IR absorption spectroscopy (NDIR, TDLAS, CRDS, FTIR) & sensing proxies

Summary

Most of the gas sensors that show up in industrial safety catalogs and climate-monitoring instruments — NDIR, TDLAS, FTIR, cavity ring-down — all work the same way: they shine infrared light through a gas and watch which colors get absorbed, because each gas molecule vibrates at its own characteristic frequency and “eats” light at that exact color.

Hydrogen (H₂) does not absorb light this way. Because it is two identical atoms locked in perfect symmetry, stretching or squeezing the H–H bond never creates the kind of lopsided electrical wobble that lets a molecule grab a photon this way — so H₂, like helium, is essentially invisible to every infrared instrument on the market, no matter how good the laser or how long the light path.

Any product marketed as an “infrared hydrogen sensor” is almost certainly a different kind of sensor (usually a thermal-conductivity sensor) or is measuring a completely different gas in the same box. Infrared spectroscopy is valuable to Hestia not for hydrogen itself but for the gases that travel with it: methane, carbon dioxide, ethane, and hydrogen sulfide are all strongly infrared-active, cheaply and precisely measurable, and essential for telling a genuine deep geologic hydrogen seep apart from ordinary biological gas production or a false alarm.

This page is about using infrared instruments for those companion gases, not as the H₂ detector itself.

Note

Context numbers used throughout this page: atmospheric H₂ background ≈ 0.53 ppm [51]; geologic H₂ soil-gas anomalies span roughly 1 ppm to low-thousands of ppm, with isolated press-released hits of 2,047–5,558 ppm [44,45]; H₂ LEL ≈ 4 vol% = 40,000 ppm [52]; UEL ≈ 75–77 vol% (unverified — not stated in [52]). This page covers the entire infrared-absorption technique family (NDIR, TDLAS/WMS, CRDS, OA-ICOS, OF-CEAS, QEPAS, FTIR, dual-comb) as a single unit because they share one governing physical fact: none of them can see H₂ or He, for the same symmetry reason. See Raman spectroscopy and Enhanced Raman for the sister optical techniques that do have a direct physical handle on H₂.

At a glance — published performance envelope

Detection limit

0.0003–50 ppm

Range

0–1,000,000 ppm

Response t₉₀

1–120 s

Recovery

1–120 s

Selectivity

high

Drift

low

Operating T

−25 … 200 °C

Humidity

0–100 % RH

Power

1–300 W

Unit cost

$1k–$150k

TRL

9

Main interferents: H2O vapor, CO₂, co-eluting, temperature · poisons: dust, H₂S, wetted-surface adsorption of polar. Ranges span the published devices/variants covered below — per-claim citations are in the body; data: src/data/modalities/ir-absorption-and-proxies.yaml.

Mechanism & physics

The selection-rule problem

Infrared absorption spectroscopy — NDIR, TDLAS, FTIR, CRDS, photoacoustic, dual-comb, all of it — works because a molecule’s vibration or rotation modulates an oscillating electric dipole moment that couples to the oscillating electric field of light. The transition rate (and hence the observable line intensity) is governed by the electric-dipole (E1) transition moment

$$\mu_{fi} = \langle \psi_f | \hat{\mu} | \psi_i \rangle, \qquad \hat{\mu} = \sum_k q_k \mathbf{r}_k$$

and the absorption cross-section integrated over a line (the HITRAN “line intensity” $S$, units cm$^{-1}$/(molecule·cm$^{-2}$), equivalently cm/molecule) scales as $S \propto \nu\, |\mu_{fi}|^2$ [4].

Molecular hydrogen, H$_2$, is the simplest case in which this mechanism fails by symmetry. It is a homonuclear diatomic ($D_{\infty h}$ point group). By symmetry, its dipole moment function $\mu(R)$ is identically zero for every internuclear separation $R$, not merely at equilibrium — stretching or compressing the H–H bond never breaks the inversion symmetry that forces $\mu=0$. Because $\mu(R)\equiv 0$, both $\mu(R_e)$ (which would allow pure rotational transitions) and $d\mu/dR|_{R_e}$ (which would allow vibrational transitions) vanish identically. This is a rigorous consequence of molecular symmetry, not a small effect — every rotational and rovibrational E1 transition of H$_2$ has exactly zero intensity in the electric-dipole approximation [1].

Helium is an even more extreme case: it is monatomic, so it has no vibrational or rotational degrees of freedom at all — there is no bond to stretch and no molecular framework to rotate, so there is no rovibrational spectrum of any kind, dipole-allowed or otherwise. Its only spectroscopic transitions are electronic, lying in the vacuum ultraviolet, irrelevant to any IR/near-IR sensing scheme.

This is worth contrasting with the molecules IR sensors do see. CO$_2$’s symmetric stretch ($\nu_1$) is dipole-forbidden for exactly the same symmetry reason as H$_2$ — but CO$_2$’s asymmetric-stretch and bend modes momentarily displace the two oxygen atoms asymmetrically about the carbon, transiently breaking the centrosymmetry and producing a net oscillating dipole; this is why the asymmetric stretch at ~2350 cm$^{-1}$ is IR-active while the symmetric stretch is not.

CH$_4$ has $T_d$ symmetry with zero permanent dipole, but its triply-degenerate $\nu_3$ (asymmetric stretch) and $\nu_4$ (bend) modes transform as the $T_2$ irreducible representation, which spans $(x,y,z)$ and is therefore dipole-allowed; only the totally-symmetric $\nu_1$ stretch is dark. The difference for a diatomic like H$_2$ is that there is no such non-totally-symmetric vibrational coordinate available at all: a diatomic molecule has exactly one vibrational mode, and for a homonuclear diatomic that mode can never break the symmetry that zeroes the dipole, so H$_2$ has no “asymmetric stretch” alternative.

This generalizes: N$_2$, O$_2$, Cl$_2$, and all homonuclear diatomics share the same E1-forbiddenness for their fundamental rovibrational bands (O$_2$’s visible absorption comes from an electronic, not rovibrational, quadrupole/magnetic-dipole transition) [1].

The escape hatch: electric quadrupole (and why it’s still nearly useless)

Higher multipole terms in the expansion of the light–matter interaction are not forbidden by the same symmetry argument. The next term is the electric quadrupole (E2) interaction, governed by the quadrupole moment operator $\hat\Theta$ and transition moment

$$\Theta_{fi} = \langle \psi_f | \hat{\Theta} | \psi_i \rangle.$$

H$_2$ does possess a nonzero, $R$-dependent quadrupole moment function $\Theta(R)$ (even a symmetric, non-polar diatomic has a quadrupole moment — think of it as two positive nuclei with an electron cloud smeared between them, an object with no net charge and no dipole but with genuine higher-order charge asymmetry).

Vibrational and rotational modulation of $\Theta(R)$ produces a weak but real absorption spectrum, first predicted and searched for by Herzberg, who observed eight lines of the H$_2$ (2,0) and (3,0) overtone bands in 1949 [2]. His paper states the practical consequence directly, anticipating the argument of this page: “The predicted intensity of this spectrum is so low that a path of the order of 10 km at atmospheric pressure is required to obtain it in absorption” [2].

Electric quadrupole selection rules differ from dipole selection rules: for a $^1\Sigma_g^+$ ground electronic state like H$_2$’s, E2 rovibrational transitions obey $\Delta J = 0, \pm2$ (giving O, Q, S branches) rather than the $\Delta J = \pm1$ (P, R branches) of dipole spectra — hence the nomenclature “S(0),” “S(1),” “O(2),” etc. used throughout the H$_2$ spectroscopic literature [2,9].

Magnetic dipole (M1) transitions are also symmetry-allowed for H$_2$ in principle but are calculated/observed to be even weaker than the quadrupole channel for the vibration–rotation bands, so E2 dominates what little spectrum exists [1].

Quantitatively: Somogyi, Yurchenko & Yachmenev’s first-principles variational calculation of E2 linestrengths for H$_2$, CO, HF and O$_2$ reports that “Electric quadrupole transition intensities are on the order of 10$^6$–10$^9$ times smaller than electric dipole transition intensities” [1]. That is a difference of six to nine orders of magnitude.

Putting a number on it: the strongest fundamental-band H$_2$ quadrupole line is the S(1) line of the (1–0) band at 4712.9 cm$^{-1}$ (2121.8 nm), with a HITRAN line intensity

$$S_{\mathrm{H_2, S(1)}} = 3.19\times10^{-26}\ \text{cm/molecule (296 K)}$$

as used in the only dedicated H$_2$ TDLAS sensor paper found in the literature [3]. Compare this to a typical strong fundamental-band line of methane’s $\nu_3$ band near 3.3 µm — the HITRAN-derived peak at 3057.69 cm$^{-1}$ for $^{12}$CH$_4$ has

$$S_{\mathrm{CH_4}} = 2.173\times10^{-19}\ \text{cm/molecule}$$

as reported from HITRAN calculations (hitran.ioa.ru interface, 15.5 °C, 1 atm) [6]. The ratio is

$$\frac{S_{\mathrm{CH_4}}}{S_{\mathrm{H_2,\,S(1)}}} = \frac{2.173\times10^{-19}}{3.19\times10^{-26}} \approx 6.8\times10^{6},$$

i.e., methane’s ordinary dipole-allowed fundamental is roughly 6.8 million times (∼6.8 orders of magnitude) stronger than the strongest fundamental-band line of H$_2$’s spectrum, within the “10$^6$–10$^9$” range quoted by Somogyi et al. [1].

