Decoupling multiphysics interference for precise, real-time piezoresistive airflow sensing by integrated laser-induced graphene system Authors: Zihao Li, Chunki Yiu, Libei Huang, Weihua Guo, Le Cheng, Hao Hu, Jianju Su, Yun Song, Mingming He, Yinger Xin, Qiang Zhang, Ge Ye, Xinge Yu, Ruquan Ye Published: 2026 ABSTRACT Airflow sensing underpins applications in environmental monitoring, industrial safety, and wearable health care. Piezoresistive airflow sensors offer a compact and energy-efficient solution, yet their accuracy is severely compromised by complex multiphysical field perturbations. Here, we report an integrated graphene-based system fabricated by in situ laser printing that simultaneously measures airflow, temperature, and humidity. We reveal that a standalone piezoresistive airflow sensor can incur substantial errors exceeding 100% under fluctuating environmental conditions. Through targeted surface functionalization and microstructural engineering, we develop a humidity-insensitive temperature sensor, a high-fidelity humidity sensor, and a piezoresistive airflow sensor. Integrating these components into a multiphysics decoupling model enables analytical separation of intertwined signals, markedly reducing airflow measurement errors to below 5%. The optimized platform achieves high-precision, low power consumption, and strong integration airflow sensing with stable output under real-time fluctuations. These capabilities further allow dual-parameter airflow communication, while the scalable, mask-free laser-printing process provides a pathway toward portable, multifunctional sensing systems. INTRODUCTION Airflow sensing plays a crucial role across a wide range of scientific and engineering domains. In traditional sectors such as mining (1), marine exploration (2), and aerospace (3), accurate and continuous monitoring of wind direction and velocity supports environmental assessment and operational security. In emerging fields such as health care and wearable technology, airflow sensors are equally indispensable (4). Real-time tracking of breathing dynamics not only enables early diagnosis and disease management but also facilitates long-term health surveillance for vulnerable populations (5–7). Despite this broad importance, current airflow sensing systems often exhibit poor environmental resilience, limited sensitivity, large form factors, and inadequate integration with portable platforms, thereby constraining their deployment in dynamic or resource-limited settings. Now, airflow sensing technologies typically fall into three main working principles: mechanical sensing (8–10), heat exchange sensing (3, 11, 12), and piezoresistive sensing (Fig. 1A) (13–15). Their performance can be evaluated in terms of power efficiency, accuracy, and compactness. Traditional airflow sensors are built upon physical transition of mechanical parts, which are bulky and less suitable for integration. Heat exchange–based sensors have the potential to differentiate different gases based on thermal conductivity, but they generally demand higher power consumption and may have a slower response rate. In comparison, piezoresistive sensors have received great attention because they can be integrated into miniaturized devices and have low energy consumption. However, piezoresistivity is highly susceptible to multiphysical field interference, especially temperature and humidity variations (16–18). These factors interfere with resistance and can cause substantial inaccuracy in airflow measurement. For instance, in the range of 10 to 50% humidity and 25° to 45°C, environmental effects can lead to measurement errors exceeding 100 to 300% (Fig. 1B). These limitations highlight an urgent demand for robust, miniaturized, and high-precision airflow sensing technologies capable of maintaining stable performance under fluctuating environmental conditions. Accurate airflow sensing under complex environmental conditions remains a critical challenge. While decoupling strategies have been explored, most existing studies focus on only two physical variables under relatively stable conditions (table S1) (3, 5, 8, 13, 14, 19–24), where simple linear calibration or regression suffices. In contrast, real-world scenarios often involve simultaneously varying temperature, humidity, and airflow, resulting in strongly coupled and nonlinear sensor responses. Consequently, accurate signal separation under such multiparameter interference remains largely unexplored. In addition, many reported sensing systems rely on multistep fabrication processes, which limit the integration capability and scalability. To address these limitations, it is necessary to develop a sensing platform that not only decouples multiphysical interference at the system level but also enables simplified and integrated fabrication. Airflow sensing principles and integrated LIG-based platform for multiphysics decoupling. (A) Comparison of three major airflow sensing principles: mechanical, heat exchange, and piezoresistive. Their performance is evaluated in terms of accuracy, power efficiency, and compactness, with piezoresistive devices particularly vulnerable to temperature and humidity fluctuations. (B) Traditional piezoresistive sensing showing strong dependence on temperature and humidity with errors above 100%. (C) Illustration of the multiphysics decoupling framework developed in this study. Variables including sensing module resistivity, temperature, and humidity are incorporated into the analytical model, enabling separation of background effects from the airflow response and reducing prediction errors to below 5%. (D) Schematic of the integrated sensing platform fabricated by in situ laser induction of graphene patterns on FPCBs. Distinct microstructures were engineered to realize airflow, humidity, and humidity-resistant temperature sensing, and planar LIG traces provided stable electrical interconnections with minimal metallic wiring. Insets show scanning electron microscopy images of the airflow, humidity, and temperature modules with scale bars of 1 mm, 500 μm, and 10 μm. In this study, we present an integrated sensing system and a multiphysics model that aim to decouple the intertwined environmental effects for precise and