Stretchable and anti-impact iontronic pressure sensor with an ultrabroad linear range for biophysical monitoring and deep learning-aided knee rehabilitation

Wearable flexible pressure sensors (FPSs) that can continuously monitor biophysical information are needed for early warning and rapid rehabilitation in fitness and healthcare. Despite progress across piezoresistive, piezocapacitive, piezoelectric, and triboelectric sensing, devices that simultaneously deliver ultrahigh linear sensing range (near 1 MPa) and high sensitivity (over 10 kPa⁻¹) remain rare. Continuous monitoring during exercise or emergencies (e.g., sudden falls) imposes high pressures, demanding both broad sensing range and acceptable sensitivity. Iontronic piezocapacitive (IPC) sensors leveraging electrical double-layer (EDL) capacitance offer high signal-to-noise ratio and broad range by increasing accessible surface area and reducing ion transport distance under pressure. However, stretchable iontronic pressure sensors (SIPSs) with both high sensitivity and broad range are scarce. Furthermore, integrating deep learning can enhance monitoring and rehabilitation by enabling in-sensor classification and prediction. Here, an on-skin SIPS with an ultrabroad linear range is developed to monitor diverse biophysical features and to enable in-sensor dynamic deep learning for knee rehabilitation using a fully convolutional network (FCN). The device achieves linear sensing up to 1 MPa, high sensitivity, high resolution, robust stretchability, and resilience to sudden impacts, supporting accurate biophysical tracking and AI-aided postoperative rehabilitation assessment.

The work situates within flexible pressure sensing, noting advances in piezoresistive, piezocapacitive, piezoelectric, and triboelectric sensors, yet highlighting the scarcity of devices combining ultrabroad linear range (~1 MPa) with high sensitivity (>10 kPa⁻¹). Iontronic piezocapacitive sensors have emerged due to EDL-induced high capacitance and SNR, with prior studies (e.g., Pan et al.) employing elastic ionic-electronic interfaces to achieve high unit-area capacitance, and Guo et al. demonstrating graded intrafillable architectures achieving unprecedented sensitivity (Smin > 220 kPa⁻¹) over 0.08 Pa–360 kPa. Nevertheless, highly stretchable iontronic sensors maintaining both sensitivity and broad range are limited. Parallel developments show deep learning integrated with wearable sensors can classify physiological signals for real-time applications (e.g., ocular motion detection, hand gesture recognition), motivating in-sensor learning for rehabilitation monitoring.

Device architecture: A thin, lightweight, stretchable iontronic pressure sensor (SIPS) consists of (i) an encapsulating soft silicone elastomer layer; (ii) a serpentine polyimide (PI, Kapton) substrate with gold and carbon–graphite flake (C-GF) composite electrodes; and (iii) a nanostructured ionic membrane based on a PVA–KOH ion gel. The serpentine interconnect geometry enables conformal skin attachment and stretchability. Working principle: Under applied pressure, ions in the PVA–KOH membrane migrate to opposing electrodes to form EDLs, greatly increasing effective capacitance versus traditional parallel-plate capacitors. The device behavior is described (based on compression theory of porous fibrous assemblies) by: C = C0 − A(P/(A + E))^a, where C0 is initial capacitance, A effective sensing area, P pressure, E equivalent elastic modulus, and a an ionic composite distribution factor. Two regimes of capacitance evolution are noted, with the chosen PVA–KOH fibric membrane operating in a favorable regime with higher capacitance variation per pressure. Materials and structure characterization: The PVA–KOH ionic film exhibits abundant ~10 µm pores that enhance accessible area under load. Tensile testing shows stretchability of the PVA–KOH film (~133% strain) and C-GF films (~25% strain). Subtractive manufacturing is used to pattern arbitrary electrode geometries and arrays on the PI/Au substrate. The device is encapsulated in silicone for mechanical protection and skin-safe wearability. Mechanical analysis: Finite element analysis (FEA) and experiments assessed biaxial stretching, conformal wrapping on hemispherical surfaces (R = 8 cm), and diagonal tensile loading. Strain localizes in serpentine connections, correlating with experiments. The device tolerates at least 20% deformation at ~9 MPa limit stress; uniaxial tests of silicone-protected SIPS show >20% deformation capability. Electrical characterization: Sensitivity S = d(ΔC/C0)/d(ΔP) is measured across pressure. Capacitance–pressure curves show minimal hysteresis. Response and relaxation times are obtained under 5 kPa dynamic loading. Low-pressure resolution is evaluated via small step loads (6.4 Pa detectable). Long-term durability is tested over 12,000 cycles at 800 kPa with negligible decay. Control devices using PDMS dielectric or copper electrodes show much lower sensitivities, attributing performance to GF-enhanced carbon electrodes and the ionic dielectric synergy. Application demonstrations: The SIPS monitors biophysical signals (pulse waves, muscle movements, eye blinks, throat vibrations) and plantar pressure. High-impact tolerance is evaluated using consecutive hammering tests simulating sudden impacts near 1 MPa during exercise or falls. For rehabilitation, in-sensor data feed a neuro-inspired fully convolutional network (FCN) to learn and predict knee joint postures, providing an AI-based assessment tool post-orthopedic surgery.

