2D-material & nanostructure FET sensors

Summary

Graphene, carbon nanotubes, MoS2, MXenes, and black phosphorus are conductors or semiconductors only one to a few atoms thick: a single adsorbed molecule can measurably shift their conductivity because there is no bulk material underneath to dilute the effect. H2 itself barely adsorbs on any bare 2D material at room temperature. A 2025 reactive-MD/TEM study of Pt-functionalized graphene found H2 dissociative chemisorption occurs only on the Pt nanoclusters, with negligible spillover onto the graphene itself [2]. Nearly every “graphene H2 sensor” in the literature is therefore a hybrid: a Pd or Pt nanoparticle runs the same catalytic chemistry used by Pd-resistive sensors and work-function devices, and the 2D lattice serves only as a high-gain electrical transducer for the resulting charge transfer.

That means this family inherits every field weakness of Pd/Pt catalytic sensing (slow recovery, humidity sensitivity, H2S poisoning) and adds its own (batch-to-batch variability, no packaged product, no pressure or long-duration data). What is genuinely new is transducer gain: room-temperature, low-power operation, and laboratory detection limits in the parts-per-billion range. Two decades of papers, since 2001 for CNTs and 2007 for graphene, have not produced a single shippable H2 instrument.

At a glance — published performance envelope

Detection limit

0.005–20,000 ppm

Range

0.005–40,000 ppm

Response t₉₀

3–73 s

Recovery

12–300 s

Selectivity

low

Drift

high

Operating T

0 … 255 °C

Humidity

0–95 % RH

Power

0.001–300 mW

Unit cost

≥ $20

TRL

3

Main interferents: H2O vapor, O₂, H₂S, CO · poisons: H₂S, prolonged high humidity, Pd. Ranges span the published devices/variants covered below — per-claim citations are in the body; data: src/data/modalities/fet-2d-materials.yaml.

How it works

Two channels, only one of which is about hydrogen. The founding 2007 graphene demonstration showed a Hall-bar could detect single molecules of NO2, NH3, H2O, or CO adsorbing and desorbing as discrete resistance steps, because each adsorbed molecule acts as a local charge that shifts carrier density [1]. That mechanism (Channel A) does essentially nothing for H2 itself: H2 is closed-shell and non-polar, and a 2025 reactive-MD/TEM study of Pt-functionalized graphene found H2 dissociative chemisorption occurs with a low barrier only on the Pt nanoclusters, with negligible spillover or chemical interaction with pristine graphene [2]. Practically every device below runs Channel B instead: a catalytic metal, almost always Pd or Pt, splits H2(g) into 2 H(ads), the same first step used by Pd-resistive and work-function sensors, and the 2D lattice reads out the resulting charge transfer with unusual gain.

Readout physics. An adsorbed or catalytic-metal-generated charge shifts the channel’s carrier density exactly as a gate voltage would:

$\Delta V_{Dirac} = -\frac{\Delta Q_{ads}}{C_{ox}}$

This is formally the same relation as the flat-band shift $\Delta V_{FB} = \Delta\Phi_M/q$ that drives Pd-gate work-function devices: a Pd or Pt nanoparticle sitting on the 2D channel builds an interfacial H dipole after dissociating H2, and that dipole locally gates the lattice the same way a continuous catalytic gate film gates a MOS capacitor. Only the geometry differs. DFT puts baseline Pd-cluster-to-graphene charge transfer at ≈0.16 e per cluster before any H2 exposure (unverified) [3].

FET readout adds gain, at a cost. Biasing near the Dirac point amplifies a fixed adsorbed charge into a large fractional resistance change, since $\sigma \propto n_s$ and $n_s \to 0$ there. A floating-gate design goes further, capacitively coupling Pd/Pt nanoparticles to a CNT channel through a thin oxide so the H-induced charge is amplified before it reaches the channel; that architecture holds this family’s LOD record (below) [6]. Reviews of the field agree this buys post-fabrication tunability at the cost of a third terminal and gate-dielectric reliability, and that back-gated lab demonstrators rarely translate into a packaged two- or three-lead field sensor [7][8].

Recovery, not sensitivity, is the field-limiting step. Coverage on the catalytic nanoparticle follows a Langmuir or Temkin isotherm at low pressure, but desorption is thermally activated ($k_d = \nu_0\exp(-E_a/kT)$) and slow at 300 K with typical Pd-H or Pt-H binding energies: minutes to tens of minutes is the typical room-temperature figure for real, non-UV/heat-assisted devices [9][10][11][12]. Mild heating (50–150 °C) or UV illumination (265–365 nm) both accelerate desorption. One graphene device is reported at 3 s response to 5 ppm H2 under 265 nm UV, and a PMMA molecular-sieve overlayer combined with UV desorption is reported to hold response roughly flat across 20–70% RH, but both figures come from search-synthesized summaries rather than a confirmed primary source (unverified).

Characteristic numbers

Quantity	Value	Ref.
Graphene intrinsic (undoped) work function	≈4.4–4.6 eV (4.55 eV SLG / 4.44 eV bilayer by Hall-bar-referenced KPFM; 4.5 eV by cavity-enhanced internal photoemission)	[13][14]
DFT charge transfer, Pd3 cluster → graphene	≈0.16 e per cluster (before H exposure) (unverified)	[3]
DFT adsorption energy, H2S vs. CH4 on vacancy-defect graphene	−2.934 eV vs. −0.154 eV (≈19× stronger for H2S)	[5]
DFT charge transfer, H2S vs. CO on PdO–graphene	0.445 e vs. 0.036 e	[4]
CNT percolation threshold (well-dispersed SWCNT in a matrix)	≈0.01–0.05 wt % (unverified)	[15]
Percolation conductivity power law	$\sigma \propto (\nu-\nu_c)^t$	[15][16]
Record CNT-FET floating-gate H2 LOD	90 ppb (RT) / 5 ppb (≈100 °C)	[6]
Record batch-fabricated CNT chemiresistor H2 LOD	890 ppb (RT) / 89 ppb (100 °C)	[17]
DOE hydrogen safety-sensor targets (for scale)	t90 <1 s at 0.1 vol% H2; LOD 0.1 vol% (1000 ppm); range 0.1–4 vol%; ≤1 W; −40 to +85 °C	[18]

Device classes

Every branch below runs the identical first-step chemistry (Pd or Pt splits H2); what differs is the 2D transducer and its failure modes.

