Desulfohalobium retbaense DSM 5692, a halophilic sulfate-reducing strain, was used as the model bacterium in this study. It can grow at pH ranging from 5.5–8.0, optimum temperature of 37°C and optimum salinity of 12% (Ollivier et al., 1991). Under anaerobic conditions, Desulfohalobium retbaense can utilize H₂ as an electron donor and sulfate as electron acceptor producing H₂S, for growth (Spring et al., 2010; Tinker et al., 2022). The organism was grown in the DSM 499 high salinity growth medium with some modifications: Base medium was: 1 g/L NH₄Cl, 0.3 g/L K₂HPO₄, 0.3 g/L KH₂PO₄, 20 g/L MgCl₂⋅6H₂O, 100 g/L NaCl, 2.7 g/L CaCl₂⋅2H₂O, 4 g/L KCl, 3 g/L Na₂SO₄, 1 mL/L trace element SL-10, 0.5 mL/L Na-resazurin solution (0.1% w/v), 0.3 g/L Na₂S⋅9 H₂O. 1.5 mL of a dense bacterial solution was inoculated in 15 mL of base media amended with 80 mM sodium lactate, 40 mM sodium acetate, 0.3% yeast extract and 0.3% peptone solution at 37°C for 7 days under anaerobic conditions (N₂ headspace). This 7-day old culture was then used and injected into the micromodel.

The experimental setup consists of a high-pressure micromodel with a Zeiss microscope (Axio Zoom. V16, Zeiss) illuminating with a S ring cold-light (CL 9000 LED) source. This microscope setup was able to capture the whole pore network (total 121 separate images) with a resolution of 4.38 µm/pixel and acquisition time of 73 s. Pore space temperature was kept constant at 37°C ± 0.5°C by circulating warm water through internal copper tubes in the chip holder. Experimental pressure was controlled at 35 bar by a high precision plunger pump (Quizix Q5000-10 K), and a back pressure regulator (EB1ZF1 Equilibar Zero Flow) connected to a pressurized 300 mL N₂ cylinder. The experimental setup has a high-precision injection system with a constant dead-volume (10 µL before the pore network). The silicon micromodel capable of withstanding pressure up to 150 bar, was designed based on a thin section from a natural sandstone and the production procedures are found in (Buchgraber et al., 2012). The pore network constitutes of 36 repetitions of a unique pattern (Figure 1). The pore network has a porosity of 0.61 and pore volume of 11.1 µL. For image segmentation, the pore network was split into 288 units and units have a porosity range between 0.59 and 0.66. Three areas of interest (a unique pattern) have been identified at inlet, middle and outlet sites (blue rectangles in Figure 1A). The bottom silicon and top borosilicate glass have strongly hydrophilic surfaces due to the fabrication procedure where silicon is oxidized to SiO₂. Two open channels connect port 1 to 2 and port 3 to 4 across the injection and production sides of the micromodel. A more detailed description of the experimental setup and micromodel properties is found elsewhere (Benali et al., 2022).

The experimental procedure consists of the following steps:

1. Preparation: The pore space was flushed with ethanol, deionized (DI) water, and H₂O₂ (ACS reagent, 30 wt% solution in water, Thermo Scientific™) to clean all surfaces and waste from previous experiments, followed by > 100 pore volumes (PVs) of DI water.

A baseline, sterile experiment (i.e., without microbial cells) was performed with the experimental protocol detailed above. Each experiment was repeated twice with comparable results, and we discuss the main results using a single microbial experiment in Section 3 below. The repeated experiment demonstrates the same governing mechanisms for wettability change, microbial-induced H₂ loss and disconnection, although with higher initial H₂ saturation and lower biofilm saturation. The reader is referred to the Supplementary Material for more detail.

Microbial growth in porous media very often leads to biofilm formation and microbial-induced pore-clogging (Liu et al., 2019), and this was observed during bacterial inoculation and H₂ drainage processes in this work. Biofilms formed predominantly in the inlet area in less than 0.1% of the pore space after 18 h of growth. This is an expected result because D. retbaense is not a strong biofilm forming bacterium. We observed that some biofilms, however, blocked the flow of H₂ gas (Supplementary Figure S1) and reduced gas injectivity occurred. For instance, during H₂ drainage, the shear stress from gas flow pushed some biofilms formed in the injection line into the inlet area, and some biofilms detached from the pore network. There was, however, no new biofilm formation during microbial growth with H₂ gas and the biofilm saturation (S_b) reduced by 24% compared to S_b after drainage and culturing in H₂ for 9 days (Figure 2). Our hypothesis is that the biofilm reduction arises from a transition of bacterial lifestyle from biofilm-mode to planktonic-mode in a H₂-rich environment (Jefferson, 2004; Chua et al., 2014; Liu et al., 2019). This hypothesis is corroborated by earlier work (Liu et al., 2019) where biofilm detachment occurred in high nutrient environments, but cannot be generalized for growth on H₂ based solely on the reported results: whether the lifestyle change is due to either reduced EPS formation or an active detachment process, and if whether cell numbers increased during this process, need to be investigated in future experiments. Our observations are only valid for the bacterial strain studied. Different microbial strains need to be investigated in the future to verify if this effect of H₂-induced biofilm detachment is common for many bacteria.

