In Situ Investigation of Fluid-Rock Interactions at Ångstrom Resolution

Abstract

The flow of aqueous phases through porous media shapes the world. It changes the courses of rivers, erodes mountainsides, dissolves historical monuments, builds caves, regulates the carbon cycle, and changes porosity in aquifers and hydrocarbon reservoirs. Despite the significance of this process both to natural phenomena and human endeavors, fundamental insight into the mechanisms that drive it at the Ångstrom scale is scarce. Here, we employ environmental transmission electron microscopy (ETEM) to visualize interactions between water and various types of porous media, including sandstone, limestone, and dolostone samples, at Ångstrom resolution in real time. Unlike other techniques presented in the literature, ETEM allows for direct visualization of both the aqueous phase and the porous media. As expected, dissolution and recrystallization were observed to significantly impact the shapes, sizes, and interconnectivities of the pores in the rocks. However, for the first time, we provide direct in situ experimental evidence that adsorption-induced strain can also significantly impact the above-mentioned parameters, especially in micropores. In some cases, adsorption was observed to reduce effective pore size by more than 60%. The magnitude of the change in pore size was found to vary with pore wall mineralogy and pore geometry. This work presents the first direct Ångstrom-resolution observations of liquid adsorption films and adsorption-induced strain.

Key Points

• Rock-fluid interactions are directly observed in situ at Ångstrom resolution

• The thicknesses of adsorbed water films on pore walls are measured

• Adsorption-induced strain is shown to alter pore size by more than 60%

1 Introduction

Comprising 75% of the surface of the Earth (Houlton et al., 2018), sedimentary rocks are one of the most abundant types of porous media. They constitute mountain ranges and ocean floors, but beyond their obvious geologic and geographic significance, sedimentary rocks are also commercially important. They have been used for millennia as construction materials and as sources of valuable minerals, including zinc, copper, lead, iron, and gold. In the past century, their exploitation has grown exponentially as repositories for greenhouse gas sequestration and reservoirs from which hydrocarbon fuels are produced.

The natural abundance of sedimentary rocks on the surface of the Earth and the nature of their use in man-made applications, such as construction, necessitate their frequent interaction with water. Aqueous phases and sedimentary rocks are thought to interact in two ways: physically through abrasion and migration and chemically through dissolution and recrystallization. Collectively, we refer to these interactions as weathering. The effects of each interaction differ depending on the type of sedimentary rock. Water is believed to interact with sandstones primarily through abrasion (McBride & Picard, 1987; Small et al., 2015) and migration. Carbonates, including limestones and dolostones, are prone to dissolution and recrystallization (Chave & Schmalz, 1966; Morse, 1986). However, at the Ångstrom scale, and even the nanoscale, in situ experimental evidence elucidating how these processes occur and how they affect the mechanical, chemical, and transport properties of sedimentary rocks is rare.

Much of what is known about weathering at the Ångstrom scale is the result of studies using atomic force microscopy (AFM). Using this technique, it has been observed that both calcite (Miyata et al., 2017) and dolomite (Hu et al., 2005) crystals submersed in water dissolve along their lattice planes. However, in studies of rock-fluid interactions, AFM is characterized by two major shortcomings: First, due to the nature of such scanning techniques, large segments of the rock cannot be characterized concurrently (Heath & Scheuring, 2019). In other words, it may not be possible to observe relatively large features, such as pores, altogether in the same time step. Second, AFM can only be used for direct (Bohr et al., 2010; Rüecker, 2018) visualization of the solid substrate, not the liquid solvent. Therefore, fluid wetting is inferred from the adhesion force. Weathering occurs bilaterally between water and rock. Thus, although significant progress has been made using AFM, many outstanding questions regarding the weathering of sedimentary rocks by water remain. For example, how do naturally occurring textural features, such as pore size, contribute to weathering?

Techniques that are not inhibited by the shortcomings of AFM are needed to answer these questions. One such technique is environmental transmission electron microscopy (ETEM). Similar to conventional transmission electron microscopy, ETEM can be used to observe the physical structures of materials at Ångstrom resolution. For example, the ETEM images in Figure 1 exhibit the textural features of dolostone lamellae at sub-Ångstrom resolution, including surface roughness and atomic structure. Unlike conventional transmission electron microscopy, ETEM can be used for in situ measurements of fluid-solid interactions at fluid pressures of up to 30 mbar. In other words, ETEM utilizes a parallel electron beam to image all elements, including solids and fluids, within the field of view simultaneously at high resolutions.

