Geochemical Assessment of Long-Term CO₂ Storage from Core- to Field-Scale Models

Abstract

Numerical simulations enable us to couple multiphase flow and geochemical processes to evaluate how sequestration impacts brine chemistry and reservoir properties. This study investigates these impacts during CO₂ storage at the San Juan Basin CarbonSAFE (SJB) site. The hydrodynamic model was calibrated through history-matching, utilizing data from saltwater disposal wells to improve predictive accuracy. Core-scale simulations incorporating mineral interactions and equilibrium reactions validated the model against laboratory flow-through experiments. The calibrated geochemical model was subsequently upscaled into a field-scale 3D model of the SJB site to predict how mineral precipitation and dissolution affect reservoir properties. The results indicate that the majority of the injected CO₂ is trapped structurally, followed by residual trapping and dissolution trapping; mineral trapping was found to be negligible in this study. Although quartz and calcite precipitation occurred, the dissolution of feldspars, phyllosilicates, and clay minerals counteracted these effects, resulting in a minimal reduction in porosity—less than 0.1%. The concentration of the various ions in the brine is directly influenced by dissolution/precipitation trends. This study provides valuable insights into CO₂ sequestration’s effects on reservoir fluid dynamics, mineralogy, and rock properties in the San Juan Basin. It highlights the importance of reservoir simulation in assessing long-term CO₂ storage effectiveness, particularly focusing on geochemical interactions.

1. Introduction

In 2023, CO₂ emissions worldwide increased by 410 million tonnes, a 1.1% rise, setting a new high of 37.4 billion tonnes. This occurred despite significant progress in renewable energy and concerted global efforts to diminish carbon footprints, according to the International Energy Agency [1]. The relentless rise in atmospheric carbon continues to propel climate change, as indicated by escalating global temperatures. Earth’s average surface temperature in 2023 was the highest recorded since 1880 [2]. Geological Carbon Sequestration (GCS), the process of injecting CO₂ into underground formations for long-term storage, is increasingly viewed as a critical strategy for significantly lowering atmospheric emissions. GCS not only addresses the potential release of carbon dioxide into the atmosphere but also plays a vital role in alleviating the ongoing climate crisis.

Geological Carbon Sequestration (GCS) offers a range of solutions by exploiting various underground formations that can securely contain carbon dioxide, standing out as one of the most effective approaches for long-term carbon storage [3,4,5,6,7,8]. These formations, which include depleted oil and gas fields, unmineable coal beds, deep saline aquifers, and sites for CO₂ storage during enhanced oil recovery, significantly vary in their characteristics and suitability for storing CO₂ [6,9,10]. Each storage type presents unique benefits and challenges, influenced by geological attributes, existing infrastructure, and their capacity for securing carbon dioxide over the long term. Saline aquifers have attracted significant interest due to their widespread availability and considerable storage capacity. Typically located deep underground, these aquifers contain brine with Total Dissolved Solid (TDS) levels above 10,000 ppm, making them unsuitable for use as drinking water sources or for other economic purposes [11]. Their characteristics, including abundant availability and current underutilization, make them excellent options for extensive CO₂ sequestration efforts.

Injecting CO₂ into deep saline aquifers triggers a complex interplay of physical and chemical events [7,12,13,14]. This sequestration process is facilitated by four main mechanisms [15,16]: first, hydrodynamic trapping, where CO₂ is trapped under impermeable cap rocks either as a gas or a supercritical fluid; second, solubility trapping, which involves CO₂ dissolving into brine, increasing its acidity, and affecting the dissolution of surrounding minerals; third, residual trapping, which occurs when CO₂ remains in pore spaces as immobile bubbles held in place by capillary forces; and fourth, mineral trapping, which happens when CO₂ reacts with brine and reservoir rocks to form stable carbonate minerals. Mineral trapping is considered the most reliable and permanent form of CO₂ storage. In this process, CO₂ is transformed into inert, stable, and immobile carbonate minerals, ensuring long-term containment. However, the specific conditions required for this mineralization—particular interactions among CO₂, water, and minerals—introduce complexity to the geochemical dynamics, making the process less predictable and intricate [17]. These interactions, critical for the effectiveness of CO₂ mineral trapping, vary widely due to the unique mineral compositions, brine chemistry, and conditions of each aquifer.

CO₂ reactive transport modeling integrates multiphase fluid flow, solute transport, and geochemical reactions to evaluate the behavior of CO₂ injected into geological formations such as saline aquifers. When CO₂ is injected into saline aquifers, it dissolves in the water and forms carbonic acid (H₂CO₃), which breaks down into bicarbonate (HCO₃⁻) and carbonate (CO₃²⁻) as seen in Figure 1. This reaction significantly lowers the pH, acidifying the environment and leading to the dissolution of silicate and oxide minerals. As these minerals dissolve, concentrations of ions like calcium (Ca²⁺), magnesium (Mg²⁺), sodium (Na⁺), silica (SiO₄⁴⁻), and iron (Fe²⁺) increase in the solution [18,19,20]. This increase facilitates the secondary precipitation of silica, silicates, and carbonates [21,22,23]. As CO₂ continues to migrate within the reservoir, these geochemical changes impact the reservoir rock’s porosity and permeability.