Collision-induced absorption: real, but requires density × path that field sensing never has

There is a second, distinct absorption mechanism available to H$_2$: collision-induced absorption (CIA). During a bimolecular collision (H$_2$–H$_2$ or H$_2$–He, H$_2$–N$_2$, etc.), the transient overlap of electron clouds induces a fluctuating dipole moment on the collision complex even though neither isolated partner has one. This is a genuine electric-dipole-allowed absorption mechanism, but its absorption coefficient scales with the square of density (it is a two-body, not one-body, process): $\alpha_{\mathrm{CIA}} \propto n_1 n_2$, whereas ordinary line absorption scales linearly with the absorber’s number density, $\alpha \propto n$.

CIA is the dominant absorption/opacity source for H$_2$-He atmospheres of giant planets, brown dwarfs, and cool white dwarfs, where pressures reach many bar to kbar and path lengths are astronomical [7,8]. Recent laboratory CIA measurements of H$_2$-H$_2$ and H$_2$-He in the fundamental band (3600–5500 cm$^{-1}$) at 120–500 K quantify this mechanism at planetary-atmosphere-relevant densities [8].

At the trace (ppm-level), near-atmospheric-pressure conditions relevant to soil-gas surveys and borehole logging, the $n^2$ scaling makes CIA entirely negligible: dropping the H$_2$ mole fraction from percent-level to ppm-level suppresses $\alpha_{\mathrm{CIA}}$ by a further four to eight orders of magnitude beyond the density reduction affecting ordinary line absorption.

CIA is a real physical phenomenon and a legitimate part of “why is H$_2$ dark” pedagogy, but it is irrelevant to any Hestia field deployment.

Computing the “absurd” pathlength for 1 ppm H2

The impracticality can be made quantitative with a simple Beer–Lambert estimate, using only the published HITRAN-class line intensity and pressure-broadening width for the strongest H$_2$ line [3]. For a Lorentzian line, the peak absorption cross-section is

$$\sigma(\nu_0) = \frac{S}{\pi\gamma}$$

where $\gamma$ is the (air-broadened, 1 atm) half-width at half-maximum in cm$^{-1}$. Using $S = 3.19\times10^{-26}$ cm/molecule and the measured air-broadening coefficient $\gamma_{\mathrm{air}} = 0.0081\ \mathrm{cm^{-1}\,atm^{-1}}$ [3]:

$$\sigma(\nu_0) = \frac{3.19\times10^{-26}}{\pi \times 0.0081} \approx 1.25\times10^{-24}\ \mathrm{cm^2/molecule}.$$

At 298 K, 1 atm, total number density $N \approx 2.46\times10^{19}\ \mathrm{cm^{-3}}$. At 1 ppm H$_2$ mole fraction, $n_{\mathrm{H_2}} = N \times 10^{-6} \approx 2.46\times10^{13}\ \mathrm{cm^{-3}}$, giving a peak absorption coefficient

$$\alpha = \sigma(\nu_0)\, n_{\mathrm{H_2}} \approx 3.1\times10^{-11}\ \mathrm{cm^{-1}}.$$

The pathlength $L$ required to reach a given minimum detectable absorbance $A_{\min} \approx \alpha L$ is then $L = A_{\min}/\alpha$:

Achievable $A_{\min}$ (fractional absorbance)	Representative technology class	Required $L$ for 1 ppm H$_2$
$10^{-4}$	Simple single-pass diode-laser direct absorption	≈ 32 km
$10^{-6}$	Good WMS-TDLAS / short cavity-enhancement	≈ 320 m
$10^{-8}$–$10^{-10}$	High-finesse CRDS-class effective path multiplication	≈ 0.3 cm – 3 m of equivalent absorption path (but built from thousands of bounces inside a compact physical cell)

This is an order-of-magnitude estimate — it ignores Doppler broadening, which for H$_2$’s light reduced mass is actually comparable to or larger than the pressure-broadened width at 1 atm (a quick estimate gives a Doppler HWHM of order 0.02 cm$^{-1}$ at 296 K near 4713 cm$^{-1}$, versus 0.0081 cm$^{-1}$ pressure HWHM), so a full Voigt treatment would somewhat lengthen these numbers, not shorten them.

It is nonetheless entirely consistent with two independent, directly-sourced anchors: Herzberg’s own 1949 estimate of “order of 10 km at atmospheric pressure” to see the overtone spectrum at all [2], and the empirical performance of the one purpose-built H$_2$-TDLAS instrument in the literature.

That instrument — Avetisov et al.’s dedicated 2f-WMS TDLAS sensor built specifically around the 2121.8 nm S(1) line, using a 12 m Herriott multipass cell — reports a precision figure of merit of 0.02 %v·m (1σ, 1 s integration), improving to 0.005 %v·m after 20 s of averaging [3]. Dividing by their actual 12 m pathlength: at 1 s integration the concentration-equivalent noise floor is $0.02\%v\cdot m / 12\,m \approx 1.7\times10^{-3}\%v \approx 17$ ppm; after 20 s of averaging it improves to roughly $4\times10^{-4}\%v \approx 4$ ppm.

In other words, a laboratory-grade instrument purpose-built around the strongest H$_2$ absorption line, with a 12-meter folded optical path and lock-in wavelength-modulation detection, still cannot cleanly resolve atmospheric background H$_2$ (0.53 ppm) or the low end of exploration anomalies (~1 ppm) without extraordinary averaging, and this is under laboratory conditions, before soil-gas dust, humidity swings, and vibration are added. This is the sourced version of the pathlength estimate above.

Why NDIR specifically cannot ever work for H2

Non-dispersive infrared (NDIR) sensing works by passing broadband IR light through a gas sample and a narrowband optical filter tuned to a specific dipole-allowed absorption band, then measuring the intensity loss on a broadband detector (thermopile, pyroelectric, or photodiode) [16]. Because H$_2$ has no dipole-allowed absorption band anywhere in the accessible IR, there is categorically no filter wavelength at which NDIR can register an H$_2$-specific signal. Multiple independent sources confirm this: NDIR sensors cannot detect hydrogen because H$_2$ does not absorb infrared light, and the same holds for other IR-inactive gases such as O$_2$ and the noble gases [12].

What is actually sold as “NDIR hydrogen sensor.” Any product marketed as an “NDIR hydrogen sensor” is, by the physics above, either (a) actually a thermal conductivity detector (TCD) rebadged with “IR” in the name for market reasons, (b) a catalytic-bead or electrochemical sensor with an NDIR module bundled alongside it for a different gas in the same enclosure, or (c) mislabeling a genuine but non-NDIR optical technique (TDLAS/CRDS targeting the E2 line, which is a real and much harder engineering feat, not “NDIR” in any meaningful sense). This mislabeling of a TCD as an IR sensor is a common source of confusion Hestia field teams are likely to encounter when vetting vendor claims — see thermal conductivity & acoustic sensors for the real (non-optical) physics behind most “IR hydrogen sensors” actually on the market.

Variants & subtypes

The IR-absorption technique family is broad; what varies is how the optical path length, wavelength selectivity, and detection scheme are engineered to push $A_{\min}$ down. None of the variants below change the fundamental H$_2$/He darkness argument above — they only change how well the technique can see the proxy gases discussed later on this page.

Detection-physics variants

• NDIR (non-dispersive infrared). Broadband IR source, bandpass filter(s) selecting a dipole-allowed absorption band, single- or dual-beam thermopile/pyroelectric detector; simplest, cheapest, most field-hardened optical gas technique; ppm-to-percent range; workhorse for CO$_2$, CH$_4$, CO in low-cost and mid-tier instruments [12,16].

• TDLAS (tunable diode laser absorption spectroscopy), direct absorption. A narrow-linewidth diode or DFB laser is tuned across a single ro-vibrational line; absorption is read directly from transmitted intensity. High selectivity because the laser addresses one line, but limited by $1/f$ and baseline noise to $A_{\min}\sim10^{-3}$–$10^{-4}$ in simple form [24,25].

• WMS-TDLAS (wavelength-modulation spectroscopy). The laser wavelength is dithered at frequency $f$ and the signal demodulated at $2f$ (or higher harmonics) with a lock-in amplifier; this shifts detection away from $1/f$ noise and typically improves $A_{\min}$ by one to three orders of magnitude versus direct absorption, at the cost of needing careful calibration or line-shape modeling for absolute concentration [3,24,25]. This is the technique used in the one dedicated H$_2$ sensor found in the literature [3], and the default choice for OEM laser gas modules (Axetris LGD Compact, mirSense multiSense) [34,35].

• CRDS (cavity ring-down spectroscopy). Light is injected into a high-finesse cavity (mirror reflectivity >99.99%) and the decay time of light leaking out is measured after the source is switched off; because decay rate rather than absolute intensity is measured, the technique is immune to laser-intensity noise and reaches extraordinary sensitivity ($\alpha_{\min}$ as low as $10^{-10}$–$10^{-13}$ cm$^{-1}$ in the best implementations) [17]. Invented by O’Keefe & Deacon in 1988 [17]; commercialized at scale by Picarro for CH$_4$/CO$_2$/CO/isotope analysis [27,28,29]. CRDS has also been used, in laboratory metrology settings, to directly measure H$_2$’s own quadrupole spectrum with noise-equivalent absorption down to $\alpha_{\min}\approx5\times10^{-11}$–$5\times10^{-12}\ \mathrm{cm^{-1}}$, resolving overtone lines with intensities as low as $1.1\times10^{-30}$ cm/molecule [9,10,11] — proof that the physics can, in principle, be pushed far enough to see H$_2$ optically. But these are neat or high-concentration H$_2$ gas-cell measurements built for spectroscopic line-list metrology, using research-grade mirror sets and long signal averaging, not field-deployable ppm-in-air trace sensors; nobody has translated this into a commercial ambient-air H$_2$ analyzer, for the practical reasons below.

