Characteristics of brown zone formed by weathering around fracture in the Hiroshima granite and estimation of iron source for iron oxide precipitation

Abstract

Brownish zones enriched in iron (oxyhydr)oxides often occur around fractures during granite weathering. The elemental distribution, color, iron oxide concentration, mineral assemblage, porosity, and iron diffusivity in the brownish zone around a fracture in the Hiroshima granite were investigated. From the fracture to the interior of the rock matrix, the iron concentration first increased with distance from the fracture, reaching a maximum at 4–10 mm, decreasing at 10–21 mm, and remaining constant in the whitish areas at greater distances. Iron oxides, thought to be mainly goethite and ferrihydrite, were distributed at the grain boundaries and concentrated in the altered portions of plagioclase. The selective iron dissolution method revealed that 0.32 wt% Fe is present as iron oxides in the brown zone. Numerical calculations of the reaction and transport of Fe near the fracture showed the following: under conditions where Fe does not precipitate, the dissolution of biotite takes less time to supply Fe for the formation of iron oxides than the diffusion of Fe from groundwater in the fracture; under conditions where Fe precipitates, the Fe that diffuses from the fracture almost completely precipitates in the vicinity of the fracture and is unlikely to be a major source of iron oxide in the brown zone. A large proportion of the brown zone Fe appears to have been derived from the dissolution of nearby biotite.

Introduction

When granite is weathered, its color often changes to yellowish or brownish compared to the color of the unweathered portions. This color change is generally considered to be due to the formation of iron (oxyhydro)oxides (hereafter abbreviated as iron oxides) (Nagano and Nakashima, 1989; Yoshida and Yamamoto, 2014). In the early stages of granite weathering, the rock matrix surrounding the fractures often becomes yellow or brown, and brown bands are often observed, particularly near the front of the colored zone. As weathering progresses, color changes spread throughout the rock. Basic knowledge of mass transport and reactions in and around the rock fractures is essential for understanding rock weathering and has also received much attention and discussion, especially in terms of its importance for the underground disposal of radioactive waste (Yoshida et al., 2003; Akagawa et al., 2004; Yu et al., 2020).

To precipitate iron oxide, dissolved iron must first be supplied. The solubility of Fe(III) is very low around pH-neutral conditions near the ground surface. Using PHREEQC (Parkhurst and Appelo, 2013) and the llnl.dat database to calculate the solubility of Fe(OH)₃ at pH = 7–9, 20°C, and equilibrium with atmospheric O₂ and CO₂, the Fe(III) concentration is about 56 ppb maximum (Fe(II) concentration is negligibly low). Thus, iron migration is presumed to occur when Fe(II) is unoxidized, when it takes time for precipitation to occur after Fe(II) is oxidized to Fe(III), or when Fe(III) migrates in the form of colloids. The dissolved Fe(III) will eventually be precipitated. There are several reaction pathways for the precipitation of iron oxides, which depend on factors such as pH, Eh (pe), and temperature (Schwertmann and Cornell, 2000). An example of the formation of iron oxides is a reaction involving the supply of O₂ and the effect of an increase in pH due to the dissolution of plagioclase and calcite within the rock (Yoshida and Yamamoto, 2014), as follows:

4Fe²⁺ + O₂ + 10H₂O = 4Fe(OH)₃ + 8H⁺

Iron oxides gradually accumulated inside the rock. Akagawa et al. (2004) displayed that the total Fe concentration in the brownish zone around the fracture of granite in Gifu, Japan, was approximately 15% greater than that of the host rock and interpreted that Fe(II) in the groundwater flowing in the fracture migrated into the rock interior and was oxidized to Fe(III) by oxygen that later flowed in from the fracture and precipitated. Similar observations and interpretations were made for the Mannari granite in Okayama, Japan (Yoshida et al., 2008).

Considering that the brown zones often look like the remains of groundwater infiltration from fractures and that the total Fe concentration in the rock matrix is often higher in the brown zones than in the unweathered area, it seems natural to assume that Fe was supplied from the fractures. However, the source of the Fe is difficult to determine. In addition to the supply from fractures, Fe may also be supplied by the in situ dissolution of primary minerals. The higher Fe concentration in the brown zone compared to the unweathered area cannot be explained by in situ dissolution alone, however it can be explained if Fe dissolves at a position distant from the fracture in the rock matrix and migrates in the direction of the fracture. A method to elucidate the source of Fe is to consider the net Fe influx and the kinetics of Fe migration and dissolution around the brown zone. However, to the best of our knowledge, studies examining the source of Fe in granite weathering from this perspective are few compared to the fairly large number of reports on brown zone. In this study, the characteristics of the brownish zone formed by weathering around a fracture in Hiroshima granite was investigated, and the source of iron oxides in the brown zone was examined using simple numerical calculations of reactive transport.

Materials and Methods

Sample

Granite and its weathered materials are widely distributed around the Chugoku District, including Hiroshima. Weathered granite is fragile, and can cause landslides. A landslide occurred at Mt. Gagara, on the Higashi-Hiroshima campus of Hiroshima University, in 2018, and borehole cores were drilled to investigate the mechanism and related phenomena, including rock degradation (Figs. 1a, b). This study was conducted as part of an effort to understand the initial weathering process of granite. The sample studied was from a 20 m-deep borehole core of granite. The granite around Mt. Gagara is composed mainly of plagioclase (37%), quartz (36%), and potassium feldspar (21%), as well as biotite, ilmenite, zircon, and other minerals (Takagi and Mizuno, 1999). According to Mizuno and Minaki (1986), the bedrock is exposed at an excavation site (elevation of ~300 m), and no fluvial or lake deposits are recognized above this elevation. Below an elevation of approximately 230 m, fluvial sediments of approximately 100–20 ka and fluvial or lake sediments of approximately 700–500 ka were identified. Therefore, it is likely that the studied rock was continuously weathered by surface water after approximately 700 ka, although it may have been temporarily covered by rivers or lakes. Observations of the cores display that unweathered whitish granite predominates at depths below approximately 11 m from the ground surface, with slightly yellowish areas around the fractures in places. From a depth of approximately 5 to 11 m, the closer to the surface, the greater the overall degree of weathering and the greater the percentage of brownish zones around the fractures. From approximately 2 to 5 m depth, the granite was considerably degraded and was almost entirely yellow to brown in color. Colored zones around a fracture located at depths of 8.00–8.15 m (Fig. 1b) were focused on, where the degree of weathering is relatively weak. The areas closer to the fractures were brownish and the areas distant from the fractures were whitish. The whitish areas are at least apparently unaffected by weathering. Similar brown zones were observed around the fractures at several other depths in the borehole core (Figs. 1c, d).