Recent experimental studies have been conducted to examine the effects of CO₂ on rock properties and investigate the reaction kinetics of CO₂ with sandstone under both static and dynamic flow conditions [24,25,26,27,28,29,30,31,32]. These experiments help confirm the suitability of sandstone as a viable repository for CO₂. Laboratory tests allow for the control of critical variables such as pressure and temperature, providing a comprehensive analysis of the interactions between CO₂, brine, and minerals [33]. Simulation models become essential in understanding the intricate dynamics of multi-component flows within subterranean reservoirs, describing geochemical reactions, and predicting mineralogical alterations over the long-term storage of CO₂ [34,35].

Xiao et al. [36] noted a decrease in porosity at the interface between the reservoir and the caprock due to mineral precipitation, which improved the caprock’s sealing properties. Research by Dai et al. [14] suggested that CO₂ interactions with rock might bolster seal integrity through the precipitation of minerals such as dawsonite, magnesite, siderite, and illite. This research underscored the significance of the mineral composition of the matrix in determining seal integrity through mineral trapping. Sun & Bourg [37] used molecular dynamics simulations to examine water–mineral wettability and assess mineral trapping in geological sequestration. Their results highlighted that the mineralization capture process is exceedingly slow, potentially taking centuries to securely store CO₂. Simulations performed by Elgendy et al. [38] found limited calcite dissolution due to pH changes caused by CO₂ plume development and noted water vaporization near the wellbore. In a study by Ma et al. [39], reactive transport modeling on sandstone demonstrated significant mineral changes resulting from CO₂ interactions. Notable findings include the dissolution of calcite and chlorite and their re-precipitation as more stable minerals like dolomite and ankerite, emphasizing the critical role of mineral kinetics and the distribution of aluminosilicate minerals in the long-term effectiveness of CO₂ sequestration in deep siliciclastic reservoirs.

Most field-scale reactive transport simulations traditionally utilize one-dimensional (1D) models [40,41,42,43]. Contrarily, this research adopts a three-dimensional (3D) model to incorporate the complex heterogeneity of the geological structures, thereby enhancing the precision and applicability of our findings in modeling reactive transport processes. A geochemical reactive transport modeling analysis was conducted, focusing on the Entrada Sandstone injection area within the San Juan Basin, New Mexico, USA. Employing the compositional numerical simulator GEM version 2024.20 by Computer Modelling Group (CMG), this study is pivotal in evaluating the long-term sustainability of CO₂ geological storage and offers insightful contributions to the understanding of CO₂ geochemical dynamics, which are relevant to similar sandstone formations across various pilot or target sites.

This study examines the fate of injected CO₂ and assesses the effects of CO₂–water–rock interactions on reservoir characteristics, with a specific focus on brine chemistry and reservoir porosities. This paper provides crucial insights into the four main CO₂ trapping mechanisms; it explores the associated processes and pertinent equations, offering a robust understanding of each mechanism’s role. Additionally, this research includes an analysis of the history-matching process, the properties of the reservoir model, and the geochemical modeling relevant to aqueous and mineral reactions observed in this study.

This study presents comprehensive findings, including history-matching, CO₂ plume containment, pH evolution, trapping mechanisms, and the impact of geochemical reactions on reservoir properties, as well as the dynamics of primary ions in solution. This manuscript is structured to comprehensively cover the study of CO₂ sequestration within the Entrada Sandstone. Section 2 outlines the methodology employed, beginning with the hydrodynamic history match, followed by core-scale simulations for geochemical history-matching. The model was then upscaled to a field-scale model prior to CO₂ injection. Section 3 and Section 4 present the results and discussion for this study. Finally, Section 5 concludes this study, summarizing the findings, their significance, and the implications they have on the field-scale CO₂ storage project. The governing equations and underlying theories related to fluid flow through porous media and geochemical reactions are provided in the Appendix A.

2. Materials and Methods

In this section, we explore the reactive transport modeling approach employed to simulate and analyze the long-term behavior of CO₂ storage within the Entrada Sandstone formation. The initialized base case model is history-matched using observed data from nearby saltwater disposal wells to validate the simulation model to be utilized in simulating geochemical reactions, informed by brine composition and X-ray diffraction (XRD) data. Figure 2 delineates the detailed steps followed in the reactive transport modeling process used in this study for the field-scale simulations.

2.1. Study Area and Geological Model

The San Juan Basin is a structural asymmetric basin located in northwest New Mexico, covering an area of approximately 7500 square miles (19,425 km²). The San Juan Basin contains a thick sedimentary section, with sedimentary fill of about 15,000 ft (4.57 km) in the deepest part of the basin. The region is highly favorable for CO₂ sequestration. The main injection and confining zones are the Entrada Sandstone and the Summerville and Todilto formations. The Entrada Sandstone consists of fine-grained interdune sands and reworked dune deposits. This sandstone is about 140 feet thick and is made up of eolian, well-sorted, fine- to medium-grained particles. The Entrada consists of quartz, feldspar, calcite, dolomite, biotite, kaolinite, illite, and other clay minerals. The porosity ranges can be as high as 25% in the shallow portion, decreasing due to compaction and cementation in the lower part. The Todilto Limestone formation is made up of carbonate, evaporite, and anhydrite and acts as an effective seal for the Entrada Sandstone. Above Todilto lies the Summerville Formation, which can act as an additional seal. The Summerville deposits consist of thin-bedded sandstones, siltstones, and mudstones with varying depositional environments [44,45]. The Carmel formation found beneath the Entrada Sandstone formation is a reddish fine-grained siltstone, and the Wingate formation beneath it consists of flat-bedded, dark-red calcareous siltstone mixed with a thin stratum of silty sandstone.

