A colorimetric ammonia sensor based on interfacially assembled porous polymer membrane: coupled hydrogen-bonding and electronic structure modulation

Abstract

Real-time, low-power, and selective detection of ammonia (NH₃) is of critical importance in semiconductor manufacturing, environmental monitoring, and occupational safety. However, current sensing technologies often fall short in achieving a balance between sensitivity, stability, device integration, and user accessibility. In this work, we introduce a structurally asymmetric porous organic polymer membrane, synthesized via a liquid–liquid interfacial acylhydrazone condensation reaction, that enables rapid (∼1 s), reversible, and visually perceptible colorimetric sensing of NH₃. The membrane exhibits dynamic keto–enol tautomerism and a well-defined anisotropic morphology—featuring a dense organic-phase side and a highly porous, fibrous aqueous-phase side—that collectively enhance molecular diffusion and optical responsiveness. Upon exposure to NH₃, the disruption of intramolecular hydrogen bonding within the membrane backbone induces a pronounced absorption red-shift and a visible color transition from pale yellow to orange. Building on these molecular-level interactions, we engineer a laminated optical sensor that leverages UV-vis absorption changes for device-level signal transduction, achieving a quantification limit as low as 1 ppm. Additionally, we implement a smartphone-assisted RGB extraction method to enable semi-quantitative and user-friendly data analysis, highlighting the potential of the membrane for intelligent, field-deployable sensing. This work establishes a new paradigm in porous organic polymer-based gas sensors by uniting dynamic covalent chemistry, interfacial nanostructuring, and accessible device engineering to meet the demands of next-generation ammonia monitoring.

---

Haonan Peng

Dr Haonan Peng is a professor in physical chemistry at the School of Chemistry and Chemical Engineering, Shaanxi Normal University. He obtained his PhD in 2015 from the Toulouse Coordination Chemistry Laboratory (LCC-CNRS), University of Toulouse, France, with a research focus on spin crossover nanomaterials. His current research interests center on the development of high-performance film-based CBRN (Chemical, Biological, Radiological, and Nuclear) sensors. He investigates the dependence of film-based sensor performance on the properties of sensing units and the structural characteristics of active layers, aiming to elucidate the underlying physicochemical mechanisms governing these effects.

1. Introduction

The semiconductor industry underpins nearly every aspect of modern technology, from computing¹ and communications² to energy³ and transportation.⁴ As device architectures shrink and process integration becomes more complex, maintaining ultra-clean environments and precise chemical control have become essential.^5,6 Among the numerous process gases involved, ammonia (NH₃) plays a pivotal role in the chemical vapor deposition (CVD) of silicon nitride and in the synthesis of various nitride-based semiconductors.⁷ However, NH₃ is also corrosive, toxic, and highly reactive,^8,9 and even trace amounts can neutralize photoacid generators in advanced lithographic processes, disrupt cleanroom air stability, and damage sensitive fabrication equipment.^10–12 These challenges necessitate the development of high-performance NH₃ sensing technologies that combine sensitivity, selectivity, speed, and compatibility with semiconductor-grade conditions.

Conventional NH₃ sensing platforms—such as electrochemical sensors, metal oxide semiconductors (MOS), and optical spectrometers—have made considerable progress but still face important limitations.^13–15 MOS-based sensors typically require elevated temperatures to function, leading to high power consumption and limited material compatibility.^16,17 Fluorescence-based sensors, on the other hand, offer excellent sensitivity and resolution,^18–20 but their long-term use is constrained by photochemical instability and the complexity of signal readout.²¹ In contrast, colorimetric sensors based on absorption spectral shifts offer inherent advantages including simplified device architectures, visual interpretability, and greater photostability.^22,23 Yet, achieving high sensitivity and chemical specificity in a scalable and mechanically robust platform remains a key materials challenge.

Porous organic polymers (POPs) have emerged as a promising class of materials for gas sensing applications due to their structural tunability, high surface area, and abundance of chemically addressable sites.^24–26 Their modular synthesis enables the incorporation of targeted functional groups capable of specific interactions with gas molecules, while the intrinsic porosity enhances diffusion and analyte uptake. These features collectively support rapid, selective, and low-energy sensing.^27–30 However, translating the molecular-level advantages of POPs into macroscopically processable and mechanically robust sensing devices—particularly those capable of reliable optical transduction—remains underexplored (Scheme 1).³¹

Scheme 1 (a) Schematic illustration of the fabrication of the BTH–DFP membrane via self-assembly at the liquid–liquid interface. (b) Visual representation of the color transition of the membrane upon exposure to NH₃ vapor under ambient daylight. (c) Conceptual diagram of the BTH–DFP membrane integrated into a laminated sensing device for NH₃ vapor detection.