Fig. 1.

(a) Map of borehole core excavation site. (b) Granite core sample analyzed (8.00–8.15 m from ground surface). (c) and (d) Other examples of similar brown zones along fractures (5.89–6.00 m; 9.49–9.62 m from ground surface).

Elemental mapping

To examine the elemental distributions around the brown zone near the fracture, a rock piece (about 2.7 cm × 3.6 ± 0.5 cm × 0.7 cm) was cut from the borehole core (Fig. 1b). A photograph of a portion of the rock piece is displayed in Fig. 2a. The area 3–13 mm from the fracture is particularly brown, and the slightly yellow area extends to approximately 18 mm. The rock piece was attached to a glass slide, cut, polished to 1 μm diamond paste, and applied to elemental mapping. Two areas were analyzed; one was 35 mm long by 4 mm wide extending perpendicular to the fracture (Fig. 2a), and another was a magnification of that area (375 μm long and 500 μm wide; Fig. 2d). Electron probe microanalyzer (EPMA) (JXA-iSP100, JEOL) was used for the elemental mapping under the conditions of acceleration voltage 15 kV, current value 200 nA, measurement speed 0.01 sec/px, and 10 μm × 10 μm/px for an area shown in Fig. 2a (3500 × 400 points) and 0.5 μm × 0.5 μm/px for an area shown in Fig. 2d (750 × 1000 points).

Fig. 2.

(a) Photograph of the rock sample analyzed by EPMA, elemental map for K, Ca, Na, Si, Mg, and Fe. The image was captured before attaching and polishing the sample on a glass slide and the left and right sides of the photo were reversed; therefore it slightly differs from the analyzed surface, though nearly identical. The positions of the minerals remain consistent. Qz: quartz, K-fs: potassium feldspar, Pl: plagioclase, Bt: biotite. (b) Average concentration profile of Fe in the lateral direction. Left profile includes Fe contained in biotite and right profile excludes Fe contained in biotite. (c) BEI and Fe distribution map of an enlarged part of yellow dashed frame (in Fe map and left photo) of panel (a). (d) BEI and Fe distribution map of the region of yellow dashed frame in panel (c). The figure right is the enlarged BEI of the red dashed frame and the results of EDS analysis at the arrowed point.

X-ray powder diffraction (XRD) analysis

For powder XRD analysis, the brown zone (distance from the fracture <15 mm) and whitish area (distance from the fracture >18 mm) of the rock fragments were cut and their powders were prepared. Powder XRD patterns were obtained using an X-ray diffractometer (Ultima IV, Rigaku). The analysis conditions were as follows: a Cu target as the X-ray source, a scan speed of 1°/min, an acceleration voltage of 40 kV, and a tube current of 20 mA.

Color measurements

The color measurement of rocks is an effective method for characterizing iron oxides in rocks (Nagano and Nakashima, 1989; Yokoyama and Nakashima, 2005). To obtain information on the type of iron oxides formed during weathering, the colors of the granite were compared with those of four iron oxides standards, goethite (α-FeOOH), ferrihydrite (Fe₅HO₈·4H₂O), lepidocrocite (γ-FeOOH), and hematite (α-Fe₂O₃) (composition from Schwertmann and Cornell, 2000). The standards were the same as those used by Yokoyama and Nakashima (2005) and were prepared by mixing the powders of each iron oxide with white amorphous SiO₂ powder in various proportions. The particle size dependence of the color of the iron oxide standard was unknown, and the effect of particle size was not considered. The rock sample used for the color measurements was taken from the same rock piece used for the EPMA analysis.

A spectrocolorimeter (CM-25d, Konica Minolta) was used to determine the color of the sample. The color was expressed using the L*a*b* color space defined by the Commission Internationale de l’Eclairage (CIE) in 1976. L* value designates brightness (black at L* = 0 and white at L* = 100), a* and b* values represent chromaticity (reddish for +a*, greenish for –a*, yellowish for +b*, and bluish for –b*) (Nagano and Nakashima, 1989). The measurement area was circular with a diameter of 8 mm, and the average L*, a*, and b* values of the area were obtained. The measurement conditions were diffuse illumination with an 8° light reception method (d/8) and the specular component exclude (SCE) mode. Each iron oxide standard was placed in a clear glass bottle and the color was measured from the bottom of the bottle. For the color measurement of the rock sample, the L*, a*, and b* values were measured every 5 mm from the brown zone to the whitish area by placing the window of the spectrophotometer on the cut surface of the sample.

Selective iron dissolution

Selective iron dissolution using citric acid and dithionite is widely used for dissolving and quantifying iron oxides. This method is estimated to dissolve more than 80% of ferrihydrite, goethite, hematite, and lepidocrocite but not halloysite or gibbsite (Parfitt and Childs, 1988 and references therein). Powdered samples were prepared for each of the brown and whitish areas from the same samples as used for XRD analysis. 5.0 g of dithionite was added to 250 mL of a solution of sodium citrate (220 g L^–1). 0.30 g of the powdered sample was added to 15 mL of this solution and shaken at room temperature for 16 h to elute iron. After dilution with pure water and centrifugation to remove the powder from the solution, the concentration was adjusted with pure water and acidified with HNO₃ (0.3 mol L^–1), and the concentration of dissolved iron was determined by an atomic absorption spectrophotometer (AA-6300, Shimadzu).

Determinations of density, porosity, and diffusivity

To determine the density and porosity of the rock, rock pieces were cut from the same location as EPMA analysis of both the brownish zone (0–18 mm from the fracture) and whitish area (>18 mm from the fracture). These samples were dried at 110°C for 24 h and the dry mass M_dry was determined (mass of brownish zone: 6.232 g; mass of whitish area: 6.414 g). The pores of the samples were saturated with water under vacuum as described by Yokoyama (2013). Subsequently, the water-saturated mass M_wet was measured (mass of brownish zone: 6.254 ± 0.001 g, mass of whitish area: 6.442 ± 0.001 g), and the volume of water-saturated sample V_sat was determined using the Archimedes method at 20°C (brownish zone: 2.37 cm², whitish area: 2.46 cm²). The value of rock density d_rock (g cm^–3) was obtained from “d_rock = M_dry/V_sat”. The porosity of pores open to the outside ϕ_open (dimensionless) was obtained from “ϕ_open = (M_wet – M_dry)/water density (0.998 g cm^–3)/V_sat”. Nishiyama et al. (1990) compiled effective diffusion coefficients D_e measured for various granites and sedimentary rocks, and proposed that D_e (cm² s^–1) and ϕ_open could be empirically related in the following form