In this study, the main microbial process with UHS implications is hydrogenotrophic sulfate reduction (Ebigbo et al., 2013; Hemme and Van Berk, 2018; Gregory et al., 2019): S O 4 2 − + 4 H 2 + H + → H S − + 4 H 2 O

The sulfate-reducing bacterium (D. retbaense) used H₂ (aq) as a source of electrons to reduce sulfate and generate H₂S. Figure 3 plots the changes of gas saturation in bacterial (solid lines) and sterilized (dash lines) experiments at inlet, middle and outlet areas (cf. Figure 1). Biofilm formation in the inlet area reduced gas injectivity (cf. Section 3.1) and initial H₂ saturation (S_g0) for the inlet area (S_g0 = 0.46) compared with the sterilized experiment (S_g0 = 0.60). Note that all plots start from the end of H₂ drainage. After 2 days of microbial growth with H₂ gas, the H₂ saturation (S_g) decreased by 29.4% due to the bacterial consumption, compared with 3.7% for the sterilized case (reduction in the sterilized case is believed to be a combined effect of system leakage and measurement uncertainty). In the following 4 days, the reduction of S_g in the bacterial experiment was low (5‧10⁻³, and similar to the sterilized case), indicating less microbial activity. At the end of experiment (9 days), 32.9% of the initially stored H₂ gas was consumed by the microbial cells. The inlet area shows a constant bacterial consumption rate (reduction of S_g over time) of 0.018 per day (constant rate for entire experiment), which is lower compared to the middle (0.03) and outlet (0.025) area for the first 2 days. On average, the reduction in H₂ saturation was one order of magnitude larger for the bacterial case (0.024) relative to the sterilized case (0.0029). Similar results were obtained from the repeated experiment (Supplementary Figure S2A) that the consumption rate in the first 2 days is much faster than the following days. Note that the initial H₂ saturation for the inlet, middle and outlet was not the same, and the differences of consumption rate were related to the distribution of bacterial solution and H₂ gas in the pore network.

The spatial distribution of S_g per unit during bacterial consumption (Figure 4) showed that the gas phase in the inlet area initially was connected, compared with predominantly disconnected H₂ bubbles in the outlet area due to the effects of snap off (Singh et al., 2017; Lysyy et al., 2022). The variation of disconnected H₂ bubbles in the pore network will be detailed in Section 3.4. The microbial consumption rate in the pore network was strongly influenced by the interfacial area between gas and brine. Smaller H₂ bubbles improved the H₂ availability for microbial consumption due to i) increased H₂ dissolution and ii) larger contact area between H₂ gas and the microbial cells. We postulate that H₂ availability is a driving mechanism for the observed consumption efficiency and rate in the pore space. Rapid bacterial consumption (within 7 h) was observed in the outlet area, occupied with disconnected H₂ bubbles with large contact area. The consumption rate is also determined by activity of the microbial cells, exemplified by a drop in consumption rate (constant S_g) observed in whole pore network and middle area beyond 2 days. We attribute the observed rate reduction to changes of the microbial growing environment. We assume that both sulfate (electron acceptor) and acetate (carbon source) were still available at high concentration in the solution. The loss in activity might therefore be a result of other unfavorable conditions. Bacterial growth in a confined system will lead to formation of bioproducts and sometimes toxins accumulated in the aqueous phase. For example, increased H₂S/HS⁻ concentrations from sulfate reduction can be toxic to cells, even to sulfate-reducing cells. This, together with a noticeable pH increase during sulfate reduction, will be unfavorable for continued bacterial growth (Liu et al., 2019). We were unable to measure H₂S in the effluent due to the small volumes associated with microfluidic experiments. Hence, the mechanisms for the reduced activity need to be investigated further. Cells in the inlet area were less affected, most likely due to the presence of biofilms that are known to protect embedded cells from environmental stresses like pH change and toxins in the environment (Flemming and Wingender, 2010).