In this study, we report a novel porous organic polymer membrane, termed BTH–DFP, synthesized via interfacial acylhydrazone condensation between 1,3,5-benzene tricarbohydrazide (BTH) and 2-hydroxy-m-xylylbenzaldehyde (DFP). The dynamic covalent reaction at the liquid–liquid interface affords a free-standing, highly crosslinked membrane with a ductile fibrous morphology.^32,33 The presence of short alkyl linkers within the DFP monomer introduces molecular flexibility into the polymer backbone, facilitating film formation, improving mechanical integrity, and promoting efficient gas diffusion. These structural features enable the BTH–DFP membrane to maintain both stability and responsiveness under ambient conditions.

Of particular interest is the ortho-hydroxyl group on the DFP unit, which can participate in intramolecular hydrogen bonding with the adjacent hydrazone linkage, leading to modulation of the π-conjugation and absorption characteristics of the polymer. In addition, spectroscopic analyses suggest the presence of reversible keto–enol tautomerism within the framework, which further contributes to the material's optical tunability.^34,35 Upon exposure to NH₃ vapor, hydrogen bonding interactions between ammonia and the polymer backbone disrupt the original hydrogen-bonding network, inducing a red-shift in the absorption spectrum that is visually observed as a color change from pale yellow to orange.

To translate this molecular response into a functional device, we integrate the BTH–DFP membrane into a layered optical sensor architecture by mounting the film onto a quartz substrate. In this configuration, changes in light transmittance through the membrane—arising from NH₃-induced spectral shifts—are quantitatively recorded via a photodetector. Unlike fluorescence-based platforms that require high-energy excitation sources and complex optics, this absorption-based design enables real-time NH₃ detection with simplified instrumentation and relatively low power consumption.^36–38 The porosity, mechanical robustness, and optoelectronic responsiveness of the membrane further support its integration into compact and deployable sensor formats suitable for semiconductor environments.

Altogether, this work presents a dynamic covalent porous polymer platform that combines molecular-level recognition with device-level optical signal transduction for colorimetric gas sensing. By addressing key challenges in selectivity, signal reliability, and integration feasibility, the BTH–DFP membrane contributes a promising strategy for developing next-generation NH₃ sensors and expands the design principles of POP-based functional sensing materials.

2. Results and discussion

2.1 Structural characterization of BTH–DFP membrane

The BTH–DFP membrane was comprehensively characterized to confirm its covalent network structure and stability (Fig. 1a–f). Fig. S1† illustrates the proposed chemical structure of the BTH–DFP framework, consisting of BTH and DFP building blocks linked via hydrazone bonds. Fourier-transform infrared (FTIR) spectroscopy of the membrane (Fig. 1a) clearly indicates the formation of these linkages. In the spectrum of the assembled membrane, the characteristic absorption bands corresponding to the –NH₂ and –CHO groups of the BTH and DFP monomers, respectively, are significantly reduced or absent. Concurrently, a prominent new absorption band appears at ∼1610 cm⁻¹, which is attributed to the C=N stretching vibration of the hydrazone (Schiff-base) linkages.^39,40 In addition, the C=O stretching band of the aldehyde group in DFP, typically located near 1700 cm⁻¹, disappears entirely after polymerization, indicating its complete consumption. A residual absorption band around 1650 cm⁻¹ remains, assignable to the C=O group of the BTH unit, which is incorporated into the hydrazone structure.

Fig. 1 (a) FTIR spectra of BTH, DFP, and the BTH–DFP membrane. (b) Survey XPS spectrum of the BTH–DFP membrane. High-resolution XPS spectra of (c) C 1s, (d) N 1s, and (e) O 1s regions. (f) Schematic illustration of keto–enol tautomerism via intramolecular proton transfer.

X-ray photoelectron spectroscopy (XPS) analysis provided further insights into the elemental composition and bonding environments of the membrane (Fig. 1b–e). The survey spectrum confirmed the presence of C, N, and O as the main elements. High-resolution C 1s spectra (Fig. 1c) were deconvoluted into three peaks at 284.6 eV, 286.3 eV, and 288.1 eV, corresponding to C=C, C=N–C, and C=O species, respectively. The N 1s spectrum (Fig. 1d) exhibited two major peaks at 400.0 eV and 400.7 eV, which can be assigned to C–N and C=N bonds, indicative of the formation of hydrazone linkages. The O 1s spectrum (Fig. 1e) showed components at 532.2 eV and 533.5 eV, corresponding to C=O and C–OH environments, respectively. The coexistence of C=O and C–OH species suggests partial enolization or hydrogen bonding interactions within the framework (Fig. 1f).⁴¹

The thermal stability of the BTH–DFP membrane was assessed using thermogravimetric analysis (TGA) (Fig. S2†). The membrane exhibited negligible mass loss below 300 °C, demonstrating excellent thermal robustness and suitability for integration into sensing devices operating under low to moderate temperature conditions.