  

$$\begin{array}{c}\hfill D_e=6.4\times {10}^{–6}{\varphi }_{\mathrm{open}}^{1.3}.\end{array}$$(1)

Nishiyama et al. (1990) used total porosity (ε_tot) instead of ϕ_open, however since their ε_tot was obtained by the water saturation method and only the pore open to the outside of the rock was measured, it is labeled ϕ_open in the present study. The values obtained using Eq. (1) was for 25°C. Given that the temperature of the spring water near the excavation site was 13.0–14.9°C as described later, the actual temperature of weathering may be lower than 25°C and is assumed to be 15°C in this study. Based on the activation energy for diffusion of ions in water reported by Nakashima (1995), 15 kJ mol^–1, the diffusion coefficient at 15°C is 0.81 times the value at 25°C, and this value was multiplied by the D_e value to correct for temperature. In addition, the D_e value depends on the solute species and the values obtained using Eq. (1) corresponds to the case of the diffusion of I^–. Considering that the diffusion coefficient of I^– at 18°C (1.72 × 10^–5 cm² s^–1) is greater than that of Fe²⁺ (5.82 × 10^–6 cm² s^–1) (Li and Gregory, 1974), to derive the diffusion coefficient of Fe, we used Eq. (1) by replacing 6.4 × 10^–6 by 2.2 × 10^–6. The resulting D_e value (cm² s^–1) at 15°C was

  

$$\begin{array}{c}\hfill D_e=1.8\times {10}^{–6}{\varphi }_{\mathrm{open}}^{1.3}.\end{array}$$(2)

The D_e value obtained from Eq. (2) was used in the reactive transport calculations.

Results

Elemental distribution profile

The elemental mapping results using EPMA are displayed in Figs. 2a–d. The distribution of elements near the fracture, brown zone, and whitish area can be seen (approximately 1 mm near the fracture was lost during sample preparation). In the elemental map shown in Fig. 2a, plagioclase corresponds to high Ca (blue to green) and Na (blue to yellow), quartz to high Si (red), potassium feldspar to high K (yellow to orange), and biotite to high Fe (red) and Mg (green to red). The Fe concentrations were particularly high in two plagioclase grains near the fractures. For these two grains, there is clear zoning from the Ca-rich core to the Na-rich rim. In the biotite, particles with localized areas of low Mg concentration (green) were found in both the brown zone and whitish area. It has been noted that Mg depletion and formation of vermiculite layers occur in the alteration of biotite at pH 5.5–7 (Acker and Bricker, 1992), and areas showing a decrease in Mg concentration are likely to be somewhat altered.

The profile on the left of Fig. 2b displays the average horizontal Fe signal intensity (Fe Lv. avg.), which includes all iron in the biotite and iron oxides. The location of the biotite in Fig. 2a clearly displays a large increase in Fe concentration (Fig. 2b, left). The brown zone from the fracture to approximately 13 mm may have a slightly higher total Fe concentration than the areas farther from the fracture; however, this is not clear as the signal from the biotite was too strong. To obtain information on iron oxides only, the right-hand profile in Fig. 2b displays the average Fe concentration after excluding Fe at the biotite location. In this calculation, the points on the Fe map with Mg Lv. greater than 15 were excluded as biotite, and points with Si Lv. less than 220 were excluded as empty areas (or resin). In addition, the average horizontal Fe signal intensity was calculated only if more than 10 points remained in the horizontal direction. The rationale for the calculation using Si was to correctly calculate the average Fe concentration, especially by excluding empty areas at both ends of the analysis area (approximately 0–2 mm and 34–36 mm from the fracture, appearing black on the Si map). As described below, since iron oxides exist together with minerals containing Si and the Si signal is not weak, it is unlikely that iron oxide would be excluded by excluding areas where the Si signal is weak. A relatively whitish region was observed near the fracture, where the Fe concentrations are similar to those in the whitish area (Fe Lv. = ~3–4). The Fe concentration increased from the fracture to a distance of approximately 4 mm, reaching a maximum (Fe Lv. = ~12–16) at distances of 4–10 mm, and gradually decreased at distances of 10–21 mm. At distances greater than 21 mm, Fe concentrations were nearly stable (Fe Lv. = ~3–4). If the two plagioclases near the fracture were excluded, the change in the Fe concentration in the brown zone would be much smaller, however the overall upward and downward trends would remain the same. A similar increase and decrease in Fe concentration with increasing distance from the fracture has been reported for the granite in Gifu (Akagawa et al., 2004) and Okayama (Yoshida et al., 2008), while the thickness of the brown zone (approximately 15 cm for Gifu; tens of centimeters for Okayama) differed from the sample in the present study. For the Gifu sample, the total Fe concentration 10 mm from the fracture was approximately 22% higher than that in the immediate vicinity of the fracture (Akagawa et al., 2004). In addition, the total Fe concentration was higher in the brown zone than in the whitish area far from the fracture in the Gifu sample, and the maximum total Fe concentration was observed near the front of the brown zone.

Figure 2c displays a backscattered electron image (BEI) and a magnified view of the Fe map in Fig. 2a (yellow dashed frame), whereas Fig. 2d shows a magnified BEI and Fe map of the yellow dashed frame in Fig. 2c. The width of grain boundaries ranges from <1 μm to approximately 20 μm. Outside the plagioclase, Fe was distributed to almost all grain boundaries within the area shown in Fig. 2c, and the Fe concentration was high in the brown zone (distance from fracture: approximately 7–13 mm) and low in the light-yellow zone (distance from fracture: approximately 13–18 mm). Continuous streaks of high Fe concentrations were observed, especially at the grain boundaries around the biotite (yellow arrows). In plagioclase, Fe is particularly abundant at the contours and interior of Ca-rich regions and is also distributed along two dominant directions that seem to correspond to the cleavage and/or twinning planes. Yoshida et al. (2008) also reported Fe enrichment within plagioclase during the weathering of granite in Okayama, Japan. They interpreted that Ca leached from the plagioclase and smectite was formed, resulting in groundwater infiltration and precipitation of iron hydroxide inside the plagioclase, or Fe was incorporated into the smectite crystal structure. Analysis of the weathering products using scanning electron microscopy (SEM) (TM3030, Hitachi) with energy dispersive spectroscopy (EDS) revealed that Fe was detected along with Al and Si (Fig. 2d, upper spectrum). In addition to the possibility that fine-grained iron oxides are present along with aluminosilicate clays, clay minerals may contain iron, although this cannot be determined from the SEM-EDS analysis.