Figure 1

Examples of surface roughness of dolostone lamellae. (a) Image Resolution = 0.191 Å per pixel. Note that atomic structure is visible in (b).

In the geosciences, micropores are defined as pores with effective diameters between 0.2 and 10 μm (Hasiuk et al., 2016). Their significance lies in their abundance. For example, French castles in Loire valley were constructed with truffeau limestone, in which microporosity can constitute up to 77% of the overall porosity (Beck et al., 2003). Likewise, 83% of pores in the red sandstone used to build the ancient Phenom Bakheng Temple in Cambodia are micropores (Wedekind et al., 2018). In nature, micropores make up 48% of the total porosity of the carbonate reservoir, Ghawar, the largest oilfield in the world (Cantrell & Hagerty, 2003). These examples in both man-made monuments and nature indicate that micropores significantly impact the geometric and, hence, morphological and transport properties of sedimentary rocks (Mehmani & Prodanovic, 2013; Okabe & Blunt, 2007). In other words, ETEM visualization of water-micropore interactions may provide more insight into sedimentary rock weathering than AFM studies on single, smooth crystals (Bohr et al., 2010; Hu et al., 2005; Miyata et al., 2017).

Here, we use ETEM to conduct in situ dynamic imaging of water interacting with dolostone, limestone, and sandstone samples. These samples were obtained from Fond du Lac, Edwards, and Berea outcrops, respectively, and their mineralogies were verified using quantitative evaluation of minerals by scanning electron microscopy (QEMSCAN). We determine the effects of water on micropore size and shape for each of the samples using a vapor-phase ETEM imaging technique. Then, we repeat the procedure using a liquid aqueous phase to determine the potential effects of the electron beam on the interactions and identify the consequences of dissolution and recrystallization. By comparing both types of experiments, we provide direct in situ experimental evidence that, in addition to dissolution and recrystallization, adsorption-induced strain can contribute significantly to weathering.

2 Materials and Methods

2.1 Sample Characterization

Core samples from the Berea, Edwards, and Fond du Lac outcrops were used to represent different types of sedimentary rock. The mineralogies of the samples were verified using a QEMSCAN 650F (Thermo Fisher Scientific). To prepare the samples for mineral analysis, core cuttings were set in Struers Epofix epoxy, after which they were ground and polished with a Struers LabPol-35 mechanical polisher. During polishing, alcohol-based lubricating fluid (DP-Lubricant Blue, Struers) was employed to keep the samples cool and to prevent swelling of the clays in the samples. After polishing, the samples were sputter coated with carbon using an EMS150R ES carbon coater (Electron Microscopy Sciences). The carbon is electron transparent and was used to prevent charging during imaging.

Imaging was carried out on the QEMSCAN at 15 kV with a spot size of 5.77, such that the beam current was optimized to within 0.05 of 10.00 nA. To determine the mineralogies of samples, the QEMSCAN uses energy dispersive X-ray scattering detectors to measure elemental concentrations. By comparing the elemental data to mineralogical definitions stored in a Species Identification Protocol, the mineralogies of the samples were identified. The Species Identification Protocol “O&G 15 kV v 3.7” was used for all three samples. The resulting mineral maps are shown in Figure 2. The Fond du Lac, Edwards, and Berea samples were observed to be dolostone, limestone, and sandstone, respectively.

Figure 2

Mineral maps of the (a) Fond du Lac, (b) Edwards, and (c) Berea samples used to prepare the lamellae. Each mineral map is 2.84 mm² in area with a resolution of 0.69 μm.

2.2 Sample Preparation

Additional rock samples from the three cores were used to prepare ETEM lamellae. We emphasize that, unlike the core cuttings used for QEMSCAN characterization, these samples were not carbon coated. Although carbon coating reduces the effects of charging such that high-quality QEMSCAN images can be generated, it can alter the surface chemistries and wettabilities of samples. Therefore, it was not employed simply to maintain the natural wettabilities of the ETEM lamellae, further preventing interference with the rock-water interactions.

