Metal Flux from Dissolution of Iron Oxide Grain Coatings in Sandstones

Iron oxide grain coatings in red sandstones contain trace metals that are released upon dissolution of the coatings. Analyses by ICPMS following acid leaching of the grain coatings show that the dissolved metals can constitute an ore-forming fluid, as hypothesized in models for sandstone-hosted ore deposits. Median compositions of 37 samples, mostly of Triassic to Devonian age, from across Britain and Ireland are 6.3 ppm copper, 2.4 ppm cobalt, 10.1 ppm vanadium, and 0.3 ppm uranium. These contents at the basin scale are adequate to form the observed range of ore deposits in red beds. The migration of hydrocarbons or brines can cause the dissolution of grain coatings and contributes to controlling the distribution of ore deposits. Future measurements should test red beds derived from uplifted, mineralized plate margins, in which sandstones may be preloaded with ore metals.


Introduction
The mineralization of sedimentary basins has made an important contribution to the planet's resources of metals [1]. Mineralization in basins requires processes that concentrate metals beyond their normal abundance by several orders of magnitude. Examples include the concentration of redox-sensitive elements by reduction sites, such as black shales, and the release of metals from clays during diagenesis. In continental basins, one of the most widespread of processes is the formation of iron oxide coatings around sand grains during early diagenesis [2]. The iron oxide derives from the alteration of mineralogically immature grains such as ferromagnesian minerals. The coatings are what make many sandstones red, when the iron oxide becomes the mineral haematite. Iron oxides are particularly significant to diagenesis because of their capacity to adsorb a wide variety of trace elements from groundwaters [3]. As diagenesis proceeds, and the volume of secondary iron oxides increases, the proportion of trace elements resident in the oxides also increases. However, the grain coatings are also susceptible to removal by dissolution into acidic pore fluids, including oilfield brines [4]. For example, dissolution is evident from the bleaching of red beds following hydrocarbon migration (e.g., [5]) and from CO 2 -rich brines [6,7], but these fluids are not always evident. The combination of trace metal uptake and dissolution is a potential pathway to produce a mineralizing fluid [8][9][10][11][12]. This possibility has been invoked to explain moderate-to large-scale ore deposits in sandstones, such as copper and silver in Kazakhstan [13,14], copper and silver in Iran [15], uranium in Australia [16], copper in Argentina [17,18], vanadium in Argentina [19], copper in Arctic Canada [20], copper in Newfoundland [21], lead-zinc in China [22], and uranium in China [23]. The occurrence of red sandstones does not, however, necessarily explain mineralization, as other sources such as underlying basement may contribute more metal in certain hydrological regimes [24].
The high abundance of red beds, in space and time [25], implicates their potential importance to the mineralization of continental basins. Traces of mineralization occur in red beds dated at least back to the Mesoproterozoic [26]. However, there are only very limited data for the trace element compositions of the iron oxide coatings that are central to the process. This study reports the compositions of grain coatings from numerous different localities and ages.
The objectives are specifically: (i) Measurement of a data base of compositions of grain coatings. Data were collected for vanadium (V), cobalt (Co), copper (Cu), selenium (Se), uranium (U), arsenic (As), and lead (Pb) in iron-(Fe-) rich grain coatings. These elements represent those which commonly occur in red bed ore deposits (ii) Assessment of whether the composition of grain coatings is related to their relative abundance, i.e., quantity of iron oxide (iii) Assessing the typical range of metal fluxes that might be expected from their dissolution

