How the sensor families actually differ: a transduction decoder

Summary

Every hydrogen sensor does two things in sequence: a receptor event (something physically happens when H2 meets the sensing material: a reaction, a metal swelling, a gas cooling a hot wire) and a transducer event (that change becomes a loggable signal: voltage, current, resistance, or a property of light). Sensors are usually named after their transducer, which is why unrelated devices sound alike. A metal-oxide chemiresistor, a palladium diode, and a fuel-cell-like electrochemical cell can all report ohms or amps while measuring three physically unrelated quantities.

This page gives the vocabulary and the arithmetic to tell the roughly eight H2-sensor families apart: what physical quantity changes in each, how it becomes a number, and which family is properly called “electrochemical” (exactly one). With it, you can usually identify the mechanism inside an unlabeled sensor from how its datasheet reports numbers.

The two-link chain: receptor and transducer

The receptor event is the actual interaction between H2 and the sensing material: a chemical reaction, hydrogen atoms dissolving into a metal lattice, or nothing chemical at all (a gas molecule carrying heat away from a hot wire). The transducer event converts that interaction into a measurable signal.

The distinction matters because the two vary independently. Two sensors can share a receptor event (hydrogen dissociating on palladium) and use different transducers, so one reports a linear voltage shift and the other an exponential current jump from the same physical trigger. Two sensors can share a transducer circuit (a Wheatstone bridge) while their receptor events are unrelated: one is a thermometer, one a plain resistor.

Table 1: the eight-way map

Family	What physically changes (the receptor)	How it’s read out (the transducer)
MOX chemiresistor	Free-electron population in a heated oxide film	Resistance in a voltage divider
Work-function / field-effect	An electric dipole layer at a buried interface	A transistor’s threshold voltage or a diode’s current
Pd resistive	The bulk resistivity, or physical geometry, of palladium metal itself	Two-terminal resistance
Catalytic pellistor	Heat released by burning H2 on a hot catalyst bead	A platinum coil’s resistance, in a bridge
Electrochemical cell	Moles of H2 chemically oxidized at an electrode	Current or voltage, tied exactly to Faraday’s law
Thermal conductivity	How fast the gas mixture carries heat away (no reaction)	A heated element’s resistance, in a bridge
Optical (Raman, IR-proxy, plasmonic, fiber)	A molecule’s polarizability, or a metal film’s optical properties	Wavelength, intensity, or phase of light

A sense of scale: what “1 ppm” is physically

“1 ppm” means one H2 molecule per million gas molecules, and a million-th of a gas is still a large absolute number. At 25 °C (298.15 K) and 1 atm (101,325 Pa), the ideal gas law gives the molar volume

$V_m = \frac{RT}{P} = \frac{8.314462618 \times 298.15}{101{,}325} = 0.024465\ \mathrm{m^3/mol} = 24.465\ \mathrm{L/mol}$

($R$ is the molar gas constant, exact under the modern SI, $8.314462618\ \mathrm{J\,mol^{-1}\,K^{-1}}$ [1]). A 1 ppm H2 mole fraction is then a number density of

$\frac{n_{H_2}}{V} = \frac{xP}{RT} = \frac{10^{-6}\times101{,}325}{8.314462618\times298.15} = 4.088\times10^{-5}\ \mathrm{mol/m^3} = 2.46\times10^{13}\ \mathrm{molecules/cm^3} .$

The scale of 1 ppm

A 1 ppm reading is about 25 trillion H2 molecules per cubic centimeter. Every detection-limit claim in this wiki (“0.53 ppm atmospheric background,” “40 ppb floating-gate transistor”) is a claim about noticing a difference of order $10^{13}$–$10^{16}$ molecules/cm³ on top of roughly $2.46\times10^{19}$ molecules/cm³ of air. That scale is a fast gut-check on implausible vendor claims. The full unit toolkit (ppmv, mg/m³, mol/L) is in Concentration units.