XRD analysis results

The powder XRD analysis results are displayed in Fig. 3. The peak positions of each mineral were assigned based on the ICDD database (Gates-Rector and Blanton, 2019). Quartz, plagioclase, potassium feldspar, and biotite, which are the primary minerals, were detected in both the brown zone and whitish area. The peak of goethite was expected to occur on either of the arrow annotated “Gt”, and only the brown zone had a weak peak near the right arrow. Therefore, although it is uncertain, goethite may have been present in the brown zone. The peak positions of illite and biotite often overlapped and were confusing. However, clay minerals containing K were detected by SEM (Fig. 2d), and it is likely that both illite and biotite were present. Kaolinite may also have been present, although it remains unclear whether the small signal near the arrow annotated “Ka” corresponds to kaolinite. Smectite (montmorillonite) and vermiculite were not present or not in sufficient quantities to be detected.

Fig. 3.

Results of powder XRD analysis of brown zone and whitish area. Qz: quartz, Ab: albite (plagioclase), Or: orthoclase (potassium feldspar), Bt: biotite, It: illite, Mm: montmorillonite, Vm: vermiculite, Ka: kaolinite, Gt: goethite. Dotted arrows are peak locations of minerals for which no clear signal was identified.

Diffusion coefficient

The obtained values of d_rock and ϕ_open were 2.63 g cm^–3 and 0.0091 for the brown zone and 2.61 g cm^–3 and 0.011 for the whitish area, respectively. The slightly lower porosity in the brown zone than in the whitish area may be due to the formation of iron oxide or clay minerals in the pores. In the following discussion, 0.01 is used as the average value of ϕ_open. By substituting ϕ_open = 0.01 into Eq. (2), the value of D_e for the sample in this study at 15°C was calculated to be 4.5 × 10^–9 cm² s^–1. There may be several dissolved species of Fe, each of which can have a different D_e; however, only the total Fe was considered as the first approximation. In the modeling described below, the apparent diffusion coefficient D_ap was used, and its relationship with D_e is as follows (c.f., Skagius and Neretnieks, 1986):

  

$$\begin{array}{c}\hfill D_e={\varphi }_{\mathrm{tra}}{D}_{\mathrm{ap}},\end{array}$$(3)

where ϕ_tra is the transport porosity and constrictivity is neglected. The transport pore is the pore that is open to the outside, however the storage pore (or dead end pore) must be excluded, therefore ϕ_tra ≤ ϕ_open. For granite, pores are reported to be continuously distributed throughout the rock and that pores exist not only at grain boundaries, but also inside mineral grains, such as feldspar (Suzuki et al., 1989). In fact, in Fig. 2c, iron oxide precipitation is observed at most grain boundaries and inside the feldspar. Therefore, it seems reasonable to assume that ϕ_tra = ϕ_open (=0.01), and from D_e and Eq. (3), D_ap = 4.5 × 10^–7 cm² s^–1 was obtained. This D_ap was for Fe diffusion as dissolved ions; however, Fe may also be present as colloids. Measurements of the diffusion coefficients of gold colloids in granite have shown that the diffusion coefficients of colloids are approximately five orders of magnitude smaller than those of weak or non-sorbing solutes, and the larger the colloidal size, the smaller the diffusion coefficient (Alonso et al., 2007). Therefore, if Fe is transported as a colloid, the diffusion coefficient is expected to be significantly lower.

Color and iron concentration of brown zone and whitish area

Figure 4a shows a*–b* diagram obtained from the color measurements of the granite piece and four iron oxide standards. The numbers next to the data points of the standards indicate the concentration (wt%) of each iron oxide in SiO₂. At the same concentration, goethite was the most yellowish of the four standards. The whitish position D had very small a* and b* values. The a* b* values for positions A and B were intermediate between those of ferrihydrite and goethite. The diffuse reflectance spectra at positions A and B (Fig. 4b) were relatively close to those of goethite, which showed a characteristic absorption band at approximately 480 nm (Nagano and Nakashima, 1989). Position B appeared to contain a large amount of goethite, although it is difficult to rule out the possible presence of hematite and lepidocrocite as there were no absorption peaks in the visible spectra of these minerals, which were clearly different from those of ferrihydrite and goethite. Compared with position B, the a* b* values for position A deviated significantly from those of goethite and were closer to those of ferrihydrite. Ferrihydrite is a low-crystallinity material that is known to transform into hematite and goethite over time in aqueous solutions (Schwertmann and Murad, 1983; Das et al., 2011). At position A, the conditions are probably suitable for the formation of ferrihydrite, and the change from ferrihydrite to goethite may also be underway. As iron is found at most grain boundaries (Fig. 2c), the contribution of iron oxides to browning is probably significant; however, the oxidation of biotite may also partially contribute to browning.

Fig. 4.

(a) A photograph of the location of color measurements (circles are measurement areas) and a*–b* diagram of the color of rock (unpolished surface) and iron oxide standards. Gray diamonds are the color of position A to D. (b) Visible diffuse reflectance spectra of position A and B shown in the panel (a) and standard iron oxides.

The Fe concentration in the brown zone and whitish area obtained from the selective iron dissolution method was 0.367 (+0.012, –0.017) wt% and 0.044 ± 0.003 wt%, respectively. The source of the slight Fe leaching in the whitish area is unknown (possibly biotite), however it could be considered a blank value in the rock matrix. Therefore, the difference between the values for the brown zone and whitish area (0.32 ± 0.02 wt%) is considered as the total amount of Fe contained in the iron oxides produced by weathering in the brown zone.

Discussion

Consideration of iron source: Influx from fracture

As a mechanism of iron oxide formation near granite fractures, Akagawa et al. (2004) proposed that initially, aqueous Fe(II) diffuses from the fracture to the rock matrix under reductive conditions. Then, as the solution in the fracture changes, oxidizing conditions are generated leading to precipitation of aqueous Fe(II) in the form of Fe(III) in the rock matrix. This results in a lower concentration of Fe(II) around the fracture, however, areas with high Fe(II) concentrations remain deep inside the matrix. Subsequently, Fe(II) diffuses from the deeper regions of the matrix toward the fracture, resulting in oxidation, precipitation, and Fe enrichment at a certain distance from the fracture. The change in the redox state of the solution in the fracture is repeated several times to allow the Fe to migrate to the interior of the matrix. This mechanism can explain the fact that the brown zone of the granite has a higher total Fe concentration than the host rock and that the Fe concentration increases near the front of the brown zone.

