Introduction

Achieving net-zero emissions requires an urgent transition from fossil fuel use. Hydrogen manufactured from coal and natural gas, for example, generated 2.4% of global CO2 emissions (900 Mt) in 2022. The need for hydrogen is hard to abate because of its use as a chemical feedstock for the production of ammonia, methanol and in metal refining1,2. Indeed, fertilizer produced from ammonia contributes to feeding half of the global population3. Low-carbon hydrogen production technologies are available, but their high cost has hampered deployment. For example, hydrogen production from the electrolysis of water uses energy from renewable sources (termed ‘green hydrogen’). Green hydrogen has low CO2 emissions (0.3 kg CO2 kg−1 H2) but the cost is high, up to US$2.5–6.5 kg−1 H2 (Fig. 1). These costs are more than double the fossil fuel-based hydrogen production technologies (‘black’ or ‘grey’ hydrogen; US$1.0–2.7 kg−1 H2), even with the additional cost of carbon capture and storage to off-set the emissions (‘blue hydrogen’; US$1.5–4.0 kg−1 H2)1,4,5 (Fig. 1). A low-carbon energy carrier, hydrogen can support the increased use of renewables and provide a suitable fuel alternative for other ‘hard-to-abate’ sectors across industry and transport. Demand for clean hydrogen could increase from 90 Mt in 2022 to 540 Mt in a net-zero world by 2050 (refs. 2,6).

Fig. 1: Comparison of commercial hydrogen with natural hydrogen.
figure 1

The cost of green hydrogen from renewable energy sources is expected to decrease by 2050 because of economy of scale and technology efficiencies. The cost and carbon footprint of natural hydrogen will be dependant on the production quality of the gas reservoir and hydrogen purity. The value of co-produced gases such as helium, found in some types of natural hydrogen system, are not included in these estimates1,4,5. The cost and carbon footprint of natural hydrogen would make it a highly competitive source of hydrogen. CCS, carbon capture and storage.

Full size image

In addition to industrial production, a substantial mass of hydrogen is generated naturally in the Earth’s crust through chemical and radioactive processes7 (Box 1). Chemical processes include oxidation of iron-bearing minerals via water–rock reactions in ultramafic rocks, and radioactive processes include radiolysis of water via the radioactive elements U, Th and K that are found in all rocks, and at higher concentrations in granites and some sediments. The Precambrian continental crust hydrogen generation rate alone is between 0.36 and 2.27 × 1011 moles H2 yr−1 (ref. 7). Over the past billion years, this production is enough to supply the energy equivalent to 170,000 years of present-day societal oil use. However, most of this hydrogen would have been consumed in the subsurface or escaped to the atmosphere. Establishing how much hydrogen has been preserved in the crust and is accessible for economic exploitation remains highly uncertain8.

Although the occurrence of natural hydrogen has been reported globally through the last half of the twentieth century7,9,10,11,12,13,14,15,16,17,18, the measurement of natural gases in different environments for hydrogen has not been routine. Nevertheless, miners throughout history have encountered explosive gas (now known to be hydrogen) in mafic and ultramafic rocks, which is the source of the eternal flames of the Chimaera ophiolite, Turkey18,19.The potential of natural hydrogen as an economic resource has been stimulated by reports of the Bourakebougou gas field in Mali in 2018, which produces hydrogen with >97% purity, although production volume data remain limited20,21. As such, natural hydrogen accumulations could be globally widespread, but the lack of consistent hydrogen sampling and measurement limits current understanding of their distribution and magnitude.

Given the minimal energy inputs required, the extraction of high-purity natural hydrogen from geological accumulations is expected to have a low-carbon footprint5, starting at ~0.4 kg CO2 kg−1 H2 and increasing with the presence of co-produced methane (for example, a gas field containing 75% H2 and 22.5% CH4 will emit 1.5 kg CO2 kg−1 H2; ref. 5). Although there are no peer-reviewed cost estimates for natural hydrogen production so far, stated values are often given as US$~0.5–1  kg−1 H2 (ref. 22), but are expected to vary depending on gas purity, flow rates, depth and the size of gas field. Therefore, if notable natural hydrogen accumulations were discovered, natural hydrogen could serve as a commercially competitive, low-carbon hydrogen source that would contribute to the energy transition.

In this Review, we summarize the current understanding of the processes that control natural hydrogen systems, including the principal natural H2 sources and their geological characteristics, the source residence time and concentrations, the mechanisms and timescales controlling primary and secondary migration, the processes of hydrogen entrapment and the preservation of hydrogen in geological structures. We focus on hydrogen generation in the continental crust and compare and contrast production from radiolytic sources with production from water–rock reactions. We do not consider subaerial volcanoes or associated active hydrothermal systems, which vent dilute H2 directly to the atmosphere. As the information on organic hydrogen sources and sinks remains nascent23,24,25,26, we focus on inorganic hydrogen in the crust. Finally, we provide an overview of the different terrane types that contain all of the key hydrogen system elements, and summarize the knowledge gaps that, if closed, might accelerate discovery of societally substantial hydrogen accumulations.

Relevant hydrogen sources

There are abundant observations of hydrogen in crustal fluids from a variety of geological settings7,10,11,12,13,14,15,16,17,27,28,29,30. Various continental-scale hydrogen generation rates have been compiled7,17. The hydrogen generation rate from the Precambrian continental crust is estimated to range between 0.36 and 2.27 × 1011 moles H2 yr−1 (ref. 7). On combustion with O2, 1.0 × 1011 moles H2 generates 286 × 1011 kJ or 7.9 × 109 kWh (7.9 TWh). This energy production is equivalent to that produced by 6.8 × 105 tonnes of oil, or 0.017% of the crude oil produced globally in 2022 (ref. 31). Even before taking into consideration the proportion of natural hydrogen generation that might be recoverable8, the estimated 0.017% value indicates that natural regeneration of hydrogen on an anthropogenic timescale is unlikely to be substantial . However, over the last 1 billion years, the Precambrian continental crust alone could have generated hydrogen from water–rock and radiolysis reactions with the equivalent energy of 170,000 years of 2022 societal oil use. Although most of this hydrogen is not recoverable, the continental settings and the key processes that enable accumulation and preservation of a portion of this inorganic hydrogen are discussed.

