In total 3 actively growing stalagmite chips (MR, PG and SS) were sampled by chipping and sawing off a small part at the top of the growth surface. PG (PG2) and SS (SS1 and SS2) refer to monitored sites “Playground” and “Skyscraper” in Kost et al. (2023). MR (MR1) refers to sample MUS (“Mushroom”) in Sliwinski et al. (2023). Active growth was confirmed by detection of the ¹⁴C bomb spike which also constrains average growth rates over the last several decades (Sliwinski et al., 2023). In addition, we examine the sample GLA, a fossil stalagmite with a hiatus and clear optical difference below and above the hiatus (“milky” and “glassy” calcite respectively) (Figure 1). XRD confirms that all samples are exclusively calcite.

The samples cover a range of different growth rates from 3 μm/yr (GLA) to 150 μm/yr (MR) (Table 1). The active samples span a range of drip rates and modeled growth seasonality, from only a single growth season (PG), to growth during both summer and winter but growth cessation during fall and spring (SS) (Kost et al., 2023). While the MR dripwater was not monitored for a full year and we cannot definitively infer its growth season, the cave ventilation in that sector is more similar to SS and it thus potentially records multiple growth seasons yearly. MR derives from the lower gallery with an active cave stream which periodically floods adjacent terraces. Additionally, active samples span a range of dripwater chemistries, from high Mg/Ca and U/Ca (PG) to lower Mg/Ca and Sr/Ca (SS) (Kost et al., 2023). Since GLA is a fossil stalagmite, dripwater chemistry and growth seasonality are not independently constrained, but its location in the cave was nearest SS.

The samples were cut in quarter sections and prepared as a double side polished 500 µm thick sections for the following set of analyses. To assess the addressed research questions a set of methods are applied to get the necessary information about the FI distribution and other crystal features affecting trace element distribution.

Transmitted light (TL) images were taken with the Keyence VHX-6000 digital microscope at ETH Zurich to reveal layering, calcite fabric including FIs (Frisia, 2015; Chiarini et al., 2017) and detrital particles. Water filled FIs appear clear and bright under the microscope, whereas empty FIs appear black due to refractive effects. The presented microscope images are scans through the whole thick section providing a focused image of all depth levels (sequential focal plane). Hence, visible features such as FIs can be deep in the sample or at the surface. This is necessary for comparison with the fluid content maps (FTIR-images) which detect absorbance through the whole section. Images taken with the analyzer using a cross polarizer show individual crystallites with different extinction angles, illustrating crystal boundaries or crystal zoning (Figure 1).

Fluorescence maps were generated with CLSM at the Scientific Center for Optical and Electron Microscopy (ScopeM) at ETH Zurich using an Olympus Fluoview 3,000. A 488 nm laser was used for excitation of the sample by detecting the fluorescence in a window of 490–555 nm. To produce a map automatic stitching of two averaged 1024 × 1024 pixels frames at 100x magnification was automatically performed by the microscope control software (CellSens). This technique is explained in more detail in Sliwinski and Stoll (2021). Fluorescence is generally attributed to the presence of fluorescent organic compounds. However, CLSM maps may also be disrupted by refractive effects in a few regions where the surface is open due to defects or voids. The CLSM images further help to distinguish between crystal defects within the 500 µm section (spotted in TL images but not visible by CLSM) and defects situated at the very surface where the laser maps are ablated. Refractive effects in the focal plane (sample surface) result in higher fluorescence, but in a characteristic fashion that makes them easy to distinguish from true fluorescence. Crystal defects seen in the TL image but not visible in the CSLM image are not situated at the top of the sample, hence should not affect LA-maps.

The fluorescence from the applied CLSM method is commonly interpreted as a first order indication of the concentration of OM in a region of the stalagmite (Sliwinski and Stoll, 2021), although the single wavelength excitation (488 nm) and a narrow detection window of 490–555 nm are sensitive to only a narrow range of compounds and therefore may not capture all OM in stalagmites (Sliwinski and Stoll, 2021; Endres et al., 2023).

