Machine learning optimises a terahertz glucose-sensor design

Researchers modelled a graphene–black-phosphorus–vanadium-dioxide spiral resonator and trained a neural surrogate on simulation data to accelerate optimisation. The article reports resonance movement across simulated glucose concentrations and a maximum sensitivity of 1.70 THz/RIU after optimisation.

Why it matters: Surrogate models can reduce expensive electromagnetic simulation cycles and accelerate advanced-material design.

Impact: This is not evidence of a validated wearable or a clinically accurate non-invasive glucose monitor; fabrication, human testing, calibration and safety remain open. Scientific Reports

Computer vision · Poland