Experimental Evidence for Self-Limiting Reactive Flow through a

Aug 15, 2012 - Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, Austin, Texas 78712, United States. § United St...
0 downloads 10 Views 4MB Size
Download PDF
Article pubs.acs.org/est

Experimental Evidence for Self-Limiting Reactive Flow through a Fractured Cement Core: Implications for Time-Dependent Wellbore Leakage Nicolas J. Huerta,*,†,§ Marc A. Hesse,† Steven L. Bryant,‡ Brian R. Strazisar,§ and Christina L. Lopano§ †

Jackson School of Geosciences, The University of Texas at Austin, Austin, Texas 78712, United States Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, Austin, Texas 78712, United States § United States Department of Energy, National Energy Technology Laboratory, Pittsburgh, Pennsylvania 15236, United States ‡

S Supporting Information *

ABSTRACT: We present a set of reactive transport experiments in cement fractures. The experiments simulate coupling between flow and reaction when acidic, CO2-rich fluids flow along a leaky wellbore. An analog dilute acid with a pH between 2.0 and 3.15 was injected at constant rate between 0.3 and 9.4 cm/s into a fractured cement core. Pressure differential across the core and effluent pH were measured to track flow path evolution, which was analyzed with electron microscopy after injection. In many experiments reaction was restricted within relatively narrow, tortuous channels along the fracture surface. The observations are consistent with coupling between flow and dissolution/precipitation. Injected acid reacts along the fracture surface to leach calcium from cement phases. Ahead of the reaction front, high pH pore fluid mixes with calcium-rich water and induces mineral precipitation. Increases in the pressure differential for most experiments indicate that precipitation can be sufficient to restrict flow. Experimental data from this study combined with published field evidence for mineral precipitation along cemented annuli suggests that leakage of CO2-rich fluids along a wellbore may seal the leakage pathway if the initial aperture is small and residence time allows mobilization and precipitation of minerals along the fracture.

■

INTRODUCTION Leakage via conductive pathways in a cemented annulus remains a significant issue in the petroleum industry.1,2 Well construction is complex and breakdowns in planning or implementation can lead to issues with zonal isolation or the well’s ability to seal permeable subsurface formations. Zonal isolation can also be lost over time, due to operations that impose temperature and pressure gradients that lead to stress cycling on the well system.3 Typically in the petroleum industry a well is worked over to seal such pathways if they are detected. However, many wells continue to have undetected leakage pathways. The goal of geologic carbon dioxide (CO2) capture and storage (CCS) is to take CO2 that is captured from a point source and to inject it into a deep formation, thus reducing anthropogenic emissions of CO2 to the atmosphere. The principle candidates for storage are oil reservoirs currently flooded with CO2 for enhanced oil recovery, depleted oil and gas reservoirs, and saline aquifers. One significant challenge to the development of CCS as an industry is for regulators, operators, and stake-holders to quantify and limit the risks and impacts of CO2 leakage.4 Time-dependent rate of leakage of CO2 via abandoned wells remains poorly constrained, yet the probability of CO2-saturated brine interacting with wells is high.5 Experiments that expose wellbore cement to CO2saturated brine have shown that carbonic acid does react chemically with the cement, but a low permeability barrier to © 2012 American Chemical Society

degradation forms, limiting degradation on the time scales relevant to CCS for a properly completed well annulus.6−8 There is also evidence from field studies for the transport of CO2-rich fluids along a well’s cement-to-earth interface.9,10 Abundant chemical reaction and precipitation occurs along the interface, with no evidence for significant CO2 leakage. Our experiments model the pathway between a cement-tocement interface as a discrete fracture. Understanding of reactive transport in fractures, especially where precipitation and dissolution can occur, has improved since Berkowitz’s11 detailed review on the topic. However, the uniqueness of each system and complexity of coupling between flow field, chemical species, and reaction rates hinders development of a robust mechanistic model. A few recent studies do provide insights relevant to our present work, and their findings are presented below. Work on coupled dissolution/precipitation experiments with large aperture fractures showed that the key controlling factors in the opening or closing of fractures was fracture roughness and chemical composition of fluid in fracture.12 Rough walled fractures, which increase local mixing and Special Issue: Carbon Sequestration Received: Revised: Accepted: Published: 269

