Injection of a CO₂-Reactive Solution for Wellbore Annulus Leakage Remediation

Abstract

Driven by concerns for safe storage of CO₂, substantial effort has been directed on wellbore integrity simulations over the last decade. Since large scale demonstrations of CO₂ storage are planned for the near-future, numerical tools predicting wellbore integrity at field scale are essential to capture the processes of potential leakage and assist in designing leakage mitigation measures. Following this need, we developed a field-scale wellbore model incorporating (1) a de-bonded interface between cement and rock, (2) buoyancy/pressure driven (microannulus) flow of brine and CO₂, (3) CO₂ diffusion and reactivity with cement and (4) chemical cement-rock interaction. The model is aimed at predicting leakage through the microannulus and specifically at assessing methods for CO₂ leakage remediation. The simulations show that for a low enough initial leakage rate, CO₂ leakage is self-limiting due to natural sealing of the microannulus by mineral precipitation. With a high leakage rate, CO₂ leakage results in progressive cement leaching. In case of sustained leakage, a CO₂ reactive solution can be injected in the microannulus to induce calcite precipitation and block the leak path. The simulations showed full clogging of the leak path and increased sealing with time after remediation, indicating the robustness of the leakage remediation by mineral precipitation.

1. Introduction

Large scale geological storage of CO₂ can significantly reduce CO₂ emissions and limit global warming [1]. Geological reservoirs are selected for the physical containment of CO₂ which guarantees permanent storage in the subsurface. However, CO₂ injection wells (and possibly old oil and gas wells) penetrating the reservoirs and the caprocks above can compromise the integrity of the storage complex. Wells have a primary structural seal of casing and annular cement (between the casing and the geological formation) and a cement plug when abandoned. Despite these seals, many oil and gas wells leak during their operational lifetime or after abandonment through leakage pathways formed by cement shrinkage or pressure and temperature fluctuations [1]. If annular cement is placed properly, the most likely leak path for CO₂ is along the well through fractures in the cement or microannuli between the cement and the casing or adjacent rock [2,3,4,5].

CO₂ leakage through microannuli will cause dissolution of CO₂ in the pore waters which acidifies the near-wellbore environment and causes cement reactivity. Reactions of cement in contact with CO₂-rich water or brine have been extensively studied with experiments and by numerical modelling [6,7,8,9,10,11,12,13]. A typical wellbore cement mineralogy consists of mainly portlandite (Ca(OH)₂) and calcium silicate hydrate (CSH), with minor phases such as aluminium-, iron-, magnesium-, or calcium- containing sulphate-, carbonate-, or silicate-hydrates [14]. In general, cement-CO₂ interaction is primarily characterized by portlandite dissolution and subsequent precipitation of calcium carbonate (CaCO₃) because of the fast reaction kinetics. The dissolving CSH phase forms additional calcium carbonate and amorphous silica gel (SiO₂) [7]. Characteristic successive, inward moving reaction zones are observed consisting of portlandite dissolution, calcite precipitation, and subsequent calcite dissolution from the rim, leaving a porous, silica-rich rim [6,10,11,12]. However, cement reactivity will most-likely only lead to cement degradation under leaching/flow conditions when reaction products are quickly removed from the reaction site. At no/low flow conditions, calcite precipitation is the dominant process rather than (re-)dissolution [4,15,16]. Depending on the initial flow and chemical conditions, continuous cement leaching occurs, or cement reactivity may actually support natural sealing of the micro annulus [17,18,19].

In the case of sustained CO₂ leakage, a corrective measures plan must be in place and appropriate remediation measures should be taken [20,21]. Intentional clogging of the near-well area with salt has been proposed as a preventative measure against CO₂ leakage [22]. This method was based on the capacity of injected CO₂ to evaporate water and precipitate pore filling salt. The process of natural sealing of the microannulus by mineral precipitation indicates a potential for chemical clogging of the annulus leak path. Clogging with calcite or silica has already been proposed for a CO₂ leak path through the caprock [23,24,25]. To induce mineralization in a leak path, it was proposed to inject a silica- or calcium-rich suspension or solution into the CO₂-containing environment. The injected solution will react in the acid environment to form a solid silica (gel) or carbonate mineral. A modelling study on leakage remediation above a leaking fault through a caprock indicated a final leakage reduction of up to 95% [25]. Experimental [23] and modelling results [24,25] for caprock leakage mitigation support the feasibility of the method for reactive clogging by injecting a CO₂-reactive solution into a high permeable leak path to form solid reactants that clog the leak path, reduce permeability and stop leakage. The main objective of this study is to assess the possibility of reactive leakage mitigation for wellbore annulus leakage.

