*4.5. Possible H2 Bubbling at Depth*

Once produced by fluid–rock interaction processes (oxidation) at depth, H2 can migrate as a dissolved component. The solubility of H2 in aqueous solutions is rather low and drops when T and P decrease when approaching the surface (Figure 7d). The possible mechanism of H2 discharge, concentrating, and transport upward to the lower T–P where H2 is less reactive is solution boiling, i.e., formation of the vapor phase coexisting with the liquid phase. The concentrations of H2 in vapors are many orders of magnitude higher than that in the liquid (Bazarkina et al., 2020). At the same time, rock permeability is much higher for gas-rich vapors than for salt-rich liquids. Bubble formation is a function of T, P, total salinity, and gas saturation. Thus, during fluid ascent upward to the lower T–P, gas bubble formation is favored (Figure 7d). Periodicity of H2 emission at the surface reported by Prinzhofer et al. (2019) could be related to the kinetics of fluid–rock interaction at depth, further time-dependent bubble accumulation, and the final periodical ejections similar to those described in geysers.

**Figure 6.** Heat flow map in the São Francisco Basin within the studied zone (modified from Alexandrino and Hamza, 2012 [52]).

### **5. Draining Fault System in the São Francisco Basin**

Deep faults serve as significant channels for deep fluids to ascend into and through the crust and the *3He*/*4He* ratio can be used to estimate the flow rate of mantle fluids through the fault zones (Kennedy et al., 1997 [53]). Since the *3He*/*4He* signature seems to be of crustal origin in the São Francisco basin (Flude et al., 2019 [17]), the migration path followed by the *H2* mostly crosses the sedimentary cover without major changes of the ratio value. This weak interaction with the Bambuí sequences could be due to a high flow rate value along the faults. This could be possible if the faults form direct drains from the basement to the surface assuming a sufficiently high value of permeability.

Several interpretations of the available seismic data have been proposed for the fault systems (Figure 3) (e.g., Romeiro-Silva and Zalán, 2005 [24]; Coelho et al., 2008 [39] and Alkmim and Martins-Neto, 2012 [40]). Nevertheless, the São Francisco basin seems to encompass different tectonic elements such as the Proterozoic rift structures, Neoproterozoic foreland f–t-belts and Cretaceous rift structures (Reis and Alkmim, 2015 [25]). The rift structure that cuts across the central portion of the basin is characterized by a system of major NW–SE faults. One could expect that the deep-rooted faults in the graben structures (Precambrian sequence), which have been reactivated during the Neoproterozoic Macaúbas basin-cycle, could cross most of the sedimentary formation. As a major fault system, they may control the drainage at all depths and delineate some morphological features observed on satellite imagery and digital elevation models (Reis et al., 2017b [42]).

If these faults cross different geological layers, mainly shales, sandstones and limestones, their permeability values can range from 10−<sup>19</sup> to 10−<sup>13</sup> m<sup>2</sup> (Donzé et al., 2020 [54]). In terms of hydraulic conductivity for the water carrying the gas and neglecting the contribution of temperature, this could correspond to a value as high as 10−<sup>6</sup> m/s. This means that in the fastest scenario, the fluid could take less than 100 years to migrate across the Bambuí sedimentary layer through a fault system.

### **6. Possible Temporary Shallow Zones of H2 Accumulation**

The pressure variation observed at 1 m depth in the São Francisco basin (Prinzhofer et al., 2019 [15]), with a momentary increase in H2 pressure, could indicate that the H2 systems are active. There is only a small temperature window where H2 may remain stable over a long time. This window corresponds to a T range where abiotic redox reactions such as thermochemical sulfate reduction or carbonate reduction (e.g., Fisher Tropsh type reaction) are slow (Truche et al., 2009 [55]), and where bacteria are inactive. Such a T range can be roughly approximated to be 100–200 ◦C. Since these temperature conditions are not met at shallow depth, H2 will probably not survive to long residence time. Despite this fact, previous studies of H2 seepages often indicate that the hydrogen systems are active, and transient accumulations of hydrogen at relatively shallow depth can be observed (Prinzhofer et al., 2018 [14], Goebel et al., 1984 [28]; Guélard et al., 2017 [26]). These observations may sugges<sup>t</sup> a constant recharge of the aquifers by H2 flowing from deeper levels of the basin.