Methods
The data set was measured from samples in Britain and Ireland, where a relatively limited area includes red sandstones from a range of stratigraphic ages, and the samples represent both first-cycle and multicycle erosion. Hence, they are a   varied assemblage of sandstones suitable to best assess the chemistry of typical red grain coatings, rather than sandstones that are all associated with ore deposits. Samples were collected from 34 localities in Britain and Ireland (Figure 1), predominantly from red sandstones of Devonian, Permian, and Triassic age and deposited in fluvial or aeolian environments. Single samples of Neoproterozoic and Silurian red sandstone were collected. In addition, a sample of Carboniferous sandstone was collected, in which reddening had occurred at a relatively late stage, below the sub-Permian unconformity which caused widespread deep oxidation [27]. Red sandstone in dykes cutting down through the same unconformity [28] was collected at Berry Head, Torquay, England. A sample of Triassic sandstone at Maghera, Ireland, was chosen because it appeared to be enriched in iron by recent "iron pan" soil mineralization. Samples of red and bleached sandstone were collected for comparison from Alderley Edge, England (Figure 2), where bleaching is implicated in copper mineralization [29,30]. Three red sandstone samples from other parts of the world (China, Australia, USA) were analysed to test if the main data base is typical of other regions. Samples were chosen that could be disaggregated readily and were not cemented by minerals that sealed in the iron oxide and prevented dissolution. Disaggregated samples were sieved to isolate the sub-500 micron or sub-250 micron fraction, depending on the predominant grain size. Sandstone from Gardenstown, Scotland, was separated into three size fractions, 500-250 micron, 250-125 micron, and sub-125 micron, to test if this factor might influence the compositions. Finally, a pure quartz sandstone with no iron oxide grain coatings (Cretaceous Lochaline Sandstone, Scotland) was analysed as a procedural blank. Samples for acid leaching were added in weighed amounts of about 150 mg to 4 mL aqua regia (3 : 1 HCl:HNO 3 , both analytical grade for trace elements) and boiled for 30 minutes at 95°C in a microwave. The liquid was then centrifuged, and the residue was subjected to a repeat of the procedure if any iron oxide remained undissolved (optical appearance). Only up to 20 wt.% of each sample was dissolved by this procedure, leaving more than 80 wt.% residue. The liquids from centrifuging were com-bined, and V, Co, Cu, Se, U, As, and Pb were measured by Triple Quad ICP-MS Agilent 8800 (Agilent Technologies), in the Trace Element Speciation Laboratory (TESLA), Department of Chemistry, at the Aberdeen University, UK. 10 μg/kg Rh in 2% HNO 3 was used as an internal standard. The total leached element masses were determined (to an accuracy of 0.001 g) and converted to concentrations of the original rock. Measurements were also taken from certified reference materials PACS-2 and BCSS-1 (marine sediments), and NIST2711a and NIST2709a (soils), and an absolute blank. Our data for V, Co, Cu, and Se for the NIST2711a and NIST2709a soils are consistent with the certificated data for the "acid-extractable" fraction. For the two sediments PACS-2 and BCSS-1, there is no certificated data for the "acid-extractable" fraction but only for the "whole rock" fraction, which also is in agreement with our data for Co, Cu, and Se, but show a higher value for V. Our acid leachable uranium, which had a recovery of 31-50% from certified totals, shows up to 70% totals from other ICPMS methods using PACS-2, and NIST SRM 2711a, 2709a [31,32]. This indicates that the reported acid leachable uranium values most likely underestimate the content of uranium in the grain coatings.
Bulk analyses of Fe contents in all samples, and V and Cu in a subset of ten Permian and Triassic sandstone samples (available from another project that did not include the other samples), were made after multiacid digestion (perchloric, nitric, hydrofluoric, and hydrochloric acids) by ICP-MS.
Thin sections of a red sandstone from the Triassic of Northern Ireland (    Optimization of the instrument was performed by ablating the glass reference NIST612. Two reference materials of MRM MASS-1 from USGS and UQAC-FeS-1 from SLIM were used as calibration standards only.

Results
Each of the red sandstone samples processed by acid removal of grain coatings was leached, indicating liberation of iron and associated trace elements to solution. Analysis of the solutions, converted to concentrations of whole rock, confirmed that trace elements had been liberated ( Table 1). The absolute blank and procedural blank yielded negligible contents (Table 1). Data for certified reference materials and all standard deviations are given in the Supplementary Information (Table S1). SEM observations confirmed that while iron oxides were removed by acid leaching, the substrate grains remained unaffected. Ranges of elements determined included 0. The bleached sandstone from Alderley Edge yielded mean acid leach values higher than the red sandstones there. All the values for Cu, Co, and Pb at Alderley Edge are much higher than in other samples. The sandstone from Maghera enriched by soil mineralization contains relatively high levels of V, Co, Cu, Pb, and U. Comparison of the grain size fractions from Gardenstown showed that the composition of the two coarser fractions that were purely sand size was similar, but the finer size fraction that included mud had higher contents of trace elements.
The subset of Permian and Triassic samples measured for both whole rock composition and grain coating leach composition have mean compositions of 11 ppm V and 7.6 ppm V and 6.1 ppm Cu and 5.2 ppm Cu, respectively.
Whole rock Fe contents range up to 4.74%. The contents of V, Co, Se, U, and Pb show moderate positive correlation with Fe content (R 2 values 0.32 to 0.69; Figure 3). There is no correlation between Fe and Cu. Distributions of values within the data set show a bias towards lower values, but in the case of Cu, there is an additional group of higher values ( Figure 4).
The LA-ICP-MS maps of a quartz grain with surrounding iron oxide coating in the thin section of red sandstone show a clear contrast between sand grain and grain coating ( Figure 5). The grains exhibit negligible (less than 1 ppm) concentrations of V and Co, while the surrounding iron oxide shows V and Co at up to 120 ppm and 60 ppm, respectively.

