Wearable Sensors for Breath Monitoring Based on Water‐Based Hexagonal Boron Nitride Inks Made with Supramolecular Functionalization

Wearable humidity sensors are attracting strong attention as they allow for real‐time and continuous monitoring of important physiological information by enabling activity tracking as well as air quality assessment. Amongst 2Dimensional (2D) materials, graphene oxide (GO) is very attractive for humidity sensing due to its tuneable surface chemistry, high surface area, processability in water, and easy integration onto flexible substrates. However, strong hysteresis, low sensitivity, and cross‐sensitivity issues limit the use of GO in practical applications, where continuous monitoring is preferred. Herein, a wearable and wireless impedance‐based humidity sensor made with pyrene‐functionalized hexagonal boron nitride (h‐BN) nanosheets is demonstrated. The device shows enhanced sensitivity towards relative humidity (RH) (>1010 Ohms/%RH in the range from 5% to 100% RH), fast response (0.1 ms), no appreciable hysteresis, and no cross‐sensitivity with temperature in the range of 25–60 °C. The h‐BN‐based sensor is able to monitor the whole breathing cycle process of exhaling and inhaling, hence enabling to record in real‐time the subtlest changes of respiratory signals associated with different daily activities as well as various symptoms of flu, without requiring any direct contact with the individual.


Introduction
[3][4][5] In particular, RH sensors are very attractive as they are typically used for breath monitoring, touch-free skin examination, non-contact switching, and spatial localization monitoring. [6]These sensors must have high sensitivity, fast response and recovery, good stability, longterm durability, high linearity response, low hysteresis, and must be easily integrated onto flexible substrates with low-cost techniques.][10][11] 2D materials are very attractive for the development of RH sensors because their large surface area enables more active sites for the gas molecules to get adsorbed, hence making them highly sensitive to changes in the environment.14] Furthermore, 2D crystals can be easily made by solution processing, [15][16][17][18] by allowing the fabrication of devices with low-cost and mass scalable methods. [19][22][23][24] However, sensitivity is not the only parameter to take into account when developing a wearable humidity sensor: the device must be robust against different environmental conditions and provide reproducible signals.It is well known that GO and rGO-based humidity sensors suffer from large hysteresis [25] and they can operate only in a small range of temperatures, as GO has limited thermal stability in air. [20]Therefore, these sensors need to be disposed after a single use.In addition, the resistive readout is also subjected to strong cross-sensitivity issues, as multiple stimuli (e.g., simultaneous change in humidity, strain, temperature, etc.) will all contribute to a change in resistance. [25]ther solution-processed 2D materials used for RH sensing include 1T tungsten disulfide (WS 2 ), [26] Mxene, [27] black phosphorous (BP), [21] and titanium disulfide (TiS 2 ); [28] however, the sensors still suffer from poor stability and low sensitivity.Table S1, Supporting Information shows the results obtained so far for RH sensors made of solution-processed 2D materials, including other materials, such as polymers.
In this work, we demonstrate a wearable and wireless humidity sensor made with pyrene-functionalized hexagonal boron nitride (h-BN)-based ink.[32] In particular, solution-processed h-BN enables simple and low-cost production of the material as well as fabrication and integration of the sensor onto flexible substrates, hence providing strong advantages as compared to h-BN grown by chemical vapor deposition.h-BN inks have been successfully used in printed capacitors [33][34][35][36][37] and transistors, [38][39][40][41] while their exploitation in sensing has been very limited because of h-BN's high resistance and poor selectivity, due to its inert nature. [8]In particular, h-BN is difficult to implement in RH sensors because the material is hydrophobic, so its sensitivity and selectivity to water molecules are very poor.So far, only a very few works [42][43][44] have reported the use of solution-processed h-BN for humidity sensing: pristine h-BN nanosheets have shown a sensitivity of 28 384% at 85% RH, but relatively slow response/recovery time and poor sensitivity (<1000%) below 50% RH. [43]To enhance the sensitivity to water, h-BN is typically mixed with a polymer, such as polyethylene oxide or polyvinylpyrrolidone, which acts as RH sensing element.Alternatively, a 3D porous h-BN-based material made by a solid-state approach with the assistance of boron trioxide has been proposed for RH sensors. [44]However, the material preparation does require sintering at the temperature of 1000 °C, which is not sustainable and compatible with many substrates.
Herein, we show a new strategy to enhance the sensitivity of the h-BN ink to water molecules, based on a simple one-pot supramolecular functionalization approach, without using any polymer.We demonstrate a wearable and wireless impedance-based humidity sensor with enhanced sensitivity (>10 10 Ohms per %RH in the range from 5% to 100%RH), fast response (0.1 ms), no appreciable hysteresis, and no temperature cross-sensitivity in the range of 25-60 °C.We finally show that this impedance-based device can be used as a wearable and wireless breathing sensor, and it is able to record signals in real-time from different individuals in daily activities as well as common symptoms of COVID-19 and flu.
The high sensitivity, in addition to lack of cross-sensitivity from changes in temperature, as well as fast response and recovery time, make the h-BN ink, produced with our one-pot supramolecular approach, extremely attractive for the next generation of wearable breathing sensors for healthcare monitoring.