What “electrochemical” actually means

Vendors call almost any small electronic H2 sensor “electrochemical.” Under the precise definition, only one family on this page qualifies.

The IUPAC Gold Book defines electrochemical measurement around a faradaic process: a reaction whose rate is tied by an exact law to an electrical current [2][3]. That law is

$Q = nFN$

where $Q$ is the charge that flows, $n$ the electrons per reacting molecule, $F$ the Faraday constant ($96{,}485.332\ldots$ C per mole of electrons) [4], and $N$ the moles reacted. Three ingredients are required for the name:

• an electrolyte (ion-conducting material) physically separating two electrodes,
• a genuine redox reaction at an electrode surface,
• an external current or voltage set quantitatively by that reaction through $Q=nFN$.

Why MOX, Pd-resistive, and work-function devices don’t qualify. A MOX sensor has one solid, no electrolyte, no counter electrode. Its reaction, $\mathrm{H_2 + O^-(ads) \rightarrow H_2O(g) + e^-}$, returns an electron to the same oxide’s conduction band; nothing leaves the device, and the measured signal is ordinary resistance. The correct class name is chemiresistive [5][6]. A Pd-resistive sensor is the same story with different chemistry: dissolved hydrogen changes the metal’s own resistivity. A work-function device reports an externally biased current modulated by a static interfacial dipole; doubling the H2 concentration does not double the current in any stoichiometric way.

Only the amperometric or potentiometric cell (the fuel-cell-like device in most soil-gas handhelds) satisfies all three conditions: H2 is oxidized at the working electrode ($\mathrm{H_2 \rightarrow 2H^+ + 2e^-}$), electrons cross an external wire while H⁺ crosses the electrolyte, and the current is exactly $nF$ times the reaction rate.

Worked example: counting electrons. A commercial amperometric cell (Membrapor H2/CA-1000 family) is specified near 28 nA per ppm H2 [7]. At 1 ppm, with the elementary charge $e = 1.602176634\times10^{-19}$ C [8]:

$\text{electrons/s} = \frac{28\times10^{-9}}{1.602176634\times10^{-19}} = 1.748\times10^{11}, \qquad \text{H}_2/\mathrm{s} = \frac{1.748\times10^{11}}{2} = 8.74\times10^{10},$

and dividing by Avogadro’s number $N_A = 6.02214076\times10^{23}$/mol [9] gives $1.45\times10^{-13}$ mol/s: 145 femtomoles of H2 oxidized per second, read directly as current. No other family has an equivalent first-principles bridge between chemical amount and electrical reading, which is why amperometric cells are the most metrologically well-behaved and genuinely linear family here.

The six transduction classes. Five of the six are non-faradaic; electrochemical is the exception, not the umbrella.

Class	What physically happens	Faradaic ($Q=nFN$)?	Ion crosses an electrolyte?	Families
Electrochemical	Redox reaction at an electrode/electrolyte interface	Yes	Yes	Amperometric & potentiometric cells
Chemiresistive	Surface or bulk chemistry changes a single solid’s own conductivity	No	No	MOX; Pd resistive (bulk & nanogap)
Field-effect / work-function	An interfacial dipole shifts an electrostatic potential	No	No	MIS/MOSFET, MIM, Schottky diode
Calorimetric (reactive)	Combustion heat raises a thermometer’s temperature	No	No	Catalytic pellistor
Thermometric (non-reactive)	A transport property of the unreacted gas mixture	No	No	Thermal conductivity
Optical	Polarizability, refractive index, or strain changes a light property	No	No	Raman, IR-proxy, plasmonic, fiber

A seventh class, gravimetric (mass-loading) devices (QCM and SAW resonators, $\Delta f = -C_f\,\Delta m$ [10]), exists in the wider literature; H2 is too light for mass loading to be a good primary route, and SAW, QCM & MEMS covers where it is used anyway.