We developed a field scale reactive transport model based on the model reported by Koenen and Wasch [18] to simulate CO₂ leakage through a microannulus, resulting in either sustained flow and cement leaching or in natural sealing and reduced leakage. For the leakage cases, we study the potential of induced CO₂ mineralization in the leak path, mitigating CO₂ leakage. The numerical modelling study includes the following processes:

• Flow of supercritical CO₂ and brine along the initially water-saturated microannulus;
• Diffusion of dissolved CO₂ into the caprock;
• Diffusion of dissolved CO₂ into impermeable cement and reactions of dissolved CO₂ and cement;
• Leakage into the aquifer overlying the caprock;
• Injection of a CO₂-reactive solution in the microannulus leak path to promote clogging by calcite formation;

In this paper, we report on microannulus leakage (versus sealing), the intentional clogging process for leakage remediation, and the post-clogging phase to assess the sustainability of the clogging procedure.

2. Materials and Methods

A reactive transport model was developed in TOUGHREACT (Version 3, Lawrence Berkeley National Laboratory, Oakland, CA, USA) [26], a simulator for coupled modelling of multiphase fluid and heat flow, solute transport, and chemical reactions by introducing reactive transport into the flow simulator TOUGH2. We use the ECO2N fluid property module to include the thermodynamic and thermophysical properties of H₂O–NaCl–CO₂ mixtures [27].

A 2D axisymmetric field-scale model (Figure 1) was adapted from the model reported in [18]. The well consists of two cemented casings: a 9 5/8” casing of 3590 m depth into the reservoir and a 13 3/8” casing to a depth of 2569 m. We assume a 0.5 cm thick casing (unreactive transport barrier) and a 3 cm thick annular cement. A 500-micron thick, porous and highly permeable micro-annulus is defined between the cement and the adjacent rock. For the rock formations, three layers are defined: a reservoir (infinite volume, representing a large storage reservoir), an impermeable caprock (550 m thick) and an overlying, permeable aquifer (3040 m thick). Simulations are performed in isothermal mode, with a (fixed) temperature gradient from reservoir to surface. To reduce simulation times, the upper 2500 m of the aquifer to the surface are removed after initializing pressure and temperature. The resulting mesh contains 70 layers and 36 columns. Vertical mesh refinement is defined around the layer interfaces and at the leakage remediation interval down to 5 cm. The mesh is refined towards the well in the radial direction, down to cells of 0.5 cm width for the annular cement. The pressure is equilibrated with depth; 25 °C and 1.5 bar at the top and 90 °C and 327 bar at reservoir level. This yields a 1180 m thick model with a radius of 150 m (Figure 1). The boundary conditions are represented by the infinite volume reservoir with a 0.8 CO₂ saturation and an infinite volume upper boundary for the microannulus. The reservoir was given a 20 bar overpressure to simulate upward leakage out of the CO₂ reservoir.

The TOUGHREACT software is developed to simulate reactive transport through porous media and does not allow open space. Therefore, we defined the microannulus as cement material with a 0.9 volume fraction porosity, with the remaining 0.1 volume fraction made up by cement mineralogy. The 500-micron microannulus hence has an effective thickness of 450 micron, but the width of the microannulus may increase by 10% in case of cement mineral dissolution. The microannulus width (aperture) can be related to its permeability by the cubic law. For an aperture of 500 micron the permeability would be 4.2 × 10⁻⁹ m² (~4200 Darcy), however, the actual hydraulic permeability will be lower as it is affected by many factors such as fracture wall roughness. To account for this, simulations have been previously run with various initial (hydraulic) permeability values between 5 × 10⁻¹³ m² (~500 mDarcy) and 8.3 × 10⁻¹⁰ m², covering values reported in literature [18]. For the present study we use a permeability range between 1 × 10⁻¹³ m² and 1 × 10⁻¹¹ m² to capture the uncertainty and natural variation in microannulus flow. For the base case, a permeability of 1.3 × 10⁻¹² m² is selected.

The flow properties of the different materials used in the model are summarized in

Table 1. Transport properties of the different materials used in the model. * In TOUGHREACT a permeability of 1.0 × 10⁻³⁰ m² represents impermeable material.