Water–rock reactions

Rocks containing Fe-rich olivine (fayalite) and orthopyroxene such as peridotite are favourable for H2 generation and associated with formation of magnetite32,33. However, the iron content alone does not control the H2 yield and the distribution of Fe(II) and Fe(III) across different mineral phases has an important role34. Formation of secondary minerals such as andradite garnet, cronstedtite and haematite enhance H2 yields34,35,36. By contrast, in clinopyroxene-rich ultramafics (pyroxinites), more iron is stabilized in reaction products such as Fe(II)-serpentine, brucite and chlorite, instead of promoting the formation of magnetite and H232,37. H2 yields predicted by thermodynamic models typically increase with temperature and peak at 300 °C, after which they decrease32,33,34,37 (Fig. 2). The influence of temperature on H2 yields from mafic rocks remains underexplored, as well as the influence of specific mineralogies and sediment types36,38,39,40. Although mafic rocks are rich in iron, modelled H2 yields are 10–100 times lower than that of peridotite, owing to the partitioning of iron into chlorite rather than magnetite34,41,42.

Fig. 2: Hydrogen generation in ultramafic rocks from water–rock reactions.
figure 2

a, Estimates of peak hydrogen generation in ultramafic rocks at ca 300 °C at different pressures and constant water to rock ratios (w:r) = 1, based on results from experiments and thermodynamical models (data from refs. 32,33,34,36). Note that ref. 36, which uses an expanded thermodynamic database and more complex reactant mineralogy, predicts higher yields at low temperatures. Generated hydrogen is not sensitive to pressure. b, Hydrogen generation could be substantially greater at lower water to rock ratios (w:r) appropriate for crustal systems (w:r ~0.01) and over a range of temperatures (data from ref. 32). Similar observations are also reported in ref. 34. The agreement across different laboratories and approaches indicates that hydrogen generation at high temperatures and high water:rock ratios are well constrained for ultramafic rocks, but more data is required for lower temperatures, low water:rock ratios and other rock systems.

Full size image

Estimates of H2 yields follow the assumption that iron oxidation and consequent H2 generation is accurately represented as a thermodynamic equilibrium. Laboratory and natural sample observations indicate that serpentinization processes are more complex, incorporating multiple reaction steps and kinetic factors40,43. Low-temperature oxidation of Fe(II)-containing minerals to Fe(III) minerals (such as ferrihydrite, goethite, lepidocrocite and haematite) also generates H235,40,44,45,46,47. Although quantifying these secondary reactions poses a substantial challenge, it implies that rocks with a more mafic composition, particularly in low-temperature settings9,35,40, could have a substantially greater capacity for H2 generation than previously recognized. Low-temperature oxidation reactions with minerals bearing a proportion of Fe(II) extends the potential for calculating hydrogen generation from mafic systems to sedimentary iron-rich formations. Banded iron formations preserve evidence for mineral oxidation from ferrous silicate to ferric iron oxi-hydroxides and from ferrous and ferric oxides (magnetite) to exclusively ferric oxides (maghemite, haematite and goethite), although reaction kinetics and efficiency of H2 generation remain to be determined48,49.

Identification of the controls on rates of reaction have similarly been focused so far on oceanic lithosphere settings such as mid-ocean ridges, passive margins or subduction zones43,50,51. In these settings, processes operating within open fracture systems promote grain boundary reactions propagated by the auto-fracturing caused by hydrated mineral volume change52,53,54,55.

Reducing conditions in fluids enhance H2 generation. As redox potential is often related to the water:rock (w:r) ratio, as the proportion of rock increases, the system transitions from an oxidant-abundant to oxidant-limited state34. Dissolved H2 concentrations increase exponentially with decreasing w:r ratios32,34 (Fig. 3), and are enhanced by strongly alkaline conditions56. Yet the commonly used thermodynamic databases are incomplete for fluid compositions associated with low w:r ratios57. Importantly, another knowledge gap exists because the lowest modelled w:r ratios in the literature are equivalent to 37% bulk porosity32,34. In the continental basement settings, crystalline rock porosity is ~1% (refs. 58,59). H2 reaction yields in the continental crust could be substantially higher (Fig. 2), despite lower reactions rates. Higher H2 reaction yields could counterbalance, to some extent, the observation that increased salinity can reduce hydrogen yield and rate51,60,61, and that less water volume is available for reaction.

Fig. 3: Water–rock and radiolytic hydrogen generation as a function of depth.
figure 3

a, The idealized crystalline basement water fracture and/or porosity as a function of depth is given by Φ  = 1.6e(−z/4.8) where Φ is porosity as a unit fraction and z is the depth from surface in metres58. b, The calculated hydrogen generation depth profile for a rock column of 1 km2 surface area and 15 km depth from water radiolysis (blue line) and water–rock reactions (yellow line) using the porosity profile from panel a. Radiolytic hydrogen production over 500 Ma is calculated following the method in refs. 12,13, using the average crustal concentrations of radioelements62, rock density of 2.7 g cm−3 and is produced in proportion to the water available in the pore space. Calculating hydrogen produced from water–rock reactions assumes a geothermal gradient of 25 °C km−1, and temperature-dependent H2 yields at water:rock (w:r) ratios of 0.2 for a harzburgite rock composition32. The w:r generated assumes total reaction of the one pore fluid water column (panel a) and reflects a combination of water availability and temperature-controlled generation efficiency. The water–rock reaction hydrogen volume does not take into account either when the hydrogen is generated or the potential for subsequent water recharge. Hydrogen generation rates for lower w:r ratios (for example, appropriate for basement rocks) are probably greater per unit volume of water than for higher w:r ratios (for example, that might correspond to mid-ocean ridge hydrothermal systems), but production will be capped by the total consumption of water by the rock.

