Mechanisms of secondary carbonate precipitation on felsic, intermediate and mafic igneous rocks: a case study for NW Scotland

Fieldwork and sampling were conducted in February and April 2022 across NW Scotland, UK. Seven sites were visited: Bad an Sgalaig, Badcall (small and main), Goat Crag, Mungasdale, Rubha Carrach and Meall an Fhir-Eoin (Figs 1 and 2). All the sampling sites listed are part of the Lewisian Gneiss Complex (BGS 2020), except for Rubha Carrach and Meall an Fhir-Eoin, which are part of the Ardnamurchan Central Complex (BGS 2020). The first group is primarily made up of meta-igneous rocks and the latter is made up of igneous rocks (Table 1).

The altitude of each of the sites is highly variable (Table 1), with the lowest being Rubha Carrach at 17 m above mean sea level (AMSL) and the highest being Meall an Fhir-Eoin at 168 m AMSL. In Northern Scotland, the average annual wind speed at 10 m AMSL is 20.8 km h^–1. The most intense winds are experienced in February (24.9 km h^–1) and the least intense winds are experienced in July (17.1 km h^–1), which may affect evaporation of water and subsequent mineral precipitation. The prevailing wind direction is SW. In the same area, minimum annual temperatures reach 4.25°C and maximum annual temperatures reach 10.60°C. In Scotland, the mean annual relative humidity is c. 80% (Met Office 2023).

Scotland is a geologically variable country resulting from deformation during a series of different tectonothermal events. The Lewisian Gneiss Complex, one of the main rock formations cropping out in the sampling localities in this study, covers a vast area across NW Scotland and the Outer Hebrides. Archean-aged felsic gneisses make up a large area of this complex, with a portion containing the Scourie dyke intrusions, which have a mafic composition (MacDonald et al. 2013). The other geological unit explored in this study was the Ardnamurchan Central Complex, which is also in the NW region. It is one of the classic sub-volcanic central complexes emplaced during the opening of the North Atlantic (Magee et al. 2013). There are three recognized sub-units (referred to as centres) for Ardnamurchan that are composed of different geology. The first is made up of minor mafic intrusions and volcaniclastic rocks (Brown and Bell 2007). The second has many gabbro intrusions of variable geometries and ring dykes (Richey and Thomas 1930; Day 1989), and the third is one large gabbro (O'Driscoll et al. 2006).

At each of the seven sampling localities, mafic, intermediate and felsic bedrocks were collected (Fig. 3) to facilitate comparisons between sites. In some cases, bedrock that was covered in carbonate was also sampled; however, this was not available in all localities. Secondary carbonates, which are minerals formed via rock weathering, were collected from the sites if available and/or accessible (Fig. 4). These carbonates are observed in various forms such as bulbous carbonates, carbonate covering the bedrock, in situ and ex situ tufas, carbonate-encrusted moss and carbonate mud. All bedrock and bedrock with carbonates were sampled from cliff faces using a sledgehammer and placed into labelled plastic bags. Secondary carbonates were selectively sampled using a Leatherman Wave®+ multi-tool pull-out knife and placed into 15 ml centrifuge tubes. Carbonate encrusted moss samples were taken from cliff faces using the claw end of a rock hammer and placed into small plastic bags. A subset of each of these samples was prepared for characterization of their geochemical and mineralogical properties.

At Badcall (main site: BCM) two types of water samples were collected. The first set was sampled from two small pools of rainwater found at the top of the cliff face on which secondary carbonates were sampled (sample ID: BCM pond A and BCM pond B). The second sample set was collected from rock fractures where the pooled water from the top of the cliff had dripped down the rock face (sample ID: BCM drip water A and BCM drip water B). BCM pond A and B samples were collected using a 30 ml syringe and transferred into a 15 ml centrifuge tube. BCM drip water A and B were sampled directly from the cliff into a 15 ml centrifuge tube. After sampling, the drip water samples (15 ml) were filtered using 0.22 µm syringe filters and acidified using 2% HNO₃ (by volume) to ensure no precipitates would form. To rule out the possibility of seawater being the sole contributor to carbonate precipitation, the Ca/Na ratios of North Atlantic seawater were compared with the ratios found in the pond and drip water samples from Badcall.

Large chunks of bedrock from each sampling site were crushed using a RETSCH jaw crusher. Each sample was fed through the crusher three times to ensure the smallest possible grain size was achieved. Afterwards, bedrock samples were triple bagged and manually smashed using a rock hammer until all particles were <8 mm in diameter. After crushing, samples from each of the sites were thoroughly homogenized prior to any analysis. Subsamples were taken (10 g aliquots) and shipped to ALS Laboratories in Ireland to be analysed for major oxides by inductively coupled plasma atomic emission spectrometry (ICP-AES). A crushed sample split (≤250 g) was ground in a ring mill pulverizer using a carbon steel ring set. The sample split was pulverized to better than 85% passing 75 $μ$m. The sample (100 mg) was added to lithium metaborate–lithium tetraborate flux, mixed thoroughly and fused in a furnace at 1000°C. The resulting melt was cooled and dissolved in 100 ml of 4% nitric acid–2% hydrochloric acid. The solution was then analysed by ICP-AES and the results were corrected for spectral inter-element interferences. The oxide concentration was calculated from the determined elemental concentration, which is the reported format in this study. Bedrock from each of the seven sites was classified as felsic, intermediate or mafic based on the XRF data. Samples with more than 65 wt% SiO₂ are considered felsic, samples between 55 and 65 wt% SiO₂ are considered intermediate and samples between 45 and 55 wt% SiO₂ are considered mafic.