To prepare the lamellae from the samples, a Helios G3 focused ion beam scanning electron microscope (FIB-SEM) with an EZ-Lift micromanipulator (Thermo Fisher Scientific) was used. Throughout the liftout procedure, a gallium ion beam was employed at 30 kV with a variety of beam currents from 0.24 pA to 64 nA. Currents below 40 pA were used for observation and guiding the micromanipulator, while currents above 40 pA were used for milling. Larger currents were used for low-spatial-accuracy bulk-out milling, while smaller currents were used for precisely thinning the lamellae. Platinum was used to solder the lamellae first to the micromanipulator needle (which was also platinum) and then to the TEM liftout grids (Ted Pella, part no. 10GC02T).

2.3 Image Processing

Fiji (ImageJ) software (Schindelin et al., 2012) was used for all image processing. For the Ceta dynamic images, Fiji was used to extract selected frames and to adjust their contrast and brightness.

Angle and length measurements of the rock features were done using Fiji's built-in measurement tools. Area measurements were made by first adjusting the contrast and brightness of the images and then adjusting the thresholding to select only the pores. To achieve suitable brightness and contrast for thresholding, Fiji's math exponential function was used to enhance contrast, after which the contrast and brightness were further manipulated by hand. In some cases, where significant enough contrast enhancement could not be achieved using Fiji's built-in features, the pore was traced pixel-by-pixel by hand. In most cases, tracing was done by hand for the frames recorded immediately after water injection, as the influx of water decreased the overall contrast of the images.

2.4 Image Acquisition

A Titan G2 environmental transmission electron microscope (Thermo Fisher Scientific) was used for all ETEM imaging. It employs a Ceta 2 camera, which allows for both static and dynamic imaging. The latter is recorded using the software feature “Ceta dynamic recording.” Using the dynamic recording feature, multiple frames were taken at specified time intervals to create video. For all of the images obtained using Ceta dynamic recording in this work, a frame rate of 1/s was employed. All imaging was carried out at 300 kV and at ambient temperature (70.0°F ± 1°F) with a single-tilt holder in the absence of a cold finger. In our previous work (Barsotti et al., 2020), it was found that employing the cold finger during experiments with water caused a significant amount of the water to freeze out onto the cold finger, lowering the maximum achievable vapor-phase pressure.

During the vapor-phase imaging, distilled water was injected directly into the ETEM column using a custom fluid injection system (Barsotti et al., 2020). This system employs a dual cylinder 5000 Series Quizix Pump (Chandler Engineering) and its associated software (PumpWorks, Chandler Engineering) first to regulate the pressure of the liquid water to below 500 mbar (the factory defined pressure cutoff on the ETEM for fluid injection) and then to inject it into the ETEM column to raise the column pressure to approximately 30 mbar (Barsotti et al., 2020). Note that the saturation pressure of water at ambient temperature was approximately 24 mbar for all experiments.

For both the liquid-phase and vapor-phase experiments, only freshly distilled (within 72 h of distillation) water was employed. Polyethylene 7 mL transfer pipets graduated to 3 mL (Globe Scientific, part no. 135030) were used for placing water droplets on the samples during liquid-phase imaging. For the liquid-phase imaging, the airlock vacuum cycle was set to a time of 5 min in order to evaporate most of the water from the sample before it was loaded into the ETEM column.

3 Results and Discussion

3.1 Rock-Surface Weathering During Vapor-Phase Experiments

During the vapor-phase experiments, ETEM column pressure was modulated using either a series of leak valves to inject fluid and increase pressure or a turbomolecular pump to evacuate fluid and decrease pressure. Starting at high vacuum (pressures below 10⁻⁵ mbar), the turbomolecular pump was disengaged from the column, and distilled water at 70°F was injected through the leak valves to increase the column pressure to approximately 28 mbar. At 28 mbar, the leak valves were closed, and the water vapor was allowed to equilibrate with each type of rock for approximately 2 h. Then, the ETEM column was opened to the turbomolecular pump, which evacuated the water, returning the column to high vacuum over the course of 1 h.