rapid detection of airflow, temperature, and humidity. Carbon-based piezoresistive sensors such as carbon nanotube (15, 25, 26) and graphene (27–29) show advantages including light weight, flexibility, high sensitivity, and high stability (30–33). However, their fabrication typically involves multistep synthesis, micropatterning, and has difficulty in integration (34–36). Instead, we leverage the laser-induced graphene (LIG) technology (37–40) to directly print different sensing modules on flexible printed circuit boards (FPCBs). LIG provides a mask-free protocol for direct conversion of carbon-rich polymer substrates into various conductive graphene structures under ambient conditions (41–43), without requiring vacuum systems, extreme temperatures, or complex chemical processing (44–46). Thus, LIG allows simultaneous synthesis and patterning, facilitating rapid fabrication of flexible, miniaturized devices suitable for integrated electronics (47–49). Although LIG has been proposed to fabricate advanced airflow sensors (13, 50), the environmental interferences have been largely overlooked. Here, we further extend this capability to construct an integrated LIG-based sensing platform for airflow detection under coupled temperature and humidity conditions. To reveal the influence of multiphysical fields on airflow sensing, we prepare three LIG-based sensors through surface functionalization and structural engineering. These three sensors are optimized showing the characteristics of humidity insensitive, humidity sensitive, and piezoresistivity sensitive. These optimized structures are then used to construct an integrated sensing platform that incorporates a LIG airflow sensor together with LIG temperature and humidity sensors. During operation, the LIG humidity sensor and the humidity-resistant LIG temperature sensor measure the environmental background parameters, which are then processed through the established multiphysics field decoupling model to analytically separate the contributions of temperature, humidity, and airflow velocity to sensor resistivity (ΔR/R0). By compensating for environmental influences in real time, the system can accurately determine airflow velocity under temperature and humidity interference, enabling precise airflow measurements with error rates below 5% (Fig. 1C). This strategy converts an environmentally coupled piezoresistive response into a quantitatively decoupled airflow output, thereby improving sensing accuracy under dynamic backgrounds. The system overcomes the poor environmental resilience of traditional airflow sensors, delivering a high-precision, low-power, and fully integrated sensing solution suitable for portable, wearable, and embedded applications. Accurate airflow sensing under such conditions also requires real-time signal processing to suppress noise and stabilize the output. Therefore, beyond sensor design and physical decoupling, the present work integrates real-time data processing to enable continuous and reliable airflow signal extraction under dynamic environments. Beyond accurate airflow monitoring, the decoupling model also enables information encryption and communication through controlled airflow modulation. By encoding information simultaneously in airflow velocity and pulse duration, the system surpasses conventional on-off Morse-type transmission and achieves one-character-per-two-pulse encoding. This dual-parameter communication framework demonstrates the feasibility of interference-free airflow-based data transfer, particularly for scenarios where conventional communication is restricted or unreliable, such as flammable environments or breath-based human-machine interfaces. RESULTS Integrated platform enabled by in situ LIG The integrated sensing platform was fabricated by in situ laser printing of graphene patterns on FPCB substrates (movie S1), where a polyimide (PI) precursor was locally converted into graphene through computer-controlled design (Fig. 1D). By tuning laser parameters such as pulse duration, power density, and scanning trajectory, distinct microstructures were rationally engineered according to the dominant sensing mechanisms of airflow, humidity, and temperature detection within the same substrate. A concentric loop scanning strategy generated a hierarchical porous network that enhanced convective interactions and enabled sensitive airflow detection, while cyclic raster scanning produced vertically aligned porous frameworks with enlarged surface area, allowing rapid humidity sensing. For temperature sensing, postfluorination of LIG endows surface superhydrophobicity, suppressing water adsorption and ensuring accurate thermal readout independent of humidity. Conductive interconnects were inscribed through fixed-pulse scanning, forming planar LIG traces that minimized metallic wiring to essential nodes. This configuration ensured a more stable signal transfer between the sensing modules and the circuit board than direct metal contact while also reducing reliance on extensive wiring and eliminating adhesive bonding. Raman spectra of the materials used for these modules both exhibit characteristic D, G, and 2D bands, confirming the formation of LIG (fig. S1) (51, 52). Comparison of the intensity ratios reveals distinct structural features associated with sensing performance. The humidity module shows the highest ID/IG ratio, indicating abundant defect sites that are beneficial to water adsorption and humidity sensing (53). The airflow module exhibits an intermediate ID/IG together with moderate I2D/IG, reflecting balanced defect density and graphitic order favorable for stable and rapid airflow sensing (54, 55). The interconnect shows the lowest ID/IG and the highest I2D/IG ratio, demonstrating the best graphene quality with minimal defects, ensuring efficient and reliable electrical conduction across the sensing platform. Detailed optimization strategies and characterization analyses for each module are provided in Supplementary Texts S1 to S4, figs. S2 to 26, and tables S2 to S4. This mask-free process integrates airflow, humidity, and temperature sensing into a single flexible platform, combining compactness with robustness, scalability, and compatibility with roll-to-roll processing for