Transmission FPA-FTIR spectroscopy maps were acquired at University of Bern using a Bruker Tensor II spectrometer with a globar infrared source, equipped with a Bruker Hyperion 3,000 microscope. The closed Plexiglas chamber was purged with dried air during measurements to limit the interference of environmental CO₂ and H₂O on measurements. Data was collected with a focal plane array (FPA) detector, composed of 64 × 64 liquid nitrogen cooled mercury cadmium telluride (MCT) detector elements on a square array in a wavenumber range of 900–4,000 cm⁻¹ with a resolution of 8 cm⁻¹ and an average over 64 scans. To improve the signal-to-noise-ratio, a binning of 4 was used, resulting in 16 × 16 pixel with a pixel size of 10.8 × 10.8 µm. To minimize organic contamination the samples have been rinsed with isopropanol and mounted on the sample stage in a dry condition.

FTIR records the energy specific absorption of vibrating, bending, and stretching atomic bonds caused by radiation passing through the sample. The absorbance of molecular H₂O and OH creates a characteristic broadband peak located at wavenumber 3400 cm⁻¹ (Stünitz et al., 2017). Additional absorbance at wavenumber of ∼2,900 cm⁻¹ is due to aliphatic C-H stretching in organic components with-CH₂ and-CH₃ groups. In lower wavenumbers, organic components at 1,540 to 1,640 cm⁻¹ due to aromatic C=C double bonds and H bonded to C=O, and 1,370 to 1,430 cm⁻¹ due to COO- symmetric stretching, may overlap with the asymmetric stretch vibration of the CO₃ group in the 1,400–1,500 cm⁻¹ range (Lebron and Suarez, 1996). The most commonly described organic components in stalagmites, such as lignin and its backbone, aromatics, and humic acids are all attributed to absorbances at wave numbers in the 1,650 to 1,265 cm⁻¹ range (Artz et al., 2008).

Here we present linear absorbance intensity maps from the 3,400 cm⁻¹ band, a 2,983 cm⁻¹ band from organics, and the linear absorbance ratio of the 3,400 cm⁻¹ to the 2,983 cm⁻¹ bands (Figure 2). The latter ratio is evaluated to show the ratio of absorption of organic bonds vs. H₂O or OH bonds. For the data processing of the FTIR-FPA maps, the atmospheric correction and concave rubber band correction with 64 points and four iterations was performed in OPUS^® version 8.5. A deconvolution of each spectrum was performed in the software package SpecXY (Gies et al., in preparation) to obtain the intensity of the water peak at 3,400 cm⁻¹ and reduce the impact of signal noise and possible overlapping peaks present in the spectra (Figure 2). FTIR features a linear relationship between measured linear absorbance and water/trace element content. However, the linear absorption coefficient needed for the conversion of absorbance to water content is not well defined for calcite. Thus, we conservatively report linear absorbance values for the 2,983 cm⁻¹ and 3,400 cm⁻¹ bands, and for the 3,400 cm⁻¹ band, relative distribution maps are calculated by normalizing the linear absorbance to the maximum linear absorbance of each sample. Additionally, we explore a quantitative conversion of the deconvoluted 3,400 cm⁻¹ broadband peak, employing the wavenumber specific calibration of (Paterson, 1982) to calculate quantified water maps in µg/g (ppm H₂O) similar to the approach of (Stünitz et al., 2017), using ϵi factor of 1/3 as the orientation factor for uniaxial crystals (Paterson, 1982).

