Highly modular conductive threads for multiplexed detection of hazardous gases

Abstract

Wearable electronics have progressed from rigid circuit boards to fibertronics, where electronic functionality is integrated directly onto textile threads, advancing their utility. However, on-body chemical hazard monitoring remains limited by the poor sensitivity, selectivity, and durability of existing fiber-based sensors. Here, we report a modular, layer-by-layer strategy to fabricate robust, crystalline, conformal, and conductive coatings of stimuli-responsive Ni-, Cu-, and Zn-coordinated hexahydroxy- or hexaimino-triphenylene metal–organic frameworks (MOFs) on cotton threads. A four-component chemiresistive array of the threads detects and differentiates five toxic gases—H₂S, SO₂, NO, NH₃, and CO—with theoretical detection limits of 43, 60, 6, 65, and 417 ppb, respectively, all below Occupational Safety and Health Administration permissible exposure limits. When embedded into personal protective equipment, the sensor patch retains functionality even under elevated humidity and after repeated washing. This modular platform technology advances the development of chemically-responsive smart garments towards multiplexed environmental monitoring of toxic chemicals.

This work achieves a generalizable strategy for the robust deposition of conductive metal-organic framework on individual threads to create a wearable MOF@thread array patch that can efficiently detect and differentiate toxic gases.

Introduction

The integration of electronic functionalities into textile fibers—an emerging field of fibertronics—is rapidly advancing the frontier of wearable electronic technologies^1,2. Embedding conductive materials directly into fibers has enabled the development of smart textiles that retain the lightweight, flexible, breathable, and wearable properties of conventional fabrics while performing advanced functions, such as logic computation^3,4, robotics⁵, energy harvesting and storage⁶, light display¹, in addition to the detection of mechanical⁷, acoustic^8,9, thermal¹⁰, magnetic¹¹, and (bio)chemical stimuli^12,13, or multifunctional combinations thereof¹⁴. Among the most urgent and compelling applications of fiber-based electronics is the real-time detection of hazardous gases, including hydrogen sulfide (H₂S), sulfur dioxide (SO₂), nitric oxide (NO), ammonia (NH₃), and carbon monoxide (CO)¹⁵. The particular challenge of this application is its demand for the seamless integration of stimuli-responsive materials into fibers that combine robust chemical and mechanical stability with highly sensitive electronic responses to specific gaseous targets. Chemically-responsive conductive materials, such as metal oxide-decorated carbon nanotubes^16,17, conductive polymers^18,19, reduced graphene oxide/metal oxide hybrids²⁰, and MXenes²¹ have shown potential for gas sensing at the fiber level. However, their deployment in practical settings has been hindered by four main limitations: i) low modularity requiring extensive optimization for specific target analytes²²; ii) poor sensitivity and selectivity at sub-ppm concentrations; iii) inadequate performance below the permissible exposure limits (PELs) defined by the Occupational Safety and Health Administration (OSHA)^16,18; and iv) reliance on hazardous reagents²⁰ or energy intensive processes for material fabrication¹⁶. Together, these challenges underscore the need for a new platform technology capable of scalable and modular integration of highly responsive materials onto fiber-based substrates that enables the detection of toxic gases with high selectivity below OSHA’s PELs and operational stability under ambient conditions.

Two-dimensional electrically-conductive metal–organic frameworks (2D cMOFs) represent an emerging class of porous crystalline materials that combine high surface area, structural modularity, and low dimensionality with intrinsic electronic conductivity²³. Their bottom-up self-assembly from highly modular molecular precursors allows for precise control over material‒analyte interactions, resulting in distinct electronic responses to specific chemical stimuli, making cMOFs particularly attractive for chemiresistive gas sensing²⁴. However, integrating these conductive crystalline materials onto mechanically flexible substrates, such as textile fibers, remains a substantial challenge²⁵. Existing strategies often fail to achieve uniform, conformal coatings at the fiber level that maintain strong adhesion, preserve crystallinity, and enable stable electrical integration on porous, woven architectures. To date, only a few cMOFs have been integrated onto textiles, with no precedent of coatings at the single-fiber level^26–31. Effective deposition of cMOFs onto textile threads must address several critical criteria, which encompass the i) preservation of the crystalline framework for maintaining chemically-precise surface functionality, ii) retention of continuous electrical contact across the fiber surface for low-power transduction of electronic signal, and iii) accommodation of the mechanical stresses associated with bending, stretching, and weaving, critical for wearable electronics. While several deposition approaches, such as solvothermal in situ growth^26,32,33, electrospinning³⁴, and oxidative restructuring^27,28 have demonstrated proof-of-concept success, their modularity, generality, and practical utility remain limited. Key barriers include preferential MOF nucleation in solution rather than on substrate³⁵, poor mechanical robustness of the MOF-fiber interface, and the absence of a broadly applicable method for depositing functionally diverse cMOFs on fiber substrates. As a result, the widespread implementation of cMOF-based textile sensors in wearable applications has yet to be realized²⁵.