Family by family: the physics and the arithmetic

MOX chemiresistors. A heated (roughly 200–450 °C) polycrystalline oxide film, historically SnO2 [5]. In air, adsorbed oxygen takes electrons from the oxide, raising resistance near grain boundaries. H2 removes that oxygen and returns the electron:

$\mathrm{H_2 + O^-(ads) \rightarrow H_2O(g) + e^-} ,$

lowering resistance. Diffusion into the porous film makes the response an empirical power law, $G \propto P_{\mathrm{H_2}}^{\,n}$ with $n \approx 0.5\text{--}1.2$ typical [5][6][11]. Worked example: a Figaro TGS821-class device with $n=0.8$ [11] and $R_s(100\,\mathrm{ppm}) = 10\,\mathrm{k}\Omega$ gives $R_s(1000\,\mathrm{ppm}) = 10\,\mathrm{k}\Omega \times 10^{-0.8} = 1.585\,\mathrm{k}\Omega$: a tenfold concentration rise drops resistance to ~16%, a strong but curved and completely non-selective signal (CO, H2S, and alcohols consume the same adsorbed oxygen). Full envelope: MOX chemiresistors.

Work-function / field-effect devices. H2 dissociates on a catalytic Pd/Pt surface and the atoms form an oriented dipole layer at a buried interface [12][13], shifting the work function by $\Delta\Phi = q N_H p/\varepsilon_0$. Reported shifts reach roughly 0.5 eV against Pd’s clean-surface work function of about 5.2 eV [14]. Three architectures read the same step:

• MISFET/MOS capacitor: threshold shift $\Delta V_T = \Delta\Phi/q$, so 0.5 eV gives 0.5 V; a directly measured device showed $\Delta V_{FB}\approx0.17$ V at 91 ppm H2 [15].
• Schottky diode, usually reverse-biased: the dipole modulates barrier height $\Phi_B$, and $J_{rev} \approx -A^{*}T^{2}\exp(-q\Phi_B/kT)$ is exponential. At 300 K, $kT/q = 25.85$ meV [16][8], so the reverse-current multiplier $\exp(\Delta\Phi_B/(kT/q))$ is:

Barrier-height drop $\Delta\Phi_B$	Reverse-current multiplier
50 meV	$\exp(1.93) \approx 6.9\times$
100 meV	$\exp(3.87) \approx 48\times$
250 meV	$\exp(9.67) \approx 1.6\times10^{4}$
500 meV (upper end reported)	$\exp(19.34) \approx 2.5\times10^{8}$

That exponential is the arithmetic behind reported signals like a Pd-nanoparticle Schottky diode’s ~$7\times10^{5}$ current change at 0.1% H2 [18]. It is also why reverse bias is the deliberate operating point: forward bias buries the barrier signal under injection current, while reverse bias isolates the exponential term against a small baseline [17]. MIM tunneling devices and the rest of the family: Work-function devices; the same catalytic-metal-plus-dipole physics underlies 2D-material FET sensors.

Palladium resistive. Same metal and dissociation step as above, different measured quantity: hydrogen dissolves into the bulk. At low loading Sieverts’ law holds, $x = K_S\sqrt{p_{\mathrm{H_2}}}$ (the dilute α phase); above $x\approx0.02$ at room temperature Pd converts to the lattice-expanded β hydride, at only about 20–30 mbar of H2 partial pressure [19]. The β lattice (4.025 Å) is 3.5% larger than α (3.889 Å) [20], and resistivity rises: bulk PdH$_x$ reaches $R/R_0 \approx 1.87$ at $x\approx0.76$ [21] (so a 100 Ω film saturates near 187 Ω). Nanostructuring flips the sign: in a discontinuous nanowire or percolation network, the ~11% volumetric swelling closes gaps between conductors and resistance drops, often by orders of magnitude, first demonstrated with electrodeposited Pd mesowires responding in under 75 ms over 2–10 vol% H2 [22]. Both signs are legitimate Pd-resistive behavior; the sign depends on film continuity. Because the law is $\sqrt{p}$, doubling total pressure at fixed volume fraction shifts the signal by $\sqrt{2}\approx1.41\times$: any elevated-pressure deployment needs an explicit correction. Full treatment: Palladium resistive sensors.