2.2 Topographic characterization of BTH–DFP membrane

The morphological and structural characteristics of the BTH–DFP membrane were comprehensively investigated using a combination of macro- and micro-scale characterization techniques. As shown in Fig. 2a, the membrane exhibits excellent free-standing capability and emits bright orange fluorescence under 365 nm UV irradiation, attributable to the incorporation of aromatic building blocks. The membrane can also be readily cut or shaped (Fig. S3†), underscoring its potential for device integration.

Fig. 2 (a) Photographs of the BTH–DFP membrane floating on water and supported by a metal ring under ambient daylight and UV light (365 nm). (b) Water contact angle measurements on the DCM side and H₂O side of the membrane. (c) SEM image, (d) magnified SEM image, (e) AFM image and (f) TEM image, of both sides of the membrane. Elemental mapping of the (g) H₂O side and (h) DCM side. (i) Cross-sectional SEM image of the BTH–DFP membrane.

Water contact angle (WCA) measurements revealed values of 56 ± 3° on the aqueous phase side and 46 ± 3° on the organic phase side (Fig. 2b), suggesting moderate hydrophilicity on both surfaces. The difference in contact angles indicates a potential asymmetry in surface structure and wettability, which may contribute to enhanced interfacial adhesion, as qualitatively supported by adhesion tests (Fig. S4†). However, given the relatively small WCA difference, this asymmetry is more convincingly demonstrated through electron microscopy. Surface morphology characterization via scanning electron microscopy (SEM) (Fig. 2c and d) revealed a clear contrast between the two sides of the membrane. The organic phase side (Fig. 2c) displays a relatively dense and smooth morphology, whereas the aqueous phase side (Fig. 2d) presents a more open, porous, and fibrous architecture.

Considering that polymerization kinetics and phase separation behavior at the liquid–liquid interface are critical factors influencing the formation of porous morphologies, we systematically tuned the composition of the reaction system to investigate their effects. Specifically, we varied the concentration of the precursor solution and the amount of acetic acid (HAc) catalyst during interfacial polymerization. As shown in Fig. S14a†, increasing the volume of HAc led to a more pronounced fibrous porous morphology, suggesting that HAc plays a pivotal role in regulating the polymerization rate and phase separation dynamics.

To further elucidate the tunability of the porous structure, we adjusted the precursor concentration. As shown in Fig. S14b†, higher precursor concentrations resulted in larger pore sizes and thicker membranes, highlighting the influence of monomer content on the structural evolution of the resulting membrane. The porous, fibrous morphology on the aqueous side is expected to enhance surface area and facilitate mass transport, while the denser organic-facing surface provides improved mechanical integrity and acts as a barrier layer.

Atomic force microscopy (AFM) (Fig. 2e) further corroborates the structural asymmetry. The aqueous phase side exhibits significantly higher surface roughness with visible nanoscale protrusions, in contrast to the smoother organic phase side. These results are consistent with the SEM observations and support the existence of an anisotropic membrane surface.

High-resolution transmission electron microscopy (HR-TEM) (Fig. 2f) and X-ray diffraction (XRD) analysis (Fig. S5†) indicate the amorphous nature of the membrane, with no evidence of long-range crystalline order. The membrane interior exhibits a disordered network-like structure, in line with the dynamic covalent assembly mechanism.

Elemental distribution was assessed by energy-dispersive X-ray spectroscopy (EDS). Mapping results on both the aqueous phase (Fig. 2g) and organic phase (Fig. 2h) sides show homogeneous distributions of C, N, and O, with no detectable phase separation or elemental aggregation, suggesting uniform chemical composition across both surfaces. Cross-sectional SEM imaging (Fig. 2i) revealed that the membrane possesses a layered architecture with a total thickness of approximately 2.8 μm.

Collectively, these results demonstrate that the BTH–DFP membrane possesses a chemically uniform yet structurally anisotropic architecture, combining a porous, fibrous side favorable for diffusion with a dense surface that offers mechanical robustness, establishing a strong basis for potential applications in interfacial sensing technologies.

2.3 Visual sensing of NH₃ vapor by BTH–DFP membrane

The colorimetric response of the BTH–DFP membrane toward ammonia vapor was systematically evaluated under ambient light conditions. As shown in Fig. 3a, the membrane exhibited a rapid and distinct visual change upon exposure to NH₃, transitioning from pale yellow to orange within 0.5 s (Video S1†). Quantitative analysis of RGB color values extracted from digital images revealed a sharp increase in red intensity, accompanied by decreases in both green and blue channels, consistent with the observed chromatic shift. The total color difference, expressed as ΔRGB (see ESI† for calculation formula), increased sharply within 0.15 s and subsequently plateaued, highlighting the ultrafast response kinetics of the membrane (Fig. 3b).⁴²

Fig. 3 (a) Time-dependent color transition of the BTH–DFP membrane upon NH₃ vapor exposure with corresponding RGB values. (b) ΔRGB response as a function of exposure time. (c) UV-vis absorption spectra before and after NH₃ exposure. (d) PCA plot illustrating selectivity toward various vapors. (e) Pseudo-color map of RGB response ratios for different vapors. Abbreviations: ACE – Acetone, DCM – Dichloromethane, MeOH – Methanol, H₂O – Water, EtOH – Ethanol, CCl₄ – Carbon tetrachloride, An – Aniline, PEA – Phenylethylamine, TEA – Triethylamine, BnNH₂ – Benzylamine, DIPA – Diisopropylamine, DEA – Diethylamine, NH₃ – Ammonia.