April 5, 2012 August 15, 2012 August 15, 2012 August 15, 2012 dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

Article

important to note the rates of reaction and species solubility are a function of fluid chemistry and cement composition, and our results are meant to highlight key types of behavior rather than reproduce exact downhole conditions. Sample Preparation. Cement cores are prepared from API Class H neat oil well cement (Lafarge, Montgomery, TX). Cement is prepared using a modified version of the API Recommended Practice 10B.19 Samples are mixed with a waterto-cement ratio of 0.38 and poured into cylindrical acrylic molds with diameter 2.54 cm and length 12.7 cm long. The samples are cured at 50 °C and ambient pressure for 3 days. The samples are then removed from their molds and allowed to cure under water and at ambient temperature and pressure for 28 days. To generate a flow path, the samples are placed into a loading frame and fractured using the Brazilian method.20 The fracture is then offset, i.e. displaced by a few millimeters in the plane of the fracture, to ensure an aperture that permits flow. The sides are sealed with latex polymer caulk (DAP KWIK SEAL) and the ends are trimmed flush. As a result of sample preparation no fracture has the exact same surface geometry or the same length. The small variation has a negligible effect on the phenomena studied here, but longer cores do increase residence time and facilitate the observation of key dynamic phenomena. Cement permeability is extremely low, and flow through the fracture is assumed to be the dominant pathway. Flow Equipment. The cement core was housed in a uniaxial Hassler cell (Phoenix Instruments), and a confining pressure was applied to the core to ensure flow through the fracture and prevent flow between the core and Viton rubber confining sleeve. To negate the effects of variable confining stress, which was an unavoidable limitation of the equipment setup, the sample fracture surface had to be conditioned to behave in a linear elastic manner. To achieve a linear elastic fracture, weak asperities were removed using an approach based on the work by Huerta et al.21 This procedure consisted of applying a cyclic loading and unloading of confining stress on the sample. We used an Enerpac manual hydraulic pump to increase confining pressure to 800 psi in 200 psi increments. Deionized water was continuously injected during the build up of confining pressure and after each increment we let the upstream pressure equilibrate. Confining pressure was backed off in similar steps down to 200 psi. The next cycle would be performed in the same manner, but we would make sure to remain below the first cycle’s maximum pressure, otherwise further inelastic fracture alteration would occur. A final cycle would be performed and then confining pressure set at the desired experiment condition (typically around 500 psi). A high pressure piston pump was used to inject aqueous fluid at constant rate into the core. A pressure transducer measured the confining and upstream pressures, while the downstream pressure was open to atmosphere and assumed to be constant. Occasionally a 5 mL sample of effluent was collected to measure pH. For all experiments, deionized water was injected initially to get a baseline measurement of the flow rate to pressure differential relationship and to get a baseline estimate of fracture conductivity. An effective hydraulic aperture (BHyd) is then estimated assuming the cubic law for flow through parallel plates holds22

residence time, promoted mineral precipitation and fracture closure. These experimental results confirm earlier numerical modeling on simple geometries.13 We wish to build on previous work and couple locally heterogeneous transport phenomena to reaction by allowing mineral dissolution/precipitation to change the local geometry. The motivation for our experiments was to understand if reactions that occur as acidic fluid leaks along a cement fracture tend to limit or enhance leakage over time. To reduce the experimental time scales we are working with a more aggressive analog system: the injection of hydrochloric acid (HCl), as opposed to carbonic acid, along a single natural fracture in a cement core. The analog system helps us define the key feedback mechanisms that control temporal evolution. More fundamentally, experiment results of a highly coupled system where reaction and transport in a heterogeneous fracture lead to complex dissolution and precipitation patterns will be useful in developing more robust models to predict leakage on the well scale.