Catalytic pellistors. Two matched beads in a Wheatstone bridge, one catalyst-loaded, one deactivated, self-heated to roughly 400–550 °C. H2 burns flamelessly on the active bead,

$\mathrm{H_2(g) + \tfrac12 O_2(g) \rightarrow H_2O(g)}, \qquad \Delta_f H^\circ = -241.83\ \mathrm{kJ/mol}\ [23],$

raising its temperature and hence its platinum coil’s resistance; the bridge reports $V_{out} \approx (V_s/4)\cdot(\Delta R_{det}/R)$. The changing quantity is bead temperature: this is a thermometer. Worked example: a Honeywell CAT25-class pellistor at ~25 mV/vol% methane with H2 relative response 107% [24] gives 26.75 mV/vol% H2, so a 100 ppm (0.01 vol%) anomaly produces 267.5 µV, a very small bridge imbalance. That is the quantitative reason pellistors bottom out around 1 %LEL (~400 ppm H2). They also need ≥10–12 vol% ambient O2 and respond to every combustible gas, differentiated only by tabulated k-factors, which vary between cartridge families (107 vs. 43 vs. 83 relative to methane = 100), so the wrong factor can be off by more than 2× [24]. Full detail: Catalytic pellistors.

Electrochemical cells. The one faradaic family: H2 oxidized at the working electrode, O2 reduced at the counter electrode, ions crossing the electrolyte [25]. A potentiostat makes every arriving H2 molecule react, so the current is diffusion-limited,

$i_L = \frac{nFAD\,C_\infty}{l} ,$

with every factor except the ambient concentration $C_\infty$ a device constant. The response is genuinely linear, which is why vendors can quote one stable nA/ppm number. The thermodynamic driver is the 1.229 V gap between the hydrogen half-reaction ($E^\circ \equiv 0$ V, the standard hydrogen electrode) and oxygen reduction ($E^\circ = +1.229$ V) [26], the same gap that powers a fuel cell. Potentiometric variants read a zero-current Nernst potential instead, about 29.6 mV per decade of H2 partial pressure at room temperature, a coarser relationship. Full envelope: Electrochemical cells; the high-temperature solid-oxide cousin: Solid-electrolyte / high-temperature sensors.

Thermal conductivity. Pellistor hardware, no reaction. For a dilute gas the Eucken relation $\lambda = (c_v + \tfrac{5}{4}R/M)\eta$ applies, and H2’s molar mass (2.016 g/mol, ~14× below air’s) gives $\lambda_{H_2}\approx0.187\ \mathrm{W\,m^{-1}K^{-1}}$, about $0.187/0.026 \approx 7.19\times$ air, the largest contrast of any common soil gas. A heated element loses heat at a rate set by λ; matched elements in a bridge report the difference. H2 is never consumed. The cost is selectivity: one degree of freedom cannot distinguish H2 from any other high-conductivity gas, and helium ($\lambda_{He}\approx0.15$) is the look-alike. For 1% He contamination the equivalent-H2 bias is roughly

$\Delta_{H_2,eq} \approx 10^{-2}\times\frac{0.15-0.026}{0.187-0.026}\times10^6 \approx +7{,}700\ \mathrm{ppm}$

Helium look-alike

1% helium contamination, plausible in exactly the deep radiogenic settings H2 exploration targets, can fake a multi-thousand-ppm hydrogen reading on an uncompensated thermal-conductivity bridge. Co-track helium wherever TC sensing is deployed: Helium & co-tracers, Thermal conductivity & acoustic sensors.