Material family	Status / mechanism	Best reported numbers	Ref
Bare graphene/GO/rGO (undecorated)	Responds well to NO2, NH3, H2O by direct charge transfer; H2 itself needs a catalyst or UV assist, and response is weak and irreproducible without one	—	[1][23]
Pd/Pt-decorated graphene/rGO (dominant branch)	Direct 2D analogue of Pd-resistive and work-function sensors. Sub-variants: flame-synthesized Pd on crumpled 3D rGO, electrodeposited flower-like Pd nanoclusters, a Pd-Ag dual-FET pairing a sensing and a reference device to cancel drift/humidity, Pt-rGO printed on a passive RFID antenna	14% response at 0.2% H2 [19]; 14.8% response, 73 s / 126 s, at 2% H2 [10]	[8][9][10][19][20][21][22]
CNT networks with Pd/Pt	Percolating chemiresistor or gated FET, with discontinuous Pd/Pt coating (a continuous film blocks H2 access). Second-oldest branch after graphene, dating to a 2001 demonstration	Floating-gate FET: 90 ppb RT / 5 ppb at 100 °C [6]; batch chemiresistor: 890 ppb RT / 89 ppb at 100 °C [17]; also wearable ripstop-fabric sheets and transformer dissolved-gas leak detection	[6][17][23][24][25][26]
TMDs: MoS2, WS2	MoS2 is the most studied, bare or Pd/Pt-decorated, as chemiresistor or heterojunction diode. WS2 physisorbs H2 at room temperature but only chemisorbs on W above ≈150 °C, a mechanistically weaker room-temperature pathway	Pd-doped MoS2/Si heterojunction: 360 ppb calculated LOD [29]	[12][27][28][29][30][31]
MXenes (Ti3C2Tx)	Pd-decorated multilayer Ti3C2Tx is “dual-functioning,” simultaneously an H2 sensor and an H2-storage medium via the same Pd-H chemistry. Oxidizes readily in humid/oxygen-rich conditions; untreated devices lose response within a week	Flexible MXene@Pd sheet survives 1000 tilts and 5000 bends	[32][33][34][35]
Black phosphorus (BP)	Chemically the least stable material here, hydrolyzing to phosphate species within hours to days in ambient air; needs noble-metal functionalization for H2 sensitivity	Pt-BP: selectivity factor 4.6 vs. other gases in N2, the family’s best explicit selectivity number [36]; Ni-decoration/encapsulation extends shelf life from ~1 week to >4 weeks [37]	[36][37]
3D/porous architectures (foam, aerogel, crumpled rGO)	Pd/Pt-decorated 3D scaffolds built for more surface area and gas access than a flat sheet	41.9% response at 3% H2 (3 nm Pd on graphene/Cu foam); 14.8% response, 73 s / 126 s, at 2% H2 (crumpled rGO) [10]	[10][11]
Wireless/RFID-tag variants	Pt- or CeO2/Fe2O3-decorated rGO printed on a passive UHF RFID antenna; H2-induced resistance change detunes the antenna for battery-free readout at meter range. Every demonstration found is an indoor benchtop test, not a field trial	LOD ≈1 ppm (Pt/rGO) [21]	[21][22]

DFT screening of Li/Na-decorated, B/N-co-doped, and MOF-functionalized graphene transistors also appears in the literature, but as computational studies only; none had field-relevant experimental H2 data at time of writing (see Open questions).

Performance envelope

How to read this table

Every row is one paper’s specific device, not a generic material class; this family has large device-to-device spread within a single material. Treat “range” as reported test range, not validated linear range.