The geological model was initially constructed utilizing Petrel software version 2024.1, after which it was imported into CMG-GEM version 2024.20 for numerical modeling. This model delineates the structure into five distinct zones, namely the Summerville, Todilto, Entrada, Camel, and Wingate formations, which are representative of the San Juan Basin.

2.2. Hydrodynamic Model History-Matching

The simulation model was history-matched using two saltwater wells within the Entrada. This history-matching covered several years and included parameters such as water injection rates and corresponding bottom-hole pressures (BHPs).

History Match Error

The degree of deviation in history-matching is calculated by comparing the relative difference between simulation predictions and actual field data for each objective function. This process assigns specific measurement errors and weightings to different types of production data from each well, influencing the calculated error for each objective function. History-matching error is thus quantified by the discrepancy between simulated results and observed field data for targeted objective functions, with data quality playing a key role. Measurement inaccuracies and importance levels are allocated to different production data types per well, contributing to the error calculation. Equations (1)–(5) were used to calculate the history-match error [46].

$${LHME}_i={\displaystyle \frac{1}{{\sum }_{j=1}^{N\left(i\right)}{tw}_{i,j}}}\times {\sum }_{j=1}^{N\left(i\right)}{\displaystyle \frac{\sqrt{\frac{{\sum }_{t=1}^{T\left(i,j\right)}{\left({mw}_{i,j,t}^m\right)}^2{\left(Y_{i,j,t}^S-Y_{i,j,t}^m\right)}^2}{{\sum }_{t=1}^{T\left(i,j\right)}{\left({mw}_{i,j,t}^m\right)}^2}}}{{Scale}_{i,j}}}\times 100\%\times {tw}_{i,j}$$

(1)

$$Global HM Error={\displaystyle \frac{\sum w_i{LHME}_i}{\sum w_i}}$$

(2)

where $i, j$ and $t$ are subscripts denoting the well, injection data type, and time step, respectively. $N_i$ represents the total number of injection data types for well $i$, and $T_{i,j}$ is the total number of measured data points for each data type. ${\mathrm{Y}}_{\mathrm{i},\mathrm{j},\mathrm{t}}^{\mathrm{s}}$ and ${\mathrm{Y}}_{\mathrm{i},\mathrm{j},\mathrm{t}}^{\mathrm{m}}$ denote the simulated and measured values, respectively. Each objective function term is weighted by ${\mathrm{t}\mathrm{w}}_{\mathrm{i},\mathrm{j}}$, which reflects the relative importance of the data type. The normalization factor ${\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}}_{\mathrm{i},\mathrm{j}}$ adjusts for differences in data scale and unit, ensuring that all production or injection data types comparably contribute to the objective function. Physically, it accounts for the range or uncertainty of measured values and is calculated based on either data variability or a defined measurement error. The point-specific weighting factor ${\mathrm{m}\mathrm{w}}_{\mathrm{i},\mathrm{j},\mathrm{t}}^{\mathrm{m}}$ allows individual data points to be emphasized or de-emphasized based on confidence or relevance, with a default value of 1. Together, these factors contribute to the objective function ${\mathrm{L}\mathrm{H}\mathrm{M}\mathrm{E}}_{\mathrm{i}}$ for a group of injection wells.

${\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}}_{\mathrm{i},\mathrm{j}}$ is the maximum of the following three quantities:

$$\mathsf{\Delta }{Y}_{i,j}^m+4\times Mer{r}_{i,j}$$

(3)

$$0.5\times min\left(\right|max\left(Y_{i,j,t}^m\right)|,|min\left(Y_{i,j,t}^m\right)\left|\right)+4\times Mer{r}_{i,j}$$

(4)

$$0.25\times max\left(\right|max\left(Y_{i,j,t}^m\right)|,|min\left(Y_{i,j,t}^m\right)\left|\right)+4\times Mer{r}_{i,j}$$

(5)

where $Y_{i,j}^m$ is the measured maximum change for well $i$ and production/injection data type $j$, and $Mer{r}_{i,j}$ is the measurement error.

2.3. Geochemical Modeling

X-ray diffraction (XRD) analysis of a core sample from the Entrada Sandstone identified quartz as the primary mineral in the sandstone grains. The sample also contained illite, calcite, chlorite, smectite, and various feldspars, including albite, plagioclase, and orthoclase, as shown in

Table 1. Mineral parameters and initial volume fractions [47].

Mineral	Chemical Formula	Initial Volume Fraction
Quartz	SiO2	0.734
Illite	K0.6Mg0.25Al2.3Si3.5O10(OH)2	0.048
Calcite	CaCO3	0.043
Albite	Na(AlSi3O6)	0.037
Plagioclase	CaAl2Si2O8	0.018
Orthoclase	K(AlSi3O6)	0.014
Chlorite	(14A)(Mg5Al2Si3O10(OH)8	0.013
Smectite	(Na,Ca)0.33(Al,Mg)2Si4O10(OH)2·(H2O)n	0.0001

. It is important to note that the sample’s location and depth may influence its mineral composition.

The initial brine composition used can be seen in

Table 2. Ion composition of formation water.

Ions	Concentration (ppm)
Na+	5245
Ca2+	24
Mg2+	13
Cl−	7633
HCO3−	336
CO32−	450
SO42−	1900

, highlighting the various ions in solution. Sodium (Na⁺) and chloride (Cl⁻) ions were the most prevalent, with concentrations measuring at 5245 ppm and 7633 ppm, respectively. The aquifer’s initial pH level was measured to be 8.37, which is basic. The pH of a system is very important as it impacts geochemical processes influencing the solubility, speciation, and reactivity of minerals [48,49]. Analysis also identified other significant ionic presences, including sulfate (SO₄²⁻), bicarbonate (HCO₃⁻), and carbonate (CO₃²⁻), with calcium (Ca²⁺) and magnesium (Mg²⁺) present in much smaller concentrations.