Results and Discussion
The h-BN nanosheets were prepared via liquid phase exfoliation (LPE) [16] in water.To note that h-BN is hydrophobic, making LPE in water very challenging.Nevertheless, it has been reported that h-BN can be exfoliated in water because the sonication treatment can turn h-BN into hydroxyl-functionalized h-BN, due to hydrolysis in aqueous solution. [45]While this functionalization route makes the material sensitive to water molecules, [43] the maximum concentration of h-BN is usually below 0.1 mg ml −1 .Similarly, ball milling in aqueous alkaline solution [46] has been used to demonstrate scalable hydroxyl-functionalized h-BN with a maximum concentration of ≈0.2 mg ml −1 in water, which is still low for practical applications.
In order to enhance the exfoliation of h-BN in water, dispersant/exfoliating molecules such as surfactants, can be added during LPE. [47]In our work, we use 1-pyrenesulfonic acid sodium salt (PS1) as a stabilizer: the pyrene core of PS1 interacts with the nanosheet, while its sulfonic group enables interaction with water molecules, hence allowing the formation of h-BN dispersions in water via electrostatic stabilization. [48]The schematic of the exfoliation process is shown in Figure 1a.This supramolecular functionalization approach enables us to prepare highly concentrated (up to 2 mg ml −1 ) as well as stable (for several months) h-BN dispersions in water.The stabilization is provided by the non-covalent functionalization with the PS1 molecules -indeed, the lack of re-aggregation of the material is fingerprint of the affinity between the functionalized h-BN nanosheets and the water molecules, and therefore of the possibility to use this type of h-BN ink for RH sensing.][50] Note that the thickness cannot be directly converted into the number of layers because of the adsorbed PS1 molecules.The residual amount of PS1 is less than 3 at.%,as measured by X-ray photoelectron spectroscopy. [48]he sensor was fabricated by drop casting ≈50 μL of the obtained h-BN dispersion onto Ni/Au interdigitated electrodes (Figure 1b), giving rise to a film of ≈50 nm in thickness.The material was left to dry in the air and then the humidity response was collected by an impedance analyzer (Figure 1b).For comparison, we also tested pristine h-BN nanosheets produced by LPE in water without using any stabilizer and commercial GO (see Experimental Section for details).
Using the equivalent RC parallel circuit (Section S5, Supporting Information), the equivalent impedance (Z) is given by the following equation. 21) where  is the frequency, C is the total capacitance, and R is the equivalent resistance.The impedance can be split into its real and imaginary parts, Z real and Z imag , respectively, as described by the following equations: (2) While the modulus of the impedance (|Z|) is: Figure 2a,b show that both Z imag and Z real are sensitive to changes in humidity produced by exhaling onto the device.The time interval for response/recovery duration of the device is 0.1 ms.The observed response and recovery time during breathing are 0.15 s and 0.09 s, respectively.In contrast, no changes in both imaginary and real parts of the impedance are observed in the case of pristine h-BN (Figure 2a,b).Furthermore, energy-dispersive Xray spectroscopy results shown in Figure S3 and Table S2, Supporting Information confirm the existence of B, N, and O elements, and clearly demonstrate that the proportion of O element increases from 5.84% to 12.19% as RH increases from 20% to 90%, suggesting the formation of hydrogen bonds between the functionalised h-BN and water molecules.These observations show that the non-covalent functionalization with PS1 enhances the h-BN sensitivity to water molecules.This can be attributed to the sulfonic group of PS1 that allows the h-BN nanosheet to interact with the water molecules via hydrogen bonding (Figure 2c).Density functional theory calculations were performed to get further insights into the sensing mechanism (Section S3, Supporting Information).As shown in Figure S4, Supporting Information, the band structure of PS1/h-BN features hybrid electronic states, with the lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) states (flat bands) of the PS1 molecule located inside the band gap of h-BN.Compared with pure h-BN, these in-gap states significantly altered the band alignment behaviors between PS1/h-BN and H 2 O in two ways: first, the HOMO state from PS1 is lower than the LUMO state from H 2 O, providing an easy pathway for charge transfer, and second, these empty states from PS1 are highly populated, generating the chemical potential for electron affinity between H 2 O and PS1/h-BN.This shows that the driving force of the electron affinity comes from the positively charged S states from PS1 and negatively charged O states from H 2 O.The calculated deformation charge density (Figure S5 and Table S3, Supporting Information) indicates the charge transfer from the H of the water molecules to the O of the PS1 molecules on the h-BN surface through hydrogen bonding.
In order to elucidate the sensing mechanism, the changes in resistance were directly measured by using a multimeter.The capacitance was measured by using the impedance analyzer, while changing the humidity in a controlled environmental chamber kept at 20 °C, and was performed at 10 kHz as this frequency provides a good signal-to-noise ratio, as shown in Figure S6, Supporting Information.Figure 2d shows that the resistance decreases by more than six orders of magnitude when the RH increases from 5% to 100%, giving rise to a sensitivity of 1.8 × 10 10 Ohms/%RH (Figure S7a, Supporting Information).Figure 2e shows that the capacitance increases exponentially starting from RH > 50%, possibly because only above this humidity level there is enough water adsorbed in the h-BN film to cause an appreciable change in the dielectric constant (to note that the dielectric constant of water is 78.5, [51] hence much higher than that of h-BN, which is ≈4 [34] ).This results in a sensitivity of 0.75 pF/%RH in the range of 50% to 100%RH (Figure S7b Using the equivalent RC parallel circuit (Figure S9, Supporting Information), the decrease in |Z| with increasing RH, Figure 2f, can be ascribed to a decrease in resistance and/or an increase in capacitance (Section S4, Supporting Information).The measurements show that the resistance decreases, while the capacitance exponentially increases for increasing RH above 50%, giving rise to a |Z| sensitivity of 3.1 × 10 5 Ohms per % RH for RH > 50% (Figure S8a, Supporting Information).The decrease of |Z| with increasing RH is in agreement with results obtained with other hydrophilic materials, such as polyethylene oxide-based thin films. [42,52]Exhaling on the device increases the RH leading to a decrease in the value of |Z|, suggesting the RH level during exhaling is near 100% (or at least much higher than 50%), in agreement with previous works. [53]n order to get further insights on the electrical readout, the equivalent RC parallel circuit was simulated by using the COM-SOL software package (see details in Table S4 and Figure S11, Supporting Information).Figure 3a shows the electric potential distribution on the surface of the interdigitated electrodes, suggesting the large effect of the electric field on sensing electrical signal output via capacitance.The calculated capacitance variation range (4-52 pf) from the COMSOL simulation is consistent with the experimental results (Figure S12, Supporting  S1, Supporting Information).d) Change of |Z| with time over 11 breathing cycles as compared with the relative theoretical impedance changes, calculated by using Equations ( 2) and ( 3).Changes in the e) real and f) imaginary parts of the impedance during deep breath monitoring by volunteer 2, showing the characteristic V shape signal associated with the changes in RH during the breathing cycle.