This work presents a generalizable, bottom-up solution-phase strategy for depositing a suite of triphenylene-based 2D cMOFs onto individual woven fibers via sequential layer-by-layer (LbL) assembly of metal-ion and organic-linker precursors. This approach yields conformal, mechanically robust cMOF@thread composites that preserve crystallinity and adhere tightly to the fiber surface. Unlike previous approaches for integrating cMOFs on textiles that lack generality and applicability beyond 1‒2 cMOFs^26–29, this approach is highly general, allowing the fabrication of at least four different cMOFs on fiber-based substrates at ambient conditions with no requirement of specialized or expensive equipment. The resulting fiber-integrated sensing array, comprising four cMOFs derived from hexahydroxy- and hexaimino-triphenylene (HHTP and HITP) linkers coordinated with nickel, copper, or zinc ions, enables electrically transduced detection of five hazardous gases: H₂S, SO₂, NO, NH₃, and CO at concentrations relevant to environmental monitoring and personal safety. The array exhibits high sensitivity, achieving limits of detection (LODs) in the parts-per-billion (ppb) range, with values of 43, 60, 6, 65, and 417 ppb for the above-mentioned gases, respectively, well below their respective OSHA PELs of 20, 5, 25, 50, and 50 ppm³⁶. In addition to its sensitivity, the cMOF@thread array demonstrates excellent selectivity, enabling both gas-specific discrimination and concentration-resolved detection using principal component analysis (PCA). Remarkably, the thread array shows efficient differentiation associated with long-term stability up to at least 10 months of aging, in air, and at elevated humidity levels reaching saturation. We further study the underlying material‒analyte interactions at the basis of these sensing responses through in situ and ex situ spectroscopic techniques. In addition, we demonstrate the utility of the sensor threads by integrating them into common personal protective equipment (PPE), including face masks and lab coats. These demonstrations highlight the potential of this platform for rapid monitoring of hazardous gas exposure in environments relevant to first responders, firefighters, and military personnel. By leveraging the advantages of sensor arrays, including multiplexed detection and facile reconfiguration for new targets, our system addresses longstanding challenges associated with rigid substrates, high power consumption, and large device footprints³⁷. The resulting lightweight, flexible electronic architecture enables non-line-of-sight detection through multiple fabric layers and maintains performance after up to three laundering cycles, underscoring its robustness and reusability. Collectively, these results establish a general, scalable platform for wearable environmental sensing, advancing the development of smart garments capable of continuous, on-body detection of toxic gases across diverse operational settings.

Results

MOF deposition rationale

We chose to deposit triphenylene-based MOFs, particularly Ni₃(HITP)₂, Cu₃(HITP)₂, Zn₃(HHTP)₂, and Cu₃(HHTP)₂, onto fibers for three major reasons. First, this class of materials has well-documented sensing capabilities towards toxic gases^26,38,39 and volatile organic compounds^40,41, in both, the bulk form and when incorporated onto textiles. We reasoned that leveraging these properties would enable the fabrication of electronic fibers suitable for compact, wearable sensing platforms with applicability to first responder protection. Second, the structural similarity^40,42 and tunability of these cMOFs, achieved through the variation of metal nodes and bridging groups, allow for modular control of material‒analyte interactions with molecular precision, correlating with diversified sensor performance⁴³. Third, the commercial availability and synthetic accessibility of their organic linker precursors, along with the relatively high earth abundance of the employed metal ions⁴⁴, make these frameworks attractive candidates for scalable and cost-effective deposition processes²⁴. Despite these advantages, their application in wearable sensing remains underdeveloped, primarily due to the lack of a generalizable deposition method capable of integrating diverse framework chemistries onto flexible fiber substrates in a robust, conformal, and mechanically-resilient manner²⁷. While two of these MOFs have been previously integrated into textile-based substrates, the integration of this class of materials onto individual fibers has not been previously demonstrated.

We employed an LbL deposition technique to coat cotton threads with these cMOF materials for two primary reasons. First, LbL assembly offers nanoscale control and simplicity compared to other MOF deposition methods^45,46, as it enables precise control over coating thickness and uniformity without requiring high power input during deposition, making it well-suited for scalable, energy-efficient fabrication⁴⁷. Second, while previous studies have successfully applied LbL methods to grow insulating MOFs, such as HKUST-1^48,49 and MIL-100(Fe)⁵⁰ on textiles, as well as cMOFs on rigid substrates, including functionalized indium tin oxide (ITO)-coated glass⁵¹, gold-patterned electrodes⁵², titanium oxide nanowire arrays⁵³, sapphire or quartz⁵⁴, the deposition of highly crystalline, conductive MOFs onto flexible fibers via LbL remains limited^25,30,31. We, therefore, hypothesized that, with targeted optimization, the LbL approach could be adapted for conformal, continuous, robust, and conductive coatings of cMOFs on individual textile threads, a capability that has not yet been demonstrated.

Synthesis and characterization of cMOF@thread composites

Rigorous optimization efforts (Supplementary Method 1, Supplementary Fig. 1, and Supplementary Tables 1-5) suggested that plasma-cleaning of threads, followed by ten sequential cycles of soaking in the metal salt solution, washing with deionized water and ethanol, followed by soaking in the ligand solution and washing again, produced uniform conformal coatings with high crystallinity (Fig. 1a-c and Supplementary Fig. 2). The resulting cMOF@thread composites showed high MOF mass loadings, reaching up to 0.75 mg per cm of thread (Supplementary Fig. 3). The conformal coatings produced conductive threads with relatively low electrical resistance values ranging between 0.9 kΩ to 10 MΩ across 1-cm segments. These values remained stable under mechanical deformation, including repeated bending and twisting (Supplementary Fig. 4). In addition, the coated threads demonstrated strong adhesion, structural integrity, and mechanical robustness during exfoliation tests (Supplementary Fig. 5). Detailed synthetic and characterization procedures are provided in Supplementary Note 1. To highlight the critical role of the LbL method in achieving crystalline, conductive, continuous, conformal, and mechanically stable MOF coatings, we also attempted deposition using a conventional solvothermal approach (Supplementary Methods 2–3 and Supplementary Figs. 6 and 7). This method yielded poor coating uniformity, weak adhesion, and non-conductive thread composites, rendering it unsuitable for practical applications.