Full size image

With this complexity, it is clear that accurate hydrogen generation rate and yield predictions require additional thermodynamic data to encompass a wider range of mineralogy and the low temperature, low porosity of continental settings. An approximation can, nonetheless, be estimated by calculating the hydrogen reaction yield at different depths in the crystalline basement for ultramafic rocks assuming water availability for reaction is controlled by the porosity (Fig. 3). Such calculations provide a reference using a fixed water–rock system with rock remaining unreacted. The addition of more water would increase hydrogen generation, illustrating how the reference system provides a conservative starting point for applying mineralogical and environment scaling to determine terrane hydrogen generation potential.

Radiolytic hydrogen generation

The primary sources of ionizing radioactivity (alpha (α), beta (β) and gamma (γ) radiation) in the subsurface are from the decay chains of uranium (238U and 235U), thorium (232Th) and potassium (40K). Radioelements are ubiquitous in the upper crust62, with higher concentrations in igneous and felsic metamorphic rocks and sedimentary systems where trace elements can be concentrated by secondary fluid migration and precipitation63,64. Radiolysis is a substantial source of H2 in Precambrian terranes7,12,13, U-enriched terranes65, deep-sea sediments38,66,67 and fractured basalts68.

Alpha particles are highly ionizing but can only penetrate up to 40 µm in fluids and minerals. Beta and gamma particles have relatively lower energy but longer travel distances69. The siting of target water molecules relative to the radioelements is therefore important for understanding radiolytic yields. Smaller grains or minerals70, and narrow fractures68 where water is in close proximity to the radioelements are more favourable for higher H2 yields71. Radiolysis can also generate H2 through the release of hydrogen atoms within the lattice of hydrated minerals72. Overall, higher bulk porosity and associated water budget in the crystalline basement is favourable for higher H2 yields64,65,73. Different sediment slurries have been shown to enhance the generation of H2 by increasing the efficiency of α and γ absorption, whereas minerals such as clays, silicates, sulfides and zeolites can efficiently react with and remove oxidizing products of the radiolysis, such as H2O2, which would otherwise react with the stabilized H238,74. The impact of fluid composition, the distribution of water and radioactive elements on micro to macro scales, as well as the role of various minerals in influencing travel distances of radioactive particles are research gaps limiting more accurate H2 generation estimates in different geological environments.

As α-particles stabilize to form 4He, there is a proportional relationship between 4He and radiolytic hydrogen generation and the pore-space environment7,65,73 (Fig. 3). Predicted radiolytic 4He:H2 ratios are observed in H2-rich gases when the full mass balance between 4He, H2 and CH4 is considered, assuming the methane to be derived from radiolytic hydrogen. Examples include deep mines across the Fennoscandian Shield, Canadian Shield and South Africa7,73, and a South African gas field where stable isotope evidence is used to show that the CH4 is derived from radiolytic hydrogen that has been taken up by methanogens65. The latter has an estimated reserve of 40–299 BCF (CH4) and a best-estimate prospective resource of 1,278 BCF (CH4)75. This system illustrates the potential for substantial H2 generation and accumulation by radiolysis alone65, but also the risk of H2 chemical reactivity and bioavailability.

The radiolytic hydrogen produced in the continental crust containing average U + Th + K concentrations with porosity decreasing as a function of depth over a 500 Ma period is compared with generation by water–rock reactions in rock with a similar porosity profile (Fig. 3). Doing so illustrates the importance of the time-integrated porosity distribution for radiolytic hydrogen generation with depth. Similar to the water–rock calculation, this approach provides a starting point for scaling and determining terrane hydrogen generation potential with appropriate U + Th + K and porosity history.

The mantle

Mantle-derived 3He is a powerful tracer of, and proxy for, mantle-derived volatiles in crustal fluids76,77,78. In all the crystalline sites globally for which large concentrations of hydrogen have been reported, including the Canadian Shield, the Fennoscandian Shield, the gold mines of South Africa and Yilgarn craton Australia, noble gas investigations have ruled out any substantial mantle-derived H2 contributions7,10,79,80,81,82,83,84.

Hydrogen speciation within Earth’s mantle is well understood. Oxygen fugacity is strongly controlled by pressure. At pressures <3 GPa (90 km depth in the mantle) the oxygen fugacity of the mantle results in molecular H2O and carbon dioxide dominating the major volcanic gases85,86. Concentration estimates in the upper mantle range from 250 to 800 ppm H2O, with the variance representing subduction recycling heterogeneity87,88,89,90,91. The deeper mantle (>200 km depth) has an equivalent H2O concentration range of 250–1,100 ppm (ref. 92), but because of the higher pressure and resulting lower oxygen fugacity, both molecular hydrogen and CH4 form the primary H species86,93,94,95. There is no evidence that these species are preserved on ascent to the near surface in ocean islands or mantle-plume volcanism.

Continental volcanic gas emissions reflect the same oxidation state as mid-ocean ridges across compositionally diverse eruptive styles (for example, carbonatites96). Continental gas compositions are independent of mantle origin (for example, mantle plumes, back arc volcanism or rifting), but can be overprinted by subduction related volatiles either directly or from accumulation in the deeper lithosphere97,98. The continental mantle-3He flux is well established and has a median value of 270 mol 3He yr−1. With a mantle 3He/H molar ratio of 1.14 × 10−10, mantle-3He will be accompanied by 2.4 × 1012 mol mantle H yr−1 (2.4 × 106 tonne H yr−1)78,97,99,100,101,102,103.

Although the absence of 3He can rule out a mantle hydrogen contribution, the presence of 3He does not provide information about the oxidation state of any associated hydrogen. Substantive H2 from the mantle is highly unlikely. The mantle hydrogen is delivered to the continental crust in its oxidized form as mineral hydroxyl or water, not molecular hydrogen. It is incumbent on any economic claims that include substantial mantle hydrogen as a source (Box 1) to provide quantitative observational evidence to support such speculation8,104.