UV-vis absorption spectroscopy was employed to further probe the interaction between NH₃ and the membrane. Following exposure, a new absorption band emerged at approximately 460 nm (Fig. 3c), indicating a change in the electronic environment of the chromophoric units within the membrane. This spectral redshift suggests an NH₃-induced structural or electronic reorganization, potentially involving modifications to conjugation length or hydrogen bonding networks.

To assess the origin of this optical transition, SEM images before and after NH₃ exposure (Fig. S7†) were compared and showed no apparent morphological changes, thereby excluding macroscopic structural collapse as a contributing factor. Further insight was obtained from FTIR spectroscopy (Fig. S8†): a redshift in the C=O stretching vibration, along with spectral changes around 1550 cm⁻¹, suggests perturbation of the original intramolecular hydrogen bonds-particularly those between hydroxyl groups and adjacent C=N linkages. This interaction may disrupt the hydrogen-bond-stabilized six-membered ring motifs and promote a more delocalized π-conjugated structure, thereby accounting for the observed absorption shift.⁴³ To support this mechanism, we constructed a molecular model of the membrane and carried out density functional theory (DFT) calculations(Fig. S9†). The results indicate that NH₃ adsorption disrupts the intramolecular hydrogen-bonding network and significantly reduces the HOMO–LUMO energy gap from 6.16 eV (unbound) to 4.92 eV (after NH₃ binding) (Fig. S10†). Moreover, simulated UV-vis spectra show the emergence of a new long-wavelength absorption feature upon NH₃ interaction (Fig. S11†), which aligns well with our experimental observations. These findings further validate the proposed sensing mechanism, in which NH₃ exposure disrupts hydrogen bonding within the membrane and induces concurrent structural and electronic changes, ultimately resulting in the observed optical response.

The selectivity toward NH₃ of the membrane was further evaluated by exposing it to various volatile organic compounds (VOCs). As shown in Fig. S12†, no comparable visual color change was observed upon exposure to potential interferents. Principal component analysis (PCA) based on RGB response data (Fig. 3d) demonstrated clear statistical separation between NH₃ and other analytes. Moreover, pseudo-color mapping of RGB intensity ratios (Fig. 3e) revealed a distinct chromatic fingerprint for NH₃, in stark contrast to those of VOCs, further reinforcing the sensor's specificity.⁴⁴

Together, these results confirm that the BTH–DFP membrane enables rapid, selective, and visually perceptible ammonia sensing under ambient conditions. The combination of ultrafast response time, well-defined chromatic transition, and analyte discrimination capability highlights its potential for real-time and field-deployable gas detection platforms.

2.4 Laminated light transmission sensors

To bridge dynamic molecular responsiveness with practical device integration, we developed a compact optical sensing platform that translates nanoscopic chemical interactions into real-time, quantifiable optical outputs. At the core of this platform is a spectroscopic detection strategy centered on the characteristic absorption band at 460 nm, which emerges uponexposure of the BTH–DFP membrane to ammonia vapor. This distinct spectral feature serves as a reliable optical signature for selective NH₃ detection.

The sensing device adopts a vertically stacked configuration (Fig. 4a), consisting of a 460 nm LED light source, an optical filter, the BTH–DFP membrane, a transparent detection window, and a wavelength-matched photodetector. This architecture ensures precise spectral isolation of the chemically responsive region and enables the selective transduction of molecular recognition events into electrical signals.^45–47 In the absence of NH₃, the transmitted light intensity through the membrane, recorded as I₀, remains constant. Upon exposure to NH₃, enhanced absorption at 460 nm reduces the transmitted intensity to I, owing to ammonia-induced modulation of the electronic structure of the membrane. The sensor response is thus defined by the ratio I/I₀, which decreases proportionally with increasing NH₃ concentration, offering a robust and real-time metric for gas quantification.

Fig. 4 (a) Schematic illustration of the optical sensor configuration based on the BTH–DFP membrane. (b) Diagram of the integrated sensing system, consisting of a control board, solenoid valve, sensor, pump, sample bag, and exhaust module. (c) Photograph of the assembled sensing system in operation.

To enable automated and reproducible operation, the optical module was integrated into a digitally controlled microfluidic system (Fig. 4b), equipped with solenoid valves, programmable pumps, and a custom-built signal acquisition board. This setup permits accurate delivery of gaseous analytes from sealed sampling bags and facilitates rapid sensor regeneration across measurement cycles. A photograph of the assembled platform is shown in Fig. 4c, illustrating its modular design and high operational precision.