large-area manufacturing of multifunctional environmental sensors. Airflow sensing under varying temperatures To confirm the thermal and humidity interferences, the airflow sensor was first evaluated at varying temperatures at a constant 50% relative humidity (RH). Figure 2A shows the temporal response of the sensor resistance, represented by the orange curve, when airflow pulses were applied. The blue curve displays the simultaneously recorded ambient temperature profile, which served as the reference for calibration. As the surrounding temperature increases, the baseline resistance of the sensor gradually shifts, revealing a clear negative temperature coefficient. Specifically, as temperature rises from 25° to 45°C, the resistance decreases in an approximately linear fashion by about 3.6% (Fig. 2B). This linear decrease is consistent with the intrinsic semimetallic nature of LIG (56–58). This temperature-dependent resistance variation is commonly observed in piezoresistive materials, where the sign and magnitude of the temperature coefficient depend on the specific material system and device design. The results confirm that temperature acts as a major external factor influencing the electrical baseline of airflow sensing. Beyond shifting the baseline, temperature also affects airflow sensing performance. As illustrated in Fig. 2C, the relative resistance change (ΔR/R0) of the airflow sensor exhibits a linear dependence on airflow velocity across all investigated temperature conditions. The linearity of the ΔR/R0-V relationship demonstrates that the LIG airflow sensor can provide predictable responses to velocity variations even in the presence of thermal disturbances. However, the slope of the ΔR/R0-V relation, which reflects sensitivity, is not constant at different temperatures. As temperature increases, the slope becomes steeper, resulting in higher ΔR/R0 values under identical flow rates. For example, at 45°C, the maximum resistance variation reaches nearly 1.8%, which is higher than the values obtained at 25°C under the same airflow amplitude. This temperature-enhanced sensitivity is likely related to more efficient carrier transport in the graphene network at elevated temperatures. Although such thermal amplification improves the apparent sensitivity, it introduces systematic errors into practical measurements. Under certain conditions, the resistance change produced by temperature drift exceeds that produced by airflow itself. This overlap complicates the direct interpretation of ΔR/R0 as a measure of flow velocity and highlights the necessity of compensating for environmental effects. Detailed analysis reveals that both the slope (sensitivity) and the intercept y parameters of the linear ΔR/R0-V relation vary monotonically with temperature (Fig. 2D). Their variation follows a linear trend with respect to the thermal input. This finding suggests that the impact of temperature can be mathematically parameterized and separated from the true airflow response. The linear temperature dependence therefore forms the basis of the decoupling model developed in this study. Temperature dependence of porous LIG-based airflow sensing. (A) Temporal evolution of airflow sensor resistance (orange) under airflow pulses with simultaneously recorded ambient temperature (blue) serving as a reference. Increasing temperature gradually reduced the baseline resistance, indicating a negative temperature coefficient. (B) Temperature-dependent baseline resistance measured at V = 0 m/s and 50% RH. (C) Linear dependence of relative resistance change on airflow velocity at different temperatures. Higher temperatures produced steeper slopes, reflecting increased sensitivity under identical airflow conditions. (D) Systematic variation of slope and intercept values of the ΔR/R0-velocity relation with temperature. Both parameters follow linear trends, providing a quantitative basis for mathematical decoupling of thermal effects from airflow response. All error bars = SD (n = 5). Airflow sensing under varying humidities To assess humidity effects, airflow sensing was examined at 25°C with varying RHs. When the sensor was exposed to a sequence of airflow pulses, its resistance varied in synchrony with the applied flow, and the underlying baseline gradually decreased as humidity decreased, as shown in Fig. 3A. This behavior is in clear contrast to the temperature results, and it highlights the distinct role of water vapor in modulating sensor behavior. Quantitative analysis reveals that as RH increases from 10 to 50%, the baseline resistance rises by about 0.6% (Fig. 3B). Although this variation is smaller than the thermally induced change, the effect is systematic and cannot be neglected. The increase can be explained by the adsorption of polar water molecules onto the porous graphene framework (59). The attached water molecules disturb conductive pathways and increase carrier scattering, thereby raising the network resistance. The approximately linear relationship observed in Fig. 3B suggests that humidity effects can be parameterized in a straightforward form, which is advantageous for later compensation modeling. Despite this baseline drift, the airflow response preserves a clear linear dependence on flow velocity under all tested humidity levels (Fig. 3C). The slope of the ΔR/R0-velocity relation becomes steeper as humidity increases, indicating that sensitivity is enhanced in more humid environments. At 50% RH, the maximum ΔR/R0 variation reaches ∼1.1%, which is much higher than at 10% RH under the same flow rates. This amplification can be attributed to the denser population of adsorbed water molecules at higher humidity. While static adsorption increases resistance, the application of airflow perturbs these molecules, inducing redistribution or partial desorption. The resulting fluctuations in local conductivity amplify the resistance variation, making the surface more responsive to flow perturbations. Humidity dependence of porous LIG-based airflow sensing. (A) Temporal response of the airflow sensor under sequential flow pulses at different RH levels. The baseline resistance increased systematically with rising humidity, reflecting adsorption of water molecules on the porous graphene