Our study characterizes geochemical processes through interactions at two levels: within the aqueous phase (homogeneous reactions) and between minerals and aqueous components (heterogeneous reactions). While a variety of equilibrium reactions could be relevant, our simulation focuses on those listed in

Table 3. Aqueous reactions employed in modeling [50].

Intra-Aqueous Chemical Equilibrium Reactions	Equilibrium Constant (log K at 25 °C)
CO2 (aq.) + H2O = (CO32−) + 2 (H+)	−16.88
(CO32−) + (H+) = (HCO3−)	10.33
H2O = (OH−) + (H+)	−14.00

. On the other hand, mineral dissolution and precipitation are represented as rate-dependent processes. The reaction kinetics are modeled using the kinetic rate law; it links reaction rates to parameters such as reactive surface area and activation energy. The kinetic parameters used for modeling mineral dissolution and precipitation are presented in

Table 4. Kinetic parameters for mineral dissolution/precipitation reactions employed in modeling.

Mineral Dissolution/Precipitation Reactions	Activation Energy J/mol	Reactive Surface Area (m2/m3)	Log 10 of Reaction Rate at 25 °C (1/s)
Quartz + 2 H2O = H4SiO4	90,900	2650.00	−13.4
Illite + 11.2 H2O = (K+) + (Mg2+) + (Al(OH)4−) + H4SiO4 + (H+)	23,600	2763.07	−10.98
Calcite = (CO3−) + (Ca2+)	14,400	2709.95	−0.3
Albite + 8 H2O = (Na+) + (Al(OH)4−) + H4SiO4	69,800	2620.00	−12.04
Plagioclase + 8 H2O = (Ca2+) + (Al(OH)4−) + H4SiO4	16,600	2760.29	−3.50
Orthoclase + 8 H2O = (K+) + (Al(OH)4−) + H4SiO4	51,700	2560.00	−10.06
Chlorite + 16 (H+) = (Mg2+) + (Al3+) + H4SiO4+ H2O	88,000	2600.00	−11.11
Smectite + 7 (H+) = (Al3+) + (Ca2+) + (Fe2+) + (Fe3+) + H2O + (K+) + (Mg2+) + (Na+) + SiO2	35,000	2350.00	−12.78

.

The Brooks–Corey function was used to create the relative permeability curves for gas and water [51], which was applied to experimental data obtained from sandstone samples from the Entrada formation. Figure 3 illustrates the relative permeability curves for the gas phase, featuring five scanning curves used for modeling hysteresis (residual trapping).

Geochemical History-Matching

The core-scale simulation is used to achieve a history match between the geochemical processes observed in the core-scale flow-through experiment and those predicted by the simulation. A 1-D model in CMG-GEM was used, as fluid flow in core flooding experiments is predominantly axial. The simulation replicated the core sample, its fluids, and the observed geochemical processes to align simulated results with experimental data [52,53].

Data categories for the model included rock properties (e.g., porosity, permeability, and mineral composition), fluid properties (e.g., viscosity, density, and composition of formation water and CO₂), and rock–fluid interaction parameters (e.g., relative permeability and geochemical reaction kinetics). Initial conditions such as pressure, temperature, and water saturation were set to match the experimental setup. The model dimensions matched the core sample size, with rock properties like porosity and permeability taken from pre-experiment measurements.

Table 5. Core characteristics and experimental conditions.

Parameter	Value
Core diameter	0.0248 m
Core length	0.0513 m
Flow rate	1.44 × 10−4 m3/day
Injection duration	23 days
Initial porosity	10.29
Initial permeability	10.34 mD

lists core sample characteristics and relevant experimental parameters.

The model was initialized with 100 percent water saturation using deionized water to match the conditions of the flow-through experiment. Two injection wells were placed at the grid ends, one for formation brine and the other for CO₂, replicating the experiment’s setup. A producer well at the opposite end simulated the experiment’s outlet. The injected formation water’s composition matched the details in Table 3, faithfully recreating the experimental conditions.

2.4. Field-Scale Modeling

After history-matching, a forecast of the core-scale simulation was performed before upscaling to the field-scale model. A summary of the input data for the model is provided in

Table 6. Initial reservoir properties for base case.

Parameter	Value
Grid dimension	164 × 116 × 25
Total number of grid blocks	475,600
Initial water saturation	100%
Initial reservoir pressure gradient	0.42 psi/ft (9.5 kpa/m)
Reference depth for initial pressure	8300 ft (2529.84 m)
Injection rate	20 MMscf/day (566,336.94 m3/day)
Fracture pressure gradient	0.62 psi/ft (14 kpa/m)
Porosity range	Up to 25%
Permeability range	Up to 220 mD

. The maximum injection pressure was set to 4631.4 psi (31.9 Mpa), corresponding to 90% of the fracture pressure. The injection rate was set to 20 MMscf/day (566,336.94 m³/day) for the 30-year period. Figure 4 shows an aerial view of the simulation model used.