Fig. 1. Synthesis and characterization of cMOF@threads.

a Schematic illustration of the layer-by-layer (LbL) technique used to synthesize the cMOF@thread series. b Structure and packing of the respective MOFs. Color code: grey, carbon; violet, nitrogen; red, oxygen; green, nickel or copper; and blue, zinc or copper. c Powder X-ray diffraction (PXRD) patterns of the cMOF series coated on cotton threads after ten cycles of the LbL technique. The X-ray source is Cu Kα (λ = 1.54 Å). Color code: green, Ni₃(HITP)₂@thread; dark yellow, Cu₃(HITP)₂@thread; blue, Zn₃(HHTP)₂@thread; pink, Cu₃(HHTP)₂@thread; black, simulated patterns; grey, bare cotton thread. d Representative SEM images of the series of cMOF@cotton thread synthesized using LbL, herein false-colored to depict the different deposited cMOFs. Color code: green, Ni₃(HITP)₂@thread; dark yellow, Cu₃(HITP)₂@thread; blue, Zn₃(HHTP)₂@thread; pink, Cu₃(HHTP)₂@thread. SEM micrographs were obtained from 4 independent synthetic batches of MOF@threads, yielding similar observations. Source data are provided as a Source Data file.

Scanning electron microscopy (SEM) images, coupled with energy dispersive X-ray (EDX) spectroscopy and elemental mapping (Fig. 1d and Supplementary Figs. 8–16), confirmed the uniform distribution of MOF coatings along the full length of the thread. Cross-sectional imaging using focused ion beam (FIB)-SEM revealed conformal coatings on the cotton fibers, with average thicknesses between 1.4‒4.1 μm for the different cMOF@threads (Supplementary Fig. 17 and Supplementary Table 6). Attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy of the cMOF@thread composites showed characteristic vibrational peaks consistent with those of the corresponding bulk MOF powders, confirming the emergence of the expected molecular structure upon LbL deposition (Supplementary Fig. 18). Survey and high-resolution (Supplementary Figs. 19 and 20) X-ray photoelectron spectroscopy (XPS) verified the chemical composition and oxidation states of the metal nodes and ligands, revealing a nearly equimolar presence of C–N and C = N or C–O and C = O species in the cMOFs, consistent with the semiquinoidal nature of the ligands in these frameworks. For the metal centers, Ni₃(HITP)₂@thread and Zn₃(HHTP)₂@thread exhibited exclusively Ni(II) and Zn(II) oxidation states, respectively (Supplementary Fig. 20). In contrast, the copper-based MOFs showed mixed-valence states with Cu(I):Cu(II) ratios of 35:65 in Cu₃(HITP)₂@thread and 25:75 in Cu₃(HHTP)₂@thread (Supplementary Fig. 20), well in agreement with previous bulk-phase studies⁴³.

We evaluated the porosity of the MOFs upon integration onto cotton thread substrates, which showed significantly enhanced Brunauer-Emmett-Teller (BET) surface area from 0.1 m² g^-1 for bare cotton thread to 33.2, 17.1, 28.8, and 11.2 m² g^-1 for Ni₃(HITP)₂@thread, Cu₃(HITP)₂@thread, Zn₃(HHTP)₂@thread, and Cu₃(HHTP)₂@thread, respectively (Supplementary Fig. 21). Thermogravimetric analysis (TGA) confirmed the thermal stability of the MOFs upon integration within thread substrates, as evidenced by minimal weight loss for up to 300 °C (Supplementary Note 2 and Supplementary Fig. 22). Collectively, these characterization results confirmed the successful deposition of conformal, crystalline, porous, and thermally stable MOF coatings on cotton threads, establishing a robust foundation for their use in wearable sensing applications.

We evaluated the long-term stability of the threads over several weeks and up to ten months. We observed that the crystallinity and conductivity of the threads decreased over time, though to varying degrees for each material (Supplementary Fig. 23). While Cu₃(HITP)₂@thread composites lost their conductivity after 3 weeks, rendering them ineffective for the intended application, Ni₃(HITP)₂, Zn₃(HHTP)₂, and Cu₃(HHTP)₂@thread retained sufficient conductivity to enable chemiresistive sensing even after ten months of ambient benchtop storage.

The interactions occurring at the MOF‒fiber interface that have previously been discussed in the literature for the deposition of MOFs on fibers include hydrogen bonding, van der Waals interactions, electrostatic, and/or covalent interactions²⁵. In attempts to provide experimental evidence for these insights, and since direct characterization through ATR-FTIR or XPS were deemed inconclusive, we performed several control experiments to infer key interactions occurring during deposition (Supplementary Fig. 24). We chose the most conductive and crystalline composite, Ni₃(HITP)₂@thread, and adapted its synthesis on other types of fibers, including silk and polyester. These efforts proved successful, likely because these substrates possess similar electron-donating nitrogen and/or oxygen atoms that provide an anchor for coordination and hydrogen bonding interactions with the metal ions and triphenylene ligands during LbL deposition. Additional experiments designed to de-populate or block the hypothesized active sites on the fiber yielded deposition of coordination polymers or MOFs with diminished crystallinity (Supplementary Note 3 and Supplementary Method 4). These control experiments confirmed the importance of available electron-rich atoms on the surface of the fiber to form robust interactions with metal ions and/or ligands for the growth of crystalline MOF coatings.

Chemiresistive sensing of cMOF@thread composites

To evaluate the chemiresistive sensing capabilities of the cMOF@threads towards five hazardous gases, we aligned 1.5-cm-long coated threads inside a custom-made Teflon chamber fitted with gold-coated pins for electrical contacts at both ends and equipped with a gas inlet and outlet (Supplementary Fig. 25). After purging the chamber with nitrogen to stabilize the baseline, we exposed the threads to 2–80 ppm concentrations of H₂S, SO₂, NO, NH₃, and CO for 20 min, followed by a 30-minute recovery period in N₂ atmosphere (Fig. 2 and Supplementary Figs. 26–48). We reasoned that these concentrations would provide a sound notion of the cMOF@thread sensing performance in exposure scenarios relevant to personal protection and occupational safety⁵⁵.