The circular depression where H2 seepage is observed (Figure 2b) could be related to a sinkhole structure resulting from a chemical dissolution process at depth. If so, these depressions will contain standing water connected with a ground-water reservoir contained in karst (De Carvalho et al., 2014 [56]). The presence of resistive carbonate and calcareous rocks was inferred from ~320 m to ~480 m followed by a layer of intercalated shales and sandstones (Solon et al., 2015 [23]). This carbonate layer, which corresponds to the Lagoa do Jacaré carbonate layer, exhibits a potential karst system according to the outcrops located East of the São Francisco Basin (Dos Santos et al., 2018 [57]).

Assuming a karst system at depth, this could imply a high level of porosity favorable for a massive storage volume of an aquifer. Since karst features are controlled by structural heterogeneities, such as faults and fractures, which influence fluid flow, they can provide preferential pathways for geofluids with the development of secondary porosity. This could agree with the fact that the circular depressions where H2 is venting are aligned along a major fault (Cathles and Prinzhofer, 2020 [16]).

### **7. Putting It All Together: A Potential H2 System within the São Francisco Basin**

The first key point is related to potential source areas, e.g., the presence of both ultramafic and U, Th and K-rich rocks. The presence of Archean greenstone belts containing ultramafic rocks, TTG, migmatites and K-rich granitic plutons represent excellent H2-producing zones either via serpentinization, or water radiolysis. Magnetic (Figure 4) and Bouguer (Figure 5) anomalies are compatible with the presence of ultramafic rock producing H2. Temperature conditions also seem favorable for the serpentinization process: with a temperature gradient of 25 ◦C/km (Alexandrino et al., 2008 [48]), the optimum range of temperature would be expected at a depth of 10 km, with possible lower rate processes at a shallower depth.

The second key point is the structural/tectonic context and the presence of faults deeply rooted in the basements capable of draining a potential deep and scattered source. All interpretations of the seismic profile of the zone of interest sugges<sup>t</sup> the presence of deep faults following the graben structures (Figure 3): they can be able to drain hydrogen produced at depth where the Pressure–Temperature conditions are optimal. Some interpretations sugges<sup>t</sup> that some of these faults could cross the entire sedimentary sequence (Coelho et al., 2008 [39]), producing gas seepages directly at the surface. Some others predict that these faults could reach some potential shallow carbonated reservoirs (Romeiro-Silva and Zalán, 2005 [24]). These deep faults could also only reach the unconformity zone which composes the boundary between the sedimentary basin and basement (Alkmim and Martins-Neto, 2012 [40]).

The third key point concerns the storage areas (i.e., reservoirs) of H2 at depth. As mentioned previously, the interface between the basement rocks and the sedimentary layers could represent a potential zone of accumulation. The interface is composed of the Macaúbas and the Espinhaço formations. The Macaúbas sequence is made up of sandstones, pelites, diamictites, carbonates, basic volcanic rocks, and metamorphosed banded iron formations (Alkmim et al., 2012 [40]), whereas the Espinhaço formation is a quartz–arenite dominated package. The presence of the Paranoá–Upper Espinhaço quartzite, which is tectonically uplifted, can facilitate the occurrence of sandstone reservoirs with appreciable permeability and porosity. Thus, potential reservoir rocks could be found among siliciclastics of the Macaúbas–Paranoá Megasequence (Solon et al., 2015 [23]).