Source rock release and accumulation

Water within rock pore spaces or fractures provides both the source of the molecular hydrogen and defines limits on the timing and amount of hydrogen that can be generated by water–rock and radiolytic reactions in crystalline rock. Rates of water–rock reactions can be varied, with both kinetics and the availability of water providing rate-limiting steps, and the latter is determined by the existing water-filled porosity and the recharge rate. Timescales range from thousands to millions of years for water–rock reactions in highly fractured rocks, but can extend to tens to hundreds of millions of years for water-limited water–rock reactions. The radiolytic source of hydrogen is determined by the time-integrated water-filled porosity, local U + Th + K concentrations and time. The timescales for substantive radiolytic hydrogen generation is tens to hundreds of millions of years, and similar to water-limited water–rock reactions. In this section, the distribution of water in the continental crust, the timescales of hydrogen production by different production mechanisms, and the conditions that might result in H2 release and gas-phase accumulation in accessible geological traps are discussed.

Water in the continental crust

The mass and distribution of liquid water in the crystalline continental crust below 2 km depth represents a large (30–45%) but under-investigated component of the global hydrosphere83,105,106,107,108,109. The Precambrian basement, about 72% of the continental crust by area, contains >8.5 million km3 liquid water83. This volume is ca. 40% of the water occupying the pore space and fracture porosity of the crust83,107.

The are few assessments of crystalline crust-fracture porosity59,74,80,83,110,111,112,113,114,115. An exponential expression originally derived for sedimentary basins of decreasing porosity with depth (Φ = 1.6e(−z/4.8) ref. 58), is consistent with the limited crystalline crust fracture-porosity. The porosity relationship provides a pragmatic reference for extrapolation to depths of 10–15 km, where the transition from brittle to ductile behaviour occurs. Application of the porosity relationship with depth to hydrogen generation for any specific terrane will encompass large variance owing to local rock properties and burial history64,105.

Porosity is assumed to be water filled and hence representative of the static mass of liquid water within a volume of rock at a point in time105. The permeability of the crystalline crust is a measure of the ability of water to flux and recharge through that volume. Like porosity, observation-based permeability estimates are few in deep-crystalline rocks116,117,118,119. One assessment was made by considering the water residence ages derived from radiogenic noble gas accumulations in fracture fluids59. Others have used the fluid flux required to replicate observed temperature distributions in hydrothermal systems116 or the time-integrated fluid flux required to precipitate observed mineralization117,118,119.

Although permeability estimates have been fitted, like porosity, to an exponential decrease with depth, the noble gas residence data between 1 and 3 km do not differentiate between an exponential decrease versus a constant value over this depth range59. At 2 km depth, the permeability range of the various studies covers nearly five orders of magnitude, from log permeability of −15 to −20 m2 (refs. 59,116,118,119). The range, in part, reflects the representative elemental volumes considered and techniques employed, as well as natural variance in different crystalline terrane105.

This permeability range also illustrates the complexity of the relationship between hydrogen resource accumulation and porosity permeability. Hydrogen generation potential increases with increased water-filled porosity of the source rock due to increased surfaces and water volumes available for reaction, whether water–rock or radiolytic (Fig. 2). In the case of water–rock reactions if the static porosity is low, a high permeability is required to replenish the reactive water supply. In contrast, low permeability is required to trap and build up the hydrogen generated. End-member examples are seen in surface-exposed ophiolites, which have relatively high permeability and water supply, but newer hydrogen is lost to the surface due to high hydrogen flux82. These contrast with deep fracture fluids in low-permeability systems, which have been isolated from the hydrogeological system, sometimes on Ma to Ga timescales79. The latter provides an environment in which to build up hydrogen from both water–rock and radiolysis, but in which heterogeneous and disconnected fracture porosity is not ideal for resource accumulation64,73. The porosity–permeability evolution is therefore key to both the timescale of the hydrogen generation and the ability of the source rock to retain the gas59,83,84,107,120. In the case of ophiolites, low-permeability trapping might occur internally or from trap configurations in overlying sedimentary rocks16. Hydrogen generated in crystalline basement systems requires transport and accumulation of the gas to reservoirs, such as overlying sedimentary basin formations121,122,123.

A sweet spot can be envisaged, different for each system. Too open (like many surface-exposed ophiolites), and there is too much potential for hydrogen loss both from transport and mixing, but also loss to microbial communities themselves stimulated by younger water ingress (for example, the Beatrix mine, South Africa124). Too ‘tight’ (for example, Kidd Creek, Canada79) and there might be substantial hydrogen, but the extremely low permeability will make detection and exploitation difficult. The facility and ability of a system to accumulate hydrogen must be met29.

Residence time in basement source rocks

The accumulation of radiogenic and inert noble gases such as 4He, 21Ne, 40Ar, 86Kr and 136Xe from U, Th and K decay can be used to determine the mean residence age of fluids. An assumption of efficient transfer of the radiogenic noble gases from the mineral lattice into the fluid space is usually made. Coherent radiogenic isotopic compositional variance from predicted generation ratios can identify whether there has been substantial diffusive gas loss or advection73,78,79,80,84,113,114,125,126,127,128,129. Although results are sensitive to mineral release efficiency, the German Deep Drilling Project found that between 81% and 96% of helium generated in the 384 Ma since rock formation had been released from the minerals down to 7 km depth130.

The Kidd Creek Observatory, Canadian Shield, provides an example of isolation, with hydrogen-rich fracture fluids sampled to 2.9 km below the surface showing mean residence ages on a 1–2 Ga timescale79,83. Fracture fluids from two mines in the Sudbury Impact Crater, Canadian Shield, sampled at 1.4 and 1.7 km below the surface, have mean residence ages of 0.5 Ga and 0.3 Ga, respectively, possibly reflecting regional fracturing of the system caused by the meteorite impact at 1.8 Ga and deeper penetration of younger fluids into the crust83,84. Neither Kidd Creek nor Sudbury samples show evidence of substantial diffusive gas loss or gain, the corollary being that any hydrogen generated in the rock in that time has remained proximal to its source.

The Witwatersrand Basin, South Africa, was also perturbed by a large impact event, at 2.02 Ga. Fluids down to 3 km depth, many hydrogen rich, show variance in age. Some fracture fluids are interpreted as water that recharged from the surface up to 100 Ma years ago113,114,126. The less saline waters show δ18O and δ2H on or near to the global mean water line, consistent with variable mixing of palaeometeoric waters but with little evidence for substantial diffusive loss or gain131. Fluids from the Moab Khotsong gold and uranium mine record heavy noble gas (Kr and Xe) mean residence ages of up to 1.2 Ga, but with extensive (ca. 80%) diffusive loss of the light noble gases and hydrogen84. The Witwatersrand Moab Khotsong fluids show that Ga-age crystalline basement fluids are not only found in the Canadian Precambrian Shield, but are probably global in occurrence.