The thick sections were glued on conventional microscope glass plates and mounted in the Laurin Technic S155 2-volume laser ablation system at ETH Zurich. A 193 nm ASI Resolution ArF excimer laser ablates in a 370 mL/min high purity He and 5.0 mL/min N₂ atmosphere. Ar (1.0 L/min) carries the sample mixture to the Agilent 8,800 Triple Quadrupole ICP-MS where it is ionized and analyzed in single quadrupole mode without reaction cell or subsequent mass filter. To obtain chemical maps, parallel lines were ablated using a 20 × 20 µm square spot with a scan speed of 20 μm s⁻¹ at an ablation rate of 10 Hz and ablation energy of 4 J cm⁻². The sampling depth is less than 10 µm (Sliwinski and Stoll, 2021). For each laser track a precleaning ablation was performed to reduce contamination. NIST- 612 was used for standardization which was performed on the Iolite 4.0 software (Paton et al., 2011). Since NIST-612 is a synthetic glass, additional reference materials (pressed carbonate powders by µ-standards: BAM RS-2-NP; ECRM-752-1*-NP; JCt-1*-NP) are used in a quality check analysis (a typical example is shown in Supplementary Figure S1). Due to the matrix mismatch of the NIST-612 standard and the sample (glass vs. calcite respectively) and different concentration ranges for some elements (e.g., Na is orders of magnitude higher in NIST-612 compared to our samples) the absolute concentrations may be biased by up to 20% as examined by the quality check analysis (Supplementary Figure S1). Nevertheless, qualitative interpretation of differences within and between samples is reasonable since offsets (if any) would be constant. We report element concentrations and additionally calculate trace element to Ca ratios normalizing to 40% Ca in the measured solid. We do not show data for any point in which Ca counts per second (CPS) were <8,00,000 to avoid mapping open voids, pores and fractures. For subsequent evaluation of trace elements controlled by processes other than the presence of detrital clays, we apply a filter with a cutoff of >6 ppm Al (Supplementary Figure S1).

The presented maps were aligned manually by visual fitting using Adobe Illustrator 2023. Prominent features visible in all produced images (cracks, edges, pores, layers, etc.) were used to reference the different images.

The IR absorbance is integrated over the whole 500 µm thick section, and therefore it does not give any information about the depth where the water is situated. While this is consistent with the detection of FIs by transmitted light microscopy (where sequential focal planes are added together), it may be at odds with CLSM and LA-ICP-MS data, which are both sampled from the uppermost ∼10 μm of the surface. One challenge is that FTIR-maps and TL images visualize FIs throughout the whole 500 µm section, however, LA-maps record <10 μm at the surface where potential FIs could be ruptured and lose fluid during sample preparation and therefore the true FI influence may be underestimated by LA scans. Within this 10 µm depth of the ablation, larger pores of dimension 10 µm or larger are more likely to intersect the surface and be exposed during polishing and potentially opened, whereas some fraction of smaller <2 µm pores are more likely to remain intact and closed within that ablation depth range. Thus, the discrepancy between LA scans and FIs estimates would be most significant for the larger FIs of similar dimension as the depth ablation crater.

Operationally, in this study FIs of <20 µm are considered; no efforts have been made to distinguish FIs from a genetic point of view (primary, secondary or pseudosecondary). This threshold is introduced as an operational limit since sample preparation opens larger pores. Both techniques employed here detect FIs (TL images) or H₂O (FTIR) throughout the whole 500 µm section (sequential focal planes). With the resolution of our TL images, we identify readily the FIs greater than about 1 µm in diameter. We differentiate between FIs and macroscopic features appearing within the calcite which we call “elongated voids”. Those are mostly not water filled but contain detrital particles and likely represent macroscopic defects in the calcite lattice and/or open FIs.

All samples contain regions with microscopic FIs visible in TL microscopy (Figures 3–7). Active samples SS1 (Figure 5), SS2 (Figure 4) and MR1 (Figure 6) show bands of higher FI density along some growth layers. Elongated FI can be as large as 20 µm in length (e.g., SS1). In all samples, the majority of visible FIs are <8 µm. In these FI-rich layers, the density of FIs is relatively high (impossible to count). Compared to the other active samples, PG2 reveals fewer and more homogeneously distributed FIs (Figure 7). Elongated voids (distinctly different than FIs and often filled with detrital particles) in PG2 are disturbing the clear appearance of calcite. GLA features clusters of inclusions between the crystallites (Figure 3).

The FTIR mapping technique identifies water in both visible microscopic as well as submicroscopic FIs (<<1 µm). Regions of high intensity of the 3,400 cm⁻¹ band (O-H bonding) also feature a high linear absorbance ratio of the 3,400 to 2,983 cm⁻¹ bands (Figure 8) as expected if spatial variations in water, rather than spatial variations in organic phases, dominated the spatial distribution of the 3,400 cm⁻¹ band. Therefore, we compare the normalized intensity maps of the 3,400 cm⁻¹ band with the distribution of FIs and here describe both as spatial variations in water content. The normalized intensity of the 3,400 cm⁻¹ band varies significantly, up to 5-fold, in different areas of a given sample (Figure 8).