Fig. 2. Sensing responses of cMOF@threads towards various gases in nitrogen.

Sensing traces of the cMOF@thread composites: Ni₃(HITP)₂@thread, Cu₃(HITP)₂@thread, Zn₃(HHTP)₂@thread, and Cu₃(HHTP)₂@thread towards (a–d) hydrogen sulfide (H₂S), e–h sulfur dioxide (SO₂), i–l nitric oxide (NO), m–p ammonia (NH₃), and q–t carbon monoxide (CO), respectively, at concentrations ranging between 2 to 80 ppm. All plotted sensing traces are averages of 3 biological replicate measurements. The shaded area between 10 and 30 min corresponds to the exposure period to each gas. The shaded error band curves in q and r correspond to the standard deviation of the 3 measurements. Color code: green, Ni₃(HITP)₂@thread; dark yellow, Cu₃(HITP)₂@thread; blue, Zn₃(HHTP)₂@thread; pink, Cu₃(HHTP)₂@thread with dark-to-light gradient colors corresponding to high-to-low analyte gas concentrations. Source data are provided as a Source Data file.

Generally, the trends in sensing responses were consistent with those previously reported for these MOFs in bulk form or integrated into other substrates^26,27,39,56. Plotting the chemiresistive responses, quantified as a negative normalized change in conductance (‒ΔG/G₀) over time revealed three key features. First, the cMOF@threads exhibited a direct concentration-dependent response with greater signal changes at higher gas concentrations. Second, the directionality of the sensing response varied depending on the cMOF and the gaseous analyte. While Ni₃(HITP)₂ and Cu₃(HHTP)₂ showed positive responses towards NH₃, Cu₃(HITP)₂ and Zn₃(HHTP)₂ exhibited negative responses to the same analyte, suggesting different interaction and/or charge transport mechanisms. Third, all sensors showed rapid initial responses, indicating fast MOF-analyte interactions, but each MOF displayed distinct saturation behaviors depending on the analyte (Supplementary Note 4 and Supplementary Tables 7–9). These observations prompted further evaluation of sensor performance, revealing i) low theoretical LODs with values of 43, 60, 5.6, 65, and 417 ppb for H₂S, SO₂, NO, NH₃, and CO, respectively, all below the PELs established by regulatory agencies³⁶ (Supplementary Note 4, Supplementary Figs. 49–52, and Supplementary Tables 10–13), and comparable to or better than other reported sensors (Supplementary Tables 14 and 15); ii) linear correlations between initial response rates (RoR) and analyte concentrations within the first minute of exposure (Supplementary Figs. 53–57); and iii) limited or no recyclability over multiple gas exposure cycles towards H₂S, SO₂ and NO, but moderate to strong reversibility towards NH₃ and CO (Supplementary Note 4 and Supplementary Figs. 44–48). The latter observations suggested a dosimetric sensing mechanism, likely driven by irreversible redox interactions and gas adsorption onto the MOF surfaces (see section Insights into material‒analyte interactions). Given the distinct and complementary responses of the four MOFs, we reasoned that combining them into an integrated sensor array should enable the selective detection and differentiation of H₂S, SO₂, NO, NH₃, and CO.

Principal component analysis (PCA)

To evaluate the selectivity and discrimination capability of the cMOF@thread sensor array, we applied PCA, an unsupervised dimensionality-reduction machine learning technique for multivariate data analysis (Supplementary Note 5)⁵⁷. As shown in Fig. 3a, b and Supplementary Fig. 58, PCA score plots revealed that the set of four thread-based sensors not only distinguished between the five chemically distinct analytes, but also differentiated them across varying concentrations (2–80 ppm). The cMOF@thread sensor array differentiated a broad range of concentrations using just the first two principal components (PCs). Though we have previously reported the utilization of supervised machine learning algorithms for binary H₂S/SO₂ mixtures on HHTP-based materials⁵⁸, the results reported herein underscore the robust capability of our cMOF@thread platform for multi-gas and concentration-specific detection in hazardous environments using PCA, an unsupervised machine learning algorithm.

Fig. 3. Principal component analysis and sensing under interference conditions.

a Principal component analysis (PCA) plots of extracted features (magnitude of response, area under the curve during exposure, and initial rate of response) from Ni₃(HITP)₂@thread, Cu₃(HITP)₂@thread, Zn₃(HHTP)₂@thread, and Cu₃(HHTP)₂@thread composites probed in triplicate upon 20-minute exposures to 2–80 ppm H₂S, SO₂, NO, NH₃, and CO in N₂. Each dot corresponds to one sample measurement, with a total of three sample measurements performed for each condition (combined in a larger circle) with larger-to-smaller size dots corresponding to high-to-low analyte gas concentrations. Color code: light brown, H₂S; turquoise, SO₂; yellow, NO; purple, NH₃; grey, CO. b Zoomed-in region of the PCA plot highlighting lower-concentration details. c Bar graph showing an exemplary cMOF, Ni₃(HITP)₂@thread, exhibiting comparable response magnitudes when exposed to the gases under air and varying humidity conditions (20–100% RH). Error bars refer to the standard deviation of 3 biological replicate measurements. The dot plot circles correspond to each value of the 3 individual measurements. Color code: white bar, N₂ background gas; patterned bar, air background gas; green bars, humid N₂ background gas with gradient light-to-dark shades indicating various humidity levels ranging from low (20%RH) to high (100%RH) humidity. Source data are provided as a Source Data file.