■

EXPERIMENTAL METHODS Well Cement. Ordinary Portland Cement (OPC) is used in most wells to support the metal casing strings and prevent migration of annular fluids. The American Petroleum Institute (API) sets standards about specific chemical composition that comprises a given type of cement. Additionally, cement is often mixed with many additives that facilitate cement placement and improve key characteristics of the cement.14 Because there are many possible formulations used in the field, we focus on the most basic type of cement, often termed neat. Even in its most basic form, unhydrated clinker contains several chemically distinct solid phases. Once hydrated, the dominant phases are calcium hydroxide (portlandite) and calcium-silicate-hydrate (C−S−H). C−S−H is an amorphous network that makes up the structural backbone of cement and comprises approximately 70% of the mass. Portlandite is a crystalline phase that forms by secondary precipitation into the pores of the C−S−H structures it hydrates, making up around 15% to 20% of the hydrated cement’s mass.14 Portlandite is thermodynamically more stable in large crystals and is known to dissolve and redeposit into larger pores or voids within the cement (Ostwald ripening), often precipitating euhedral crystals in open pore space and in open fractures of cement that is not fully cured. Important secondary phases are hexacalcium aluminate trisulfate hydrate (ettringite) and calcium aluminate monosulfate (AFm). Acid Attack on Cement. Acid attack on cement is principally influenced by the acid concentration and subsequent precipitation of salts.15,16 We chose to use HCl to allow control of pH over a wide range. In both carbonic acid and hydrochloric acid attack, the process is initiated when a low pH fluid invades the high pH fracture fluid. Solubility of hydrated cement phases increases as pH decreases and cations are liberated (e.g., Ca, Al, Fe). Precipitation then occurs when aqueous phase cations mix with high pH pore fluid ahead of the acid front allowing precipitation. In the case of carbonic acid exposure, calcite has been shown to be the dominate precipitate.6−9 In the HCl case, poorly crystalline phases rich in calcium and aluminum were identified.17 Calcite precipitation has also been previously observed with the source of carbonate assumed to be from atmospheric CO2.18 Thus the mechanism for attack and possibility of salt precipitation occurs in both the down hole and the model system. However is it

⎡ Q L ⎤1/3 BHyd = ⎢12μ ⎥ ⎣ ΔP W ⎦ 270

(1)

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

Article

Table 1. Experiment Parameters and Key Results

The aqueous phase viscosity (μ) is assumed to be 10−3 Pa-s, flow rate (Q) is constant during each experiment, pressure differential (ΔP) is measured from the upstream pressure transducer (as the downstream is open to atmospheric pressure), and sample length (L) and sample width (W) are assumed constant and set to the bulk dimensions of the core. Thus, changes in upstream pressure are related to changes in flow geometry (true flowing width and aperture distribution). Fracture Surface Analysis. Upon completion of the experiment, the sample core halves were separated and stored in sealed containers until analysis. Macroscopic images were acquired with a digital camera. Electron microscopy was then used to study fracture surface alteration for changes in cement chemistry and microstructure. An FEI-Quanta 600 FEG Environmental SEM (ESEM) equipped with an Oxford INCA energy dispersive spectroscopy (EDS) system with a Large-field Secondary Electron (LFD) detector and a Back Scatter Electron detector (BSED) were used to collect images. The unpolished and uncoated samples were analyzed in low vacuum mode with an acceleration voltage of 20 keV and nominal working distance of ∼10 mm for EDS mapping. Care was taken to analyze surface areas that were as flat as possible. Images and EDS spot chemical analyses were collected throughout various regions of the fracture surface (varying working distance where necessary to account for surface roughness). Additionally elemental EDS maps were collected over several hour periods to better monitor the overall chemical alteration of the cement along the fracture pathway. Higher magnification EDS spot analyses were used to confirm trends seen in the maps to ensure that chemical trends were not a result of surface roughness effects.