In case of PG2 the more homogenous intensity of the 3,400 cm⁻¹ band confirms the observations made in TL images with a generally low FI density homogeneously distributed without clusters or bands of FIs as observed in the other samples (Figure 6). In MR1 and SS2, visible bands of high FI density coincide with bands of higher IR absorbance of the 3,400 cm⁻¹ band (normalized FI distribution in Figures 4, 6). However, even within these samples, the bands with highest FI density do not necessarily indicate highest IR absorbance of the 3,400 cm⁻¹ band (MR1 or SS2). The amplitude of variations in water content is high, with some areas featuring 10 times less water than the maximum in MR1 (Figure 6). Also, the size of the FIs does not seem to affect total IR absorbance of the 3,400 cm⁻¹ band. For example, the elongated FIs along a growth-layer parallel textural discontinuity in SS2 do not yield the highest IR absorbance of the 3,400 cm⁻¹ band in this sample; rather highest IR absorbance of the 3,400 cm⁻¹ band coincides with other zones with finer FIs (Figure 4). It is possible that large FIs lost part of their water either during sample preparation (cutting and polishing) or enclose air trapped during closure of the pore.

A slightly different presentation is seen in SS1 (and SS2), which shows patches of high IR absorbance of the 3,400 cm⁻¹ band not obviously correlated with higher visible FI density, as well as FI trails along growth layers with rather low 3,400 cm⁻¹ band absorbance (Figures 5, 6). SS1 features smaller variations with a relative water content ranging only between 30% and 100% of the maximum water density. The appearance of the FTIR image looks like a honeycomb pattern. The visible FI bands in SS1 are not controlling the IR absorbance of the 3,400 cm⁻¹ band but rather clear compact calcite with low visible FI density reveals high IR absorbance suggesting water in compact calcite (Figure 5). Consequently, the normalized intensity map of the of the 3,400 cm⁻¹ band suggests water where it is not expected from the visual FI distribution.

From both normalized intensity maps of the 3,400 cm⁻¹ band and TL, high water concentrations are observed in GLA along the hiatus, 10 times higher than the minimum (Figure 3). The elevated IR absorbance fingers above the hiatus coincide with higher visible FI density and often follow crystallite boundaries (Figure 3). The quantification of water content (Supplementary Figure S4, Supplementary Text S1.2) suggests water contents in the range of 0.05%–0.25% in most samples.

All our samples show the presence of detrital particles in TL microscopy, and these correspond with regions of very high Al content in LA-ICP-MS maps (Supplementary Figure S1). Al shows a bimodal distribution, with high Al zones reaching 10–20 ppm at maximum and concentrations in the visibly cleaner calcite closer to 1 ppm.

Detrital minerals are concentrated in the hiatus, along fractures, and in pores. The most prominent detrital contributions are recorded in GLA which represents a hiatus spanning thousands of years (Supplementary Figure S1). In GLA, detrital layers occur only along this hiatus horizon and just below it in infilled voids. In SS2 there are distinct layers of detrital enrichment marked by high Al content and detrital particles in TL; one such layer coincides with a change in calcite fabric and appears to represent a discontinuity in growth conditions (Supplementary Figure S1). MR1 features a layer of enriched Al along the fracture, and also in some layers in the lower portion of the sample which coincide with FI-rich bands. Detrital enrichment in fractures is also seen in PG2 (in growth direction and opening into pore, Supplementary Figure S1) and SS2 (fissure fracture within crystallite seen Supplementary Figure S1). In contrast, the fracture along the crystal boundary detected in SS2 or the fracture in SS1 (Supplementary Figure S1) is not very prominent in Al (no detrital contamination, hence fracture likely originates from sample preparation). Furthermore, detrital particles and Al enrichment are noted in inter- and intra-crystalline pores.

While FTIR has been employed previously on aggregate stalagmite samples or decarbonated stalagmite powder (Gázquez et al., 2012; Martínez-Pillado et al., 2020), the present study illustrates the novel application of FTIR microscopy to map the spatial distribution of water (3,400 cm⁻¹ band) and organic molecules (2,983 cm⁻¹ band) in a speleothem section.