Due to the low solubility of H₂ in brine, H₂ loss from diffusion and dissolution will be small compared to microbial-induced loss (see the Supplementary Material) in agreement with previous numerical estimations (Carden and Paterson, 1979; Anikeev et al., 2021). Microbial consumption, however, is a major risk for H₂ storage at industrial scale, and specifically for the microbial sulfate reduction in our work. As sulfate compounds are common in subsurface reservoirs, deriving from the aqueous dissolution of mineral phases like anhydrite (CaSO₄) (Hemme and Van Berk, 2018), H₂ injection can trigger hydrogenotrophic sulfate reduction due to accelerated activity of sulfate-reducing microorganisms. H₂ can be microbially converted to H₂S (aq), resulting in H₂ loss (up to 32.9% in this study) and reservoir souring. Our results show that H₂ consumption rates vary over time due to changes in active microbial cells and mass transfer at the interface. Therefore, microbial induced H₂ loss pertaining to underground storage is determined by the reservoir conditions (T, p, pH), microbial cells (activity and density) and mass transfer at the interface. Previous theoretical studies have shown that consumption can vary from 0%–17% of the injected H₂, depending on the aquifer conditions and the involved microbial mechanisms (Thaysen et al., 2021). We measured H₂ consumption rates at high initial nutrient content, high initial cell numbers and optimal physicochemical conditions for the studied strain, hence, investigating a worst-case scenario for microbial H₂ loss for the oligotrophic subsurface. On the other hand, in the subsurface some effects will favor microbial consumption, for example buffering minerals (carbonates) which will stabilize pH leading to a prolonged microbial activity. Also, under certain flow conditions in the subsurface fresh nutrients and cells might be introduced into the storage area, resulting in a faster consumption rate in the late stage of gas storage compared to our confined system.

The drainage of H₂ into the pore network containing two solutions with the same salinity resulted in non-identical gas distribution (Figure 4; Supplementary Figure S3) as a result of microbial induced wettability alteration. The wettability changes were investigated by measuring the contact angle in a three-phase (gas-aqueous-grains) system, which was quantified with image analysis using open-source ImageJ software. The gas-liquid interfacial area reduced to near minimal values when the H₂ gas was exposed to the bacterial solution for 1 day (see images in Figure 5, Supplementary Figure S2B). Results indicate that the bacterial cells altered the surface wettability from an initial water-wet (41° contact angle) to a neutral-wet (average 96° contact angle), compared with an unchanged contact angle (28°) for the sterilized experiment. We assume that bioproduct and bacterial adhesion to the solid surfaces are the dominant mechanisms for the contact angle shift to increased hydrophobicity. Microbial cells growing within the water film around the H₂ gas phase can easily attach to the SiO₂ surface of the micromodel. The cell surface of microbes has wetting properties more neutral wet than the SiO₂ surface; thus, the wettability is altered depending on the nature of the bacterial cell’s membrane structure. Similarly, the increased pH due to microbial growth can also result in increased hydrophobicity, that can be explained by the protonation state of the polar molecules and, thus the number of charged surface species at the interface of H₂/aqueous/SiO₂ (Santha et al., 2019).

Wettability is one of the key parameters that determines the storage capacity and gas recovery (Bai and Tahmasebi, 2022; Pan et al., 2022) for underground gas storage. With the increase of contact angles induced by microbes, the capillary entry pressure decreases resulting in more favorable drainage displacement efficiency. The contact angle near 90° also lead to minimal interfacial surface area and favorable gas recovery due to reduction in capillary pressure (Lin et al., 2019; Lin et al., 2021). This in turn leads to decreased, snapped-off, and disconnected gas phase and an increase in H₂ relative permeability during imbibition. Hence, a large percentage of gas can potentially be recovered from porous reservoir rock when all other factors are equal. Accordingly, microbial growth induced minimal surface areas in the hydrogen-brine-SiO₂ system may enhance the gas storage efficiency of UHS operations.