Another subsample (3 g aliquots) of the crushed bedrock from each site was milled for 7 min at a frequency of 25 s^–1 using a RETSCH MM 400 Ball Mill. This material was passed through a 45 $μ$m sieve to ensure particles were a suitable size for X-ray diffraction (XRD). XRD patterns for the bedrock samples were collected at the University of Glasgow on a Malvern Panalytical Empyrean X-ray diffractometer with PIXcel3D-Medipix3 1 × 1 detector using Cu K$α$ radiation (wavelength 1.541874 Å) in Bragg–Brentano reflection geometry. Data collection and mineral identification were completed using Highscore Plus software.

Rock chips were cut from each of the original bedrock samples (Table 1) using a Logitech CS10 thin section cut-off saw to fit within the dimensions of a standard thin section slide (26 mm × 46 mm). These chips were then mounted with Buehler EpoHeat heat resin (4:1 resin to hardener) and vacuumed five times prior to curing for 24 h at 60°C. After the resin hardened, samples were cut again using the same saw to achieve a thickness of c. 1 cm. Glass thin section slides were added to a Logitech LP30 Production Lapping and Optical Polishing Machine and frosted using aluminium oxide abrasive (23 µm) from Lapmaster Wolters. The 1 cm thick rock chips were then added to the lapping machine for 30 min or until the surface of the rock was exposed from the resin. Afterwards, these samples were mounted onto the frosted thin section slide using Buehler EpoThin 2 resin (2:1 hardener to resin). Samples cured for 48 h and were then ground using a Buehler Petrothin thin section system to obtain a total thickness of c. 100 µm for the samples. Samples were then added back onto the lapping machine until thin sections were c. 30 µm thick. To ensure the correct thickness was achieved samples were assessed using a Michel–Levy colour chart under cross-polarized light (XPL) and also measured using a digital caliper. These thin sections were individually analysed using a Leica DM750P microscope at ×2.5 and ×5 magnification. Images were captured using a Leica ICC50 W microscope camera with Leica LAS EZ software.

All carbonate samples (3–5 g aliquots) were dried under ambient conditions prior to milling, which used the same protocol that was followed during milling of bedrock samples. XRD patterns for carbonate samples (2 g) were collected at the University of Glasgow using a benchtop Rigaku MiniFlex 6G X-ray diffractometer equipped with a Cu sealed tube (K$α$1 and K$α$2 wavelengths; 1.5406 and 1.5444 Å, respectively), consistent with the methods of Khudhur et al. (2022). Data collection and mineral identification were completed using Rigaku SmartLab Studio II software (Rigaku 2022).

After milling, aliquots (180–220 µg) of carbonate samples were weighed using an A&D Weighing BM-252 Ion Semi-Micro Balance for stable carbon and oxygen isotopes. Values of $δ$¹³C and $δ$¹⁸O of the various carbonate samples were analysed by continuous flow isotope ratio mass spectroscopy at the University of Glasgow. Samples were acidified using phosphoric acid (≥1.90 SG) and heated for 1 h at 60°C on an Elementar GasBench and analysed on a Isoprime 100 mass spectrometer. Samples were run in triplicate and the average was reported as the result with 1 SD. Values were calibrated to Vienna PeeDee Belemnite (VPDB) using NBS-18 and IAEA-603 reference standards. A secondary standard, IA-RO2 (Iso-Analytical Ltd) was used to validate the calibration linearity for more depleted values of $δ$¹³C and $δ$¹⁸O than the primary standards have. Analytical uncertainties of 0.4‰ on $δ$¹³C and 0.8‰ on $δ$¹⁸O were obtained on measurements of IAEA-603 (n = 20). Calibration data for $δ$¹³C and $δ$¹⁸O are available in the Supplementary material Table S1.

Water samples were analysed at the University of Glasgow by inductively coupled plasma optical emission spectrometry (ICP-OES) on an Agilent 5900 SVDV system fitted with an AVS 7 valve for 5800/5900ICP-OES. The samples were introduced to the instrument using an SPS 4 Autosampler. The gas is zero grade argon, and 2% nitric acid (analytical reagent grade) was used as the rinse solution. The square of the correlation coefficient was higher than 99% for the calibration. Additional information on this dataset is available as Supplementary material Table S2. A subsample of these water samples was taken and pH was measured using a Orion Star™ A211 Benchtop pH Meter.

Secondary carbonate deposits were observed on and surrounding bedrock at all seven sampling localities in NW Scotland. These carbonates primarily occurred as bulbs whereas the rest of the carbonates appeared as mud, carbonated moss, scaling across bedrock or as a tufa (Fig. 4). Thin coatings of carbonate scaling (Fig. 4a) were noted on the bedrock, some in small patches only centimetres wide whereas others stretched for many metres. Solid carbonate bulbs (Fig. 4b) were found in rock fractures and/or on the bottom of a cliff overhang. Tufa was either found in situ on the cliff face (Fig. 4c) or ex situ, on the ground near a cliff face (Fig. 4d). Carbonate encrusted moss was consistently found at the bottom of cliffs where rocks and different organic matter were in contact with one another (Fig. 4e). In some cases, carbonate mud was observed in rock fractures similar to the bulbs; however, it was often found in damp areas on the cliff face where rainwater had been dripping (Fig. 4f).

