Plasmonic & metamaterial sensors

Summary

Metal nanoparticles behave unusually when light hits them: the free electrons at the surface slosh back and forth together in a collective wave called a plasmon, tuned to a specific color that depends on the particle’s size, shape, and what is happening right at its surface. Build those nanoparticles out of palladium and the effect becomes useful for sensing, because palladium splits H2 molecules apart and dissolves the resulting hydrogen atoms into its own crystal lattice, swelling slightly and changing its own electron count as it does. That swelling-and-electron-count change nudges the color the nanoparticle prefers to absorb or scatter, so a beam of light shone onto a chip covered in engineered palladium nanostructures — and read out at the far end of an optical path, potentially meters or kilometers away down a fiber — can report a hydrogen concentration with no electronics anywhere near the gas itself. That is the appeal for Hestia: a sensing physics that is intrinsically spark-free and could, in principle, put zero active electronics downhole.

Palladium is a poor plasmonic metal compared to gold or silver, with high optical loss, so the field has spent two decades either accepting a weak, broad, slow signal from pure palladium nanostructures, or building two-part devices where a narrow-linewidth gold antenna reports on a separate palladium patch doing the hydrogen chemistry nearby.

The best laboratory demonstrations now reach sub-ppm sensitivity with sub-second response, covering Hestia’s ~1 ppm–to–percent exploration window. But almost none of this exists outside a spectrometer-equipped optics bench: no dust-and-moisture-hardened field instrument, no high-pressure/high-temperature-tested downhole version, and, as of this writing, the one company that ever shipped a commercial hydrogen sensor product built on this physics is selling that business off entirely.

At a glance — published performance envelope

Detection limit

0.25–5,000 ppm

Range

0.25–40,000 ppm

Response t₉₀

0.4–2,400 s

Recovery

1.5–2,400 s

Selectivity

medium

Drift

medium

Operating T

−30 … 130 °C

Humidity

0–98 % RH

Power

n/a

Unit cost

n/a

TRL

5

Main interferents: O₂, CO, H₂S, relative humidity · poisons: CO, H₂S, NOx. Ranges span the published devices/variants covered below — per-claim citations are in the body; data: src/data/modalities/plasmonic-metamaterials.yaml.

Mechanism & physics

Localized surface plasmon resonance — the readout physics

A metal nanoparticle much smaller than the wavelength of light supports a localized surface plasmon resonance (LSPR): the incident field drives the conduction-electron gas into a collective oscillation against the fixed ionic background. In the quasistatic limit for a sphere of radius $a$ in a host medium of permittivity $\varepsilon_m$, the induced dipole polarizability is

$$\alpha(\omega) = 4\pi\varepsilon_0 a^3\,\frac{\varepsilon(\omega)-\varepsilon_m}{\varepsilon(\omega)+2\varepsilon_m},$$

which diverges — the Fröhlich condition — when $\mathrm{Re}\,[\varepsilon(\omega)] = -2\varepsilon_m$. The extinction cross-section is

$$\sigma_{\mathrm{ext}}(\omega) = \frac{9\,\omega\,\varepsilon_m^{3/2}}{c}\,V\, \frac{\varepsilon_2(\omega)}{\big[\varepsilon_1(\omega)+2\varepsilon_m\big]^2+\varepsilon_2(\omega)^2},$$

with $V$ the particle volume and $\varepsilon = \varepsilon_1 + i\varepsilon_2$ the metal’s complex dielectric function [1][2]. Two facts follow directly: resonance position tracks $\varepsilon_1$ (local refractive index — the basis of ordinary refractometric biosensing), while resonance linewidth (FWHM) tracks $\varepsilon_2$, the metal’s intrinsic optical loss. A gas-sensing figure of merit falls out immediately:

$$\mathrm{FOM} = \frac{S}{\mathrm{FWHM}}, \qquad S \equiv \frac{\Delta\lambda}{\Delta c}$$

where $S$ is sensitivity — peak shift per unit concentration or pressure [3][4]. Because Pd/Pd-alloy sensing works by changing the dielectric function of the metal itself rather than a surrounding medium’s refractive index, $S$ and FWHM are coupled through the same material physics. Maximizing $S$ while minimizing FWHM, in a system where one lossy metal often has to supply both the sensing chemistry and the optical readout, is the central design problem this page keeps returning to.

Gold and silver support narrow LSPR peaks because their conduction electrons behave close to an ideal free-electron (Drude) gas, $\varepsilon(\omega) = \varepsilon_\infty - \omega_p^2/(\omega^2+i\gamma\omega)$, with small damping $\gamma$. Palladium does not: its d-band edge sits near the Fermi level, so strong interband ($d\to sp$) transitions overlap the visible/near-UV range and contribute a large, frequency-dependent $\varepsilon_2$ exactly where a plasmon resonance wants to sit.

Nanoparticulate Pd shows a broad extinction feature with its LSPR riding on an interband continuum rather than standing apart from it, which is why the field split into two architectures: use Pd directly and accept a broad, low-Q resonance, or separate the sensing chemistry (Pd) from the optical readout (Au or Al), which can then keep its narrow, low-loss resonance.

Direct and indirect sensing

When H2 dissociatively chemisorbs on a Pd (or Pd-alloy) nanoparticle and the resulting H atoms populate octahedral interstitial sites, two competing effects shift the LSPR: electron donation to/from the lattice shifts the plasma frequency $\omega_p=\sqrt{ne^2/\varepsilon_0 m^*}$ and hence $\varepsilon_1$, while the lattice physically expands (several percent linear strain at full hydride formation), changing interparticle spacing and near-field coupling.

These effects can push the resonance in opposite directions; the net result for small Pd nanoparticles is typically a small blue shift (~1 nm) on hydride formation. Either way the shift rides on Pd’s broad, lossy background, which is why FOM rather than raw peak shift is the metric that matters.

The indirect nanoplasmonic sensing (INPS) architecture, introduced by Langhammer, Larsson, Kasemo, and Zorić, sidesteps Pd’s poor optical quality entirely [5]. A 2-D array of Au (or Al) nanodisks — a good, narrow-linewidth antenna — sits under a thin dielectric spacer (commonly SiO2, 10–30 nm), on top of which the actual sensing material (a granular Pd film or discrete ~5 nm Pd “satellite” nanoparticles) is deposited.

The antenna’s near-field, which decays over tens of nanometers matching the spacer thickness, is highly sensitive to the satellites’ complex refractive index on hydride formation, so hydrogen is read out as a shift of the Au resonance, never the Pd resonance directly.

Follow-on work using PdAu satellites (~25 at % Au) is reported in the Chalmers indirect-sensing literature to give an 8-fold absolute low-pressure sensitivity enhancement and sub-second response across 0–40 mbar relative to pure-Pd satellites. The general platform has also been validated as an “ultrasensitive experimental platform for nanomaterials science and optical nanocalorimetry” beyond hydrogen [5]. Bimetallic Pd–Au nanoantennas — alloying inside a single scattering nanoantenna rather than pairing a separate Au antenna with Pd satellites — combine directional scattering with hydrogen sensitivity, a hybrid between the two philosophies [6].