Information).The calculated Z real increases by four orders of magnitude in the range from 0% to 68%RH, reaches the maximum value at ≈68%RH and then decreases by two orders of magnitude in the range from 68% to 100%RH (blue line in Figure 3b).The maximum sensitivity of Z real is obtained at ≈68% RH (blue line in Figure S8b, Supporting Information).The calculated Z imag decreases by three orders of magnitude between 50% to 100% RH, giving rise to a sensitivity of 3.1 × 10 5 Ohms per %RH in this range.These results indicate that changes in Z real are dominated by the changes in resistance for RH < 68%, while changes in both resistance and capacitance affect Z real for RH > 68%.In contrast, Z imag is dominated by the changes in capacitance.Figure 3d and Figure S13, Supporting Information show the measured (line) and calculated change (black points) in |Z|, Z real , and Z imag during 11 cycles of breathing, demonstrating very good agreement between experimental and calculated results.
Because of the particular dependence of Z real on both resistance and capacitance, the Z real signal shows a characteristic "V" shape (Figure 3e) in every breathing cycle.The "V" shape is caused by the changes of RH in the exhaling-inhaling breathing cycle: exhaling causes the RH to go above 68%, hence causing Z real to first increase from the initial value at RH ≈20% to 68% and then to decrease till the maximum RH is achieved in the exhaling process.In contrast, in the inhaling process (causing RH < 68%) Z real increases when RH decreases from maximum RH to 68%, and then decreases after RH goes below 68%.The "V" shape is therefore fingerprint of the exhaling-inhaling process, and hence, it can be used to monitor human breathing during different activities or conditions.In contrast, the "V" shape is not observed when monitoring Z imag during the breathing cycle (Figure 3f).
In order to test any temperature cross-sensitivity, we conducted experiments at fixed RH while changing the temperature.Figure S14, Supporting Information demonstrates that when the temperature changes (in the range between 25 and 60 °C), the sensor still works well.Figure S15, Supporting Information indicates that the humidity sensor can detect moisture during breathing without being affected by motion, airflow, and temperature (see Video S1, Supporting Information).
We finally compared our results with those obtained by replacing the PS1-functionalized h-BN with GO. Figure S16, Supporting Information shows that for the change of Z imag , h-BN provides much higher sensitivity than GO and Z real does not show appreciable change.This comparison shows that PS1functionalized h-BN nanosheets do provide higher sensitivity and enhanced signal stability, as compared to GO.Furthermore, the device shows almost no hysteresis as compared to the one made by GO (Figure S17, Supporting Information).We also compared the surface area of h-BN and GO through Brunauer-Emmett-Teller surface area measurements.The surface area of h-BN and GO are 51 and 47 m 2 g −1 , respectively, indicating that the surface area cannot explain why the humidity sensing ability of h-BN is better than that of GO.
Our results show that the supramolecular functionalization of h-BN nanosheets with PS1 enables to achieve an enhancement of at least two orders of magnitude in the RH sensitivity, as compared to the best h-BN-based humidity sensor.This sensitivity is also much higher than that of sensors made with other 2D materials produced by LPE (e.g., BP, vanadium disulfide, molybdenum disulfide, molybdenum diselenide, titanium disulfide, Mxene), as well as other solution-processed materials, including those produced by Li intercalation, crosslinking (e.g., polymers), and Hummer's method (e.g., GO, rGO).Table S1, Supporting Information provides a comparison of all results and Figure 3c shows the key performance parameters of our device (i.e., sensitivity, defined as S = ΔZ/ΔRH, response and recovery time) as compared to the state-of-the-art, showing the excellent performance of the sensor.
Breath monitoring has attracted strong attention in the field of medical diagnosis and disease monitoring due to the presence of biomarkers associated with different diseases.Even the simple moisture sensing of exhaled breath can provide important physiological information on an individual, related to cardiac, neurological, and pulmonary conditions [54,55] as well as certain types of illness. [56,57]We integrated the humidity sensors on a face mask's external layer, close to the nose, to monitor a person's health status in real-time by recording breathing signals (Figure 4a).The h-BN-based humidity sensor shows a higher response than the one made by GO (Figure 4b).The wearable humidity sensor made by h-BN can record in real-time the respiration signals of different individuals and different daily activities.The real-time breathing response curves of the five volunteers show that each individual differs in the inhale-exhale cycle of the moisture and the duration of one breathing cycle (Figure S18, Supporting Information).Different human activities, including deep breathing, reading, watching a video, swallowing, and running also exhibit different breath curves (Figure S19, Supporting Information).The device can also be easily coupled to a wireless transmission module to make the wearable sensor more practical (Figures S20 and S21 and Video S2, Supporting Information).An app, named ST BLE Sensor, was used to receive and demonstrate the data in a graphical way for users on the mobile phone (Figure S20, Supporting Information).
Figure 4c,d demonstrates that the breathing sensor is an effective device to monitor several common symptoms of COVID-19 and flu such as coughing, fever, and runny and stuffy nose.During coughing, breath is short and discontinuous.When one gets a fever, the breath is slightly faster than that of the normal state.A runny nose induces high moisture concentration in the nose (high RH) and enhanced inhaling to avoid the mucus coming out of the nostrils, and thus the imaginary part of impedance hardly goes back to the baseline (low RH) for every breathing cycle.A stuffy nose leads to difficulty in breathing, thus the breath is slow and the imaginary part of the impedance is smaller than that of a normal state.The smaller "V" shape in the real part of the impedance monitoring curve also indicates that the maximum RH value is lower than that in the normal state.

