detect humidity and NH3, we demonstrated an impressive self-humidity calibration for accurate NH3 detection in a humidity-variable environment. These favorable performances of the device inspired a portable system enabled by Pt-PPy+ and PPy0-based LC sensor, capable of food freshness monitoring, POC diagnosis, and hazardous leakage alarm. In consideration of the limited samples and testing environment, more statistics are needed to show the reliability of the aforementioned applications.

Methods

Simulations

  1. (1)

    Materials simulations: All DFT calculations were performed using Vienna ab initio simulation package, including geometrical relaxation and self-consistent field calculations. The projector-augmented wave pseudopotentials were used for interpreting the interactions between ions and electrons, while the Perdew-Burke-Ernzerh of functional within the generalized gradient approximation was used to analyze the exchange correlation interaction. The van der Waals correction of optPBE-vdw is used for the description of the weak interactions of sensing materials and gas. Plane-wave basis set (cutoff energy of 500 eV) can be used for expanding the one electron wave function. The DFT calculations were performed based on PPy with 3Py oligomers (See Supplementary Data 1). For the geometrical optimization process, a Gamma-centered 2 × 1 × 1 k-grid mesh sampling was applied in the software, while a denser 4 × 2 × 1 k-point sampling for all successive calculations. The calculations were terminated when the energy and the Hellmann-Feynman force were equal to the convergence criteria of 10−6 and 0.02 eV Å−1. The charge transfer (Q) analysis between sensing materials and gas molecules was realized by the Bader charge analysis process developed by Henkelman group. The adsorption energy of NH3 molecules on the material can be expressed as: ΔEads = \({E}_{{\rm{NH}}_{3}+material} - {E}_{{\rm{NH}}_{3}} - {E}_{{\rm{material}}}\), where ΔEads, \({E}_{{\rm{NH}}_{3}+material}\), \({E}_{{\rm{NH}}_{3}}\) and Ematerial are the adsorption energy of NH3 gas molecules on the material, the total energy of NH3 adsorbed on the sensing material, the total energy of NH3, and sensing material, respectively.

  2. (2)

    Devices simulations: we performed the electrical simulations of the device in a high-frequency structure simulator (ANSYS Electronics Desktop 2018, ANSYS, USA). The maximum number of passes and maximum ΔS were set as 6 and 0.02, respectively. The dimension of the simulation boundary was set as 300 × 300 × 300 mm3. The lumped port was designated as the excitation type.

Materials synthesis

To obtain PPy+, 25 μL of pyrrole (Aladdin, USA) and 50 mg of sodium dodecyl sulfate (Aladdin, USA) were mixed in 50 mL of deionized water (DI) with sonication for 15 min. Then ferric chloride (Aladdin, USA) solution (0.1411 g FeCl3 + 10 mL DI) was slowly added into the above solution and then magnetically stirred at 5 °C for 3 h. After that, the resultant product was rinsed with DI and ethanol (Sinopharm, China) alternately for four times and then dried in a vacuum at 50 °C for 5 h. To synthesize 10 Pt-PPy+ and PPy0 with one-step process, 10 mg of PPy+ was mixed with 10 mL of 95% ethanol (Sinopharm, China) and sonicated for 25 min. 0.625 mL of 8 mM chloroplatinic acid (Aladdin, USA) and 30 μL of 35 mg/100 mL sodium borohydride (Sinopharm, China) solution was added into the above PPy+ solution dropwise under vigorous stirring. After stirring for 13 h, the precipitation was washed with DI and 95% ethanol alternately 4 times, and dried in a vacuum for 5 h to obtain 10 Pt-PPy+ and PPy0. Similarly, 5 Pt-PPy+ and PPy0 and 30 Pt-PPy+ and PPy0 were prepared with the same manner by adjusting the volumes of chloroplatinic acid and sodium borohydride solution (5 Pt-PPy+ and PPy0: 0.3125 mL, 15 μL; 30 Pt-PPy+ and PPy0: 1.875 mL, 90 μL). The prepared Pt-PPy+ and PPy0 samples were then mixed with 95% ethanol and sonicated for 30 min. After that, the Pt-PPy+ and PPy0 solution was sprayed on the interdigital electrodes of the devices with a custom-made mask, and then dried in a vacuum for 5 h.