3. Results

3.1. History Match

After history-matching the field-scale model, geochemistry was calibrated using core-scale simulations and results from flow-through laboratory studies. The model was then refined and upscaled. The history-matching process focused on two saltwater disposal wells (Figure 5 and Figure 6), with particular attention to aligning simulation outputs with historical water injection rates due to operational constraints. Both wells showed an exact match in water rate, as it was the controlling parameter, although some variability was noted in the BHP simulation. The overall global error for both wells was recorded at 7.83 percent, indicating robust model performance. This history-matching validated the model for CO₂ injection.

3.2. Laboratory and Core-Scale Simulation Analysis

The graphs in Figure 7 illustrate a comparative analysis of ion concentrations measured in flow-through experiments on a core sample against those obtained from corresponding simulations across various brine volumes. The ions analyzed include Na⁺, Fe²⁺, K⁺, Mg²⁺, Ca²⁺, and Al³⁺, with each subplot (a–f) detailing how the concentrations of these ions change over the course of the experiment.

The sodium ion (Na⁺) concentration, shown in Figure 7a, remains relatively stable in both laboratory and simulation data, with only minor discrepancies at the beginning of the brine volume range. Aluminum (Al³⁺), presented in Figure 7b, exhibits an initial spike followed by uniformly low levels throughout the experimental range in both datasets. Potassium (K⁺), shown in Figure 7c, does not exhibit a sharp decline in the experimental data but rather a gradual reduction, indicating slower release or exchange reactions; this behavior is overestimated in the simulation. In Figure 7d, magnesium (Mg²⁺) demonstrates a significant initial decrease followed by stability, which both the laboratory data and simulation capture effectively, indicating a well-calibrated model for Mg²⁺ interactions. However, Figure 7e highlights calcium (Ca²⁺) with notable differences; laboratory results show multiple fluctuations in calcium concentration, likely due to calcite dissolution and precipitation, while the simulation shows a single peak followed by a reduction, suggesting calcite dissolution followed by precipitation. Finally, iron (Fe²⁺), in Figure 7f, maintains consistently low concentrations across all laboratory measurements, closely matched by the simulation due to the very low smectite concentration.

The results of the forecasting, as shown in Figure 8, illustrate the changes in mineral moles over a 365-day period. Quartz and calcite are observed to precipitate out of the solution, while the remaining minerals undergo dissolution.

3.3. CO₂ Plume Containment

The plume results are illustrated in a series of diagrams in Figure 9. During injection, the CO₂ forms an ascending plume that quickly reaches the reservoir–cap rock boundary. The initial average CO₂ saturation within this plume is about 28%, with local peaks above 40% near the wellbore. Post-injection, CO₂ spreads laterally, continuing for nearly 100 years before stabilizing into a mushroom-shaped plume, as shown in Figure 9. This lateral migration gradually slows, approaching stability between 500 and 1000 years. The simulations confirm that CO₂ remains securely confined within the aquifer without leakage into the cap rock, validating the reservoir’s effectiveness for CO₂ storage. By the end of the 1000-year period, CO₂ saturation in the lower part of the plume decreases from 20% to 15%, mainly due to CO₂ dissolution and dissociation. Near the wellbore, at the top of the plume, CO₂ saturation shows a smaller reduction from 43% to 41%, while its lateral extent beneath the cap rock increases.

3.4. Changes in pH Due to CO₂ Injection Across Different Distances and Depths

Laboratory studies show that the initial pH of the Entrada Sandstone region within the San Juan Basin is 8.37, as listed in Table 4. When CO₂ injection begins, it reacts with formation brine to form carbonic acid (H₂CO₃), which dissociates into hydrogen ions (H⁺) and carbonate ions (CO₃²⁻), as described in Equation (1). The resulting hydrogen ions lower the pH within the aquifer. Shortly after injection, a rapid pH drop is observed near the wellbore, falling from an initial level of 8.37 to an acidic level of 5, as shown in Figure 10. Over time, as the CO₂ plume migrates, it interacts with surrounding rock minerals, triggering buffering effects and mineral dissolution that alter the pH. This reduces the CO₂ concentration around the injection well and induces pH changes across a broader area, as illustrated in Figure 10b,d.

Figure 11 illustrates the changes in pH levels and hydrogen ion (H⁺) concentrations in response to CO₂ injection, as shown in Figure 11a,b, which depict the temporal evolution of these parameters at the injection well and at varying distances from it. In Figure 11a, a sharp decrease in pH is observed at all monitored locations (the wellbore, 1000 feet, and 2000 feet laterally) during CO₂ injection, with the most significant decline near the wellbore, where CO₂ concentration is initially highest. This acidification results from CO₂ dissolution in formation water, which increases H⁺ concentration (Figure 11b). During injection, H⁺ concentration rises sharply, followed by a marked decrease, mainly due to CO₂ migration and geochemical reactions within the aquifer. The dissolution of minerals, such as calcite, consumes H⁺ ions and releases divalent cations like calcium. These cations can precipitate as carbonate minerals, sequestering CO₂ from the aqueous phase and reducing acidity. Additionally, non-carbonate minerals in the formation, such as chlorite and smectite, contribute to the neutralization process through their own dissolution, which consumes H⁺ ions. This buffering mechanism stabilizes pH over time after the initial CO₂ injection.

Thirty years after CO₂ injection ceased, pH levels show considerable spatial variability due to the migration and concentration of the CO₂ plume. At 1000 feet from the wellbore, pH is higher than directly at the wellbore, primarily due to a rapid decline in H⁺ ion concentrations near the wellbore, resulting from the intense initial migration and subsequent diffusion. At 2000 feet from the wellbore, pH remains consistently higher compared to closer distances, highlighting the reduced influence of the CO₂ plume at greater lateral distances. The lower CO₂ concentration in this region has a diminished impact on pH relative to areas nearer the injection well.