Compositions of Grain Coatings.
Previous proposals of a model of mineralization from leached grain coatings are based on the circumstantial occurrence of bleached sandstones in ore fields [14,17,18], distal reprecipitation of leached metals [6,7], or whole rock data [11]. Measurements of some trace elements in iron oxide grain coatings were made by [33], including U and V and rare earth elements but not Cu, Co, or Se. The data set reported here for grain coatings is more comprehensive than hitherto. Variations in compositions of grain coatings are assessed here in terms of the iron content and the stratigraphic age. The relatively high values of several metals within the soil-influenced sample from Maghera indicate that enrichment in iron oxides can occur on a geologically short time

Geofluids
scale. Similar enrichment at the surface is the basis of mineral exploration using metal anomalies in iron oxide precipitates in the surface environment [34,35]. The early incorporation of trace elements is consistent with the model for metal release by mineral alteration during early diagenesis [2,9].
Most sandstones have Fe contents in the range of 0.17 to 4.74% (mean 1.27%). The mean concentrations of aqua regia leachable elements from the unmineralized red sandstones are 14.0 ppm Cu, 12.0 ppm V, 0.4 ppm U, and 3.1 ppm Co. The median values are comparable, except in the case of copper which has a much lower median value of 6.3 ppm. In contrast, the pure quartz sandstone contains negligible contents of iron or trace elements (less than 1% of the mean values). The broad correlation of iron and other metals in the whole rock data (Figure 3) suggests that the residence for much of the trace metal is within the iron oxide grain coatings. This is consistent with the LA-ICP-MS maps, which show trace elements only in the grain coatings and not in the substrate grains. The lack of correlation between Cu and Fe ( Figure 3) may reflect the incorporation of Cu into other phases such as aluminosilicates, as observed elsewhere in red sandstones [17,36].
The measurements made are of metal that is readily leachable from the sandstone, rather than the total metal content of the sandstones. Some metal within detrital grains may not be readily accessible to the leaching fluids, whereas the metal in iron oxide grain coatings is amenable to leaching. However, a comparison of whole rock data and grain coating leach data for a subset of ten red sandstone samples analysed for V and Cu showed that over 70% and 85%, respectively, of the trace elements reside in the grain coatings and is leachable. Notably, the mean values of metals in the grain coatings are not anomalously high compared to typical rock compositions, but the occurrence of the metals in abundant thin coatings makes them relatively available.

Geofluids
The aqua regia used to remove the iron oxide grain coatings is a more acidic and oxidizing solvent than would occur in nature, and the dissolution temperature (95°C) in the laboratory is higher, so shows the total potential that can be released, rather than what might be released during a single fluid flow event. Nevertheless, given that ore deposits may form over a period of millions of years (Brown, 2009), the total potential is an appropriate value to measure. Red bed sandstones commonly exhibit zones that are completely bleached, indicating that the iron oxide grain coating is completely removed over time.
Comparison of metal values with stratigraphic age shows that values of V and U are higher in the red sandstones of pre-Permian age than in those of Permian and Triassic age (Figures 6 and 7). The mean values for V are 17.7 ppm (n = 13) and 6.7 (n = 21), respectively. The mean values for U are 0.54 and 0.28, respectively. In contrast, values for the other elements show no relationship with stratigraphic age. The range of pre-Permian values for Cu is greater than for Permo-Triassic values, but the mean values are similar (Figure 7). The pre-Permian samples are mostly from northern Britain, where the provenance is in the Caledonide basement and sandstones represent first cycle erosion. They are therefore more likely to contain mineralogically immature grains than the Permo-Triassic sandstones which are commonly derived from multiple cycles of erosion, especially the aeolian Permian sandstones. The immature grains would have yielded trace metals during diagenetic alteration, whereas quartz-rich sandstones have much less potential. However, the higher contents in pre-Permian sandstones also reflect higher Fe contents ( Figure 6). In fact, the V/Fe, Se/Fe, and U/Fe ratios are lower in the pre-Permian samples. The lower ratios in older sandstones could indicate that some of the trace elements have been released from the grain coatings, at a higher thermal maturity than in the younger sandstones. With thermal maturation, the crystallinity of haematite increases [37], and the capacity to adsorb trace metals would decrease with recrystallization [38], although changes in surface area during alteration could complicate the effect on adsorption. The contrast with age may be more evident in the case of V, Se, and U because of very ready redox-controlled mobility [39,40], which is also evident in the predominance of these elements in diagenetic reduction spheroids [41,42]. On the other hand, vanadian haematite containing percent level V is identified in some ore deposits, so this can occur as a stable phase [43,44].