Sensing for aged samples and under interference conditions

To elucidate the ability of the MOF@thread array in retaining performance upon long-term storage or in complex environments, we performed various sensing experiments on aged samples, in air, and in humid environments. MOF@thread composites stored under ambient conditions and aged for ten months showed normalized sensing responses that were mostly comparable to the sensing traces of freshly synthesized threads (Supplementary Figs. 59–63). Notable exceptions include a decrease in the sensitivity of some cMOFs to certain gases, such as the case for Zn₃(HHTP)₂@thread towards H₂S and SO₂. This diminished performance may be attributed to the possible saturation of active sites by water adsorbed from the surrounding atmosphere, as well as to the partial degradation of the material over time, as evidenced by its decreased crystallinity⁵⁹. Despite that, superimposing the features from these sensing traces onto the PCA plot provided efficient differentiation of the gas identity, although with somewhat diminished precision in concentration-dependent differentiation (Supplementary Figs. 59–63).

To mimic typical ambient conditions, we performed sensing experiments on cMOF@threads in an air background (Fig. 3c and Supplementary Figs. 64–68). While some cMOF@threads exhibited a decrease in sensing magnitude or reverse in direction, likely due to oxygen binding and competitive interactions⁶⁰, PCA analysis revealed the ability of the thread array to maintain its distinguishing performance (Supplementary Figs. 64–68). Following that, we evaluated the performance of the thread-array under varying humidity levels (Supplementary Figs. 69–73) of 20, 40, 60, 80, and 100% relative humidity (RH). Overall, most sensing responses remained robust despite the strong interference from water vapor, as illustrated for Ni₃(HITP)₂@thread in Fig. 3c. However, some responses decreased in magnitude—for example, Cu₃(HITP)₂@thread toward NO at 60% RH—or reversed sign, as observed for Cu₃(HHTP)₂@thread toward H₂S, SO₂, and NH₃ across several humidity levels. The reduced responses likely arose from competitive adsorption, where water molecules occupy open metal sites that serve as gas-binding and signal-generating centers. The sign reversals in sensing signals align with previous reports where conductivity may be perturbed by proton doping or transport^28,59–61. Despite these effects, PCA of the array outputs retained the ability to differentiate the gases, with the spatial clustering of superimposed signals mapping cleanly onto gas identities.

Insights into material‒analyte interactions

To gain insights into the interactions occurring between the cMOFs and gaseous analytes, we conducted a series of microscopic and spectroscopic characterizations on the cMOF@thread composites after a 2-hour exposure to 1% concentrations of each gas in nitrogen (Supplementary Figs. 74–88), as well as in situ diffuse reflectance infrared transform spectroscopy (DRIFTS) on their bulk counterparts (Supplementary Figs. 89–93). We summarized key material‒analyte interactions, deduced from various techniques—including some previously reported in the literature^{27,28,38,43,58,62}—and complemented these insights with additional experiments on previously unreported systems, as detailed in Supplementary Tables 16–20. PXRD patterns obtained after the sensing experiments showed retainment of cMOF@thread crystallinity (Supplementary Fig. 74). SEM micrographs confirmed the retention of the morphological structures of the MOFs upon exposure to the high concentration of gaseous analytes (Supplementary Figs. 75–78), and EDX elemental mapping revealed the adsorption of analytes onto the MOFs (Supplementary Note 6 and Supplementary Figs. 79–83). To better understand the chemical nature and oxidation-state changes resulting from gas exposure, we performed ex situ XPS analyses of the cMOF@threads.

Upon exposure to H₂S, XPS analysis revealed the formation of various sulfur species on the cMOF@thread surfaces, including copper (I) sulfide (Cu₂S at 161.8 eV)⁶³, copper-deficient non-stoichiometric sulfide (Cu_xS_y at 162.8 eV)²⁷, zinc sulfide (ZnS₂ at 161.8 eV and ZnS at 162.8 eV)^64,65, polysulfide (S_x at 163.8 eV)⁶⁶, H₂S‒M (at 165.5 eV)⁶⁷, sulfite (SO₃^2- at 167.2 eV)⁶⁸, and sulfate species (SO₄^2- at 168.7 eV)⁶⁸, shown in Fig. 4a. These redox transformations were accompanied by i) significant reductions in the copper oxidation state for the Cu-based MOFs with decreased Cu(II):Cu(I) ratios; ii) reduction of ligands in HITP-based MOFs with declined C = NH:C–NH ratio; and iii) oxidation of ligands in the HHTP-based MOFs with increased C = O:C–O ratios (Supplementary Fig. 84). In contrast, exposure of cMOF@threads to SO₂ resulted in the formation of SO₃^2- and SO₄^2- (Fig. 4b)⁶⁸, as the sulfur-containing species. Accounting for this redox change, the copper nodes and HITP moieties underwent partial reduction (Supplementary Fig. 85). DRIFTS experiments on the powder MOFs showed strong decreases in background absorbance upon exposure to both sulfurous gases, indicating significant electronic changes to the material. These changes were accompanied by perturbations to HITP- and HHTP-ligand vibrational bands, which supported the redox changes observed in complementary ex situ XPS analyses on the coated fibers (Supplementary Figs. 89 and 90 and Supplementary Tables 16 and 17).

Fig. 4. Insights into material‒analyte interactions.

High-resolution X-ray photoelectron spectroscopy (XPS) spectra at selected regions for all cMOF@thread samples before and after exposure to 1% of the gaseous analyte in N₂ for 2 h: a S 2p region after H₂S exposure, b S 2p region after SO₂ exposure, and c N1s region after NO exposure. Source data are provided as a Source Data file.