During CO₂ injection, the pH impact across different depths closely resembles the effects seen across lateral layers, with both showing a sharp and immediate reduction in pH. The pH change is especially pronounced in the uppermost layer, as shown in Figure 12. This effect is due to the significant density difference between formation brine and injected CO₂, which leads to rapid buoyant segregation. As a result, CO₂ rises toward the top layers, where it becomes trapped beneath the impermeable seal, creating more acidic conditions in the upper layers.

3.5. Temporal Dynamics of CO₂ Trapping Mechanisms

As shown in Figure 13, during the first year of CO₂ injection, structural trapping significantly increases to dominate other trapping mechanisms. By the end of year one, structural trapping accounts for 81.57% of total CO₂ storage, overshadowing other mechanisms. Residual trapping is minimal at 0.1%, and dissolution trapping stabilizes at 17.39%. The precipitation of calcite sets mineral trapping to 0.94%.

By the fifth year of injection, structural trapping peaks at 84.38% before gradually declining, stabilizing at 69.20% by the end of injection. This decline is due to an increase in residual trapping, which reached 19.10%. Dissolution trapping contributes 11.70%, while mineral trapping remains negligible due to how slow the mineralization process is. A decade after injection ends (30–40 years from the start), residual trapping rises by approximately 1.6 times, reaching 30.01% of the total CO₂ trapped, and further increases to 32.34% at the 125-year mark. Structurally trapped and dissolved CO₂ fractions are 56.72% and 10.93%, respectively, with minimal mineral trapping (0.01%).

The trend in residual trapping growth is attributed to relative permeability hysteresis. For dissolution trapping, dissolved CO₂ levels decline until 20 years post-injection (50 years since simulation start), after which they increase again and stabilize toward the end of the 500-year simulation. This resurgence in dissolved CO₂ is likely driven by Rayleigh–Taylor instabilities [54], where the denser, CO₂-rich brine at the reservoir top descends, and lighter brine beneath rises, enhancing dissolution through mixing and reducing structurally trapped CO₂. Structurally trapped CO₂ continues its decline until the end of the simulation, while mineralized CO₂ remains negligible throughout.

At the conclusion of the 500-year simulation, structural, residual, dissolved CO_2, and mineral-trapped CO₂ account for 45.43%, 42.95%, 11.59%, and 0.03%, respectively. This trend persists until the 1000-year mark. The model’s mineralization calculations were confined to carbonate minerals, primarily calcite, and excluded changes in non-carbonate minerals, which do not contribute to mineral trapping of CO₂ [48].

3.6. Characteristics of Mineralization and Its Influence on Flow Properties

CO₂ sequestration through mineralization represents a relatively small portion of CO₂ trapping compared to dissolution, residual trapping, and structural CO₂. This is because CO₂ mineralization primarily involves converting dissolved CO₂ into stable solid forms such as calcite (CaCO₃), magnesite (MgCO₃), siderite (FeCO₃), and dawsonite (NaAlCO₃(OH)₂), depending on the geochemical environment and available cations. In the San Juan Basin, calcite is the only carbonate, comprising 4.3% of the initial mineral volume. While non-carbonate minerals do not directly contribute to carbonate trapping, their dissolution and precipitation can alter porosity and permeability, indirectly affecting fluid flow and reservoir behavior. This highlights a broader geochemical and geological influence beyond direct carbonate mineralization. Figure 14 illustrates all precipitation and dissolution reactions of minerals over the 1000-year period.

Figure 15 shows long-term changes in calcite precipitation over 1000 years. The graph indicates a continuous increase in calcite moles, reflecting a steady progression toward chemical stabilization in the system. Initially, high precipitation rates suggest an abundance of Ca²⁺ ions within the first 200 years; afterward, the rate of precipitation slows as Ca²⁺ availability declines due to plume stagnation. This pattern suggests that the brine composition and CO₂ injection conditions promote calcite formation.

Beyond calcite, Figure 16a–d show the behavior of other minerals over 1000 years, shedding light on the reservoir’s evolving geochemistry. Quartz exhibits consistent precipitation throughout the period, which tends to reduce formation porosity as the mineral occupies previously open pore spaces. In contrast, chlorite and illite show steady dissolution over the same timeframe, potentially increasing formation porosity and partially offsetting the effects of quartz and calcite precipitation. This balance between mineral precipitation and dissolution illustrates the complex geochemical adjustments within the reservoir.

Feldspars, including albite, plagioclase, and orthoclase, consistently dissolve when exposed to CO₂ over the 1000-year timeline. Albite and orthoclase dissolve more significantly than plagioclase, and their breakdown further enhances porosity, as these minerals constitute a major part of the rock matrix, and their dissolution frees up space. Smectite remains largely unchanged due to its minor initial presence, making up only 0.0184% of the total composition, and thus has minimal impact on the reservoir’s overall geochemical evolution.

Figure 17 illustrates the effect of geochemical reactions on porosity. Changes in porosity primarily occur in areas where pH variation happened, with notable alterations observed within the first 30 years, as shown in Figure 17a. During this initial period, porosity decreases due to the dominance of precipitation of minerals. As the CO₂ plume migrates over time, porosity changes expand across a larger area, as depicted in Figure 17a–d. Although the dissolution of certain minerals attempts to counterbalance the porosity decrease from quartz and calcite precipitation, the overall effect of mineral reactions results in an approximate 0.09% net decrease in porosity. Despite instances of dissolution, precipitation remains the predominant process throughout.