Transport Properties	Unit	Reservoir	Cement	Caprock	Microannulus	Aquifer
Porosity	φ (-)	0.2	0.3	0.1	0.9	0.2
Permeability	k (m2)	2.0 × 10−13	1.0 × 10−30 *	1.0 × 10−30 *	1.3 × 10−12	2.0 × 10−13
Tortuosity	τ (-)	0.4	0.01	0.05	0.4	0.4
Capillary Curve
Van Genuchten	λ (-)	0.2	-	-	-	0.2
	Slr (-)	0.1	-	-	-	0.1
	1/P0 (-)	0.002	-	-	-	0.002
	P0 (Pa)	1.0 × 108	-	-	-	1.0 × 108
	Sls (-)	1	-	-	-	1
TRUST	P0 (Pa)	-	1.0 × 106	1.0 × 106	-	-
	Slr (-)	-	0.2	0.2	-	-
	η (-)	-	0.4	0.4	-	-
	Pe (Pa)	-	4.0 × 106	4.0 × 106	-	-

. For relative permeability, we use the equation from Corey’s curve with a residual liquid and gas saturation of 0.02 and 0.1. The diffusion coefficient is calculated in TOUGHREACT by multiplying a standard diffusion coefficient (1 × 10⁻⁹ m²/s) by the tortuosity, porosity and liquid saturation. Hence the effective diffusion coefficient will change over time as the porosity and saturation develop due to fluid flow and mineral reactions.

The cement consists of portlandite, CSH_1.6 (CSH of a 1.6 Ca/Si ratio), monosulfoaluminate and hydrotalcite (

Table 2. Initial mineralogy of cement and aquifer/caprock and possible secondary minerals.

Mineral-Formula	Cement (Volume Fraction)	Aquifer/Caprock (Volume Fraction)
Portlandite-(Ca(OH)2)	0.2	-
CSH(1.6)-(Ca1.60SiO3.6:2.58H2O)	0.6	-
Monosulfoaluminate-(Ca4Al2SO10:12H2O)	0.1	-
Hydrotalcite-(Mg4Al2O7:10H2O)	0.1	-
Quartz-(SiO2)	-	0.7
Calcite-(CaCO3)	-	0.01
Amorphous silica-(SiO2)	-	-
Anhydrite-(CaSO4)	-	0.01
Dolomite-(CaMg(CO3)2)	-	0.01
Gibbsite-(Al(OH)3)	-	-
Microcline-(K(AlSi3)O8)	-	0.1
Albite-(NaAlSi3O8)	-	0.1
Kaolinite-(Al2Si2O5(OH)4	-	0.07

). The secondary minerals for cement are calcite, amorphous silica, anhydrite, dolomite and gibbsite. A clastic rock was included consisting of quartz, albite, microcline, kaolinite, anhydrite, dolomite and calcite. We used the thermodynamic database Thermoddem (version 1.10, 6 Jun 2017, BRGM, France) to model chemical reactions [28]. The reaction of minerals is kinetically controlled using a rate expression of Lasaga et al. [29]. Mineral kinetics are listed in

Table 3. Kinetic rate parameters (*^a [32], *^b [33] and *^c [34]).

Mineral	Acid Mechanism			Neutral Mechanism		Carbonate/Base Mechanism
	Log (k25 °C) (mol/m2/s)	ΔH (kJ/mol)	n	Log (k25 °C) (mol/m2/s)	ΔH (kJ/mol)	Log (k25 °C) (mol/m2/s)	ΔH (kJ/mol)	n
Portlandite *c	−3.10	75	0.6	−7.66	75	-	-	-
C1.6SH *c	−7.23	23	−0.28	-	-	-	-	-
Monosulf. *b	−3.09	74.9	0.6	−11.2	15	-	-	-
Hydrotal. *b	−7.23	15	0.28	−17.8	15	-	-	-
Quartz *a	−7.52	56.9	0.5	−4.55	56.9	-	-	-
Calcite *a	−0.30	14.4	1	−5.81	23.5	−3.48	35.4	1
Silica(am) *a	-	-	-	−9.42	49.8	-	-	-
Anhydrite *a	-	-	-	−3.19	14.3	-	-	-
Dolomite *a	−3.19	36.1	0.5	−7.53	52.2	−5.11	34.8	0.5
Gibbsite *a	−7.65	47.5	0.99	−11.5	61.2	−16.7	80.1	−0.78
Albite *a	−10.2	65	0.46	−12.6	69.8	−15.6	71	−0.57

. The specific surface area is assumed 0.98 m²/kg, except for the C-S-H phases and clays for which a value of 100 m²/kg is assumed. The CO₂-reactive solution to stimulate microannulus clogging is designed by equilibrating lime with a sodium chloride brine at surface conditions. The different fluid compositions are listed in

Table 4. Initial composition of the different fluids, dissolved species in mol/L.