Device	LOD (H2)	Tested range	Response time	Recovery time	Op. temp	Notes / ref.
Graphene-Pd-Ag dual-gate MEMS FET	not stated as ppm	qualitative H2 response curve	not stated	not stated	25–254.5 °C (best window)	Reference FET on the same chip for drift/humidity cancellation [9]
Pd-decorated graphene (eco-friendly synthesis)	tested down to 0.2% (2000 ppm)	0.2%–	not stated precisely	not stated	room temperature	14% relative conductance change at 0.2% H2, 50% RH [19]
Flame-synthesized Pd-decorated crumpled 3D rGO	tested from 0.0025% (25 ppm)	0.0025–2%	73 s	126 s	room temperature	14.8% response at 2% H2 [10]
Chemically deposited Pd NPs on graphene	not stated as ppm	qualitative/device demo	not stated	not stated	room temperature	Mechanism/fabrication demonstration [20]
Pd NP-decorated transfer-free 3D graphene	not stated as ppm	tested at 3%	not stated	not stated	30 °C	41.9% response, 3 nm Pd optimum
rGO on DEP-positioned AZO electrodes	~100% sensitivity at pure H2, 1 V	pure H2 (qualitative)	~10 min exposure protocol	not stated	ambient	RH flagged as “critical factor” [38]
Pt/rGO on passive RFID tag (wireless)	1 ppm	sub-ppm to several ppm (nonlinear below 1 ppm)	not stated	not stated	room temperature	Battery-free readout via antenna detuning [21]
Pd-MWCNT chemiresistor (air)	2000 ppm	—	not stated	not stated	room temperature	[39]
Pt-Pd-MWCNT chemiresistor (air)	400 ppm	—	not stated	not stated	room temperature	[39]
Pd-MWCNT + Pt layer (N2 ambient)	5 ppm	—	not stated	not stated	room temperature	~2 orders better than air-ambient (O2-competition effect) [39]
Flexible Pd-MWCNT/ripstop-fabric chemiresistor	0.1 mol% (~1000 ppm)	tested at 20 ppm in N2	not stated	reversible without O2 purge or heat	room temperature	Wearable/textile form factor [25]
Batch-fabricated CNT chemiresistor (solution-derived)	890 ppb (RT); 89 ppb (100 °C)	ppb–hundreds ppm	~7 s (at 311 ppm)	not stated precisely	RT–100 °C	First sub-ppm RT report for CNT H2 sensors [17]
Floating-gate FET, CNT channel + Pd NP + Y2O3 gate insulator	90 ppb (RT); 5 ppb (~100 °C)	ppb–%	not fully stated	not fully stated	RT–100 °C	>1 order-of-magnitude gain vs. resistance-type CNT sensors [6]
Pd-nanoparticle-decorated multilayer MoS2	not stated as ppm	tested at 100 ppm	not stated	not stated	150 °C	55% max response at 100 ppm, 150 °C [12]
Pd-nanocube-decorated 1T@2H-MoS2	not stated as ppm	0.01–4% (100–40,000 ppm)	5 s	not stated	room temperature	”Fast and reliable room-temperature” framing [28]
Pd-doped MoS2/Si heterojunction	360 ppb (calc., 3σ/slope)	tested to 20,000 ppm	24.6 s (at 10,000 ppm)	12.2 s (at 10,000 ppm)	room temperature	Response ≈6.4×10³% at 10,000 ppm; 51% response change 50→80% RH; 14-cycle repeatability, no weeks/months data [29]
Pt NP-decorated MoS2 (unverified)	not stated as ppm	qualitative	not stated	not stated	not stated	[27]
WS2 (bare, room temp)	not stated as ppm	qualitative	not stated	not stated	RT (physisorption); >150 °C for chemisorption on W	Mechanistically weaker RT pathway than Pd/Pt-catalyzed devices [30]
Pd-decorated multilayer Ti3C2Tx MXene	not stated as ppm	qualitative + storage function	not stated	not stated	RT (2 V applied)	Optimal 1.1 at% Pd; dual H2-sensing/H2-storage [33]
MXene@Pd CNC flexible sensor	not stated as ppm	tested to 4% (40,000 ppm)	32 ± 7 s	not stated	RT (flexible/wearable)	Response (23 ± 4)% at 4% H2 [34]
Ni-decorated black phosphorus	not stated for H2 specifically	qualitative	not stated	not stated	room temperature	Ambient-stability focus, not raw H2 sensitivity [37]
Pt-incorporated black phosphorus	”most sensitive” vs. prior 2D H2 sensors (no ppm figure)	qualitative	not stated	not stated	RT (N2 background)	Selectivity factor 4.6 vs. other gases in N2 [36]
Undecorated CVD graphene, UV-assisted	not stated as ppb/ppm	tested at 5 ppm	3 s (under 265 nm UV)	improved but not quantified	room temperature (UV-assisted)	Represents the “no catalytic metal” ceiling case: response collapses without UV assist (unverified)

Measured against the DOE targets in the table above (t90 <1 s at 0.1 vol%, LOD 0.1 vol%, ≤1 W, −40 to +85 °C [18]), several devices individually clear the LOD target by three to six orders of magnitude and the power target trivially, but none has demonstrated that LOD together with DOE’s humidity/temperature range, high-concentration response time, and multi-week stability at once.

Figures of merit

The universal figures of merit are defined on the Figures of merit page; units (ppm, ppb, vol%) on Concentration units. Two modality-specific rows matter most here: transconductance/Dirac-point shift, and the humidity coefficient.

Figures-of-merit table for this modality (10 rows)