surface. (B) Linear increase of baseline resistance with RH, confirming that moisture adsorption acts as a predictable factor influencing the electrical baseline. (C) Linear dependence of relative resistance change on airflow velocity under different humidity levels. Higher humidity produced steeper slopes and greater resistance variations, indicating enhanced sensitivity. (D) Systematic variation of slope (sensitivity) and intercept parameters of the ΔR/R0-velocity relation with RH, providing a quantitative basis for incorporating humidity effects into the analytical decoupling model. All error bars = SD (n = 5). From a sensing perspective, the dual role of humidity, which raises baseline resistance while simultaneously enhancing airflow sensitivity, captures the complex balance between damping and responsiveness. On one hand, higher humidity increases the susceptibility of the sensor to low flow rates, which can improve detection limits. On the other hand, the accompanying drift complicates direct interpretation, since humidity-induced resistance changes may obscure the true airflow contribution, particularly when the velocity signal is weak. Careful modeling is therefore required to disentangle the overlapping effects. Consistent with the temperature results, both the slope and intercept y of the linear ΔR/R0-V relation in Fig. 3D vary monotonically with humidity, and their dependence is well approximated by linear functions. This finding is important because it demonstrates a parallel between temperature and humidity responses, which highlights the necessity of a unified analytical approach that treats temperature and humidity together rather than separately. Both environmental variables alter baseline and sensitivity in a predictable fashion that can be described by linear relations. This linearity is critical as it transforms environmental interference from a source of unpredictable error into a set of deterministic parameters that can be embedded in analytical models, thereby ensuring both accuracy and scalability in practical sensing. Humidity sensing and temperature compensation To provide accurate RH input for airflow correction, a dedicated humidity sensor was fabricated by cyclic raster scanning, producing vertically aligned porous graphene structures for RH detection. The structure exhibits superhydrophilicity, as evidenced by the spreading of a water droplet across the surface with a contact angle approaching 0° (Fig. 4B and movie S2). This superhydrophilic morphology promotes efficient adsorption of water molecules, enabling rapid resistance responses under dynamic humidity variations. The resistance increases linearly with increasing RH, confirming the suitability of the porous LIG structure for humidity detection (Fig. 4D). At 25°C, the relative resistance change is ∼13% when RH increases from 22 to 40%, whereas at 50°C, the variation is closer to 8%. In addition, the slope and intercept of the ΔR/R0-RH relation are temperature dependent (fig. S27). As temperature increases, the baseline resistance shifts downward, and the slope becomes slightly steeper. The enhanced sensitivity at elevated temperature can be attributed to the combined effects of thermally activated carriers and reduced adsorption damping. Moreover, a clear correlation between surface wettability and humidity sensitivity was observed, indicating that more hydrophilic surfaces promote stronger interactions with water molecules and higher sensing responses (Fig. 4E and fig. S28). The dependence of sensor resistivity Rhum response on both humidity and temperature can be expressed by a bilinear model (Eq. 1) Rhum(T,RH)=a1(T)×RH+a2(T)(1) Humidity and temperature sensing modules based on tailored LIG surface. Contact angle measurements (conducted at 25°C and ∼50% RH) illustrating the transition from (A) pristine PI (76°) to (B) superhydrophilic porous LIG after cyclic raster scanning (≈0°) and to (C) superhydrophobic Fluoro-LIG after fluorination modification. The extreme wettability contrast validates selective surface functionalization. (D) Humidity sensing curves of porous LIG under different temperatures, showing higher RH leading to larger resistance changes. (E) Correlation between surface wettability and humidity sensing (50°C), indicating that more hydrophilic surfaces promote stronger interactions with water molecules and higher sensing responses. (F) Fluoro-LIG–temperature sensing curves at different RH levels, showing nearly overlapping linear relations and indicating negligible RH influence on thermal response. (G) Nonfluorinated porous LIG temperature sensing, where temperature responses are markedly perturbed by humidity, underscoring the necessity of hydrophobic modification for stable thermal sensing. All error bars = SD (n = 5). where both a1 and a2 vary linearly with T, as supported by the regression analysis shown in fig. S27 (R2 > 0.98 for both parameters). This leads to the generalized expression (Eq. 2)Rhum=C0+CRH×RH+CT×T+CT·RH (T×RH)(2) The excellent linearity and high fitting coefficients indicate that the bilinear model reliably captures the coupled effects of temperature and humidity. By incorporating accurate temperature readings, the model enables correction of thermal influences and ensures that the humidity sensor provides precise input values for subsequent airflow decoupling and compensation. Humidity-resistant temperature sensing A fluorinated LIG sensor was fabricated to obtain reliable ambient temperature readings independent of humidity. Fluorination modifies the porous graphene surface into a superhydrophobic state (Fluoro-LIG), producing a contact angle of ∼150.9° (Fig. 4C). When subjected to slight lateral airflow perturbations, droplets readily detach from the sensing module, and the entire process of contact angle measurement and droplet removal induces almost no change to the Fluoro-LIG morphology (movie S3). Water droplets retain spherical shapes and roll off the surface with minimal adhesion, confirming strong hydrophobicity. The reduction in surface affinity effectively suppresses moisture adsorption, which otherwise perturbs resistance readings. The Fluoro-LIG sensor exhibits a consistent linear ΔR/R0-temperature relationship across a wide range