	Cement	Aquifer/Caprock	CO2 Reservoir	CO2-Reactive Solution
pH (-)	10.8	6.2	3.5	13.8
Ca2+	1.30 × 10−2	3.15 × 10−2	3.10 × 10−2	0.68
Mg2+	1.69 × 10−8	9.32 × 10−3	9.16 × 10−3	-
Na+	1.0	0.98	0.98	1.03
K+	-	8.62 × 10−3	8.48 × 10−3	-
H4SiO4	4.10 × 10−6	8.57 × 10−4	8.46 × 10−4	-
HCO3−	-	8.29 × 10−3	1.78	-
SO42−	2.07 × 10−4	3.25 × 10−2	3.20 × 10−2	-
Al3+	4.14 × 10−4	7.24 × 10−8	7.12 × 10−8	-
Cl−	1.0	1.0	0.98	1.03

.

Porosity changes due to water-rock reactions are calculated in TOUGHREACT using mineral molar volumes. Porosity change can be related to permeability change by porosity-permeability relations, but they contain highly uncertain and material specific input parameters [30]. The Verma–Pruess relation (Equation (1)) is an extension of a Power Law relation considering a critical porosity at which the permeability is assumed zero. This relation is often used for mineral precipitation processes given the high impact of the critical porosity on porosities decreasing during mineral precipitation, while the relevance of critical porosity in the mathematical expression diminishes for porosities increasing during dissolution [30]. We used the porosity-permeability relationship of Verma and Pruess [31] and addressed the uncertainty of the input parameters by a sensitivity study. The base case values are a 0.80 critical porosity and a power law component of 6.

$$\frac{k}{k_i}={\left(\frac{\phi - {\phi }_c}{{\phi }_i - {\phi }_c}\right)}^n$$

(1)

where k is the permeability, k_i the initial permeability, φ the porosity, φ_i the initial porosity, φ_c the critical porosity below which the permeability is assumed zero, and n is a power law exponent.

3. Results of CO₂ Leakage and Natural Sealing

3.1. CO₂ Leakage and Cement Reactions

The reactive transport simulations show upward flow of CO₂ through the microannulus. With the applied 20 bar overpressure, the flow rate of gaseous CO₂ through the microannulus is 2.1 × 10⁻⁵ kg/s. During upward flow, a fraction of the CO₂ gas dissolves into the cement and caprock pore water and diffuses horizontally into the cement and the caprock, thereby lowering the pH of the pore waters. Cement minerals react with the carbonized water and start to dissolve to buffer the pH. Similar to e.g., Kutchko et al. [6], we observe inward progression of reaction zones in the horizonal direction, which is limited by diffusion (Figure 2). The reaction zones are characterized by: portlandite dissolution and calcite formation, CSH dissolution with amorphous silica and calcite precipitation, monosulfoaluminate and hydrotalcite dissolution with dolomite (max. 0.003 volume fraction), gibbsite (max. 0.0025 volume fraction) and anhydrite (max. 0.009 volume fraction) precipitation. The further up from the reservoir level, the less advanced the horizontal progression of these zones is, since the reactions only start as soon as the upward migration of the CO₂ reaches that level and the pH decreases. After 2 years of CO₂ leakage, approximately 1 cm of the cement just above reservoir level is affected by horizontal CO₂ diffusion and related reactions (Figure 2). The silicate minerals of the caprock do not show siginificant reactions within the simulated time. There is a minor increase of calcite in the caprock adjacent to the microannulus.