Thermodynamics and kinetics

Pd-hydrogen thermodynamics follow the same pressure–composition–temperature framework as the resistive Pd sensor family (see Palladium resistive & nanomaterial sensors): a dilute α phase transitions, above a critical hydrogen content, into a concentrated, lattice-expanded β (hydride) phase, with a first-order boundary described by a van’t Hoff relation,

$$\ln p_{\mathrm{eq}} = \frac{\Delta H}{RT} - \frac{\Delta S}{R}.$$

For individual Pd nanocrystals, $\Delta H \approx -33\ \mathrm{kJ\,(mol\,H_2)^{-1}}$ and $\Delta S \approx -82\ \mathrm{J\,K^{-1}(mol\,H_2)^{-1}}$ [7]. Below the critical temperature, absorption and desorption plateau pressures differ — hysteresis — because each transformation must nucleate and propagate a new phase against a strain penalty, a gap that widens with particle size [7].

For a sensor this is a problem: the same optical signal can correspond to two different hydrogen concentrations depending on history, and eliminating it was a major goal of the 2015–2019 literature [8][9].

The standard fix — alloying Pd with ~25 at % Au — works because Au donates electron density into Pd’s d-band, altering the H–H and H–metal energetics that drive phase separation; at sufficient Au content the miscibility gap closes below room temperature, so absorption becomes a continuous solid solution and the optical response becomes linear and hysteresis-free [8][9]. This is bulk thermodynamic engineering rather than simply adding a catalyst, and it is the physical basis of every hysteresis-free variant below.

Once thermodynamics allows hydrogen in, response speed is set by dissociative chemisorption plus diffusion into the particle, which in a diffusion-limited regime scales as $\tau \sim L^2/D_{\mathrm H}$ for transport length $L$. This quadratic size dependence is why nearly every high-performance device here is a nanoparticle, nanodisk, or ultrathin film: shrinking $L$ from micrometers to nanometers can move $\tau$ from minutes to milliseconds, credited directly for sub-second response in alloy nanoparticles with “optimized volume-to-surface ratio” [9][8].

It also explains why the record-LOD metasurface below, which deliberately uses a thick (300 nm) polymer membrane, pays for sensitivity with a ~40-minute response [3]; sensitivity, speed, and selectivity compete for the same design budget throughout this field.

Variants & subtypes

Direct Pd/Pd-alloy nanoparticle LSPR sensors. The baseline architecture: nanodisks, nanoparticles, or nanoholes of Pd (or a Pd alloy) whose own LSPR peak is tracked. The original demonstration used Pd nanodisks made by hole-mask colloidal lithography, establishing the basic remote, real-time metal-hydride readout scheme [10]. This family carries pure Pd’s full set of thermodynamic drawbacks — hysteresis, slow kinetics, CO/NOx/sulfur deactivation — unless alloyed or coated [11].

Indirect (nanoantenna + satellite) sensors. As above: a high-Q Au or Al nanoantenna array under a dielectric spacer, decorated with Pd satellite nanoparticles that do the hydrogen chemistry, decoupling antenna optical quality from transducer chemistry [5][6]. The tradeoff is an extra fabrication layer and near-field sensitivity that falls off with spacer thickness.

Alloy engineering. A major design lever in this field:

• PdAu (~25 at % Au) is the hysteresis-suppression workhorse and the transducer in the metal-polymer hybrid sensor below [8][9]. A direct thin-film-vs-nanoparticle comparison found the two morphologies respond differently, confirming geometry is an independent design variable from composition [12].
• PdCu alloying suppresses CO poisoning even at high concentration — Pd95Cu5 resists deactivation to 5000 ppm CO [11] — trading some hysteresis benefit for chemical robustness.
• PdAuCu ternary alloys combine both. Pd65Au25Cu10 is reported as the optimum balance of hysteresis-free, sensitive response with deactivation resistance to 500 ppm CO and good long-term stability [13][14], reflecting an explicit “trade-off between the sensitivity enhancement and the CO-poisoning resistance effects provided by Au and Cu alloyants, respectively” [14].
• Ta–Pd, a non-coinage-metal alloy, gives hysteresis-free sensing over seven orders of magnitude in pressure with sub-second response [15], showing that Au/Cu are not the only route to closing the miscibility gap.

Metal–polymer hybrid nanomaterials. Nugroho et al.’s 2019 result pairs a PdAu nanoparticle transducer with a tailored polymer overlayer, reporting 3 ppm H2 detection with ~1 s response, hysteresis suppression, and operation in “demanding chemical environments” without long-term deactivation [9]. The mechanism is molecular sieving: the polymer lowers the apparent activation energy for hydrogen transport while a tandem membrane blocks poisons [9].

The same philosophy — PMMA and related overlayers as size-selective membranes passing small H2 while blocking larger interferents or liquid water — recurs in a PMMA-coated Pd/graphene hybrid built for selectivity [16] and in the inverse-designed metasurface below [3].

Metasurfaces and inverse design. A periodic 2-D array of plasmonic nanoparticles (typically Pd or Pd-alloy nanodisks) can produce a surface lattice resonance (SLR) — narrower than a single particle’s LSPR — when array pitch is tuned near the Rayleigh anomaly condition.

Because FOM depends on both sensitivity and lattice-narrowed linewidth, current state of the art uses computational inverse design: Nugroho, Baldi, and co-workers used particle-swarm optimization with FDTD simulation to find an optimal Pd-nanodisk metasurface (124 nm diameter, 20 nm height, 376 nm pitch, 300 nm PMMA coating) reaching a measured 250 parts per billion detection limit — the lowest reported for an optical hydrogen sensor at publication — at the cost of a ~40-minute response set by diffusion through the thick membrane [3]. A related paper generalizes inverse design to alloy composition rather than geometry [17].

Perfect absorbers and metal–dielectric–metal (MDM) metamaterials. A metal–dielectric–metal perfect-absorber cavity — Pd top layer, dielectric spacer, back-reflector — is engineered so nearly all incident light is absorbed at a cavity resonance, with losses deliberately localized in an ultrathin Pd layer. Hydride formation shifts the minimum-reflectance wavelength; one design reports a ~60 nm shift at 4 vol % H2 with easy, large-area fabrication [18]. A related nanostructured perfect absorber operates in the visible range specifically for hydrogen sensing [19].

Chiral plasmonic sensors. Three-dimensional chiral Pd(-Au) nanohelices, made by glancing-angle deposition, show circular dichroism that shifts measurably on hydrogen uptake — an additional readout channel for rejecting non-chiral optical artifacts. Au alloying reduces hysteresis and “significantly improv[es] the linearity of the circular dichroism” [20]. This remains niche and fabrication-intensive, with no demonstration near the ppb–ppm sensitivity of the leading planar designs.

Hot-electron / on-chip photonic variants. A newer direction integrates plasmonic transduction with silicon photonics via plasmon-induced hot-electron–molecule interactions for “ultrasensitive and rapid” on-chip sensing [21], with related SOI-waveguide proposals pointing toward CMOS-photonic co-integration, though none exist beyond lab chip demonstrations.

Adjacent optical-Pd variants (context, not core nanoplasmonics). Two related but mechanistically distinct technologies should not be conflated with true LSPR/metasurface sensing: propagating SPR/thin-film sensors, e.g. a Pd-coated fiber “ball lens” reading reflectivity change (not a resonance shift) from a 40 nm film, with a 10 ppm (1σ) detection limit over 0–1% H2 [22]; and Pd-film colorimetric sensors offering naked-eye reflectance change [23]. Both are optically read and Pd-hydride-based but rely on no nanostructured resonance, making them closer to the resistive/mechanical family (see Palladium resistive & nanomaterial sensors) than to this page’s focus.