For the detection of organics, one advantage of FTIR spectroscopy is that it is less sensitive to the topography of the polished stalagmite surface and near-surface reflection effects compared to CLSM, as observed in our sample MR1 (Figure 6). Although not exploited here, the technique could have the potential to distinguish an array of organic components through full deconvolution of the organic peaks in the in the 1,650 to 1,265 cm⁻¹ wavenumber range, as completed for analysis of peat samples (Artz et al., 2008). This contrasts with standard CLSM which typically uses a narrow excitation/emission pair.

In this study, CLSM was limited to detecting organic components excited at 488 nm and fluorescing in the window of 490–555 nm, a small fraction of the total fluorescent organic constituents delivered to speleothems in this cave(Endres et al., 2023), and potentially emphasizing components such as lignin (Sliwinski and Stoll, 2021). From FTIR, we focused on the 2,983 cm⁻¹ band of the asymmetric CH₂ component which is expected in fats, waxes, and lipids. Our initial, qualitative comparison of CLSM and the FTIR 2983 cm⁻¹ band suggests that they are generally elevated in the same regions of the speleothems, although they likely target differing classes of organic material. One explanation is that various classes of organic components are co-located in the same zones of the stalagmite, i.e., bands high in lignin are also high in lipids and waxes. Future studies in precisely co-referenced maps could quantitatively compare the intensity variations of the 2,983 cm⁻¹ band and other FTIR organic bands in the 1,650 to 1,265 cm⁻¹ wavenumber range with the CLSM fluorescence maps.

For the detection of water, one advantage of FTIR spectroscopy is that it has the potential to detect submicroscopic FIs smaller than those resolved by standard transmitted TL microscopy. Indeed, we find that through the 3,400 cm⁻¹ band, it is possible to detect zones with high density of microscopic FIs but also identify high water content in areas where no microscopic FIs were recognized (e.g., SS1; Figure 5). Additionally, the linear absorbance readily quantifies the scale of changes in water abundance in different regions of the sample.

One challenge with FTIR microscopy is that the 3,400 cm⁻¹ band detects OH which may be present as free water (e.g., in FIs), as interstitial H₂O in the crystal lattice or as OH groups bound to organic molecules (and other phases). In previous FTIR application on bulk samples, the 3,400 cm⁻¹ band is routinely attributed to water in biominerals (Engin et al., 2006) but in abiogenic minerals it is commonly attributed to OH groups in organic compounds (Gázquez et al., 2012). Here, to compare the distribution of the OH peak with organic components, we also quantified the distribution of an organic-attributed band (2,983 cm⁻¹). Because of the interference of the CO₃ group in the region of the most commonly described organic components of stalagmites in the 1,650 to 1,265 cm⁻¹ wavenumber range (Artz et al., 2008; Martínez-Pillado et al., 2020), we focused on the linear absorbance of the 2,983 cm⁻¹ band of the asymmetric CH₂ component which is expected in fats, waxes, and lipids.

As a first order qualitative test of whether sectors of the stalagmite high in 3,400 cm⁻¹ band linear absorbance were only located in areas of high abundance of organics, we mapped the ratio of the 3,400 cm⁻¹ band linear absorbance to 2,983 cm⁻¹ band linear absorbance. If all elevated 3,400 cm⁻¹ band linear absorbance were proportionally due to elevated 2,983 cm⁻¹ band, then this ratio would be constant spatially across the stalagmite. Yet, we observe that regions of high 3,400 cm⁻¹ band linear absorbance are also characterized by a high 3,400 cm⁻¹ band to 2,983 cm⁻¹ band ratio, implying increased OH stretching relative to CH₂ stretching of aliphatics. We have interpreted this as evidence that a significant portion of the 3,400 cm⁻¹ band linear absorbance signal reflects free water. This interpretation assumes that the 2,983 cm⁻¹ band linear absorbance is representative of the abundance distribution of the broad suite of organic components which could feature OH groups. In future work, such an interpretation could be more rigorously tested by a full deconvolution of the organic peaks in the 1,650 to 1,265 cm⁻¹ wavenumber range.

