Natural hydrogen potential and basaltic alteration in the Asal–Ghoubbet rift, Republic of Djibouti.

Dihydrogen (H₂) is a key element in the energy transition and is often used as an energy carrier (e.g., for storage) rather than as a primary energy source (Moretti, 2019; Lapi et al., 2022). At present, H₂ is produced primarily from hydrocarbons through a process that emits high volumes of CO₂; this is known as “grey” hydrogen (Rigollet and Prinzhofer, 2022). However, lower emission sources of H₂ exist and are more appropriate in the context of current global environmental conditions. Dihydrogen is referred to as “blue” when CO₂ emissions are captured and stored during grey hydrogen processing and as “green” when it is produced by the electrolysis of water using a decarbonized source of electricity. “White” dihydrogen is H₂ in its natural state (Rigollet and Prinzhofer, 2022) and is generated mainly through fluid rock interactions reactions (Smith et al., 2005; Gaucher, 2020; Moretti and Webber, 2021).

This natural H₂ has been observed in various geological contexts (Truche et al., 2020; Zgonnik, 2020), including mid-oceanic ridges (Charlou et al., 1998, 2002: Kelley et al., 2001; Kelley, 2005); ophiolitic massifs such as the Sultanate of Oman, New Caledonia, or the Philippines (Neal and Stanger, 1983; Abrajano et al., 1988; Sano et al., 1993; Deville and Prinzhofer, 2016; Vacquand et al., 2018); and Archean and Proterozoic cratonic basins such as those in Namibia, Mali, Russia, Australia, and Brazil (Larin et al., 2015; Prinzhofer et al., 2018, 2019: Boreham et al., 2021; Moretti et al., 2021, 2022: Malvoisin and Brunet, 2023). Steam at the high-temperature geothermal sites of Italy, Iceland, and Djibouti (Oskarsson, 1984; Ármannsson et al., 1989; Leila et al., 2021; Pasquet et al., 2021, 2022: Combaudon et al., 2022) has also been shown to contain H₂, possibly from multiple sources.

Many H₂ emission zones have been discovered, yet the production of natural hydrogen (as in Mali; Prinzhofer et al., 2018) remains rare and natural H₂ exploration is still in its infancy. The geological and geothermal context of the active rift in the Afar Depression is favorable for natural H₂ production, with several sources proposed previously: alteration of iron minerals in basaltic material, pyritization, or volcanic degassing (Holloway and O’Day, 2000; Klein et al., 2020; Pasquet et al., 2021). Following surface gas surveys, analysis of cuttings from two wells is presented below and aims to improve our understanding of basalt alteration and Fe²⁺/Fe³⁺ evolution at depth across the Asal–Ghoubbet rift in order to assess its potential H₂ resource.

Many reactions can produce H₂ (Worman et al., 2020), depending on the protolith, fluid, and physical parameters (i.e., temperature, pressure) of the environment; some of these reactions are oxidation–reduction processes. The ferrous iron contained in a mineral is oxidized when in contact with water to become ferric iron, whereas the water is reduced to form H₂. The more ferrous iron a rock contains, the greater its ability to produce H₂ during alteration processes. Basalt is composed mainly of plagioclase, with smaller amounts of pyroxene and olivine (olivine typically less than 20%). Olivine, including the iron-rich end-member fayalite, is highly unstable at the surface and will deteriorate more rapidly than the other minerals present in basalt (Stevens and McKinley, 2000; Sissmann et al., 2014; Kularatne et al., 2018). In contrast, plagioclase, the major phase, has less or no iron and is kinetically more stable. Accordingly, the production of H₂ from a basaltic protolith is typically limited by its high plagioclase content.

Despite this limitation, volcanic zones such as Iceland (Combaudon et al., 2022) offer other advantages for natural hydrogen production. These include the relatively regular renewal of crustal material (and thus of potentially oxidizable minerals), high temperature gradient, significant water resource (meteoric, glacial, or marine), and production of H₂S due to volcanic activity.

The present study will focus on several oxidation–reduction equations that generate H₂: oxidation of Fe²⁺ from fayalite to magnetite (eq. (1)); oxidation of Fe²⁺ from magnetite to hematite (eq. (2)); oxidation of Fe²⁺ from mesostasis to hematite (eq. (3)); oxidation from H₂S to SO₂ in the gas phase (eq. (4)); iron sulphides pyritization (eq. (5)); oxidation of H₂S to form pyrite from hematite (eq. (6)) and magnetite (eq. (7)) (Arrouvel and Prinzhofer, 2021); and oxidation of chlorite from Fe²⁺ to Fe³⁺ (eq. (8)) (Steudel et al., 2016; Lempart et al., 2018).

Additionally, different clay phases can be formed by basalt alteration at different temperatures, with smectites at low temperatures and chlorites at higher temperatures. This argillization of basalt does not necessarily lead to direct production of H₂, because the iron remains ferrous (Reed and Palandri, 2008). However, once chlorite is formed, it can oxidize (Brindley and Youell, 1953); here, we discuss whether this could represent an intermediary phase for H₂ production. As in hydrocarbon systems, in addition to defining the source rock (here, the reactive minerals) for H₂ generation, it is also necessary to define a reservoir, migration pathways, and a seal.

The Asal–Ghoubbet rift is part of a large spreading system, the East African Rift (EAR), that extends from the Red Sea in the north to Mozambique in the south. In the Republic of Djibouti (Fig. 1), in the north of the magmatic east branch of the EAR, this rift is an emergent segment of the Aden Ridge (Stieltjes et al., 1976). Although continental extension started earlier, the first evidence of oceanic crust accretion in the Gulf of Aden has been dated to 20–17 Ma (Nonn et al., 2017, 2019). In the Red Sea, extension started approximately 25 Ma and ocean formation approximately 4 Ma. In the southern Red Sea, Gulf of Aden, and Afar region, after various phases of extension, volcanism, and thinning (Barberi et al., 1972, 1975b), a large series of fissural basaltic eruptions resulted in the Afar Stratoïd Series (Barberi et al., 1975a). This series is composed of basalts with an alkaline tendency (Varet, 1978; Fournier et al., 1984) at its base and rhyolitic rocks at its top and has been dated to 4–1 Ma. These eruptions were followed by the Axial Series, from 1 Ma to the present day; this series comprises more evolved tholeiitic and porphyritic basalts. These series represent the accretionary segments of the present axial volcanic ranges, linked together by transform faults, such as the Asal–Ghoubbet rift (Barberi and Varet, 1970; Stieltjes et al., 1976; Barberi and Varet, 1977). The N130° oriented Asal–Ghoubbet rift has an opening rate of approximately 1.5–2 cm/year (Delibrias et al., 1975) with a N40° direction. The last major eruption in the rift zone was that of the Ardoukoba shield volcano in 1978 (Demange et al., 1980). The rift comprises three main zones: the axial volcanic zone, the inner margins, and the outer margins. The lavas evolve from porphyritic on the margins to aphyritic in the center and from magnesian to ferrous towards the center of the rift zone (Stieltjes et al., 1976; Varet, 1978).

