Metasurface-enabled multifunctional single-frequency sensors without external power

The study addresses the challenge of enabling maintenance-free, battery-free IoT sensors capable of sensing multiple physical parameters while conserving scarce frequency resources. Conventional metasurface-based sensors map each sensed physical property to a distinct resonant frequency, preventing independent detection of multiple parameters at a single frequency and conflicting with regulated spectrum allocations. The authors propose metasurface-based sensors that exploit time-domain scattering responses—rather than frequency-domain resonances—to sense multiple quantities at one operating frequency. By integrating photocells and temperature-dependent capacitors within diode bridge circuits in metasurface unit cells, they break harmonized oscillations, enabling independent inference of temperature and light intensity from time-domain reflections at a fixed frequency. Machine learning (random forest regression) is used to predict physical parameters from the scattered waveforms, targeting maintenance-free and sustainable next-generation wireless systems.

Prior metasurface sensors leverage subwavelength unit-cell resonances to detect physical properties such as light and temperature by embedding materials/components like vanadium dioxide, MEMS, thermistors, and photocells, with each property tied to a separate resonant frequency. Spectrum allocation constraints motivate moving beyond frequency-multiplexed sensing. Time-varying metasurfaces have been explored for wavefront control, RISs, nonreciprocal propagation, and RF/optical devices, often requiring external power. More recent passive, pulse-width-dependent (waveform-selective) metasurfaces exploit diode-bridge rectification and transient RC dynamics to produce waveform-dependent responses at constant carrier frequency. This work builds on those passive time-domain concepts to realize multifunctional sensing without external power, targeting simultaneous detection of temperature and light intensity at a single frequency.

Design theory: The metasurface unit cell consists of a ground plane, dielectric substrate (Rogers 3003), and patterned conducting patches forming a diode bridge across a gap. Inside the bridge, a parallel RC network is inserted. The diode bridge rectifies the incident sinusoid to |sin|, concentrating energy near DC and inducing transients in the RC network. The time constant τ of the transient is τ = (C R_C R_d)/(R_C + R_d). For R_C >> R_d, τ ≈ C R_d, implying temperature-dependent capacitors primarily control transient duration, while the steady-state response depends on the effective resistance R_C (photocell-dependent). By integrating temperature-dependent capacitors (C) and photocells (R_C) in parallel within the bridge, time-domain reflections encode temperature and light intensity, respectively. The approach can be extended with inductive elements to shape richer temporal responses tied to additional circuit parameters. Numerical simulations: A periodic unit cell model was simulated with C=1 nF, R_C=10 kΩ, input power 0 dBm. Frequency-domain reflectance was evaluated for short pulses (50 ns) versus CW, showing a pronounced pulse-only absorption near ~3.9 GHz. Time-domain reflectance at 3.9 GHz was computed for various C (10 nF, 100 nF) and R_C (1 kΩ, 10 kΩ), demonstrating that increasing C lengthens the transition time, while decreasing R_C lowers steady-state reflectance. Geometrical and material parameters follow Rogers 3003 substrate and realistic lumped elements (details in Methods/Supplementary). Experiments: A 15×15 unit-cell metasurface was fabricated. The sample was mounted on a programmable hot plate to set temperature and illuminated by a PWM-controlled light source to set light intensity. A horn antenna transmitted TE-polarized signals at 30° incidence; a second horn received reflections at 30°. Initial conditions were 22.1 °C and 328 lux, yielding effective C and R_C close to 10 nF and 10 kΩ. Frequency-domain reflectance in free space showed reduced reflectance for short pulses near ~3.76 GHz. Time-domain measurements isolated parameter effects by (i) using temperature-dependent capacitors with fixed resistors (10 kΩ) at 3.82 GHz, and (ii) using fixed capacitors (1 nF) with photocells at 3.82 GHz. Combined sensing used temperature-dependent capacitors plus photocells at 3.76 GHz. Machine learning via random forest regression was applied to the time-domain scattering features to predict temperature and light intensity (details referenced).

• The metasurface exhibits waveform-selective behavior: strong absorption for short pulses but high reflection for CW at the same frequency, enabling time-domain discrimination near ~3.8–3.9 GHz.
• Simulation: Increasing C from 10 nF to 100 nF shifts the transition to longer times (larger τ), while reducing R_C from 10 kΩ to 1 kΩ lowers steady-state reflectance, consistent with τ ≈ C R_d and steady-state dependence on R_C.
• Experiment (temperature effect): With temperature-dependent capacitors and fixed 10 kΩ resistors at 3.82 GHz, increasing metasurface temperature from 23.5 °C to 65.0 °C decreased transition time, indicating reduced capacitance with temperature and a smaller time constant.
• Experiment (light effect): With fixed 1 nF capacitors and photocells at 3.82 GHz, increasing light intensity from 3 lux to 1970 lux decreased steady-state reflectance from −30.9 dB to −37.5 dB, consistent with reduced photocell resistance producing stronger absorption.
• Experiment (combined): Using temperature-dependent capacitors plus photocells at 3.76 GHz, raising temperature from ~40 °C to 70 °C shifted reflectance curves to shorter times (smaller τ), and increasing light intensity from ~300 lux to ~2000 lux decreased steady-state reflectance from about −23 dB to −33 dB.
• Machine learning results: Random forest regression on time-domain scattering enabled accurate, simultaneous estimation of temperature and light intensity with determination coefficients exceeding 0.96.

By breaking harmonized time-domain oscillations using diode-bridge rectification and an internal parallel RC network, the metasurface encodes multiple physical parameters into distinct aspects of the time-domain reflection at a single carrier frequency. Temperature primarily modulates the transient time constant via the temperature-dependent capacitance, while light intensity modulates steady-state absorption via the photocell resistance. Simulations and experiments corroborate the theoretical model: larger capacitance leads to longer transients; lower resistance enhances steady-state absorption. The consistent trends across numerical and experimental results demonstrate the feasibility of single-frequency, battery-free multifunctional sensing. The use of random forest regression translates the time-domain scattering profiles into quantitative estimates of temperature and light intensity with high accuracy (R² > 0.96), validating the approach for practical IoT sensing under constrained spectrum. The design concept is extensible by adding reactive elements (e.g., inductors) to construct richer temporal signatures for additional sensing dimensions.

The work introduces a passive metasurface sensor architecture that achieves multifunctional sensing at a single operating frequency without external power. By leveraging waveform-selective, time-domain responses from diode-bridge RC transients, the system independently senses temperature and light intensity. Simulations and free-space experiments confirm control of transient and steady-state reflectance through capacitance and resistance, respectively, and machine learning delivers accurate parameter estimation (R² > 0.96). This approach addresses maintenance and spectrum constraints for large-scale IoT and can be generalized to additional physical quantities by integrating further circuit elements (e.g., inductors) and optimized layouts. Future research may focus on expanding the set of sensed parameters, robust feature extraction and ML models, integration with communication protocols, environmental robustness, and miniaturization for deployment in smart infrastructures.

• The demonstrated multifunctional sensing is limited to two physical quantities (temperature and light intensity) in this study.
• Simulations used idealized components (ordinary capacitors/resistors) and periodic boundaries, while measurements in open space exhibited overall lower reflectance and minor frequency shifts due to incidence angle and parasitic effects.
• Some experimental configurations simplified the internal circuits (either temperature-dependent capacitors with fixed resistors, or fixed capacitors with photocells) to isolate effects, which may differ from fully integrated operation.
• The approach relies on training machine learning models for parameter estimation, which requires calibration data across operating conditions.
• The resistance of the diodes (R_d) is not adjustable, constraining independent tuning of the time constant relative to capacitance.