Throughout each 3 h experiment, the water-rock interactions were observed in real time using Ceta dynamic recording. Selected frames from the recordings are shown in Figures 3-5 for dolostone, limestone, and sandstone, respectively. In transmission electron microscopy, an electron beam is transmitted through the sample. For vacuum or gas, the beam is not attenuated or is hardly attenuated. For liquids, the beam is somewhat attenuated. And for solids, the beam can be partially or fully attenuated. In other words, image greyscale is directly related to material density. As shown in the figures, the material with the highest greyscale intensity (i.e., the lightest color) is the pore, which for the purposes of this study, is occupied by empty space (i.e., vacuum) or gas. The material with the lowest greyscale intensity (i.e., darkest color) is the solid mineral matrix. Adsorbed films, which are closer in density to liquids, possess an intermediate greyscale, as will be shown in subsequent figures.

Figure 3

Weathering of a Fond du Lac dolomite sample by water. Water was injected at 1 min 13 s. Water was removed from the sample by subjecting the microscope column to vacuum at 1 h 47 min 00 s. Image resolution is 3.563 Å per pixel. (a) 0 min 0 sec; 1.83x10-6 mbar; 0.16 μm², (b) 3 min 43 sec; 27.5 mbar; 0.14 μm², (c) 8 min 43 sec; 27.5 mbar; 0.13 μm², (d) 23 min 52 sec; 27.0 mbar; 0.11 μm², (e) 39 min 51 sec; 26.4 mbar; 0.083 μm², (f) 1 hr 44 min 27 sec; 24.9 mbar; 0.060 μm², (g) 1 hr 51 min 22 sec; 4.15x10-1 mbar; 0.078 μm² (h) 2 hr 17 min 42 sec; 2.19x10-4 mbar; 0.094 μm² (i) 3 hr 35 min 6 sec; 4.29x10-5 mbar; 0.10 μm².

Figure 4

Weathering of an Edwards limestone sample by water. Water was injected at 0 min 52 s. Water was removed from the sample by subjecting the microscope column to vacuum at 1 h 48 min 5 s. Image resolution is 3.563 Å per pixel. (a) 0 min 0 sec; 6.77x10-6 mbar; 0.60 μm², (b) 8 min 7 sec; 28.4 mbar; 0.56 μm², (c) 13 min 15 sec; 28.4 mbar; 0.54 μm² , (d) 47 min 56 sec; 26.7 mbar; 0.52 μm², (e) 1 hr 47 min 02 sec; 25.1 mbar; 0.41 μm² , (f) 1 hr 48 min 25 sec; 1.16 mbar; 0.43 μm² , (g) 1 hr 54 min 21 sec; 1.41x10-1 mbar; 0.51 μm² , (h) 2 hr 1 min 23 sec; 5.48x10-4 mbar; 0.56 μm², (i) 3 hr 16 min 5 sec; 6.70x10-5 mbar; 0.63 μm².

Figure 5

Weathering of a Berea sandstone sample by water. Water was injected at 0 min 40 s. Water was removed from the sample by subjecting the microscope column to vacuum at 1 h 46 min 58 s. Image resolution is 3.563 Å per pixel. (a) 0 min 0 sec; 8.36x10-6 mbar; 0.092 μm², (b) 2 min 13 sec; 28.3 mbar; 0.090 μm², (c) 12 min 51 sec; 28.8 mbar; 0.061 μm², (d) 46 min 9 sec; 27.1 mbar; 0.059 μm², (e) 1 hr 46 min 47 sec; 25.5 mbar; 0.063 μm², (f) 1 hr 47 min 15 sec; 1.62 mbar; 0.067 μm², (g) 1 hr 48 min 55 sec; 3.28x10-1 mbar; 0.085 μm², (h) 2 hr 21 min 39 sec; 2.26x10-4 mbar; 0.10 μm², (i) 2 hr 53 min 45 sec; 1.01x10-4 mbar; 0.11μm² .

All three rock types exhibited a similar trend: Within the first 10 min of water injection, a layer of adsorbed water formed on the micropore walls. As the adsorbed water interacted with the pore walls, the pore sizes decreased. By measuring the effective pore size directly from the images in Figures 3-5, it can be seen that at 2 h, immediately preceding water evacuation, the sizes of pores in the dolostone, limestone, and sandstone decreased with respect to the initial pore sizes by 62.5%, 31.7%, and 31.5%, respectively. After 2 h, when the water was evacuated from the ETEM column, the pore sizes increased. In this way, the final pore size of the dolostone was 33.9% smaller than its initial size, while the final sizes of the limestone and sandstone increased by 3.4% and 17.3%, respectively. Interestingly, the two carbonates behaved differently from each other. This difference is discussed in more detail in the latter part of this section.