parameters were flow rate, injected acid concentration, and total acid injected. Parameters unique to each sample were width (typically 2.54 cm), sample length, and initial hydraulic aperture size. The hydraulic aperture was calculated using core parameters, flow rate, and pressure differential (eq 1). Initial average velocity was calculated assuming flow across the maximum available flow area. Total acid injected was calculated from the product of injected acid concentration, flow rate, and experiment time. Upstream pressure evolution, effluent pH history, postexperiment fracture surface images, and electron microscopy of characteristic experiments are presented in this section. Upstream Pressure Evolution. In each experiment, the upstream pressure followed one of three general trends (Table 1). The plot in Figure 1 shows an example of the most common behavior (case A), which was a pressure spike and decay that flattened out but remained above the initial upstream pressure. There is an initial period of constant pressure, which typically occurs before 0.2 mmol HCl are injected and is uncorrelated to flow rate or initial hydraulic aperture. It is most likely related to the displacement of some dead volume and before the acid begins to significantly interact with the cement. Upstream pressure begins to rise very quickly until it reaches some maximum pressure, often more than twice the initial pressure. At this point upstream pressure begins to decay (with a much slower slope). The pressure curve eventually flattens out or has a small negative slope but always remains above the initial pressure. Maximum height, width, and curve smoothness are uncorrelated to any single parameter. The most significant rise and fall of inlet pressure typically occurs before 1.0 mmol acid are injected, and most curves have gentle pressure decay for several more mmol of HCl injected. The plot in Figure 2 shows an example of case B, with a significant monotonic increase in pressure over time. In this example the pressure increase was so high that the pressure transducer’s limit was reached and the experiment had to be

■

RESULTS AND DISCUSSION The results for twelve samples are shown in Table 1. Results are grouped based on flow behavior. Controlled experimental 271

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

Article

spanned up to 3 orders of magnitude. First order correlation to onset of pressure increase indicates low pH, slow flow rate, and small aperture tend to cause earlier pressure buildup. The plot in Figure 3 shows the final behavior type (case C), a small decay in pressure drop that flattened out or had a shallow

Figure 1. Sample PN-1. Image on the top is of a fracture surface. Acid injection is from left to right. Red box is the area analyzed using electron microscopy (see Figure 4 and the Supporting Information). This is an example of an experiment that developed a distinct reaction channel, which was a subset of the available fracture surface. Red arrows highlight examples of ‘islands’ of unreacted cement within a reacted channel. Plot on the bottom shows pressure differential (solid blue line) and effluent pH (dashed red line) as a function of acid injected. This plot shows an example of case A pressure history.

Figure 3. Sample BSD1. Fracture surface image shows a tan coloration across most of the surface. Note that the crack across the hsurface occurred post experiment. The plot shows an example of case C behavior, with an early pressure differential decrease (solid blue line). Effluent pH also shows a very fast breakthrough in this example (dashed red line).

decreasing slope. This behavior occurred in three experiments. In all three experiments there was a correlation to high flow rate, relatively high injected acid concentration, and medium to large initial aperture size. Effluent pH History. We chose to track effluent pH evolution as an indication of reaction progression as it is bound between the pH of fluid fully equilibrated with cement (initial value) and the pH of the injected acid (value when all reaction has completed). Prior to acid flow, the effluent pH was measured from nonacidified water flow through the fracture. For all experiments this pH reached a steady state of ∼10. This is assumed to be a function of injected fluid pH, injection rate, residence time, and diffusion from pore fluid and cement solids. In the absence of flow, the pH of static fluid filling the crack is assumed to be ∼12.3 based on equilibrium with solid portlandite in the cement.6 Effluent pH history for acid injection tests fell into two types of behavior. The first type occurred in all experiments with flow profiles described in case B above, and an example is shown in Figure 2. The effluent pH remains high, near what we would expect the pH of cement’s equilibrated fracture fluid to be. The three experiments that retained an elevated pH are the same that showed a significant and sustained increase in pressure differential over time. The most common effluent pH behavior occurred in all of the case A and C experiments. They had a stable initial pH, then a period of sharp decay, which approaches the injected acid pH. In none of the experiments did effluent acid concentration reach injected acid concentration; despite large volumes injected. The shape of the pH curve has no obvious correlation to parameters (Figure 1 and Figure 3). Some