For nitrogen-based gases, we detected adsorbed NO (peak at 401.8 eV)⁶⁹ and identified the formation of nitrate (NO₃^‒ at 406.7 eV)⁷⁰ on HHTP-based MOFs and the partial conversion of C = NH to NO₃^‒ in HITP-based MOFs upon NO exposure, as confirmed by both XPS and DRIFTS (Fig. 4c, Supplementary Figs. 86 and 91, and Supplementary Table 18). However, the oxidation states of the metal centers in all MOFs showed minimal changes (Δ0‒3%) after NO exposure (Supplementary Fig. 86). Conversely, exposure to NH₃ resulted in a reduction of the HITP-based frameworks, as evidenced by a substantial decrease in their C = NH:C–NH ratios (Δ11‒15%), while HHTP-based MOFs exhibited new N 1 s peaks at 397.7 and 399.7 eV, corresponding to dissociated NH_x (x = 1, 2) species⁷¹ and to Lewis acidic metal–ammonia (M‒NH₃) interactions⁴³, respectively (Supplementary Fig. 87). DRIFTS experiments during NH₃ exposure showed the presence of Lewis acid-site (M‒NH₃) interactions for all four MOFs, but basic sites for the HHTP-based MOFs, complementing the XPS results (Supplementary Fig. 92 and Supplementary Table 19)⁴³.

Finally, CO exposure led to partial oxidation of the HHTP ligands in both HHTP-based MOFs, reflected in an increased C = O:C–O ratios of Δ6‒8% (Supplementary Fig. 88), which we attributed to CO adsorption at the metal centers⁷². Likewise, DRIFTS experiments showed perturbations to the ligand for HHTP-based MOFs, unlike the immeasurable change in signal for HITP-based MOFs (Supplementary Fig. 93 and Supplementary Table 20). Due to the lack of sensitivity of the HITP-based MOFs towards CO, we did not perform post-exposure analyses on their respective threads. Collectively, these in situ and ex situ analyses revealed that the chemical reactivity of the MOF ligands and metal centers governed the gas-specific redox transformations, establishing a clear structure‒property relationship between the framework composition and its selective sensing behavior toward toxic analytes.

Sensor array prototype

To evaluate the potential of integrating the cMOF@thread composites into wearable sensors, we fabricated a multi-thread sensor array onto a cotton textile and incorporated into a face mask or lab coat, by sewing the sensor threads with silver-plated conductive threads (Supplementary Note 7 and Fig. 5a). We ensured stable Ohmic electrical connections between the sensor and the metal-plated thread to allow for reliable transduction of electronic signals across ends (Supplementary Figs. 94–95). Integrating multiple sensor arrays into a single reference/counter electrode configuration retained Ohmic contacts, demonstrating design versatility (Fig. 5b and Supplementary Figs. 96–99). We placed the threaded patch into an enclosed plastic chamber, equipped with a gas inlet and outlet, connected the silver-plated threads to a potentiostat (Supplementary Fig. 100), and exposed the device to 40 ppm of each analyte gas. The observed trends and magnitudes of response closely matched those obtained in the controlled Teflon chamber setup (Supplementary Fig. 101), underscoring the adaptability and consistency of sensing properties across different configurations within this technological platform. The sensor-array patch recovered its sensing performance towards H₂S, SO₂, and NO after saturation upon a water-soaking and drying treatment for three consecutive cycles (Fig. 5c, Supplementary Note 8, and Supplementary Figs. 102–105), despite the increased resistances of the threads due to the irreversible redox processes occurring upon interactions with analytes, as detailed in the material‒analyte interaction study. Other laundering methods, which have the potential to further enhance the regeneration of irreversible sensing signals from Cu₃(HHTP)₂@textile swatches, include washing with detergent and strong oxidants, such as hydrogen peroxide, and bleach²⁷. This result not only highlights the structural robustness of the thread array upon soaking in water, but also affirms its practical reusability via simple signal regeneration.

Fig. 5. Prototype design, wearability, and performance.

a Prototype array device patch of the four different cMOF@thread sensors sewn onto a face mask and a lab coat. b Various connections showing the versatility of the thread array sewn on cotton textile swatches. Bar graphs comparing sensing responses towards 40 ppm of NO c upon washing the device in water for two times, and d a hidden prototype, shown as an inset, which offered non-line-of-sight detection. Error bars represent the standard deviation of 3 biological replicates, with dots showing each measurement. Color code: green, Ni₃(HITP)₂@thread; dark yellow, Cu₃(HITP)₂@thread; blue, Zn₃(HHTP)₂@thread; pink, Cu₃(HHTP)₂@thread. e PCA structure from the MOF@thread base dataset generated in this study (Fig. 3a), onto which the calculated data points of the MOF@thread array sensing responses in the prototype format after 20 min of exposure to 40 ppm of gases in N₂ are superimposed (see Supplementary Note 5). Legend: H₂S (diamond), SO₂ (square grid), NO (yellow right-pointed triangle), NH₃ (purple upward-pointed triangle) in prototype format, in addition to NO in non-line-of-sight detection (star), NO after first wash (black down-pointing triangle), and NO after second wash (yellow down-pointing triangle). Source data are provided as a Source Data file.

We further demonstrated the prototype's capability for non-line-of-sight detection, wherein the thread array exhibited responses comparable to NO even when concealed between layers of textile (Fig. 5d and Supplementary Fig. 106), thus highlighting its functionality when embedded within wearable fabrics. Under these various conditions, the thread array retained its discriminatory power, as evident from the distinct clustering of responses in the overlaid PCA plot upon washing, under humidity, and within non-line-of-sight detection, thus affirming the robustness and reliability of this device format (Fig. 5e).