The geodynamic activity of this rift has prompted numerous geothermal exploration campaigns since the 1970s, and temperatures observed in exploration wells have been encouraging with an average thermal gradient of 180 °C/km to 200 °C/km in the axis as on the margins. In 1975, the BRGM (Bureau de recherches géologiques et minières) drilled two wells on the southwestern outer margin of the rift: wells A1 (260 °C) and A2 were drilled to depths of 1154 and 1554 m, respectively (Abdillahi, 2014). A second phase of drilling, carried out by Aquater between 1987 and 1988, created three new wells on the outer margins of the rift: wells A3 (T_max = 265 °C), A4 (T_max = 344 °C), and A6 (T_max = 281 °C) were drilled to depths of 1316, 2011, and 1761 m, respectively, while well A5 (located in the center of the rift) was drilled to 2105 m (T_max = 359 °C) (D’Amore et al., 1997; Abdillahi, 2014). Since 2016, further drilling has been undertaken in the rift: the Gale le Goma 1 (Glc1) well was drilled to 600 m on the southern outer margin near A3 and A1 (T_max = 138 °C) (Ahmed, 2018); and, more recently, wells F1 (T_max = 363 °C), F2 (T_max = 352 °C), and F3 (T_max = 363 °C) were drilled around the Fiale Caldera to depths of 2475, 2237, and 2402 m, respectively (Turk et al., 2019).

Seismic velocity anomaly studies have identified a magma chamber beneath the Fiale Caldera at a depth of more than 5 km (Doubre et al., 2007). Combined velocity anomaly data and well data indicate cold marine water intrusions between 0 and 1 km depth in the center of the rift (Zan et al., 1990; Doubre et al., 2007; Doubre and Peltzer, 2007). Seawater flows from Ghoubbet Bay to Lake Asal through numerous NW–SE faults. This temperature inversion at shallow depth has been noted in wells A5, F1, F2, and F3, i.e., at the center of the rift (Zan et al., 1990; Doubre et al., 2007; Turk et al., 2019). At the rift margins, these hypersaline fluids are warmer and the flow is weaker (Mlynarski and Zlotnicki, 2001).

Well data and a field campaign (2019) across the rift highlighted the presence of H₂ at depth in the area (Fig. 2), in some geothermal wells along the southwestern margin (wells A3 and A6; D’Amore et al., 1997), and in the fumaroles along the southern inner margin near well A5 and the Fiale Caldera (Pasquet et al., 2021). The well data indicate H₂ percentages of ∼0.25% (in dry gas, excluding air and water vapor); however, no H₂ sources were measured at the surface in the vicinity of the wells during the field campaign. Along the inner margins and in the axial volcanic zone, although no gas data were acquired at depth, the intense hydrothermal activity present at the surface made it possible to measure gas contents: after correction for air, H₂ contents were found to be up to 3%, with an average value of 1% and a mean value of 0.77% (Pasquet et al., 2021). Gas analysis following the 1978 Ardoukoba eruption (in the volcanic core zone) also indicated 0.15% H₂ in the magma gas (Allard et al., 1979). At the surface, H₂ concentrations were found to increase towards the center of the rift (Pasquet et al., 2021).

Several hypotheses have been proposed for the origin of this H₂: (1) alteration of basalts by hot hypersaline fluids circulating at depth; (2) alteration of basalts at lower temperatures due to marine intrusion at shallow depth in the rift axis; (3) pyritization; and (4) volcanic degassing as magma ascends through major active faults rooted near the magma chamber (Pasquet et al., 2021).

In summary, H₂ has been measured at depth in geothermal wells along the margins of the Asal–Ghoubbet system but has not yet been detected at the surface in these areas. Conversely, in the center of the rift, H₂ has been measured at the surface in fumaroles but no subsurface data are available.

We first conducted a preliminary analysis of surface gases in the axial volcanic zone. Gas measurements were acquired in the field during two campaigns in 2019 and 2020 using the BIOGAS 5000 analyzer; some of this dataset has been published previously (Pasquet et al., 2021). Soil samples were acquired primarily during the first campaign in 2019, although one sample was acquired in 2021.

The drill cuttings from the Fiale 1 (F1) and Glc1 wells were studied based on total rock analysis: ICP–OES (Inductif Coupled Plasma Optical Emission Spectrometry), XRD (X-ray diffraction) and Mössbauer spectroscopy. The ICP–OES analyses were carried out by the Service d’analyse des roches et des mineraux of the CRPG at Nancy.

Surface gas measurements were acquired on three different days. The studied site is located near the Asal 5 (A5) and F1 geothermal wells, along the inner margins of the rift along a major N140°-oriented fault that runs between Ghoubbet Bay and Lake Asal (Pasquet et al., 2021; Pasquet, 2023). This site includes a set of fumaroles with hydrothermal alteration around them (Fig. 3). Steam is present over approximately 70 m wide and the Afar people living in this area have installed tubes to induce condensation and take advantage of the drinking water contained in the hot steam. To obtain measurements, a perforated stainless steel tube was inserted into the fumaroles; drilling tools were not necessary as the soil was soft. The tube was then connected by silicone tubes to the BIOGAS 5000 field analyzer. Also, to increase gas capture for fumaroles, a funnel can be placed in the vent and connected directly to the analyzer. Gas contents were monitored in real time, although only peak values for each 2-min interval were recorded. CH₄, CO₂, and O₂ were measured in % and H₂, CO, and H₂S in ppm. The remaining unanalyzed gases were grouped under the term “Balance” and measured in %. This balance is often considered to be N₂ but might also include, in this volcanic context, gases such as SO₂ or HCl. H₂O_(g) was not analyzed owing to the presence of a water vapor filter. Measurements uncertainties were as follows: ±0.5% of the CH₄ and CO₂ contents; ±1.0% of the O₂ content; ±2.5% of the H₂ content; and ±2.0% of the CO and H₂S contents.

As H₂ contents have been shown to vary both within and between days in some geological settings (Prinzhofer et al., 2019; Moretti et al., 2021, 2022), we recorded the time each measurement was obtained and will present the results versus time. Although samples were obtained from various locations along the leaking zone (Fig. 3), no SE–NW trends in concentrations were noted; accordingly, we did not attempt to map H₂ concentrations spatially.