At Badcall (main site) water ponding was observed at the top of the cliff. At the time of sampling, water was slowly streaming down the cliff face and into the rock fractures where many carbonates were found. Similar observations were also made at Mungasdale and Meall an Fhir-Eoin; however, pooled water could not be sampled owing to limited accessibility. Each of the sites was within a few kilometres of the North Atlantic Ocean, and in some cases, experienced substantial amounts of sea spray. During sampling at Badcall (small and main) there were severe winds with mist coming from the sea, which may have had an influence on the composition of water samples.

Bedrock from Goat Crag was classified as felsic owing to its high abundance of SiO₂ (65.60 wt%). It contained major abundances (>1.0 wt%) of Al₂O₃ (17.25 wt%), Fe₂O_3tot (total iron as ferric; 4.61 wt%), CaO (4.92 wt%), MgO (1.67 wt%) and Na₂O (4.84 wt%). All other elements were present in minor (0.1–1.0 wt%) or trace abundances (<0.1 wt%) and loss on ignition (LOI) is 1.18% in the Goat Crag bedrock (Table 2). The alkali content was plotted against the silica content using a total alkali–silica (TAS) diagram to help aid in the further classification of bedrock, in which this sample was identified as dacite (Fig. 5).

In thin section the bedrock from Goat Crag (Fig. 6) contained very abundant amounts of biotite, which exhibited moderate relief and a deep brown colour and strong pleochroism in plane-polarized light (PPL). Across the thin section a sieve texture is observed where pyroxene is replaced by hornblende and quartz. Hornblende is abundant in different areas across the thin section, exhibiting a green colour and moderate pleochroism in PPL with distinctive cleavage intersecting at 120°/60°. Quartz was very abundant across the thin section as shown by its low relief in PPL and lack of cleavage planes. Plagioclase was abundant, but to a lesser extent than the previously described phases, and was identified by its characteristic polysynthetic twinning. Where plagioclase was present in lower modal abundances, sericite (mica group) was more abundant as it probably replaced plagioclase during weathering.

Two bedrock samples were classified as intermediate; one from the main Badcall site and the other from Mungasdale, as these samples contain 57.80 and 59.30 wt% SiO₂, respectively. Both samples were dominated by major abundances of Al₂O₃, Fe₂O_3tot, CaO, MgO and Na₂O, and the Mungasdale sample also had major amounts of K₂O. All other elements were present in minor or trace abundances and LOI is 2.06 and 1.53% in the respective Badcall and Mungasdale bedrock (Table 2). These samples were classified using the TAS diagram as trachyandesite and andesite, respectively (Fig. 5).

In thin section, both intermediate samples had an abundant amount of quartz and biotite, similar to the felsic sample from Goat Crag. In the Badcall main bedrock (Fig. 7a and b) there is an abundant amount of plagioclase and clinopyroxene, which was identified by its 90° angle between cleavage planes. Clinopyroxene was visible in the Mungasdale samples (Fig. 7c and d) as well, albeit in lower abundances. Within the Mungasdale sample olivine was present in sparse amounts, recognized by its distinctive fracturing. Plagioclase and sericite are also abundant. Hornblende was visible in sparse amount across both samples.

Badcall small and main, Bad an Sgalaig, Rubha Carrach and Meall an Fhir-Eoin were classified as mafic as they contain 52.20, 50.70, 52.00, 48.50 (n = 2) and 47.60 wt% SiO₂, respectively. Both samples were dominated by major abundances of Al₂O₃, Fe₂O_3tot, CaO, MgO and Na₂O, and the Rubha Carrach sample also had major amounts of K₂O. The LOI for each sample is given in Table 2. All mafic bedrock samples were classified as basalt, with the exception of Badcall small, which was classified as basaltic andesite (Fig. 5).

The Badcall (small site) sample (Fig. 8a and b) is dominated by subhedral or anhedral clinopyroxene, evident by its distinctive 90° cleavage planes and pleochroism in PPL. Plagioclase is also abundant but is replaced by clay in some areas across the sample. Hornblende and quartz (<5%) are also present in sparse abundances. Badcall (main site; Fig. 8c and d) was abundant in laths of plagioclase, where some areas had been replaced by sericite, as well as hornblende, which is identified by its pleochroism from green to blue–green and distinctive cleavage. Orthopyroxene and clinopyroxene are available in sparse amounts. A granoblastic texture can be observed in the mafic bedrock sample, as well as the mafic bedrock sample covered in carbonate (Fig. 8e and f). Mafic Bad an Sgalaig bedrock (Fig. 9a and b) and Rubha Carrach bedrock (Fig. 9c and d) is dominated by laths of plagioclase, with sparse minor amounts of clinopyroxene present in the former and more abundant amounts present in the latter. Carbonate covering is visible in the Rubha Carrach sample. Meall an Fhir-Eoin (Fig. 9e and f) contains an abundance of plagioclase and clinopyroxene with subhedral olivine crystals throughout the sample. The XRD results corroborate the petrography observations, as plagioclase was the most dominant mineral phase present amongst the bedrock samples, followed by clinopyroxene, olivine, quartz, hornblende, chlorite and biotite mica (Table 3).

All carbonate samples are dominated by calcite, except for one bulbous carbonate sample from Goat Crag that showed intense peaks for monohydrocalcite (CaCO₃.H₂O). Bulbous carbonates from Bad an Sgalaig, Badcall and Meall an Fhir-Eoin, and carbonate moss from Badcall showed minor aragonite (CaCO₃) peaks. Minor gypsum peaks were also detected in bulbous carbonate samples from Bad an Sgalaig, Badcall, Goat Crag and Mungasdale. A full list of identified minerals is available as Supplementary material Table S3. Carbonate that was covering a large area of the surface of bedrock at Badcall was observed in thin section. Calcite was identified under XPL owing its low-order birefringence colours and rhombohedral cleavage, corroborating the XRD results. No other carbonate samples were examined in thin section owing to difficulties preserving the sample, as they typically were polished away during preparation stages.

