The occurrence of molecular hydrogen in the soils of Carolina bays was discovered after noting the remarkable geomorphological similarity to depressions that emit H2 in the EEC in Russia (Sukhanova et al. 2013; Larin et al. 2014). In both cases, relatively high H2 concentrations were found in the soil gas of the bay-like features. This study suggests that Carolina bays are geomorphological features related to the occurrence of H2 rather than the results of simple depositional processes along coastal plains (Stolt and Rabenhorst 1987). Therefore, the causal link between the origin of these morphological features and their association with molecular hydrogen must be established. This bay–hydrogen association suggests a connection between fluid seepage from depth towards the atmosphere (deep geological control) and the surficial geomorphic expression of Carolina bays. A possible interpretation might be that Carolina bays are not only focal points for groundwater recharge, as suggested by Grant et al. (1998), but also a morphological expression of fluid seepage from depth toward the atmosphere.
Origin of hydrogen
Carolina bays are natural depressions, commonly filled with water and organic deposits (peat). Consequently, we should consider whether the source of H2 might be attributable to superficial biological activity under reducing conditions (Sugimoto and Fujita 2006). However, our study clearly shows that the highest concentrations of H2 systematically occurred in the sand deposits on the rims of the bays, where there was no peat cover (Figs. 2, 4, and 7). We also detected H2 at a depth of 2 m, 120 cm below the level of the peat (Fig. 5), which reduces the likelihood of a very-near-surface origin for the H2. Furthermore, the H2 concentrations measured in the wetlands were low compared with the measurements made in sand. The H2 concentrations also increased with depth (Fig. 5, Additional file 1: Table SI-1), and the maximum concentration measured, 0.37 vol.%, was detected at a depth of 5 m below ground level in sandpit 2 (Additional file 1: Table SI-1), in a clay and sand substrate. Although slight background H2 associated with biological activity cannot be excluded, these observations make a purely superficial origin for the H2 highly improbable. It is also generally believed that the free H2 produced by bacteria remains at low concentrations because it is rapidly consumed by methanogenic bacteria and soil enzymes (Conrad and Seiler 1981). Indeed, the biological production of H2 occurs within bacterial consortia, where this valuable chemical energy source is constantly consumed and consequently remains at low concentrations (high H2 concentrations inhibit H2-producing bacteria). H2 is considered to be the most energetic substrate, able to sustain lithoautotrophic ecosystems in subsurface environments, where it is readily consumed (Nealson et al. 2005). Therefore, such high concentrations in soil gases are more probably linked to geochemical H2-producing processes.
Inventories of the geological controls on the sources of natural hydrogen have been made by Apps and Van De Kamp (1993) and Smith et al. (2005). The natural settings for hydrogen include hydrocarbon-bearing basins, young organic-rich sediments, coal beds, fault zones, extrusive igneous rocks, alkaline igneous complexes, geothermal fields, crystalline basements, potash-bearing strata, salt-bearing strata, and ultramafic rocks. The geologically controlled sources of natural H2 can be grouped according to four main families of processes: (1) water hydrolysis processes (several processes that include the oxidation of ferrous minerals, radiolysis, cataclasis, and metamorphism); (2) organic matter decay (including thermal maturation); (3) methane and/or ammonia decomposition during metamorphism; and (4) deep Earth degassing. The alteration of Fe(II)-bearing minerals is the most commonly reported source of natural H2 seepages on Earth, notably at mid-oceanic ridges and in ophiolitic massifs, where mafic and ultramafic are altered. Moreover, a recent study suggested that the H2 production from the Precambrian continental lithosphere has been underestimated (Sherwood Lollar et al. 2014).
Several ultramafic suites in the eastern Piedmont Province of NC are interpreted as parts of ophiolite sequences (Butler 1989). In particular, the Halifax County complex, described in detail by Kite (1982) and Kite and Stoddard (1984), and interpreted as ophiolitic, is locally overlain by Coastal Plain deposits. Kite and Stoddard (1984) also proposed that an extensive ophiolitic belt might occur beneath the Coastal Plain (also see Lawrence and Hoffman 1993). However, the available deep drilling data are insufficient to define exactly where these rocks are located (and at what depth). The alteration of peridotite, in both oceanic and continental contexts, produces H2 with the reduction of water by Fe(II), contained especially in fayalite (olivine ferric phase). This process, which is associated with serpentinization, consumes water and is responsible for mineral hydration. Other minerals, as well as olivine, contain Fe(II). In particular, clays (ferrous illites, chlorites, or smectites) can release Fe2+ ions in solution under certain conditions. The oxidation of dissolved ferrous ions by water produces H2, and this type of process could be a potential source of H2 if the sedimentary pile is clay-rich and provides a substantial reservoir of Fe(II).