of RH values (Fig. 4F). The slope and intercept of the fitted line remain stable, with relative SDs below 3.5% as shown in fig. S29. This low variability demonstrates that humidity exerts a negligible influence on the temperature response. The ΔR/R0-T curves are nearly superimposed, yielding comparable sensitivities of about −1% °C−1. In contrast, when the same porous LIG structure is used without fluorination, the temperature response (such as resistance variations and sensitivity) is strongly perturbed by ambient humidity, as shown in Fig. 4G and fig. S30. This comparison highlights the necessity of hydrophobic modification for stable thermal sensing. The reliable linearity and humidity tolerance of the Fluoro-LIG sensor establish it as an accurate temperature probe for the decoupling framework. By supplying stable temperature input unaffected by ambient RH, the sensor ensures that humidity corrections for the airflow sensor can be performed with minimal uncertainty. Together with the porous LIG humidity sensor, this component provides the necessary environmental references for multifield interference compensation. Multiphysics decoupling of airflow signal With accurate temperature and humidity readings obtained from the dedicated LIG-based environmental sensors, a multivariable model was constructed to decouple the airflow contribution from environmental interference. Airflow sensing differs fundamentally from humidity or temperature sensing because the resistance variation arises not only from flow velocity but also from concurrent environmental factors. For humidity sensing, the resistance change can typically be described by a direct linear relationship between resistance and RH at a fixed temperature. In contrast, the airflow sensor exhibits responses that are simultaneously influenced by velocity, temperature, and humidity. This coupling precludes the construction of a simple one-dimensional calibration curve, as any variation in ambient temperature or humidity alters the apparent airflow contribution. The above systematic experiments reveal that the resistance variation depends jointly on these three variables and that their effects can be captured by first-order interactions. The governing relation can therefore be expressed as Eq. 3Rairflow=a1T×V+a2RH×V+a3T×RH×V+a(T,RH)(3) where the first three terms represent the multifield coupling between airflow velocity and environmental conditions, while a(T, RH) denotes the baseline offset arising from temperature and humidity. This formulation highlights that airflow-induced resistance changes cannot be isolated without considering environmental cross-terms. To capture this behavior in a tractable form, the airflow sensor response is reformulated as Eq. 4, with accurate temperature and humidity readings obtained from the dedicated LIG-based environmental sensorsR=F1(T,RH)×V+F2(T,RH)(4) where F1(T, RH) accounts for the effective coupling between velocity and environmental fields, and F2(T, RH) captures the residual offset. Both terms increase monotonically with temperature and are well approximated by a bilinear function (Eq. 5)F(T,RH)=C0+CT×T+CRH×RH+Cm×T×RH(5) where C0 corresponds to the baseline resistance offset in the absence of environmental fluctuations, CT represents the linear temperature sensitivity, and CRH quantifies the humidity dependence. The cross-term coefficient Cm captures the synergistic interaction between heat and moisture, which modifies both the effective slope and intercept of the airflow response. By feeding the experimental database (R, T, RH, and V) into the model, the values of these coefficients are quantitatively extracted through regression fitting, as summarized in table S5. These constants provide a compact representation of how environmental variables modulate sensor output, enabling reliable decoupling of airflow contributions from temperature and humidity interferences. The model reproduces the experimental dataset with an error below 5% across the full range of studied conditions (Fig. 5A and fig. S31). This high accuracy confirms that the model is sufficient to capture the essential physics of the system without resorting to higher-order nonlinearities. Airflow velocity can therefore be obtained from the corrected expression (Eq. 6)V=[R−F2(T,RH)]/F1(T,RH)(6) Analytical modeling and real-time performance of robust LIG-based airflow sensing system. (A) Comparison of model-calculated and reference airflow velocities, showing relative errors below 5% across the experimental dataset (at 45°C and 50% RH). (B) Three-dimensional simulation database of airflow velocity constructed from temperature, humidity, and resistance variation. (C) Real-time airflow measurements without temperature-humidity compensation, showing large deviations with errors exceeding 100 to 160%. (D) Real-time airflow measurements using the decoupling framework, where the computed velocity closely follows the reference with deviations consistently below 5%. This formulation allows direct computation of velocity using the measured resistance change rate together with the independently determined temperature and humidity values. By inputting prescribed values of temperature, RH, and resistance variation into the analytical model, the airflow velocity under multiphysical conditions can be quantitatively determined, enabling construction of a three-dimensional simulation database (Fig. 5B), where the color encodes the velocity. With increasing temperature and humidity, part of the resistance change is consumed by environmental effects, leading to substantial deviations between the corrected velocity and that obtained without compensation. The predictive capacity extended across the entire parameter space tested, confirming that the same analytical model is robust to diverse environmental backgrounds without requiring separate calibrations. Real-time airflow measurement The integrated sensing system was further evaluated in a dynamic environment where airflow varied under dynamic environmental conditions. As shown in Fig. 5D and movie S4, the resistance variations (fig. S32) produced by the airflow sensor are continuously processed through the decoupling model and converted into velocity signals in real time. The