Note that throughout this work, we employ effective pore size, which we define as the total or absolute pore size less the adsorbed film. We present the effective pore size rather than the absolute pore size because adsorbed phases are commonly theorized to be immobile (Dolinar & Trček, 2019). Therefore, the adsorbed film does not contribute to fluid transport (Tokunaga, 2011).

The changes in pore size observed in Figures 3-5 may affect the transport properties of the rock. For example, some correlations show absolute (Nishiyama & Yokoyama, 2017) and relative permeability to depend exponentially upon pore size (Burdine, 1953). Permeability controls the rate by which fluid can flow through disordered pore space (McKinley & Warke, 2007). Furthermore, during multiphase flow, the adsorbed water layer may significantly impact wettability alteration, for example, the aging process in hydrocarbon reservoirs. Wettability alteration also plays a key role in controlling the multiphase flow properties of porous media.

Based on the observed adsorbed film, we attribute these changes in pore size to adsorption-induced strain. Micropores are particularly vulnerable to adsorption-induced strain due to their large surface-to-volume ratio (Gor et al., 2017). This contraction and expansion of pores due to adsorption and desorption, respectively, has been studied indirectly both at the macroscale, using techniques such as dilatometry and ellipsometric porosimetry (Kim & Guyer, 2014; Van Opdenbosch et al., 2016), and at the nanoscale, using techniques such as small-angle X-ray scattering (Gor et al., 2017; Kim & Guyer, 2014; Van Opdenbosch et al., 2016). However, until now, this process has not been experimentally characterized in situ at the Ångstrom scale.

By measuring the adsorbed film thicknesses and effective pore sizes directly from Figures 3-5, we can correlate the adsorbed film thickness to both the water pressure and the pore size. An example of this for the sandstone micropore is shown in Figure 6. In Figure 6a and 6b, we plot the thicknesses of the adsorbed film at three different locations in the sandstone micropore against the water pressure and the effective pore size, respectively. The sites at which the film thicknesses were measured and example film thickness measurements are given in Figure 6c and 6d, respectively. We chose to measure the adsorbed film thickness at three randomly chosen locations on the pore wall because, although the ultimate thickness of the adsorbed film depends on global properties of the system, like pore size and water pressure, it may also be affected by local properties of the rock, such as surface chemistry, surface roughness, and surface curvature. Thus, even though the thickness of the adsorbed film may vary from one location to another, the general trend of the adsorbed film thickness in relation to pressure should be consistent. In accordance with our hypothesis, the pore size and adsorbed layer thicknesses are inversely proportional.

Figure 6

Water adsorption film thickness on Berea sandstone: (a) variation of adsorption film thickness and water pressure throughout the entire experiment; (b) variation of adsorption film thickness and pore size throughout the entire experiment; (c) the three sites at which adsorption film thickness was measured; (d) sample adsorption film thickness measurements. Image resolution is 3.563 Å per pixel. Note that the curves for adsorption film thickness are identical in (a) and (b); they are repeated in both plots to emphasize their relationship to water pressure and effective pore size.

During adsorption, the sandstone micropore gradually closed along its diagonal axis, becoming more slit-like. This behavior was expected, as the adsorbed water films in the corners of the pore were in closer proximity to one another than the films along the sides of the pores. Thus, the attractive forces between the pore corners were larger. However, in both the dolostone and limestone, the adsorption process occurred violently, such that the adsorbed water pulled different parts of the pore walls toward the centers of the pores at different rates. Even though the distortions to the pore walls of both carbonates were nonuniform, the process occurred more intensely in the limestone, leading to the formation of bubble-like structures from the nonuniformly stretched parts of the pore wall. Because the adsorbed water stretched the pore walls of the limestone more than the dolostone, the limestone walls were more elastic. Therefore, when the water was evacuated from the sample, the limestone walls rebounded, allowing the pore to more closely regain its original size and shape than the dolostone micropore. This explains the differences in the magnitudes of the changes to pore size between the two carbonates during adsorption and desorption. Nevertheless, the distortions to both the dolostone and limestone pore walls made it difficult to distinguish between the rock samples and the adsorbed phases. To illustrate this, measurements of the adsorbed film thicknesses prior to and during distortion of the pore walls are shown in Figure 7. Because these distortions to the pore walls were observed as quickly as 3 min after water injection, plots similar to Figure 6 could not be generated for the limestone and dolostone samples.