H₂ can be an electron donnor and an energy source for many aerobic, sulfate-reducing, (homo)acetogenic and methanogenic microorganisms. We identified the microbial families present and considered the metabolic processes associated and described in the literature, in particular their ability to consume H₂. Extraction of genomic DNA was performed on 250 mg of soil using the DNeasy PowerSoil Pro kit (Qiagen). gDNA was quantified by fluorescence, using the Quant-it kit (Invitrogen), following the supplier’s instructions. A Synergy HTX multi-mode reader (BioTek) microplate reader using the Gen5 software was used. Amplifications of 16S rRNA genes were performed by PCR for soil diversity analysis by Illumina MiSeq sequencing. Universal primers 515F-ill (CTTTCCCTACACGACGCTCTTCCGATCTGTGYCAGC- MGCCGCGGTA) and 928R-ill (GGAGTTCAGACGTGCT- CTTCCGATCT CCCCGYCAATTCMTTTRAGT) (region V4–V5; Wang et al., 2015) were used, containing the tags required for sequencing, according to the following PCR cycle: 94 °C-2 min; (94 °C-30 s; 65 °C-30 s; 72 °C-40 s) × 30; 72 °C-7 min. The core PCR kit (Roche) supplemented with 1 mg/mL (final concentration) of BSA (Bovine Serum Albumin, supplied by NEB-B9200S) was used, with a 2720 Thermal Cycler (Applied Biosystems). The amplicons were sequenced using the MiSeq Illumina 2 × 250 bp technique, via the GenoToul sequencing platform (Toulouse, France) and analyzed with the FROGS analysis pipeline as described in Haddad et al. (2022).In order to monitor metabolic activity based on gas phase variations in soil samples from the 2021 Djibouti campaign, soil samples weighing 2 g were introduced into sterile 100 mL penicillin vials. Two conditions were defined: in anaerobiosis and in aerobiosis. The anoxic vials were hermetically sealed (butyl stopper, aluminum cap) and flushed with nitrogen; then, 10 mL of sterile nitrogen-degassed distilled water was added to the soil. Finally, these bottles were flushed with H₂/CO₂ (80/20; 1 bar). The oxic flasks were treated in the same way except for the initial gaseous phase, which was air and sterile distilled water that had not been degassed in order to maintain aerobic conditions. H₂/CO₂ was added under slight overpressure in the sealed bottles (1 bar). The vials were then incubated at 95 °C. After 10 days of incubation, the composition of the gaseous phase was again analyzed to compare the gaseous phases versus T0; in this way, we were able to monitor the evolution of the system (GC-μTCD, Micro GC Fusion, Chemlys, France). The cultures were placed in an oven at 37 °C for 30 minutes before analysis and the pressure of the gaseous phase quantified before and after analysis by a manometer (Digitron 2022P) in order to quantify the moles of gases present under the same conditions.

Materials from two field campaigns (in 2019 and 2021) have been studied. Soil samples were collected in triplicate at Site 4 (Fig. 3). During the 2021 campaign, three locations along the leaking zone were sampled and named zones 13, 15 and 16.

The well cuttings were reduced to powder to obtain XRD measurements. The analyses were carried out at the Université de Pau et des Pays de l’Adour using a D2 PHASER diffractometer (Bruker). XRD allowed us to obtain information about the crystalline phases of our rock samples and thus to determine both qualitatively and quantitatively the presence of specific minerals. For this analysis, the powdered samples were mounted on a base and the intensity of scattered X-rays analyzed to determine mineral composition.

The analyses were performed under conditions of 10 mA, 300 W, and 30 kV and a PSD (Position Sensitive Detector) of 1.6 s. The front slot, blade, and back slot had openings at 0.2, 1.0, and 3.0 mm, respectively. The observed 2θ angle was between 10 and 80° with an X-ray wavelength of 1.54184 Å.

Once the mineral phases were identified, their Fe²⁺/Fe³⁺ ratios were measured to determine potential for oxidation to produce H₂. For this purpose, ⁵⁷Fe Mössbauer spectroscopic analyses were carried out at the “Resonance Spectroscopies” facility of the Institute of Condensed Matter Chemistry of Bordeaux (ICMCB). This analysis is based on the Fe oxidation state, magnetic state, and crystalline chemistry of iron-containing minerals. In particular, it allows characterization of the valence of iron and the sites at which iron is present within crystals.

The measurements were carried out at room temperature using a constant acceleration spectrometer operating in transmission geometry. The gamma emitter used in the experimental setup is a ⁵⁷Co radioactive source (activity: 1.85 GBq) embedded in a rhodium matrix kept at room temperature (293 K). Thin absorbers were prepared with about 250 mg of sample powder and placed between the source of gamma rays and a NaI(Tl) scintillation detector. The recorded ⁵⁷Fe Mössbauer spectra are then the sum of various components (sub-spectra) associated with the different iron-bearing minerals in the analyzed rock samples. Each sub-spectrum is characterized by Mössbauer hyperfine parameters: isomer shift (δ, in mm/s, relative to α-Fe0 at room tempearture), quadrupole splitting (Δ, in mm/s) or quadrupole shift (2ε, in mm/s), hyperfine magnetic field (Bhf, in T) and resonance line width (Γ, in mm/s). Refinement of hyperfine parameters and relative absorption areas were performed with the WinNormos^® software (Wissenschaftliche Elektronik GmbH).

Here, we present our results based on the analysis of in situ gas measurements obtained at Site 4 (Fig. 3). This fumarole site (T > 96 °C) was measured on three different days (one in November 2019 and two in August 2020), with measurements taken at regular intervals from 09:30 to 17:00 (UTC +3) to record diurnal variations in H₂ activity. In November 2019, measurements were collected throughout the study site; accordingly, the measurement values for this period (i.e., orange dots in Fig. 4) were obtained from slightly different locations. To consider conditions at depth, we discuss here the results from drill cuttings obtained from the F1 and Glc1 geothermal wells; we collected 37 samples down to 1118 m and 10 samples down to 458 m for F1 and Glc1, respectively.

Site 4 is an active hydrothermal site along the inner margins of the Asal–Ghoubbet rift, located to the southwest of the Fiale Caldera near wells F1 and A5 (Fig. 2). Fumaroles escape from a major rift fault with a gas temperature exceeding 96 °C. This site has been shown to exhibit the highest surface H₂ concentrations in the area, with values exceeding 1000 ppm (Pasquet et al., 2021). The temporal evolution of H₂ and CO₂ at this site is illustrated in Figure 4, with data points color coded according to sampling day.

For our uncorrected dataset, bell trends in in situ H₂ concentrations were observed, with the highest values (900–1000 ppm) were typically observed between 13:00 and 15:00 (Fig. 4a). In contrast, in situ CO₂ measurements increased with time to a maximum value of approximately 6% (Fig. 4b).

A major problem encountered in this region, as in many volcanic emissions, is that the sampled gas is a mixture of volcanic gas and atmospheric air. The volcanic gases can be diluted with air by the thermal contrast between the emanations and the atmospheric air, by the lava corridors formed during previous eruptions, or by the large faults in the vegetation-free landscape, which induce high vertical and horizontal permeability. The field analyzer measured CH₄, CO₂, O₂, H₂, CO, H₂S, and the balance. As already stated, a water filter was placed upstream of the analyzer to obtain the dry gas concentration; volcanic gas typically contains approximately 80% water vapor. Then, having measured the O₂ concentration and if the balance is considered to contain mostly N₂ (i.e., if the balance/O₂ ratio is higher than that of the atmosphere), the measured gas can be corrected. To do this, we subtracted the O₂ compound, which we considered to be outside the system, and removed the equivalent volume of N₂ from the balance according to the atmospheric ratio N_2(atm)/O_2(atm) = 3.727. Similarly, argon was removed from the balance according to the ratio Ar_(atm)/O_2(atm) = 0.044. We observed a temporal trend in the balance/O₂ for samples taken on two days (i.e., blue and orange points in Fig. 4c), which allowed us to make this correction. This ratio was higher than 3.7, indicating that the balance contains other gases than atmospheric N₂ (Ar, He, SO₂…). In contrast, a different temporal trend was observed for the other sampling day (i.e., gray points in Fig. 4c): we observed a constant decrease of the balance/O₂ ratio between 4.3% and 3.8%, associated with an increase in O₂ concentration from 18.2% to 20.5% (Fig. 4c).