Conclusion
This work shows that the non-covalent functionalization of h-BN nanosheets with pyrene derivatives allows to exploit this material as a sensing element in impedance-based, wireless, and wearable sensors for healthcare monitoring.The supramolecular functionalization with PS1 allows to prepare highly concentrated and stable dispersions, and to enhance the affinity of h-BN to water.The resistance shows more than six orders of magnitude change in the RH range from 5% to 100%, while strong changes in capacitance are recorded above 50% RH.This produces a characteristic signal shape in the real part of the impedance that allows fingerprinting of the breathing process of exhaling and inhaling, hence enabling us to use the device to record in real time the subtlest changes of respiratory signals in different daily activities as well as various symptoms of flu, without any direct contact.

Experimental Section
Preparation of the h-BN Dispersion: The dispersions were prepared via pyrene-assisted liquid-phase exfoliation by using h-BN powder (purchased from Sigma-Aldrich, >1 μm, 98%).This was added to de-ionized (DI) water at a concentration of 3 mg ml −1 , mixed with 1 mg ml −1 PS1 (purchased from Sigma-Aldrich, > 97%).The mixture was sonicated using a 600 W Hilsonic HS1900/HIlsonic FMG 600 bath sonicator at room temperature for 5 days, followed by the centrifugation at 3500 rpm (g factor = 903) for 20 min using a Sigma 1-14K refrigerated centrifuge to remove the unexfoliated bulk material.Afterward, the upper 2/3 supernatant was collected and further centrifuged at 13 800 rpm for 2 h to wash away the excess PS1.After washing, the precipitate was collected and diluted with water until reaching a concentration of 2 mg ml −1 .The concentration was determined by UV-vis spectroscopy using an absorption coefficient of 1000 L g −1 m −1 for h-BN measured at 550 nm. [48]GO water dispersion (oxygen content of 41-50 wt%) was purchased from Graphenea and diluted to ≈2 mg ml −1 .
Sensor Fabrication: 0.05 ml of the prepared h-BN aqueous solution was directly drop cast onto the electrodes, which were fabricated by sputtering nickel and gold (Ni: 3 μm thick; Au: 0.07 μm thick) onto polyimide (PI).The film thickness was ≈50 nm.The electrode array had a dimension of 10 × 10 mm, while the interdigitated electrodes were 100 μm in width, 0.75 mm in length, and a gap of 100 μm between electrodes.The number of electrodes was 10.
Sensor Characterization: The impendence measurements were carried out using a Zurich Instruments MFIA Impedance Analyser with a frequency range from 5 MHz to 0.5 Hz with two input interfaces and two output interfaces.The voltage was fixed at 1 V and the frequency was fixed at 10 KHz.The controlled humidity environments were achieved at a constant temperature (≈20 °C).The environment temperature and humidity were monitored by using a hygrometer (TESTO 608-H2).The normalized response and sensitivity, which were defined by Re = Z 0 /Z x and S = ΔZ/ΔRH, respectively, were used as figures of merit to evaluate the performance of the humidity sensor, where Z 0 and Z x are the impedance of the sensor at 0% RH and at x %RH, respectively, ΔZ is the change in impedance, and ΔRH is the RH change.

