Sweat monitoring beneath garments using passive, wireless resonant sensors interfaced with laser-ablated microfluidics

The study addresses the need for noninvasive, on-body monitoring of sweat loss and composition to inform hydration status, particularly for individuals wearing thick PPE (e.g., firefighters) who are at high risk of dehydration. Existing methods (e.g., body mass change) are confounded by absorbed water in clothing and sweat pooling on skin, and sweat rates vary with fitness, body type, acclimation, and anatomical region. Wearable sweat sensors face challenges for under-clothing use due to line-of-sight requirements (optical) or bulky electronics. The research question is whether a passive, wireless resonant sensor integrated with microfluidics can simultaneously and robustly measure sweat volume (rate) and conductivity through clothing/PPE, enabling low-cost, multi-region monitoring under real-world conditions.

Wearable sweat systems can be categorized by sampling (natural ventilation, wicking, microfluidics) and interrogation (potentiometry, capacitance, impedance). Natural ventilation best captures sensible and insensible sweat but is bulky, especially under PPE. Wicking and microfluidics are more wearable-friendly; modern systems leverage miniaturized electronics and wireless links (Wi‑Fi, Bluetooth, NFC). Under-clothing perspiration monitoring remains difficult due to optical line-of-sight constraints and bulk/instrumentation complexity. Resonant (LC) sensors, whose resonant frequency depends on the surrounding permittivity, have been applied in biomedical contexts (implants, vital signs, tissue dielectric characterization, biofluids, enzyme activity) and offer a simple, passive platform potentially suitable for sweat analysis beneath garments.

Device design and fabrication: The sweat sensor is a conformable sticker combining a laser-ablated PDMS microfluidic channel with a planar Archimedean spiral resonator etched from copper-coated polyimide (DuPont Pyralux). Layer stack (bottom to top): PDMS (1 mm), laser-ablated channel (~0.4 mm width; depth set by laser parameters), plasma-bonded PDMS cap (1 mm), resonator attached with 100 μm acrylic adhesive. Total thickness ~2.16 mm. The resonator outer diameter is ~40 mm, designed for bare f0 ~85 MHz, below the 250 MHz limit of the portable VNA (MetroVNA). The spiral geometry was chosen for inductive coupling efficiency and fabrication simplicity.

Laser ablation characterization: PDMS channels were fabricated using a CO2 craft laser cutter (GlowForge Plus). Eighty-eight straight channels were ablated over a range of laser powers (8–24 W) and speeds (0.008–0.04 m/s). Dimensions were measured with a 3D digital microscope. A depth model based on mass/heat flux balance was fit to the data, yielding c1 = 0.357 µm/(J/m^2) (±0.063, 95% CI) and c2 = 18.5 nm (±4.51), R^2 = 0.943, RMSE = 45.10 µm. Width correlated linearly with depth: W = 0.073 D + 11.0 (RMSE = 17 µm, r = 0.635). The method supports aspect ratios >1:4 and sweat volumes >14 µL for 1 m channel length.

Reader and interrogation: A portable VNA (MetroVNA Deluxe 250 MHz) with a custom 3D-printed holder and dual-loop interrogation coils served as the wireless reader. Transmission loss (TL) magnitude and resonant frequency (f0) were recorded. Five scans were averaged per measurement to reduce noise. TL responses were analyzed in Matlab (quadratic fitting around the peak).

Benchtop characterization: The resonator was tested against different dielectric media (air, deionized water, NaCl solutions, corn oil) to observe changes in TL peak magnitude and f0. For conductivity calibration, NaCl solutions (0.01–0.1 M, ~1–10 mS/cm) served as artificial sweat. TL change (ΔTL) versus conductivity was modeled quadratically from the zero-concentration reference. Separation effects were studied by varying distance between reader and resonator; PPE measurements were performed through thick firefighter PPE (Morning Pride TAILS), and a linear correction for separation-induced ΔTL offset was derived. Volume sensing was evaluated by filling channels incrementally (eighth-turn marks) with solutions via a syringe pump and tracking Δf0 versus distance filled (d). Experiments were conducted with direct reader contact and through PPE.

On-body feasibility and orientation studies: A small human cohort was used to evaluate variability from local tissue dielectric heterogeneity and reader–sensor orientation. Measurements were taken on multiple body sites (e.g., forearm, stomach, lower back, thigh/calf), with repeated repositioning to assess intra- and inter-subject variability. Additional controlled orientation tests (translations in x/y, rotations) mapped TL and f0 dependence across the reader head (9×6 cm area).

Mechanical tests: Effects of bending on channel cross-sectional area were measured by imaging channels bent over blocks of known curvature and analyzing area changes. Cyclic bending fatigue tests were performed; frequency shift responses were periodically recorded during repeated elastic deformation.

Data and software: VNA data acquisition used WRJ software; analysis scripts (Matlab) fit quadratic models to extract TL peak magnitude and f0.