Altogether, this integrated system exemplifies a generalizable approach for developing portable optical gas sensors by coupling material-intrinsic spectral responsiveness with scalable optoelectronic interfaces. Beyond ammonia detection, this strategy offers a versatile blueprint for the development of real-time environmental sensing technologies targeting a broad range of volatile analytes.

2.5 Sensing performance of membrane BTH–DFP for NH₃

The ammonia sensing performance of the BTH–DFP membrane was systematically investigated under ambient conditions. To establish optimal operational parameters, both the sample injection duration and gas flow rate were first optimized. As shown in Fig. 5a, a 1.0 s injection time yielded a rapid and stable response, suggesting this duration as an appropriate operational setting. Additionally, a gas flow rate of 4 mL s⁻¹ (Fig. S13†) provided reproducible and uniform delivery of analyte vapor and was therefore used in all subsequent experiments.

Fig. 5 (a) Optimization of sampling time for the BTH–DFP membrane. Kinetic response curves of the (b) H₂O side and (c) DCM side of the membrane to the same NH₃ vapor concentration. (d) Sensor response intensities at various NH₃ concentrations, showing a linear relationship from 1 to 100 ppm (n = 3). (e) Selectivity performance of the sensor exposed to saturated vapors of interfering VOCs and NH₃. (f) Reproducibility evaluation using five independently prepared membrane batches. (g) Repeatability test showing stable cycling performance over 85 exposure–recovery cycles.

Given the asymmetric architecture of the membrane, we compared the sensing response when NH₃ vapor was introduced from either the aqueous-phase side or the organic-phase side. As shown in Fig. 5b, exposure from the aqueous-phase side led to a rapid signal drop within ∼1 s, followed by gradual recovery to baseline within 97 s upon purging with ambient air. In contrast, exposure from the organic-phase side (Fig. 5c) resulted in significantly attenuated signal changes and negligible recovery. This directional discrepancy is attributed to morphological differences: the aqueous-phase side features a porous, fibrous structure that facilitates gas diffusion and exchange, whereas the organic-phase side is denser and less permeable, hindering analyte transport.⁴⁸ To further elucidate the structure–performance relationship, we evaluated the effect of pore structure on sensing behavior. Laminated sensors fabricated from membranes with varying pore sizes were exposed to a fixed concentration of NH₃ vapor. As shown in Fig. S14c†, membranes with larger pores exhibited stronger optical response intensities but lower recovery efficiencies. This trade-off likely arises from the increased membrane thickness associated with higher precursor concentrations, which enhances gas uptake but concurrently impedes desorption, thereby delaying signal recovery.

Under optimized conditions, the response of the membrane to NH₃ at varying concentrations was evaluated (Fig. 5d). The response signal—quantified as the relative change in transmitted intensity—decreased progressively with increasing NH₃ concentration from 1 to 100 ppm, displaying excellent linearity (R² = 0.997). These results confirm the potential of the membrane for reliable quantitative analysis. Compared with recently reported high-performance NH₃ sensors (Table S1†), the BTH–DFP-based sensor exhibits a fast response time (1 s) and a low quantification limit (1 ppm), outperforming many chemiresistive and fluorescence-based sensors in terms of speed and sensitivity.

To evaluate selectivity, the membrane was exposed to saturated vapors of 21 representative volatile organic and inorganic compounds (Fig. 5e). Only NH₃ produced a significant change in the signal, while all other analytes yielded negligible responses, indicating high chemical specificity. This selectivity may originate from the strong Lewis basicity and steric characteristics of NH₃.

Sensor reproducibility was confirmed by testing five independently fabricated membrane batches under identical conditions (Fig. 5f). All batches exhibited consistent response magnitudes, demonstrating excellent batch-to-batch uniformity. Long-term operational stability was assessed through 85 consecutive exposure–recovery cycles using 50 ppm NH₃ (Fig. 5g), during which the membrane maintained stable performance without significant signal drift or degradation, highlighting its outstanding reversibility and durability.

2.6 Smartphone-assisted system enabled on-site NH₃ vapor detection

The BTH–DFP membrane also enables real-time, in situ monitoring of NH₃ vapor through a concentration-dependent colorimetric response that is easily observable under ambient lighting. As shown in Fig. 6a, the membrane color progressively changes from pale yellow to orange as the NH₃ concentration increases. This transformation is readily distinguishable to the naked eye and becomes perceptible at concentrations as low as 5 ppm, highlighting the high sensitivity andpractical detection utility of the membrane. To overcome the subjectivity of visual perception and enable quantitative analysis, a smartphone-assisted detection strategy was employed. Photographic images of the membrane were acquired using a standard smartphone camera, and the RGB values were digitally extracted. The red (R) channel intensity exhibited a pronounced increase with rising NH₃ concentration, while the green (G) and blue (B) channels concurrently decreased. The derived R/B and G/B intensity ratios showed strong linear correlations with NH₃ concentration in the range of 0–100 ppm (Fig. 6b and S15†, R² = 0.997), enabling semi-quantitative estimation of analyte concentration based solely on image data.