• OA-ICOS (off-axis integrated cavity output spectroscopy). A fourth-generation cavity-enhanced technique: the laser beam traverses a high-finesse cavity off-axis so that many longitudinal cavity modes are excited simultaneously, avoiding the need to actively lock the laser to a single cavity resonance (as CRDS effectively requires) [18]. This trades a little sensitivity for much greater mechanical/thermal robustness and lower cost, which is why Los Gatos Research (now ABB) built an entire commercial analyzer line around it [30,31]. Effective path lengths of kilometers are obtained from physical cavities tens of centimeters long; original demonstration reached a minimum detectable absorption of $\sim1.4\times10^{-5}$ over a 4.2 km effective path, i.e., detection sensitivity $\sim3.1\times10^{-11}\ \mathrm{cm^{-1}\,Hz^{-1/2}}$ [18].

• OF-CEAS (optical-feedback cavity-enhanced absorption spectroscopy). A variant in which optical feedback from the cavity itself locks and narrows the laser linewidth, giving CRDS-like sensitivity with faster (1 Hz) update rates; this is the technique underlying LI-COR’s LI-7810/LI-7815 trace-gas analyzers [26].

• QEPAS (quartz-enhanced photoacoustic spectroscopy) and conventional PAS. Instead of measuring transmitted light, a modulated laser is tuned to a gas absorption line and the periodic heating/cooling of the gas is detected as a pressure wave — in QEPAS, by a quartz tuning fork resonating at its ~32 kHz mechanical resonance, acting as an extremely narrow-band, low-noise acoustic transducer [19,20]. QEPAS cell volumes are ~mm$^3$, making it attractive for compact, rugged OEM modules (mirSense’s multiSense is a QCL + heated photoacoustic cell) [35,36]. Zero-absorption background is intrinsically near-zero (no light reaches a detector at all in the ideal case), giving good long-term stability. Because it detects a heat/pressure signal rather than transmitted photons, PAS/QEPAS requires a still, definable gas volume and generally needs a buffer/carrier gas with adequate heat capacity — it is not a pure line-of-sight open-path technique like NDIR/TDLAS/dual-comb.

• FTIR (Fourier-transform infrared). A Michelson interferometer scans an optical path-length difference while a broadband IR source illuminates the sample; a single scan yields the full mid-IR spectrum, allowing simultaneous multi-component analysis of tens of species from one interferogram via post-hoc classical least-squares fitting against reference spectra [33]. Trades per-line sensitivity for enormous multiplex/simultaneity advantage; workhorse of stack-emissions and multi-gas process monitoring (Gasmet DX4000: up to 50 gas compounds simultaneously) [33].

• Dual-comb spectroscopy (DCS). Two phase-locked frequency combs with slightly different tooth spacing beat against each other, mapping the entire optical spectrum onto an RF comb that can be digitized directly — effectively “FTIR without moving parts,” at far higher spectral resolution and acquisition speed [21]. First demonstrated coherently at NIST in 2008 [21]; field open-path demonstrations have measured CO$_2$, CH$_4$, H$_2$O, HDO over 2 km round-trip open-air paths with 0.14% (0.57 ppm) precision on CO$_2$ and 0.35% (7 ppb) on CH$_4$ over two-week campaigns [21], and multispecies agricultural flux (CH$_4$, NH$_3$, CO$_2$, H$_2$O) with fluxes agreeing to within 6% of a co-located CRDS instrument [23]. Currently the most exotic/expensive end of the technique ladder, mostly deployed in research-grade open-path and airborne-controlled-release contexts (e.g., CH$_4$/C$_2$H$_2$ leak quantification within 10-50% of known flux) [22].

Deployment topology

• Extractive. Gas is pumped through a sample cell (most NDIR, CRDS, OA-ICOS, FTIR, QEPAS instruments); allows conditioning (drying, particulate filtration, pressure/temperature control) at the cost of pump power, tubing lag, and potential surface losses of sticky gases (H$_2$S, NH$_3$).

• Open-path. A collimated or retroreflected beam traverses ambient air directly (dual-comb, some TDLAS); no sample handling artifacts or lag, but subject to atmospheric turbulence, beam-steering losses, dust/fog attenuation, and cannot easily isolate a specific soil-gas probe volume.

• Flux-chamber (static or dynamic). A chamber is placed over the soil surface and headspace gas concentration change is tracked over time (closed/static) or continuously flushed and measured (dynamic/open); the standard method for quantifying areal gas flux (not just concentration) in soil-gas and GHG work, and directly compatible with any extractive analyzer plumbed to the chamber headspace (LI-7810/LGR/Picarro product literature explicitly targets this mode) [26,31].

Performance envelope

All rows below describe the proxy-gas-sensing performance of the technique, not H2 sensing (which the mechanism section above rules out). Every quoted figure is drawn from a fetched datasheet, paper, or search-verified snippet.

Technique	Typical target gas(es)	LOD (as reported)	Range	Response time (T90)	Selectivity	Notes / source
Low-cost NDIR	CO2, CH4	~50 ppm + 3–5% of reading (consumer/IAQ grade)	0–5,000 to 0–10,000 ppm	3–120 s	Medium (broadband filter overlap possible)	Winsen/DFRobot/Prana-class modules, search-verified
OEM TDLS/NDIR-class laser module (Axetris LGD Compact-A)	CH4	LDL ≤0.4 ppm (≤0.15 ppm with 10 s averaging); CH4/CO2 dual: CH4 LDL ≤0.4/0.8 ppm, CO2 LDL ≤2,000 ppm (≤600 with 10 s avg)	0–100 / 0–2,000 / 0–40,000 ppm (CH4); 0–100,000 ppm (CO2)	T90 ≤1.8 s (CH4, 2 l/min)	High	[34]
Portable direct-absorption TDLAS (dedicated H2 lab prototype, for contrast)	H2 (not a field product)	~4–17 ppm-equivalent (12 m cell, 1–20 s averaging, derived above)	0–10 %v validated	~1 s per scan	High (single line)	[3]
WMS-TDLAS handheld (Aeris MIRA Pico family)	CH4, C2H6, CO2, CO, N2O, CH2O, OCS, C2H2	”sub-ppb precision” claimed; models differ by species	ppb to thousands of ppm	1–5 Hz acquisition	High	Vendor page; exact per-gas numbers not independently published [32]
OF-CEAS (LI-COR LI-7810)	CH4, CO2, H2O	CH4: 0.6 ppb (1 s) / 0.25 ppb (5 s) @ 2 ppm; CO2: 3.5 ppm (1 s)/1.5 ppm (5 s) @ 400 ppm	CH4 0–100 ppm; CO2 0–10,000 ppm; H2O 0–60,000 ppm	≤2 s (CH4, 0→2 ppm)	High	Full datasheet fetched [26]
CRDS (Picarro G2201-i, isotope mode)	CO2, CH4 (+ δ13C)	CO2: 200 ppb + 0.05% (1σ, 30 s); CH4: 1 ppm + 0.25% (12C, isotope-only mode)	Ambient to % level (model-dependent)	Seconds (interleaved sampling)	High	Datasheet [27]
CRDS/GasScouter (Picarro G4302)	CH4, C2H6	Ethane/methane mode: CH4 30 ppb (1 s)/3 ppb (100 s); C2H6 10 ppb (1 s)/1 ppb (100 s). CH4-only mode: 3 ppb (1 s)/0.3 ppb (100 s)	ppb–ppm	~1 s	High (isotopic/ratio-capable)	Datasheet + search [29]
OA-ICOS (ABB/LGR UGGA)	CH4, CO2, H2O	CH4: 2 ppb (5 s)/0.6 ppb (100 s); CO2: 300 ppb (5 s)/100 ppb (100 s)	CH4 0.01–100 ppm (to 10% w/ extended range); CO2 200–20,000 ppm	0.01–1 Hz selectable	High	Full datasheet fetched [31]
FTIR (Gasmet DX4000)	Up to 50 gases simultaneously (CO2, CH4, CO, H2S, NH3, VOCs, etc.)	ppm to percent, species-dependent (not itemized in datasheet)	ppm–% (path-length selectable: 2.5/5.0/9.8 m)	T90 <120 s typical	Medium–high (multi-component least-squares fit, some cross-interferences)	Full datasheet fetched [33]
QEPAS	CH4	1.3 ppb (1000 s integration); ~18–90 ppb (0.1–1 s) reported across studies	ppb–ppm	0.1 s–1000 s depending on target LOD	High	[39]
QEPAS (THz, pulsed)	H2S	~360 ppb (10 s integration)	ppb–ppm	Tens of seconds	High	[40]
QCL-WMS (ATEX 3-channel prototype)	H2S, CH3SH, COS	H2S ≈0.3 ppmv; CH3SH ≈60 ppbv; COS ≈5 ppbv	ppb–ppm (process-stream)	Seconds	High	[38]
Dual-comb spectroscopy (open path)	CO2, CH4, H2O, HDO (2 km round-trip)	0.14% (0.57 ppm) CO2; 0.35% (7 ppb) CH4 over 2-week campaign	Ambient background to elevated plumes	Seconds to minutes (averaging-dependent)	High (full spectral fingerprint)	[21]