• Orthogonal sensing: The sticker simultaneously resolves sweat conductivity via TL peak magnitude and sweat volume (distance filled) via resonant frequency shift, enabling dual-parameter monitoring through PPE.
• Conductivity calibration: ΔTL versus NaCl concentration (0.01–0.1 M) fit a quadratic model with R^2 = 0.964 (model constants reported as C = 0.19 and d = −1.08). A consistent offset in PPE measurements was attributed to reader–resonator separation; the separation-induced ΔTL change was linear with slope −1.345 ΔTL per mm, yielding an offset of −1.2 ΔTL for 0.9 mm PPE thickness and improving prediction through PPE for conductivities ≥0.001 M.
• Volume (fill distance) calibration: Resonant frequency shift followed a quadratic relation with distance filled, Δf0 = p1 d + p2 d^2. On-reader fit: p1 = 7.2×10^-4 kHz/m^2, p2 = −0.27 kHz/m, R^2 = 0.754. Through PPE fit: p1 = 4.9×10^-2 kHz/m^2, p2 = −0.55 kHz/m^2, R^2 = 0.934. Filling-induced changes to ΔTL (conductivity readout) were insignificant, supporting orthogonality of the two measurements.
• Human feasibility and SNR: In on-body tests with repeated repositioning, average standard error was ~156 kHz for f0 and ~0.08 dB for ΔTL. Given dynamic ranges (~300 kHz for Δf0 and ~6 dB for ΔTL), estimated SNRs were ≥20 (frequency) and ≥10 (magnitude), suggesting robust conductivity sensing; sweat rate (frequency) is more sensitive to orientation.
• Orientation sensitivity: Large TL and frequency variations occurred with x-direction translation and rotation (>10 dB TL change and ~2 MHz frequency shift), while y-direction variations were smaller (<1.2 dB TL, <0.7 MHz). Accurate, repeatable orientation is critical for frequency-based volume readouts.
• Laser ablation models: Depth model achieved R^2 = 0.943 (RMSE 45.10 µm); width–depth linear relation had RMSE 17 µm. Rapid, low-cost laser ablation enables tuning channel dimensions to sweat rate needs.
• Read-through capability: The passive resonant sticker was interrogated wirelessly through thick, non-metal PPE, enabling under-garment monitoring with a low-cost portable VNA.

The prototype demonstrates that a passive, chipless resonant sensor integrated with laser-ablated PDMS microfluidics can monitor sweat conductivity and volume beneath clothing/PPE by mapping TL magnitude and resonant frequency, respectively. Conductivity measurements through PPE can be corrected using a simple linear offset accounting for reader–sensor separation. However, the volume (frequency) readout is highly sensitive to reader–sensor orientation and local dielectric environment, complicating calibration under different garment types and placements. Strategies to mitigate orientation sensitivity include mechanical alignment guides (tabs, magnets, hall sensors with indicators), embedded sensor arrays for positional correction, and increasing signal dynamic range by reducing the fluid–resonator separation (current separation ~0.75 mm), thereby strengthening EM coupling. Improving proximity of the microfluidic channel and optimizing resonator geometry could enhance both frequency and amplitude responses, making the system more robust to positional variations. Overall, the work validates the feasibility of orthogonal, wireless sensing of sweat properties under PPE and outlines practical engineering steps required for reliable field deployment and larger-scale human studies.

This work introduces a low-cost, passive, wireless resonant sticker with laser-ablated microfluidics that can be interrogated through thick, opaque garments to simultaneously monitor sweat conductivity and volume. Key contributions include: (1) an orthogonal sensing modality (TL magnitude for conductivity; f0 for volume), (2) calibration models for conductivity and volume with read-through PPE, including a linear separation-distance correction, and (3) feasibility evidence from benchtop and small on-body studies. Future work should: (i) increase EM coupling by bringing the microfluidic channel closer to the resonator, (ii) implement robust alignment/positioning aids or sensor arrays to control orientation effects, (iii) calibrate across garment materials and thicknesses, and (iv) extend the platform to selective, multiplexed detection of additional sweat biomarkers for comprehensive hydration and health monitoring.

• Orientation sensitivity: TL and f0 strongly depend on reader–sensor position and angle; large variations can overshadow volume-induced frequency shifts without strict alignment control.
• Garment/material effects: While a linear offset corrects conductivity (ΔTL) for separation distance, volume/frequency responses lack a simple universal correction and require calibration across clothing types and thicknesses.
• On-body variability: Local tissue dielectric heterogeneity and site-to-site variability contribute to measurement variance; studies to quantify and compensate are needed.
• Fabrication variability: Laser ablation depth exhibited RMSE ~45 µm, likely from PDMS thickness non-uniformity affecting laser focus; this may introduce channel-to-channel variability.
• Small cohort and benchtop proxies: Human testing involved a small number of participants and used NaCl solutions as sweat proxies; performance with real sweat under active sweating conditions remains to be established.
• Mechanical durability: While bending and cyclic tests were initiated, comprehensive fatigue and deformation impacts on sensor calibration were not fully detailed.