Figure 7

Water adsorption film thickness on Fond du Lac dolostone (a and b) and Edwards limestone (c and d). Prior to significant distortion of the pore wall, the water film was clearly visible on both types of rock (a and c). However, once the pore walls were distorted, the adsorption film could no longer be distinguished (b and d). In other words, the stretching of the pore walls precluded distinction of the interfaces between the adsorbed phase and the pore wall and between the adsorbed phase and the bulk or free water. Without the accurate distinction of both interfaces, adsorption film thickness measurements are extremely challenging to undertake. Image resolution is 3.563 Å per pixel.

Interestingly, the adsorbed film thicknesses that we measured on the carbonates before distortion are approximately one order of magnitude larger than those measured elsewhere using indirect methods, such as X-ray reflectivity (Bohr et al., 2010). For example, whereas we measured an adsorbed film thickness of 12.6 nm on the limestone sample in Figure 7a, Bohr and coworkers measured a film thickness of approximately 1.5 nm on a calcite crystal (Bohr et al., 2010). Likewise, we observed the thicknesses of the adsorbed films to depend strongly upon the amount of water (i.e., pressure), whereas Bohr and coworkers’ measurements were only weakly dependent on relative humidity. One plausible explanation for this difference may be the types of samples used. While we employed rock samples, Bohr and coworkers employed single, smooth calcite crystals (Bohr et al., 2010). Therefore, the greater abundances of surface roughness and porosity in our samples may also affect the adsorbed film thickness. It has been reported in the literature that surface roughness (Chiang et al., 2016), porosity (Thommes et al., 2015), and pore wall curvature (Coasne & Pellenq, 2004) may lead to the accumulation of thicker adsorbed films than smooth, nonporous surfaces. Such differences between studies on pure crystals and rock samples emphasize the need for experiments with more natural or realistic materials to bridge the knowledge gap between the weathering of single minerals and sedimentary rocks.

Finally, for the vapor-phase experiments, dissolution was not found to significantly impact weathering. This is evident from two observations: First, the pore size of the sandstone changed considerably, even though the effect of dissolution on sandstone is known to be insignificant. Second, the changes to the pore sizes of the sandstone and limestone samples were similar in magnitude. In other words, if dissolution were the dominant factor controlling the change in pore size, we would expect the changes in the pore sizes of the carbonates to be more pronounced compared to that in the sandstone. Therefore, adsorption-induced strain is the more dominant factor during vapor-phase imaging.

The comparative insignificance of dissolution in these experiments is likely due to the low pressure at which the water was introduced to the sample. Even though we introduced water to the microscope column above its saturation pressure, it persisted as a vapor, likely due to electron beam heating (Rykaczewski et al., 2011). The vapor phase is evident from the relative clarity of the images. If the ETEM column were filled with liquid, the density of the liquid would obstruct the electron beam, preventing the observation of sharply distinguishable rock features. The presence of a vapor, rather than a liquid, reduces the concentration gradient that drives dissolution. It has been reported in the literature, that in gaseous environments, carbonate dissolution strongly depends upon relative humidity (López-Arce et al., 2011). However, in cases where water is more abundant, such that a stronger concentration gradient exists, dissolution can also significantly affect pore size and shape, as is discussed in the subsequent section.

3.2 Dissolution and Recrystallization During Liquid-Phase Experiments

During dry imaging, beam damage was not observed to significantly affect pore size and shape. However, the electron beam was observed to preclude the condensation of the bulk water in the vapor-phase experiments even at experimental pressures above the saturation pressure of the water. Therefore, to further explore the impacts of the amount and phase of the water on weathering, liquid-phase experiments were performed.