Environmental limits (T/P/RH), drift, power, cost class, drawn from the same sources:

• Operating temperature: portable extractive analyzers typically −25 to 45 °C (LI-7810 [26]); OEM laser modules −10 to 55 °C depending on target gas configuration [34]; FTIR sample cells are actively heated to 180–200 °C specifically to prevent condensation of “sticky”/soluble species and handle corrosive streams [33], while the electronics enclosure itself is only rated 0–40 °C short-term [33].
• Humidity: LI-7810 rated 0–85% RH ambient (non-condensing) and up to 99.9% RH in the sample line itself (non-condensing) because it corrects for water-vapor dilution/broadening in real time using its own measured spectra [26]; Gasmet and most extractive analyzers require non-condensing sample gas at all times, with heated lines/cells the standard mitigation [33].
• Pressure: LI-COR OF-CEAS rated 50–110 kPa depending on configuration [26]; Gasmet reports a linear ~1% measurement error per 1% sample-pressure change, actively compensated [33]; TDLAS/WMS techniques in general require pressure control or correction because line width (and hence WMS 2f signal amplitude) is pressure-dependent [3,24].
• Drift: LI-7810 CH4 channel drift <1 ppb per 24 h [26]; ABB/LGR OA-ICOS is marketed explicitly on “negligible zero and span drift… significantly reduced need for regular calibration” versus older CRDS/direct-absorption designs [30]; Gasmet FTIR specifies zero-point drift <2% of range per 24 h calibration interval and states “sensitivity drift: none” between calibrations [33] — these optical techniques are, as a class, markedly lower-drift than electrochemical or catalytic-bead sensors.
• Power: OEM laser modules can run on a few watts (Axetris LGD Compact is designed for “low power consumption for battery-powered applications” [34]); portable analyzers run 18–70 W (Aeris MIRA Pico 18 W steady-state [32]; LI-7810 22 W steady-state, up to 100 W during battery-charging warmup [26]; LGR UGGA 60 W [31]); bench FTIR draws up to 300 W with a hot (180 °C) sample cell and Peltier-cooled MCT detector [33].
• Cost class (qualitative — public list pricing is essentially never published by these vendors; classes below are inferred from product positioning, not verified invoice figures, and should be treated as approximate): OEM bare laser/photoacoustic modules for embedding in a third-party instrument — lowest capital tier, order $10^3$; handheld/portable single- or dual-species laser analyzers (Aeris MIRA-class, Axetris-integrated products) — mid tier, order $10^4$; multi-gas cavity-enhanced or FTIR bench/portable analyzers (Picarro, ABB/LGR, Gasmet) — highest commercial tier, order $10^4$–$10^5$; open-path dual-comb systems remain largely research infrastructure rather than a purchasable product line.

Figures of merit

The universal comparison axes — sensitivity, LOD, selectivity, response time, drift, and the rest — are defined on the Figures of merit page; this section only specializes them to what an IR-absorption spec sheet actually reports. Because H$_2$ itself has no usable IR line (see the mechanism section above), every figure below describes the proxy gases (CH$_4$, CO$_2$, CO, C$_2$H$_6$, H$_2$S); concentrations are quoted in the units defined on the Concentration units page.

Figures-of-merit table for this modality (11 rows)

Figure of merit	Symbol / formula	What it means here	Typical value for this technology	How to read it
Sensitivity (Beer–Lambert slope)	$\dfrac{\partial A}{\partial C} \approx \sigma(\nu_0)\,L = \dfrac{S}{\pi\gamma}L$	How much absorbance $A$ grows per unit proxy-gas concentration — fully determined by the gas’s intrinsic line strength $S$ and broadening $\gamma$, then multiplied up by engineered path length $L$.	Not usually quoted as a standalone number; it is the product of the Line strength and Path length rows below, and shows up indirectly as the instrument LOD.	Unlike an electrochemical cell’s nA/ppm, there is no separate calibration slope to specify — for IR absorption, “sensitivity” is line strength times path length, so the only two engineering levers are picking a stronger line and building a longer path.
Line strength $S$ (cm/molecule)	HITRAN line intensity $S$	Intrinsic quantum-mechanical strength of one absorption line — fixed by the target molecule’s dipole transition moment, not by instrument design.	CH$_4$ $\nu_3$ fundamental at 3057.69 cm$^{-1}$: $S=2.173\times10^{-19}$ cm/molecule (unverified — secondary HITRAN-derived source [6], not a raw HITRAN export). Contrast H$_2$’s strongest quadrupole line, $S=3.19\times10^{-26}$ cm/molecule [3] — about $6.8\times10^6\times$ weaker [1].	This is why direct H$_2$ IR sensing is impractical and proxy sensing is easy: CH$_4$/CO$_2$/CO all have ordinary dipole-allowed fundamentals in the $10^{-17}$–$10^{-19}$ cm/molecule class, six-plus orders of magnitude stronger than anything H$_2$ offers, and no amount of instrument engineering changes this number.
Noise-equivalent absorption, $\alpha_{\min}$ (cm$^{-1}$·Hz$^{-1/2}$)	$\alpha_{\min}$	Smallest absorption coefficient detectable per unit bandwidth — the standard way to compare cavity-enhanced techniques’ raw sensitivity independent of the specific path length used.	CRDS, best-in-class implementations: $10^{-10}$–$10^{-13}$ cm$^{-1}$ [17]. OA-ICOS: $\sim3.1\times10^{-11}$ cm$^{-1}$·Hz$^{-1/2}$ over a 4.2 km effective path [18].	Lower is better; must be paired with the path length actually used (next row) to convert to a concentration LOD via Beer–Lambert — quoting $\alpha_{\min}$ alone without $L$ is like quoting a car’s horsepower without its weight.
Path length $L$ (physical / effective)	$L$	How far, or how many times, light traverses the gas sample — the main engineering lever for pushing detection limits down without changing the gas’s line strength.	Simple single-pass TDLAS: order cm. Herriott multipass cells: 5–12 m physical, folded [3,33]. Cavity-enhanced (CRDS/OA-ICOS): km-scale effective path from physical cavities tens of cm long via >99.99% mirror reflectivity [17,18].	Longer effective $L$ directly lowers LOD for a fixed line strength, but cavity techniques trade this gain for mirror-fouling fragility — dust, aerosol, or condensate on a >99.99%-reflectivity mirror degrades sensitivity fast (see Environmental covariance, below).
Limit of detection (LOD), proxy gas, at stated integration time	$\mathrm{LOD}=k\,\sigma_{blank}/S_{\text{instrument}}$ (see Figures of merit for the competing $k$ conventions)	Smallest proxy-gas concentration change reliably resolved above noise.	CH$_4$: 0.6 ppb (1 s)/0.25 ppb (5 s), OF-CEAS [26]; 2 ppb (5 s)/0.6 ppb (100 s), OA-ICOS [31]. CO$_2$: 200 ppb+0.05% (1σ, 30 s), CRDS [27]. Low-cost NDIR: ~50 ppm + 3–5% of reading, consumer/IAQ grade (integration time not stated).	Every LOD above is only meaningful with its integration time: going from 1 s to 100 s buys roughly an order of magnitude via averaging until drift takes over (the Allan-deviation floor described on the Figures of merit page). Compare two vendor numbers only at the same τ.
Selectivity / cross-sensitivity	$K_{ij}=S_j/S_i$	How well the instrument’s chosen line or filter distinguishes the target proxy gas from interferents — mainly H$_2$O and CO$_2$/CH$_4$ spectral overlap.	Axetris CH$_4$/C$_2$H$_6$ module: ethane reliably distinguishable only at C$_2$H$_6$:CH$_4$ ratios of roughly 1:200 to 1:10, flagged “not detectable” outside that window (worked example: 50 ppm C$_2$H$_6$ in 20,000 ppm CH$_4$ = 1:400, not detectable) [34]. H$_2$O is the dominant cross-interferent across this entire technique family, corrected via simultaneous H$_2$O measurement in OF-CEAS [26].	Laser/cavity techniques rate “High” selectivity because they address one line, but that rating assumes the interferent stays inside a tested ratio window; check it against expected field CH$_4$:C$_2$H$_6$ ratios and humidity swings.
Response / recovery time ($t_{90}$)	$t_{90}$	How fast the reading catches up to a real concentration step (and falls back after).	Axetris LGD Compact-A, CH$_4$: $t_{90}\le1.8$ s [34]. LI-7810: $t_{90}\le2$ s (CH$_4$, 0→2 ppm) [26]. Gasmet FTIR: $t_{90}<120$ s [33].	Spectroscopic techniques are gas-exchange-limited (sample-cell flush/purge time), not surface-chemistry-limited — $t_{90}$ tracks cell volume and flow rate more than detection physics, which is why extractive designs are slower than open-path by this exact mechanism.
Dynamic / linear range	$N_{\text{decades}}=\log_{10}(C_{\max}/C_{\min})$	Span over which one sensitivity curve holds.	OA-ICOS CH$_4$: 0.01–100 ppm, extendable to 10% v [31]. CO$_2$: 200–20,000 ppm [31]. FTIR: path-length-selectable (2.5/5.0/9.8 m) trades range for sensitivity [33].	Unlike MOX/FET sensors, Beer–Lambert absorption is genuinely linear in concentration over many decades (until saturation), so a quoted dynamic range is close to the true linear range — a rare case where the two nearly coincide.
Drift / stability	$d_Z,\ d_S$ (zero, span drift per unit time)	How much the zero-point or slope creeps between calibrations.	LI-7810 CH$_4$ channel: <1 ppb/24 h [26]. Gasmet FTIR: zero drift <2% of range/24 h, “sensitivity drift: none” between calibrations [33]. ABB/LGR OA-ICOS marketed on “negligible zero and span drift” [30].	As a class, these optical techniques run markedly lower-drift than electrochemical or catalytic-bead sensors — but “negligible” and “none” are vendor claims, not warranted numbers; budget a field re-zero cadence regardless.
Power	—	Continuous electrical draw — mostly laser/QCL + thermoelectric cooler + sample pump, not a heated catalytic surface.	OEM laser modules: a few W, “designed for battery-powered applications” [34]. Portable analyzers: 18–70 W (Aeris MIRA Pico 18 W [32]; LI-7810 22 W steady-state, up to 100 W during charging [26]; LGR UGGA 60 W [31]). Bench FTIR: up to 300 W with a 180 °C heated cell [33].	Power scales with instrument class (OEM module → portable → bench FTIR), tracking the LOD–response time–power triangle from the Figures of merit page — the lowest-power options are also the least sensitive and slowest.
Operating envelope (T/P/RH)	$C_{\text{reported}}=C_{\text{raw}}\cdot f_T(T)\cdot f_{RH}(RH)\cdot f_P(P)$	The T/P/RH window inside which the stated accuracy, not just “produces a number,” actually holds.	Portable extractive analyzers: −25 to 45 °C [26]. OEM modules: −10 to 55 °C [34]. FTIR sample cell heated to 180–200 °C, electronics enclosure rated 0–40 °C [33]. LI-7810: 0–85% RH ambient, up to 99.9% RH in-sample-line (non-condensing) [26]; pressure 50–110 kPa, with FTIR reporting ~1% error per 1% sample-pressure change (actively compensated) [33].	Soil-gas and downhole T/P/RH swings routinely exceed a portable analyzer’s compensated range even while it stays “operating” and keeps producing a number — cross-check against the Environmental covariance section above before trusting a reading at the extremes.