The trend in Figure 18 indicates that the topmost grid blocks (70, 51, 11; 72, 51, 11; 74, 51, 11; 76, 51, 11) experience a more significant reduction in porosity than the deeper layers, largely due to CO₂′s buoyancy. Upon injection, CO₂ rises towards the top because of its lower density, leading to higher concentrations in the upper layers, which accelerates mineral dissolution in these zones. In contrast, deeper layers, exposed to lower CO₂ concentrations, exhibit a lesser reduction in porosity, revealing a vertical gradient in CO₂–rock interactions where the top layers undergo more active dissolution.

An initial rapid decline in porosity across all layers, followed by a plateau, possibly reflects the equilibrium state reached between dissolution and precipitation reactions over time. Our analysis identifies a distinct spatial trend in porosity changes associated with CO₂ injection. Initially, the most significant porosity reduction occurs near the wellbore, where CO₂ concentrations are highest, facilitating intense and immediate mineral reactions. As CO₂ migrates laterally away from the wellbore, it accumulates in more distant reservoir regions, leading to a gradual decrease in the rate of porosity change in these areas, which eventually stabilizes at lower levels than those near the wellbore.

3.7. Impact of Geochemical Reactions on Ions

The behavior of initial ions like Na⁺, Ca²⁺, Mg²⁺, and Cl⁻ in a geochemical system is heavily influenced by CO₂-induced mineral reactions. Figure 19 shows the concentration changes of these ions over 1000 years, illustrating the dynamic interactions between mineral dissolution and precipitation and the resulting ion distribution.

The concentration of Na⁺ steadily increases over time, as seen in Figure 19a. This trend suggests ongoing dissolution of sodium-bearing minerals, specifically albite and smectite. The consistent increase in Na⁺ concentration by over 1500 ppm indicates that Na⁺ is released into the system faster than it is consumed by any secondary mineral formations, which aligns with the minimal precipitation observed in smectite compared to the more significant dissolution of albite.

Figure 19b displays a continuous decline in Ca²⁺ concentration, reflecting the precipitation of calcium-bearing minerals. This reduction in Ca²⁺ is primarily attributed to calcite precipitation, as shown in Figure 15.

Similar to Na⁺, the concentration of Mg²⁺, shown in Figure 19c, gradually increases over time, indicating the dissolution of Mg-bearing minerals such as illite and chlorite. The slower increase in Mg²⁺ concentration could be due to the lower solubility of these minerals compared to sodium-bearing minerals or a slower reaction rate with CO₂.

The behavior of Cl⁻, depicted in Figure 19d, shows an initial decrease followed by a steady increase after approximately 400 years. It is noteworthy that these changes are minimal, under five ppm. Since Cl⁻ is generally treated as a conservative ion in this study and does not participate in mineral reactions, these shifts are likely due to dilution effects or changes in the water mass of the system, as illustrated in Figure 20, rather than direct geochemical reactions.

Just like the primary ions, the behavior of the secondary ions, minerals that were not present in the initial brine composition, also gets affected by the mineral dissolution and precipitation. Al²⁺ and K⁺ experience an increase in concentration, which is attributed to the dissolution of illite, chlorite, and orthoclase, while the behavior of Fe²⁺ and Fe³⁺ corresponds to the dissolution and precipitation behavior of smectite (Figure 21a–d).

These plots collectively demonstrate the intricate interactions between CO₂ and the ions in a geochemical system. The dissolution of specific minerals releases ions into the solution, while the formation of secondary minerals can sequester them, thus modulating their concentrations over time.

4. Discussion

The results demonstrate a strong agreement between the simulated data and laboratory measurements, affirming the model’s effectiveness in replicating dynamic geochemical processes. While the experiments were conducted over short durations, the calibration enhances the model’s validity, supporting its application for predicting these processes over much longer timescales. The close matching of ion concentrations in most cases underscores the capability of our numerical model to accurately perform reactive transport simulations.

However, certain deviations were noted in the concentrations of specific ions, notably Ca²⁺ as seen in Figure 7e. These discrepancies could potentially be attributed to various factors beyond the scope of the initial model settings. For instance, the mineral composition of the rock used in laboratory experiments, which can significantly influence ion interactions and behavior, may differ from the average values used in the simulation.

The field-scale simulation results provide a detailed view of CO₂ plume containment during and after the injection phase. Initially, the CO₂ forms an ascending plume that rapidly reaches the boundary between the reservoir and the cap rock. With CO₂ saturation exceeding 40% near the wellbore, the plume subsequently spreads laterally, forming a characteristic mushroom shape as seen in Figure 9a,b. The confinement of CO₂ within the aquifer, without fluid movement into the cap rock, underscores the reservoir’s effectiveness in secure CO₂ storage. The decrease in CO₂ saturation from 20% to 15% in the lower part of the plume is primarily due to the buoyant nature of the gas, allowing it to rise upward. Near the wellbore, a minor reduction in CO₂ saturation from 43% to 41% suggests relative stability and limited lateral spread beneath the cap rock.