Nanofabrication.

• Hole-mask colloidal lithography (HCL) — self-assembled colloidal spheres as an evaporation mask — produces wafer-scale nanodisk arrays at low cost [24][25].
• Electron-beam lithography gives precise control needed for periodic SLR metasurfaces but is “very costly and not really suited for mass fabrication,” typically limited to ~100×100 µm² areas [25].
• Colloidal wet-chemical synthesis of PdAu/PdAuCu nanoparticles trades vacuum deposition for better composition/size control and scalability [26].
• Glancing-angle deposition builds the chiral nanohelices above.

Readout simplification. Beyond full spectrometer-based spectroscopy:

• single-wavelength LED-plus-photodiode readout, framed as “purely optical” and low-cost [27]
• naked-eye/colorimetric readout via hydrogen-induced nanoparticle aggregation [28]
• machine-learning-accelerated readout, where a neural network or transformer trained on the raw transient signal extracts concentration faster than a classical fit — pushing humid-air LOD to 100 ppm [29] and accelerating response up to 40× while removing an intrinsic pressure dependence in inert-gas environments [30]

No dedicated hydrogen-specific smartphone-camera demonstration was found, though the same simplification logic exists for other gases.

Performance envelope

Figures below are as reported in the cited source; test conditions (carrier gas, temperature, pressure vs. concentration units) vary between papers and are not directly comparable without normalizing for those conditions.

Variant / device	LOD	Range tested	Response	Recovery	Notes	Ref
Pure Pd nanodisk, direct LSPR (original scheme)	not quantified as ppm LOD	qualitative isotherms across α/β transition	seconds-scale (qualitative)	—	First nanoplasmonic H2 sensing demonstration; hysteresis present	[10]
PdAu alloy nanoparticle (hysteresis-free)	—	1 mbar–1 bar H2 pressure (~3 orders of magnitude)	sub-second	—	Hysteresis eliminated at ~25 at % Au	[8]
PdAu/polymer metal-polymer hybrid	3 ppm	demonstrated with poisoning-gas exposure	~1 s	—	No long-term deactivation observed; hysteresis suppressed	[9]
PdAuCu ternary (Pd65Au25Cu10)	low ppm	5 orders of magnitude H2 pressure	sub-second (RT)	—	Deactivation-resistant to 500 ppm CO; long-term stable	[13][14]
Ta–Pd alloy	—	7 orders of magnitude in pressure	sub-second	—	Hysteresis-free	[15]
Inverse-designed Pd-nanodisk metasurface (PMMA-coated)	250 ppb	low-ppb to ppm regime	~40 min (diffusion-limited by 300 nm PMMA)	—	FOM 0.11; lowest optical H2 LOD reported at publication; needs added PVA layer to block O2 interference	[3]
Pd70Au30 + neural-network readout	100 ppm	tested to 80% RH	not separately quoted	—	Operated 80–130 °C; passes ISO 26142:2010 stability at 0.06% H2, 80% RH; no loss after 142 h continuous operation	[29]
Plasmonic + transformer (LEMAS) readout	outperforms classical LOD at ppm-scale leak rates	≤0.1 vol % H2, inert-gas background	1.6–3.6 s vs. 50–85 s classical (21–40× faster)	—	Targets N2/Ar-encapsulated installations; removes intrinsic pressure dependence	[30]
Catalytic-plasmonic Pt nanoparticle sensor	30–50 ppm	0–80% RH	fast (“rapid,” t90 not quoted)	—	Operated 80–100 °C; response increases with humidity above a threshold — opposite of typical Pd degradation; stable 143 h at 80% RH	[31]
Chiral Pd/PdAu nanohelices (CD readout)	not quantified as ppm LOD	qualitative concentration dependence	not separately quoted	—	Hysteresis reduced with Au alloying; CD linearity improved	[20]
MDM Pd perfect absorber	—	up to 4 vol % H2	not separately quoted	—	~60 nm cavity-minimum shift at 4 vol % H2; near-total absorption at resonance	[18]
Fiber-tip Pd ball-lens SPR (adjacent, thin-film)	10 ppm (1σ)	0–1% H2 in air	not separately quoted	—	40 nm Pd film on 350 µm fiber-tip ball lens; reflectivity-based, not LSPR	[22]
General Pd-nanostructure device family (cross-technology review)	0.02 ppm – 5000 ppm across surveyed devices	mostly 0.1–1 vol % H2 test conditions	0.068–60 s (t90)	1.5–570 s (t10)	Illustrates the field’s spread, not a single device	[11]
Insplorion NPS-P2 (commercial product)	not a trace-ppm product	0.4–4 vol % H2	t90 < 5 s	not specified	Requires 1–5 vol % O2 carrier gas; accuracy <0.2 vol % or ±25% of reading; ATEX zone 1 certification in progress	[32][33]
2Witech ARPA-E target (in development)	target 20 ppb	mobile, “varied environmental backgrounds”	not yet reported	not yet reported	Federally funded target, not yet a demonstrated result	[34]

Cost, power, drift. HCL chips are positioned as “low-cost and large-area” relative to EBL, which is “very costly and not really suited for mass fabrication” [24][25], but no per-chip dollar figure is published. LED-plus-photodiode readout removes the need for a benchtop spectrometer [27], implying sub-$100-class optoelectronics at the detection-principle level.

A certified instrument is a different cost class: Insplorion’s NPS-P2 is a 24 VDC, IP64, ATEX-track industrial instrument with no published unit price [32]; one order for its separate research-instrument line totaled roughly 750 kSEK including accessories, not a per-unit list price [35].

The longest continuous-operation windows found anywhere in this family are 142 h (Pd70Au30 + neural network, 80% RH, no loss) [29] and 143 h (catalytic-plasmonic Pt, 80% RH) [31]; no source reports multi-month or multi-year field drift data for any plasmonic H2 sensor.

Figures of merit

The universal figures of merit below are defined in full, with worked arithmetic, on the Figures of merit page; this modality also carries a literal, named figure of merit — $\mathrm{FOM}=S/\mathrm{FWHM}$ — that falls directly out of the plasmon lineshape itself (see Mechanism & physics above). Concentration values below follow the conventions on Concentration units.