It has been suggested that H2 forms during rock–water reactions (e.g., cataclasis) between fresh rock surfaces containing radicals (Sugisaki et al. 1983) and by the redox conversion of hydroxyls to peroxy groups in silicates (Freund et al. 2002). H2 can also be formed in uranium-, thorium-, and potassium-rich geological settings, by the radiolysis of water and/or organic matter, and also by the reaction of water with newly formed elements (Savchenko 1958; Lin et al. 2005). However, the quantities of H2 that can be produced by these mechanisms are limited, and therefore they cannot explain the H2 flows estimated in this study.
Another possible source of H2 is the decay of solid or dissolved organic matter, either by bacterial decomposition in sediments during diagenesis and/or later during thermal maturation, which produces H2 during the ultimate cracking of organic matter. Thermal maturation can be ruled out on the coastal plain of eastern USA because the sedimentary cover of NC and SC is not thick enough to produce the burial depth necessary for H2 production by thermal maturation.
The decomposition of methane and/or ammonia at high temperatures (above 600 °C) during metamorphism can produce H2. Such reactions can also produce dinitrogen. In the present case, we observed a diffusive flow of H2 in the soil, but because it was already mixed with atmospheric components, it was difficult to evaluate the gas components that evolved with it, which may help to clarify the origin of this molecular hydrogen.
The degassing of the Earth’s mantle is usually ruled out based on our present understanding of the oxidation state of the upper mantle, and it is therefore not compatible with the high concentrations of dihydrogen detected here. However, because the amount of hydrogen that was originally incorporated into the Earth’s interior is unknown, the oxidation state of the upper mantle may actually differ from the conclusions drawn from surface observations, which are necessarily indirect. A recently revised geochemical concept (Toulhoat et al. 2015), initially proposed by Larin (1993), suggests that the Earth’s interior is enriched in hydrogen, which is supported by calculations and experiments with iron hydride (Isaev et al. 2007). Such a model is consistent with Ohmoto’s suggestion (Ohmoto 1997) that the oxygenation of the atmosphere is linked to the evolution of the continental crust. Under these conditions, deeply stored hydrogen could slowly seep to the surface.
The daily H2 flows estimated in this study are quite important. For the largest structure, Jones Lake, they varied from 1120 to 2740 m3/day. If the total number of Carolina bays spread along the Atlantic coast of the USA (about 500,000) is considered, it is clear that a large-scale process is involved, and the mechanism that produces this hydrogen must be efficient enough to sustain the observed H2 flow.
Geometry and origin of Carolina bay structures
Whatever the source of this hydrogen, the subsurface migration of the ultimate reduced gas must induce reactions with oxidized subsurface rocks and fluids. These reactions and the seepage routes taken might be associated with the initial formation of the elliptical geomorphic structures that are characteristic of Carolina bays. These landscape features develop exclusively in areas where unconsolidated sediments are present at the surface. This is the case in both Carolina and Russia, where the weathered bedrock below the soil consists of unconsolidated granular sediments (Larin et al. 2014). These structures are rarely seen in river valleys, where they are likely to be obscured by fluvial processes. However, a creek that crosses Smith Bay (Fig. 4) does not seem to greatly affect the form of the bay. This may indicate either the recent age of the creek or of Smith Bay.
A variety of compounds can form with the hydrogenation of rocks along the migration pathways of H2, including water, hydrocarbons, and acids. All these compounds are susceptible to mobilization and to migration out of the reaction zone. In this way, hydrogen might induce an increase in the bulk porosity of the rocks during its vertical migration, and as a consequence, it is likely to create its own channel for vertical migration. All the associated processes (degassing, dewatering, and volume loss at depth) will generate subsidence at the land surface, thus forming rounded (circular or elliptical) depressions.
A study of similar depressions in the Ukraine suggested that water seepage from them is possible (Bixio et al. 2002). Therefore, water with dissolved hydrogen might be discharged onto the surface, creating swamps in some bays, such as the newly formed, small structure in Jones Lake Bay. This small depression, which appeared during the late 2000s, suggests a still active mechanism underlying the formation of some of the bays. Satellite images taken in 2008 show no sign of this structure, whereas 1 year later in 2009, it had become clearly visible (Fig. 6), with the loss of all the trees within it, within less than 1 year, which was probably related to the flooding of the area. The same rapid formation of a new structure, within only several years, has been observed in the Moscow region of Russia (Larin et al. 2014).