Device fabrication

The fabrication process of the LC resonator was assisted by Shenzhen V-Layers Printed Circuit Technology Co., Ltd. A layer of Cu seeds was first sputtered on a piece of polyimide substrate (~12.5 μm). Then, a layer of ~9 μm Cu was formed on the substrate through an electroplating process and electrodes were imaged. Further, a layer of ~2 μm nickel and a layer of ~0.025 μm gold were sequentially deposited on the Cu layer.

Characterizations

  1. (1)

    Devices: The enlarged photo of the device and the thickness of the sensing film were captured using a digital microscope (VHX-5000, KEYENCE, Japan).

  2. (2)

    Materials: SEM (Ultra Plus, Carl Zeiss, Germany) and high-resolution Transmission electron microscopy (Tecnai G2 F20, FEI, USA) were used to study the morphologies of sensing materials. X-ray diffraction (D8 ADVANCE, Bruker, Cu Kα, Germany) was used for studying the phase of sensing materials. FTIR (Nicolet 6700, THERMO FISHER, USA) was used to analyze the IR spectra of the sensing materials. Further, a Raman spectrometer (inVia, Renishaw, England) was used to obtain the vibrational properties of the sensing materials. The chemical compositions of the sensing materials were investigated using XPS (Axis UltraDLD, Kratos, Al Kα, 1486.6 eV, Japan). The conductivity property of the sensing materials was studied using a semiconductor analyzer (ST-2722, Jingge, China).

Measurements

  1. (1)

    Radio frequency performance evaluation: The measurements of S11-f curves of the device at different working distances and working radius are arranged on a customized fixture (Supplementary Fig. 11). In the test, the bending radius of the device, and the distance between the device and the antenna are changed by moving the fixture and the micrometer. The network analyzer (E5063A, Keysight, USA) can be used for recording the S11-f curves of the device.

(2) Gas sensing performance evaluation: the response of the device to gas was measured on a customized gas test system. The device is fixed inside the chamber and communicates with an external antenna. The network analyzer processes the transduced signal acquired by the antenna. The network analyzer further connects to the computer, in which the real-time response of the device is displayed in a labview software (NI, USA).

Demonstration of applications

  1. (1)

    Meat freshness evaluation: shrimp, beef, and chicken were selected to validate the portable use of our device in real-time meat freshness monitoring. First, every kind of meat was weighed for 40 g and prepared for two portions which were stored in 20 °C and 7 °C, respectively. The portable system includes a Mini network analyzer (Nano VNA-V2, SYSJOINT, China) and a smart package (~3 L airtight crisper with our device). For the meat freshness test, the meat was put into the crisper and tightened for monitoring the NH3 concentration released by spoiled meat every 24 h through our device. The responses of the device were interrogated by the antenna out of smart package. All meat is placed in a ~ 0.5 L open box for transferring. We used laboratory instruments (LD-SP01, LANENDE, China) to analyze the real TVB-N concentrations during spoilage of three types of meat (weight for every test: 0.5 g).

  2. (2)

    Exhaled gas evaluation: the measurement system includes a POC tester (a 0.25 L chamber attached to our device) and a Mini network analyzer. The POC tester has inlet and outlet valves functioning for input-maintain-output the gas. We first collected the exhaled gas of healthy people (Male, 30 years old) with hygiene airbags and mixed them with 1000 ppm NH3 in a certain proportion to obtain 1000 ppb, 2000 ppb, 3000 ppb, and 5000 ppb simulated exhaled gas. Before the analysis, the chamber is vacuumed and then the 250 mL simulated gases were gradually injected into the POC tester till the response of the device was stable. The responses of the device were interrogated by the antenna out of POC tester. After that, the gas chamber is vacuumed for the next test.

  3. (3)

    Early leakage detection: we use micro-syringe to mimic ultra-low concentration NH3 leaks. By moving the device in a certain direction near the leak point, the characteristic curves of the device were recorded by the network analyzer. For leak frequency detection, we manually move micro-syringes back and forth to mimic a certain frequency of leakage and simultaneously record the characteristic curves of the device.

Ethics declarations

All research involving human participants was approved by human participants and the Institutional Review Board of Shanghai Jiao Tong University.