Fluids recovered from the 2.5 km Outokumpu deep borehole from the Precambrian Fennoscandian Shield in Finland record fluid residence ages from 20–50 Ma (ref. 80). These match well a regional estimate of 40 Ma fluid residence-age derived from fitting a diffusion profile to 4He water concentrations in 34 boreholes with samples recovered between 200 and 1,000 m depth more regionally, and 100% 4He mineral release assumed132.

The fluids supplying major hydrogen seeps from surface-exposed ophiolites are relatively modern, consistent with their shallow hydrogeological setting (for example, Chimera, Turkey18; Samail, Oman82; Tablelands, Canada; and Cedars and Aqua de Ney, USA133). 4He groundwater ages available from the Samail ophiolite, Oman, place the oldest fluids in the 20–220 Ka range134. Evidence for hydrogen accumulation has also been developed for the Bulqizë ophiolite, Albania16, but no observations yet exist to place limits on the gas residence age.

It is tempting to rank entire shield environments in order of regional permeability and associated mean residence age of deep fluids. There are certainly events such as the Witwatersrand impact at 2.02 Ga, or Sudbury Basin impact at 1.8 Ga, which will have had a regional effect on the architecture of the regional basement rocks84. We can speculate how a terrane might develop more interconnected fracture networks and elevated permeability (for example, the Fennoscandian Shield), how fluid flow responds to regional tectonic stress-fracture and fault propagation11,59,105,120,135 or, for example, the detail of foliation-induced anisotropy and regional stress fracturing. Yet the number of case studies on only three continents remain few, and more observational constraints are needed.

Tracking secondary migration

Helium serves as a tool for understanding the regional release mechanisms and flux of hydrogen and other deeper crustal gases such as nitrogen123,125,136. Formed in crystalline systems through radioactive decay, helium (4He) is often co-produced with hydrogen from the same source rock region. In contrast, helium is not consumed by chemical or biological processing and provides a clear resolvable signal in near-surface crustal fluids78. With similar physical characteristics, such as solubility and diffusivity, helium provides a relevant proxy for understanding hydrogen source-rock retention, expulsion and transport efficiencies.

The theoretical steady-state crustal helium flux provides a useful reference and is predicated on the rate of release of helium to the surface being equivalent to the rate of whole-crustal radiogenic 4He generation. From average continental U and Th concentrations, the steady-state helium generation rate is 1.47 × 10−6 mol 4He m−2 yr−1 (2.81 × 1010 atoms 4He m−2 s−1)137,138. This value is indistinguishable from the observed global log-normal mean surface flux of 2.2 × 10−6 mol 4He m−2 yr−1 (4.18 × 1010 atoms 4He m−2 s−1)139 (Fig. 4). This includes 4He data from n = 271 Precambrian Shield lakes from Saskatchewan (n = 101) and Labrador (n = 170), which provide an average degassing flux across two shield environments, also indistinguishable from the steady-state reference139.

Fig. 4: Tectonic setting controls deep-crustal degassing and helium flux.
figure 4

Crustal helium flux versus mantle helium flux (data from ref. 139) can be used as a proxy for understanding the processes controlling regional basement hydrogen release to the near surface. A broad trend of increasing crustal 4He flux with mantle volatile flux measured in groundwater and lake samples is associated with rifting (‘tectonic strain’, shaded blue region) and associated with regional thermal perturbations. Helium flux measured from volcanic crater lakes and calderas shows additional focusing of the mantle ‘volcanic’ helium flux (shaded red region), two to three orders of magnitude greater than would otherwise be predicted from groundwater and lakes in tectonic regions. Data compiled by ref. 139, with the addition of data for Taiwan101; Yellowstone147; and East Marengo Basin (EMB), California142. The dashed line is the log-normal mean continental degassing flux (4.18 × 1010 4He atoms  m−2 s–1), which lies within uncertainty of the rate at which helium is generated in the continental crust (‘steady state’)139. The correlation between the average global flux and generation rate shows that continental crust loses its gas, on average, as fast as it is generated. Regions of high tectonic strain, shown by the presence of mantle helium, show degassing fluxes higher than ‘steady state’, probably caused by more efficient release of crustal gases accumulated during periods of quiescence. GAB, Great Artesian Basin, Australia; GHP, Great Hungarian Plain, Hungary.

Full size image

Globally the surface helium flux is too great to occur via diffusion alone from the deeper crust, and an advective and episodic transport mechanism is required125. This point is reflected in the high flux variance78,125,139,140,141,142. Low helium flux is observed in several stable tectonic settings, such as the Molasse Basin143,144 and the Great Lakes145. Higher fluxes are found correlated with rock deformation and seismicity146, high heat flow or volcanic caldera such as Yellowstone, crater lakes or intrusions139,147,148; and with basement faults (San Andreas Fault, California100; East Texas Basin149; Irpina Fault, Italy150; and Four Corner Region USA148,151) (Fig. 4). These systems are additional indications that the magnitude and migration pathways for hydrogen and helium produced within the crystalline continental crust, and the accumulation potential, can be predicted from the local geological context.

Gas-phase formation and trapping

A key component in assessing accumulation potential is the ability of the system to generate a free gas phase. It is not usually commercially feasible to recover fully dissolved gas owing to the large volumes of water that need to be produced. For a gas phase to form, the gas concentration in the groundwater must exceed the solubility limit or ‘bubble point’, in turn controlled by the groundwater temperature, pressure, salinity and mixture of gases123. Gas-phase formation can also concentrate the gas through buoyancy migration into geological trapping structures. Processes applied to hydrocarbon gas field integrity, such as reservoir porosity and the gas column height that can be sustained by the seal considering pore throat size, capillary threshold and gas buoyancy, then provide a critical assessment of resource target capacity152.