Ore Fluids.
These measurements allow us to calculate the volume of rock required to supply metal to a large ore deposit ( Table 2). For a sandstone of density 2.5.10 3 kgm -3 (values typically range from 2.2 to 2.8.10 3 kgm -3 ), a metal availability of 10 ppm equates to 1 kg of metal from 40 m 3 (1 Mt from 40 km 3 ) or 2.5 kg of metal from 100 m 3 (2.5. 10 4 tonnes from 1 km 3 ) rock. Aquifer volumes for Triassic red beds, measured for potential CO 2 storage, range from hundreds to thousands of km 3 [45][46][47]. For 10 ppm metal availability, an aquifer of  [14]. The mean size of sediment-hosted Cu deposits is also 22 Mt [48], equivalent to 140% of the Cu available in 10 3 km 3 . Previous calculations for the Triassic red beds of central England concluded that 1 ppm Cu from 340 km 3 could yield over 0.8 Mt Cu [49], and 2 ppm Cu from 3300 km 3 Triassic sandstone could yield 15 Mt Cu [50].
Most sandstone-hosted U deposits contain 0.1 to 10 kt U [51]. This represents 0.01 to 1.4% of the U available from 10 3 km 3 ( Table 2). Most sandstone-hosted V deposits contain 1 to 100 kt V [52], which would require 0.004 to 0.4% of the V available from 10 3 km 3 ( Table 2). Most types of Co deposit contain 5 to 500 kt Co [53], requiring 0.09 to 8.5% of the Co that could be released in 10 3 km 3 .
The data for red and bleached sandstones at Alderley Edge, where Cu mineralization is attributed to the dissolution of grain coatings by reducing brines including hydrocarbons [30,50], show variable patterns for different elements. The bleached sandstones have lower values of V and Co, but much higher values of Cu, As, and Pb. However, the high Cu values in the bleached sandstones indicate that the Cu was not exported from the system and has another residence.
The bleached sandstones have a lower iron content, but they contain a range of secondary Cu minerals [29]. The lower values of V, at least, might reflect uptake by migrating hydrocarbons.
Multiple factors will control metal release, including pH, Eh, temperature, and pore fluid chemistry [10]. The model considered here envisages the dissolution of grain coatings by acidic fluids, particularly migrating oil and gas. Largescale hydrocarbon migration, evidenced by widespread oil residues, therefore, contributes to the exploration for sandstone-hosted ore deposits [16,17]. Most oil is generated <130°C, and it is likely that related removal of iron oxide grain coatings occurs at shallower depths and lower temperatures. This implies the removal of coatings and liberation of trace metals at depths of less than 4 km. At these depths, there is still adequate porosity (typically up to 10%) to allow intergranular fluid flow through sandstones and export of diagenetic alteration products.
The compositions of the grain coatings measured here are unexceptional, because the sandstones did not have a known provenance in preexisting metalliferous ore deposits. If the sand and/or grain coating was derived from a metalrich source, red beds could be preloaded with metal that could more readily be leached to form an ore deposit. This is most likely where sand was eroded from an orogenic belt, in which metalliferous phases had been unroofed following   10 Geofluids uplift. In this way, arc-based deposits such as Cu porphyries, V-rich magnetite, and Se-rich volcanic soils could be recycled into sandstone ore feedstock. A mineralized plate margin mountain chain with subaerial erosion products (red beds) and a foreland basin (potential hydrocarbon source rocks) could provide a scenario where hydrocarbons strip enriched grain coatings from the red beds to create an ore fluid ( Figure 8). For example, in the Tethyan region (Iran, Iraq, and adjacent areas), where red bed-hosted copper deposits are widespread [15], the Zagros and other orogenic mountain chains are fringed by red beds and hydrocarbon sources. In the Andes, copper mineralization in red beds has been attributed to the unroofing of arc volcanics [54]. Examples like these could be tested by mass balance studies of metals in source regions and ore deposits. In particular, the alteration of titanomagnetite is a suggested source of metals in red bed deposits [55], and there is evidence of such alteration in mineralized red beds [54,56] which could be quantified. In the Andes, the magnetite in volcanic rocks is rich in vanadium [57], altered magnetite in the volcanic rocks shows the release of vanadium [58], and red beds are rich in magnetite grains [59], suggesting that this region would make a good case study. While the vanadium in titanomagnetite ores can be processed and reprecipitated by microbes in the soil zone (e.g., [60,61]), the contribution of microbial activity in the whole mineralization cycle has yet to be evaluated.