During H₂ drainage, gas-phase continuity can be broken by snap-off, generating isolated gas bubbles and trapping gas in the pore space (Lysyy et al., 2022). Due to the low H₂ drainage rate (the capillary number is 2.04‧10⁻⁸ and Pelect number equals 4.76) in our study, the initial saturation of isolated gas bubbles in the pore network was less than 0.01 after H₂ injection and mainly located in the outlet area (Figure 3). As discussed above, the microbial effects of bacterial consumption and the increase of contact angle can reallocate the residual H₂ phase in the pore network (see Figure 6). Initially, H₂ preferentially occupied the large, connected pore clusters with large pore throats. After bacterial consumption for 1 day, most of the residual H₂ remained in the same pores (red) but some movements into the neighboring pores were also observed (yellow). As the capillary entry pressure was reduced with the increase of contact angle, more and more H₂ bubbles were able to enter the nearby pores with smaller pore throats. The disconnection of gas phase (S_gi) in the pore network was measured by dividing the number of pixels of disconnected bubbles with the number of pixels of pore space. The plot in Figure 6A indicates that the disconnection of the H₂ phase in the pore network increased up to 72.8% for 5 h, while the sterilized experiment increased only 22.6% and stayed at a constant value until the end of experiment.

Bacterial consumption can break the continuity of the gas phase and generate isolated gas bubbles, which further increases the contact area of H₂ and microbial cells and then accelerated the gas consumption rate. Therefore, a decrease of S_gi was observed after 7 h as a result of the fast consumption of isolated gas bubbles. The same trend was observed in the repeating experiment (see Supplementary Figure S2C). Our pore-scale results reveal that microbial consumption and wettability alteration are the two main mechanisms that determine disconnection and entrapment of the gas phase in the pore network. The increase of disconnection of H₂ gas will reduce gas mobility dramatically in the receiving reservoir (Alhosani et al., 2021). Furthermore, the blockage of gas bubbles at the entrance will increase the flow resistance (Jian et al., 2022), causing a low gas recovery. Most of the trapped H₂ bubbles in the high permeability zones will be immobile in the subsequent injection cycles, hence, are difficult to recover.

The microbial effects on disconnection of H₂ gas can be summarized as the following: the increase of contact angles induced by microbes can reduce the capillary entry pressure and decrease the snapped-off and disconnected gas phase during imbibition; the stored H₂ phase, however, can be consumed by microbial cells forming new disconnected H₂ bubbles in the pore.

The aim of UHS is to store large amounts of H₂ gas from surplus production of renewable electricity to mitigate seasonal fluctuations in energy supply and demand (Amid et al., 2016; Heinemann et al., 2021b; Dopffel et al., 2021). Because H₂ is an import electron donor for microorganisms, assessment of microbial effects for UHS is an inevitable step to estimate the potential risks for gas injectivity, H₂ loss and recovery for large-scale H₂ geo-storage operations (Anikeev et al., 2021). In our study, we investigated one of the major H₂ consuming processes in the subsurface, bacterial sulfate reduction under high salinity conditions. Overall, four microbial effects were observed in our experiments: microbial induced-clogging, H₂ loss from bacterial consumption, wettability alteration, and increased residual trapping of the H₂ phase.

Microbial-induced wettability alteration appears to be a positive effect of microbial growth in a H₂ storage reservoir: increased hydrophobicity resulted in the presence of minimal gas-liquid surface areas, and high recovery efficiency, due to reduction in capillary pressure and high gas relative permeability. In contrast, the other observed microbial effects are expected to affect adversely the storage efficiency.

Biofilm formation in the near-well regions may reduce H₂ injectivity and change the subsurface transport properties. The risk of H₂ loss from dissolution and diffusion in the brine is limited due to low solubility, but permanent loss from microbial consumption must be considered for economic reasons, particular for storage in aquifers due to the abundance of sulfate-reducing microorganisms and sulfate in the formation water (Zivar et al., 2021). The production of hazardous gases, such as H₂S would lead to contamination of the injected H₂. The increase of disconnected H₂ bubbles in the pore network will also speed up the microbial consumption process due to increased interfacial area and increase the flow resistance, resulting in a low gas recovery.

Whether all these microbial effects will be at play during H₂ storage will depend on field specific factors like the indigenous microbial community, cell numbers, available nutrients and sulfate. In two major H₂ storage field trials (Underground Sun Project, Austria, and HyChico, Argentina) microbial activity was reported as an important factor for H₂ loss, but without indications of active sulfate-reducing microbes because both reservoirs were low in sulfate and the major microbial activity was methanogenesis (H₂ + CO₂ = CH₄) (Bauer and Austria, 2017). However, a recent study showed significant microbial hydrogen consumption with sulphate reduction in original brine from a gas reservoir (Dohrmann and Krüger, 2023). The overall risk will depend on the field specific microbial communities and field characteristics. That means that further field trials need to show if the mechanisms described in our work will be observed in the field. Our results help in understanding and interpreting field test data and to avoid significant H₂ losses.