For the liquid-phase experiments, each rock sample was first imaged dry and then removed from the ETEM column. Without removing the rock sample from the ETEM holder, one droplet of water was placed upon it. The water and rock were allowed to equilibrate for 1 min at atmospheric pressure in air at 70°F. Then, the sample was loaded back into the ETEM. The ETEM's airlock vacuum and turbomolecular pumps were used to evacuate all water from the sample before the column valves were opened. We emphasize that vacuum alone was used to dry the sample to prevent any contact or contamination with foreign particulates that could affect the natural water-rock interaction. In all cases, imaging was not carried out until the ETEM column pressure decreased below 5 × 10⁻⁶ mbar. This vacuum condition typically required approximately 20 min to achieve. Achieving this vacuum level prior to opening the column valves allowed us to observe the aftermath of the water-rock interaction without introducing any beam-water effects (Barsotti et al., 2020; Laanait et al., 2015).

To obtain multiple frames with which the progression of the interaction could be observed, the same procedure was repeated for each sample, employing subsequent water-rock equilibrium times of 2, 5, and 10 min. These times were chosen arbitrarily but were consistent across all three rock types. Note that equilibrium times longer than 10 min were attempted but resulted in total dissolution of the limestone samples, due to the strong interaction of the water with calcite and the small quantity of the rock employed. The more pronounced dissolution effects in the liquid-phase experiments compared to the vapor-phase counterparts are due to the larger amounts of water (i.e., repetitive contact with fresh distilled water at atmospheric pressure rather than 28 mbar) used in the former. In all cases, the samples in order of dissolution rate from fastest to slowest were limestone, dolostone, and sandstone, respectively.

Selected frames from the liquid-phase images for the limestone, dolostone, and quartz samples are shown in Figures 8, 9, and 10, respectively. The images are presented in terms of cumulative rock-water equilibrium time. For example, in Figure 8l, where a cumulative equilibrium time of 8 min is given, the sample was equilibrated with water outside of the ETEM column three separate times for 1, 2, and 5 min. Likewise, in cases where a cumulative time of 18 min is reported, the sample was equilibrated with water four separate times for 1, 2, 5, and 10 min.

Figure 8

Liquid-phase imaging of Edwards limestone. In (a–d), bubbling and nanoparticles are observed; resolution of 2.256 Å per pixel. In (e–h), dissolution and nanoparticles are observed; resolution of 1.095 nm per pixel. (i–l) Image resolution of 2.256 Å per pixel. Note that the sample used for (i–l) was fully dissolved after 8 min.

Figure 9

Liquid-phase imaging of Fond Du Lac dolostone. The resolution of (a–d) is 1.095 nm per pixel. The resolution of (e and f) is 2.895 Å per pixel. The resolution of (g and h) is 3.563 Å per pixel.

Figure 10

Liquid-phase imaging of Berea. The resolution of the images is 3.563 Å per pixel. (a) 0 min, (b) 1 min, (c) 3 min, (d) 18 min.

Figure 8d confirms our observation of limestone bubbling in the vapor-phase experiments while also exhibiting other types of interactions. Dissolution is observed to increase the size and interconnectivity of pores. For example, as shown in Figure 8h, dissolution at the grain boundaries provides extra throats to connect pores. Likewise, in Figure 8l, dissolution increases pore size by dissolving the pore walls. However, in some cases, widening of the pores by dissolution was counteracted by recrystallization. In Figure 8d and 8h, recrystallization of the dissolved rock upon evaporation of the water decreased pore size via the growth of new crystals on the pore walls.

Dolostone, shown in Figure 9, behaved similarly. In Figures 9a–9d, the pore size was observed to change, seemingly due to dissolution and water adsorption. In Figures 9e–9h, distortion of the rock and recrystallization are also shown. However, although dissolution occurred for both the limestone and dolostone samples, its mechanism differed between the two. For the limestone, dissolution was observed to take place along the calcite planes of cleavage, resulting in surface roughness at approximately 102 and 78 degrees, shown in Figure 11a, which is typical of the 1014 calcite structure. Likewise, as shown in Figure 11b, recrystallization on the calcite surface resulted in the growth of rhomboidal crystals exhibiting similar lattice angles. Conversely, well-defined dissolution planes and crystals were not observed for dolostone. For example, in Figure 11c, no distinct lattice angles were observed along the dissolution plane, while recrystallization (Figure 11d) of the dolostone occurred in the growth of oblong particles. Side-by-side comparison of limestone and dolostone dissolution and recrystallization are shown in Figure 11. These results are in agreement with the work of Liu and coworkers, who classified the type of large, oblong crystal shown in Figure 11d as protodolomite (Liu et al., 2019).