After the CO₂ within the microannulus reached the top of the caprock, the pH is around 4.5 throughout the microannulus. In the cement affected by CO₂ interactions-adjacent to the microannulus-the pH is roughly 5 and in the unaltered cement nearly 11 (Figure 3). The pH decrease causes full dissolution of portlandite in the microannulus and adjacent cement cells and gradual dissolution of the other cement phases such as CSH, with decreasing amount of dissolution from the reservoir level upwards. As a result, secondary calcite is also highest at the level close to the reseroir (Figure 3). Within the microannulus calcite that precipitated is re-dissolved due to the low pH and flow conditions which allow quick removal of dissolved species. The permeability of the microannulus is predicted to increase from 1.3 × 10⁻¹² m² to 1.9 × 10⁻¹² m² after 2 years of CO₂ flow and corresponding cement alteration.

3.2. Microannulus Leakage Versus Natural Sealing

Four permeability values were selected to simulate different initial leakage rates and to assess the chemical processes within a leaking or self-sealing annulus. Results are discussed after half a year of leakage simulation. The different permeabilities yield different levels of gas saturation, with a higher gas saturation for a higher permeability (Figure 4a). For all scenarios, the flow of CO₂ and dissolution of CO₂ in the initially water saturated microannulus result in complete dissolution of portlandite within the microannulus—which was 10% cement filled—and subsequent precipitation of calcite (Figure 4b). Above the reservoir, re-dissolution of calcite can be observed. Only for the lowest permeability of 1 × 10⁻¹³ m², CO₂ gas does not reach the top of the caprock. This is due to natural sealing of the microannulus, with complete calcite clogging in the middle part of the microannulus. The front of calcite precipitation is characterised by a peak in calcium content (Figure 4c) and an increase in pH (Figure 4d). The permeability of the microannulus depends on the initial permeability and the dissolution and precipitation reactions. There is a high permeability near the reservoir where the calcite content is lowest (Figure 4e). The 1 × 10⁻¹¹ and 5 × 10⁻¹¹ m² scenarios show only little permeability change in the upper part of the microannulus due to portlandite dissolution and calcite precipitation. The initial permeability value 1 × 10⁻¹² m² is more affected by calcite precipitation and the 1 × 10⁻¹³ m² scenario shows complete permeability impairment due to calcite clogging.

4. Results of CO₂ Reactive Leakage Remediation

4.1. Injection of the CO₂-Reactive Solution with Different Injection Pressures

The 1 × 10⁻¹² m² permeability scenario is selected to model initial leakage and subsequent leakage remediation. Upward CO₂ leakage through the microannulus was simulated for half a year before leakage remediation was applied. To remediate leakage, a CO₂-reactive solution (composition given in Table 4) was injected into the CO₂ containing microannulus. The solution is injected at a depth of 3048 m, which is 8 m below the top of the caprock. The CO₂-reactive, lime-saturated solution is injected in order to react with dissolved CO₂ to form calcite (CaO + CO₂- > CaCO₃). To inject the CO₂-reactive solution in the model, we used a fixed pressure cell adjacent to the microannulus to allow for pressure-controlled injection, instead of defining a fixed injection rate. The sensitivity of leakage remediation to flow rates was asessed by variying the pressure of injection with 1, 5 and 10 bar overpressure for the microannulus pressure of 324 bar.

For all three injection pressures, the injection of the CO₂-reactive solution leads to an initial increase followed by an interruption of the upward CO₂ flow in the microannulus, but flow recovers (after t = 0 in Figure 5, showing the results of 5 bar overpressure injection). Calcite starts to precipitate in the microannulus at the level where the CO₂-reactive solution is injected. Gradually the permeability of the microannulus reduces with a corresponding decrease of the CO₂ and water flow rate (Figure 5). The rate of solution injection reduces as well, since the permeability decreases and the pressure is fixed. Figure 6a shows the initial higher injection rate with a higher overpressure and the decrease of injection rate with time. The moment of complete flow impairment occurs at the time when the microannulus adjacent to the injection cell is nearly filled with calcite, reducing the permeability to zero (Figure 5). All overpressures yield complete calcite clogging (Figure 6b), but with differences in the development of calcite precipitation due to the differences in the balance of calcium and CO₂ supply. Within 10 days, all injection pressure scenarios predict clogging the microannulus with calcite precipitation, preventing further leakage of the CO₂.