Figure of merit	Symbol / formula	What it means here	Typical value for this technology	How to read it
Sensitivity (transconductance / Dirac-point shift)	$g_m = \partial I_D/\partial V_G$; $\Delta V_{Dirac} = -\Delta Q_{ads}/C_{ox}$	How much channel current, or the Dirac/minimum-conductivity point, shifts per unit of Pd/Pt-mediated adsorbed charge; FET geometry lets the operator bias into the high-gain region near the Dirac point where $\lvert\Delta R/R\rvert \approx \lvert\Delta n_s/n_s\rvert$ diverges	No universal $g_m$ figure is reported across this family. Devices instead quote fractional response: 14% at 0.2% H2 [19], 14.8% at 2% H2 [10], up to ≈6.4×10³% at 10,000 ppm for the Pd-doped MoS2/Si heterojunction [29]; floating-gate CNT-FET reports “>1 order of magnitude” gain over resistance-type readout [6]	Compare normalized (%) response, not raw current or $g_m$, since channel geometry and gate dielectric differ device to device; the “order-of-magnitude gain” claim reflects real transducer physics
Limit of detection (room-temperature, ppb claims)	$\mathrm{LOD} = k\,\sigma_{blank}/S$	Smallest H2 concentration statistically distinguishable from baseline noise	90 ppb RT, floating-gate CNT-FET [6]; 890 ppb RT, batch-fabricated CNT chemiresistor [17]; 360 ppb (calculated, 3σ/slope), Pd-doped MoS2/Si heterojunction [29]	Check the measurement atmosphere before trusting a ppb figure. A matched Pd-MWCNT comparison shows ~400× worse LOD in air (2000 ppm) than in N2 (5 ppm) [39]; many of this family’s ppb-class headline LODs were measured in inert-gas gloveboxes and may be optimistic by 1–3 orders of magnitude relative to true open-air field performance
Selectivity / cross-sensitivity	$K_{ij} = S_j/S_i$; selectivity $=1/\max_j(K_{ij})$	Discrimination of H2 against co-evolved CH4, H2S, CO, and humidity	Best explicit device-level figure in the whole family: BP/Pt’s selectivity factor of 4.6 vs. other gases in a N2 background [36], far below the >10³–10⁶ selectivity claims typical of electrochemical or well-engineered Pd-resistive products; DFT shows H2S binds ≈19× more strongly than CH4 on graphene by adsorption energy [5], implying a real but device-unverified H2S cross-sensitivity/poisoning risk	Treat every selectivity number in this family as a single-study result rather than a validated multi-gas matrix; no H2-vs-CH4-vs-H2S-vs-CO device-level selectivity study was located for this page
Response time	$t_{90} = 2.303\,\tau$ (single-pole approximation)	Time to reach 90% of signal after a step increase in H2	3 s at 5 ppm under 265 nm UV assist (unverified) [device table]; 5 s at concentrations up to 4% for Pd-nanocube MoS2 [28]; 73 s at 2% H2 for flame-synthesized Pd-rGO [10]; 24.6 s at 10,000 ppm for the Pd-doped MoS2/Si heterojunction [29]	Fast numbers cluster at high (%) concentrations; response time at ppm-relevant, Hestia-scale concentrations is rarely reported separately, so these figures should not be assumed to hold at 1–10 ppm
Recovery time (slow, desorption-limited)	$k_d = \nu_0\exp(-E_a/kT)$	Time to return to baseline after H2 is removed; governed by thermally activated desorption of chemisorbed H from Pd/Pt, not by the 2D channel itself	Minutes to tens of minutes is the typical room-temperature figure for real (non-UV/heat-assisted) devices [9][10][11][12]; 126 s at 2% H2 for flame-synthesized Pd-rGO is one of the few directly quantified values [10]	This, not LOD, is usually the field-limiting number for this family. The two engineered fixes (mild heating, 50–150 °C, or UV illumination, 265–365 nm) both erode the room-temperature/low-power advantage the family is marketed on, and the UV-assisted fast-recovery claims are themselves flagged (unverified) pending primary-source confirmation
Dynamic range	$N_{\text{decades}} = \log_{10}(C_{\max}/C_{\min})$	Span of concentration over which a device produces usable output	Individual devices span from the ppb floor (floating-gate FET, ppb–% range [6]) to 4% / 40,000 ppm (MXene@Pd CNC [34]). No single device has been tested across that full span	Response near the Dirac point is inherently log-linear, not a fixed-slope linear range like an electrochemical cell (see Figures of merit); “tested range” in the Performance envelope table above is reported test range, not validated linear range
Drift / long-term stability	$R_0(t) = R_0(0) + d_Z t$; $S(t) = S(0)(1+d_S t)$	Baseline and sensitivity creep over deployment duration	No H2-specific dataset longer than a lab test run (hours) exists for any device in this family (unverified/absent); the best analog is non-H2 MXene gas sensing, where pristine Ti3C2Tx loses essentially all response within 1 week absent mitigation, extending to 150 days with pre-oxidation/compositing [35]	Assume unknown drift for every device in this table until a multi-week, H2-specific, variable-RH dataset is published; this is the highest-value missing measurement for this page (see Open questions)
Humidity coefficient	$C_{\text{reported}} = C_{\text{raw}}\cdot f_{RH}(RH)$	RH-driven shift in baseline and/or response amplitude, layered on top of the true H2 signal	51% loss of response amplitude moving 50→80% RH at fixed 20,000 ppm H2, Pd-doped MoS2/Si heterojunction [29]; GO/rGO baseline conductivity can shift up to six orders of magnitude across 10–95% RH independent of any H2 exposure; RH flagged as “critical” without being quantified for several other devices [9][19][38]	The most repeated confound in this family; a serious problem for soil-gas use without an independent RH sensor and correction model, or a molecular-sieve overlayer (both unproven outside single-study demonstrations)
Power / operating envelope	—	Electrical power draw, and the temperature/RH/pressure window over which stated performance actually holds	Room-temperature, µW–mW-class operation (no heater) is the family’s headline advantage; best-characterized temperature window is 25–254.5 °C for one MEMS variant [9]; zero devices in this family have been characterized above 1 atm	The power/no-heater advantage is real and genuinely differentiates this family from MOX and pellistor devices, but it has never been demonstrated as part of a validated compensated envelope; no pressure data exists at all, ruling out downhole consideration on this axis alone
Contrast vs. bulk Pd-resistive (shared chemistry, different gain)	$\Delta R/R$ (bulk Pd-resistive) vs. $\lvert\Delta n_s/n_s\rvert$ (FET/2D, diverging near the Dirac point)	Both families run the identical first-step chemistry, H2(g) → 2 H(ads) on Pd/Pt; only the transduction gain differs	Bulk Pd-resistive $\Delta R/R$ reaches up to ≈0.8–0.87 (see Figures of merit and Palladium resistive sensors); the 2D/FET floating-gate architecture instead reports record ppb-level LODs (90 ppb RT [6]; 890 ppb RT [17]) that beat typical Pd-resistive LODs by orders of magnitude on paper	The transducer gain is genuine physics, but it inherits every Pd/Pt chemistry weakness (humidity, slow recovery, H2S poisoning) unchanged; higher gain does not fix chemistry it did not change

What dominates

Humidity and recovery time decide whether this family is field-usable, not LOD: several devices already clear DOE and soil-gas anomaly thresholds by orders of magnitude on paper [6][17]. The same catalytic nanoparticles that give the ppb-level LODs are the ones most exposed to humidity competition and slowest to desorb, so low LOD comes with no corresponding gain in stability or selectivity. Until a device shows humidity-compensated stability over weeks rather than hours, a ppb-level datasheet number should not drive a procurement decision.

Cross-sensitivity & environment

Humidity is the dominant, unsolved covariate for this family, with H2S poisoning and batch-to-batch reproducibility close behind. All three are downstream of chemistry shared with Pd-resistive and work-function sensors; the 2D transducer does not change them.