Figure 2. Sample AAT-15. On the image (top of the figure) caulk has been identified to distinguish between white precipitate. The left side of the surface (inlet) has a more tan color, while the right side (outlet) is whiter. A case B example is shown the plot, where the pressure is flat until the point where is begins a very sharp increase (solid blue line). The effluent pH shows no breakthrough (dashed red line).

terminated. The flow path is assumed to be significantly reduced at this point, and the reaction would have likely sealed the flow path if the flow would have been driven by a constant pressure drop as expected in the field. All three experiments that yielded this pressure history had a slow flow rate and low acid concentration but had a range of initial apertures. Onset to pressure build up varied widely (from ∼0.02 to 1 mmol acid injected), and upstream pressure increases over initial pressure 272

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

Article

practical considerations to note when looking at the shape of the curve are that the time between switching from deionized water flow to acid flow allows fluid in the fracture to equilibrate with the cement and determine the initial effluent pH. Higher flow rate and shorter cores allow faster fluid breakthrough. Interpretation of Fracture Surface. Color images of the fracture surface were collected after the acid experiment to get an initial description of the reaction. Flow is from left to right in all images (Figure 1, Figure 2, and Figure 3). Unreacted cement is gray. Reacted surfaces have a tan to brown coloration, consistent with previous studies in static reactors.17,18 A white mineral precipitate is often associated along reacted boundaries between unreacted cement and the dominant flow paths but is not to be confused with caulk along sample edges. The nature of the reacted pathway along the surface of the fracture is quite complex. Every experiment yielded a unique reaction pathway, and three examples are shown in Figure 1, Figure 2, and Figure 3. The nature of the pathways is different and allows us to organize the patterns into the following three groups described below and in Table 1. In the first group, there was limited visual evidence for reaction (Figure 2). In the samples that fell into this classification there was either no apparent change or a small length of reacted surface that did not penetrate the entire fracture length (as in the presented image). There is always evidence for white material across the surface but a lack of reacted pathway development. The second type is the broad reaction pathway (Figure 3). This is evidenced by a general tan coloration across large areas of the fracture surface in the example shown. White material is present but less visible except along the sides. The third and most prevalent type is a distinct reaction pathway across a subset of the total fracture surface (Figure 1). The channel can be a single pathway, or multiple smaller channels, and is generally wider at the inlet and narrows toward the outlet. There is often a white material along the edges of the channel boundary. Images of the fracture surface yield a general sense of the reaction pathway; additional microscopic understanding of the structural and elemental changes is key to describing reaction mechanism in our system. The red box in Figure 1 was an area studied using SEM and EDS. While these techniques are qualitative in nature due to sample roughness, care was taken to ensure that the trends seen were not surface roughness artifacts. General trends have been identified (described below) and were consistent across the different samples analyzed. Figure 4 shows a backscatter image of the red boxed area shown in Figure 1, and there are several distinct textures and zones that can be identified. Inside the orange band, the unreacted cement has a fine grained uniform texture. In between the orange and red bands is an intermediate zone with larger, well-formed crystals and more grainy texture, corresponding to the location of the white band seen in the Figure 1 image. Above the red line is a zone of reacted material presumed to be the dominant flow path composed of plate-like fine grained material, separated by subpolygonal gaps. EDS mapping (see the Supporting Information) in the same area shows sharp element gradients across the fracture surface. Calcium is present in both the unreacted and intermediate area but is significantly depleted in the reacted zone. Higher magnification EDS spot analysis confirmed these trends. Iron and silicon are more prevalent in the reacted channel and are likely more apparent due to leaching of calcium from the

Figure 4. Sample PN-1. BSE image of fracture surface showing three distinct zones (Figure 1 − red box). Zone 1 is within the orange line. Zone 2 is between the orange line and the red line. Zone 3 occupies most of the area outside the red line.