Figures-of-merit table for this modality (15 rows)

Figure of merit	Symbol / formula	What it means here	Typical value for this technology	How to read it
Sensitivity ($S$)	$S \equiv \Delta\lambda/\Delta c$	How far the resonance wavelength moves per unit change in H2 concentration or pressure — the raw slope before normalizing for linewidth	Not reported as one absolute number across sources; the indirect Au-antenna + PdAu-satellite architecture gives an ~8× enhancement over pure-Pd satellites at low pressure, 0–40 mbar [5]; isolated small Pd nanoparticles show only a small net blue shift (~1 nm) at hydride formation because electron-donation and lattice-strain effects partly cancel	Raw shift alone is not decisive: a broad, lossy Pd resonance can post a large $S$ and still lose to a narrow Au-antenna design with smaller $S$; see FOM below
Refractive-index sensitivity (nm/RIU)	$S_{RIU} \equiv \Delta\lambda/\Delta n$	The classic biosensing metric for peak shift per unit change in surrounding refractive index (RIU)	Not the operative mechanism here and not reported in this unit by any source: Pd/Pd-alloy H2 sensing works by changing the metal’s own dielectric function ($\varepsilon_1$) as hydride forms, not a surrounding medium’s bulk index [1][2]	Do not import nm/RIU numbers from biosensing papers to estimate this modality’s H2 sensitivity; read $\Delta\lambda/\Delta c$ (row above) instead
Literal figure of merit, $\mathrm{FOM}=S/\mathrm{FWHM}$	$\mathrm{FOM} = S/\mathrm{FWHM}$	This modality’s namesake metric — sensitivity normalized by resonance linewidth, so a narrow but weakly-shifting resonance can beat a broad but strongly-shifting one	0.11 for the record inverse-designed Pd-nanodisk metasurface (124 nm disks, 300 nm PMMA coating, 250 ppb LOD) [3]; no other source quotes a numeric FOM	Higher is better; because $S$ and FWHM both derive from the same lossy Pd dielectric function, pushing one up tends to push the other up too, so FOM rather than raw shift is what separates competing designs
Resonance linewidth / Q-factor (FWHM)	FWHM (nm); $Q\sim\lambda_0/\mathrm{FWHM}$	How spectrally sharp the tracked resonance is — narrower means an easier-to-resolve shift and higher FOM for the same $S$	Isolated Pd-nanoparticle LSPR is broad (~650 nm FWHM typical for the uncoupled case); the surface-lattice-resonance metasurface narrows this to 296 nm via periodic-array coupling, the change that raises FOM to 0.11 [3] (this specific FWHM pair is not tabulated elsewhere on this page; treat as a dossier-sourced detail)	Pure Pd is a poor plasmonic metal (large $\varepsilon_2$ from interband transitions), so linewidth narrowing has to come from array/antenna engineering, not from the metal itself
Resonance shift per %H2 (or per mbar)	$\Delta\lambda/\Delta(\%\mathrm{H_2})$	A concentration-normalized version of raw sensitivity, useful for comparing designs tested at different H2 levels	~60 nm cavity-minimum shift at 4 vol % H2 for a Pd metal–dielectric–metal perfect absorber [18]; ~1 nm net blue shift at hydride formation for isolated small Pd nanoparticles, unquantified against concentration in the source	Not directly comparable across device classes: a cavity-resonance shift and an LSPR peak shift are different observables; treat each number as valid only within its own paper’s test conditions
Limit of detection (LOD)	$\mathrm{LOD}=k\,\sigma_{blank}/S$ (see Figures of merit)	The smallest H2 concentration reliably distinguished from zero — but only at whatever response/integration time the number was measured at	250 ppb at ~40 minute response (PMMA-membrane metasurface) [3] down to 3 ppm at ~1 s response (PdAu/polymer hybrid) [9]; a cross-technology review of Pd-nanostructure devices spans 0.02 ppm–5000 ppm [11]	Report the 250 ppb figure together with its 40-minute integration time; this pairing is a clear example of the LOD–response-time–power triangle
LOD in ppm·√s	$\mathrm{NEC}\propto\sigma(\tau)\sqrt{\tau}$	A noise-equivalent-concentration metric normalizing LOD by averaging time, standard for laser-absorption/photoacoustic sensors (see Enhanced Raman, IR absorption)	Not reported in this unit for any plasmonic/LSPR device surveyed (unverified) — the LSPR literature instead reports a single LOD at a stated response time rather than an Allan-deviation-derived floor	Do not assume a ppm·√s figure exists or can be trivially derived from the LOD/response-time pairs above; no published noise curve for a plasmonic H2 sensor was found
Selectivity / cross-sensitivity	$K_{ij}=S_j/S_i$	How much humidity, CO, O2, or H2S masquerade as an H2 signal	Humidity is the dominant, historically unresolved confound — response degrades above ~40% RH in most Pd-based designs [11]; two 2024–2025 heated (80–130 °C) designs reach 100 ppm LOD at 80% RH [29] or show response that increases with humidity [31]. CO is suppressed by Cu/Au alloying to 500–5000 ppm depending on composition [11][13][14]; the record-LOD metasurface needed an added barrier layer to block O2 interference [3]	A lab LOD measured in dry N2 is not the field-effective LOD; re-derive against the humidity and CO/H2S background actually expected on site before trusting a headline ppb number
Response time	$t_{90}=2.303\,\tau$	How fast the reading catches up after a real concentration step	Sub-second for alloy-nanoparticle architectures (PdAu [8], PdAuCu [13][14], Ta–Pd [15], metal-polymer hybrid ~1 s [9]); ~40 minutes for the record 250 ppb metasurface, diffusion-limited by its 300 nm PMMA membrane [3]; a cross-technology review spans 0.068–60 s $t_{90}$ [11]	Fast and ultra-low-LOD have not been demonstrated in the same device; expect to pick one per the LOD–response-time–power triangle
Recovery time	family-specific, desorption-governed	How fast the reading falls back to baseline once H2 is removed — usually slower than response and less often reported	Not reported separately for any single device in this dossier; the only figure available is a cross-technology survey range of 1.5–570 s ($t_{10}$) across mixed Pd-nanostructure devices [11] (unverified for any specific plasmonic design)	When a recovery time is not quoted, assume it is the throughput bottleneck until measured otherwise
Hysteresis (PdAu)	branch-to-branch shift at fixed $p_{H_2}$	Whether the same optical signal means two different H2 concentrations depending on absorption/desorption history	Present in unalloyed Pd and severe enough to rule it out for quantitative sensing (original Pd-nanodisk scheme) [10]; eliminated by alloying ~25 at % Au, which closes the α–β miscibility gap below room temperature [8][9]; Ta–Pd achieves the same hysteresis-free behavior via a non-coinage-metal route over 7 orders of magnitude in pressure [15]	Treat any device description as incomplete until it states whether it is pure Pd (assume hysteresis) or an alloy engineered to close the miscibility gap
Dynamic range	$N_{\text{decades}}=\log_{10}(C_{\max}/C_{\min})$	How wide a concentration span the device covers with one calibration	3 orders of magnitude in H2 pressure for PdAu (1 mbar–1 bar) [8], up to 5 orders for PdAuCu [13][14] and 7 orders for Ta–Pd [15] in lab demonstrations; the one commercial product, Insplorion NPS-P2, covers only 0.4–4 vol % H2 — roughly one order of magnitude, and the wrong end of Hestia’s target window [32]	Lab dynamic range and the commercially available range are not the same thing: only the lab alloys cover ppm-to-percent, while the shipping product covers percent-level safety thresholds only
Drift / long-term stability	$R_0(t),\,S(t)$ vs. time	Whether the baseline or the calibration slope creeps over weeks to years of deployment	Longest continuous-operation windows found anywhere in this family are 142 h (Pd70Au30 + neural network, 80% RH, no signal loss) [29] and 143 h (catalytic-plasmonic Pt, 80% RH) [31] — under six days each; no source reports multi-month or multi-year field drift data for any plasmonic H2 sensor (unverified beyond ~6 days)	Treat any deployment-duration claim beyond about a week as unsupported by the current literature
Power	heater + optical source + readout electronics	What it costs to run the light source, any heater, and the readout electronics	Not quantified at the sensing-element level in any source; LED-plus-photodiode readout is framed as removing the need for a benchtop spectrometer, implying sub-$100-class optoelectronics at the detection-principle level [27]; the one certified commercial instrument (Insplorion NPS-P2) runs from a 24 VDC bench supply with no separately published power draw [32] (unverified)	A “spark-free, zero-downhole-electronics” pitch is about the sensing element only; the surface-side light source, spectrometer or photodiode, and any heater still draw real, unpublished power
Operating envelope	storage ⊃ operating ⊃ compensated	The temperature/pressure window where the device survives, produces output, and where stated accuracy actually holds	No source tests any nanoplasmonic or metasurface device at borehole-representative high-temperature/high-pressure conditions; two 2024–2025 designs deliberately run heated at 80–130 °C to buy humidity tolerance [29][31], trading passive ambient-temperature operation for an active heater power budget	HTHP survivability of the nanostructures and their polymer coatings is entirely untested (unverified); treat any downhole application as TRL 2–3 regardless of ambient-lab performance