An interesting characteristic of the deep topography is that the seepage zone H2G is located near the apex of the basement rock in the central part of the basin (see Figure 3). As the H2 charged fluid reaches the Macaúbas/Espinhaço formations, it migrates along the unconformity toward this highest point before escaping to the surface in the green seepage zone (Figure 2). On its way to the surface, H2 can also be temporarily trapped in the Sete Lagoas formation and at a shallower depth, inside the Lagoa do Jacaré formation. The permeability value of the Sete Lagoas Larst aquifer formation is estimated to range between 10−<sup>14</sup> m<sup>2</sup> and 10−<sup>9</sup> m<sup>2</sup> (Galvão et al., 2015 [58]). As for the Lagoa do Jacaré formation, very low permeability and porosity values were found in the Petrobras well 1-RF-1-MG. The presence of faults, possibly connecting all these reservoirs with the surface could explain the apparent structural control of the distribution of the known gas seepages (Curto et al., 2012 [22]). Nevertheless, the presence of sinkholes in the H2G seepage area suggests the existence of a shallow local karst formation, which could constitute a temporary reservoir for H2.

**Figure 7.** Conecputal model of the H2 cycle in the Sao Francisco Basin. (**a**) Interpretated seismic section (Martins-Neto, 2009). (**b**) Zoom of the upper part of the Bambui sequence. (**c**) Possible presence of a karst structure according to the presence of sinkholes (Figure 2). (**d**) Calculated solubility of H2 in H2O vs. depth (Bazarkina et al., 2020 [59]).

Thus, surface seepages may be either in connection with the source rock or with intermediate leaking reservoirs since these two configurations are present in this area. A summary of H2 migration from sources to seeps in the São Francisco basin is presented in Figure 7.

### **8. Discussing the H2 Production from Radiolysis and Hydration Reactions in the São Francisco Basin**

Combining the H2 production rate from water radiolysis and hydration reactions assessed in the previous sections, we obtain an estimate of 1.01 to 6.43 × 10<sup>8</sup> mol·yr<sup>−</sup><sup>1</sup> H2 production, i.e., ~200 to 1300 tons·yr<sup>−</sup><sup>1</sup> (Table 3). Cathles and Prinzhofer (2020 [23]) considered the local flux rate in the H2G seepage zone (Figure 2b) to range from 7000 to 178,000 m<sup>3</sup> per day. At a temperature of 21 ◦C and a pressure of 1 atm, these values correspond to 0.105 to 3.68 × 10<sup>9</sup> mol·yr<sup>−</sup>1, i.e., 213 to 5400 tons·yr<sup>−</sup><sup>1</sup> (Table 3). Since the expulsion rate of H2 is almost certainly not steady, the episodic rate measured in the H2G vent might be overestimated.


**Table 3.** Estimated H2flow rate production (tons·yr<sup>−</sup>1).

In comparison, on the Mid-Atlantic Ridge (MAR), the total H2 discharge at the Rainbow hydrothermal field is estimated to be ~10<sup>8</sup> moles H2 per year, i.e., ~200 tons·yr<sup>−</sup><sup>1</sup> (Charlou et al., 2010 [60]) (Table 3). At a larger scale, the H2 flux from all high-temperature basaltic vents along the MAR has been estimated at ~109–1010 mol·yr<sup>−</sup>1, whereas the H2 flux from high-temperature ultramafic vents along the Mid-Oceanic Ridge (MOR) has been estimated at ~1010–1011 mol·yr<sup>−</sup><sup>1</sup> (Table 3) (Keir, 2010 [61]; Cannat et al., 2010 [62]), i.e., 20,000 to 200,000 ton.yr−1.

According to our calculations, which are based on the model developed by Sherwood Lollar et al. (2014 [29]) for the Precambrian continental lithosphere, the maximum H2 production rate from the basement rocks of the São Francisco Basin is within the same order of magnitude as the H2 flux of one sinkhole of 500 m in diameter (H2G zone). This latter H2 venting site would also represent from 0.047% to 7.5% of the global estimated H2 production from the Precambrian continental Lithosphere. This large discrepancy in the results leads us to conclude that there is a need to increase the accuracy of hydrogen flux estimates through long term monitoring of soil gas migration according to different methodologies and/or to revisit the global models.