channel. Aluminum appears to have increased concentration in the intermediate zone. Classification of Fracture Surface. Based on the fracture surface images and electron microscopy we can identify three distinct zones. Zone 1 − Unaltered zone. This zone has typical cement phases and elemental compositions. On some samples there is indication of carbonation of the fracture surface likely due to dissolved CO2 in the water use to store the sample. The unaltered zone typically remains present in experiments with little acid injected or along the sides of the fracture. When this zone sits within a reacted channel, we infer the local aperture was small and thus not exposed to significant flux of acidic water. Zone 2 − Precipitation zone. This zone corresponds to the white material on the macroscopic images. It can occur as diffuse bands on fracture surfaces that show no or diffuse alteration. Most commonly this zone separates zone 1 and zone 3. This zone is characterized by distinct euhedral mineral precipitation, rich in calcium and aluminum. Qualitative identification of minerals using micro X-ray diffraction (μXRD) shows this zone contains calcite and brownmillerite (see the Supporting Information). Zone 3 − Dissolution zone. This zone is characterized by varying shades of tan to brown and can be diffuse or form distinct channels. Dissolution of calcium rich minerals from Zone 1 and Zone 2 leads to a surface that is fine grained but has what look like polygonal cracks and is relatively enriched in silicon and iron. Degree of coloration may be an indication of depth of penetration and progressive dissolution of initial cement phases.6,17 Proposed Mechanism for Channeling. The occurrence of reacted channels that occupy only a fraction of the fracture surface area is remarkable. For single phase flow, there is no reason a priori to expect fluid to be excluded from any portion of the fracture volume. The effective apertures determined from hydraulic conductivity measurements are consistent with this expectation. Single phase unreactive flow experiments using a tracer dye independently confirm this expectation. Thus a reasonable expectation for the reactive transport response is the emergence of classical, essentially one-dimensional (in the direction of flow) zonation of chemical compositions.23,24 273

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

Article

According to classical theory, the zonation could approach a steady state, or it could evolve with time, possibly with fronts propagating in the direction of flow for suitable ratios of flow velocity and reaction rate.12 Contrary to expectation, we observe zonation that appears to be controlled by fluid dynamics, rather than by geochemical processes. Channels confine the dissolution zone; unreacted areas lie outside the channels and sometimes even form “islands” between some channels (Figure 1). The fronts that bound the reacted channel appear to be propagating transverse to the main flow direction, if they are moving at all. The zonal development is quite similar to observations from previous studies of diffusion dominated reaction into cement.6−8,17,18 The source of this unexpected complexity must be the coupling between acid/cement reactions and the flow field. Here we outline features of this coupling that contribute to the emergence of these patterns. As low pH HCl is injected into the flow path it interacts with the hydrated (and any remaining unhydrated) cement phases. Portlandite is the first to be attacked and totally dissolves, releasing calcium and hydroxide ions. The next phase to be attacked is C−S−H, followed by ettringite, and then Afm.17,18 These phases also release calcium but leave behind insoluble reaction products rich in silicon, iron, and aluminum. This solid is more porous than the unaltered cement8 and might undergo some volumetric contraction.25 Any minor amount of carbonation on the fracture surface will also be dissolved, releasing calcium and carbonate ions. As the active flow path is reacted, and available is calcium leached, the acid begins to diffuse through the amorphous silicate layer to attack the cement below. Ahead (downstream) of the injected acid front (before acid breakthrough) and on the lateral edges of the active flow channel a large gradient in pH, cation, and anion concentrations occurs. The pH of cement pore fluid is significantly higher (pH ∼12.3) than the injected acid, and as the fluids mix several phases become insoluble and precipitation occurs along this interface. The precipitated phases are rich in calcium and aluminum and form distinct crystals that are either euhedral rhombs or elongate crystals. Minerals identified in the intermediate zone were calcite and brownmillerite, with little evidence for portlandite. The source of carbonate for calcite precipitation was assumed to be either minor carbonation on fracture surface or from atmospheric equilibrium of injected fluid. Because the fracture is rough-walled, the aperture distribution is spatially heterogeneous. Thus rapid precipitation at the beginning of the experiment can create numerous local dead ends near the flow inlet, diverting subsequently injected acid elsewhere. The dissolution/precipitation process yields a moving front that propagates at a speed proportional to local flow speed. Thus the diversion process is self-reinforcing and quickly leads to a channel;26,27 the leading edge of the channel moves rapidly because the constant injection rate forces fluid to keep moving downstream, while the edges of the channel transverse to the flow direction move slowly. If the flow field is slow enough and gradient in concentration between acid and cement small, we propose that precipitation at the front of the flow path is sufficient to cause an elevated pressure spike as in case B. If acid concentration is high, residence time low, and flow path large, then a small decrease in upstream pressure was seen (case C). This decrease could be the result of unblocking of flow restrictions via portlandite dissolution, aperture increase via volumetric contraction of the remnant amorphous silicate,25