- Ultrahigh linear sensing range: up to 1 MPa with highly linear ΔC/C0 vs pressure (R² ≈ 0.995). - High sensitivity: 23.10 kPa⁻¹ below 325 kPa; 12.43 kPa⁻¹ from 325 kPa to 1 MPa. - High pressure resolution: 6.4 Pa detectable at low pressures. - Fast dynamics: response time 14.2 ms and relaxation time 13.9 ms at 5 kPa. - Durability: no observable decay over 12,000 loading cycles at 800 kPa. - Stretchability and mechanical robustness: maintains integrity and function under bending, twisting, and stretching; operational up to at least 20% deformation; tolerates sudden impact pressures near 1 MPa. - Minimal hysteresis between capacitance and applied pressure. - Mechanism attribution: EDL capacitance in porous PVA–KOH ionic membrane and stable serpentine C-GF/Au electrodes with high surface area and conductivity. - Controls: PDMS dielectric and copper-electrode variants show much lower sensitivities (0.0424 kPa⁻¹ and 0.0002 kPa⁻¹), underscoring material and structural choices. - Applications: Accurate real-time monitoring of pulse, muscle movement, plantar pressure. FCN-based in-sensor learning enables accurate prediction of knee joint postures for rehabilitation assessment.

The SIPS addresses the central need for a pressure sensor that combines high sensitivity with an ultrabroad linear range to withstand real-world, high-impact scenarios while capturing subtle biophysical signals. The EDL-based iontronic mechanism in a porous PVA–KOH membrane, together with mechanically resilient serpentine C-GF/Au electrodes and silicone encapsulation, yields linear, low-hysteresis, high-SNR responses from a few pascals to 1 MPa. This broad, linear range allows one device to monitor both low-pressure physiological signals (pulse, muscle activity) and high-pressure events (plantar loading, sudden impacts) without saturation or damage. Rapid response and relaxation support dynamic monitoring. Long-cycle durability and stretchability enable on-skin use during movement. Integrating an FCN to process in-sensor data demonstrates a pathway to intelligent rehabilitation tools capable of predicting knee posture, potentially improving postoperative assessments and personalized therapy. Collectively, these outcomes enhance the practicality and clinical relevance of wearable pressure sensing systems.

This work presents a stretchable, anti-impact iontronic pressure sensor with ultrabroad linear range (to 1 MPa), high sensitivity (up to 23.10 kPa⁻¹ at low pressures; 12.43 kPa⁻¹ up to 1 MPa), high resolution (6.4 Pa), fast response, and strong durability (>12,000 cycles). The sensor reliably captures diverse biophysical signals and tolerates sudden high-pressure impacts. Coupled with an FCN, it enables accurate, in-sensor prediction of knee joint postures, illustrating potential for AI-assisted rehabilitation after orthopedic surgery. Future research may include multi-sensor arrays for spatial pressure mapping, integration into fully wireless systems, extended clinical validation across larger cohorts, and optimization of materials and mechanics for greater stretchability and long-term on-skin biocompatibility.