However, assuming an influx of Fe from the fractures, as in the above mechanism, how much Fe entered the matrix from the groundwater in the fracture of the sample in the present study? Consider an area of 1.5 cm × 1 cm × 1 cm (1.5 cm³) (red box in Fig. 5a). The case of various values of brown zone thickness α is discussed later, and Fig. 5a corresponds to the case of α = 1.5 cm. The first important factor to consider when estimating the influx of Fe is the Fe concentration in the fracture. As a reference value, spring water was sampled (twice) approximately 16 m downstream from the excavation point on Mt. Gagara, and Fe concentrations totaling 0.05–0.06 mg L^–1 (13.0–14.9°C; pH 6.6–6.7) were obtained. In addition, Iwatsuki et al. (2005) reported a maximum total Fe concentration of 0.086 mg L^–1 (most data are ≤0.02 mg L^–1) in groundwater of the granite area in Gifu, Japan (depth (mbgl): 96.9–995.3 m; 17.5–25.0°C; pH 8.1–9.2; Eh –280–0 mV (pe –4.8–0); only data with granite geology were selected). Although the exact composition of the solution in the fracture at Mt. Gagara is unknown, it is suspected to have been an intermediate composition between the spring water data (composition of oxidative solutions near the ground surface) and the data of Iwatsuki et al. (2005) (composition of reductive solutions in the subsurface). Thus, it seems reasonable to assume that the Fe concentration in the fracture is approximately 0.1 mg L^–1 (= 1.8 × 10^–9 mol cm^–3). Next, we will discuss the speed at which Fe migrated into the rock when groundwater in the fracture is considered the main Fe source. The two model cases are discussed: one in which Fe diffusion and precipitation occur alternately, and the other in which precipitation occurs simultaneously with Fe diffusion. The first model assumed that Fe diffusion and precipitation occurred alternately, as in the aforementioned mechanism proposed by Akagawa et al. (2004). Diffusion of Fe without precipitation can occur, for example, if the solution in the fracture and rock matrix is reductive and Fe(II) is the dominant species with no precipitation of Fe(III). The time variation of the Fe concentration in the pore water can be calculated using Fick’s second law:

Fig. 5.

(a) Schematic overview of the region where iron concentration profile and accumulated influx were considered. (b) Binarized image of Mg distribution map used for estimating reactive surface area of biotite. (c)–(f) Results of reactive transport analysis. (c) Time variation of the concentration profile of iron in the red box region (relative to the concentration in the fracture) for the case of α = 15 mm assuming no iron precipitation occurs. (d) Elapsed time vs. the ratio of average Fe concentration within the brown zone to Fe concentration in the fracture. (e) Time required for the Fe concentration to reach 99% of the fracture concentration (0.1 mg L^–1) by diffusion from the fracture, T_sup vs. the thickness of brown zone, α. The red dashed lines indicate the time required to dissolve biotite to reach an Fe concentration of 0.1 mg L^–1 for SI values <–2, –0.1, and –0.01, respectively. (f) Quasi-steady state profiles of Fe concentration relative to that at the fracture and the relative amount of goethite precipitation under conditions of simultaneous Fe diffusion and precipitation.

  

$$\begin{array}{c}\hfill \frac{\partial c}{\partial t}=D_{\mathrm{ap}}\frac{{\partial }^{2}c}{\partial x^2},\end{array}$$(4)

where c is the Fe concentration and x is the distance from the fracture. For all the calculations below, the PHREEQC and llnl.dat databases were used to simulate 1D reactive transport. With weakly basic and reductive groundwater in the fracture in mind, initial Fe concentration of zero, pH = 8, and pe = –1 at x = 0–15 mm were assumed, and Fe concentration in the fracture c_fra of 0.1 mg L^–1, pH = 8, and pe = –1 at x = 0, a closed boundary at x = 15 mm, and D_ap = 4.5 × 10^–7 cm² s^–1 for all dissolved species, were assumed. Fe(II) was predominant under these conditions. The time variation of Fe concentration profile is shown in Fig. 5c, and the average concentration c_avg obtained for each time is plotted against time in Fig. 5d (case α = 15 mm). The time required for c_avg to become approximately equal to c_fra (c_avg = 0.99c_fra), T_sup, is 103 d. If Fe moves as a colloid, the diffusion coefficient would be smaller, resulting in a longer T_sup. Although the influx by diffusion eventually stops when c_avg reaches c_fra (Fig. 5c), if an oxidative solution flows through the fracture and the inside of the matrix also becomes oxidative, Fe precipitates and the dissolved Fe concentration decreases in the matrix. Diffusion then resumed when the reducing, high-Fe-concentration solution flowed through the fracture. Thus far the case where the brown zone thickness α is 15 mm has been considered, however in reality α did not remain the same throughout the weathering period; it likely started as zero and increased with time. The same calculations as above were performed for various α from 0.5 mm to 15 mm, and the results are shown in Fig. 5e (black filled circles). The smaller α is, the shorter T_sup is. The T_sup is used later in the discussion. Let us now consider the amount of Fe that migrates via diffusion. The amount of Fe that can flow in during one cycle from the start of diffusion until the end is maximized when α is 15 mm, which is 2.7 × 10^–11 mol (= 1.8 × 10^–9 mol cm^–3 × 1.5 cm³ × 0.01 (porosity)). In contrast, from the results of the selective iron dissolution, the concentration of Fe present as iron oxide in the brown zone was 0.32 wt%. This means that the amount of Fe present in the 1.5 cm³ area is 2.3 × 10^–4 mol (= 2.63 g cm^–3 (rock density) × 1.5 cm³ × 0.0032/55.8 g mol^–1 (Fe atomic mass)). Notably, this value is 8.5 × 10⁶ times the amount of Fe that can flow through one diffusion cycle. This implies that to accumulate the required amount of Fe, a series of processes (influx by diffusion, precipitation, and changes in the redox state) must be repeated numerous times. Therefore, the model is unlikely to be realistic.

