Materials
Two experiments were conducted using cylindrical peridotite samples (diameter = 25 mm, length = 25 mm), one of which possessed a slit and the other possessed a fracture. The peridotite sample employed herein was a harzburgite specimen obtained from the Horoman Peridotite Complex in Hokkaido, Japan. The peridotite samples were mainly composed of olivine (Mg1.82Fe0.14)SiO4 (82%), orthopyroxene (Cr0.02Mg1.74Fe0.11Al0.09)(Si1.98O6), and clinopyroxene (Cr0.04Fe0.05Ca0.84Mg0.88Al0.13)(Si1.98O6) (13%), with lower amounts of Cr-rich spinel (Mn0.01Mg0.73Fe0.32)(Al1.03Cr0.92)O4 (2.5%), serpentine ( < 1%), and magnetite (<1%) components (Table S1). Serpentine (antigorite) (Al0.02Mg2.54Fe0.18)(Si2.12O5)(OH)4 was also detected in the form of thin veins. The chemical compositions of the minerals were determined by electron microprobe analysis (EPMA), as described in the Analyses subsection.
The slit sample was prepared by cutting along the diameter, and the resulting flat surfaces were polished using diamond powder with a particle size of 1 μm. Subsequently, one of the flat surfaces was channeled using a disk polishing machine to create a canal with a width of 16 mm and a depth of 0.4 mm. The fractured sample was prepared using a Brazilian fracture test in which a tensile fracture was induced parallel to the cylinder axis (i.e., along the macroscopic flow direction during the flooding experiment).
An aqueous GLDA solution was prepared from a 40 wt% aqueous solution (pH ~12.7) of GLDA-Na4 (C9H9NNa4O8) purchased from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan). The prepared solution contained GLDA-Na4 at a concentration of 20 wt% ( ~ 0.6 M), and its pH was adjusted to pH 4 using HNO3 (60–61%, Kanto Chemical Co., Japan).
Experimental procedures
The experimental setup and procedures employed in the current study were described in detail previously24,25,26,27,30. The chelating agent flooding experiments were conducted using the experimental setup shown in Fig. S4. In each trial, a Viton-sleeved sample with end plugs attached to both the inlet and outlet faces was placed horizontally in the pressure vessel. Water was pumped into the vessel at a constant pressure, subjecting the sample to a confining pressure of 15 MPa. A temperature of 200 °C was maintained using a mantle heater. The initial permeability of the fractured sample was determined using Darcy’s law by pumping pure water at a constant flow rate of 5.00 mL·min−1. The permeability (k) was calculated as follows:
$$k=frac{Qmu L}{pi {r}^{2}varDelta P}$$
(2)
where Q is the flow rate, μ is the dynamic viscosity of the injected fluid, L and r are the length and radius of the sample, respectively, and ΔP is the differential pressure of the fluid between the inlet and outlet faces of the sample. The viscosities of water and GLDA solution used for the permeability calculations were 134 and 304 µPa·s, respectively27. During the flow-through experiments, water flowed out of the sample through a backpressure regulator adjusted to 5 MPa. After determining the initial permeability, the chelating agent solution was injected into the sample at the same flow rate for 2 h. Two experiments (named Run_frac and Run_slit) were conducted by injecting the chelating agent solution into the fracture and slit samples at flow rates of 5.00 and 1.00 mL·min−1, respectively. For the Run_Frac sample, the permeability was calculated at discrete steps every 10 min (Fig. 4b) using Darcy’s Law. For this purpose, 2 min-averaged differential pressure data were employed, along with the corresponding fluid dynamic viscosities depending on whether water (time <0 min) or the GLDA solution (time > 0 min) was injected. For the Run_slit sample, the differential pressure was not measured because the sample possessed an unmeasurably high permeability owing to the open space (0.4 mm thickness). Furthermore, the flow rate of the Run_slit experiment was set at 1.00 mL·min−1 to ensure consistent conditions and allow comparison with a previous study27.
Aliquots of the effluent samples were collected throughout the experiment, including the water injection stage (corresponding to time = 0 min) and thereafter GLDA solution injection (time > 0 min). In the case of the Run_frac experiment, effluent samples (10 mL) were collected every 2 min during the initial 20 min of GLDA solution injection, and every 10 min thereafter. For the Run_slit experiment, the effluent samples were collected at 10-min intervals, including both stages of injection, i.e., water and GLDA solution injection.
Total GLDA solution volumes of 120 and 600 mL were injected into the Run_slit and Run_frac samples, respectively. These injected volumes corresponded to fluid-to-rock mass ratios of ~10:1 and 50:1. These ratios were selected to ensure sufficient fluid availability for reactive transport during the experiments.
Analyses
Solution chemistry
The effluent aliquots were collected after flowing through the samples. More specifically, these aliquots were collected at room temperature and were rapidly filtered (0.45 μm filter) and diluted (10 vol% of collected sample, 85 vol% of milliQ water, and 5 vol% HNO3) to prepare for chemical analysis. The concentrations of Al, Ca, Cr, Fe, Mg, and Si were analyzed using inductively coupled plasma–optical emission spectroscopy (ICP-OES; Agilent 5100). Instrumental calibration was performed using in-house standards at 0, 1, 5, 10, 20, 50, and 100 ppm for all elements, with an additional standard at 200 ppm for Si, Mg, and Fe. Analytical blanks (0 and 5 ppm) were run every 15 samples, and each sample was analyzed three times automatically by the equipment, with the average value being reported. Water and GLDA solution blanks (i.e., before injection) were prepared to provide baseline elemental values for subtraction, allowing the elements eluted from the rock samples to be accurately reported.