computed velocity closely traces the reference measurement obtained from a calibrated flowmeter, with deviations consistently below 5%. This agreement confirms that the decoupling framework effectively removes the confounding effects of temperature and humidity, leaving a signal that directly reflects airflow velocity. In contrast, when the same piezoresistive LIG airflow sensor is evaluated without temperature-humidity compensation, using only the fixed linear ΔR/R0-V calibration (25° and 30% RH), substantial drifts occur with error levels that can exceed 160% (Fig. 5C). The comparison underscores the necessity of environmental compensation and highlights the robustness of the present approach. To further assess the reproducibility of the proposed system, both sample-to-sample and batch-to-batch variations were examined. For the sample-to-sample comparison, multiple sensing modules with identical design were fabricated and tested on the same FPCB platform. For the batch-to-batch comparison, fully assembled devices with the same design were fabricated on separately prepared FPCB substrates and evaluated under identical airflow sensing conditions. As shown in fig. S33, the corrected airflow signals obtained from different samples and fabrication batches remain consistent with the applied airflow inputs, with measurement errors generally around 4%, demonstrating good system reproducibility. To evaluate the long-term stability of the sensing system, airflow sensing measurements were repeated after 6 months under the same experimental conditions. As shown in fig. S34, the airflow velocity obtained through the multiphysics decoupling model remains in good agreement with the actual airflow input. The corresponding measurement error remains below 5%, comparable to that of the original measurements (Fig. 5D). These results indicate that the integrated sensing platform maintains stable airflow sensing performance over extended periods. To benchmark the proposed system, key performance metrics are compared with representative airflow sensing technologies (14, 15, 25, 60–67) in table S6. Our systems show high linearity, low detectable velocity, and stable output. Notably, the effective airflow sensitivity is adaptively modulated with temperature and humidity, enabling accurate sensing under varying environmental conditions. In addition, the mask-free laser-induced fabrication allows direct and integrated device construction, further simplifying the fabrication process. To further evaluate the independence and cross-sensitivity of the integrated sensing system, the simultaneously recorded resistance outputs of the three sensing channels were analyzed under independently varied airflow, temperature, and humidity conditions. Under stepwise airflow variation at constant temperature and humidity, the airflow sensor showed distinct step-like resistance changes, whereas the temperature and humidity sensors remained nearly stable, indicating that airflow excitation mainly affects the airflow sensing channel (fig. S35). Under temperature variation at constant humidity and without airflow excitation, all three sensing channels showed resistance variations, confirming that temperature can shift the electrical baselines of both the humidity and airflow sensors and should be included in the correction model (fig. S36). Under humidity variation at constant temperature and without airflow excitation, the temperature sensor remained nearly unchanged, while the humidity and airflow sensors exhibited humidity-dependent resistance variations, indicating that humidity directly affects the humidity sensing channel and also introduces baseline interference into the airflow sensing channel (fig. S37). The cross-sensitivity characteristics and correction strategies are summarized in table S7. These results further support the physical basis of the multiphysics correction model, in which independently measured temperature and humidity are used to compensate environmental interference and accurately extract airflow velocity from the coupled resistance response. Building on the accurate airflow velocity sensing achieved through the decoupling multiphysics interference framework, the integrated airflow system demonstrates the potential for information encryption and transmission. Unlike conventional Morse code–based methods that rely solely on the duration of on-off pulse (typically requiring three to four pulses per letter and limited to alphabetic transmission), the proposed decoding scheme simultaneously uses airflow pulse duration and airflow velocity as dual parameters. This enables the robust delivery of one alphanumeric character per two airflow pulses, supporting all 26 letters and digits from 0 to 9 (the detailed encoding strategy and operational definitions are provided in tables S8 and S9). As illustrated in Fig. 6 (A to C and F to H) (35°C and 50% RH) and fig. S38 (A to C and F to H) (25°C and 30% RH), the proposed decoupling model effectively eliminates environmental interference during communication. Under compensated conditions, the decoded waveforms precisely match the theoretical information-encoded waveforms, achieving fully accurate message reconstruction for both “CITYU” and “CALL911.” In contrast, when decoding is performed using only the fixed linear ΔR/R0-V calibration (45°C and 50% RH) without temperature-humidity compensation (Fig. 6, D, E, I, and J, and fig. S38, D, E, I, and J), the inaccurate velocity estimation leads to distorted or lost information. These results confirm that the decoupling model not only ensures high-fidelity airflow measurement but also guarantees reliable information transmission even under varying environmental conditions. To further evaluate this capability under more complex airflow inputs, additional array-based experiments were conducted using both single-direction and simultaneous airflow excitations. In all cases, the airflow was introduced at normal incidence, i.e., perpendicular to the sensor surface (θ = 90° relative to the device plane). All measurements were conducted at 36.5°C and 95% RH, simulating typical human exhalation conditions. In a local sensing region, turbulent or complex airflow can be conceptually decomposed into multiple directional