Factor	Effect	Design response	Ref
Humidity	Water competes for adsorption sites on the Pd/Pt nanoparticles, and separately drives GO/rGO baseline conductivity by up to six orders of magnitude across 10–95% RH independent of H2. The Pd-doped MoS2/Si heterojunction loses 51% of response moving 50→80% RH at 20,000 ppm H2	Needs an independent RH sensor and correction model, or a molecular-sieve overlayer; unproven outside single-study demonstrations	[8][9][19][29][38]
Oxygen	Not required (Pd/Pt dissociate H2 in inert gas too), but O2 measurably interferes: Pd-MWCNT LOD is ~400× worse in air (2000 ppm) than N2 (5 ppm), attributed to competing PtO2 surface chemistry that must be reduced by H2 first	Treat any number not explicitly stated “in air” as an inert-atmosphere best case, optimistic by 1–3 orders of magnitude	[8][39]
CH4	Genuine, physically founded advantage: CH4 binds far more weakly than H2S to graphene by DFT (−0.022 to −0.154 eV vs. −0.7 to −2.9 eV); devices report “insensitive to CH4” at 75–150 °C	Not systematically validated across the family or at Hestia-relevant concentration ratios	[5]
H2S	The recurring poison, as in the Pd-resistive and work-function families; DFT ranks H2S binding far stronger than CH4/CO on graphene, the same asymmetry dedicated graphene H2S sensors exploit	No dedicated H2S-poisoning aging study for a 2D-material H2 sensor was located; priority gap given Hestia’s co-evolved-gas context	[4][5]
CO	Intermediate cross-sensitivity by DFT (0.036 e vs. 0.445 e for H2S on PdO-graphene); named among the reducing gases that cross-respond generally	No dedicated H2-vs-CO selectivity-ratio study for this family was found	[4]
CO2, He	No family-specific cross-sensitivity data located	Low interference is the physically reasonable expectation by analogy with Pd-resistive sensors, but treat as an assumption	—
Temperature	Characterized only at room temperature or up to ~150–254 °C for the handful of heated/MOS-hybrid variants; none below 0 °C or at sustained downhole-relevant temperatures	Rules out downhole consideration on this axis alone until it changes	[9][12]
Pressure	No device in the surveyed literature was characterized above 1 atm	Structural evidence gap; rules out downhole consideration until filled	—
Batch-to-batch variability	Named repeatedly as a first-order commercialization barrier independent of any device’s headline numbers; CVD graphene’s grain-boundary defects and the lack of a reliable coating/deposition process are the cited causes	No quantitative coefficient-of-variation study across ≥10 devices was found	—
Material-specific aging	MXenes oxidize in humid storage (untreated Ti3C2Tx loses essentially all response within a week); black phosphorus hydrolyzes to phosphates fastest of any material here; Pd/Pt nanoparticle sintering at elevated temperature or after repeated dosing is a known general failure mode by analogy with Pd-resistive sensors	Encapsulation/pre-oxidation (MXene) and Ni-decoration/encapsulation (BP) are the leading fixes	[35]

H2S in detail. DFT ranks H2S as binding far more strongly to graphene and graphene composites than CH4 or CO does, the same asymmetry that dedicated graphene-based H2S sensors exploit for their own sensitivity [4][5]. For a H2 sensor this is a mirror-image problem: any H2S co-evolved at a field site is likely to bind more strongly and persistently than H2 to the same Pd/Pt/graphene sites, producing both a confounded signal and progressive, possibly irreversible sensitivity loss. No dedicated H2S-poisoning aging study for a 2D-material H2 sensor was located, a priority gap given Hestia’s co-evolved-gas field context.

Use for geologic H₂

Field verdict

Every advantage here (room temperature, low power, no heater, flexible form factor) is a transducer-level advantage. Every weakness (humidity, drift, incomplete selectivity, batch variability, and no pressure or long-duration data) is a materials/systems-engineering weakness on top of chemistry that is not itself novel. At best this is a candidate future low-power front-end for qualitative soil-gas screening; it should not currently be considered for downhole or unattended long-duration monitoring.

Soil-gas surveys. Best fit on paper: room-temperature, battery-powered, spot-measurement operation matches the soil-gas survey profile better than heated MOX or pellistor instruments. But soil pore gas runs near saturation with large diurnal RH swings, and the best-characterized device loses roughly half its response amplitude over a 50→80% RH excursion [29], while several others flag RH as the dominant confound without quantifying it. No H2-vs-CH4/H2S selectivity-ratio field validation exists either. A defensible role today is qualitative screening only, paired with an independent RH sensor and correction model, unproven outside single-study lab demonstrations.

Continuous monitoring stations. Superficially attractive for a battery- or solar-powered station given the low-power, no-heater profile, but no H2-specific multi-week stability dataset exists for any device; the closest analog is a different target gas entirely [35]. A station built on this technology today would have no basis for predicting its drift. The one narrow opportunity is a battery-free, disposable RFID-tag sensor read wirelessly and replaced often rather than trusted long-term [21][22], but that has only been demonstrated indoors on a benchtop.

Downhole logging. Ruled out as currently demonstrated. No device has been characterized above 1 atm; temperature coverage tops out near 150–254 °C [9][12], below deep-exploration downhole targets, and no study ran longer than a lab test (hours, not the weeks to months downhole needs). Every device, chemiresistor or FET, also needs local active electronics at the sensing point, the opposite of the no-downhole-electronics preference this role wants.

Bottom line. Genuine transducer-gain physics on top of chemistry that is not novel, and no shippable H2 product after two decades of papers. Track this family as a research candidate, not a procurement option. See Why H2 is hard and Geologic context.

Vendors & products

No vendor ships a hydrogen sensor built on graphene, CNT, TMD, MXene, or black-phosphorus technology. Every company that surfaces in a “graphene sensor” search sells something else.