Changes in pH levels due to CO₂ injection are key indicators of the geochemical interactions taking place within the aquifer. The initial sharp decline in pH from 8.37 to around 5 near the wellbore is a direct consequence of CO₂ dissolution, which leads to the formation of carbonic acid in the water. This acidification not only lowers the pH but also promotes a number of reactions with minerals in the surrounding rock. These reactions involve the dissolution of feldspars (albite, plagioclase, and orthoclase), phyllosilicates (chlorite and smectite), and mica-type clay (illite), releasing species like sodium ions, potassium ions, magnesium ions, orthosilicic acid, and aluminum ions into the brine, and the precipitation of tectosilicates (quartz) and carbonates (calcite), which consumes calcium and carbonate ions. Figure 16a–d show the trend of these minerals, while Figure 19a–d and Figure 21a–d show the trend of the ions. As the acidified fluid moves further from the wellbore, buffering effects mitigate this acidification, stabilizing pH levels. Spatial variability in pH, with higher levels at greater distances from the wellbore, indicates dispersal and dilution effects of the CO₂ plume. The pronounced pH reduction in the uppermost layers, due to buoyant segregation of CO₂, highlights the vertical migration and trapping behavior of CO₂ within the aquifer.

Evaluating the temporal dynamics of CO₂ trapping mechanisms over the 1000-year simulation period shows significant shifts in the dominant processes. Initially, dissolution trapping is predominant, but its role diminishes as structural trapping becomes more significant. The increase in residual trapping after CO₂ injection ceases underscores the importance of long-term CO₂ retention. The decline in mineral trapping, primarily due to continuous calcite dissolution, indicates complex interactions between geochemical reactions and CO₂ trapping. The resurgence in dissolution trapping driven by Rayleigh–Taylor instabilities highlights ongoing dynamic processes influencing CO₂ sequestration over extended periods.

The comparison between the core-scale simulation and the field-scale model shows consistent trends, with quartz and calcite precipitating while other minerals dissolved. The key difference lies in the magnitude of changes, as the core-scale’s smaller size and limited CO₂ injection led to much smaller reactions. Despite this, the model accurately captured these trends at both scales, demonstrating its reliability in forecasting geochemical behavior.

Mineralization processes, although representing a smaller fraction of CO₂ trapping, play a significant role in altering the reservoir’s flow properties. The consistent precipitation of quartz and calcite, along with the dissolution of other minerals in the reservoir, impacts reservoir porosity (Figure 17). Although there is a reduction in porosity due to the dominance of precipitating minerals, these changes are less than 1%, rendering their effect on flow properties and permeability largely negligible. This minimal alteration in porosity means that, while precipitation and dissolution processes are occurring, they are unlikely to significantly impact fluid movement, CO₂ injection efficiency, or long-term storage capacity within the reservoir.

The behavior of primary ions such as Na⁺, Ca²⁺, Mg²⁺, and Cl⁻ provides insights into the geochemical system’s response to CO₂ injection. The steady increase in Na⁺ and Mg²⁺ concentrations indicates ongoing dissolution of sodium- and magnesium-bearing minerals. The initial spike and subsequent stabilization of Ca²⁺ concentration reflect dynamic dissolution and precipitation processes involving calcium-bearing minerals. The complex pattern observed for Cl⁻, with initial fluctuations followed by stabilization, likely indicates dilution effects and interactions with other ions rather than direct geochemical reactions. These trends highlight the intricate interactions between CO₂, mineral dissolution, and ion concentrations, emphasizing the importance of considering geochemical reactions in assessing the long-term impacts of CO₂ sequestration.

5. Conclusions

This study presents a comprehensive assessment of CO₂ sequestration within a deep saline aquifer, incorporating both laboratory coreflood validation and long-term field-scale simulation. The numerical model demonstrated strong consistency with laboratory data, particularly in replicating ion transport and geochemical interactions, thereby supporting its application for long-term reactive transport prediction.

Over the 1000-year simulation period, the model showed effective confinement of CO₂ within the reservoir, supported by lateral plume expansion and vertical restriction beneath the caprock. Although no CO₂ or brine flux into the caprock was observed in the simulation, this result reinforces the structural sealing capability of the formation.

A major focus was the evolution of pH, which showed a sharp drop near the injection zone—from 8.37 to around 5—due to carbonic acid formation. This acidification initiated extensive mineral dissolution, particularly of feldspars (e.g., albite), phyllosilicates (e.g., smectite, chlorite), and mica-type clays, while promoting precipitation of quartz and calcite. These reactions had a measurable but modest effect on porosity (less than 1%), indicating limited impact on injectivity and flow over the modeled period.

In terms of mineral trapping, calcite dynamics were particularly important. Although the amount of trapped CO₂ via mineralization was small compared to residual and solubility trapping, its role in buffering pH and participating in long-term geochemical cycles is critical. Notably, no evidence of extreme mineral buildup leading to pore-blocking was observed in our results, contradicting generalized concerns about injectivity loss. However, the temporal analysis did reveal that trapping mechanisms evolve: solubility trapping dominated early stages, while residual and structural trapping gained importance over time.

Our findings also reveal mineral-specific behaviors that influence ion concentrations. For example, Na⁺ and Mg²⁺ increased due to feldspar and chlorite dissolution, respectively, while Ca²⁺ showed non-monotonic trends reflecting the complex interplay of dissolution and precipitation. These trends emphasize the importance of accounting for mineral heterogeneity and kinetic rates in long-term forecasts.

Ultimately, this study advances our understanding of the geochemical and physical transformations induced by CO₂ storage. Future work should explore broader mineral suites, variable salinity regimes, and coupled mechanical effects to improve the robustness of storage predictions under diverse reservoir conditions.