4.2. Sensitivity to Porosity-Permeability Relation Input Parameters

The clogging process and specfically the calculation of the permeability based on the porosity development depends on highly uncertain porosity-permeability parameters [30]. A sensitivity study is performed reflecting the range in values for the critical porosity (φ_c) and power law (n) component as probed by Verma and Pruess [31] (1 ≤ n ≤ 6, 0.8φ ≤ φ_c ≤ 0.9φ) and Xu et al. [35] (4 ≤ n ≤ 13, 0.88φ ≤ φ_c ≤ 0.94φ). The initial leakage and subsequent remediation simulations (with a 1 × 10⁻¹² m² permeability) were repeated for six combinations of 2 critical porosity values of 0.8 and 0.88 (80 and 88% reduction of the original porosity) with three different power law components of 2, 6 and 10.

The calcite plug formed using the different porosity-permeability relation input parameters is very similar (Figure 7). A peak of calcite precipitation is observed next to the cell from which the reactive solution is injected. Above this interval, the microannulus is only partially clogged due to calcite precipitation. The relative insensitivity of the characteristics of the calcite plug to the porosity-permeability relationship can be explained by the nature of the remediation process. The process is designed to inject the reactive solution up to full clogging, meaning that with all parameters a full permeability reduction will be simulated. There is a small difference in the calcite formed in the upper part of the plug, with the 0.88φ_c–2n model yielding the most precipitation and a combinitation of 0.8φ_c and 10n the least. The used porosity-permeability parameters do have a large impact on the predicted time that is required to achieve full clogging. The time it takes to perform the remediation method ranges from 7 to 113 days (

Table 5. The predicted time of remediation up to full clogging of the microannulus.

Scenario (φc–η)	0.8–6	0.8–2	0.8–10	0.88–2	0.88–6	0.88–10
Remediation Time (d)	39	7	48	11	45	113

). This indicates the importance of the porosity-permeability parameters for the prediction of the duration of the remediation procedure and for the asessment of the related costs and overall feasibility.

4.3. Stability of the Plug in Time

After the remediation procedure, the reactive transport simulation is continued for 1 year to assess the stability of the calcite plug with time. This allows for equilibration of the system and continuation of diffusion and possibly flow. To assess the stability of the plug and the chemical evolution within the microannulus, two porosity-permeability scenarios were selected representing the most (0.88φ_c–2n) and least calcite precipitation (0.80φ_c–10n).

Throughout the microannulus, the pH remains around 4.6 in the post-remediation phase, indicating the wellbore environment is still acidic and is not buffered by the cement within the year after leakage remediation that was simulated. The caprock minerals show no significant reactions in this time period. The main observed process is the increase in calcite volume fraction throughout the microannulus (Figure 8a), the thin original plug is plotted for comparison. After the remediation method stops CO₂ leakage, the process of natural sealing becomes dominant throughout the micoannulus adjacent to caprock due to the absence of flow. This results in clogging by mainly calcite precipitation and minor amorphous silica precipitation. After one year of reactions following the remediation procedure, the 0.80φ_c–10n scenario yields full clogging of the microannulus (Figure 8a,b). The 0.88φ_c–2n scenario does have two sections of partial clogging, but full clogging of the rest of the microannulus (Figure 8b). The partial clogging above the original plug is due to absence of CO₂ which has been completely consumed, yielding a zero gas saturation (Figure 8c). The gas saturation is highest just below the plug and just above the CO₂ reservoir. The zero permeability of the plug continues to block the CO₂ flow and causes the gas saturation above the clogged level to decrease with time as CO₂ migrates and is consumed by reactions. The permeability impairment forms a pressure block, with the microannulus below the plug approaching the CO₂ reservoir pressure and the microannulus above the plug retaining hydrostatic pressure.

The two porosity-permeability scenarios both yield sigificant natural sealing after clogging by injection of the remediation fluid. This indicates that reduction of leakage due to the remediation procedure enhances the natural capability of the wellbore system to seal and form a barrier against future leakage.

5. Discussion and Conclusions

Despite the function of annular cement as a seal preventing oil, natural gas or stored CO₂ to migrate to aquifers or to the surface, wells are known to leak due to microannuli formed by processes such as cement shrinkage or pressure and temperature fluctuations [2,5]. The width of a microannulus formed tends to increase for larger temperature differences between the produced or injected fluid and the rock formation, which is especially relevant for cold CO₂ injection [36]. The higher risk of microannulus formation during CO₂ injection combined with high abandonment pressures asks for an assessment of CO₂ microannulus leakage and methods for leakage remediation. Due to the high reactivity of cement with carbonated brine, the chemical processes are key. A field-scale wellbore model was developed which successfully incorporates CO₂ migration by two-phase flow through a microannulus and diffusion of dissolved CO₂ into the adjacent caprock and cement. This enables the simulation of the complex reactive transport processes of a storage system, including a storage reservoir, wellbore cement with a continuous microannulus from reservoir to caprock, and caprock overburden. Despite the large scale of the model, it was successful in predicting the well-known small-scale reaction characteristics as found in experimental and (small scale) modelling studies [6,8,9,10,11,12].