Company	What they actually make	Hydrogen gas-sensing product?	TRL / status	Source
Grapheal (Grenoble, France; CNRS spin-off)	Graphene FET biosensors for wound monitoring (WoundLAB), diagnostics (TestNPass, GraphLAB assay kits), transparent conductive films	No: biomedical/diagnostic focus, not gas sensing	Product-stage (CES 2022 award) but for a completely different application space	[40]
Paragraf (UK)	Graphene Hall-effect sensors (GHS series) for magnetic-field sensing; separately markets “Graphene Molecular Sensors (GMS)” for gas/ion/biosensing markets	Graphene Hall sensors are magnetic, not gas; GMS gas-sensing line exists in marketing material but no H2-specific datasheet, SKU, or spec sheet was found	Hall-sensor line commercial/shipping (TRL 8–9) for magnetic sensing; GMS applicability to H2 unverified	[41]
Archer Materials (Australia, ASX:AXE)	“Biochip”, a graphene FET (gFET) biosensor for medical diagnostics (e.g., potassium-ion sensing for chronic kidney disease, partnered with Paragraf)	No: biomedical point-of-care diagnostics, not gas sensing	Proof-of-concept/pre-commercial	[42]
C2Sense (Cambridge, MA; MIT/Swager lab spin-out)	CNT-network chemiresistor with a “selector” chemistry layer for ethylene (fruit ripeness) and amine/ammonia (meat/fish freshness) detection, marketed as “the first commercially available CNT sensor”	No: explicitly a food-freshness sensor, not H2	Commercial product (food industry)	[43]
Grolltex (San Diego, CA)	CVD graphene material supplier and foundry/contract-manufacturing services; markets graphene as an enabling material for gas sensors (NO2 examples cited) and holds a patent on a multi-modal “super sensor” concept	No specific H2 product found; sells raw graphene material and foundry services, not a finished calibrated H2 sensor	Materials supplier (TRL for their own material is high; TRL of any resulting H2 sensor product is unestablished)	[44]
MSE Supplies (materials/lab-equipment vendor)	Sells a generic “GFET Sensor Chip” marketed as “optimal for gas and volatile compound detection”, a bare, uncalibrated research die	No H2-specific calibration, datasheet spec, or application note found	Research-grade component, not a sensor product	(unverified)
Established H2 gas-sensor incumbents for comparison: Teledyne, Honeywell, H2scan, Figaro, SGX Sensortech, Alphasense	Portfolio hydrogen detectors. Honeywell’s 2025 Hydrogen Leak Detector (HLD) uses thermal-conductivity detection claiming leaks down to 50 ppm; Alphasense’s HB/H2 lines are electrochemical; H2scan uses Pd-alloy resistive/MIS hybrids	Yes, but none of these incumbent commercial hydrogen-detector product lines use graphene/2D-material/CNT-FET technology in any product identified	TRL 8–9 (shipping, calibrated, certified products)	see Work-function devices, Palladium resistive sensors, Thermal conductivity & acoustic sensors

Materials suppliers (Grolltex, MSE Supplies) sell the enabling graphene/GFET substrate, not a calibrated H2 instrument; the rest target magnetic sensing (Paragraf), biomedical diagnostics (Grapheal, Archer), or food freshness (C2Sense). Every established H2 detector line uses thermal-conductivity, electrochemical, catalytic-bead, or Pd-alloy resistive/MIS technology instead, exactly the technologies covered elsewhere in this wiki. The mechanism has been published continuously since 2001 (CNT [24]) and 2007 (graphene [1]) without converting into a single named, shippable H2 SKU.

Open questions

1. 1000-hour (or longer) cycling-RH exposure test for any Pd/Pt-decorated graphene, CNT, MoS2, or MXene device, the highest-value missing dataset for this page. None was found for hydrogen specifically.
2. A device-level H2-vs-CH4-vs-H2S-vs-CO2-vs-He selectivity matrix at Hestia-relevant concentration ratios. Current evidence is DFT-computed adsorption energies (a proxy, not a measurement) or single-gas superlatives [5][36].
3. Any pressure-characterized (above 1 atm) measurement for any device in this family. Zero were found; this alone rules out downhole consideration until it changes.
4. A field, non-benchtop pilot deployment of any device in this family for hydrogen, even a few days. All positive results surveyed were laboratory bench, cylinder-gas, or inert-atmosphere tests.
5. Whether the ~400× air-vs-inert-atmosphere LOD gap [39] is general across the family or an outlier. If general, every ppb-class headline LOD measured in N2/glovebox conditions should be discounted 1–3 orders of magnitude for open-air use.
6. A batch-to-batch reproducibility study (coefficient of variation across ≥10 nominally identical devices). Repeatedly named as a barrier in review literature, never quantified for hydrogen sensing specifically.
7. Verification of the UV-assisted room-temperature recovery claims (3 s response to 5 ppm H2 under 265 nm UV; PMMA maintaining response across 20–70% RH). Both were retrieved as search-synthesized summaries without a confirmed primary source; treat as (unverified).
8. Cost data. No specific $/unit price was found for any graphene/2D-material/CNT hydrogen sensor component or module.
9. Track the wafer-scale CMOS-integrated graphene sensor line [45] and the floating-gate CNT-FET line [6][17] as the two most credible near-term paths to a manufacturable product; revisit this page if either produces a packaged, calibrated hydrogen SKU.

Related pages

• Why H2 is hard
• Geologic context
• Work-function devices
• Palladium resistive sensors
• MOX chemiresistors
• Catalytic pellistors
• Thermal conductivity & acoustic sensors
• Electrochemical cells
• Plasmonic metamaterials
• Soil-gas surveys
• Downhole deployment
• Calibration & QA/QC
• Environmental covariance

References

Full reference list (45 entries)