The $δ$¹³C and $δ$¹⁸O values of the suite of carbonates (n = 16) range from −7.3 to 14.9‰ VPDB and from −5.2 to 0.5‰ VPDB, respectively (Fig. 10). In general, bulbous carbonates collected from different sites are the most enriched in ¹³C (4.4–14.9‰), followed by carbonate covering the bedrock (−7.3 to 14.0‰), ex situ carbonates (−1.0‰), carbonate mud (−2.8 to −2.4‰), carbonate encrusted moss (−3.5‰) and tufas (−5.5 to −4.9‰). The $δ$¹⁸O values ranged from −4.45 to 0.45‰ for bulbous carbonates, −5.6 to −3.01‰ for carbonates covering the bedrock, −0.99‰ for the ex situ tufa, −3.20 to −2.41‰ for carbonate mud, −2.76‰ for the carbonate encrusted moss and −3.21 to −0.76‰ for in situ tufas.

Both pooled water samples collected at the top of cliff faces show consistently lower concentrations of Ca, Mg and Na than their respective drip water samples (Fig. 11). BCM drip water A had four times the amount of Ca, twice the amount of Mg and an equal amount of Na, respectively. Similarly, BCM drip water B had twice the amount of Ca and Mg and 1.5 times the amount of Na compared with the pooled samples. The pH of water samples ranged from 7.7 to 8.4.

The mineralogical and geochemical composition of the target bedrock is a critical property that governs its ability to mineralize and sequester CO₂. In this study, basalt and basaltic andesite were the dominant rock types sampled (six of nine) followed by trachyandesite, andesite and dacite, which make up the remaining samples (Fig. 5). The use of mafic rocks, such as basalt, for safe and long-term storage of CO₂ has gained significant recognition from CarbFix and CarbFix2. During the first stage of the CarbFix project approximately 190 tons of CO₂ was mineralized within 2 years of its injection into Icelandic basalt (Matter et al. 2016; Snaebjornsdottir et al. 2017, 2018) and the project has since been upscaled in an attempt to capture and remove more than half of the CO₂ and all of the H₂S emissions produced from the Hellishedi power plant since 2014 (Gunnarsson et al. 2018; Clark et al. 2020). CarbFix basalt has a desirable mineralogy (Pogge von Strandmann et al. 2019) and is generally made up of olivine and plagioclase, which have relatively fast dissolution rates (Pokrovsky and Schott 2000; Thom et al. 2013) compared with minerals dominating intermediate and felsic rocks (Gudbrandsson et al. 2014). The basalt and basaltic andesite samples analysed in this study have comparable mineralogy with the Icelandic basalt as both contain olivine, clinopyroxene (i.e. diopside) and plagioclase (Table 3). As a result, there are similar abundances of Ca and Mg present; however, the abundance of SiO₂ and Na + K is higher (≤4 and ≤1–2 wt%, respectively) within samples in this study. Although the basalt from CarbFix is carbonated under higher temperature and pressure compared to the basalt in this study, which has been weathered under Earth surface conditions, both types of rock share similar mineralogical and geochemical characteristics (Fig. 5), making them suitable feedstocks for CO₂ sequestration.

The suitability of basalt and basaltic andesite as a feedstock for CO₂ mineralization heavily depends on the dissolution rate of the minerals that make up these rocks. Olivine and plagioclase, which were two dominant minerals in these samples, are sources of Mg (sometimes Fe) and Ca, respectively. Samples contained 9.28–16.30 wt% Fe₂O_3tot, 8.63–13.55 wt% CaO and 5.04–6.66 wt% MgO (Table 3), which is very similar to the composition of highly desirable basalt used in CarbFix that contains 7.0–13.0 wt% Fe₂O₃, 7.0–10 wt% CaO and 5.0–6.0 wt% MgO (Alfredsson et al. 2013; Xing et al. 2022). Carbonation of Mg- and Ca-silicate minerals is thermodynamically favourable (Lackner et al. 1995), making minerals such as olivine (Chizmeshya et al. 2007; Daval et al. 2011) and plagioclase (Gislason et al. 2010; Munz et al. 2012) an ideal feedstock for long-term CO₂ removal.

The log dissolution rate for forsterite (Mg end-member of olivine) is −13.8 mol cm^–2 s^–1 at the pH of rainwater (pH = 5.6; Pokrovsky and Schott 2000; Oelkers et al. 2018), the most likely source of water in this study owing to high levels of precipitation in the area, whereas the log dissolution rate for anorthite (Ca end-member of plagioclase feldspar) is −15.7 mol cm^–2 s^–1 at the same pH (Declercq and Oelkers 2014). In this study, the primary carbonate phase identified was calcite; however, minor abundances of aragonite were also detected in some samples. In experiments testing calcite v. aragonite precipitation, De Choudens-Sánchez and González (2009) found that in solutions with low Mg/Ca ratios, calcite was the dominant mineral phase present, with aragonite forming in smaller quantities. As the Mg/Ca ratio in solution increased, aragonite became the most abundant mineral phase, eventually becoming the only carbonate precipitating. This indicates that precipitation of calcite, a stable sink for CO₂, may be inhibited as Mg/Ca ratios in solution increase (De Choudens-Sánchez and González 2009). The presence of calcite indicates that plagioclase and clinopyroxene probably supplied Ca during mineral dissolution, as they are the primary sources of Ca, other than potentially seawater. The water chemistry data from Badcall provides evidence of bedrock weathering, as there are increased concentrations of Ca, Mg and Na in the waters sampled from the cliff face compared with the source pond (Fig. 11). The Ca/Na ratios in these drip waters are consistently higher than that of seawater, indicating that Ca within the carbonates was not solely sourced from seawater. The average Ca/Na ratio from seawater in the North Atlantic Ocean is 0.038 (n = 3; Summerhayes and Thorpe 1996), whereas the drip water from this study is 5–108% higher than the former ratio.