airflow components with different intensities and temporal variations. As shown in fig. S39, different excitation configurations produce distinct spatial resistance patterns, with dominant responses consistently localized at the airflow-exposed sensors. In the single-direction excitation case (fig. S39B), the sensor directly exposed to airflow exhibits a pronounced response, while adjacent sensors show only minor variations, indicating limited lateral coupling. In contrast, under simultaneous excitation (fig. S39, C and D), dominant responses emerge at the corresponding sensor positions, with clear spatial differentiation preserved despite overlapping airflow inputs. These spatially resolved response features enable direct identification of airflow direction and distribution from the array outputs. Despite overlapping inputs and minor temporal fluctuations, the relative amplitude hierarchy among sensors remains stable and reproducible across repeated cycles, and the dominant response regions remain well defined. This indicates that the sensing mechanism is governed primarily by localized airflow-sensor interactions, rather than relying on global flow uniformity. These simultaneous excitation experiments provide a simplified and controlled demonstration of how multiple directional airflow components influence the sensor array, revealing asymmetric and spatially localized airflow-sensor responses. The corresponding decoupling and waveform reconstruction results are shown in figs. S40 and S41. The airflow velocities extracted from the dominant-response sensors follow the applied airflow profiles in both amplitude and temporal evolution, while the reconstructed waveforms remain within the predefined ranges (table S8), which preserves correct decoding despite complex excitation. This agreement is consistently maintained across various excitation scenarios, demonstrating that the decoupling process effectively isolates the target airflow contribution from spatially overlapping and dynamically perturbed inputs. From an array sensing perspective, when multiple airflow components are present, the primary airflow direction can be determined by identifying the sensors with the largest response amplitudes, corresponding to regions of strongest local airflow interaction. The signals from these dominant sensors are then directly used as inputs for the multiphysics decoupling model. On the basis of these results, the approach is expected to enable reliable airflow detection and information decoding under more complex, spatially distributed flow conditions. It should be noted that turbulent airflow is chaotic and location dependent, and each airflow sensor captures the local airflow state at its installation position. Therefore, to obtain the global direction or spatial distribution of a turbulent airflow field, multiple airflow sensors or sensor arrays should be deployed at different locations, so that locally measured airflow information can be spatially combined to infer the broader airflow distribution. While conventional communication techniques, such as radio frequency or optical communication, are facile for standard applications, the proposed airflow-based information decoding system targets critical niche environments where traditional methods fall short. For instance, in highly flammable or explosive settings (e.g., chemical plants and underground mines), airflow communication eliminates the spark risks associated with electronic devices, offering an intrinsically safe alternative. Furthermore, this technology holds great promise for health care, greenhouse, building, and industrial manufacturing, where precise airflow detection and control are important. For example, in health care, our technology can enable reliable, breath-controlled silent communication for patients with severe motor impairments. In these scenarios, our decoupling model effectively eliminates the severe interference caused by the high humidity and temperature of human breath, ensuring accurate information decoding. Information encryption and decoding through airflow communication. Information transmission and decoding process under 35°C and 50% RH for the messages (A to E) CITYU and (F to J) CALL911. (A and F) Sensor resistance variation (ΔR/R0). [(B) and (G)] Airflow decoding with the proposed decoupling model. [(D) and (I)] Airflow decoding using fixed linear ΔR/R0-V calibration (45°C and 50% RH). Comparison between measured and information-encoded waveforms, demonstrating correct [(C) and (H)] or failed [(E) and (J)] decoding results under compensated and uncompensated conditions. The detailed encoding strategy and operational definitions are provided in tables S8 and S9. DISCUSSION In this work, we developed a fully integrated airflow sensing system based on LIG patterned with functional microstructures. The platform was fabricated directly on FPCBs using in situ laser induction, enabling the selective formation of porous, superhydrophilic, and fluorinated hydrophobic microstructures optimized for airflow, humidity, and temperature sensing, respectively. The inclusion of dedicated humidity and temperature modules provided accurate environmental inputs, which were incorporated into a multiphysics decoupling model. This model analytically separated the contributions of temperature, humidity, and airflow velocity to the resistance variation, reducing measurement errors to below 5% even under real-time dynamic conditions. This performance demonstrates the capability of the proposed system to handle strongly coupled and dynamically varying multiphysical conditions beyond conventional calibration scenarios. The system exhibited rapid response and recovery, reaching subsecond levels under optimal conditions and remaining within a few seconds across the full velocity range. Stable outputs and clear advantages over uncompensated LIG sensors, which typically suffer from errors exceeding 160%, were also demonstrated. These findings validate environmental decoupling as an effective strategy for achieving high-precision airflow measurements. Beyond demonstrating a robust sensing mechanism, the proposed approach highlights the scalability and versatility of LIG for multifunctional sensing. The fabrication process