1. Schedin, F., Geim, A.K., Morozov, S.V., Hill, E.W., Blake, P., Katsnelson, M.I., Novoselov, K.S. “Detection of individual gas molecules adsorbed on graphene.” Nature Materials 6, 652–655 (2007). DOI: 10.1038/nmat1967. https://www.nature.com/articles/nmat1967
2. “Data-Driven Molecular Dynamics and TEM Analysis of Crystal Growth and Hydrogen Sensing in Pt-Functionalized Graphene Chemiresistive Sensors.” arXiv:2504.05438 (2025). https://arxiv.org/abs/2504.05438
3. “Reversible doping of graphene field effect transistors by molecular hydrogen: the role of the metal/graphene interface.” 2D Materials (2019/2020). DOI: 10.1088/2053-1583/ab0b23. Preprint: https://arxiv.org/pdf/2003.02346 (does not support the ≈0.16 e Pd3-cluster→graphene charge-transfer figure cited to it in the Characteristic Numbers table)
4. “Theoretical Study of CO, NO, NO2, Cl2, and H2S Adsorption Interactions with PdO–Graphene Composites for Gas Sensor Applications.” (2024). https://pmc.ncbi.nlm.nih.gov/articles/PMC11767997/
5. “Performance of Intrinsic and Modified Graphene for the Adsorption of H2S and CH4: A DFT Study.” (2020). https://pmc.ncbi.nlm.nih.gov/articles/PMC7075213/
6. “Sub-10 parts per billion detection of hydrogen with floating gate transistors built on semiconducting carbon nanotube film.” Carbon (2021). https://www.sciencedirect.com/science/article/abs/pii/S0008622321004656
7. Recum, P., Hirsch, T. “Graphene-based chemiresistive gas sensors.” Nanoscale Advances (2024), RSC. DOI: 10.1039/D3NA00423F. https://pmc.ncbi.nlm.nih.gov/articles/PMC10729924/
8. “Recent Advances in Chemoresistive Gas Sensors Using Two-Dimensional Materials.” (2024). https://pmc.ncbi.nlm.nih.gov/articles/PMC11397462/
9. “MEMS based highly sensitive dual FET gas sensor using graphene decorated Pd-Ag alloy nanoparticles for H2 detection.” Scientific Reports (2018). https://www.nature.com/articles/s41598-018-24324-z; https://pmc.ncbi.nlm.nih.gov/articles/PMC5897527/
10. “Hydrogen Sensing at Room Temperature Using Flame-Synthesized Palladium-Decorated Crumpled Reduced Graphene Oxide Nanocomposites.” ACS Sensors (2020). DOI: 10.1021/acssensors.0c01040. https://pubs.acs.org/doi/abs/10.1021/acssensors.0c01040
11. “Highly responsive hydrogen sensor based on Pd nanoparticle-decorated transfer-free 3D graphene.” Sensors and Actuators B: Chemical (2023). https://www.sciencedirect.com/science/article/abs/pii/S0925400523016313
12. “Pd-Nanoparticle-Decorated Multilayered MoS2 Sheets for Highly Sensitive Hydrogen Sensing.” Chemosensors 11(11), 550 (2023). https://www.mdpi.com/2227-9040/11/11/550
13. “Direct Measurement of Intrinsic Dirac Point and Fermi level at Graphene/Oxide interface and Its Band Alignment by Cavity Enhanced Internal Photoemission.” NIST. https://www.nist.gov/publications/direct-measurement-intrinsic-dirac-point-and-fermi-level-grapheneoxide-interface-and
14. “Standardization of surface potential measurements of graphene domains.” Scientific Reports (2013), srep02597. https://www.nature.com/articles/srep02597; and atmospheric H2 background (~0.5–0.53 ppmv): “Hydrogen in the atmosphere: Observations above a forest canopy in a polluted environment,” J. Geophys. Res. Atmos. (2003), https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2001JD001199; NOAA Global Monitoring Laboratory context, https://gml.noaa.gov/aggi/aggi.html
15. “Theoretical Prediction of Electrical Conductivity Percolation of Poly(lactic acid)—Carbon Nanotube Composites in DC and RF Regime.” (2023). https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10420098/ (does not support the claim)
16. “Electrical percolation thresholds of semiconducting single-walled carbon nanotube networks in field-effect transistors.” Physical Chemistry Chemical Physics (2015), RSC. https://pubs.rsc.org/en/content/articlelanding/2015/cp/c4cp05964f
17. Xiao, M., Liang, S., Han, J., Zhong, D., Liu, J., Zhang, Z., Peng, L. “Batch Fabrication of Ultrasensitive Carbon Nanotube Hydrogen Sensors with Sub-ppm Detection Limit.” ACS Sensors 3(4), 749–756 (2018). DOI: 10.1021/acssensors.8b00006. https://pubs.acs.org/doi/10.1021/acssensors.8b00006
18. “High-Performance Nanostructured Palladium-Based Hydrogen Sensors—Current Limitations and Strategies for Their Mitigation.” ACS Sensors 5(11) (2020). https://pmc.ncbi.nlm.nih.gov/articles/PMC7735785/ (citing U.S. DOE EERE Fuel Cell Technologies Office, Multi-Year Research, Development, and Demonstration Plan, 2011–2020, for the DOE hydrogen safety-sensor performance targets)
19. “Fully eco-friendly H2 sensing device based on Pd-decorated graphene.” Sensors and Actuators B: Chemical (2016). https://www.sciencedirect.com/science/article/abs/pii/S0925400516312710
20. “Chemically deposited palladium nanoparticles on graphene for hydrogen sensor applications.” Scientific Reports (2019). https://www.nature.com/articles/s41598-019-40257-7
21. Lee, J.S. et al. “Wireless Hydrogen Smart Sensor Based on Pt/Graphene-Immobilized Radio-Frequency Identification Tag.” ACS Nano 9(8) (2015). DOI: 10.1021/acsnano.5b02024. https://pubs.acs.org/doi/abs/10.1021/acsnano.5b02024
22. “High conductivity graphite paste for radio frequency identification tag with wireless hydrogen sensor based on CeO2–Fe2O3–graphene oxide.” RSC Advances (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12013617/