Next, we considered the case in which precipitation occurred simultaneously with the influx of Fe via diffusion. The precipitate assumed to be of goethite only. There is no need to consider the changes in the redox state, as in the previous case. In this case, the reaction term was added to Eq. (4), which takes the forms:

  

$$\begin{array}{c}\hfill \frac{\partial c}{\partial t}=D_{\mathrm{ap}}\frac{{\partial }^{2}c}{\partial x^2}+S_{\mathrm{BET}}{M}_{m}m{k}_T\left(1–{10}^{SI}\right),\end{array}$$(5)

  

$$\begin{array}{c}\hfill k_T=k_{25}{e}^{–\frac{E}{R}\left(\frac{1}{T}–\frac{1}{298.15}\right)},\end{array}$$(6)

where S_BET is the BET specific surface area of goethite, M_m is the molar mass, m is the mole of goethite in contact with the solution (mol kg_water^–1), k_T (mol m^–2 s^–1) is the reaction rate constant at temperature T (T = 288.15 K), k₂₅ is the rate constant at 25°C, SI is the saturation index (precipitation occurs for SI >0 and dissolution occurs for SI <0), E (= 86.5 kJ mol^–1) is the activation energy, and R is the gas constant (= 8.314 kJ mol^–1) (c.f., Palandri and Kharaka, 2004; Parkhurst and Appelo, 2013). PHREEQC and the same input format for kaolinite presented by Marty et al. (2015) was used with modifications for goethite using the following values: S_BET = 37 m² g^–1 (Pedersen et al., 2005), M_m = 88.8 g mol^–1, and k₂₅ = 10^–7.94 mol m^–2 s^–1 (Palandri and Kharaka, 2004). As a first approximation, a constant Fe concentration of 0.1 mg L^–1 at x = 0, and constant pH of 8 and the solution in equilibrium with atmospheric O₂ pressure for x = 0–15 mm were assumed. Under these conditions, Fe existed mostly as Fe(III). Initial amount of goethite = 0 and Fe concentration = 0 mg L^–1 for x = 0–15 mm. The charge was adjusted with Na⁺ or Cl^–. The calculated quasi-steady-state profiles of the Fe concentration and amount of goethite precipitated are displayed in Fig. 5f. The concentration gradient and precipitation occurred only in the immediate vicinity of the fracture (x = ~0 mm). The result was almost unchanged, for example, when setting pH = 7 at 0 ≤ x ≤ 2 mm and pH = 9 at 2 < x ≤ 15 mm, or setting a lower oxygen concentration (pO₂ = –40 at 0 < x ≤ 15 mm). This means that owing to the rapid precipitation of Fe, the Fe that flows in from the solution in the fracture tends to precipitate in the immediate vicinity of the fracture. However, in reality, the concentration of iron oxides in the vicinity of the fracture (distance <2 mm in Fig. 2b) was lower than that at greater distances from the fracture. Compared to the case where Fe diffusion and precipitation occur alternately, the Fe inflow rate is greater when Fe precipitation and diffusion occur simultaneously as a large concentration gradient is maintained. However, since Fe precipitation is limited to the vicinity of the fracture, it appears that Fe sources other than groundwater in the fracture need to be considered.

Consideration of Fe source: Dissolution of nearby biotite

A possible Fe source other than the fracture is biotite around the brown zone. Here the amount of time needed to dissolve biotite to bring the porewater Fe concentration to the same as the groundwater in the fracture (0.1 mg L^–1) is considered. According to Acker and Bricker (1992), the dissolution rates of biotite (grain size 149–420 μm) at pH 5–6.7 and 22°C are 5.8 × 10^–17–1.3 × 10^–16 mol_bio cm^–2 s^–1 (1 mol of biotite contained 2.24 mol of Fe). Assuming that an activation energy of biotite dissolution is the same as that for muscovite (22 kJ mol^–1; Nagy, 1995), the dissolution rate at 15°C is approximately 0.8 times the value at 22°C. Based on these values, 2 × 10^–16 mol_Fe cm^–2 s^–1 was used as the dissolution rate of Fe from biotite at 15°C. To estimate the reactive surface area, the distribution map of Mg (Fig. 2a) was used and binarized to extract biotite (Fig. 5b; small amount of Mg-bearing minerals other than biotite were also counted as biotite), and measured the perimeter with ImageJ software (Schneider et al., 2012). The result yielded a total perimeter of 9.3 cm for a 3.5 cm × 0.4 cm area. Converting this perimeter value to a 1.5 × 1 cm area yields 10 cm. Assuming that the total perimeter is the same as in any section parallel to the paper surface in Fig. 5b, 10 cm² was obtained as the reactive surface area of biotite in the 1.5 cm³ area (red box in Fig. 5a). Using this value, the time required to dissolve the biotite and bring the iron concentration in the pore water to the same as the groundwater in the fracture (0.1 mg L^–1) by dissolving biotite was calculated to be 0.16 d (= 1.8 × 10^–9 mol cm^–3 × 1.5 cm³ × 0.01 (porosity)/(2 × 10^–16 mol cm^–2 s^–1 × 10 cm²)/(60 × 60 × 24 s)). The surface area was calculated only from the perimeter, ignoring the areas at the top and bottom of the grains. If these factors were considered, the surface area would be somewhat larger, which could result in a smaller time estimate. In addition, a potentially significant effect is the SI dependence of the dissolution rate. In fact, there are many unknowns in the solution composition inside the rock matrix, making it difficult to estimate SI values. However here, an attempt to derive a rough estimate of SI was undertaken. According to solution data obtained in situ at underground excavation sites with granite geology (Iwatsuki et al., 2005), most solutions have a pH of 8–9 and an SI of approximately zero for illite. Based on this, the SI for annite (KFe₃AlSi₃O₁₀(OH)₂) and phlogophite (KMg₃AlSi₃O₁₀(OH)₂) was estimated, the two end components of biotite, using PHREEQC under the conditions of SI = 0 for illite, total Fe concentration = 0.1 mg L^–1, pH 8.5, and 15°C, and the charge balance achieved by Na⁺ or Cl^–. The SI of phlogopite was not affected by pe value and was –9.4 under the above conditions. In contrast, SI for annite decreased with increasing pe; SI was below –1.2 (down to –39) at pe = 0–13, and SI was from –1.2 to 0.0 at pe from 0 to –1. The value of pe is the most unknown factor; however, considering the possible presence of ferrihydrite in the brown zone, it is likely that pe is greater than 0.5 (SI for Fe(OH)₃ is greater than zero under the above conditions). In that case, the SI of biotite in the brown zone was estimated to be less than –1.2, and there would be little decrease in the dissolution rate, assuming that the dissolution rate of biotite is proportional to (1–10^SI) as in Palandri and Kharaka (2004). Figure 5e displays the time required to increase the Fe concentration to 0.1 mg L^–1 by dissolution of biotite (red dashed lines), calculated for SI of <–2, –0.1, and –0.01. It is assumed that biotite, which can dissolve, is present everywhere in the brown zone because the amount of Fe, including biotite, is greater than the amount of Fe excluding biotite, even in the brown zone (Fig. 2b). Therefore, the red lines are independent of the brown zone thickness. The Fe supply by dissolution was faster than diffusion from the fracture when the thickness of the brown zone is greater than 4–5 mm, even when the solution was nearly in equilibrium with biotite, as at SI = –0.01. Therefore, except when the thickness of the brown zone was a few millimeters or less, the supply of Fe by dissolution was likely greater than that through fracture. Indeed, the brown zone (Fig. 2c) shows many streaks of high Fe concentrations (yellow arrows), presumably because the Fe dissolved from the biotite and precipitated nearby.