Although the log dissolution rate of plagioclase is slower than that of olivine, it may have been compensated by the presence of labile Ca; this includes the loosely bonded Ca ions from the surface of the target mineral (Stubbs et al. 2022). Similar observations have been noted in other studies where there is a rapid initial dissolution rate in Mg silicate minerals that is consistent with the dissolution rate of highly reactive hydroxide phases (Vanderzee et al. 2019). Although this initial dissolution does not last for long periods of time, it may provide enough Ca for secondary carbonate precipitation. There is substantially more CaO (8.63–13.55 wt%) than MgO (5.04–6.66 wt%) in the studied basaltic samples, which suggests that there could be a higher quantity of labile Ca than Mg. Although the olivine dissolves orders of magnitude faster, there would be a greater quantity of Ca available for secondary calcite precipitation. A potential consideration is the possibility for Ca to be adsorbed to exchangeable sites where it would be consumed by other sinks rather than carbonate minerals (Tipper et al. 2021). For example, Ca could be exchanged on rock particles or in soils, rather than in carbonates. In some instances, Mg may be rapidly removed into clay minerals (Oelkers et al. 2019); however, clay minerals were not identified by XRD in the investigated samples, suggesting that this would not hinder carbonate precipitation.

Igneous and metamorphic intermediate and felsic rocks are much more abundant on the Earth's land surface (Amiotte Suchet et al. 2003) but have received much less attention for carbon capture and storage than mafic and ultramafic rocks. It is widely known that these rocks contain minerals with dissolution rates that are orders of magnitude slower than those of minerals making up mafic or ultramafic rocks (Declercq and Oelkers 2014); however, their abundance may offer currently untapped potential for CO₂ mineralization. Whole-rock geochemistry shows that felsic and intermediate rocks from this study contained 4.92–8.35 wt% CaO, 3.98–5.84 wt% Fe₂O₃ and 1.65–1.88 wt% MgO, which is consistent with values for other rocks in NW Scotland (Weaver and Tarney 1981; Guice et al. 2018, 2020). Field photographs (Fig. 4) show evidence of secondary mineralization, as extensive amounts of calcite were found among cliff faces of felsic and intermediate bedrock. In this region, there are no other known sources of Ca or Mg, which suggests that these were sourced from plagioclase, or other Ca-bearing minerals (e.g. clinopyroxenes) hosted in the bedrock. Although plagioclase has a relatively slow dissolution rate (Gudbrandsson et al. 2014) compared with olivine (Pokrovsky and Schott 2000) found in the mafic bedrock, it still supplies desirable cations for carbonation. For example, Power et al. (2013) reviewed different carbon sequestration technologies and estimated that the carbonation of felsic mine tailings could store 2.5 times more carbon than ultramafic tailings owing to their sheer abundance. Gaining a further understanding of the geochemical mechanisms driving carbonate precipitation in these rocks may have significant implications for the development of carbon capture and storage technologies.

Carbonates analysed in this study shared different morphologies (Fig. 4; e.g. bulbous v. carbonated moss) with different respective isotopic compositions making it important to break down the $δ$¹³C and $δ$¹⁸O values for the different types of carbonates to help gain an understanding of their precipitation process. There are five possibilities influencing the isotopic compositions of the carbonate samples: (1) the carbonates may contain mixed sources of carbon rather than forming from only one source (e.g. solely atmospheric or solely organic); (2) temperature; (3) CO₂ degassing and evaporation; (4) nearby waters; (5) sea spray (Fig. 12).

The primary water source for the detected carbonates was determined using known fractionation factors for calcite (O'Neil et al. 1969) and a range of temperatures that have been measured in NW Scotland (lows of 0°C, highs of 18°C, average annual temperature  8°C; Statistica 2021). Each solid carbonate oxygen value (⁠$δ$¹⁸O VPDB) was inverted to estimate a fluid value (⁠$δ$¹⁸O VSMOW), allowing for comparison with known values for meteoric water sources and seawater in Scotland. Calculated values ranged from −9.1 to 0.3‰ VSMOW (Fig. 13), with the more positive value resulting from higher temperatures (18°C), which are not common, and the more negative values resulting from cooler temperatures (0°C). Although maximum temperatures of 18°C were used, it is not likely that carbonates would be forming at these temperatures owing to (1) cooler temperatures recorded in the previous years, (2) limited days a year with these temperatures and (3) the occurrence of many of the carbonates in rock fractures, which are much cooler. As a result, the average annual temperature in Scotland (8°C) was also used, resulting in calculated values ranging between −7.0 and −2.1‰ VSMOW (Fig. 13). It should be acknowledged that the more enriched values are probably not accurate, as these carbonates were found inside rock fractures, where temperatures are near or below freezing (e.g. 2.1‰ VSMOW would be equivalent to −4.2‰ VSMOW at 0°C). Hoogewerff et al. (2019) studied the ¹⁸O composition of meteoric freshwater in Scotland and demonstrated a range of $δ$¹⁸O values across the country, from −9.8 to −3.0‰ VSMOW (Fig. 13), similar to the measured values in the present study. Global data for the oxygen isotope composition of seawater were presented by LeGrande and Schmidt (2006), where the $δ$¹⁸O values in NW Scotland range from 0.3 to 0.5‰ VSMOW (Fig. 13), failing to fall within the range presented in both the present study and that of Hoogewerff et al. (2019), further demonstrating that seawater is not a primary water source contributing to carbonate precipitation.