Figure 11

Dissolution in Edwards (a), recrystallization in Edwards (b), dissolution in Fond du Lac (c), and recrystallization in Fond du Lac (d). Note that the crystals in Fond du Lac have no distinct structure. The resolutions of the images in (a–d) are 2.256 Å per pixel, 4.463 Å per pixel, 1.095 nm per pixel, and 4.463 Å per pixel, respectively.

Our observation of fast (starting at 3 min cumulative water equilibration time in Figure 9g) and abundant recrystallization on the dolostone surface at room temperature lies in contrast to some studies presented in the literature. Experimental observations of dolomite recrystallization on dolostone surfaces at room temperature are notoriously difficult (Morrow, 1982). For example, in studies employing AFM of dolomite crystals in supersaturated aqueous solutions, dolomite recrystallization has been observed to be inhibited at temperatures below 120°C, (Berninger et al., 2017) resulting in unsuccessful attempts to recrystallize dolomite at (25°C) with some attempts yielding no crystal growth at all even after timespans as long as 32 years (Land, 1998). Furthermore, at elevated temperatures, when recrystallization does occur, it occurs in smooth layers that mimic the original, smooth dolomite surface, (Berninger et al., 2017) rather than as large oblong particles, such as we observe here.

The difference between the results that we present here and those available in the literature may be due to differences in experimental methodologies. In our work, the same rock sample is used both for dissolution and recrystallization, with evaporation of the saturated water being used to drive crystal growth. In the literature, a fresh dolomite crystal was used for recrystallization experiments with artificially created supersaturated aqueous solutions providing the ion sources for recrystallization (Berninger et al., 2017; Land, 1998). The results that we present are similar to those recently published by Lui and coworkers, who used evaporation to crystalize dolomite on clay (Liu et al., 2019).

New crystal growth was also observed for the sandstone in Figure 10. Interestingly, Figures 8-10 all show that, after initial recrystallization, subsequent exposures to water did not re-dissolve the recrystallized particles. Furthermore, for the limestone, with each successive interaction with water, new crystals, once nucleated, grew larger as additional solutes were deposited onto them.

Because recrystallization and dissolution were significantly more pronounced in the liquid-phase experiments than their vapor-phase counterparts, they also significantly affected pore size and shape in the former tests. Our direct visualizations of dissolution along planes of cleavage and of recrystallization through the physical growth of new crystals prove that the changes to pore size and shape can be reasonably attributed to them. Taking the results of both the liquid-phase and vapor-phase experiments into consideration, we conclude that dissolution, recrystallization, and adsorption-induced strain can impact the shapes, sizes, and interconnectivities of pores in sedimentary rocks. The relative significance of each interaction depends on the interplay of pore geometry, mineralogy, and the thermodynamic state of the water.

4 Conclusions

Direct observations of the weathering of limestone, dolostone, and sandstone samples were reported for the first time at Ångstrom resolution in real time and then at selected times using vapor-phase and liquid-phase experimental techniques, respectively. These two types of experiments provide direct evidence that adsorption-induced strain, dissolution, and recrystallization can all alter the sizes and shapes of pores in sedimentary rocks. Although dissolution and recrystallization have long been identified as contributors to the weathering of sedimentary rocks, we present the first direct in situ experimental evidence for adsorption-induced strain as a source of weathering. In our experiments, during the adsorption of water, we observed the pore sizes in various sedimentary samples to change by up to 62.5% with the magnitude of the change varying depending on pore wall mineralogy and pore geometry. Such large changes in pore size characteristic of adsorption-induced strain may alter the transport properties of the rock, including permeability; while the adsorbed layer, itself, may affect wettability during multiphase flow processes.

Acknowledgments

We gratefully acknowledge the financial support of Hess Corporation, Thermo Fisher Scientific, and the School of Energy Resources at the University of Wyoming. We thank Henry Plancher of Piri Research Group at the Center of Innovation for Flow through Porous Media for providing freshly distilled water on a daily basis. We extend our gratitude to Chuck Moe and Peter Stokes from Thermo Fisher Scientific for their technical support on the ETEM.

Open Research

Data Availability Statement

Research data associated with the article can be accessed at https://doi.org/10.5281/zenodo.3677444, https://doi.org/10.5281/zenodo.3677365, and https://doi.org/10.5281/zenodo.3677290.