Discussion

In summary, we have reported the successful growth of a suite of conductive MOFs onto fiber thread substrates with high crystallinity and unprecedented adhesion via an LbL deposition method. The resulting cMOF@threads exhibited excellent chemiresistive sensing performance toward five toxic gases, achieving high sensitivity and low ppb-level LODs, all below OSHA’s PELs. Using four distinct cMOF@threads as a sensor array, we demonstrated effective detection and discrimination of these gases across the broadest concentration range with just two PCs and provided mechanistic insights into their molecular interactions with the cMOFs through comprehensive in situ and ex situ analyses. We further validated the practical applicability of these sensors by integrating the cMOF@threads into a flexible textile swatch. The prototype maintained sensing functionality in various conditions, including non-line-of-sight detection, post-washing cycles, and in complex environments containing air and humidity. We also demonstrated the platform's high modularity by integrating multiple designs and configurations into common PPE.

This work showcases a cMOF-coated thread array capable of both “smelling” (by selectively detecting and differentiating analyte gases) and “communicating” (by electrically transducing signals through silver-plated conductive threads). This platform represents a step forward in the multifunctionality of fiber-based devices. By retaining high sensing performance in a miniaturized footprint (0.5-cm thread segment), this system exemplifies Moore’s law for fibers⁷³, achieving greater functionality per unit length and advancing the push toward ultra-compact, efficient sensing fibers. Altogether, this work establishes MOF-coated threads as a transformative platform for next-generation wearable sensors, offering modular miniaturization, molecular-level selectivity and sensitivity, and seamless integration into PPE for advanced environmental monitoring and protection.

Methods

Materials

2,3,6,7,10,11-Hexahydroxytriphenylene (HHTP, CAS number 4877-80-9) and 2,3,6,7,10,11-hexaaminotriphenylene hexahydrochloride (HATP•6HCl, CAS number 1350518-27-2) were purchased from Ambeed. Nickel (II) acetate tetrahydrate (Ni(OAc)₂•4H₂O; CAS number 6018-89-9) was acquired from TCI America, while copper (II) acetate anhydrous (Cu(OAc)₂; CAS number 142-71-2) and zinc (II) acetate dihydrate (Zn(OAc)₂•2H₂O; CAS number 5970-45-6) were purchased from Thermo-Fischer Scientific. Anhydrous sodium acetate (NaOAc; CAS number 127-09-3) was purchased from VWR. Reagents, including alcohol or ethanol and N, N-dimethylformamide (DMF), were purchased from Fisher, and deionized (DI) water was obtained from a Milli-Q (IQ 7000) water filtration system. White natural cotton threads were purchased from Gütermann. Threads were initially cleaned in a Harrick Basic Plasma Cleaner (PDC-32G) prior to MOF deposition. Analyte gases (H₂S, SO₂, NO, NH₃, and CO) were purchased from Airgas as tanks constituting 10,000 ppm gas analyte in nitrogen. All figures were drawn using Origin 9.0 or Origin 2025b. PCA plots were generated through Origin 2025b (see Supplementary Note 5 for details).

PXRD

Powder X-ray Diffraction (PXRD) diffraction patterns were recorded on a Rigaku sixth-generation MiniFlex X-ray diffractometer with a Cu Kα (600 W, 40 kV, 15 mA, λ = 1.54 Å) radiation source using MiniFlex Guidance software. PXRD analysis was performed on a PDXL2 software interface. Three 3-cm-long threads were each cut into three pieces. Onto a zero-background diffraction silicon holder, petroleum jelly was used to aid in adhering the MOF-coated threads (a total of 9 1-cm threads) to the plate. The threads were then discarded (not used for further testing or characterization).

SEM and EDX

Scanning Electron Microscopy (SEM) images were taken on a Thermo Scientific™ Helios™ 5 CX DualBeam microscope. Energy Dispersive X-ray (EDX) spectra were recorded using an Ultim Extreme 100 Oxford Instrument X-ray detector.

ATR-FTIR

Attenuated total reflectance (ATR) Fourier Transform – Infrared (FTIR) spectra were collected on a Nicolet iS50 FT-IR spectrophotometer using OMNIC software.

BET surface area analysis

Nitrogen sorption experiments were conducted on a 3Flex (Micrometrics, Norcross, Georgia) instrument at 77K, with the data collected on 3Flex version 5.02. BET measurements were performed on a total mass of around 50 mg of cMOF@thread composites. For these measurements, MOF coatings were deposited on 10-cm long threads rather than the regular conditions of 3-cm long threads for easier handling within the BET tubes. All synthesis conditions and parameters were similar to the optimized conditions.

TGA

Thermogravimetric Analysis (TGA) plots were recorded on a TGA 55 instrument from 25 to 900 °C with a ramp rate of 5 °C/min under nitrogen as purge gas using the TRIOS software.

XPS

X-ray photoelectron spectroscopy (XPS) experiments were performed on a Kratos Axis Supra XPS at a pressure of 10^-9Torr, using a 110 μm aperture to narrow the detector focus. The XPS data were analyzed using ESCApe software.

Pretreatment of threads

Initial cleaning of threads consisted of soaking and sonicating 3-cm cotton threads in water, then ethanol, then acetone, followed by drying in an oven at 45 °C. Threads were then subjected to air plasma cleaning at a high-power setting for one minute.

MOF deposition on threads through optimized LbL procedure

Coating of MOFs on threads (10 × 3 cm) using LbL consisted of 10 consecutive full cycles of: 1) soaking in metal ion solution (detailed below) for 2 min, 2) drying in oven at 45 °C for 15 min, 3) washing via rigorous dipping in water then in ethanol, 4) drying in oven at 45 °C for 5 min, 5) soaking in ligand solution (detailed below) for 2 min, 6) drying in oven at 45 °C for 30 min, 7) washing via rigorous dipping in water then in ethanol, and 8) drying in oven at 45 °C for 5 min.