What dominates. Line strength $S$ is fixed by molecular physics and cannot be engineered — it is the reason proxy-gas sensing is easy (CH$_4$/CO$_2$/CO fundamentals are six-plus orders of magnitude stronger than any H$_2$ transition) and direct H$_2$ IR sensing is not viable. Given that floor, what actually decides field usability is the trade between path length/noise-equivalent absorption on one side and power and mirror-fouling robustness on the other: cavity-enhanced techniques (CRDS, OA-ICOS) buy huge LOD headroom by turning a compact physical cell into a kilometers-long effective path, but they pay for it in mirror-fragility risk in dusty, wet soil-gas probes and in tens-of-watts continuous power draw — a direct instance of the LOD–response time–power triangle on the Figures of merit page. The second-order decider is selectivity: broadband or coarse-filter instruments (NDIR, FTIR) are vulnerable to H$_2$O and CO$_2$/CH$_4$ spectral overlap in exactly the humid, high-methane soil-gas settings Hestia works in, which is why single-line laser techniques (TDLAS/WMS, CRDS) are preferred whenever the C$_2$H$_6$/CH$_4$ ratio itself is the thing being measured.

Environmental covariance & cross-sensitivity

Because every technique in this family is fundamentally a spectroscopic measurement, its failure modes are qualitatively different from electrochemical or catalytic sensors, but they are real and field-relevant.

• Water vapor. H$_2$O has an extremely dense, strong rovibrational spectrum across the entire mid-IR and much of the near-IR, and it is a common cross-interferent across every technique in this table. Two distinct mechanisms matter: (1) direct spectral overlap — H$_2$O lines can sit under or near the target line, requiring either careful line selection, drying, or simultaneous H$_2$O measurement and correction (LI-7810 measures H$_2$O explicitly and reports CH$_4$/CO$_2$ on a dry-mole-fraction basis without physical drying [26]); (2) pressure/collisional broadening — H$_2$O is an efficient broadening partner, so uncorrected humidity swings shift both the linewidth and the peak height used by a WMS-2f measurement, biasing concentration retrievals if not modeled. Field soil-gas humidity swings (near-saturation in wet soils, near-zero in arid conditions) make this the dominant real-world confounder for any of these instruments deployed without drying or explicit H$_2$O correction.

• Temperature and pressure. Line intensity $S$ itself is temperature-dependent (via the Boltzmann population of the lower state and partition function $Q(T)$, both tabulated in HITRAN [4,5]), and linewidth $\gamma$ scales with both temperature (as $T^{-n}$, typically $n\approx0.5$–1) and pressure (linearly, near-ambient). An instrument calibrated at one T/P and deployed at another without correction will show a systematic concentration bias.

  Gasmet’s FTIR datasheet quantifies this directly: <2% of range error per 10 K temperature change, and ~1% measurement error per 1% sample-pressure change (the latter actively compensated in their firmware) [33]. This matters acutely for soil-gas and downhole work, where diurnal soil temperature swings and barometric pumping are large and well-documented drivers of gas-flux variability — the Harvey/Perth Basin comparison study explicitly attributes short-duration H$_2$ pulses (up to 598 ppm against an 8–14 ppm baseline) to temperature/pressure/soil-moisture-driven flux transients that point sampling alone could not resolve [43].

• CO2 and broadband hydrocarbon overlap. CO$_2$ has a dense, strong spectrum with bands that can overlap CH$_4$ features at coarse (filter-based NDIR) resolution; this is a standard reason multi-gas NDIR and even some laser instruments require careful filter/line selection or simultaneous CO$_2$ measurement and subtraction. Axetris explicitly documents this class of interference for its CH$_4$/C$_2$H$_6$ dual-gas module: ethane is only reliably distinguishable from a methane matrix at C$_2$H$_6$:CH$_4$ ratios of roughly 1:200 to 1:10, and becomes not detectable outside that ratio window at their reference conditions (their worked example: 50 ppm C$_2$H$_6$ in 20,000 ppm CH$_4$, ratio 1:400, is flagged “C2H6 not detectable”) [34]. This is directly relevant to Hestia’s C$_2$H$_6$/CH$_4$-ratio thermogenic-vs-biogenic source discrimination workflow — high absolute methane can blind the ethane channel.

• H2S and corrosive/sticky gases. H$_2$S is both spectroscopically detectable via IR/QCL techniques and chemically aggressive to instrument wetted surfaces, mirrors, and gaskets over time; the ATEX-certified petrochemical QCL sensor explicitly needed automated gas conditioning, purge/pressurization, and intrinsically safe electronics to operate reliably in H$_2$S-bearing hydrocarbon streams [38]. “Sticky” polar/soluble gases (H$_2$S, NH$_3$, HF) also adsorb onto sample-line and cell walls, producing long memory effects/response tails unless lines are heated and inert-coated — a known driver of Gasmet’s heated (up to 180–200 °C), rhodium/gold-coated sample-cell design [33].

• Particulates and mirror/window fouling. Cavity-enhanced techniques (CRDS, OA-ICOS) depend on mirror reflectivity exceeding 99.99%; any dust, aerosol, or condensate deposit on the mirrors directly degrades cavity finesse and hence sensitivity — this is precisely why Avetisov et al. note that cavity-enhanced techniques for H$_2$ sensing specifically “faced mirror contamination issues in industrial settings,” motivating their choice of a simpler, robust TDLAS/WMS approach instead [3]. Soil-gas probes pulling gas through dusty, wet, biologically active ground are a worst-case environment for any mirror-based instrument without aggressive upstream filtration.

• O2/N2/inert bulk-matrix effects. Because none of these techniques require oxygen for their detection chemistry (unlike electrochemical or catalytic-bead sensors, see electrochemical cells and catalytic pellistors), bulk matrix composition (N$_2$, O$_2$, or even anoxic CO$_2$/N$_2$-dominated soil gas) does not poison the measurement the way it can poison combustion- or oxidation-based sensors. The main matrix-related effect is collisional (pressure) broadening from the bulk gas, a correctable/predictable line-shape effect rather than a poisoning mechanism [3,4].

Poisoning and failure modes

Mirror/window fouling by dust, aerosol, or condensation (cavity techniques); wetted-surface adsorption and long response tails for sticky gases (H$_2$S, NH$_3$) without heated lines; laser mode-hop or wavelength drift with temperature (mitigated by TEC stabilization, standard in all these instruments); pump/flow-path leaks introducing ambient-air dilution in extractive systems; and, for open-path/dual-comb systems, beam-steering loss or complete signal dropout in fog, heavy dust, or heavy rain.