Limitations to Leaching
Model. This study focused on red sandstones, as sandstones are potential aquifers for fluid flow and diagenetic reactions including dissolution of grain coatings. Red mudstones may contain higher contents of iron oxides and associated trace metals [50], as shown by the data from different grain size fractions at Gardenstown. Some trace metals in mudstones would be released into compactional brines, but they are not commonly pathways for the hydrocarbon migration envisaged to leach iron oxides from red sandstones. An exception is in mudstones which show evidence for overpressuring, which allows the flow of brines and hydrocarbons along bedding planes, a circumstance that may be promoted by impermeable evaporite beds that occur in some red bed sequences.
Leaching data reported by [8] are based largely on very young mudrocks of Holocene-Pliocene age in which the iron oxide crystallinity was not fully developed and allowed element liberation more readily than from fully crystalline oxides. Their data are nonetheless interesting in indicating a relative ease of leaching in which Co > Cu > Fe > U > V. The relatively high availability of Cu recorded in the present study and by [8] may partially reflect the high adsorption (log K values) of copper onto iron oxides [62,63]. V and U (and Se) might be harder to desorb if initially adsorbed as oxyanions [63]. A calculation based on leaching data for zinc in the red mudrocks concluded a volume of 6.10 4 km 3 rock was required to source a sedex deposit [8,64], at least an order of magnitude more than the volume required to source a large copper deposit. The contrast in volumes required may reflect differential adsorption of the elements onto iron oxide [62,63], whereby copper outcompetes zinc for adsorption, and consequently, desorption occurs at greater rates and possibly under different pH conditions.
Bleached sandstones at Alderley Edge locally contain secondary Cu minerals, indicating that leached metals are not necessarily exported from the system but may be reprecipitated by variable redox conditions. This is borne out by the high Cu contents in the bleached samples at Alderley Edge. This difference in mobility may also contribute to the poor correlation of Cu and Fe contents. Copper accumulation in bleached facies has similarly been observed in the Kupferschiefer [65], and it is evident as metal-enriched cores in reduction spheroids in many red beds [42]. The potential for leaching to create an oreforming fluid clearly requires a sedimentary basin scale and fluid flow system that are large enough to allow continued, cumulative transport of metals.

Conclusions
Measurement of trace elements resident in iron oxide grain coatings in red sandstones confirmed that they are a source of metals during diagenetic dissolution. In particular (i) They represent a significant proportion of the metals available from the sandstones to ore-forming fluids (ii) The quantities of metal in the grain coatings are adequate to form sandstone-hosted ore bodies. Only small proportions of available metal would be required to form U, V, and Co deposits, but the available Cu is still adequate to form moderate-size ore bodies if the process is efficient (iii) A broad correlation between Fe content and the content of most trace metals confirms the predominant residence of the trace elements in the grain coatings, and hence, their availability to ore fluids Future research should explore basins where red sandstones have a provenance which includes anomalous metal concentrations, for example, sandstones containing erosion products of mineralized granites or arc volcanics.  Figure 8: Schematic evolution of red bed ore deposit by erosion of magmatic arc rocks into red beds, followed by hydrocarbon migration through red beds to concentrate metals from metalbearing grain coatings.

Data Availability
The manuscript is a data self-contained article, whose results were obtained from the laboratory analysis, and the entire data are presented within the article. However, if any additional information is required, these are available from the corresponding author upon request to the e-mail j.parnell@abdn.ac.uk.

Conflicts of Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.