Previous simulations showed an initial critical CO₂ leakage velocity of 0.1 × 10⁻⁵ m/s below which calcium released from dissolving cement minerals can diffuse towards the microannulus where calcite forms which clogs the microannulus and prevents further flow [18]. For our models, a low enough leakage rate to facilitate natural sealing was achieved with an initial microannulus permeability of 1 × 10⁻¹³ m². The possibility for natural microannulus clogging at low leakage rates is similar to the self-sealing potential of fractures in cement as demonstrated by e.g., Huerta et al. [16]. They discussed the critical residence time for the CO₂ fluid to be present in a cement sample for a fracture to close. However, natural sealing is not solely dependent on the flow rate as determined by the systems permeability. Nonuniformity of the microannulus geometry can lead to local changes in flow velocity affecting the sealing process [19]. The specific chemistry of the cement and rock formations has a large impact, with, for example, a high potential for calcite forming in the microannulus when the host rock is a carbonate [17]. Previous modelling showed anhydrite clogging of the microannulus related to relatively high sulphate concentration in the formation water of the caprock [18], whereas the chemical characteristics of the cement and surrounding formation water in our model led to dominant calcite clogging. The thermodynamics and kinetics of the cement phases still pose uncertainty in the chemical processes in the microannulus [17]. Our kinetic parameters for CSH and silica may be considered conservative, yielding only minor amorphous silica in the microannulus. In addition, the uncertainty and variability in the reactive surface area of cement phases may further affect the predicted leaching of cement and natural sealing process. A dedicated uncertainty assessment with varying parameters for dissolution/precipitation kinetics and mineral reactive surface areas was out of the scope of this study but would be needed to assess the sensitivity to natural sealing of the microannulus.

For high leakage rates when natural sealing is not predicted to occur, the process of microannulus clogging can be induced by adding calcium to the system [22,23,24]. We injected a calcium-rich brine to react with the dissolved CO₂ in the microannulus, yielding a full permeability decrease due to calcite precipitation, as graphically represented in Figure 9. There are large uncertainties in the clogging process regarding the porosity-permeability relation of mineral precipitation in a microannulus. As discussed by Ito et al. [23] and Druhan et al. [24], the porosity-permeability relation is of utmost importance for predicting effective leakage remediation. However, compared to our previous numerical modelling study [22] in which we injected the CO₂-reactive solution in an aquifer above a caprock leak, leakage remediation in the microannulus was predicted to be more successful and far less sensitive to the porosity-permeability relation. This is due to the confined nature of a microannulus and the more difficult placement of a plug above a caprock leak path. In our study, the uncertainty of the porosity-permeability relation was primarily expressed in the time it takes for remediation and not in the success of remediation. A longer remediation time is related to the larger amount of mineral precipitation and porosity reduction that is required to achieve full permeability reduction when using a more conservative porosity-permeability relationship. Hence, accurate design of the remediation procedure requires additional data on the porosity-permeability behaviour of a microannulus. The previous study [22], indicated the significance of the leakage rate on the success of leakage remediation. The design of the remediation procedure would require knowledge on the actual leakage rate or a numerical sensitivity study on the possible range in initial microannulus permeability and resulting leakage rate. With a higher initial leakage rate, injection of the reactive solution might require a higher injection pressure.

Permanent leakage remediation, considering long-term CO₂ storage, requires a chemically stable plug in the leak path. Our model results indicate that the formed calcite plug does not only remain stable, but that cessation of flow enables natural sealing in the microannulus at the level below the plug. The increase in sealing of the microannulus enhances the potential for intentional clogging as a remediation method. Future work could focus on the sensitivity of the intentional and subsequent natural clogging process to the chemistry of the CO₂ reactive solution and the cement and formation rock chemistry. The chemical nature of the plug is of less importance when subsequent natural sealing can take over the barrier function, even if a placed plug would degrade in time.