or widening of the reacted channel. However, these phenomena must be balanced with the fact that in most experiments where a channel still forms (case A) the net result is a significant increase in upstream pressure. Time Dependent Leakage Flux. The practical result from our present work is that even though we inject an acid into cement, fracture opening and wormhole development did not occur. In fact an amorphous silicate phase remains which prevents fracture opening and becomes a diffusive boundary that slows reaction into the cement. Further, evidence for selflimiting behavior is seen when calcium rich water interacts with high pH fluid in front of the channel tip and to the sides of the flow path. Where this mixing occurs, distinct minerals precipitate which could act as a restriction in flow. In the cases with small decrease in upstream pressure, we hypothesize that the flow path was large and that any precipitation occurring during the duration of the experiment was not significant enough to constrict flow or that flux of fluid (and core length) did not allow significant residence to promote sufficient mineral precipitation to block the flow path. The experiments reported here are for constant flow rate. Thus even when pressure gradient increased significantly as precipitated mineral accumulated, it was possible to break through such an accumulation mechanically. The expected downhole conditions differ from these experimental conditions in several key respects: significantly less acid concentration (pH between 4 and 5), smaller fluid flux driven by a constant pressure gradient (as opposed to a constant flow rate), the presence of carbonate anion, and much longer residence time within the fracture. Further the complexity of specific cement formulations, downhole conditions (e.g., pressure and temperature), and brine composition affects the coupled system. As a result, reaction rates, mineral solubility, and local channel development could vary. More work is needed to understand how more realistic downhole conditions affect wellbore leakage. However, based on our work self-limiting or even self-sealing behavior seems likely due to local dissolution precipitation reactions in the fracture.

■

ASSOCIATED CONTENT

S Supporting Information *

EDS maps of the fracture surface, BSED images of key minerals with EDS spot measurements, and μXRD analysis of secondary minerals. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Partial support for this work comes from the Geological CO2 Storage Industrial Affiliates Program at the Center for Petroleum and Geosystems Engineering at UT-Austin. This work was also supported by the Carbon Sequestration program of the U.S. DOE National Energy Technology Laboratory. We would also like to thank Karl Jarvis for the abstract image, Barbara Kutchko for discussions on cement chemistry and carbonic acid attack of cement, and Corinne Disenhof for help with μXRD. 274

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275

Environmental Science & Technology

■

Article

(24) Walsh, M. P.; Bryant, S. L.; Schlechter, R. S.; Lake, L. W. Precipitation and dissolution of solids attending flow through porous media. AIChE J. 1984, 30, 317−328. (25) Matteo, E. N.; Scherer, G. W. Experimental study of the diffusion-controlled acid degradation of Class H Portland cement. Int. J. Greenhouse Gas Control 2012, 7, 181−191. (26) Sherwood, J. D. Stability of a plane reaction front in a porous medium. Chem. Eng. Sci. 1987, 42, 1823−1829. (27) Chadam, J.; Hoff, D.; Merino, E.; Ortoleva, P.; Sen, A. Reactive infiltration instabilities. IMA J. Appl. Math. 1986, 36, 207−221.