For hydrogen released from its source, the gas flux, timescale and system architecture become important in generating and preserving a gas phase. For example, in the Western Canadian Williston basin, a flux of N2 and 4He averaging steady state from the basement can generate a free gas phase in the lithology overlying the basement at ca. 150 Ma despite substantial loss (>94%) from diffusion through the sealing strata. It is evident from the models that it is the retardation of the diffusional gas flux by the whole sedimentary column that generates the observed helium-rich gas fields in this example, rather than consideration of a single sealing unit123.

To generalize, systems with high flux require the least competent sealing architecture (for example, Tanzanian East African Rift122 and Bourakebougou gas field, Mali20,21). In contrast, seals such as salt or evaporite lithologies are particularly effective for low-flux helium and hydrogen systems (for example, Paradox basin, USA153). Effective seals account for the hydrogen and helium gas field discoveries in the tectonically quiescent Amadeus Basin, Australia27,154. The identification of the trade-off between gas release rate from the basement, or flux, and the overlying quality of the sealing system enables the development of a range of exploration concepts, tailoring hydrogen generation rates and location in the crust to transport style (diffusion versus advection) and gas-phase accumulations in different geological terrane.

Type geological hydrogen terranes

Hydrogen originates and accumulates throughout the Earth’s continental crust in a variety of tectono-stratigraphic settings. The hydrogen system requires four conditions: hydrogen sources, accumulation within the source rocks or migration of hydrogen into a reservoir, a sealed trap, and preservation by isolation from microbial and reactive elements. There is a wide geological variety, both in time and space, of inorganic hydrogen systems (Fig. 5). Whereas the basic elements of the natural hydrogen systems have remained constant through time, the tectono-stratigraphic setting of those same systems has changed as the Earth has cooled and tectonic behaviour changed155,156,157,158. These hydrogen systems have a globally widespread distribution today (Fig. 6). In this section, the common hydrogen systems and their evolution through time are discussed in chronological order from the Archaean to the Phanerozoic (Fig. 5).

Fig. 5: Five potential source-to-sink natural hydrogen systems.
figure 5

a–j, Geological terrane settings where potential commercial natural hydrogen systems could be found (letters in circles): ophiolite water–rock source (a), with migration into overlying clastic reservoirs within rift structures (b); radiogenic granites with radiolysis source (c), trapped in salt basin structures (d); flood basalt and water–rock source (e), trapped in clastic sediments near salt and inversion structures (f); radiogenic granites with radiolysis source (g), trapped within fractured granite (h); and ultramafic water–rock source (i), trapped in overlying carbonate and clastic rocks and sill traps (j). The source rock and trap pairs illustrated across different terrane form only exemplar hydrogen systems, with different permutations providing more possibilities, especially at terrane boundaries. TTG, tonalite–trondhjemite–granodiorite.

Full size image
Fig. 6: The global distribution of hydrogen-potential terranes.
figure 6

Hydrogen-potential terranes (described in Fig. 5) are widely distributed across all continents. The extent of sedimentary basins (yellow), including those containing evaporitic facies (pink), are shown to highlight regions that could contain potential trap structures. Examples of each terrane type with known notable hydrogen accumulations are circled in black: Oman, Ophiolite Complex; Australia, Amadeus Basin (an alkaline granite complex); USA, mid-continental rift (a large igneous province); Mali, tonalite–trondhjemite–granodiorite (TTG) and/or greenstone belt. Distribution data are from ophiolites194, large igneous provinces195, basement terrane196, sedimentary basins197 and evaporite basins198. Most of the continents contain widespread coverage of the terrane types that might contain hydrogen systems, and have hydrogen accumulation potential. The regional details of generation, migration, trapping and preservation are required to determine whether commercial accumulations are viable within these terrane.

Full size image

Archaean greenstone belts and TTG batholiths

Archaean greenstone belts occupy large volumes of continental cratons. Frequently deformed, the ultramafic and mafic volcanic greenstone assemblages, interleaved with tonalite–trondhjemite–granodiorite (TTG) batholiths, again deformed to varying degrees, represent two very different but connected sources of hydrogen. The extrusive ultramafic and mafic rocks of the greenstone belts are susceptible to water–rock reactions (Fig. 5i). In contrast the potassium lean, grey TTG gneisses have sufficient proportions of U and Th to generate widespread and long-lived hydrogen by radiolysis (Fig. 5g). Together, both rock assemblages have been, and continue to be, important potential sources of natural hydrogen10,80,120,159.

Trapping configurations for hydrogen formed in these tectonic environments require either an overlying sedimentary basin cover (Fig. 5j) or ‘within basement’ trapping (Fig. 5g,h). The former occurs as sedimentary basins within cratons160 and, more commonly, sedimentary basins developed at cratonic margins. Upon cratons, sedimentary basins tend to be either a shallow cover or the reflection of deep-basin formation within a craton that can be considered a craton margin121,123,136.

Most craton margins are defined by sedimentary basins or deformed metasedimentary basins younger than the underlying craton. Where these margins are transitional, rather than abrupt edges, there is extensive scope for clastic and carbonate reservoirs and seal combinations (Fig. 5d,f). This scope is demonstrated in the Mali Bourakebougou hydrogen field on the West African craton20,152. In this case the preservation of a hydrogen charge or presence of an active charge are the key risk factors. In contrast to shallow systems, the potential for generating and trapping hydrogen within the greenstones or fractured TTG batholiths depends on the local porosity–permeability evolution. To extend the identification of hydrogen-rich fracture fluids in mining locations to explore for targets distal to those locations in the basement rocks represents one of the major technology challenges of hydrogen exploration in these settings.

LIPs

Large igneous provinces (LIPs) are large volumes (~106 km3) of dominantly mafic igneous rocks that occur periodically in the continental crust from the Mesoproterozoic onwards (Fig. 5). They were initially regarded solely as extrusive, continental flood basalts and intrusive gabbros161,162, but increasingly a silicic aspect has been recognized163. Such large volumes of mafic rock are a potential source of water–rock hydration and the creation of large volumes of hydrogen (Fig. 2a). Their origin is driven by sublithospheric processes such as mantle plumes and time-equivalent rifting164,165.