Fig. 6 (a) Photographs of the BTH–DFP membrane under ambient daylight showing color change in response to increasing NH₃ concentrations (0–1000 ppm), with corresponding RGB values. (b) Linear correlation between the R/B ratio and NH₃ concentration in the range of 0–100 ppm. (c) Schematic illustration of real-time NH₃ vapor quantification using smartphone-assisted RGB analysis. (d) Simulation of NH₃ leakage detection in a pipeline scenario using the smartphone-based sensing platform.

Importantly, this method eliminates the need for specialized analytical instrumentation. Users can simply photograph the membrane with a smartphone and upload the image to an app-based system (Fig. 6c), which automatically extracts RGB values and estimates the corresponding NH₃ concentration based on pre-calibrated colorimetric models. This image-based quantification strategy enhances the sensor's portability, accessibility, and ease of use, making it suitable for point-of-care and field applications.

To demonstrate its practical utility, a simulated NH₃ leak scenario was established by delivering 50 ppm ammonia vapor through micro-perforated rubber tubing using a miniature air pump. The BTH–DFP membrane was positioned near the leak site, and a smartphone image was captured following exposure. As shown in Fig. 6d, RGB parameters extracted from the image yielded an estimated local NH₃ concentration of approximately 20 ppm via the smartphone app. This result confirms the effectiveness of the membrane for real-world applications such as on-site ammonia leak detection and supports its integration into portable, user-friendly sensing systems.

2.7 Blind and sealed detection of NH₃ using BTH–DFP membranes

In order to investigate the reliability of the sensing membranes in real detection situations, we designed blind testing experiments. As shown in Fig. 7a, a blind-box test was performed using ten sealed vials (A1–A10), among which three (A2, A4, and A9) were randomly injected with a fixed amount of NH₃ vapor. The test was conducted by an independent operator with no prior knowledge of the sample identities. Both the freestanding colorimetric membrane (Fig. 7b) and the laminated sensor (Fig. 7c) accurately identified the NH₃-containing vials, with detection results perfectly matching the true labels. This confirms the sensor's high reliability and selectivity under blind testing conditions.

Fig. 7 (a) Photograph of ten sealed vials (A1–A10), with three vials containing NH₃. (b) Colorimetric sensing results of the membranes. (c) Detection performance of the laminated sensor. (d) Three sealed cartons containing different items: carton 1 with trace NH₃, carton 2 with a lab coat, and carton 3 with a laboratory notebook. (e) Photograph of the gas sampling process. (f) Sensing results of the laminated sensor for boxes 1–3; inset images show the corresponding colorimetric responses result.

To further simulate a real-world scenario, we carried out a sealed-carton test (Fig. 7d). Carton 1 contained trace NH₃ vapor, while cartons 2 and 3 contained a lab coat and a notebook, respectively, serving as common non-target packaging materials. Gas samples were extracted from the cartons via syringe and analyzed using the sensor (Fig. 7e). Both sensing formats successfully detected NH₃ in carton 1, with no false positives in the control cartons (Fig. 7f). These findings highlight the great potential of this sensor for practical deployment in field monitoring and packaging safety applications.

3. Conclusions

In summary, we developed a structurally asymmetric ammonia-sensing membrane via a liquid–liquid interfacial acylhydrazone condensation strategy. The resulting BTH–DFP membrane features a dense organic-phase side and a porous, fibrous aqueous-phase side, enabling efficient mass transport and directional gas diffusion. The colorimetric sensing mechanism is governed by hydrogen-bond-mediated modulation of the hydrazone chromophore, wherein NH₃ disrupts intramolecular hydrogen bonding and induces a red-shift in optical absorption, manifesting as a rapid and visually perceptible color change from pale yellow to orange.

By coupling the intrinsic optical responsiveness of the membrane with a laminated transmission-type device architecture, we constructed a compact optical sensor capable of real-time ammonia detection with a response time of 1 s and a quantification limit as low as 1 ppm. Furthermore, integration with a smartphone-based RGB analysis platform enables semi-quantitative evaluation of NH₃ concentrations under ambient conditions, eliminating the need for specialized instrumentation. This combination of dynamic covalent chemistry, interfacial nanoengineering, and portable digital readout establishes a versatile and scalable framework for next-generation colorimetric gas sensors, with strong potential in on-site environmental monitoring and industrial leak detection applications.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the National Key Research and Development Program of China (2022YFA12055002), National Natural Science Foundation of China (22272101, 22427802), the Program of Introducing Talents of Discipline to Universities (111 project, B14041).