All bulbous carbonate samples were collected from rock fractures and the base of cliff overhangs, similar to a cave environment (Fig. 4b), and yielded positive $δ$¹³C values ranging from 4.4 to 14.9‰ VPDB. Numerous studies have reported stable carbon and oxygen isotope data for cryogenic cave carbonates (CCC); however, these carbonates typically form in the presence of ice (Munroe et al. 2021), yet ice was not observed at the time of sampling in this study. Žák et al. (2008) analysed powder carbonates found in or on ice in Scărișoara Cave, Romania, and reported positive $δ$¹³C values ranging from 1.0 to 12.0‰ VPDB, and Žák et al. (2004) reported $δ$¹³C values up to 6.0‰ VPDB for calcite on cave ice in Slovakia. Although the values reported in these earlier studies are from carbonates surrounded by ice, it is possible that ice was originally present at this study's sampling localities but melted prior to collection. Historical weather data from the Meteorological Office provide evidence that every sampling site reaches temperatures that are low enough to allow for ice formation during winter months (December, January, February).

At weather stations across Scotland, minimum and maximum temperatures (°C), days of air frost and other climatic parameters have been recorded from 1991 to 2020. In Altnaharra, Northern Scotland (44 km inland to the east from the Badcall sites) minimum temperatures reach below 0°C in the winter months, and in July maximum temperatures reach 18°C (Met Office 2023), which could facilitate freeze–thaw cycles. On the coast of the Atlantic in Aultbea, NW Scotland (10 km SW from Goat Crag and Mungasdale; 17 km SW from Bad an Sgalaig) the coolest days are also experienced in the winter months; however, minimum temperatures remain around 3°C. Despite this area rarely experiencing temperatures below zero, there are approximately 5 days per winter month that experience air frost, where the air temperature may be at or below the freezing point of water at a height of at least 1 m above the ground (Met Office 2023). In the spring and autumn, there are also between 0.5 and 3 days per month when air frost is observed, and in the summer maximum temperatures reach 17°C. In Dunstaffnage, NW Scotland (55 km inland to the SE from Rubha Carrach and Meall and Fhir-Eoin) minimum temperatures reach around 3°C in the winter, and in July maximum temperatures reach 18°C. Similar to Aultbea, this area also experiences several days of air frost in the winter (5–6 days per month) with an average of 26 days with air frost per year, allowing for possible ice formation.

Žák et al. (2008) stated that ice and CCC form when surface snowmelt occurs in the spring and water enter caves that have cooled to below 0°C during the preceding winter. Drip waters typically freeze into thin layers, which facilitates the rapid release of CO₂ from the solution in an open system. This often leads to high $δ$¹³C with a large range of values, similar to what is observed in the bulbous carbonates of this study. Extremely high $δ$¹³C values of CCC, up to 17.0‰, are rare in non-ice environments (Žák et al. 2008), suggesting that ice probably had an impact on the bulbous carbonates in this study, given that winter temperatures and days of air frost would allow this to happen. These samples generally fall within the same range as fine-grained CCC delineated by Žák et al. (2008) as opposed to coarse-grained CCC, which exhibit more negative $δ$¹³C values (Supplementary material Fig. S1; Clark and Lauriol 1997; Žák et al. 2004).

Equilibrium fractionation, which is temperature dependent, is likely to play an important role in the positive $δ$¹³C values measured in the bulbous carbonate samples. At lower temperatures equilibrium fractionation is greater, and in the case of this study, the air and water temperature are both low, which would result in more fractionation and subsequently greater ¹³C. For example, Socki et al. (2001) found an enrichment in ¹³C of residual bicarbonate and the precipitated calcite while testing fractionation during rapid freezing of bicarbonate solutions. The greatest enrichment in $δ$¹³C values was related to calcite that precipitated from repeated freezing and melting cycles (Socki et al. 2001), which is the most likely scenario based on the historical meteorological data.

Additionally, the shift towards higher $δ$¹³C values typically suggests that carbonates have formed in an environment where CO₂ degassing and equilibrium fractionation have occurred (Žák et al. 2008). The stable carbon isotope composition of carbonates is dependent on the waters, which will be affected by the total distance travelled; however, this information is unknown in the present study. If waters have travelled a long way, the initial $δ$¹³C could be high. However, if the water is meteoric, which is the likely scenario for this study, then waters have probably not travelled far. Upon initial exposure to the atmosphere, degassing would begin, causing an enrichment in ¹³C. For example, the precipitation of calcite speleothem driven by CO₂ degassing may cause the HCO₃⁻ reservoir in the cave drip water to undergo ¹³C enrichment if C isotope exchange reactions between dissolved inorganic carbon and CO₂(g) in the cave are relatively low. The enrichment in ¹³C is a result of the low $δ$¹³C value of the CO₂ that is lost from cave drip water during degassing (Mickler et al. 2006). Similarly, Power et al. (2014b) stated that Mg-carbonate sediments also had positive $δ$¹³C values (ranging from 2.0 to 8.0‰ VPDB) as a result of CO₂ degassing.