The interest in LIPs for hydrogen exploration is largely with continental flood basalts that offer a very large source of hydrogen (Fig. 5e) beneath overlying sedimentary rock formations that could act as reservoir, seal and trap (Fig. 5f). The pace and intensity of the water–rock reaction is controlled by the temperature and the rock surface area accessible to the water. It is feasible that fine-grained, low-permeability, sedimentary layers within the basaltic flows could also act as seals, but their exploration would be technologically challenging from both an imaging and drill testing perspective. The United States mid-continental rift with extensive basalt infill formed at 1.1 Ga has received much attention with three wells (called Scot#1, Heins#1 and Sue Duroche#2) showing high hydrogen, but with limited flow, on the proximal (65 km) Nemeha High structure166,167,168,169,170,171.

Continental alkaline granite terranes

These are a widespread feature of the continental crust that have developed since the palaeo-Proterozoic and reflect a process of hydrogen generation purely from crustal-derived radiogenic granites12,13,172. The connection between the radioactive decay of K, U and Th and the presence of water results in the radiolysis of water (Box 1). Of the many complexities arising from this process, the controls on the molecular generation of H2 is reasonably well understood.

The hydrogen produced by this process is substantial owing to the great time frame in which it has been active. Its residence time within fractured granites has proven to be large. The exploration challenge then is understanding the migration, trapping and preservation of such long-accumulated volumes of H2. The trapping could occur in sealed zones of fractured granites (Fig. 5c) or in overlying sedimentary basins (Fig. 5d) (for example, Mt Kitty hydrogen discovery well, Amadeus Basin, Australia27,173).

Continental margin ophiolite complexes

Ophiolites are a tectonic signature of collisional and convergent plate tectonics174 (Figs. 5 and 6). They comprise large slices of mafic and ultramafic rock suites originally of marine origin, emplaced upon, beneath and within continental crust (Fig. 5a). The subsequent hydration of olivine and pyroxene rich ultramafic and mafic rocks is well described as serpentinization, releasing large volumes of hydrogen (Box 1). The large volumes of associated mafic rock occurring as gabbro, dolerite dykes and basalts experience similar hydration and are also a source of hydrogen.

Large ophiolite complexes, mostly preserved from the last 500 Ma of Earth’s history, are distributed along continental and palaeo-continental margins such as the Tethyan-collisional belt and its eastwards extension into the Himalayas and the active margins of the Pacific Ocean. Exploration for the H2 generated from these belts requires extensive hydration and the association of a reservoir and trapping configuration. Ophiolites exposed at the surface are not obviously prospective, as fracturing typically creates somewhat more ‘open’ systems11,84,105. However, others are covered by sedimentary rocks or trap hydrogen internally to make these ophiolites a potentially relevant commercial source of hydrogen16 (Fig. 5b).

Hydrogen accumulation preservation

Preservation of a hydrogen reservoir falls into three categories: physical preservation of the gas within the trap, including trap competence over time; chemical sinks or dilution; and biological consumption. Although hydrogen is more mobile relative than, for example, methane, hydrogen diffusion, solubility and other physiochemical characteristics provide for well-constrained gas behaviour. Within traps in sedimentary rocks there is a trade-off between seal quality, flux, and gas residence or preservation. For higher hydrogen fluxes, approaches familiar to the hydrocarbon industry are appropriate to determine reservoir and seal competence in geologically contemporary (up to tens of million years ago) accumulations152.

Helium provides an analogue to assess hydrogen residence and trapping efficiency in higher-permeability sedimentary traps in low-flux settings. In these environments, diffusional loss, dissolution into the groundwater and risk of lateral flushing must be considered when gas-phase development can take tens to hundreds of million years ago123,136,153. Hydrogen residence within some source-rock trapping environments is possible on Ga timescales79,83,84. Within systems that have faster generation rates and rely on traps in sedimentary rocks, analogy with hydrocarbon trapping is useful where the risk of reservoir and trap failure from regional changes (for example, faulting, uplift, flushing and pore-space loss) increases with time. In these cases, as with petroleum systems, the newer the accumulation, the more likely hydrogen will be preserved175.

Reactive gases such as hydrogen are impacted by chemical and biological sinks in the source rocks and reservoirs, as well as the near surface159,176,177. Investigations into the nature of methane and hydrogen in crystalline basement settings have demonstrated that increasing microbiological activity in the subsurface creates a major increase in the ratio of biotic–abiotic methane, and a correlated decrease in the hydrogen accumulations that reflects the loss of hydrogen to subsurface metabolic activity11,124,176,178. These shifts are typically correlated with deeper penetration of younger groundwaters such as more hydrogeologically open systems11,74,105,131,179. Subsurface microbial communities, such as those which reduce sulfate and methanogens, consume hydrogen, ultimately converting it back to water or methane180,181,182.

The rates of such metabolic activity at depth remain a subject of active investigation, but are typically thought to be substantially lower than in the near-surface biosphere183,184,185. Where microbial activity is not limited, degradation of gas within a reservoir can be on a decadal timescale186,187. In the case of the Welkom, Virginia gas field, South Africa, microbial processing probably accounts for the loss of hydrogen to methane in the shallow (<1 km) methane, helium and nitrogen dominated system65. Yet at only slightly greater depths, there are many proximal mines that report both high fluid residence (up to 168 Ma) and substantial hydrogen in fracture fluids84,113,114,126. Neither depth nor age are the main controls, but rather the degree of penetration of younger palaeometeoric fluids114,179 or, conversely, isolation of much older groundwaters in more hydrogeologically closed systems11,84,105.

Abiotic reactions are also important. Hydrogen that is exposed to subsurface unsaturated hydrocarbons will be lost to those systems188, accounting for the limited co-occurrence of hydrogen with long-chain hydrocarbons. Pyrite in the presence of H2 can be reduced to pyrrhotite at temperatures >90 °C, whereas sulfate can be reduced to H2S by thermochemical sulfate reduction189,190. It has been suggested that H2 might only be preserved on geological timescales within a narrow temperature window ranging from 100 to 200 °C where abiotic reactions remain kinetically limited, but above that favoured by microbes191. Evidence from ancient hydrogen-bearing mine fluids at lower temperatures, however, point to a wider but admittedly still poorly defined preservation window73,79,84,105.