Optical families. The signal carrier is a photon. H2’s symmetry gives it zero permanent dipole, so ordinary infrared absorption (the mechanism behind cheap CO2 and CH4 sensors) is blind to it (Why H2 is hard to sense). Two workarounds:

• Raman scattering needs polarizability, not a dipole: the induced moment $\mu_{ind}=\alpha E$ changes as the bond vibrates, and H2’s strongly anisotropic bond polarizability makes it a good Raman scatterer despite being IR-dark. See Raman spectroscopy and Enhanced Raman.
• The electric-quadrupole absorption pathway is symmetry-allowed but roughly six to nine orders of magnitude weaker than dipole absorption; the strongest H2 line has HITRAN intensity near $3\times10^{-26}$ cm/molecule [27] vs. $10^{-21}$–$10^{-19}$ for strong IR absorbers. Commercial IR instruments therefore never target H2 directly and instead sense co-evolved proxies (CO2, CH4): IR absorption and proxies.

A second optical channel uses palladium’s hydride chemistry rather than H2 spectroscopy: plasmonic sensors track how PdH$_x$ shifts a nanostructure’s resonance (Plasmonic metamaterials), and fiber sensors track how the lattice expansion strains a grating or shifts an evanescent path (Fiber-optic sensors). Neither moves an electron through a circuit; the readout is wavelength, phase, or intensity. Emerging optical approaches: UV/EUV & fluorescence, Exotic optical methods.

Reading an unlabeled sensor: a five-question test

Asked in order against a vague datasheet:

1. Liquid/gel/membrane electrolyte between two or three electrodes, sensitivity in nA/ppm or µA/ppm? → Electrochemical cell, the only family whose sensitivity constant comes from Faraday’s law.
2. Continuous heater power to 200–450 °C, a bare oxide film or bead, an explicit O2 dependency? → MOX chemiresistor.
3. Two beads on wire coils in a Wheatstone bridge, one catalyst-loaded, output in mV per vol% methane? → Catalytic pellistor.
4. A bare, unheated (or lightly heated) resistor, wire, or nanowire of Pd or Pd alloy, no catalyst-on-oxide structure, no bridge? → Pd resistive; check whether resistance rises (bulk film) or falls (nanogap/mesowire).
5. A transistor, capacitor, or diode with a catalytic-metal gate, with “threshold voltage,” “flat-band voltage,” or “barrier height” in the datasheet, or explicitly run in reverse bias? → Work-function/field-effect. Most often confused with #4 because both use palladium; the tell is a genuine junction rather than a two-terminal resistor.

If none fits and the device involves a laser, fiber, or spectrometer, it is optical: Raman (laser + spectrometer, targets H2 directly), IR/NDIR (targets a proxy gas), or fiber/plasmonic (a coated tip or chip with no gas-phase optical path). A heated wire or MEMS membrane in a bridge marketed as poison-immune and O2-free is thermal conductivity.

Common mistakes

Watch for these

1. Calling every small electronic H2 sensor “electrochemical.” Only amperometric/potentiometric cells with a real electrolyte qualify.
2. Treating “MOS sensor” as unambiguous: it can mean a MOX chemiresistor bead or the MOS capacitor/transistor structure in the work-function family, where “oxide” is the gate insulator. Check which oxide a document means.
3. Assuming Pd-gate and Pd-resistive devices are the same technology because both say palladium. One measures a surface dipole’s effect on a junction, the other the bulk resistivity of the palladium itself.
4. Converting a %LEL pellistor reading to ppm H2 with a generic factor. Cartridge-specific k-factors differ by more than 2×; use the instrument’s own calibration.
5. Assuming every H2 sensor needs oxygen. MOX, pellistors, and standard amperometric cells do; work-function, Pd resistive, thermal-conductivity, and most optical methods do not. An O2-dependent sensor in anoxic soil reads falsely low or zero with no fault flag.
6. Trusting a product-line name over the datasheet body. Vendors sell MOX, catalytic, and electrochemical sensors under one marketing category; confirm the principle from the technical specification.
7. Assuming reverse bias means a Schottky device is off or faulty. For work-function diodes it is the deliberately chosen, most sensitive operating point.
8. Comparing LOD figures without checking the statistical definition. The IUPAC 3σ convention and the SNR=1 convention differ by roughly 3× on identical noise [28][29]; see Reading the specs and Figures of merit.
9. Assuming linearity from a straight-looking plot. MOX and TC are power-law or ratio-based, Pd resistive follows a square-root law, and only amperometric cells are linear by construction.
10. Conflating Pd-gate “poisoning” with Pd-resistive “embrittlement.” The first is a catalysis failure (CO/H2S blocking the surface), the second a materials failure (lattice-volume fatigue from α↔β cycling); alloying addresses the second, alternate metals or overlayers the first.