The highest concentrations of H2 occurred on the external slopes of the sandy rims of the Carolina bays, with moderate concentrations inside the bays and no H2 recorded at some distance from the bay limits (Figs. 2, 4, and 7). This distribution of H2 is similar to that in the structures studied in the EEC (Larin et al. 2014). The sand rims that border the bays (1–3 m higher than surrounding surface elevation; visible on LiDAR images in Fig. 3 and highlighted in Additional file 1: Figures SI-3 and SI-4) often showed the highest concentrations of H2. These zones appear to be the preferential drainage sites for H2-rich fluids moving toward the surface. Although the mechanisms of their formation and their association with higher concentrations of H2 are not yet understood, they may be linked to fault networks around the features or their peculiar petrological properties (notably the porosity of the rocks surrounding Carolina bays).
Field studies in the EEC have shown that bay-like features emitting H2 gas sometimes occur along structural trends, very probably corresponding to basement faults (Larin et al. 2014). Many studies have suggested that H2 anomalies are commonly related to faults (Wakita et al. 1980; Jones and Pirkle 1981; Ware et al. 1985; Sato et al. 1986; Shcherbakov and Kozlova 1986; McCarthy and McGuire 1998; Rogozhin et al. 2010), which act as fluid conduits. This suggests that H2-emitting features are genetically related to structural features of the crystalline basement. The available information on the deep geology of NC is insufficient to exactly define the locations of faults or their distances from these depressions (Lawrence and Hoffman 1993). This theory, together with the potential alignment of the bays along structural trends, suggests a close relationship between Carolina bays and the observed molecular hydrogen seepages from these potential geological structures.
The elliptical shape of Carolina bays is the feature that most significantly distinguishes them from the hydrogen-seeping structures in the EEC, which usually have rounded shapes. The elliptical shape of Carolina bays could be interpreted as a consequence of the local stress regime. The long and short axes of the bays appear to occur parallel to the minimum and maximum horizontal stresses, as is the case for many calderas. The formation of stress-induced oval structures is well documented in calderas around mud volcano systems (Bonini 2012). Stress discharge will determine the predominant orientation of the vertical (or subvertical) fractures of crystalline basement rocks (Lawrence and Hoffman 1993). Indeed, on geological maps, faults are almost always parallel to the bays’ orientation (Brown 1985). Consequently, the initially round (isometric) shape of the hydrogen stream would gradually become elliptical as it ascends through the upper layers of the lithosphere.
When Carolina bays are compared with the structures in the EEC (Larin et al. 2014), the chemical composition of the gases seeping from them and the flow rates of the gases are quite similar. The H2 concentrations range from tens to hundreds of parts per million. Small quantities of CH4 and its close homologues (C2+) are also sometimes present locally. We observed similar links between the geochemistry, geomorphology, and the distributions of the H2 concentrations in the Carolina bays and the Russian structures: the highest concentrations commonly occurred on the external shores of these structures, with high H2 concentrations inside the structures and almost zero H2 outside the depressions. According to our estimates, larger bays showed greater absolute H2 flows, whereas smaller bays showed greater flows per unit of surface.
When soil scientists studied the hydrogen-seeping features in the EEC in Russia (Sukhanova et al. 2013; Polyanskaya et al. 2014), they found that molecular hydrogen seeps from these structures and that this seepage affects the soil layers by disturbing the vegetation and the microbial biomass. In areas of H2 seepage, the humus content decreases by a factor of 2–3 and the optical density of the humic acids is lower than in the surrounding areas. The fertility of arable lands decreases significantly, and they often become unsuitable for cultivation (Sukhanova et al. 2013). The total quantities and biomasses of bacterial cells and fungal spores and the lengths of fungal and actinomycete mycelia decrease (Polyanskaya et al. 2014). The soil bleaching associated with Carolina bays could also be similar to the bleaching phenomenon associated with the bay-like features in the EEC.
The size distribution of the Carolina bays indicates that the number of bays decreases as the surface area of the bays increases (Semlitsch 2000). In this, their size distribution has the same characteristics as the bay-like features in the EEC (Larin et al. 2014), indicating that they may have a common origin.
In summary, the rounded (elliptical) depressions in the Atlantic Coastal Plain, Province of the USA, and the rounded structures in the EEC show similar geomorphic features and size distributions, and they emit H2 at similar flow rates. All these similarities suggest a common origin and the same mechanism of formation for these features. We interpret them as the surficial marks of pathways of hydrogen-rich fluid migration. We interpret these bays as the results of local structural collapses associated with the rock alterations induced by H2-enriched fluid flowing from the crust or deeper. Similar features can be seen in satellite images on all continents (except ice-covered Antarctica), suggesting that other analogous structures exist elsewhere in various settings.