Mineral distributions and chemical compositions of the minerals
The flat, diamond-polished side of the slit sample was carbon-coated and subjected to quantitative elemental analysis using EPMA (JEOL JXA-8200) before the experiment, to determine the chemical compositions of the minerals. The natural and synthetic standards used for calibration were wollastonite (Si and Ca), rutile (Ti), corundum (Al), hematite (Fe), manganosite (Mn), periclase (Mg), albite (Na), feldspar (K), and eskolaite (Cr). The counting times for the peak and the background were 10 and 5 s, respectively. The measurements in the slit sample were conducted quantitatively at an accelerating voltage of 15 kV and a beam current of 120 nA, with randomly dispersed points measured for each mineral in the rock and averaged to obtain the chemical composition (Table S1). Furthermore, a normalized atoms per formula unit was calculated and included in Table S1. The carbon coating was then removed by rapid polishing (e.g., ~10 s using 1-µm diamond powder) after the EPMA measurements and scanning electron microscopy (SEM) observations (described in the following paragraph). Furthermore, the slit sample was also analyzed using µXRF (TORNADO M4), without a carbon coating, to map the elemental distribution at the surface. A tube voltage of 300 kV was employed, along with a resolution of 20 µm and a dwell time of 30 ms.
Variation in the surface chemical composition of the slit sample
To determine the variations in the surface elemental distributions and evaluate possible preferential element extraction by the chelating agent, the Si, Mg, and Fe contents on the slit surfaces were visualized before and after the dissolution experiments using µXRF mapping (TORNADO M4, Bruker Co., parameters as detailed above). Furthermore, quantitative EPMA was conducted after the dissolution experiments (applying a carbon coating layer before measurement and following the calibration procedure previously described) to determine the mineral compositions at comparable locations to those examined before the experiments. These measurements were conducted once on each point reported in Table S2. Finally, SEM (Hitachi SU8000) observations were performed at a voltage of 15 kV with a working distance of 8.00 mm, and a carbon coating applied to the slit surface, to observe changes in the mineral textures before and after the dissolution experiments.
X-ray CT scanning
X-ray CT (ScanXmate D225RSS270) was employed to analyze the effects of chelating agent injection on the samples, scanning both the fractured and slit samples before and after the experiment. For this purpose, the following settings were employed for all scans: tube voltage = 120 kV, tube current = 150 µA, voxel size = 20 μm, and number of projections = 2000. For the fractured sample, Ilastik software43 was used to enhance the visualization of the phases present in the X-ray CT images (Fig. 5a). Additionally, Molcer Plus 3D image visualization and processing software (White Rabbit Corp., Tokyo, Japan)24 was used to compute the distributions of pores and fracture apertures (i.e., voids) with sizes larger than 8 voxels. The slices before and after the experiment were manually correlated by observing the unreacted minerals (e.g., spinel) to ensure that comparable pore volumes were located in similar positions. Blender 3D software was used to visualize the three-dimensional (3D) distributions of the voids. Furthermore, the porosity was calculated as a function of distance from the inlet side by dividing the volume of voxels corresponding to voids (air) by the volume of the sample. For this purpose, the porosity in each X-ray CT slice (thickness of 20 μm) was determined. The void volume was measured by subtracting the volume integral of the voxels that corresponded to the solid body (i.e., exceeding a solid/air threshold of 15,000 in 16-bit TIFF images) from the volume of the sample.
For the slit sample, Molcer Plus 3D software was used to produce 3D models of both the flat and cut sides before and after the dissolution experiment, aligning the models using their boundaries and the unreacted minerals (e.g., spinel) as anchor points. The System for Automated Geoscientific Analyses (SAGA) geographic information system (GIS)44 was used to create topographic maps from the 3D models of the exposed surfaces. The void map (i.e., of the voids formed through mineral dissolution) was then calculated by subtracting the topographic maps after and before GLDA solution injection (i.e., to discard preexisting voids) and categorizing the voids according to the elemental maps obtained by μXRF.
Flow simulations
Flow simulations were employed to analyze the qualitative relationship between the flow path and the aperture distribution. Using the porosity calculated for each slice of the fractured sample (i.e., from the X-ray CT scans detailed above), two-dimensional (2D) maps (resolution = 20 μm) of the aperture distribution were generated across the fracture plane (i.e., yz) before and after the flow-through experiment. These 2D maps were then employed to simulate the fluid flow by applying the Reynolds equation, which models the steady-state laminar horizontal flow of a viscous fluid in a fracture, as follows45,46:
$$frac{partial }{partial y}left(frac{{{e}_{y,z}}^{3}}{12mu },frac{partial {P}_{y,z}}{partial y}right)+frac{partial }{partial z}left(frac{{{e}_{y,z}}^{3}}{12mu },frac{partial {P}_{y,z}}{partial z}right)=0$$
(3)
where ey,z represents the local aperture, µ denotes the fluid dynamic viscosity, and Py,z represents the local fluid pressure. The boundary conditions for this equation were set as Q = 0 at y = 0 and y = 22 mm, 5.06 MPa at z = 0 mm, and 5.00 MPa at z = 22 mm (flow from top to bottom, Fig. 5c). The z-axis grid was cropped to exclude the wide fractures near the inlet and outlet sides of the sample. Furthermore, the differential pressure was set to 0.06 MPa to match the initial condition of the sample before injection of the GLDA solution, and to directly compare the flow rate improvement of the sample after the experiment. The viscosity of water was set to 135 μPa·s.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.