How this shapes sensor choice

Field verdict

The taxonomy is the right organizing principle, but three practical filters decide purchases: (a) does the site ever go anoxic (buried soil, purged downhole, sealed headspace), which eliminates O2-dependent families outright; (b) what poisons this specific catalytic metal, electrolyte, or oxide, with H2S tolerance first for geologic work; (c) under what statistical convention and averaging time the quoted detection limit was measured.

The buyer’s checklist, in priority order:

1. What physical quantity changes, and is it linear, power-law, square-root, or exponential in concentration? This decides whether one calibration point suffices (linear: electrochemical only) or a multi-point curve is mandatory.
2. Does it need O2, and will the deployment ever be anoxic?
3. What poisons the specific formulation, and is that interferent present at the site?
4. What convention and averaging time define the quoted LOD?
5. Is the product name the sensing principle, or a marketing umbrella?

For the exploration-specific version (which family wins for soil-gas, continuous monitoring, or downhole), see Sensor selection and Comparison charts. Gas-phase vs. aqueous-phase H2 sensing covers how these mechanisms adapt to dissolved H2. Gas chromatography and mass spectrometry sit outside the taxonomy as bench-reference techniques, the cross-checks behind mud-gas logging and dissolved-gas groundwater work. Field discipline lives in Soil-gas surveys, Downhole deployment, Calibration & QA/QC, and Environmental covariance; pairing readings with co-tracers, isotopes, and geologic context turns a number into a decision; sensor arrays with machine learning combine families to cross-cancel individual cross-sensitivities.

Cheat sheet

Family	Physical quantity that changes	Readout	Governing relation	Faradaic?	Needs O2?
MOX chemiresistor	Conduction-band electron population	Resistance	$G\propto P_{H_2}^{\,n}$, $n\approx0.5$–1.2	No	Yes
Work-function / field-effect	Interfacial H-dipole → electrostatic potential	Threshold voltage / diode current	$\Delta V_T=\Delta\Phi/q$; $J_{rev}\propto\exp(-q\Phi_B/kT)$	No	No
Pd resistive (bulk)	Pd bulk resistivity (PdH$_x$ formation)	Two-terminal resistance (↑)	Sieverts: $x=K_S\sqrt{p_{H_2}}$	No	No
Pd resistive (nanogap)	Pd geometric swelling closing gaps	Two-terminal resistance (↓)	Same Sieverts physics, geometric readout	No	No
Catalytic pellistor	Bead temperature (combustion heat)	Pt-coil resistance, bridge	$V_{out}\approx(V_s/4)(\Delta R_{det}/R)$	No	Yes (≥10–12 vol%)
Electrochemical cell	Moles H2 oxidized at an electrode	Faradaic current / potential	$i_L = nFAD\,C_\infty/l$	Yes	Yes (counter electrode)
Thermal conductivity	Bulk gas thermal conductivity λ	Heated-element resistance, bridge	Eucken: $\lambda=(c_v+\tfrac54 R/M)\eta$	No	No
Optical (Raman)	Molecular polarizability	Scattered-photon wavelength	$\mu_{ind}=\alpha E$	No	No
Optical (IR-proxy / plasmonic / fiber)	Proxy-gas dipole / Pd optical constants	Absorbed/reflected light	Quadrupole ~10⁶–10⁹× weaker than dipole IR	No	No

