Comparison charts

Four synthesis charts built from the wiki’s normalized modality data (src/data/charts/modalities.json) and the enhanced-Raman dossier. Every number behind them is cited on the corresponding modality page; ranges reflect the published spread across each family, so treat them as landscape orientation, not datasheet values. Regenerate after data edits with node ops/generate-charts.mjs.

The sensitivity landscape

Log-log scatter of H₂ detection limit versus response time for 17 sensing technology families, colored by family. Dashed lines mark the 0.53 ppm atmospheric background and the 4 vol% LEL. GC (RGD/PDHID), MOX and work-function devices sit lowest; only GC and the best enhanced-Raman results resolve below background with quantitative credibility.

The lower-left corner (fast and trace-sensitive) is essentially empty: sensitivity costs either time (GC cycles, cavity integration) or selectivity (MOX’s sub-ppm claims come with severe covariance; see the matrix below). The sensor-selection guide walks the same trade space use-case by use-case.

The confounder matrix

Severity matrix (none/low/medium/high) of eight environmental confounders (temperature, humidity, O₂ dependence, CO, CH₄, H₂S, CO₂, pressure) for each technology family.

By row, optical scattering techniques (Raman) are nearly clean; anything palladium-catalytic shares the H₂S column; MOX and electrochemical cells carry the humidity and CO burden that dominates the soil-survey artifact debate (environmental covariance has the coefficients).

Temperature envelopes vs. the downhole window

Operating temperature range per technology family against shaded surface-operations and 80–300 °C downhole bands.

The 100–200 °C band, where most candidate wells sit, is served by almost nothing deployable today; that gap and its workarounds are the subject of downhole & borehole sensing.

The Raman enhancement ladder

Best demonstrated H₂ detection limit per Raman variant, log scale, from 2100 ppm (commercial process analyzer) to 60 ppb (feedback-locked multipass diode).

Spontaneous Raman’s ~10⁻³⁰ cm²/sr cross-section is the physics floor; every rung below ~200 ppm relies on an enhancement technique (pressure, multipass, cavities, hollow-core fiber), each adding its own complexity, detailed on enhanced Raman.

Why H₂ is a hard target: two physics pictures

Two bar panels comparing thermal conductivity and speed of sound for common gases; hydrogen and helium are outliers on both.

Hydrogen and helium are the only gases far from the rest on both transport properties. That outlier status is what thermal-conductivity and acoustic sensors exploit, and why helium is their worst confounder.

Log-scale bars comparing H₂'s electric-quadrupole line strength against dipole-allowed IR absorbers, spanning five to seven orders of magnitude.

Hydrogen’s only infrared coupling is the electric-quadrupole mechanism, five to seven orders of magnitude weaker than the dipole-allowed bands behind every cheap NDIR CO₂ or methane sensor. This is why direct IR sensing of H₂ is a dead end and where why H₂ is hard starts.