Once these air corrections were made, H₂ concentration was found to remain relatively constant throughout the day for samples collected in November 2019 and on August 12th, 2020, with values between 0.57% and 0.89%. However, after correction, the H₂ concentrations obtained on August 13th, 2020 exhibited a strong positive diurnal trend, increasing from 0.31% to 3% over the course of the day (Fig. 4d). This could be explained by a higher concentration of O₂, and so a bigger correction on air (raw data are available in Supplementary material 1).

Our corrected CO₂ concentrations exhibit no notable trend for the samples collected in November 2019. Concentrations measured on August 12th, 2020 increased from 21% to 36% over the course of the day, while the strongest positive correlation was observed on August 13th, 2020 (Fig. 4e). Finally, we observed a positive correlation between H₂ and CO₂ (Fig. 4f), as shown previously by Pasquet et al. (2021).

Following microbial analysis of soil samples collected at Site 4 in 2019, samples were collected in three different locations in 2021. DNA samples were extracted from these soils and the 16S rRNA genes were amplified and sequenced to taxonomically identify the prokaryotes present. The number of sequences analyzed (following high-throughput sequencing and the removal of low-quality reads and chimera) was as follows: 19,038, 15,467 and 17,338 sequences for Site 4 2019; 15,359, 17,993, and 18,144 sequences for Site 4 2021 (zone 13); 18,454, 19,663 and 18,368 sequences for Site 4 2021 (zone 15); and 17,614, 18,462 and 13,611 sequences for Site 4 2021 (zone 16). The analysis of these sequences indicated a very high dissimilarity between the communities collected in 2019 and 2021, as illustrated by the Heatmap (Fig. 5).

As shown in Figure 6, the microbial community of 2019 included few archaea (>2%), while the 2021 community was dominated by these prokaryotes (with 91 ± 13% for Site 4 2021 [zone 13] and 93 ± 3% for Site 4 2021 [zone 15]). Site 4 2021 (zone 16) revealed another equilibrium between bacteria and archaea (37 ± 4%), which could be explained by the presence of abundant chloroplasts (28 ± 3%). The archaea belonged primarily to the phylum Crenarchaeota, including the classes Nitrososphaeria and Thermoprotei, and to the orders SCGC AB-179-E04 (from 33 ± 4% to 79 ± 13%) and Geoarchaeles (from 4 ± 0% to 19 ± 16%), respectively. No archaea similar to methanogens were identified at this site during either sampling campaign. In addition, enrichments carried out in the dark under anoxic and oxic conditions showed no significant consumption of H₂ or CO₂ present in the gas phase, during incubation carried out for 10 days.

Regarding bacteria, the community from Site 4 in 2019 included mainly the following orders: Kapabacteriales (15 ± 1%), Thermoflexales (family of Thermoflexaceae; 8 ± 3%), Thermomicrobiales (Thermomicrobiaceae; 8 ± 4%), Cyanobacteriales (Nostocaceae; 14 ± 3%), Brevibacillales (Brevibacillaceae; 10 ± 4%), Paenibacillales (Paenibacillaceae; 5 ± 2%), and Symbiobacteriales (Symbiobacteraceae; 3 ± 2%). The presence of organisms affiliated to the class of Hydrothermae (14 ± 3%) should also be noted. For the second sampling campaign, in zones 13 and 15, bacteria represented only 6–8% of the relative abundance of the communities present. Only the community at Site 4 2021 (16) exhibited a relatively high abundance, with 63% of bacteria belonging to the orders Pseudonocardiales (Pseudonocardiaceae; 7 ± 1%), Ktedonobacterales (Ktedonobacteraceae; 5 ± 1%), and Gemmatales (Gemmataceae; 7 ± 2%).

The total rock analyses performed by ICP–OES for the F1 and Glc1 wells (Tab. 1) demonstrate elemental variations with depth in the inner and outer margins. For the F1 well, SiO₂ concentrations remained relatively constant with depth, typically in the range 44–47%, although SiO₂ values closer to 50% were observed at a few depths. In contrast, sample 15 (304 m) was particularly poor in SiO₂ (31.86%). For Glc1, SiO₂ concentrations were more variable: typically, these were lower than 45% for the first 280 m and higher than 45% in the rest of the cuttings, reaching up to 66% for sample 10 (458 m).

A transition is apparent in well F1 at 464 m depth (Fig. 7). Al₂O₃ concentrations were higher at depths of 0–464 m, typically varying between 18.86 and 25.32%, compared to 11.95–21.21% in the rest of the well. Most other parameters analyzed generally exhibited lower values at shallower depths (i.e., 0–464 m), including total iron (4.45 to 8.81%), TiO₂ (0.67 to 1.47%), Na₂O + K₂O (2.18 to 2.46%), and loss on ignition (LOI; 1.31 to 5.69). However, concentrations of CaO were higher at these depths (12.18 to 15.39%) than elsewhere. Values at 518–1118 m were as follows: 5.51–15.83% for Fe₂O₃; 0.97–3.48% for TiO₂; 2.67–6.20% for Na₂O + K₂O; 2.07–9.37% for LOI; and 5.12–13.54% for CaO.

In well Glc1, a transition was also observed at 280 m depth (Fig. 7), with lower SiO₂ contents and higher Fe₂O₃, MgO, CaO, and TiO₂ contents at depths of 0–280 m relative to the rest of the well. In contrast, Al₂O₃ contents remained relatively constant with depth (∼13 ± 1.5%).

A global decrease in LOI with depth was found for well Glc1, whereas LOI was found to increase slightly with depth in well F1. In both wells, Fe₂O₃ peaks were associated with TiO₂ peaks (e.g., at 250 m and 280 m for Glc1; at 616 m, 894 m, 954 m, and 1016 m for F1).

Sample 15 from well F1 is totally different from the surrounding basalts, with SiO₂, CaO, and LOI values of 31.86%, 30.28%, and 25.87%, respectively.

On a total alkali–silica (TAS) diagram (Fig. 8), the F1 well cuttings are concentrated in the area of basalts with relatively homogeneous SiO₂ concentrations (44% < SiO₂ < 47%). However, some of the samples exhibited higher silica concentrations and could be described as hawaiite-type trachybasalts or andesitic basalts (SiO₂ > 50%). All of the basalts are distributed close to the straight line separating alkaline basalts from tholeiitic basalts, with samples from proximal levels in the tholeiitic domain and the deeper samples in the alkaline domain. A few samples exhibited SiO₂ concentrations below 45%, placing them closer to picro-basalts and basanites, which are more basic and richer in normative olivine.

The data from the Glc1 well cuttings are more heterogeneous, despite remaining predominantly within the basaltic domain. Samples from shallower levels tend towards alkaline basalts, whereas those from deeper levels are typically tholeiitic basalts. However, point 10, from 548 m, is particularly rich in SiO₂ (66%) and is therefore more acidic and located within the dacite zone.

In addition to the major primary minerals, such as plagioclase (mainly labradorite and bytownite), clinopyroxene (solid solution between diopside and augite), the XRD analyses revealed the appearance at depth of secondary minerals (Fig. 9). Some minerals, such as olivine or zeolites, which are present in low amounts, were difficult to detect due to poor because the signal-over-noise ratio is too low to distinguish their weak concentrations. In addition, many peaks overlap each other.