Related pages

• Concentration units: the ppmv/mg/m³/mol-L toolkit behind the scale section above
• Gas-phase vs. aqueous-phase H2 sensing: how these mechanisms adapt (or don’t) to dissolved H2
• Figures of merit: reading LOD, response time, drift, and selectivity claims for any family above
• Why H2 is hard to sense: the molecular-symmetry reason optical sensing needs a workaround
• Sensor selection
• Comparison charts
• Reading the specs
• MOX chemiresistors
• Work-function devices
• Palladium resistive sensors
• Thermal conductivity & acoustic sensors
• Catalytic pellistors
• 2D-material FET sensors
• Electrochemical cells
• Solid-electrolyte / high-temperature sensors
• SAW, QCM & MEMS (gravimetric)
• Raman spectroscopy
• Enhanced Raman
• IR absorption and proxies
• UV/EUV & fluorescence
• Fiber-optic sensors
• Exotic optical methods
• Plasmonic metamaterials
• Sensor arrays & machine learning
• Gas chromatography
• Mass spectrometry
• Mud-gas logging
• Dissolved H2 in groundwater
• Geologic context
• Soil-gas surveys
• Downhole deployment
• Calibration & QA/QC
• Environmental covariance
• Helium and co-tracers
• Isotope fingerprinting
• Standards & regulations
• Glossary

References

Full reference list (29 entries)

[1] NIST. CODATA recommended value: molar gas constant, $R = 8.314462618$ J/(mol·K) (exact, 2019 SI redefinition). https://physics.nist.gov/cgi-bin/cuu/Value?r

[2] IUPAC. “faradaic current” (F02321). IUPAC Compendium of Chemical Terminology (Gold Book). https://goldbook.iupac.org/terms/view/F02321

[3] IUPAC. “electrochemical method of detection” (E01944). Gold Book. https://goldbook.iupac.org/terms/view/E01944

[4] NIST. CODATA recommended value: Faraday constant, $F = 96{,}485.33212\ldots$ C/mol (exact). https://physics.nist.gov/cgi-bin/cuu/Value?f

[5] Hestia Energy internal research dossier, mox-sensors.md, citing inter alia Yamazoe, N. “Receptor Function and Response of Semiconductor Gas Sensor.” J. Sensors 2009, 875704. DOI: 10.1155/2009/875704.

[6] Barsan, N.; Weimar, U. “Conduction Model of Metal Oxide Gas Sensors.” J. Electroceramics 7, 143–167 (2001). https://link.springer.com/article/10.1023/A:1014405811371

[7] Membrapor AG. “H2/CA-1000” gas sensor datasheet (sensitivity 28 ± 11 nA/ppm), as reviewed in electrochemical-cells.md.

[8] NIST. CODATA recommended value: elementary charge, $e = 1.602176634\times10^{-19}$ C (exact). https://physics.nist.gov/cgi-bin/cuu/Value?e

[9] NIST. CODATA recommended value: Avogadro constant, $N_A = 6.02214076\times10^{23}$ /mol (exact). https://physics.nist.gov/cgi-bin/cuu/Value?na

[10] Sauerbrey, G. “Verwendung von Schwingquarzen zur Wägung dünner Schichten und zur Mikrowägung.” Zeitschrift für Physik 155, 206–222 (1959).

[11] Figaro Engineering Inc. “TGS 821 – Special Sensor for Hydrogen Gas,” product datasheet (sensitivity-slope specification 0.6–1.2 decade/decade). https://www.figarosensor.com/product/docs/TGS821.pdf

[12] Lundström, I.; Shivaraman, S.; Svensson, C.; Lundkvist, L. “A hydrogen-sensitive MOS field-effect transistor.” Appl. Phys. Lett. 26, 55–57 (1975). DOI: 10.1063/1.88053.