Notes and references

1. H. Ran, Z. Ren, J. Li, B. Sun, T. Wang, D. Gu, W. Wang, X. Hu, Z. Dong, Q. Song, L. Wang, S. Duan and G. Zhou, Adv. Funct. Mater., 2024, 35, 2418113.
2. Y. De Koninck, C. Caer, D. Yudistira, M. Baryshnikova, H. Sar, P.-Y. Hsieh, C. I. Özdemir, S. K. Patra, N. Kuznetsova, D. Colucci, A. Milenin, A. A. Yimam, G. Morthier, D. Van Thourhout, P. Verheyen, M. Pantouvaki, B. Kunert and J. Van Campenhout, Nature, 2025, 637, 63–69.
3. T. Wang, B. Yang, Z. Zhou, Y. Wu and Z. Jin, Small, 2025, 21, 2501128.
4. L. Xu, S. Lv, Q. Chen, P. Gao, G. Gu, Y. Wang, X. Min, H. Kim and L. Cai, J. Hazard. Mater., 2025, 486, 137083.
5. F. Schüler, A. Keller, M. Grebe, A. Kent, M. Buttery and M. Holynska, Sci. Rep., 2025, 15, 15407.
6. W. Wang, N. Clark, M. Hamer, A. Carl, E. Tovari, S. Sullivan-Allsop, E. Tillotson, Y. Gao, H. de Latour, F. Selles, J. Howarth, E. G. Castanon, M. Zhou, H. Bai, X. Li, A. Weston, K. Watanabe, T. Taniguchi, C. Mattevi, T. H. Bointon, P. V. Wiper, A. J. Strudwick, L. A. Ponomarenko, A. V. Kretinin, S. J. Haigh, A. Summerfield and R. Gorbachev, Nat. Electron., 2023, 6, 981–990.
7. N. Hegedüs, K. Balázsi and C. Balázsi, Materials, 2021, 14, 5658.
8. R. Zhang, N. Liu, Z. Lei and B. Chen, Chem. Rev., 2016, 116, 3658–3721.
9. J. Shin, K. Kim, I. S. Min, M. Sang, J. Y. Lee, K. Hwang, Y. Kang, J. Kim and K. J. Yu, Adv. Funct. Mater., 2024, 35, 2419110.
10. X. Wang, L. Gong, Z. Li, Y. Yin and D. Zhang, J. Mater. Chem. A, 2023, 11, 7690–7701.
11. Y. Wang, L. Huang, T. C. Zhang, Y. Wang and S. Yuan, Chem. Eng. J., 2024, 482, 148811.
12. K. M. Yoo, J. Midkiff, A. Rostamian, C. J. Chung, H. Dalir and R. T. Chen, ACS Sens., 2020, 5, 861–869.
13. Q. Liu, Q. Sun, J. Shen, H. Li, Y. Zhang, W. Chen, S. Yu, X. Li and Y. Chen, Coord. Chem. Rev., 2023, 482, 215078.
14. R. Zhu, Q. Zhou, Q. Tian, S. Zhao, W. Qin, X. Wu, S. Xu and Y. Zhang, ACS Sens., 2024, 9, 4286–4294.
15. S. Sonwal, K. S. Ranjith, S. Han, Y.-K. Han, M.-H. Oh and Y. S. Huh, J. Mater. Chem. A, 2024, 12, 11004–11019.
16. A. Sharma, S. B. Eadi, H. Noothalapati, M. Otyepka, H. D. Lee and K. Jayaramulu, Chem. Soc. Rev., 2024, 53, 2530–2577.
17. W. Gao, X. Chang, O. Ola, Z. Wang, Q. Liao, X. Zhu, J. Li, Y. Jiang, D. Wang and S. Sun, Chem. Eng. J., 2025, 509, 161482.
18. R. Huang, T. Liu, H. Peng, J. Liu, X. Liu, L. Ding and Y. Fang, Chem. Soc. Rev., 2024, 53, 6960–6991.
19. H. Peng, L. Ding and Y. Fang, J. Phys. Chem. Lett., 2024, 15, 849–862.
20. C. Lyu, C. Zhao, M. Wang, J. Li, Z. Cai, X. Dou and B. Zu, Aggregate, 2025 DOI:10.1002/agt2.70053.
21. J. Wang, R. Wen, J. Kong, J. Liu, Z. Liu and Y. Fang, Angew. Chem., Int. Ed., 2025, 64, e202502020.
22. D. Chang, J. Li, R. Liu, M. Liu, K. Tram, N. Schmitt and Y. Li, Angew. Chem., Int. Ed. Engl., 2023, 62, e202315185.
23. K. Jin, D. Moon, Y. P. Chen and J. Park, Adv. Mater., 2024, 36, e2309570.
24. Y. Zhu, P. Xu, X. Zhang and D. Wu, Chem. Soc. Rev., 2022, 51, 1377–1414.