In well F1 (37 samples; XRD patterns in Supplementary material), the secondary minerals found include smectite at 208 m and calcite at 464 m; quartz was detected in sample 21 at 518 m, chlorite at 536 m, and epidote at 854 m. Pyrite began to appear at 694 m, although in very low concentrations. Magnetite-rich sections were found between 518 m and 616 m depth and from 894 m to 1118 m. Zeolites were very rare.

In well Glc1 (10 samples; XRD profiles in Supplementary material), kaolinite, calcite and quartz are present in the samples extracted from the top of the well. No secondary minerals were observed between 250 and 280 m. Analcime zeolites were observed in the first 160 m of the well, whereas smectite began to appear below 280 m. Secondary minerals typical of higher temperatures, such as chlorite, epidote, and pyrite, were not detected by XRD.

⁵⁷Fe Mössbauer analyses were performed on 20 samples from well F1 and 10 samples from well Glc1 (Tab. 2). These data allowed the Fe²⁺/Fe³⁺ ratios in the rocks to be determined and allowed iron-bearing minerals to be identified. The main minerals found were ferrihydrite (Fe³⁺₅HO₈.4H₂O), clinopyroxene ((Ca,Na,Mg,Fe²⁺,Al,Ti)₂(Si,Al)₂O₆), olivine ((Fe²⁺,Mg)₂SiO₄), ilmenite (Fe²⁺TiO₃), hematite (Fe³⁺₂O₃), magnetite (Fe²⁺O. Fe³⁺₂O₃), and maghemite (γ-Fe³⁺₂O₃) in the first part of the well (i.e., depths up to 464 m). At greater depths, a Fe-rich chlorite like chamosite (iron end-member of chlorite: (Al,Fe²⁺,Fe³⁺,Mg)₆(Si₃Al)O₁₀(OH)₈) was found in place of olivine and maghemite disappeared (Tab. 2).

The phase mixtures in these rocks are quite complex and the Mössbauer spectra exhibit various components (quadrupole doublets and magnetic sextets) associated with different Fe-bearing minerals. The analyses were carried out at room temperature; accordingly, there is some uncertainty in the attribution of some components. Indeed, the Fe³⁺ doublets characterized by isomer shift (δ) values ranging from 0.35 to 0.38 mm/s and rather large quadrupole splitting (0.60 ≤ Δ ≤ 0.84 mm/s) could correspond to either ferrihydrite or Fe³⁺-chlorite (Fig. 10). Moreover, this component often exhibits broad resonance lines (Г > 0.60 mm/s) that could hide a low-intensity doublet associated with pyrite (Fe²⁺S₂) which is characterized by δ ≈ 0.30 mm/s and Δ ≈ 0.5 mm/s. Similarly, the ⁵⁷Fe Mössbauer hyperfine parameters of the Fe²⁺ doublets due to olivine and Fe²⁺-chlorite are very similar. Further analysis at different temperatures would allow these Fe³⁺ or Fe²⁺-bearing phases to be distinguished. This study has taken the approach of interpreting these doublets and sextets considering the results of XRD analyses and especially, the detection of chlorite XRD pattern in the diffractograms. Thus, the concentration of chlorite is likely to be overestimated in comparison with olivine and the proportion of pyrite likely underestimated.

Fe³⁺ was found to be carried by ferrihydrite, maghemite, and hematite at depths of 0–464 m but primarily by Fe-rich chlorite and magnetite at greater depths (Tab. 2). An additional threshold was observed at 518 m, where ferrihydrite, maghemite, and Fe-clinopyroxene were replaced by Fe-chlorite. Despite the highest concentrations of Fe_total being observed below 464 m depth, the percentage of Fe³⁺ in the rock was significantly lower (24–55%) at greater depths than at depths less than 464 m (63–77%; Fig. 11).

Here, we discuss how our results allow us to better define the dihydrogen system of the Asal–Ghoubbet rift; this includes inferring the H₂ source rock, migration pathways, reservoir rock, and a seal in the Fiale area.

We produced simplified lithological logs (Fig. 12) by comparing our XRD and bulk rock analysis results; the resolution of these logs is based on the spacing of the well cuttings, i.e. 20–50 m for F1 and 20–90 m for Glc1. Unsurprisingly, the rocks in these two wells are mainly basalts. These basalts are transitional, essentially composed of plagioclase (solid solution between labradorite and bytownite) and clinopyroxene (solid solution between diopside and augite). They can be divided into two domains: basalt with a tholeiitic tendency (0–464 m for F1) and basalt with an alkaline tendency (464–1118 m for F1) (Figs. 7 and 8). The first domain corresponds to the basalts of the Asal Series (1 Ma to present) and the second to the Afar Stratoïd Series (4–1 Ma). The basalts of the Asal Series are richer in plagioclase so richer in CaO%.

The transition between the two series, although marked in both wells, is less pronounced in well Glc1. Indeed, the composition of the lavas here are known to have varied with time, with the basalts of the Asal Series along the external rift margins exhibiting compositions that are more intermediate than tholeiitic (Stieltjes et al., 1976). In our dataset, the transition from the basalts of the Asal Series to the Stratoïd Series was observed to occur at 280 m. The Asal Series is thicker along the rift axis.

Basalts with a Fe₂O₃ (total iron) composition above 15% (Ox%) are considered to be ferrobasalts (Stieltjes et al., 1976) and are associated with high TiO₂ concentrations (>3.48%) (Tab. 1). Many of these basalts are rich in iron (>10%) and titanium (>2%) but cannot be classed as ferrobasalts because their Fe₂O₃ content is lower than 15%. This correlation suggests high levels of ilmenite and titano-magnetite (Fe²⁺(Fe³⁺,Ti)₂O₄). SiO₂ peaks are often correlated with Na₂O + K₂O peaks, corresponding to more acidic rocks such as andesite basalts or trachybasalts.

LOI can be used as a proxy for the alteration rate of basalts; here, many samples exceeded 5%, which indicates argillization/chloritization of the rock. At shallow depths in well Glc1 (160–200 m), LOI values above 15% are considered to indicate hyaloclastites, which were formed under seawater. Conversely, the first cuttings from well F1 (i.e., from depths up to 250 m) exhibited LOI values lower than 2.5%. These basalts can be considered unaltered or slightly altered; the XRD data support this interpretation, with less pronounced clay signals.

Finally, a sedimentary layer was observed at 300 m depth in F1; this layer was composed primarily of calcium carbonates and quartz. Thus, the high Fe³⁺ concentration in the hyaloclastites at the margins indicates almost complete oxidation of the submarine basalts, in contrast to the open-air conditions along the axis. The underlying Stratoïd Basalts, richer in total iron and also more altered, indicate a reconcentration of iron after alteration.

Secondary minerals provide information about the alteration degree of the basalts and the physicochemical conditions at depth. Based on our knowledge of conditions and the mineral phases present, it is possible to consider potential redox reactions leading to H₂ formation and to determine whether a rock has already produced or could produce H₂.

For well F1 (axial zone), temperatures reach a maximum of 210 °C at 694 m depth, and 130 °C at 400 m for Glc1 (external margins). Then, in the axial zone, temperature decreases from this 210 °C maximum to approximately 60 °C (see Tab. 1) owing to marine water intrusion (Zan et al., 1990; Varet, 2014).