What dominates

Two figures of merit decide whether this technology is usable for a geologic-H2 job: the LOD–response-time pairing, since the best LOD (250 ppb) only comes with a ~40-minute read while fast (sub-second) devices sit at ppm-level LOD, which is a direct instance of the LOD–response-time–power triangle rather than a solvable engineering gap; and humidity cross-sensitivity, historically a major unresolved confound and only now being engineered around by trading passive room-temperature operation for an actively heated (80–130 °C) design.

Hysteresis is a solved problem (alloy with ~25 at % Au, or use Ta–Pd) and raw sensitivity is not the bottleneck, since the physics comfortably covers Hestia’s ppm-to-percent window, but no device has demonstrated low LOD, fast response, humidity tolerance, and HTHP survival simultaneously in one package, which is why this modality sits at TRL 2–3 for field deployment despite TRL 6–7 lab/commercial maturity for its narrow, percent-level safety niche.

Environmental covariance & cross-sensitivity

Temperature. Because the α–β plateau pressure is governed by the van’t Hoff relation above, raising temperature accelerates diffusion-limited kinetics. Two 2024–2025 papers exploit this deliberately: the neural-network Pd70Au30 sensor runs at 80–130 °C to buy humidity tolerance [29], and the catalytic-plasmonic Pt sensor runs at 80–100 °C for the same reason [31], both trading ambient-temperature simplicity for actively heated operation. Alloying is itself a temperature lever: sufficient Au content pushes the α–β critical point below room temperature, the mechanistic reason 25 at % Au removes hysteresis at ambient conditions [8][9].

Relative humidity. Historically one of the field’s main unresolved cross-sensitivities. A comprehensive review states plainly that “none of the fast [Pd-based] sensors have ever achieved acceptable humidity resistance in up to 95% RH, as required by the US DoE,” with response degraded beyond roughly 40% RH from competitive water adsorption blocking Pd sites [11].

Two very recent results attack this directly: the neural-network Pd70Au30 sensor reaches 100 ppm LOD at 80% RH, passing ISO 26142:2010 [29]; and the catalytic-plasmonic Pt sensor’s response increases with humidity above a threshold concentration — “the opposite of other hydrogen sensing technologies that deteriorate in high humidity” [31]. Both are 2024–2025 and neither fix has reached a commercial product yet.

Oxygen. Unlike catalytic-bead sensors (see Catalytic pellistors), Pd hydrogen chemistry does not consume or require O2. It still matters operationally: the record-LOD metasurface needed an added PVA barrier layer “to block oxygen interference” to reach 250 ppb [3], and Insplorion’s NPS-P2 specifies a required “1–5 vol % O2” carrier gas [32] — validated for inerted, not fully anoxic, backgrounds. The ARPA-E transformer/LEMAS work is explicit that plasmonic sensors carry an “intrinsic pressure dependence” in inert-gas environments removed by machine-learning post-processing rather than by the sensor physics itself [30]; this is a real confound currently patched in software.

CO. A classical Pd poison competing for dissociation sites. Cu alloying suppresses CO poisoning “even at 5000 ppm” for Pd95Cu5 [11], and the PdAuCu ternary is engineered and validated for deactivation resistance to 500 ppm CO with good long-term stability [13][14].

H2S / sulfur. Directly relevant given H2S co-evolves with natural hydrogen seeps. Nanoscale analysis of sulfur poisoning on single Pd nanoparticles finds SOx preferentially adsorbs at the particle rim, slowing kinetics and lowering capacity; it also binds less strongly to the flat β-phase (hydride) surface than to Pd or the rough rim, so hydrogen exposure itself partially, selectively desorbs sulfur from the center while it persists at the rim [36], a path-dependent picture that a single-point calibration will miss.

CH4, CO2, He. No source reports a dedicated cross-sensitivity test against these gases. None dissociatively chemisorb on Pd at typical operating temperatures, so expected cross-response is low by the same mechanism argument used for the resistive Pd family — but this is an inference, not an independently confirmed measurement for the optical/plasmonic pathway, and should be flagged as such.

NOx is listed alongside CO and sulfur species as a Pd-poisoning gas in the comprehensive review, without device-specific optical/plasmonic figures [11].

Pressure. The field’s core measurement principle is literally a pressure–composition isotherm, so every thermodynamic and kinetic relation is pressure-dependent by construction. No source tests these sensors at elevated total ambient pressure (as opposed to varying H2 partial pressure at ~1 atm) — HTHP behavior of the nanostructures and their polymer coatings is an open question below.

Failure modes distinct from the resistive family. An optical readout introduces failure modes with no chemiresistor analog:

• condensation, dust, or biofouling on a window or fiber facet can attenuate the signal independent of any real chemistry
• misalignment or shock to a free-space path can introduce drift
• photobleaching of polymer membranes over long deployments is plausible but unquantified beyond the 142–143 h windows already noted [29][31]

The same hydride-cycling lattice strain that drives embrittlement in resistive Pd films applies identically here; the alloys that suppress hysteresis also mitigate this damage.

Use for geologic H2

Field verdict

The underlying physics is well matched to Hestia’s exploration window: lab-demonstrated sensitivity spans 250 ppb to low ppm with sub-second response achievable in some (not all, simultaneously) configurations, and the readout is intrinsically spark-free with a credible path to zero downhole electronics. But no field-ruggedized, dust/moisture-tolerant, ppm-capable instrument exists, no nanoplasmonic device has been tested at borehole-representative pressure or temperature, and the one company that shipped a commercial product on this physics is actively exiting the hydrogen-sensor business. Watch closely; do not specify this technology for a 2026 field campaign.

Application-specific fit

Soil-gas surveys. The lab sensitivity ceiling — 250 ppb [3]; 100 ppm at 80% RH with a practical protocol [29] — comfortably covers Hestia’s ~1 ppm-to-percent window and reaches below the ~0.53 ppm atmospheric background [37]. But no field-ruggedized, battery-powered, dust/condensation-tolerant plasmonic soil-gas instrument has been identified.