25. S. Wang, H. Li, H. Huang, X. Cao, X. Chen and D. Cao, Chem. Soc. Rev., 2022, 51, 2031–2080.
26. S. Guo and T. M. Swager, J. Am. Chem. Soc., 2021, 143, 11828–11835.
27. W. Song, Y. Zhang, C. H. Tran, H. K. Choi, D.-G. Yu and I. Kim, Prog. Polym. Sci., 2023, 142, 101691.
28. M. Asad, M. Imran Anwar, A. Abbas, A. Younas, S. Hussain, R. Gao, L.-K. Li, M. Shahid and S. Khan, Coord. Chem. Rev., 2022, 463, 214539.
29. P. Srinivasan, K. Dhingra and K. Kailasam, J. Mater. Chem. A, 2023, 11, 17418–17451.
30. W. Chen, P. Chen, G. Zhang, G. Xing, Y. Feng, Y. W. Yang and L. Chen, Chem. Soc. Rev., 2021, 50, 11684–11714.
31. Z. Li, J. R. Askim and K. S. Suslick, Chem. Rev., 2019, 119, 231–292.
32. Y. Shu, Y. Luo, H. Wei, L. Peng, J. Liang, B. Zhai, L. Ding and Y. Fang, Angew. Chem., Int. Ed. Engl., 2024, 63, e202402453.
33. K. Tiwari, P. Sarkar, S. Modak, H. Singh, S. K. Pramanik, S. Karan and A. Das, Adv. Mater., 2020, 32, e1905621.
34. W. J. Zhang, W. Z. Sun, J. T. Xu, B. Q. Ou, W. Q. Zhou, L. Chen, J. W. Ye and M. Pan, J. Am. Chem. Soc., 2025, 147, 17486–17496.
35. Y. Xu, T. Sun, T. Zeng, X. Zhang, X. Yao, S. Liu, Z. Shi, W. Wen, Y. Zhao, S. Jiang, Y. Ma and Y. B. Zhang, Nat. Commun., 2023, 14, 4215.
36. F. Chun, B. Zhang, Y. Gao, X. Wei, Q. Zhang, W. Zheng, J. Zhou, Y. Guo, X. Zhang, Z. Xing, X. Yu and F. Wang, Nat. Photonics, 2024, 18, 856–863.
37. A. Schlachter, P. Asselin, A. Chatelain and P. D. Harvey, Adv. Funct. Mater., 2024, 34, 2404111.
38. M. D. Mehare, C. M. Mehare, H. C. Swart and S. J. Dhoble, Prog. Mater. Sci., 2023, 133, 101067.
39. J. Tang, Z. Liang, H. Qin, X. Liu, B. Zhai, Z. Su, Q. Liu, H. Lei, K. Liu, C. Zhao, R. Cao and Y. Fang, Angew. Chem., Int. Ed. Engl., 2023, 62, e202214449.
40. J. Tang, B. Zhai, X. Liu, J. Liu, C. Zhao and Y. Fang, J. Colloid Interface Sci., 2022, 612, 516–524.
41. X. Liu, R. Huang, L. Peng, J. Yang, J. Yan, B. Zhai, Y. Luo, C. Zhang, S. Tan, X. Liu, L. Ding and Y. Fang, Angew. Chem., Int. Ed. Engl., 2025, 64, e202414472.
42. J. Feng, H. Chang, X. Zhou, S. Zhang, L. Ding, T. Liu and Y. Fang, ACS Appl. Mater. Interfaces, 2025, 17, 14623–14633.
43. A. Y. Sheikh, A. Mattei, R. Miglani Bhardwaj, R. S. Hong, N. S. Abraham, G. Schneider-Rauber, K. M. Engstrom, M. Diwan, R. F. Henry, Y. Gao, V. Juarez, E. Jordan, D. A. DeGoey and C. W. Hutchins, J. Am. Chem. Soc., 2021, 143, 17479–17491.
44. X. Miao, C. Wu, Y. Xia, S. Yu, F. Li and M. Zhang, Nat. Commun., 2024, 15, 10641.
45. Z. Zhou, L. Zhang, L. Peng, Y. Li, X. Zhu, Y. Wu, Z. Qiu, G. He, M. Qin, H. Peng and Y. Fang, Aggregate, 2024, 5, e629.
46. K. Liu, C. Shang, Z. Wang, Y. Qi, R. Miao, K. Liu, T. Liu and Y. Fang, Nat. Commun., 2018, 9, 1695.
47. Y. Ke, Y. Liu, B. Zu, D. Lei, G. Wang, J. Li, W. Ren and X. Dou, Angew. Chem., Int. Ed. Engl., 2022, 61, e202203358.
48. W. Mo, Z. Zhu, F. Kong, X. Li, Y. Chen, H. Liu, Z. Cheng, H. Ma and B. Li, Nat. Commun., 2022, 13, 5189.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta03808a

‡ Zebiao Qiu and Qianchi Xiong contributed equally to this work.

---