Optimized metal ion solutions were prepared at a concentration of 100 mM using the respective metal acetate salts in water. Optimized ligand solutions were prepared at a concentration of 10 mM of the respective ligand in (1:1) H₂O:DMF with 100 eq of sodium acetate for Ni₃(HITP)₂@thread; in (1:1) H₂O:DMF with 100 eq of sodium acetate for Cu₃(HITP)₂@thread; in (1:1) H₂O:EtOH with 100 eq of sodium acetate for Zn₃(HHTP)₂@thread; and in (1:1) H₂O:EtOH for Cu₃(HHTP)₂@thread. For more information, see Supplementary Method 1.

Sensing setup and experimental details

Prior to sensing experiments, the coated threads were heated in an oven at 50 °C overnight. First, threads were fitted into the Teflon chamber to form a stable electrical connection with the gold pins upon enclosure (Supplementary Fig. 25a). The Teflon chamber was then fitted into the sensing setup through which the analyte gas was delivered at a controlled concentration. To achieve desired concentrations of analyte gas, we controlled the flow of purchased analyte gas tanks through a low-flow Micro-Trak (Sierra Instruments, Inc.) mass flow controller (MFC), mixed with the background gas passed through a high-flow Smart-Trak (Sierra Instruments, Inc.) MFC (Supplementary Fig. 25b). By controlling the rate at which each of the MFCs was delivering the respective gas, specific concentrations of analyte gas can be attained. Sensing experiments were recorded as current values over time, which were collected using a PalmSens MUX8-R2 potentiostat on PSTrace 5.8 or 5.9 software. Sensing under humid conditions was achieved using a KIN-TEK Analytical FlexStream™ Gas Standard Generator System composed of a base module (FlexBase) and a humidification module (FlexHG).

The threads were first equilibrated with a stream of nitrogen gas for 2 to 6 hours until a straight baseline was achieved prior to recording. Each recorded measurement included 10 min of background gas flow, followed by 20 min exposure to the analyte gas at the desired concentration, and a 30-minute recovery period where the analyte mass flow controller was switched off, and the threads were only subjected to nitrogen gas. All measurements were performed at a constant driving voltage of 1 V. Sensing measurements in the chamber consisted of six thread sensors, while measurements using the prototype design consisted of four thread sensors, with an average device yield of approximately 70% from which reasonable sensing responses could be taken into account (i.e. with low signal-to-noise ratio, consistent connectivity throughout entire measurement, and reliable magnitudes of sensing responses). In this work, for the sake of consistency in regards to principal component analysis measures (see Supplementary Note 5), only three responses for each measurement were taken into consideration for averaged values, their standard deviations, and features extracted. For sensing under humid conditions, the N₂ background gas flow was directly humidified through the gas generator system, and the percentage relative humidity (% RH) values were in the error range of ± 3%.

The change in current of different threads was normalized according to following equation: $-△\mathrm{G}/{\mathrm{G}}_0=-\frac{\mathrm{I}-{\mathrm{I}}_0}{{\mathrm{I}}_0}\times 100$, where ‒ΔG/G₀ (in %) is the normalized response, I (in µA) represents the current at a certain time, and I₀ (in µA) is the initial current directly before gas exposure.

DRIFTS setup and experimental details

Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS) data were acquired using a Nicolet 6700 FT-IR spectrometer and collected on the OMNIC software. For data collection, a homogeneous composite of potassium bromide (KBr) with the respective MOF powder was placed in a sample cup in an air-tight, sealed steel chamber with gas inlet and outlet ports and KBr windows to allow an infrared beam path. The composite was activated by heating to 110 °C under nitrogen flow for about 1h in the sample cup, followed by cooling. Each MOF/KBr composite was then subjected to a single beam (unsubtracted), which was used as the initial spectrum, while accounting for a background spectrum of dry KBr. Then, difference spectra (Kubelka-Munk) were collected by subtracting spectra collected during exposure from the initial spectrum. Each of these spectra were collected with 32 scans from 400 to 4000 cm^-1. The reported spectra were taken every 5 min for 30 min of exposure to gaseous analyte, followed by recovery spectra while purging with nitrogen gas after 30 min of recovery in N₂. Due to low signal of gases at the concentrations at which sensing experiments were performed, we opted to use higher concentrations of gases for the DRIFTS experiments, by which 1% of H₂S, SO₂, NO, NH₃, and CO in nitrogen were utilized. For NO experiments, NO gas was passed through a dry tube containing soda lime (activated overnight at 80 °C under nitrogen) to eliminate/decrease the presence of nitrogen dioxide impurities from the NO gas tank, as they can lead to the formation of undesired potassium nitrate that overlaps with the IR peaks under study.

Human participants research

The Committee for the Protection of Human Subjects at Dartmouth College determined that this project falls under Not Human Subjects Research. The human research participant in Fig. 5a has given informed consent and agreed to model the wearability of the prototype.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

E.O.S. and K.A.M conceived and designed the project. E.O.S. optimized the LbL synthesis conditions, performed the sensing experiments, designed the prototype, and wrote the manuscript. E.O.S. and P.D. performed the characterization and material‒analyte interpretations. F.J.M. assisted with LbL deposition and prototype optimization. K.A.M. guided and supervised the research, and contributed to data interpretation and manuscript preparation. All authors discussed the results and data interpretation and provided feedback on the manuscript at all stages.

Peer review

Peer review information

Nature Communications thanks Meng Gao, Shun Mao, Young-Moo Jo, and Hongye Yuan for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data generated in this study are provided with this paper.

Competing interests

K.A.M. declares a financial interest in MRIA, a start-up company focused on the commercialization of sensor technologies based on metal–organic frameworks. E.O.S., P.D., and K.A.M. also declare financial interest as co-inventors on pending patents pertaining to this work.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72267-1.