Use for geologic H₂

Field verdict

For sensing H₂ itself, none of these techniques are usable. Longer cells, better lasers, and colder detectors do not change this: H$_2$’s only accessible transitions are $10^6$–$10^9\times$ weaker than an ordinary dipole-allowed line [1], and even a purpose-built, lab-grade instrument with a 12 m multipass cell cannot cleanly resolve ppm-level concentrations without extraordinary averaging times [3]. Budget IR-absorption instrumentation into every soil-gas, downhole-adjacent-surface, and continuous-monitoring campaign as the companion-gas backbone (CH$_4$, C$_2$H$_6$, CO$_2$, CO, H$_2$S); it is mature, sensitive, low-drift, and essential for source attribution and false-positive rejection. Do not specify it, or accept a vendor’s claim of it, as the H$_2$ (or He) sensing channel itself.

This section applies the physics above to field practice.

Direct optical H2 detection has been demonstrated only in narrow, non-field contexts. Laboratory CRDS metrology has resolved H$_2$’s own quadrupole overtone lines at extraordinary sensitivity ($\alpha_{\min}\sim5\times10^{-11}$–$10^{-12}\ \mathrm{cm^{-1}}$) [9,10,11], and one dedicated 2f-WMS TDLAS sensor targeting the S(1) fundamental line has been published and validated over 0–10% v H$_2$ with a 12 m cell [3]. Neither is a field-deployable, cost-competitive, robust instrument for ppm-level geologic H$_2$ exploration: the CRDS work is spectroscopic-physics metrology on neat/high-concentration H$_2$ samples, and the TDLAS sensor’s own reported noise floor sits at several-to-tens of ppm even under benign lab conditions, an order of magnitude above the atmospheric background (0.53 ppm) and comparable to or worse than the low end of reported soil-gas anomalies (tens to low-hundreds of ppm) [42,44,45]. This is corroborated by field practice: the most detailed publicly available comparative soil-gas survey methodology paper found explicitly used an electrochemical sensor for H$_2$ and H$_2$S, and a separate infrared sensor only for CO$_2$, in its autonomous monitoring rig [43] — i.e., practitioners in the field already default to non-optical H$_2$ sensing (see electrochemical cells) and reserve IR specifically for the proxy gas (CO$_2$), exactly the division of labor this page recommends. Similarly, handheld multi-gas meters used in the same study (iBrid Mx6, GA5000) are believed to rely on electrochemical/catalytic/TCD cells for H$_2$, not IR (unverified — [43] confirms both instruments measured H$_2$ and CO$_2$ but does not itemize their internal sensor technology).

Soil-gas surveys. H$_2$ concentrations in point soil-gas surveys over known/candidate seeps have been reported from 0 (background, <1 ppm, essentially atmospheric [51]) up to values exceeding 1,000 ppm, with individual company press releases reporting isolated hits at 2,047–5,558 ppm along fault structures [44,45]; continuous autonomous monitoring at one comparative-methods site recorded baseline emissions around 8–14 ppm with a transient pulse to 598 ppm [43]. This is squarely the regime IR-absorption cannot reach for H$_2$ itself, but comfortably within reach of every optical proxy technique in this page for CH$_4$, CO$_2$, and (where relevant) H$_2$S and C$_2$H$_6$, which are the standard co-interpreted gases in the natural-hydrogen-exploration literature: hydrogen seepage is characteristically accompanied by He (crustal-radiogenic signature), N$_2$, CO$_2$, and CH$_4$, and correct source attribution (deep/geologic vs. shallow biogenic) explicitly requires the associated-gas geochemistry, not H$_2$ concentration alone [36,41,42] — see geologic context for H₂ exploration. Practically: deploy an OA-ICOS or CRDS unit (LGR UGGA, Picarro G-series) or a portable TDLAS (Aeris MIRA) on CH$_4$/CO$_2$/C$_2$H$_6$ alongside whatever non-optical H$_2$/He sensor is doing the primary target-gas measurement; use the C$_2$H$_6$/CH$_4$ ratio and CO$_2$ co-variation to help rule in/out a thermogenic or biogenic (rather than abiotic/geologic) source for any observed methane, keeping the Axetris cross-sensitivity ratio limit (above) in mind when methane is very high relative to ethane.

Flux-chamber deployments for quantifying areal H$_2$/CH$_4$/CO$_2$ flux (not just point concentration) are directly supported by extractive OF-CEAS/OA-ICOS/CRDS instruments, several of which are explicitly marketed for “chamber flux, soil studies” duty [26,31] — again, for the companion gases, with H$_2$ itself measured by whatever non-optical sensor is plumbed into the same chamber headspace loop.

Downhole/borehole logging. This is where these techniques face the hardest constraints. Downhole environments combine high temperature (often >150–250 °C bottom-hole) [50], high pressure (thousands of psi, occasionally tens of thousands) [50], corrosive chemistry including H$_2$S and CO$_2$ [50], and — critically for Hestia’s stated design preference — a strong operational bias against active electronics downhole (reliability, cost of retrieval on failure, intrinsic-safety/explosion-proofing burden). None of the laser/cavity/photoacoustic/FTIR techniques in this page can be run with “no downhole electronics”: every one of them requires an active light source (laser diode, QCL, or IR emitter), a photodetector or acoustic transducer, and real-time signal processing (lock-in demodulation for WMS, ring-down timing for CRDS, interferogram capture for FTIR) co-located with the optical cell. The only genuinely “passive-downhole” optical-adjacent approach identified is fully passive fiber-optic sensing (fiber Bragg gratings for T/P, or distributed Raman/Brillouin scattering for distributed temperature/strain), where “no electronics or power source are required in the well — only passive fiber and a surface interrogator” [46,47] — see fiber-optic sensors. Distributed temperature sensing (DTS) via spontaneous Raman scattering in an optical fiber is noted in the literature as well suited to high-temperature applications for exactly this reason [47], and a related but distinct line of research — stimulated Raman spectroscopy (SRS) in specialty micro/nanofibers — has been proposed specifically for hydrogen leak detection in underground gas storage, reporting noise-equivalent detection limits around 100–200 ppm and demonstrated dynamic range from a few ppm to 100% in one nanofiber-enhanced laboratory implementation [48,49]. This is a genuinely different physical mechanism from the IR absorption covered in the rest of this page (it exploits the H–H stretch’s large, well-characterized Raman cross-section — Raman scattering has entirely different selection rules than IR absorption, and homonuclear diatomics are strong Raman scatterers precisely because bond stretching does strongly modulate their polarizability even though it can never modulate their dipole moment; see Raman spectroscopy) — it is flagged here because it is the one optical-family technique that could plausibly reach downhole H$_2$ without downhole electronics, but as of this research it remains a laboratory/underground-storage-leak-detection demonstration (ppm-to-percent LOD), not a qualified geologic-hydrogen-exploration logging tool, and its LOD (~100+ ppm) would not resolve the low end of the anomaly range discussed above. All-digital, non-electronic downhole pressure sensing (binary-state ADC-style transducers surviving 69 MPa/250 °C) demonstrates the broader “no downhole electronics” engineering pattern Hestia should look for as this space matures [50], but no equivalent all-optical, no-electronics H$_2$ concentration sensor was identified.

Continuous monitoring stations. Fixed-site OA-ICOS/CRDS/OF-CEAS analyzers (Picarro, ABB/LGR, LI-COR) are the natural choice for continuous, long-term, low-drift companion-gas monitoring at a surface station, given their sub-ppb-to-ppb CH$_4$/CO$_2$ precision, <1 ppb/24 h drift, and demonstrated ability to run for multi-week unattended campaigns in open-path configurations [21,26,31]. They are not weatherproof, ruggedized field instruments out of the box (indoor/enclosure operation is standard practice) and require mains or substantial battery/solar power (tens of watts continuous) — a real logistics constraint for remote continuous stations, though well within the capability envelope of a solar-plus-battery system sized for a soil-flux or exploration monitoring pad.

Vendors & products

Vendor claim to flag explicitly: any product literature using the phrase “NDIR hydrogen sensor” or “infrared H2 sensor” without specifying TDLAS/CRDS/quadrupole-line targeting should be treated as almost certainly describing a TCD (thermal-conductivity), catalytic-bead, or electrochemical sensor mislabeled for marketing reasons, or an IR sensor for a different gas bundled in the same housing — see the mechanism discussion above and the Harvey/Perth Basin field study’s own instrument choice (electrochemical for H2/H2S, infrared reserved for CO2) as a real-world confirmation of this pattern [43].