23. “Graphene-Based Hydrogen Gas Sensors: A Review.” Processes 8(5), 633 (2020), MDPI. https://www.mdpi.com/2227-9717/8/5/633
24. Kong, J., Chapline, M.G., Dai, H. “Functionalized Carbon Nanotubes for Molecular Hydrogen Sensors.” Advanced Materials 13(18), 1384–1386 (2001). https://advanced.onlinelibrary.wiley.com/doi/abs/10.1002/1521-4095(200109)13:18%3C1384::AID-ADMA1384%3E3.0.CO;2-8
25. “Hydrogen Sensors Based on Flexible Carbon Nanotube-Palladium Composite Sheets Integrated with Ripstop Fabric.” ACS Omega (2019/2020). https://pmc.ncbi.nlm.nih.gov/articles/PMC6964304/ ; https://pubs.acs.org/doi/10.1021/acsomega.9b03023
26. “The Functionalized Single-Walled Carbon Nanotubes Gas Sensor With Pd Nanoparticles for Hydrogen Detection in the High-Voltage Transformers.” Frontiers in Chemistry (2020). https://pmc.ncbi.nlm.nih.gov/articles/PMC7154150/
27. “Highly sensitive two-dimensional MoS2 gas sensor decorated with Pt nanoparticles.” Royal Society Open Science 5(12), 181462 (2018). https://royalsocietypublishing.org/doi/10.1098/rsos.181462 (does not support the claim; this paper reports NH3/H2S sensing only, not H2)
28. “Pd nanocube-decorated MoS2 hybrid nanostructures for fast and reliable room-temperature hydrogen detection.” International Journal of Hydrogen Energy (2025). https://www.sciencedirect.com/science/article/pii/S0360319925049535
29. Ma, E., Xu, Z., Sun, A., Yang, S., Jiang, J. “High-Performance Hydrogen Gas Sensor Based on Pd-Doped MoS2/Si Heterojunction.” Sensors 25(15), 4753 (2025). DOI: 10.3390/s25154753. https://pmc.ncbi.nlm.nih.gov/articles/PMC12349308/
30. “Hydrogen Sensing Mechanism of WS2 Gas Sensors Analyzed with DFT and NAP-XPS.” (2023). https://pmc.ncbi.nlm.nih.gov/articles/PMC10224176/
31. “Photoresponsive and Gas Sensing Field-Effect Transistors based on Multilayer WS2 Nanoflakes.” Scientific Reports (srep05209). https://www.nature.com/articles/srep05209
32. “Unveiling the potential of Ti3C2Tx MXene for gas sensing: recent developments and future perspectives.” Materials Advances (RSC). DOI: 10.1039/D3MA00631J (2023). https://pubs.rsc.org/en/content/articlehtml/2023/ma/d3ma00631j
33. “Palladium nanoparticle-decorated multi-layer Ti3C2Tx dual-functioning as a highly sensitive hydrogen gas sensor and hydrogen storage.” (2021). https://pmc.ncbi.nlm.nih.gov/articles/PMC8694963/
34. Zhu, Z., Liu, C., Jiang, F., Liu, J., Ma, X., Liu, P., Xu, J., Wang, L., Huang, R. “Flexible and lightweight Ti3C2Tx MXene@Pd colloidal nanoclusters paper film as novel H2 sensor.” Journal of Hazardous Materials 399, 123054 (2020). DOI: 10.1016/j.jhazmat.2020.123054. https://www.sciencedirect.com/science/article/abs/pii/S0304389420310438
35. “Development of Long-Term Stable MXene-Based Gas Sensing Material.” (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12655386/
36. Cho, S.-Y., Koh, H.-J., Yoo, H.-W., Jung, H.-T. “Tunable Chemical Sensing Performance of Black Phosphorus by Controlled Functionalization with Noble Metals.” Chemistry of Materials 29(17), 7197–7205 (2017). https://pubs.acs.org/doi/abs/10.1021/acs.chemmater.7b01353
37. “Air Stable Nickel-Decorated Black Phosphorus and Its Room-Temperature Chemiresistive Gas Sensor Capabilities.” (2021). https://pmc.ncbi.nlm.nih.gov/articles/PMC8461602/
38. “A route to detect H2 in ambient conditions using a sensor based on reduced graphene oxide.” Sensors and Actuators B: Chemical (2020). https://www.sciencedirect.com/science/article/abs/pii/S0924424719313226
39. “Carbon Nanotube-Based Chemiresistive Sensors.” Sensors 17(4), 882 (2017). https://pmc.ncbi.nlm.nih.gov/articles/PMC5424759/ ; https://www.mdpi.com/1424-8220/17/4/882
40. Grapheal company/product information: https://grapheal.com/ ; “Grapheal raises €1.9m to develop graphene-based biosensors,” Silicon Republic, https://www.siliconrepublic.com/start-ups/grapheal-biosensors-graphene; “New product for seamless graphene biosensing,” Graphenea, https://www.graphenea.com/blogs/graphene-news/new-product-for-seamless-graphene-biosensing
41. Paragraf GHS (Graphene Hall Sensor) product line: https://www.paragraf.com/hall-sensors/ ; https://store.paragraf.com/collections/ghs; GHS-series brochure: https://www.paragraf.com/wp-content/uploads/2022/07/GHS-series-brochure.pdf
42. Archer Materials Biochip: https://www.archerx.com.au/biochip; “Archer Materials completes proof-of-concept biochip graphene sensor technology,” Graphene-Info, https://www.graphene-info.com/archer-materials-completes-proof-concept-biochip-graphene-sensor-technology
43. C2Sense: Tracxn company profile, https://tracxn.com/d/companies/c2sense/__ZKaMXhYdcd5FxuzXtdi9ZmzONjILUZtVfRhK-x_9pLQ; “C2Sense wins award for ‘Selective CNT Gas Sensors’,” Safar Partners, https://www.safar.partners/portfolio-companies/c2sense-wins-award-for-graphene-2d-materials-most-significant-advancement-in-graphene-cnt-or-2d-material-commercialization/ ; MIT STEX profile, https://startupexchange.mit.edu/startup-features/c2sense
44. Grolltex graphene gas sensor applications: https://grolltex.com/applications-graphene-gas-sensors/ ; https://grolltex.com/graphene-sensors/
45. “Towards wafer-scale 2D material sensors.” 2D Materials (2025). DOI: 10.1088/2053-1583/adac73. https://iopscience.iop.org/article/10.1088/2053-1583/adac73