A single nanosensor is rarely selective — a metal-oxide film or a functionalized nanoparticle coating responds, to some degree, to many different vapors. A cross-reactive nanosensor array turns this "weakness" into a strength: it combines several sensor elements, each with a different, overlapping affinity profile, so that every analyte produces its own unique multidimensional response pattern — a chemical "fingerprint" — even though no single sensor is analyte-specific.
Element response: ΔR/R₀ (i) = k_i(analyte) · C · [1 + ξ]
k_i(analyte) — affinity of sensor i to the analyte (its coating chemistry)
C — analyte concentration (ppm)
ξ — zero-mean sensor noise, amplitude set by the slider
Classification (correlation / nearest-neighbor):
r(analyte) = Pearson correlation of the measured pattern
against each analyte's reference pattern
identified analyte = argmax_analyte r(analyte)
- Pillars — one nanosensor element each (metal-oxide, gold/thiol, silver/citrate, carboxylated carbon-nanotube, PEDOT polymer, ZnO nanowire, Pd-doped-CNT, amine-graphene coatings). Height and color encode that element's live resistance change.
- Concentration — scales every element's response roughly linearly at these low-ppm levels (Langmuir-type adsorption is linear far from saturation).
- Sensor noise — real chemiresistors have baseline drift and thermal noise; more noise degrades pattern shape and lowers classification confidence.
- Active elements — using fewer array members removes information from the fingerprint, which is why real electronic-nose/tongue arrays use many partially-selective sensors rather than one "perfect" one.
This is the operating principle behind electronic-nose and colorimetric sensor-array devices (cf. Suslick-style array noses, nanowire/CNT chemiresistor arrays): identification comes from the shape of the response vector across the array, not from any single reading.