Current practice instead relies on portable electrochemical/pumped analyzers (e.g. GA5000-class instruments) [38] and dedicated in-soil stations such as MONHyTOR, non-plasmonic, built for month-scale unattended deployment [39]. Insplorion’s NPS-P2 is calibrated for the wrong end of the range — 0.4–4 vol %, an LEL/safety range — and requires a minimum O2 background [32] not guaranteed in a soil-gas headspace. The underlying physics fits the application, but the one fielded product does not.

Downhole / borehole logging. Here the “no downhole electronics” preference and nanoplasmonic sensing’s optical nature are structurally well matched in principle: a passive chip at a fiber tip, read from a surface source/detector, needs no downhole power, architecturally similar to the fiber-tip Pd ball-lens sensor already demonstrated (a thin-film SPR device, not true LSPR) [22].

That said, no source demonstrates a nanoplasmonic or metasurface sensor surviving, or even tested at, borehole-representative HTHP conditions, and the polymer coatings central to the best-performing variants are an unproven point of failure at elevated downhole temperatures where polymer glass-transition and permeability change substantially. Treat this as TRL 2–3 at best for the downhole application, despite the attractive architecture-level match. See Fiber-optic sensors for the broader “electronics at the surface only” philosophy this would need to fit into.

Continuous monitoring stations. The closest match to commercial reality: Insplorion’s NPS-P2 targets exactly this — storage systems, compressor rooms, fuel cell enclosures, pipe-in-pipe leak monitoring [32] — fixed-point, percent-level, safety-driven monitoring with sub-5-second response and pending ATEX zone 1 certification [32][33]. For a wellhead or facility safety backstop rather than trace exploration, this is the most mature entry point available, with the commercial caveat below.

Outlook

Show-stoppers

1. No dust/moisture-hardened, ppm-capable field instrument exists; record LODs come from lab-bench, spectrometer-read, temperature-controlled setups [3][29].
2. The best LOD (250 ppb) carries a ~40-minute response and needs an O2-blocking barrier [3]; it is not fast and small at the same time, and the fast (sub-second) results are ppm-level rather than ppb [8][9].
3. Commercial signal is weak and moving backward: Insplorion — the only vendor shipping a plasmonic hydrogen sensor product — reached a conditional agreement in February–March 2026 to sell its entire hydrogen-sensor business, including NPS-P2, IP, and patents, to Mann Teknik AB for SEK 5.5 million, citing the market’s “early maturity phase” [40][41].
4. HTHP survivability of the nanostructures and their polymer coatings is entirely untested.
5. CH4/CO2/He cross-sensitivity, while plausibly low by shared Pd-hydride mechanism, has not been independently confirmed for the optical/plasmonic pathway.

Where it could still win

The case for a future soil-probe or downhole-logging plasmonic sensor rests on three counts no other modality combines:

• intrinsic spark-free operation — only light needs to reach the sensing element [11]
• demonstrated sub-ppm-to-ppb sensitivity with sub-second response in at least some configurations [3][8][9]
• a natural fit for remote, multiplexed fiber readout with zero downhole electronics, mirroring Hestia’s own preference exactly

Against NASA’s TRL framework [42]:

Component	TRL
Most cited work	3–4
PdAu/PdAuCu material system as a whole	arguably 5
Insplorion NPS-P2 (percent-level safety application)	roughly 6–7 [32][33][43], a foothold its own vendor judged insufficiently sustainable to remain a standalone business

Vendors & products

Vendor	Product / program	Technology	Key specs	Status (mid-2026)	Ref
Insplorion AB (Gothenburg, Sweden; Chalmers spinout, founded 2010)	NPS-P2 hydrogen detector	NanoPlasmonic Sensing (NPS); Pd-alloy nanoparticle LSPR shift	0.4–4 vol % H2 range; t90 < 5 s; accuracy <0.2 vol % or ±25% of reading; 24 VDC, serial TTL / analog 4–20 mA; IP64; ATEX zone 1 in progress	Launched Q2 2025; multiple pilot orders and a HydraSpecma/MannTek collaboration; hydrogen-sensor business (NPS-P2, IP, patents) subject to a Feb–Mar 2026 conditional sale agreement to Mann Teknik AB for SEK 5.5M, pending shareholder approval	[32][33][40][41][43][44]
Insplorion AB	M8 8-Channel Analyser; S2 Flow; X1 (custom)	Same NPS platform, fiber-optic chips, applied to battery/electrochemistry R&D	Up to 8 simultaneous channels for operando battery sensing	Retained after the hydrogen-business divestment; core surviving line	[35][46]
Mann Teknik AB (MannTek) (Sweden)	(post-acquisition) hydrogen sensing in coupling/refueling hardware	Acquiring NPS-P2 IP/product line	30-year background in cryogenic/liquid-hydrogen coupling hardware	Deal conditional on an extraordinary general meeting; not yet closed	[41][45]
2Witech Solutions (Andover, MA, USA)	“Hybridized High-Q Plasmonic Hydrogen Sensor” (ARPA-E H2SENSE, $1,000,000)	Plasmonic metasurface with an H2-permeable self-cleaning coating	Target: first mobile, 20 ppb-level, self-cleaning system	Federally funded R&D; performance not yet published	[34]
Syzygy Plasmonics (Houston, TX; Rice University spinout)	Photothermal/plasmonic catalytic reactors	Plasmonic photocatalysis — not gas sensing	Not applicable to this page’s scope	Flagged for disambiguation only, given the shared “plasmonic” name	—
Academic groups (no independent spinout beyond Insplorion)	—	Chalmers (Langhammer): HCL, INPS, PdAu/PdAuCu, metal-polymer hybrids, ML readout. VU Amsterdam (Baldi): inverse-designed ppb metasurface. TU Delft (Bannenberg/Dam): Ta-Pd	—	Source of essentially all IP above; Insplorion is the only commercialization vehicle identified	[3][5][6][8][9][10][13][14][15][20][24]

Reading the landscape. As of mid-2026, exactly one company has shipped a hydrogen-sensing product built on nanoplasmonic LSPR physics, and it is exiting the hydrogen business entirely, selling the product line to a coupling-hardware manufacturer [40][41]. The most credible successor signal is ARPA-E H2SENSE seed funding for 2Witech’s more-sensitive (20 ppb) mobile device, not yet published [34].

Within that same $18M ARPA-E cohort, only one of nine funded projects uses plasmonic sensing at all — the rest use:

• photoacoustic/Raman spectroscopy
• non-plasmonic fiber-optic sensing
• laser backscatter imaging
• piezoelectric resonators
• tracer-gas/AI methods
• dielectric-excitation sensing [34]

This is a directly sourced signal that plasmonics is a minority bet even within a dedicated federal hydrogen-sensing program.