For well F1, we observed the successive appearance of smectites, calcite, quartz, chlorite, and then epidote with increasing temperature (Fig. 13); this is similar to the characteristics of well A5 as described previously (Zan et al., 1990). Indeed, these minerals can be utilized to define temperature domains according to their dominance (Franzson et al., 2008): 50 to 200 °C for smectite–zeolites; 220 °C to 240 °C for chlorite; and 240 °C to 280 °C for the chlorite–epidote mixture. Chlorite was first observed by XRD and Mössbauer spectroscopy at 190 °C, according to the well temperature profile. However, high-temperature minerals such as chlorite and epidote are not affected by this sudden geothermal water temperature drop. Therefore, it can be inferred that the secondary minerals are not currently in equilibrium with the fluid.

Secondary magnetite is a good proxy for H₂ production, but we assume that the titano-magnetite is principally of primary origin here, because there is a strong and uniform positive correlation between TiO₂ and Fe₂O_{3 (Total)} in our data (Fig. 14). Little pyrite was found in our samples. For well Glc1, a clear zeolite layer is evident in the first 160 m of depth. Deeper than 310 m, smectite and mixed-clay layers start to appear.

When considering H₂ production, both the observation of these alteration minerals and the alteration ratio are key to constraining potential cap rocks at depth. Despite the presence of a clay-rich layer around 850 m in well F1, it has not been possible to confirm that this layer is a good seal based on the LOI values (Tab. 1 and Fig. 7). However, at the external margin of the rift in well Glc1, fully weathered layers were observed over almost the first 400 m of depth; despite the presence of some unweathered layers, these locations may represent a suitable cap rock.

As already stated, continuous H₂ flow has been measured at the surface next to the Fiale Caldera, and some H₂ content has been detected in wells F1 and A5. The relative H₂ concentrations have been found to be 0.57–0.89% of the dry gas phase. This concentration is higher than that measured at depth in well A3 (0.25% of the dry gas phase) and also higher than a purely volcanic degassing such as that during the Ardoukoba eruption in 1978 (0.145% of the dry gas phase; Allard et al., 1979). After the volcanic degassing, this surface fumarolic H₂ may have been enriched by other processes, including redox reactions between iron-rich mineral phases and other liquid or gaseous phases.

Microbial communities in continuous H₂ flow have been studied in order to characterize their in situ taxonomic diversity and their potential to use H₂ as a source of energy and electrons from enrichment cultures. Indeed, it is known that microbial ecosystems can be hydrogen-based (Stevens and McKinley, 1995) with sulfate, carbon, dioxide or oxygen, for example, as acceptor electron. Given the environmental conditions at this site, in particular the high temperatures, our study of taxonomic diversity unsurprisingly revealed the presence of microorganisms known to be extremophiles; this includes the archaea, which are very well represented but not very diversified at this location. All of the archaea identified belong to the orders SCGC AB 179 E04 and Geoarchaeales, which have been described in thermal springs in the US (Havig and Hamilton, 2019). Similarly, the bacteria identified also belong to taxonomic groups already known to populate hot springs, such as the Pseudonocardiaceae (Ningsih et al., 2019), Kapabacteriales (Guajardo-Leiva et al., 2021), Thermoflexaceae (Dodsworth et al., 2014), Thermomicrobiaceae (Al Saade et al., 2022), Ktedonobacteraceae (Arif et al., 2021), Nostocaceae (Ansary et al., 2017), Brevibacillaceae (Inan et al., 2016), Paenibacillaceae (Panosyan et al., 2020), and Hydrothermae (Smrhova et al., 2022). Much affiliation remains vague at the order or even class level, since hot springs are extreme environments from which few microorganisms can be cultured and studied (López-López et al., 2013; Marín-Paredes et al., 2021); this justifies the need to multiply the studies of these extreme environments In order to better understand them. Our results illustrate variation in taxonomic diversity both spatially (i.e., between Site 4 2021 [zone 13], Site 4 2021 [zone 15], and Site 4 2021 [zone 16]) and, particularly, temporally (i.e., between 2019 and 2021). This seems to suggest that unstable environmental conditions are strongly influencing the structuring of communities. Most of the microorganisms identified are aerobic and the presence of cyanobacteria and chloroplasts (algaes) suggest the influence of oxygenic photosynthesis as a key metabolism of two of the communities (Site 4 2019 and Site 4 2021 [zone 16]). The dominance of aerobic microorganisms explains the absence of functional groups such as sulfate-reducers and methanogenic archaea, which harbor numerous hydrogenotrophs. The absence of methanogens is consistent with the non-detection of methane during gas analyses. Laboratory tests have shown no H₂ consumption in either anaerobiosis or aerobiosis. The absence of photosynthetic or hydrogenotrophic autotrophic organisms raises the question of maintenance in such ecosystems, in particular for the communities at Site 4 2021 (zone 13 and zone 15). However, some members of the Nitrososphaeria class have been shown to be thermophilic autotrophic ammonia-oxidizers (Hatzenpichler et al., 2008). This capacity would explain the very strong dominance of these kinds of archaea (up to 79% of the relative abundance), since they would be key players in the microbial communities, capable of using ammonium as an electron donor, oxygen as electron acceptor, and CO₂ as a carbon source.

It remains to evaluate the internal reactions of the iron-rich minerals of different basalts in order to conclude whether enrichment of the H₂ phase at depth is likely to be possible. Based on the Mössbauer data, we were able to establish remnant H₂ generation using the calculations of Geymond et al. (2022). These results give an idea of the maximum H₂ potential of the rock if the corresponding minerals undergo the associated reaction. From the reactions presented in the introduction and based on the minerals present in the basalts of the Asal–Ghoubbet rift, two main groups of reactions can be isolated: pyritization reactions involving H₂S oxidation and Fe³⁺ reduction; and reactions involving H₂O reduction and Fe²⁺ oxidation.

Fe³⁺ reduction and pyritization:

Fe²⁺ oxidation:

According to Figure 15, H₂ production is low in the upper part of the wells, i.e., in the basalts of the Asal Series, with only pyritization processes occurring (up to 90 mmol/kg rock). Conversely, H₂ concentrations are much higher in the Stratoïd Basalts. Pyritization of magnetite (up to 486 mmol/kg rock) and oxidation of chlorite (up to 1332 mmol/kg rock) are the two processes leading to the highest H₂ production. It can be seen that olivine alteration and hematization do not offer high hydrogen generation potential in this setting, with olivine content being very low in these basalts. The presence of iron-rich chlorite at depth also offers interesting potential for the production of H₂, if of course all of the Fe²⁺ becomes Fe³⁺. To investigate this possibility further, we assumed the maximum point at 1332 mmol/kg of rock to be equivalent to approximately 32.04 L/kg of H₂ (at 20 °C and 1 atm). If we consider a redox cycle where iron is both reduced and oxidized under optimal conditions, where all iron reacts to form H₂, the remnant H₂ production of these basalts from the F1 and Glc1 wells is, respectively 211.5 and 197 mmol/kg for the Asal Series basalts and 1192 and 304.2 mmol/kg for the Stratoïd Series basalts.