The formation mechanism of the pores in calcite, which may contain water, is still debated (e.g., Chiarini et al., 2017; Frisia et al., 2018). Molecular scale studies have shown growth defects deviating from the ideal spiral growth as one source of voids (De Yoreo and Vekilov, 2003). Laboratory precipitation experiments suggested higher dripwater DOM promotes more defects and pores (Chalmin et al., 2013; Green et al., 2016; Pearson et al., 2020), while trace element additives similarly introduce morphological deformations that may give rise to entrapped pores (Meldrum and Hyde, 2001; Wasylenki et al., 2005). Crystal defects could also be promoted by organic coating or impurities in the calcite lattice, for example, by clays. While it has been proposed that faster growth results in more crystal defects, Chiarini et al. (2017) suggest that more porous calcite forms at slower growth rates due to incorporation of elements inducing crystal lattice defects.

Our set of samples does not reveal a clear relationship of pores or water with average decadal-scale stalagmite growth rate defined by the bomb spike (Table 1). We do not recognize a greater number of pores or microscopic FIs in the faster growing active stalagmites (MR1, SS) compared to the slower PG2. We do observe that the more homogeneously distributed FIs occur in PG2, which precipitates calcite during a single season (Kost et al., 2023). In contrast, in SS and MR1 FIs are concentrated in certain growth layers. We do not resolve a fluid rich and fluid poor alternation within each yearly increment, as seen in some studies (Wassenburg et al., 2021). It is possible that the multiple growth seasons in SS condition the formation of inclusion rich layers during some environmental conditions. However, our data do not constrain the mechanism. Because we cannot resolve growth rate differences among individual growth layers or phases, it is possible that within-sample growth rate variations do correlate with FI density.

One important observation is an enrichment of visible and elongated FIs from TL images superimposed on condensed phases of growth as revealed, for example, by GLA or SS2 (Figures 3, 4). The higher IR absorbance at 3,400 cm⁻¹ observed along hiatuses and growth discontinuities with detrital layers (GLA, SS2) is likely due to water bound to clays. At the same time, elongated voids around the GLA hiatus, potentially triggered by detrital minerals, also coincide with visible FIs indicating discrete free water regions (Figure 3). Similarly, SS1 and SS2 do reveal a higher water content in organic rich areas (e.g., upper part in SS2; Figure 4), which seems to be consistent with the observations of Pearson et al. (2020) suggesting DOM to promote crystal defects. However, in SS1, the areas of higher water content paradoxically coincide with the most transparent, microscopic FI and void free calcite.

Maps of FTIR absorbance at 3,400 cm⁻¹ also suggest regions of higher water content in stalagmites which do not coincide with microscopic FIs visible in TL images (e.g., honeycomb pattern in SS1; Figures 5, 12). We interpret these regions must contain submicroscopic FIs as suggested by Sliwinski and Stoll (2021), Borsato et al. (2007) and Chiarini et al. (2017). Such submicroscopic FIs undetected by TL microscopy dominate the water distribution in SS1 and PG2 and potentially influence on the distribution of the most incompatible trace elements.

In PG2, dominated by dispersed submicroscopic FIs, high water content coincides with void-rich calcite (Figure 7). This pattern is evident in other samples, with the exception of SS1 (low water in elongated voids). This suggests that water remains in crystal defects or is adsorbed to the calcite surface in some crystal defects or fractures. However, in some major fractures, i.e., in SS2, the water content is low, suggesting that the stage of fracturing may affect water distribution and fractures generated during sampling or preparation do not affect water distribution. Meanwhile, some larger inter- (between crystallites) and intra-crystal pores (within crystallites) pores with visible detrital contamination do not yield higher IR absorbance in other sections (e.g., SS1, PG2; Figures 5, 7) and may have lost their water during closure or sample preparation.

The average partition coefficients for Na for some of our actively growing stalagmites (SS1 and SS2), using median Na of the laser maps (Figure 9), are higher than estimated from previous experimental calcite precipitation (Füger et al., 2019). Either the laboratory determined partition coefficients are not representative of Na incorporation into calcite in these speleothem growth settings, and/or simple partitioning from the drip water to calcite does not account for the main process hosting Na in the speleothem calcite. Because the absolute partition coefficient of Na is more than an order of magnitude lower than the other measured trace elements, Na would be the element most sensitive to contribution from microscopic or sub-microscopic FIs.