Overview of iron supply and transport around brown zone

From the above discussion, it seems reasonable to consider that the amount of Fe supplied by diffusion through the fracture is small and that dissolution from inside the matrix is the main source. Figure 6 displays a conceptual diagram of the reactions and the amount of Fe in the solid created by considering the observations of this study and the results of the numerical simulations of simplified granite weathering by Lebedeva et al. (2007). Not all simulation results are applicable to the sample in the present study as the differences in mineral assemblage (albite, FeO, goethite, kaolinite, and quartz in the simulations), mineral fractions, and porosity; however, the following simulation results may be applicable: in the vicinity of the fracture, the amount of precipitated goethite is small and the pH increases sharply (from 5 to 9 in the simulation); in the brown zone, dissolved Fe is present predominantly as Fe(III) and its concentration is generally extremely low, and the Fe(II) concentration increases sharply near the front of the brown zone. Since biotite with a reduced Mg concentration that seems to have been altered is present not only in the brown zone but also in the whitish area (Fig. 2a), the dissolution of Fe occurs not only in the brown zone but also in the whitish area to some distance from the front of the brown zone, at least locally. Thus, the Fe dissolved in the whitish area likely migrated to the brown zone, resulting in an increase in the amount of total Fe in the brown zone compared to the whitish area. In addition, a large amount of Fe precipitation has often been found near the front of the brown zone (Akagawa et al., 2004), which may be related to the intensive precipitation of Fe at the front (Lebedeva et al., 2007) and the effects of nucleation and the Ostwald theory (Ortoleva et al., 1986). In the present analysis, however, the increase in Fe in the front part was not noticeable (Fig. 2b), partly as the large amount of Fe in the plagioclase interior. Therefore, we did not discuss the phenomena occurring in front of the brown zone in detail.

Fig. 6.

Conceptual diagram of the reaction involving Fe and the amount of Fe in solids (iron oxides + biotite).

Because browning along fractures, as observed in the present granite sample, is universally observed, it is likely that a similar interpretation can be applied in many cases. However, this phenomenon is probably influenced by differences in the rock type and fracture solution composition. Evaluating of the extent to which these factors affect this phenomenon will be a future task. Nevertheless, it is important to recognize that fractures are not necessarily the main Fe source of the iron oxides in the brown zone.

Conclusion

Characterization of the brown zones around a fracture formed by granite weathering displayed that iron oxides, possibly a mixture of ferrihydrite and goethite, were distributed throughout the mineral grain boundaries and within the plagioclase. The numerical calculations of the supply of Fe by diffusion from groundwater in the fracture and the supply of Fe from the dissolution of biotite suggest that a large proportion of Fe in the brown zone was supplied by the biotite present near the brown zone. When considering phenomena that occur around rock fractures, not just iron, it may not be uncommon to consider the supply of elements from the fractures as the primary source; however, the results of this study indicate that caution is required when dissolution from the primary minerals is not negligible.

Acknowledgments

We thank three anonymous reviewers for their constructive comments. This work was partially supported by JSPS KAKENHI (Grant Number JP22H01312) awarded to TY. Elemental mapping was performed at the Natural Science Center for Basic Research and Development (N-BARD), Hiroshima University.