This difference between the two series can be explained by the high Fe³⁺ content of the basalts of the Asal Series, even though the alteration index is lower. The concentrations of magnetite and hematite (maghemite and ferrihydrite, intermediate iron oxides), where Fe³⁺ is contained, are lower than Fe²⁺-minerals in basalts; accordingly, the reduction reactions of this iron that can generate H₂ are lower than the oxidation reactions. It can be argued that the latter basaltic units were formed in a more oxidizing environment in direct contact with the atmosphere. However, in the Asal Series, the remaining Fe²⁺ is mainly concentrated in Fe-clinopyroxene, which is has much slower H₂ formation kinetics than olivine or oxides and does not easily integrate Fe³⁺ in its structure.

A hydrogen system, as in a petroleum system, is defined by the source rock, reservoir, a seal and the migration pathways between these elements. Here, we discuss likely candidates for each of these components.

There is not a single source; we were able to distinguish a primary volcanic source produced by degassing of the magma chamber under the Fiale Caldera, with a summit identified at 3–5 km depth in the rift axis (Doubre et al., 2007). In volcanic areas, the major gases present are H₂O, CO₂, and S phases (Giggenbach, 1996). Thus, during degassing, H₂ is controlled by the SO₂–H₂S equilibrium and formed mainly via the reaction H₂O + H₂S = SO₂ + H₂ (Klein et al., 2020). Little H₂S has been observed at the surface in fumaroles (Pasquet et al., 2021), suggesting that it has interacted with water to form H₂ and SO₂ or sulfides. Pyrite is observed in our sample 30 and in all other Asal wells (Zan et al., 1990). Otherwise, because SO₂ and H₂S are extremely soluble in the aqueous phase, SO₄ may have been formed.

The two other possible sources are related to rock–fluid interactions. The remnant potential of these sources has been calculated for an ideal case where the whole iron-rich mineral undergoes a redox reaction. This provides an indication of the maximum potential remaining for H₂ generation. Fe³⁺ reduction (or pyritization), as described in the previous section, involves iron minerals such as hematite (and maghemite) and magnetite reacting with H₂S partly from degassing to form pyrite and H₂. The presence of pyrite as revealed by XRD and Mössbauer analyses (Fig. 9) supports this hypothesis, although the small quantities measured suggest that this is not the major source of H₂ (or the source with the greatest potential). In sample 30 from well Fiale 1, Fe²⁺-pyrite represents 17% of the iron content (Tab. 2), with a total iron concentration of 8.62 wt% in the sample, pyrites represent 1.47 wt% of the sample.

Secondly, hydrothermal reactions involving olivine, magnetite, or chlorite alteration by water reduction are associated with the oxidation of Fe²⁺ to Fe³⁺. Although not measured (H₂ content) at depth in the rift axis wells during drilling, this source (and particularly the oxidation of Fe-chlorite in the Stratoïd Basalts deeper than 552 m) has the greatest hydrogen generation potential (with an average of 1192 mmol H₂/kg rock in the Stratoïd Series basalts) (Fig. 15). Dehydrogenation chlorite processes occur systematically and differ based on the chlorite type, one oxidize Fe²⁺ gives one H⁺ (Steudel et al., 2016). Optimum dehydrogenation for Fe²⁺-rich chlorite, like our case, were calculated on a range between 390 °C and 410 °C (Lempart et al., 2018). This zone corresponds to the geothermal propylitic zone (chlorite–epidote assemblage). Olivine does not play a major role here, in contrast to what has been observed at the Mid Atlantic Ridge or in other ophiolitic contexts (Vacquand et al., 2018). In the Asal Series, according to the cuttings, olivine represents 0.1 to 0.5 wt% of the rocks and 1.6 wt% for sample 30. However, in terms of volume, it should be noted that the volume of H₂ that could be generated in this area is considerable. Based on an average value of 500 mmol H₂/kg, i.e. 1 g H₂/kg of rock (by averaging the values obtained from the reactions in Fig. 15) over a thickness of 500 m (according to the occurrence of chlorite by XRD in Figure 13, to be in the propylitic zone in well Fiale 1),. Even a “small” area, such as the Fiale Caldera between Ghoubbet Bay and Lake Asal (∼25 km²), could produce up to 20 Mt of H₂. Globally, one may then consider that the main focus of future exploration should be the detection of reservoirs and seals.

A good reservoir must have a well-connected porosity and some permeability.

As already noted, different volcanic units have been identified at depth, ranging from the Asal Series Basalts to the Stratoïd Basalts and Dahla Basalts (Varet, 1978). Although mineralogically similar, these transitional fissural basalts have a rather high primary porosity of approximately 10% (Aquater, 1989). This porosity can be explained by degassing during the emplacement of the lava flow but it does not imply that it is connected. Pasquet et al. (2021) also showed that there was porosity in the vicinity of the phenocrysts in the microlithic texture, through which the primary gases could flow forming these flames like hematite bodies in olivines.

Physical parameters including porosity and large scale permeability were measured during the various drilling campaigns. In the rift axis, at wells A4, A5, F1, F2, and F3, permeability was found to be rather low (Houssein and Axelsson, 2010), with values <1.5 mD in well A5 (Elmi, 2005). Nevertheless, cold fluid circulation (∼60 °C) has been observed at depths between 600 and 1200 m in the Stratoïd Basalts, in the axis wells (Zan et al., 1990), and has been interpreted as marine intrusion. Greater permeability has been measured at the outer margins, in wells A3 and A6 (Abdillahi, 2014): between 10 and 50 mD in well A3 (near well Glc1) at depths of 1016–1316 m which corresponds to the propylitic zone according to the XRD (Elmi, 2005; Houssein and Axelsson, 2010), and thus within the Dahla Series (D’Amore et al., 1997). This permeability originated from the fractures and the interbeds between the lava flows (Zan et al., 1990). Temperatures measured in these permeable zones are approximately 260 °C (Zan et al., 1990; Elmi, 2005). Furthermore, the extensional activity of the rift induces local high vertical permeability, as evidenced by the surface fumaroles on the major rift faults (Pasquet et al., 2021).

The various wells studied to date, both in the rift axis and along the margins, exhibit a strong increase in temperature associated with low resistivity (1.9 Ω.m) at depths between 240 and 600 m (Zan et al., 1990; Varet, 2014). This layer is interpreted as a relatively continuous clay-rich layer that may represent the seal of a geothermal reservoir (Varet, 2014). Figure 13 highlights the sharp increase in temperature from 30 °C to 210 °C over these 360 m, that means a low thermal conductivity zone. It is in agreement with the rheology since the basalts have a conductivity around 3 W.m⁻¹K⁻¹ and the shales around 1.7 W.m⁻¹K⁻¹. In the likely hypothesis of a thermal field mainly due to the fluid advection, cold as hot, this sharp increase suggests, as the low resistivity, a low permeability zone. This thermal field is not in steady state as confirmed by the fact that the cold zone is a consequence of the Ardoukoba event (1978–1987 period), also recent magmatic injections between 1991 and 2001, resulted in some heating events (Varet, 2014). The XRD and bulk analysis results reported in this study indicate the presence of smectite at these depths. The LOI results confirm this smectite-rich horizon in well F1 along the rift axis, at depths between 0 and 500 m (LOI_smectite < LOI_chlorite [Sun et al., 2009]). In Glc1, a very altered layer dominated by hyaloclastites was observed at the top of the well with smectites below, with LOI decreasing progressively to 500 m depth (the limit of the well). With a thermal cap, a very low gradient zone is observed, at 130 °C (Fig. 13). In the literature the geothermal argillic zones are characterized by a dominant smectite composition with associated minerals such as calcite, zeolites or hematite, marked by temperatures below 180 °C and resistivities below 10 Ω.m (Henley and Ellis, 1983; Stimac et al., 2008). It corresponds to the geothermal reservoir seal.