is rapid, mask free, and compatible with roll-to-roll production, making it suitable for large-area manufacturing. The integration of accuracy, compactness, and environmental resilience establishes a pathway toward next-generation portable sensing systems. Sensor design, multiphysics decoupling, and real-time signal processing together form a robust system-level framework capable of reliable airflow sensing under complex and dynamically varying environmental conditions. Furthermore, by capitalizing on its high-resolution and interference-free airflow detection capability, the system was further adapted as a dual-parameter communication platform, capable of encoding and decoding alphanumeric information through controlled airflow pulses. The integration of multifunctional sensing and robust communication underscores the broader applicability of the proposed LIG system, not only as a precision environmental sensor but also as a foundation for airflow-mediated information transmission and encryption. With targeted optimizations, the platform’s capabilities could be extended from respiratory monitoring and wearable health care to industrial process control and environmental surveillance, where stable airflow sensing is essential. MATERIALS AND METHODS Fabrication of LIG-based sensing system PI films provided by Zeman Tape Material Technology (thickness, ∼120 μm) were used as precursors for laser induction. Graphene patterns were directly written using a CO2 laser processing platform (Minsheng Laser, #MSDB-FM60 CO2 Laser Marker, 60 W, 10.6-μm wavelength) in ambient conditions. By varying laser scanning trajectories and parameters, different microstructures were created in situ on FPCB (fig. S42). For airflow sensing, concentric loop scanning was used to produce a porous three-dimensional LIG network. The laser processing was in vector mode, and the power, frequency, scanning speed, pulses/dot, and line spacing were set as 12 W, 3000 Hz, 500 mm/s, 2, and 200 μm, respectively. For humidity sensing, cyclic raster scanning was applied to generate vertically aligned porous graphene filaments. The laser processing was in raster mode, and the laser power, frequency, scanning speed, and line spacing were adjusted to 15 W, 3000 Hz, 500 mm/s, and 400 μm, respectively, to optimize the microstructure for moisture adsorption. For temperature sensing, the porous LIG was further fluorinated by drop-casting a 1 wt % ethanol solution of trichloro(1H,1H,2H,2H-perfluorooctyl)silane (purchased from Aladdin, China) until the sensor surface was fully covered, followed by drying treatment. In addition to the sensing units, fixed-pulse scanning was used to fabricate conductive interconnects between the sensor modules and the readout circuit. The power, frequency, scanning speed, pulses/dot, and line spacing were set as 4.5 W, 3000 Hz, 500 mm/s, 10, and 0.05 μm, respectively, ensuring reliable electrical conductivity for signal transmission. The electronic design of the FPCB, including the microcontroller unit (MCU), ADCs, and wireless module, is shown in fig. S43. The sensors were connected in series with fixed resistors to convert impedance variations into potential signals. These signals were acquired by three 24-bit analog-to-digital converters (ADS1220), providing high-resolution sampling. The digitized data were collected by a MCU (STM32F410) and transmitted wirelessly to a computer-based station via a Bluetooth module (WH-BLE106). Characterization of the integrated sensing system Morphologies of the porous LIG and fluorinated structures were examined by scanning electron microscopy (QUATTRO S Scanning Electron Microscope from Thermo Fisher Scientific). Raman spectra were collected with a 532-nm excitation laser to confirm the graphene structure (WITec RAMAN alpha 300R). Surface wettability was assessed by a contact angle tester. Untreated PI exhibited a hydrophilic angle of ∼76°, porous LIG showed superhydrophilicity with complete spreading (∼0°), and fluorinated LIG displayed superhydrophobicity. Airflow sensing tests were conducted in an acrylic glove box with controllable temperature and humidity (Sanjiang Acrylic Experimental Instruments Co. Ltd., China). Argon was delivered through a polytetrafluoroethylene tube with an inner diameter of 3 mm. The flow rate was regulated by mass flow controllers [Alicat Scientific Inc., 0 to 50 sccm; Asert Instruments (Beijing) Co., Ltd., 10 to 200 sccm], and the gas velocity was calculated as v = Q/A, where Q is the volumetric flow rate and A = π × r2 is the cross-sectional area of the tube. A custom clamping fixture was used to fix the relative position and orientation between the airflow outlet and the sensing device, enabling pulsed loading of the gas stream to the sensing area at a controlled incident angle. During the standard airflow sensing measurements, the holder was used to maintain the airflow direction perpendicular to the sensing surface unless otherwise specified. Dynamic and real-time testing For dynamic evaluation, airflow pulses with varying amplitudes were applied while maintaining relatively stable temperature and humidity conditions. The resistance signals from the airflow sensor were continuously processed through the decoupling model and converted in real time into velocity, temperature, and humidity outputs. To suppress local signal fluctuations, an adaptive noise calibration was performed before each measurement cycle. A short preactivation stage was used to statistically characterize the instantaneous noise amplitude under current environmental conditions. During real-time processing, signal variations within this dynamically determined noise band were suppressed, while effective responses exceeding this threshold were preserved, ensuring noise reduction without distorting transient dynamics. Response and recovery times were extracted from transient resistance traces under stepwise airflow. In this study, these times denote the complete durations required for the resistance to reach its peak or return to baseline after the onset or cessation of airflow. Movie S4 shows the real-time acquisition of airflow velocity, temperature, and humidity, obtained through the integrated sensing system with multiphysical field decoupling.