[13] Eriksson, M.; Lundström, I.; Ekedahl, L.-G. “A model of the Temkin isotherm behavior for hydrogen adsorption at Pd–SiO2 interfaces.” J. Appl. Phys. 82(6), 3143–3146 (1997). https://pubs.aip.org/aip/jap/article-abstract/82/6/3143/178806

[14] US Patent 7,622,080, “Hydrogen gas sensitive semiconductor sensor” (Helmholtz dipole model; Pd work function ≈5.2 eV). https://image-ppubs.uspto.gov/dirsearch-public/print/downloadPdf/7622080

[15] “Sensing properties of palladium-gate MOS (Pd-MOS) hydrogen sensor based on plasma grown silicon dioxide.” ScienceDirect. https://www.sciencedirect.com/science/article/abs/pii/S0925400599000696

[16] NIST. CODATA recommended value: Boltzmann constant, $k = 1.380649\times10^{-23}$ J/K (exact). https://physics.nist.gov/cgi-bin/cuu/Value?k

[17] “Improved Hydrogen-Sensing of TiO2 Schottky Device Through Schottky Barrier Height Modulation.” Sensors 26(4), 1400 (2026). DOI: 10.3390/s26041400. https://www.mdpi.com/1424-8220/26/4/1400 (a Pt-electrode TiO2 Schottky device; cited here for barrier-height-modulated H2 sensing in reverse bias)

[18] “Highly sensitive hydrogen sensor based on graphite-InP or graphite-GaN Schottky barrier with electrophoretically deposited Pd nanoparticles.” PMC3212005. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3212005/ (reports a current-change ratio of ≈7×10⁵ at 0.1% H2)

[19] Hestia Energy internal research dossier, palladium-resistive.md (β-phase two-phase plateau at ≈20–30 mbar H2 partial pressure, room temperature).

[20] Wikipedia, “Palladium hydride” — compiled crystallographic lattice constants (α-phase 3.889 Å, β-phase 4.025 Å).

[21] Resistance-vs-composition study of bulk PdHx (R/R0 through the α/β transition, maximum ≈1.87 at x≈0.76), as reviewed in palladium-resistive.md.

[22] Favier, F.; Walter, E.C.; Zach, M.P.; Benter, T.; Penner, R.M. “Hydrogen Sensors and Switches from Electrodeposited Palladium Mesowire Arrays.” Science 293(5538), 2227–2231 (2001). DOI: 10.1126/science.1063189.

[23] NIST Chemistry WebBook. “Water” — standard enthalpy of formation, gas phase, $\Delta_fH^\circ = -241.83$ kJ/mol. https://webbook.nist.gov/cgi/cbook.cgi?ID=C7732185&Mask=1

[24] City Technology / Honeywell. “CAT 25 CiTipeL® Combustible Gas Sensor” datasheet (relative-response table including H2 = 107 vs. CH4 = 100). https://prod-edam.honeywell.com/content/dam/honeywell-edam/sps/siot/ru-ru/products/sensors/gas-sensors/other-combustibles/documents/sps-ast-citytech-cat-25-datasheet.pdf

[25] LaConti, A.G.; Maget, H.J.R. “Electrochemical Detection of H2, CO and Hydrocarbons in Inert or Oxygen Atmospheres.” J. Electrochem. Soc. 118(3) (1971).

[26] Chemistry LibreTexts. “Standard Reduction Potential” (O2 + 4H⁺ + 4e⁻ → 2H2O, $E^\circ=+1.229$ V vs. SHE at 25 °C). https://chem.libretexts.org/Bookshelves/Analytical_Chemistry/Supplemental_Modules_(Analytical_Chemistry)/Electrochemistry/Redox_Chemistry/Standard_Reduction_Potential

[27] 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. 155, 214303 (2021). DOI: 10.1063/5.0063256. https://pubs.aip.org/aip/jcp/article/155/21/214303

[28] IUPAC. “limit of detection” (L03540). Gold Book. https://goldbook.iupac.org/terms/view/L03540

[29] Saffell, J.R.; Martin, N.A. “Measurements of the Limit of Detection for Electrochemical Gas Sensors.” J. Testing and Evaluation 52(5), 2675–2684 (2024). DOI: 10.1520/JTE20230675.