Thus, despite the notable contribution of argillization to decreasing porosity and permeability, we cannot confirm that a seal for a geothermal reservoir is also a seal for a H₂ reservoir. But if so, such a seal would be more likely present in well Glc1, along the outer margin, over 450 m deep (LOI > 11%) than in well F1 and therefore in the rift axis.

Resistivity and temperature analyses have indicated a potential seal between 0 and 500 m. The main issues yet to be resolved include the definition of the minimum clay content required to form a H₂ seal and the confirmation of whether a geothermal reservoir seal can be considered equivalent to an H₂ seal. So far 3 types of hydrogen reservoir seal have been described, namely the dolerite in the Mali wells (Prinzhofer et al., 2018), the clays of the Mid-Rift system in Kansas (Coveney et al., 1987; Guélard et al., 2017) and the salt layer in the Amadeus basin (Leila et al., 2022).

The poorly connected porosity and low permeability of basalts makes the circulation of gases complex. Gases will typically seek preferential paths along major or minor active faults (i.e., rather vertical permeability; Fig. 16) or along the strata boundaries of basalt flows (i.e., horizontal permeability). The presence of vertical fast migration pathways is evidenced by the fumaroles on the surface along the faults and the absence of H₂ in the basaltic floor of the rift axis (Fig. 2) (Pasquet et al., 2021).

Many authors working on MORs have highlighted the differences between the basalts and the olivine-rich upper mantle in terms of H₂ generation and have shown that hydrothermal vent fluid composition is largely controlled by oceanic lithosphere composition (see, for instance, Wetzel and Shock (2000) and references therein). Studies of the evolution of fresh basalts with depth at MORs are less numerous and the one presented here both allows a closer view of alteration over the first few million years and discusses the influence of deep fluids versus those arising from water/rock interactions within the final gases.

Our data proves that the gases in the Asal–Ghoubbet rift are still mainly volcanic in origin, corresponding to the degassing of the magmatic chamber (Holloway and O’Day, 2000); CO₂ makes up the majority, but H₂S reduction is likely to generate some H₂ in that process. The elevated heat flows in the rift induce rapid circulation of water and steam, which enhances basalt weathering. The resulting water/rock interactions will result in water reduction, although low quantities of olivine are present in the basaltic series of the Asal Rift. However, the alteration of basalt leads to the formation of clay; the resulting Fe-rich chlorite may then be oxidized and generate additional H₂.

The alteration profile of the basalt of the Asal–Ghoubbet Rift is consistent with typical alteration in high-temperature geothermal fields. Furthermore, the first hundred meters of rock below the surface are already altered at the rift margins, with LOI reaching 17% for basalts with ages of less than 1 Ma, and slightly weathered along the axis. This indicates that seawater was present and then removed during the accretion of the subsequent basalt flows. In this context, we have provided an overview of underwater, meteoric, and atmospheric alteration/oxidation over depths of a few kilometers. Here, the contacts between the basalt and the air and the oxygen-rich seawater have exerted a strong influence over this evolution.

The data recorded from Site 4, an emitting fracture related to the major faults that border the rift to the southwest, indicate the presence of H₂. The concentration of H₂ here varies from 400 to 1000 ppm, although our sensor saturated at 1000 ppm; after air correction, our data indicate 0.5 to 3% H_2. This value is not uniform; however, global variations in the value are smaller than the variations detected in soil affected by microseepage of H₂ in the continental domain (Prinzhofer et al., 2019; Moretti et al., 2021, 2022). This is coherent with the differences between micro- and macro-seepages observed for the natural gas, which reflect differences in the mode of transport of hydrogen through soils and fractures (Schumacher, 1996). The velocity of the fluid within fractures is greater than that in soil; faster fluid flow reduces the potential for interaction and, in our case, reduces the rate of H₂ consumption by chemical or microbiological reactions. This means that a few measurements may be sufficient to determine if a fracture emits H₂ and that long-term monitoring is not mandatory, as it is for the H₂-emitting fairy circles in the cratons.

In this case study, microorganisms can be considered to have an indirect impact on H₂. Indeed, it is known that NH₃ and NH₄ are present in volcanic gases and that, via interaction with water and/or iron, redox reactions occur and can form both N₂ and H₂. The presence of these gases in fumaroles may explain the presence of ammonia-oxidizing archaea; further investigation could confirm the mechanisms underlying the high N₂ content of the emission zone at Site 4 (if we assume that the balance consists primarily of N₂).

The gas data demonstrates the presence of H₂ in the well on the margin of the rift. However, although the fumaroles at the center of the rift contain H₂, no H₂ was detected at depth in this area (i.e., in the A5 and F1 wells) during drilling, because of the absence of measurements. At the opposite, on the margin near Glc1, H₂ concentrations were very low at the surface. The other difference between the central axis and the margins relates to the presence of a geothermal gradient inversion due to the flow of seawater from Ghoubbet Bay to Lake Asal. In addition, alteration of the basalts along the rift margins led to the formation of hyaloclastites and smectites, which may constitute a potential seal for both geothermal reservoirs and H₂. Although diffuse, the H₂ potential (in terms of theoretical volume) is large: more than 500 mmol H₂/kg rock. The majority of this H₂ may have been lost by continuous leakage to the surface. However, accumulations (even temporary ones) may still occur in cases where a reservoir and seal are present. Such a reservoir would be preferentially located at depths of 1000–1300 m on the southwestern rift margin, in the Dahla Basalts, zone with the best permeability (10–50 mD). This zone is located in the propylitic geothermal zone (chlorite–epidote assemblage) and it’s also a potential source of H₂ through Fe²⁺ oxidation and chlorite dehydrogenation (Fig. 16). The outer margins of North Ghoubbet can also be considered prospective in this context but data are still scarce in this area without well.

Thermodynamic calculations must now be carried out to better constrain the reaction kinetics in these ferrous chlorites. Furthermore, despite the surface and well data on H₂ in the margins, can this poorly connected porosity make these basalts a good reservoir. Further experimental studies are also needed to improve the documentation of H₂ seals.

G.P.’s PhD was funded by the Isite E2S; the fieldwork was funded by the Carnot ISIFOR. Amin Mohamed Idriss is now studying in UPPA thanks to Campus France.

We thank Isotope Editing for editing a draft of this manuscript. We thank ODDEG for sharing with us their knowledge of Djibouti geology and for allowing us access to the cuttings from the two studied wells. This collaboration, ODDEG-UPPA, was initiated by K Moussad, J Varet, and I Moretti.

I would like to thank Alain Prinzhofer, Yves Geraud and an anonymous reviewer for their constructive comments which helped to clarify this article.