Overall, within any hydrogen system tens of million years old, reservoir competency factors already familiar to the hydrocarbon industry enable a physical preservation and volumetric assessment. Hydrogen loss through diffusion-controlled processes and the dissolution of the gas phase into the groundwater remain the key preservation risks in older (tens to hundreds of million years old) hydrogen systems. Microbial consumption intersects with dissolution risk, with biological hydrogen loss enhanced by surface-water contact. Although preservation of hydrogen from inorganic reactions favours lower temperatures, the mineralogy of target systems can be used to mitigate this risk. Together, these can be used in developing a coherent exploration strategy for accumulation preservation.

Summary and future perspectives

The two dominant processes creating natural hydrogen gas in the continental crust are water–rock reactions with iron-bearing minerals, and radiolysis of water from the natural decay of uranium, thorium and potassium7. The Precambrian crust alone has generated hydrogen volumes over the last billion years equivalent to approximately 170,000 years of present-day societal oil use. However, defining how much natural hydrogen has been preserved and is accessible remains highly uncertain8. For example, gas-phase formation and accumulation with preservation from dissolution (redissolving), diffusional, microbial or chemical loss remain key risk factors. The Earth’s mantle does not contribute substantial H2 mass to the tectonically quiescent continental crust. Continental systems do not provide a regenerating hydrogen system on anthropogenic timescales — therefore, natural hydrogen should not be considered as a renewable resource. The production of natural hydrogen from geological accumulations will have a low-carbon footprint5, but notable knowledge gaps need to be closed to accelerate the discovery of societally important natural hydrogen accumulations.

With this Review, we provide the foundations of a geological framework for natural hydrogen exploration, with a particular focus on continental systems. Natural hydrogen generation and trapping requires a source rock, water and a trap with integrity to retain the hydrogen. Indeed, a stable tectonic context could be the best environment suited to the formation of material radiolytic or slowly generated hydrogen because of the long-geological timescales required to accumulate hydrogen in these settings. We conclude that although large accumulations of hydrogen are feasible, it is likely that very high concentrations of hydrogen (such as that found in Mali) will be exceptional. Gas concentrations should not nevertheless be conflated with the quantity of natural hydrogen. Mixed resources involving helium, nitrogen and other gases could contain substantive hydrogen. A better understanding of the timing of water–rock reactions and hydrogen preservation will be key to finding economically extractable volumes of hydrogen.

For exploration to be successful, predicting the amount and timing of hydrogen generated will be critical. Calculating the generative capacity of potential hydrogen source rocks from radiolytic sources is reasonably well developed, with field examples matching theory, and scaling a function of radio-element enrichment, porosity and age12,13,38,65,83. The hydrogen generated by water–rock reactions is subject to more uncertainty, but with enough information to provide basic information for exploration. Critical uncertainties for hydrogen produced by water–rock reactions include identification of how much water has been exposed to a lithology, how much rock has reacted, the mineralogy of the rock and the timing of hydrogen generation. The latter is controlled by both water availability and kinetics. Estimating the water–rock hydrogen potential of different terrane is currently limited by a lack of suitable thermodynamic data to calculate accurate volumes of hydrogen generated for low water–rock ratios and low temperatures relevant to the crystalline continental crust, in addition to data for only a limited range of lithologies beyond ultramafic mineral assemblage9,36,82. Notably, this absence is because much previous work has focused on mid-ocean ridge hydrothermal systems with high water–rock ratios and fast reaction kinetics. New investigations focusing on low water:rock ratio systems and rock types representative across different crustal-source rock types are now needed. These investigations will include cross-calibration of field and laboratory observations with thermodynamic and reaction-kinetic modelling across relevant continental systems. Scaling the laboratory and modelling results to field application will probably present the greatest technical challenges.

Play- or system-level predictions of terrane porosity and permeability, tectonic controls on source timescales, and fault, lithology or lateral controls on migration pathways are also essential for determining exploration potential. For example, hydrogen-rich fluid residence times within some source rocks can be on the order of 100 million-year to billion-year timescales. By contrast, regional deep gas fluxes derived from helium analyses show crust release rates in many regions where the continental crust is releasing gases to the near surface at the same rate they are being generated (‘steady state’). In other regions, tectonic activity releases a pulse of deep continental gas over tens of million years ago, that has accumulated in the deeper crust over hundreds of millions of years.

Geophysical detection of the magnitude and location of rock hydration through density contrast, resistivity channels or magnetic change caused by magnetite mineral formation will be important for system-level understanding191,192,193. The trapping efficiency of hydrogen in relatively high flux environments conforms to well-understood gas behaviour in porous media. Quantifying hydrogen accumulations and retainment in geological trap structures filled over longer timescales (10s to >100 Myr) requires development of appropriate models that consider the impact the entire lithological column has on diffusive gradients and multigas systems rather than single-seal trapping efficiency123,136. Recognition of the temporal elements that can operate on timescales much longer than typical hydrocarbon systems requires a cultural change in exploration approaches. We need to better take into account the role of basement deep-water history and transport controls to the near surface, diffusion, gas phase formation and groundwater flow in key hydrogen play types.

Substantial work remains to better understand the risk of both abiotic and microbiological hydrogen sinks. Assessing abiotic risk will probably form part of the same field, laboratory and modelling efforts discussed above, with focus on continental water:rock ratios and mineralogy. By contrast, assessing microbial risk, while clearly identifying low temperatures and contact with shallow groundwater, has more work to do when considering, for example, nutrient limits and how this may map onto accumulation targets in different terrane. Fusing geological data to understand and forecast the geochemical and geomicrobiological variables discussed here is the route to understanding the exploration risk and to rank different exploration opportunities. This data fusion will help to quantify and predict the discovery of economic, natural hydrogen accumulations. That predictability will be proportional to the technology and, in particular, the quality of geology, geophysics and geochemistry applied to the search, and the economic and regulatory environment that defines hydrogen exploration landscapes.