The distribution of water in stalagmites suggests that FIs could control spatial variations in Na within some samples, but not others. Within SS1 and SS2, FTIR suggests water is dominantly in submicroscopic FIs in compact calcite, in which polishing during sample preparation would impart minimal disruption of FIs at the scale of laser ablation depth (<10 μm), so these very small FIs may be more likely to be intersected by the laser path and record an appreciable trace element signature. Yet, the high water enriched zones in SS1 and SS2 do not show elevated Na. This suggests that in this sample, most of the Na is hosted in the calcite lattice or interstitially (i.e., not in the fluids). Additionally, cave monitoring suggests very limited (<10%) temporal variation in the Na concentration of an individual drip over a seasonal cycle probably due to strong buffering of concentration by cation exchange in soils (Tadros et al., 2019; Kost et al., 2023). At length scales similar to annual growth increments, and laterally along the same growth increment, there are much larger (>2-fold) spatial variations in Na concentrations in stalagmites. These larger variations do not correspond to changes in the water content in most cases, and therefore likely reflect variable effective incorporation in the calcite itself. As shown by Füger et al. (2019), Na incorporation depends on growth rate. Therefore, it is possible that changes in growth rates along a growth surface can result in such lateral variations.

In contrast, in MR1, detailed imaging shows two bands of high 3,400 cm⁻¹ band absorbance with coincident microscopic inclusions visible in TL (Figure 12). High Na spots are more frequent within these bands and especially corresponding to elongated voids in the TL image. In GLA, the only sample in which the studied region is comprised of multiple crystallites, the 3,400 cm⁻¹ band absorbance suggests concentration of water between the crystallites, although visible FIs in TL are not concentrated at these boundaries (Figure 3). Where FTIR indicates water between crystallites, high Na and U concentrations occur in these boundary areas (Figure 10). Also, in GLA, very locally-distributed clusters of high FI density with twice the water absorbance are enriched in Na by a factor of 2 at the ablation surface while other high water regions do not affect Na likely because the FIs are situated deeper in the sample than the laser ablation depth.

To reconcile these observations of diverse response in MR1 and GLA vs. SS1 and SS2, we propose that FIs do enrich the Na concentrations measured by LA-ICP-MS. However, this effect is more easily detectable in stalagmites with initially low calcite Na concentration. For instance, MR1 has at least fourfold lower Na incorporation and Na concentration than SS1 and SS2. Although partitioning cannot be calculated for the fossil stalagmite GLA, due to lack of corresponding dripwater, it features the lowest average Na concentration of all stalagmites, roughly twofold lower than MR1 and fivefold lower than SS1 and SS2 (Figure 15). In these samples, the FIs and/or the associated voids locally enhance Na.

In contrast, we propose that other factors raise the background Na partitioning into calcite in SS1 and SS2 and therefore the additional Na contribution from FIs is a negligible proportion of the total measured Na. Thus, paradoxically, FI Na is not responsible for the anomalously high Na partition coefficients in SS1 and SS2 compared to laboratory calcite experiments (Figure 15). The higher Na partitioning in the SS1 and SS2 calcite is not likely attributable to growth rate, since its growth rate is similar to MR1 (Table 1). SS1 and SS2 form under faster drip rates and feature much higher fluorescence, organic content as well as Y concentration. The causes of high Y concentration in this sample are detailed in (Sliwinski et al., 2023). The very high concentration of Y³⁺ may increase Na⁺ incorporation in order to maintain charge balance via coupled substitution (Voigt et al., 2017), potentially explaining the water-unrelated enrichment of Na in SS1 and SS2.

If the quantitative estimates of water content, in the 0.1 to 0.25 volume % range (Supplementary Figure S3,S4, Supplementary Text S1.2) are accurate, then a water-free effective Na partition coefficient two orders of magnitude lower in MR1, PG2, and GLA than in SS1 or SS2 would be consistent with the observed water content and bulk Na concentration. Assuming such partition coefficients, the spatial variation in water content in MR1 could lead to over 2 to 3-fold changes in the total (water+calcite) Na concentration of the sample. A similar magnitude spatial variation in water content would lead to only 5% and 2% spatial variations in Mg concentration in MR1 and PG2, respectively. Spatially varying water content would induce <1% variation in Sr concentrations.