Open questions

1. No published per-unit price for any plasmonic H2 sensor product or chip — needed before a real cost-class comparison against electrochemical/resistive/pellistor alternatives can be made.
2. No multi-month or multi-year field drift/lifetime data exists; the longest continuous-operation windows found are 142–143 hours [29][31].
3. HTHP (downhole) survivability of nanostructured Pd/PdAu transducers and their polymer coatings is completely untested — the single biggest gap between the attractive “no downhole electronics” concept and a deployable product.
4. CH4/CO2/He cross-sensitivity is inferred, not measured, for the optical/plasmonic pathway and should not be presented to field teams as confirmed.
5. Post-acquisition fate of the NPS-P2 product line is unresolved; the Insplorion→MannTek transaction was still conditional on shareholder approval at last report [41] — revisit once the deal closes or fails.
6. Fano-resonant geometries for hydrogen sensing are flagged as an interesting variant but no dedicated Fano-resonance H2-sensing paper was located — treat as unverified.
7. Whether hole-mask colloidal lithography can hit a $/sensor cost point compatible with a distributed soil-gas survey grid, rather than a single lab chip, is an open manufacturing-economics question.
8. PdAuCu and Ta-Pd have not been cross-tested against H2S specifically — the rim-poisoning mechanism was studied on unalloyed Pd [36] — worth checking whether alloying that fixes CO poisoning and hysteresis also helps (or hurts) H2S resistance.

Related pages

• Why H2 is hard
• Geologic context
• Palladium resistive & nanomaterial sensors
• Work-function devices
• Fiber-optic sensors
• Enhanced Raman
• Catalytic pellistors
• Thermal conductivity & acoustic sensors

References

Full reference list (46 entries)