Vendor	Product / line	Technique	Target species	Key specs (as sourced)	Datasheet / source
Picarro	G2201-i Isotopic Analyzer	CRDS (WS-CRDS)	CO2, CH4, δ13C	CO2: 200 ppb+0.05% (1σ,30s); CH4: 1 ppm+0.25% (12C, isotope-only)	[27]
Picarro	G2401	CRDS	CO2, CO, CH4, H2O	Multi-species simultaneous, ppb-class	[28]
Picarro	GasScouter G4302	CRDS, portable/battery	CH4, C2H6, H2O	CH4 3–30 ppb; C2H6 1–10 ppb (mode-dependent); 25 lb, 35 W, 6 h battery	[29]
ABB (ex-Los Gatos Research)	LGR-ICOS GLA132-GGA / UGGA	OA-ICOS	CH4, CO2, H2O	CH4 0.6–2 ppb; CO2 100–300 ppb; 15 kg, 60–66 W; extended range to 10% CH4	[30,31]
Aeris Technologies	MIRA Pico family	Direct-absorption TDLAS, mid-IR laser	CH4, C2H6, CO2, CO, N2O, CH2O, OCS, C2H2	<3 kg, 18 W steady-state, 60 cc cell, 1–5 Hz, “sub-ppb precision” (vendor claim, not independently itemized per gas)	[32]
LI-COR Biosciences	LI-7810	OF-CEAS	CH4, CO2, H2O	CH4 0.25–0.6 ppb; CO2 1.5–3.5 ppm; −25 to 45 °C; 0–85% RH; <1 ppb/24h drift	[26]
LI-COR Biosciences	LI-7815	OF-CEAS	CO2, H2O	High-precision CO2 variant of the LI-7810 platform	Product page (unverified — search-verified, not independently datasheet-fetched)
Gasmet Technologies	DX4000	FTIR	Up to 50 gases (incl. CO2, CH4, CO, H2S, NH3, VOCs)	8/4 cm⁻¹ resolution, 900–4200 cm⁻¹ (2.4–11.1 µm), 5 m (or 2.5/9.8 m optional) multipass cell heated to 180 °C, T90 <120 s, 150 W avg/300 W max, 13.9 kg	[33]
Axetris (Leister Group)	LGD Compact-A (CH4; CH4/CO2; CH4/C2H6; NH3/H2O; C2H2/H2O variants)	TDLS (tunable diode laser spectrometry)	CH4, CO2, C2H6, NH3, H2O, C2H2	CH4 LDL ≤0.4 ppm (≤0.15 with 10 s avg); T90 ≤1.8 s; −10 to 55 °C depending on variant; 3.7–30 VDC	[34]
mirSense	multiSense	QCL + photoacoustic cell	Hydrocarbons incl. CH4, NH3, NO, NO2, SF6, CF4	”Matchbox”-sized, sub-ppm LOD (vendor claim, unverified)	[35,36]
mirSense	MIRCHIP X	QCL-based OEM multigas module	Multi-gas (customer-specified)	Compact OEM analyser module	[35]
nanoplus	DFB laser diodes, 2200–2600 nm family	Component (laser source, not a complete analyzer)	Enabling component for CH4, CO, N2O, HF, and (in the specific 2121.8 nm case used by [3]) H2 quadrupole-line sensors	Custom wavelength to 0.1 nm precision across 760 nm–14 µm	[37]; also directly used as the source in the dedicated H2-TDLAS prototype [3]
— (illustrative, non-Hestia-specific)	ATEX-certified 3-channel QCL prototype	QCL-WMS	H2S, CH3SH, COS (petrochemical process streams)	H2S ≈0.3 ppmv; CH3SH ≈60 ppbv; COS ≈5 ppbv	[38]

Open questions

1. Verify the exact CH4 line-intensity figure independently against a primary HITRAN export. The 2.173×10⁻¹⁹ cm/molecule figure used here for comparison [6] comes from a secondary source reporting values pulled through the hitran.ioa.ru web interface, not a directly-fetched HITRAN line file. It is consistent with general knowledge of CH4 ν3-band intensities but should be re-derived from a raw HITRAN .par extract before being treated as load-bearing for any formal spec sheet.
2. Doppler/Voigt correction to the pathlength estimate above. The worked calculation uses a pure-Lorentzian approximation with only the pressure-broadened width; given that H2’s Doppler width at 1 atm is comparable to or larger than its pressure width, a proper Voigt-profile recalculation would sharpen (and likely modestly worsen) the quoted pathlength/sensitivity numbers. Worth a dedicated half-day exercise if this number needs to appear in a customer-facing document.
3. No commercial or demonstrated all-optical, no-downhole-electronics H2 concentration sensor was found. The closest analog (stimulated-Raman micro/nanofiber H2 leak detection [48,49]) has LOD in the 100+ ppm class and targets underground gas storage leak detection, not geologic-H2 exploration logging. This is a genuine capability gap; worth tracking as a technology-scouting item, and worth a direct inquiry to groups active in this space about roadmap toward lower LOD and field packaging.
4. Independent verification of Aeris MIRA Pico per-gas LOD numbers. The vendor page states “sub-ppb precision” as a general capability claim but does not itemize LOD by target gas model; before specifying a MIRA unit for a campaign, request the gas-specific datasheet/spec sheet from Aeris directly.
5. Cost figures above are qualitative/approximate, not vendor-quoted. No vendor in this space publishes list pricing; before budget planning, get direct quotes for the 2–3 shortlisted instruments per campaign rather than relying on the order-of-magnitude cost classes used here.
6. Cross-check the Axetris CH4/C2H6 cross-interference ratio table [34] against Hestia’s expected field CH4:C2H6 ratios for candidate seep sites before relying on that module for thermogenic/biogenic source discrimination — if background CH4 is very high (as it can be near active seeps), the ethane channel may silently fail per the documented 1:200–1:10 detectable-ratio window.
7. Follow up on whether any vendor’s OA-ICOS/CRDS platform has a validated, field-published H2S channel (most product literature emphasizes CH4/CO2/C2H6; H2S is more commonly addressed via QCL-photoacoustic or QCL-WMS niche/prototype instruments [38,40] rather than the mainstream CRDS/OA-ICOS product lines surveyed here) — relevant if Hestia campaigns are in H2S-co-occurring (sour) settings.

Related pages

• Why H₂ is hard to sense — the compendium-wide framing of the same symmetry problem
• Raman spectroscopy — the sister optical technique that does have a direct physical handle on H₂, via polarizability rather than dipole moment
• Enhanced Raman — cavity/fiber/plasmonic tricks pushing Raman’s H₂ sensitivity toward ppb
• UV/EUV and fluorescence — the other native optical handle on H₂, blocked by atmospheric absorption
• Fiber-optic sensors — the leading passive, no-downhole-electronics alternative for the same wellbore niche IR absorption cannot fill
• Thermal conductivity & acoustic sensors — the real physics behind most “IR/NDIR hydrogen sensors” actually sold
• Catalytic pellistors — another non-optical technique frequently confused with or bundled alongside IR modules
• Electrochemical cells — the field-proven non-optical workhorse for the H₂ channel itself
• Geologic context for H₂ exploration — why companion-gas geochemistry (He, N₂, CO₂, CH₄) is required for source attribution

References

Full reference list (0 entries)

[1] Somogyi, W., Yurchenko, S.N., Yachmenev, A. “Calculation of electric quadrupole linestrengths for diatomic molecules: Application to the H2, CO, HF and O2 molecules.” J. Chem. Phys. (2021); arXiv:2111.05840. DOI: 10.1063/5.0063256. https://arxiv.org/pdf/2111.05840

[2] Herzberg, G. “Quadrupole Rotation-Vibration Spectrum of the Hydrogen Molecule.” Nature 163, 170 (1949). https://www.nature.com/articles/163170a0

[3] Avetisov, V., Bjørøy, Ø., Wang, J., Geiser, P., Paulsen, K.G. “Hydrogen Sensor Based on Tunable Diode Laser Absorption Spectroscopy.” Sensors 19(23), 5313 (2019). DOI: 10.3390/s19235313. https://pmc.ncbi.nlm.nih.gov/articles/PMC6928617/

[4] Gordon, I.E., Rothman, L.S., Hargreaves, R.J., et al. “The HITRAN2020 molecular spectroscopic database.” J. Quant. Spectrosc. Radiat. Transfer 277, 107949 (2022). DOI: 10.1016/j.jqsrt.2021.107949. https://www.sciencedirect.com/science/article/pii/S0022407321004416

[5] HITRAN isotopologue metadata (H2, CH4 abundances / partition functions). https://hitran.org/docs/iso-meta/

[6] “Methane infrared absorption” (HITRAN-derived CH4 ν3-band line intensity, via hitran.ioa.ru). https://climateauditor.com/methane-infrared-absorption/

[7] “Rototranslational collision-induced absorption by H2-H2 pairs at temperatures from 600 to 7000 K.” NASA Technical Reports Server. https://ntrs.nasa.gov/citations/19950049089

[8] “New experimental measurements of the Collision Induced Absorptions of H2-H2 and H2-He in the 3600–5500 cm−1 spectral range from 120 to 500 K.” (2024). https://www.sciencedirect.com/science/article/pii/S0022407324003625

[9] Kassi, S., Campargue, A. “The absorption spectrum of H2: CRDS measurements of the (2-0) band, review of the literature data and accurate ab initio line list up to 35 000 cm−1.” Phys. Chem. Chem. Phys. (2012). https://pubs.rsc.org/en/content/articlelanding/2012/cp/c1cp22912e (PDF: https://www.fuw.edu.pl/~krp/papers/camparge.pdf)

[10] “Electric quadrupole transitions and collision-induced absorption in the region of the first overtone band of H2 near 1.25 μm.” J. Mol. Spectrosc. https://www.sciencedirect.com/science/article/abs/pii/S0022285214000836

[11] “Cavity ring-down spectroscopy of the electric quadrupole transitions of H2 in the 784–852 nm region.” J. Mol. Spectrosc. https://www.sciencedirect.com/science/article/abs/pii/S0022285214000629

[12] NevadaNano. “NDIR Sensor: Are There Disadvantages to Consider?” https://nevadanano.com/ndir-sensor-are-there-disadvantages-to-consider/

[13] Horiba. “Thermal Conductivity Detector Method (TCD) for Measuring Hydrogen.” https://www.horiba.com/int/process-and-environmental/measuring-principles/thermal-conductivity-detector/

[14] H2scan. “Hydrogen Detection Systems for Safer Processing.” https://h2scan.com/hydrogen-detection-systems-cross-sensitivity-could-shut-down-your-process/

[15] MTL / Eaton. “K522 Thermal Conductivity (TCD) Gas Sensor for Hydrogen (H2), Helium (He), Nitrogen and Argon.” https://www.eaton.com/us/en-us/catalog/intrinsic-safety/mtl-k522-thermal-conductivity-tcd-gas-sensor-for-hydrogen-helium.html (link updated — original mtl-inst.com URL now redirects to a generic Eaton landing page, not the specific product)

[16] Hodgkinson, J., Tatam, R.P. “Optical gas sensing: a review.” Meas. Sci. Technol. 24, 012004 (2013). DOI: 10.1088/0957-0233/24/1/012004.