Figure 1 .
Figure 1.a) Schematic of pyrene-assisted liquid phase exfoliation of h-BN.b) Schematic of the h-BN-based humidity sensor fabrication and testing by impedance analyzer.

Figure 2 .
Figure 2. a) Change of the imaginary part (ΔZ imag ) and b) real part (ΔZ real ) of the impedance versus time obtained for the sensor made with h-BN nanosheets exfoliated without PS1 (black and green line) and h-BN exfoliated with PS1 (red and blue line).RH was switched between 20% and 60% for five times.c) Schematic of the sensing mechanism of the h-BN-based humidity sensor, showing hydrogen bonding interactions between the functionalized nanosheets and water molecules.d) Dependence of the resistance versus %RH measured at 20 °C.e) Dependence of the capacitance versus %RH.f) |Z| as a function of the %RH, calculated by R from (d) and C from (e) using Equation (4).

Figure 3 .
Figure 3. a) Simulated potential over the interdigitated sensor.b) Absolute values of real (|Z real |, blue line) and imaginary parts (|Z imag |, red line) of the impedance as a function of %RH, measured at 10 kHz.c) Comparison of the sensitivity (given by ΔR/R) versus response time versus recovery time for our sensors as compared to the state-of-the-art (TableS1, Supporting Information).d) Change of |Z| with time over 11 breathing cycles as compared with the relative theoretical impedance changes, calculated by using Equations (2) and (3).Changes in the e) real and f) imaginary parts of the impedance during deep breath monitoring by volunteer 2, showing the characteristic V shape signal associated with the changes in RH during the breathing cycle.

Figure 4 .
Figure 4. Wearable sensors for human breathing real-time monitoring.a) Photo of the wireless humidity sensor integrated onto a mask.b) Comparison of breathing sensors made by h-BN (red) and GO (green) used by volunteer 2. c) Changes in the: c) Imaginary and d) real parts of the impedance from volunteer 2 with common symptoms of flu, such as coughing, fever, and runny and stuffy nose.