1. “Drift-corrected nanoplasmonic hydrogen sensing by polarization.” Nanoscale 7, 10963 (2015). DOI: 10.1039/C5NR01818H. https://pubs.rsc.org/en/content/articlehtml/2015/nr/c5nr01818h
2. “Propagating and localized surface plasmon resonance sensing — A critical comparison based on measurements and theory.” Sensing and Bio-Sensing Research. https://www.sciencedirect.com/science/article/pii/S2214180416300034
3. Nugroho, F. A. A.; Bai, P.; Darmadi, I.; Castellanos, G. W.; Fritzsche, J.; Langhammer, C.; Gómez Rivas, J.; Baldi, A. “Inverse designed plasmonic metasurface with parts per billion optical hydrogen detection.” Nature Communications 13, 5737 (2022). DOI: 10.1038/s41467-022-33466-8. https://www.nature.com/articles/s41467-022-33466-8
4. “Design Principles for Sensitivity Optimization in Plasmonic Hydrogen Sensors.” ACS Sensors (2020). DOI: 10.1021/acssensors.9b02436. https://pubs.acs.org/doi/abs/10.1021/acssensors.9b02436
5. Langhammer, C.; Larsson, E. M.; Kasemo, B.; Zorić, I. “Indirect Nanoplasmonic Sensing: Ultrasensitive Experimental Platform for Nanomaterials Science and Optical Nanocalorimetry.” Nano Letters 10(9), 3529–3538 (2010). DOI: 10.1021/nl101727b. https://pubs.acs.org/doi/abs/10.1021/nl101727b
6. Shegai, T.; Johansson, P.; Langhammer, C.; Käll, M. “Directional Scattering and Hydrogen Sensing by Bimetallic Pd–Au Nanoantennas.” Nano Letters 12(5), 2464–2469 (2012). DOI: 10.1021/nl300558h. https://pubs.acs.org/doi/10.1021/nl300558h
7. “Hydride formation thermodynamics and hysteresis in individual Pd nanocrystals.” https://livrepository.liverpool.ac.uk/2024759/1/Hydride%20formation%20thermodynamics%20and%20hysteresis%20in%20individual%20Pd%20nanocrystals%20with.pdf
8. Wadell, C.; Nugroho, F. A. A.; Lidström, E.; Iandolo, B.; Wagner, J. B.; Langhammer, C. “Hysteresis-Free Nanoplasmonic Pd–Au Alloy Hydrogen Sensors.” Nano Letters 15(5), 3563–3570 (2015). DOI: 10.1021/acs.nanolett.5b01053. https://pubs.acs.org/doi/abs/10.1021/acs.nanolett.5b01053
9. Nugroho, F. A. A.; Darmadi, I.; Cusinato, L.; Susarrey-Arce, A.; Schreuders, H.; Bannenberg, L. J.; Fanta, A. B. da Silva; Kadkhodazadeh, S.; Wagner, J. B.; Antosiewicz, T. J.; Hellman, A.; Zhdanov, V. P.; Dam, B.; Langhammer, C. “Metal–polymer hybrid nanomaterials for plasmonic ultrafast hydrogen detection.” Nature Materials 18, 489–495 (2019). DOI: 10.1038/s41563-019-0325-4. https://www.nature.com/articles/s41563-019-0325-4
10. Langhammer, C.; Yuan, Z.; Zorić, I.; Kasemo, B. “Hydrogen storage in Pd nanodisks characterized with a novel nanoplasmonic sensing scheme.” (2007). PubMed ID: 17850168. https://pubmed.ncbi.nlm.nih.gov/17850168/
11. “High-Performance Nanostructured Palladium-Based Hydrogen Sensors—Current Limitations and Strategies for Their Mitigation.” ACS Sensors (2020). DOI: 10.1021/acssensors.0c02019. https://pmc.ncbi.nlm.nih.gov/articles/PMC7735785/
12. “Direct Comparison of PdAu Alloy Thin Films and Nanoparticles upon Hydrogen Exposure.” ACS Applied Materials & Interfaces (2019). DOI: 10.1021/acsami.8b22455. https://pmc.ncbi.nlm.nih.gov/articles/PMC6498406/
13. “Rationally Designed PdAuCu Ternary Alloy Nanoparticles for Intrinsically Deactivation-Resistant Ultrafast Plasmonic Hydrogen Sensing.” ACS Sensors (2019). DOI: 10.1021/acssensors.9b00610. https://pubs.acs.org/doi/10.1021/acssensors.9b00610
14. “Optimization of the Composition of PdAuCu Ternary Alloy Nanoparticles for Plasmonic Hydrogen Sensing.” ACS Applied Nano Materials (2021). DOI: 10.1021/acsanm.1c01242. https://pubs.acs.org/doi/10.1021/acsanm.1c01242
15. Bannenberg, L. J. et al. “Tantalum–Palladium: Hysteresis-Free Optical Hydrogen Sensor Over 7 Orders of Magnitude in Pressure with Sub-Second Response.” Advanced Functional Materials (2021). DOI: 10.1002/adfm.202010483. https://advanced.onlinelibrary.wiley.com/doi/10.1002/adfm.202010483
16. “A Highly Sensitive Hydrogen Sensor with Gas Selectivity Using a PMMA Membrane-Coated Pd Nanoparticle/Single-Layer Graphene Hybrid.” ACS Applied Materials & Interfaces (2015). DOI: 10.1021/am5073645. https://pubs.acs.org/doi/abs/10.1021/am5073645
17. “Computational Design of Alloy Nanostructures for Optical Sensing of Hydrogen.” ACS Applied Nano Materials (2022). DOI: 10.1021/acsanm.2c01189. https://pubs.acs.org/doi/10.1021/acsanm.2c01189
18. “Hydrogen Sensing Using Thin-Film Perfect Light Absorber.” ACS Photonics (2019). DOI: 10.1021/acsphotonics.9b00764. https://pubs.acs.org/doi/10.1021/acsphotonics.9b00764
19. “Palladium-Based Plasmonic Perfect Absorber in the Visible Wavelength Range and Its Application to Hydrogen Sensing.” Nano Letters. DOI: 10.1021/nl202489g. https://pubs.acs.org/doi/full/10.1021/nl202489g
20. Matuschek, M. et al. “Chiral Plasmonic Hydrogen Sensors.” Small 14(7), 1702990 (2018). DOI: 10.1002/smll.201702990. https://onlinelibrary.wiley.com/doi/abs/10.1002/smll.201702990
21. “On-chip ultrasensitive and rapid hydrogen sensing based on plasmon-induced hot electron–molecule interaction.” Light: Science & Applications (2023). DOI: 10.1038/s41377-023-01123-4. https://www.nature.com/articles/s41377-023-01123-4
22. “An Optical Fiber Hydrogen Sensor Using a Palladium-Coated Ball Lens.” Journal of Lightwave Technology 33(12), 2535 (2015). https://opg.optica.org/jlt/abstract.cfm?uri=jlt-33-12-2535
23. “Pd films on soft substrates: a visual, high-contrast and low-cost optical hydrogen sensor.” Light: Science & Applications 7, 8 (2018). DOI: 10.1038/s41377-018-0114-x. https://www.nature.com/articles/s41377-018-0114-x
24. Fredriksson, H.; Alaverdyan, Y.; Dmitriev, A.; Langhammer, C.; Sutherland, D. S.; Zäch, M.; Kasemo, B. “Hole–Mask Colloidal Lithography.” Advanced Materials 19(23), 4297–4302 (2007). DOI: 10.1002/adma.200700680. https://advanced.onlinelibrary.wiley.com/doi/abs/10.1002/adma.200700680
25. “Hole-mask colloidal nanolithography combined with tilted-angle-rotation evaporation: A versatile method for fabrication of low-cost and large-area complex plasmonic nanostructures and metamaterials.” Beilstein Journal of Nanotechnology 5, 682–691 (2014). https://pmc.ncbi.nlm.nih.gov/articles/PMC4077317/
26. “Robust Colloidal Synthesis of Palladium–Gold Alloy Nanoparticles for Hydrogen Sensing.” ACS Applied Materials & Interfaces (2021). DOI: 10.1021/acsami.1c15315. https://pubs.acs.org/doi/10.1021/acsami.1c15315
27. “Low-Cost Hydrogen Sensor in the ppm Range with Purely Optical Readout.” ACS Sensors (2019). DOI: 10.1021/acssensors.9b02314. https://pubs.acs.org/doi/abs/10.1021/acssensors.9b02314
28. “Eye-readable plasmonic hydrogen sensor based on Au@Pd nanoparticle aggregation.” ACS Sensors (2022). DOI: 10.1021/acssensors.2c01471. https://pubs.acs.org/doi/10.1021/acssensors.2c01471
29. Tomeček, D.; Klein Moberg, H.; Nilsson, S.; Theodoridis, A.; Darmadi, I.; Midtvedt, D.; Volpe, G.; Andersson, O.; Langhammer, C. “Neural network enabled nanoplasmonic hydrogen sensors with 100 ppm limit of detection in humid air.” Nature Communications 15, 1208 (2024). DOI: 10.1038/s41467-024-45484-9. https://pmc.ncbi.nlm.nih.gov/articles/PMC10853499/
30. Martvall, V.; Klein Moberg, H.; et al. “Accelerating Plasmonic Hydrogen Sensors for Inert Gas Environments by Transformer-Based Deep Learning.” ACS Sensors (2025). DOI: 10.1021/acssensors.4c02616. https://pubs.acs.org/doi/full/10.1021/acssensors.4c02616
31. “A Catalytic-Plasmonic Pt Nanoparticle Sensor for Hydrogen Detection in High-Humidity Environments.” ACS Sensors (2025). DOI: 10.1021/acssensors.5c03166. https://pmc.ncbi.nlm.nih.gov/articles/PMC12670998/
32. Insplorion AB. “Hydrogen detection: improved safety & reliable applications” (NPS-P2 product page). https://www.insplorion.com/en/hydrogen-sensor-p2/
33. Insplorion AB. “Insplorion officially launches the Insplorion NPS-P2 hydrogen detector.” https://www.insplorion.com/en/news/insplorion-officially-launches-the-insplorion-nps-p2-hydrogen-detector/
34. ARPA-E. “H2SENSE Exploratory Topic — Project Descriptions” (2025). https://arpa-e.energy.gov/sites/default/files/2025-01/Project%20Descriptions_H2SENSE.pdf
35. Insplorion AB. “Instrument: Insplorion M8 8-Channel Analyser.” https://www.insplorion.com/en/products/insplorion-8-channel-analyzer/
36. “Nanoscale Analysis of Sulfur Poisoning Effects on Hydrogen Sorption in Single Pd Nanoparticles.” ACS Nano (2025). https://pmc.ncbi.nlm.nih.gov/articles/PMC12574217/
37. “Atmospheric composition and climate impacts of a future hydrogen economy” (preprint). Atmospheric Chemistry and Physics (Copernicus), cited for the ~0.53 ppmv present-day tropospheric H2 background. https://acp.copernicus.org/preprints/acp-2023-29/acp-2023-29.pdf
38. “Assessing the Portable Gas Analyzer GA5000 for Accurate H2 Monitoring in Geological Exploration.” https://www.researchgate.net/publication/385938959_Assessing_the_Portable_Gas_Analyzer_GA5000_for_Accurate_H_Monitoring_in_Geological_Exploration
39. “In-soil hydrogen concentration monitoring using MONHyTOR: application for natural hydrogen exploration.” Advances in Geosciences 67, 57–65 (2025). https://adgeo.copernicus.org/articles/67/57/2025/
40. Insplorion AB. “Insplorion has completed its strategic review and entered into an agreement in principle regarding the sale of its hydrogen sensor business to Mann Teknik AB for SEK 5.5 million” (Feb 2026). https://www.insplorion.com/en/news/insplorion-has-completed-its-strategic-review-and-entered-into-an-agreement-in-principle-regarding-the-sale-of-its-hydrogen-sensor-business-to-mann-teknik-ab-for-sek-5-5-million/
41. FuelCellsWorks. “Insplorion to Sell Hydrogen Sensor Business to Manntek” (March 2026). https://fuelcellsworks.com/2026/03/04/news/insplorion-to-sell-hydrogen-sensor-business-to-manntek
42. NASA. “Technology Readiness Levels Demystified.” https://www.nasa.gov/aeronautics/technology-readiness-levels-demystified/
43. FuelCellsWorks. “Insplorion Secures Three New Orders for NPS-P2 Hydrogen Sensors” (Oct 2024). https://fuelcellsworks.com/2024/10/21/clean-technology/insplorion-secures-three-new-orders-for-nps-p2-hydrogen-sensors-worth-180-ksek
44. Insplorion AB. “Quarterly report Q2 2025.” https://www.insplorion.com/en/news/quarterly-report-q2-2025/
45. FuelCellsWorks. “Insplorion and MannTek Sign LOI to Integrate Hydrogen Sensing Tech Into Refueling Nozzles” (Oct 2025). https://fuelcellsworks.com/2025/10/23/green-technology/insplorion-and-manntek-sign-loi-to-integrate-hydrogen-sensing-tech-into-refueling-nozzles
46. Insplorion AB. “About us: We are NanoPlasmonic Sensing.” https://www.insplorion.com/en/about-us/