The positive $δ$¹³C values for the bulbous carbonates suggest that the only carbon source is from the atmosphere as there is no primary carbonate in the bedrock, and if there was an organic carbon source the isotopic composition would have been much more negative. This suggests that atmospheric CO₂ is being stored in these carbonate deposits; however, it is extremely difficult to quantify as its difficult to estimate the total amount of calcite inside the rock fractures.

In situ tufas that were sampled from Goat Crag and Mungasdale shared similar negative $δ$¹³C values that ranged from −5.5 to −5.0‰ VPDB whereas ex situ tufa from Bad an Sgalaig had a C isotope composition of −1.0‰ VPDB. Carbonate crusts, mud and encrusted moss had $δ$¹³C values that ranged from −3.5 to −1.8‰ VPDB, except for one crust sample that had a $δ$¹³C value of −7.3‰ VPDB. Within these samples, the carbon isotope composition could have been affected by three of the possibilities previously outlined: (1) mixed carbon sources; (2) CO₂ degassing; (3) proximity to source waters.

The in situ tufa, carbonate moss and carbonate mud were all found within close proximity to the ground (0.9–1.5 m) whereas the ex situ tufa sample was found surrounded by various types of grass and moss, which would have a negative carbon isotope signature relative to an atmospheric source (Boutton 1991). For example, Garnett et al. (2004) measured the stable carbon and oxygen isotope composition of British Holocene tufas and recorded negative $δ$¹³C values between −8.5 and −10.5‰ at the top of the tufa deposit. These values suggest an input of isotopically light soil-zone CO₂ while also being mixed with isotopically heavier carbon derived from limestone aquifer dissolution and equilibration of the aquifer and spring water with atmospheric CO₂ (Garnett et al. 2004). In this study there was no evidence of nearby limestone in the field or on geological maps (Fig. 1), which suggests that the tufa-like carbonates would have formed via bedrock weathering and a mix of isotopically heavier atmospheric carbon and lighter organic carbon from decomposing organic matter nearby.

The distance between the tufa-like deposit and the initial water source (meteoric) has an important control on the $δ$¹³C values of carbonates. In this study the most probable water source is meteoric owing to the high level of precipitation this area in Scotland receives. It is possible that water with dissolved C had accumulated in a certain location to form pools and over time began to stream downhill and over the cliff face, which was observed in the field. Although the exact starting location of these pools is unknown, CO₂ degassing probably influenced the $δ$¹³C values of carbonates precipitating downstream. Andrews (2006) observed that preferential degassing of ¹²CO₂ results in higher $δ$¹³C values of dissolved inorganic carbon and tufa downstream. Within many metres from the spring, degassing results in rapid calcite precipitation and disequilibrium of the carbonate isotope system (Usdowski et al. 1979; Dandurand et al. 1982). In the present study, $δ$¹³C values are not reflective of solely organic carbon despite their proximity to organic sources. This suggests that CO₂ degassing may have increased the $δ$¹³C values downstream, resulting in an isotopic composition that reflects a mixture of both organic and atmospheric carbon.

The tufa samples have $δ$¹⁸O values ranging from −3.2 to −0.8‰ VPDB. The carbonate mud and moss samples ranged from −3.2 to −2.4‰ VPDB, which is similar to one of the in situ tufa samples. The oxygen isotope composition of these tufa carbonates is much greater than at other tufa localities, which are primarily affected by continentality (Žák et al. 2002; Makhnach et al. 2004), where signatures are affected by the local temperature and precipitation on landmasses. The bulbous carbonates found in rock fractures had $δ$¹⁸O values ranging from −4.5 to 0.5‰ and the carbonate crusts found near these samples ranged from −5.2 to −0.2‰ VPDB, which is much greater than for other carbonates found in cave environments (Clark and Lauriol 1997; Žák et al. 2004). There are four primary factors that could have an influence on the stable oxygen isotopes of these carbonates: (1) evaporation; (2) water temperature; (3) the $δ$¹⁸O value of the water that carbonates precipitate from; (4) sea spray (Andrews 2006).

The precipitation of secondary carbonate minerals on land is primarily driven by the evaporation of waters (Acero et al. 2007; Bea et al. 2012; Stubbs et al. 2022). An increase in evaporation typically results in an increase in the $δ$¹⁸O value of surface streams; however, it is more of a problem that affects the oxygen isotope composition of carbonates in semi-arid regions (e.g. Egypt; Smith et al. 2004; Horton et al. 2016) or areas with pronounced dry seasons (e.g. northwestern Australia; Ihlenfeld et al. 2003). As the water in this study (meteoric) comes into contact with the bedrock it would promote mineral dissolution and subsequently release cations needed for carbonate precipitation. Over time, the evapoconcentration of ions (e.g. Ca²⁺ and CO₃²⁻) will drive the precipitation of calcite; however, this might occur at a slower rate than in an arid or semi-arid region. In NW Scotland it is unlikely that prolonged periods of evaporation will occur, so its effect on $δ$¹⁸O may be negligible in comparison with the other identified factors.

Both water temperature and the $δ$¹⁸O value of the water that the carbonate precipitates from will affect the stable oxygen isotope composition of the final carbonate. For example, Andrews (2006) stated that under equilibrium conditions a 1°C temperature change results in a 0.2‰ change in the $δ$¹⁸O of the tufa carbonate. The temperature of the groundwater, which is determined by local climatic conditions, influences spring and stream water temperature, which is further affected by solar insolation, leading to downstream warming. In this study, the exact distance the water travelled is unknown, making it difficult to understand how much the temperature fluctuated. However, both the air and water temperature at the time of sampling were low (mean annual minimum = 4.25°C, mean annual maximum = 10.60°C; Met Office 2023), which will result in less negative $δ$¹⁸O of the carbonate given that equilibrium fractionation is greater at lower temperatures. In this scenario, it is likely that the waters had high initial $δ$¹⁸O values owing to the cooler temperatures in a temperate maritime climate.