REFERENCES

(1) Bourgoyne, A. T., Jr.; Scott, S. L.; Manowski, W. A review of sustained casing pressure occurring on the OCS; Technology Assessment & Research Program 008DE; Minerals Management Service: Washington, DC, March 1999. (2) Watson, T. L.; Bachu, S. Evaluation of the potential for gas and CO2 leakage along wellbores. SPE Drill. Completion 2009, 24, 115− 126. (3) Ravi, K.; Bosma, M.; Gastebled, O. Safe and economic gas wells through cement design for life of the well; SPE 75700: Richardson, TX, 2002. (4) Bachu, S. CO2 storage in geological media: Role, means, status and barriers to deployment. Prog. Energy Combust. Sci. 2008, 34, 254− 273. (5) Zhang, M.; Bachu, S. Review of integrity of existing wells in relation to CO2 geological storage: What do we know? Int. J. Greenhouse Gas Control 2011, 5, 826−840. (6) Kutchko, B. G.; Strazisar, B. R.; Dzombak, D. A.; Lowry, G. V.; Thaulow, N. Degradation of well cement by CO2 under geologic sequestration conditions. Environ. Sci. Technol. 2007, 41, 4787−4792. (7) Kutchko, B. G.; Strazisar, B. R.; Lowry, G. V.; Dzombak, D. A.; Thaulow, N. Rate of CO2 attack on hydrated class H well cement under geologic sequestration conditions. Environ. Sci. Technol. 2008, 42, 6237−6243. (8) Duguid, A; Scherer, G. W. Degradation of oilwell cement due to exposure to carbonated brine. Int. J. Greenhouse Gas Control 2010, 4, 546−560. (9) Carey, J. W.; Wigand, M.; Chipera, S. J.; Wolde Gabriel, G.; Pawar, R.; Lichtner, P. C.; Wehner, S. C.; Raines, M. A.; Guthrie, G. D. Analysis and performance of oil well cement with 30 years of CO2 exposure from the SACROC Unit, West Texas, USA. Int. J. Greenhouse Gas Control 2007, 1, 75−85. (10) Crow, W.; Williams, D. B.; Carey, J. W.; Celia, M.; Gasda, S. Wellbore integrity analysis of a natural CO2 producer. Energy Procedia 2009, 1, 3561−3569. (11) Berkowitz, B. Characterizing flow and transport in fractured geological media: A review. Adv. Water Res. 2002, 25, 861−844. (12) Singurindy, O.; Berkowitz, B. The role of fractures on coupled dissolution and precipitation patterns in carbonate rocks. Adv. Water Res. 2005, 28, 507−521. (13) Dijk, P.; Berkowitz, B. Precipitation and dissolution of reactive solutes in fractures. Water Resour. Res. 1998, 34, 457−470. (14) Nelson, E. B. Well Cementing; Schlumberger Educational Services: Sugarland, TX, 1990. (15) Zicica, V.; Bajza, A. Acidic attack of cement based materials − a review. Part 1. Principles of acidic attack. Constr. Build. Mater. 2001, 15, 331−340. (16) Beddoe, R.; Dorner, H. Modelling acid attack on concrete: Part I. The essential mechanisms. Cem. Concr. Res. 2005, 35, 2333−2339. (17) Chandra, S. Hydrochloric acid attack on cement mortar − An analytical study. Cem. Concr. Res. 1988, 18, 193−203. (18) De Ceukelaire, L. The effects of hydrochloric acid on mortar. Cem. Concr. Res. 1992, 22, 903−914. (19) API. Recommended Practice for Testing Well Cements; API Recommended Practice 10B; American Petroleum Institute: Washington, DC, 1997. (20) Guo, H.; Aziz, N. I.; Schmidt, L. C. Rock fracture-toughness determination by the Brazilian test. Eng. Geol. 1993, 33, 177−188. (21) Huerta, N. J.; Bryant, S. L.; Strazisar, B. R.; Kutchko, B. K.; Conrad, L. C. The influence of confining stress and chemical alteration on conductive pathways within wellbore cement. Energy Procedia 2009, 1, 3571−3578. (22) Witherspoon, P. A.; Wang, J. S. Y.; Iwai, K.; Gale, J. E. Validity of cubic law for fluid flow in a deformable rock fracture. Water Resour. Res. 1980, 16, 1016−1024. (23) Helfferich, F. G. The theory of precipitation/dissolution waves. AIChE J. 1989, 35, 75−87. 275

dx.doi.org/10.1021/es3013003 | Environ. Sci. Technol. 2013, 47, 269−275