[17] Wikipedia. “Cavity ring-down spectroscopy.” https://en.wikipedia.org/wiki/Cavity_ring-down_spectroscopy

[18] Baer, D.S., Paul, J.B., Gupta, M., O’Keefe, A. “Sensitive absorption measurements in the near-infrared region using off-axis integrated-cavity-output spectroscopy.” Appl. Phys. B 75, 261–265 (2002). https://ui.adsabs.harvard.edu/abs/2002ApPhB..75..261B/abstract

[19] Patimisco, P., Scamarcio, G., Tittel, F.K., Spagnolo, V. “Quartz-Enhanced Photoacoustic Spectroscopy: A Review.” Sensors 14, 6165–6206 (2014). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4029643/

[20] Kosterev, A.A., Tittel, F.K., et al. “Applications of quartz tuning forks in spectroscopic gas sensing.” Rev. Sci. Instrum. 76, 043105 (2005). https://ui.adsabs.harvard.edu/abs/2005RScI…76d3105K/abstract

[21] Coddington, I., Newbury, N., Swann, W. “Dual-comb spectroscopy.” Optica 3, 414–426 (2016). https://www.nist.gov/publications/dual-comb-spectroscopy; https://pmc.ncbi.nlm.nih.gov/articles/PMC8201420/

[22] Giorgetta, F., et al. “Open-Path Dual-Comb Spectroscopy for Multispecies Trace Gas Detection in the 4.5–5 µm Spectral Region.” Laser Photonics Rev. 15, 2000583 (2021). https://onlinelibrary.wiley.com/doi/full/10.1002/lpor.202000583

[23] “Precise multispecies agricultural gas flux determined using broadband open-path dual-comb spectroscopy.” Sci. Adv. https://pmc.ncbi.nlm.nih.gov/articles/PMC8011971/

[24] Reid, J., Labrie, D. “Second-harmonic detection with tunable diode lasers — Comparison of experiment and theory.” Appl. Phys. B 26, 203–210 (1981). https://link.springer.com/article/10.1007/BF00692448

[25] “Improvement of the Detection Sensitivity for Tunable Diode Laser Absorption Spectroscopy: A Review.” Front. Phys. 10, 853966 (2022). https://www.frontiersin.org/articles/10.3389/fphy.2022.853966/full

[26] LI-COR Biosciences. “LI-7810 CH4/CO2/H2O Trace Gas Analyzer Specifications.” Doc 982-17412 (03/2023). https://www.ekotechnika.cz/sites/default/files/pdf/7810-Spec-Sheet-17412.pdf

[27] Picarro. “G2201-i Analyzer Datasheet.” https://www.picarro.com/environmental/g2201i_analyzer_datasheet

[28] Picarro. “G2401 CO2, CO, CH4 and H2O Gas Concentration Analyzer Datasheet.” https://www.picarro.com/sites/default/files/product_documents/Picarro_G2401%20Datasheet_211029.pdf

[29] Picarro. “GasScouter G4302 CH4, C2H6 and H2O Analyzer Datasheet.” https://www.picarro.com/sites/default/files/product_documents/DataSheet_G4302_Picarro_GasScouter_CH4_C2H6_June17.pdf

[30] ABB. “OA-ICOS Laser Gas Analyzers — LGR-ICOS Technology.” https://new.abb.com/products/measurement-products/analytical/laser-gas-analyzers/lgr-icos-technology

[31] Los Gatos Research / ABB. “Ultraportable Greenhouse Gas Analyzer (CH4, CO2, H2O) Datasheet.” http://www.lgrinc.ru/docs/analyzers/gas/GHG/LGR_Portable_GGA_Datasheet.pdf

[32] Aeris Technologies. “MIRA Pico.” https://aerissensors.com/products/mira-pico/

[33] Gasmet Technologies Oy. “DX4000 FTIR Gas Analyzer Technical Data v1.6.” https://www.anatrac.com/wp-content/uploads/2019/07/Gasmet_DX4000_Technical_Data_v1.6.pdf

[34] Axetris (Leister Group). “OEM Gas Sensing Module LGD Compact-A Series — Product Datasheet Rev J.” https://cdn-assets.leister.com/pim/medias/17/47/66//174766.pdf

[35] mirSense. “Compact OEM gas analyser modules based on quantum cascade laser technology.” Envirotech Online. https://www.envirotech-online.com/news/environmental-laboratory/7/mirsense/compact-oem-gas-analyser-modules-based-on-quantum-cascade-laser-technology/47984

[36] mirSense. “Industry & Environment.” https://mirsense.com/industry-environment/

[37] nanoplus Nanosystems and Technologies GmbH. “Distributed Feedback Lasers 2200 nm – 2600 nm.” https://nanoplus.com/products/distributed-feedback-laser/2200nm-2600nm

[38] “ATEX-Certified, FPGA-Based Three-Channel Quantum Cascade Laser Sensor for Sulfur Species Detection in Petrochemical Process Streams.” Sensors 25(3), 635 (2025). https://www.mdpi.com/1424-8220/25/3/635

[39] “Rapid ppb-Level Methane Detection Based on Quartz-Enhanced Photoacoustic Spectroscopy.” Anal. Chem. (2025). https://pubs.acs.org/doi/10.1021/acs.analchem.5c00154

[40] “N2-cooled THz quartz-enhanced photoacoustic sensor operating in pulsed mode for hydrogen sulfide detection in the part-per-billion concentration range.” https://www.researchgate.net/publication/338957406

[41] Zgonnik, V. “The occurrence and geoscience of natural hydrogen: A comprehensive review.” Earth-Sci. Rev. 203, 103140 (2020). DOI: 10.1016/j.earscirev.2020.103140.

[42] “Exploring natural hydrogen hotspots: a review and soil-gas survey design for identifying seepage.” Geoenergy, Lyell Collection (2023). DOI: 10.1144/geoenergy2023-014.

[43] Davies, K., Haines, P.W., Thomas, C., Normore, L. “Natural hydrogen soil gas emissions near Harvey, Perth Basin: a comparative study of survey methods.” Sci. Technol. Energy Transition 80, 46 (2025). DOI: 10.2516/stet/2025026. https://www.stet-review.org/articles/stet/full_html/2025/01/stet20250031/stet20250031.html (does not itemize the iBrid Mx6/GA5000 internal sensor technology)

[44] “Primary Hydrogen finds 665 ppm and 531 ppm H2 in NL soil gas survey.” StockTitan. https://www.stocktitan.net/news/HNATF/primary-hydrogen-reports-anomalous-hydrogen-values-from-phase-1-soil-gvcxp9are6v1.html

[45] “QIMC Announces Major Natural Hydrogen Discovery in Nova Scotia… Record 5,558 ppm.” Resource World Magazine. https://resourceworld.com/qimc-announces-major-natural-hydrogen-discovery-in-nova-scotia-with-multiple-high-grade-soil-gas-samples-including-record-5558-ppm/

[46] Fibercore (Humanetics Group). “Fiber Optic Sensing for Downhole Monitoring in Oil & Gas.” https://fibercore.humaneticsgroup.com/perspectives/fiber-optic-sensing-downhole-monitoring-oil-gas

[47] “Temperature dependent characterization of optical fibres for distributed temperature sensing in hot geothermal wells.” arXiv:1206.1853. https://arxiv.org/pdf/1206.1853

[48] “A safe and high-precision detection method for hydrogen leakage analysis of underground gas storage based on stimulated Raman spectroscopy of micro-nanofiber.” Fuel (2025). https://www.sciencedirect.com/science/article/abs/pii/S0016236125014681

[49] “Nanofiber enhanced stimulated Raman spectroscopy for ultra-fast, ultra-sensitive hydrogen detection with ultra-wide dynamic range.” Optica 6(5), 570 (2019). https://opg.optica.org/optica/fulltext.cfm?uri=optica-6-5-570&id=409432

[50] “All-digital Sensor System for Distributed Downhole Pressure Monitoring in Unconventional Fields.” OSTI Technical Report. https://www.osti.gov/biblio/2370516

[51] “Molecular hydrogen in atmosphere” (tropospheric H2 background ~0.53 ppmv). https://www.researchgate.net/publication/229528201_Molecular_hydrogen_in_atmosphere; Weybourne Atmospheric Observatory H2 record: https://www.tandfonline.com/doi/full/10.3402/tellusb.v64i0.17771

[52] Evikon MCI. “%LEL, %LFL, % vol, ppm – different units in gas detection” (H2 LEL ≈4% confirmed on this page; UEL ≈75–77% does not appear on this page). https://www.evikon.eu/news/lel-lfl-vol-ppm-different-units-in-gas-detection-a-15/