All samples were collected on or surrounding cliff faces within 0–5 km from the Atlantic Ocean where there are frequently periods of high winds, which result in various amounts of sea spray. Andrews (2006) suggested that sea-spray contribution to meteoric waters in coastal settings will move both $δ$¹³C and $δ$¹⁸O towards less negative values. In Scotland, tufa formation on the Isle of Lismore is promoted by evaporation of sea spray (Faulkner and Crae 2022) and in the Shetland Islands travertine contaminated by sea spray results in an enrichment in ¹⁸O (Flinn and Pentecost 1995). In this study, data suggest that sea spray is not a primary contributor but in regions closer to the sea this potential avenue can not be ignored.

In this study, secondary carbonates were detected surrounding natural igneous and meta-igneous crystalline rocks that contained several fractures. Given the cool climate of NW Scotland and limited reactive surface area (e.g. compared with a crushed rock) it is not likely that the carbonation rates would be ideal to consider this as a scalable CDR technology. However, the abundance of detectable carbonates suggests that in areas where these rocks (ranging from felsic to intermediate) are fractured, atmospheric CO₂ may already be stored. Although it is not feasible to consider increasing the fracture network of these rocks to enhance CDR, it might be of interest to further try and quantify the abundance of these pre-existing carbonates that could potentially be a sink for CO₂. To accelerate CDR in Scotland, various technologies that utilize a similar rock feedstock should be considered.

Within the last few years, enhanced weathering has gained a substantial amount of interest especially in the UK, where rock is crushed and spread over agricultural land. As it is not feasible to consider extracting natural rock and crushing it down to a powder, future research should make use of pre-existing natural rock that has already been converted into a reduced grain size from industrial processes (e.g. quarry fines). Attention has recently been devoted to the use of ultramafic mine wastes as a feedstock for CO₂ sequestration (Stubbs et al. 2022; Zeyen et al. 2022; Paulo et al. 2023) as mines produce hundreds of thousands of tonnes of fine-grained rock annually. In Scotland, there are numerous quarries that produce several tonnes of fine-grained felsic, intermediate and mafic rock through their processing technology. Attempting to make use of this material would be of economic and environmental benefit to the UK.

In situ carbon mineralization may offer a promising avenue for carbon capture in Scotland, given the presence of basalt and other volcanic rocks within the country, the existing infrastructure from the oil and gas industry in the North Sea (Chapman 1976) and pre-existing transportation networks. In this process, rather than carbonation occurring under ambient conditions, CO₂ is injected deep underground where it reacts with calcium- and magnesium-bearing silicates to form stable carbonates (Sigfusson et al. 2015), preventing CO₂ from re-entering the atmosphere. Although Scotland's substantial volcanic deposits may present opportunities for CDR, several challenges and considerations must be addressed before implementation. These include the permeability and porosity of the feedstock for efficient CO₂ injection, potentially slow reaction rates, induced seismicity, economic viability and the need for thorough monitoring and verification. Further investigation of these challenges is beyond the scope of this study. However, conducting detailed geological surveys to identify optimal sites, initiating pilot projects to test and refine the technology and developing a regulatory framework to ensure safe and effective implementation are recommended to assess future opportunities.

Carbonate minerals investigated in this study have formed from the weathering of felsic, intermediate and mafic bedrock. In the sampling region, the only available source of Ca is in the bedrock and sea spray; however, $δ$¹⁸O values suggest that seawater did not have a substantial effect on the formation of these carbonates, indicating that the majority of cations are a result of silicate weathering. The stable carbon and oxygen isotope data are highly variable, with bulbous carbonates found within rock fractures showing positive (4.4–14.9‰ VPDB) $δ$¹³C values whereas other carbonates (e.g. tufas, carbonate mud and moss) have negative (−7.3‰ to −1.0) $δ$¹³C values. Five factors were previously identified as possibilities for influencing the composition of these samples. However, it appears that the most likely factors are (1) carbonates containing mixed sources of carbon (e.g. solely atmospheric or solely organic), (2) temperature, (3) CO₂ degassing and evaporation and (4) nearby waters (Fig. 12), with sea spray being an unlikely factor for these specific samples. Attempting to estimate the total CO₂ mineralization potential of these carbonates is extremely challenging owing to the variability in abundance between sites; however, this study demonstrates that secondary carbonate minerals are forming under ambient conditions from rocks traditionally ignored in CO₂ removal studies. This research offers a new understanding on the formation of various natural carbonates, while bringing hopeful new insights for CO₂ removal in NW Scotland using highly abundant albeit less reactive rocks that have previously been overlooked in the field.

We thank C. Brolly for assisting with thin section preparation, C. Wilson for providing training and assistance with XRD, and C. Slaymark for assistance with isotope analysis.

ARS: conceptualization (lead), data curation (lead), formal analysis (lead), funding acquisition (lead), investigation (lead), methodology (equal), project administration (equal), resources (lead), software (lead), supervision (equal), validation (equal), visualization (equal), writing – original draft (lead), writing – review & editing (lead); JM: investigation (supporting), methodology (supporting), supervision (lead), writing – original draft (supporting), writing – review & editing (supporting); IN: data curation (supporting), validation (supporting).

A.R.S. was funded by IAPETUS2 as part of the Natural Environment Research Council (NERC) DTP2 process.

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.

All data generated or analysed during this study are included in this published article and its supplementary information files.