References

•  Acker,  J. G. and  Bricker,  O. P. (1992) The influence of pH on biotite dissolution and alteration kinetics at low temperature. Geochim. Cosmochim. Acta 56, 3073–3092. https://doi.org/10.1016/0016-7037(92)90290-Y
•  Akagawa,  F.,  Yoshida,  H.,  Yogo,  S. and  Yamamoto,  K. (2004) Redox reaction along a ground water conducting fracture and secondary elemental migration in granitic rock: —Analogue of long-term contaminant migration and fixation in geological environment—. J. Geol. Soc. Japan 110, 671–685 (in Japanese with English abstract). https://doi.org/10.5575/geosoc.110.671
•  Alonso,  U.,  Missana,  T.,  Patelli,  A.,  Rigato,  V. and  Ravagnan,  J. (2007) Colloid diffusion in crystalline rock: An experimental methodology to measure diffusion coefficients and evaluate colloid size dependence. EPSL 259, 372–383. https://doi.org/10.1016/j.epsl.2007.04.042
•  Das,  S.,  Hendry,  M. J. and  Essilfie-Dughan,  J. (2011) Transformation of two-line ferrihydrite to goethite and hematite as a function of pH and temperature. Environ. Sci. Technol. 45, 268–275. https://doi.org/10.1021/es101903y
•  Gates-Rector,  S. and  Blanton,  T. (2019) The powder diffraction file: A quality materials characterization database. Powder Diffr. 34, 352–360. https://doi.org/10.1017/S0885715619000812
•  Iwatsuki,  T.,  Furue,  R.,  Mie,  H.,  Ioka,  S. and  Mizuno,  T. (2005) Hydrochemical baseline condition of groundwater at the Mizunami underground research laboratory (MIU). Appl. Geochem. 20, 2283–2302. https://doi.org/10.1016/j.apgeochem.2005.09.002
•  Lebedeva,  M. I.,  Fletcher,  R. C.,  Balashov,  V. N. and  Brantley,  S. L. (2007) A reactive diffusion model describing transformation of bedrock to saprolite. Chem. Geol. 244, 624–645. https://doi.org/10.1016/j.chemgeo.2007.07.008
•  Li,  Y.-H. and  Gregory,  S. (1974) Diffusion of ions in sea water and in deep-sea sediments. Geochim. Cosmochim. Acta 38, 703–714. https://doi.org/10.1016/0016-7037(74)90145-8
•  Marty,  N. C. M.,  Claret,  F.,  Lassin,  A.,  Tremosa,  J.,  Blanc,  P.,  Madé,  B.,  Giffaut,  E.,  Cochepin,  B. and  Tournassat,  C. (2015) A database of dissolution and precipitation rates for clay-rocks minerals. Appl. Geochem. 55, 108–118. https://doi.org/10.1016/j.apgeochem.2014.10.012
•  Mizuno,  K. and  Minaki,  M. (1986) Stratigraphy of the Quaternary deposits in southern part of the Saijo basin, Hiroshima Prefecture, Japan. Bull. Geol. Surv. Japan 37, 183–200 (in Japanese with English abstract).
•  Nagano,  T. and  Nakashima,  S. (1989) Study of colors and degrees of weathering of granitic rocks by visible diffuse reflectance spectroscopy. Geochem. J. 23, 75–83. https://doi.org/10.2343/geochemj.23.75
•  Nagy,  K. L. (1995) Dissolution and precipitation kinetics of sheet silicates. Chemical Weathering Rates of Silicate Minerals. Rev. Min. 31, (White, A. F. and Brantley, S. L., eds.), 173–233, Mineralogical Society of America.
•  Nakashima,  S. (1995) Diffusivity of ions in pore water as a quantitative basis for rock deformation rate estimates. Tectonophysics 245, 185–203. https://doi.org/10.1016/0040-1951(94)00234-Z
•  Nishiyama,  K.,  Nakashima,  S.,  Tada,  R. and  Uchida,  T. (1990) Diffusion of an ion in rock pore water and its relation to pore characteristics. Mining Geol. 40, 323–336 (in Japanese with English abstract). https://doi.org/10.11456/shigenchishitsu1951.40.323
•  Ortoleva,  P.,  Auchmuty,  G.,  Chadam,  J.,  Hettmer,  J.,  Merino,  E.,  Moore,  C. H. and  Ripley,  E. (1986) Redox front propagation and banding modalities. Physica 19D, 334–354. https://doi.org/10.1016/0167-2789(86)90063-1
•  Palandri,  J. L. and  Kharaka,  Y. K. (2004) A compilation of rate parameters of water-mineral interaction kinetics for application to geochemical modeling. U. S. Geological survey open file report 2004–1068.
•  Parfitt,  R. L. and  Childs,  C. W. (1988) Estimation of forms of Fe and Al: A review, and analysis of contrasting soils by dissolution and Mossbauer methods. Aust. J. Soil Res. 26, 121–144. https://doi.org/10.1071/SR9880121
•  Parkhurst,  D. L. and  Appelo,  C. A. J. (2013) Description of input and examples for PHREEQC Version 3—A computer program for speciation, batch-reaction, one-dimensional transport, and inverse geochemical calculations. U. S. Geological Survey Techniques and Methods, Book 6, chap. A43, 497. http://pubs.usgs.gov/tm/06/a43
•  Pedersen,  H. D.,  Postma,  D.,  Jakobsen,  R. and  Larsen,  O. (2005) Fast transformation of iron oxyhydroxides by the catalytic action of aqueous Fe(II). Geochim. Cosmochim. Acta 69, 3967–3977. https://doi.org/10.1016/j.gca.2005.03.016
•  Schneider,  C. A.,  Rasband,  W. S.,  Eliceiri,  K. W. (2012) NIH Image to ImageJ: 25 years of image analysis. Nature Methods 9, 671–675. https://doi.org/10.1038/nmeth.2089
•  Schwertmann,  U. and  Cornell,  R. M. (2000) Iron Oxides in the Laboratory. Wiley-VCH, Weinheim.
•  Schwertmann,  U. and  Murad,  E. (1983) Effect of pH on the formation of goethite and hematite from ferrihydrite. Clays Clay Min. 31, 277–284. https://doi.org/10.1346/CCMN.1983.0310405
•  Skagius,  K. and  Neretnieks,  I. (1986) Porosities and diffusivities of some nonsorbing species in crystalline rocks. Water Resour. Res. 22, 389–398. https://doi.org/10.1029/WR022i003p00389
•  Suzuki,  T., Nakashima, S., Nagano, T., Kita, H. (1989) Micropore distributions in granite as material migration pathways. Mining Geol. 39, 349–354 (in Japanese with English abstract). https://doi.org/10.11456/shigenchishitsu1951.39.218_349
•  Takagi,  T. and  Mizuno,  K. (1999) Geology of the Kaitaichi district, with geological sheet map at 1:50,000. Geol. Surv. Japan (in Japanese with English abstract).
•  Yokoyama,  T. (2013) Characterization of the reaction and transport properties of porous rhyolite and its application to the quantitative understanding of the chemical weathering rate. Geochim. Cosmochim. Acta 118, 295–311. https://doi.org/10.1016/j.gca.2013.05.011
•  Yokoyama,  T. and  Nakashima,  S. (2005) Color development of iron oxides during rhyolite weathering over 52,000 years. Chem. Geol. 219, 309–320. https://doi.org/10.1016/j.chemgeo.2005.03.005
•  Yoshida,  H. and  Yamamoto,  K. (2014) Ferric oxyhydroxide in underground geological environments and high-level radioactive waste disposal: Analysis of influence on nuclide migration scenarios. J. Geol. Soc. Japan 120, 327–343 (in Japanese with English abstract). https://doi.org/10.5575/geosoc.2014.0038
•  Yoshida,  H.,  Yamamoto,  K.,  Tanaka,  S.,  Yogo,  S. and  Milodowski,  A. E. (2003) Redox front development and related secondary elemental migration —An analogue of long-term chemical contaminant fixation in geological environment—. J. Geol. Soc. Japan 109, 548–558 (in Japanese with English abstract). https://doi.org/10.5575/geosoc.109.548
•  Yoshida,  H.,  Nishimoto,  S.,  Cho,  A.,  Yamamoto,  K. and  Katsuta,  B. (2008) Alteration of subsurface granitic rock in Okayama area, Japan. J. Japan Soc. Eng. Geol. 49, 256–265 (in Japanese with English abstract). https://doi.org/10.5110/jjseg.49.256
•  Yu,  C.,  Drake,  H.,  Dideriksen,  K.,  Tillberg,  M.,  Song,  S.,  Mørup,  S. and  Åström,  M. E. (2020) A combined X-ray absorption and Mössbauer spectroscopy study on Fe valence and secondary mineralogy in granitoid fracture networks: Implications for geological disposal of spent nuclear fuels. Environ. Sci. Technol. 54, 2832–2842. https://doi.org/10.1021/acs.est.9b07064