Distribution of chlorine and fluorine in benthic foraminifera

Abstract. Over the last decades a suite of inorganic proxies based on foraminiferal calcite have been developed, of which some are now widely used for paleoenvironmental reconstructions. Studies of foraminiferal shell chemistry have largely focused on cations and oxyanions, while much less is known about the incorporation of anions. The halogens fluoride and chloride are conservative in the ocean, which makes them candidates for reconstructing paleoceanographic parameters. However, their potential as a paleoproxy has hardly been explored, and fundamental insight in their incorporation is required. Here we used 5 nano-scale secondary ion mass spectrometry (NanoSIMS) to investigate, for the first time, the distribution of Cl and F within shell walls of four benthic species of foraminifera. In the rotaliid species Ammonia tepida and Amphistegina lessonii Cl and F were highly heterogeneous and correlated within the shell walls, forming bands that were co-located with the banded distribution of phosphorus. In the miliolid species Sorites marginalis and Archaias angulatus the distribution of Cl and F was much more homogeneous without discernible bands. In these species Cl and P were correlated, whereas no correlation was observed 10 between Cl and F or between F and P. Additionally, their F content was about an order of magnitude higher than in the rotaliid species. The high variance in the Cl and F content in the studied foraminifera could not be attributed to environmental parameters. Based on these findings we suggest that in the rotaliid species Cl and F are predominately associated with organic linings. We further propose that in the miliolid species Cl may be incorporated as a solid solution of chlorapatite or associated with organic molecules in the calcite. The high F content together with the lack of correlation between Cl and F or P in the miliolid 15 foraminifera suggests a fundamentally different incorporation mechanism. Overall, our data clearly show that the calcification pathway employed by the studied foraminifera governs the incorporation and distribution of Cl, F, P and other elements in their calcite shells.

Because the primary ion beam was positive, calcium had to be detected as 40 Ca 16 Oand not as 40 Ca + . However, because the Ca:O stoichiometry in calcite with trace amounts of organics is fixed, we assumed that the measured distribution of 40 Ca 16 O well approximated the distribution of Ca. This assumption was supported by the good correlation between the secondary ions 40 Ca 16 Oand 16 Odetected from the calcite ( Figure A2). 31 P was measured as a tracer for organics in the calcite (Geerken et al., 2019), whereas 12 C was measured to help distinguish between resin and calcite. To ensure that the detection of 35 Cl 105 was not influenced by possible isobaric interferences such as 16 O 18 O 1 H and 34 S 1 H, we used sufficiently high mass resolution power (MRP > 5113) and additionally measured the isotope 37 Cl as well. The obtained 37 Cl/ 35 Cl ratio differed from the natural abundance ratio of 0.320 by no more than 0.015, confirming that the influence of isobaric interferences for the detection of Cl was negligible. Similarly, separation of 19 F from the possible interference by 18 O 1 H was achieved by using MRP > 2214, whereas interferences from molecules such as 12 C 7 Li, 13 C 6 Li or 16 O 1 H 3 are highly unlikely. 110 Before each analysis the area of interest was pre-sputtered for 10-15 min until secondary ion counts stabilized. Subsequently, ion count images were acquired by rastering the primary beam (dwell time of 1 ms pixel −1 ) over the sample surface using the diaphragm and slit settings listed in Table A2. The primary beam current on the sample surface ranged between 0.5-2 pA depending on the size of the imaged area. The spatial resolution ranged between 50-100 nm pixel −1 . All analyses employed an e-gun to avoid charging of the sample surface. Because some of the secondary ion counts were very low, the imaged areas 115 were measured multiple times (250-1000, depending on the sample) and the signals from the individual planes were aligned and accumulated.

Data processing
NanoSIMS data were processed with the freeware program Look@NanoSIMS (Polerecky et al., 2012). After alignment and accumulation of the image data, regions of interest (ROIs) corresponding to foraminiferal calcite were drawn by hand based 120 on the 40 Ca 16 Oimage. With the additional use of the 12 Cand 35 Climages, areas of exposed resin or pores within the foraminiferal walls were identified and excluded from the final analysis. because the secondary ion counts detected by NanoSIMS are linearly proportional to the concentration of the corresponding element, and because all the measurements were done in a similar biogenic calcite matrix using the same pre-sputtering and measuring protocol.
To ensure that the El/Ca ratios were not affected by insufficient pre-sputtering, depth profiles of the El -/ 40 Ca 16 Oion count ratios were inspected for each ROI, and the planes where the ratios showed a significant trend with depth were excluded from 130 the final analysis. Lateral profiles of the El -/ 40 Ca 16 Oion count ratios perpendicular to the shell surface were extracted from the accumulated NanoSIMS images. The width of the profile line, which corresponds to the amount of averaged lateral profiles, was set to 20 pixel to increase the signal-to-noise-ratio.
To investigate the spatial correlation of El -/ 40 Ca 16 Oion count ratios, ROIs were drawn on the NanoSIMS images in Look@NanoSIMS in such a way, that regions of higher and lower ion count ratios on the foraminifera were separated into 135 different ROIs to separate the spatial variability. That way, 40 to 47 separate data points (ROIs) per species, grouped per specimen, were extracted from the NanoSIMS images. Subsequently, correlation matrices were calculated for the accumulated ion count ratios in those ROIs using the corrplot package (Wei and Simko, 2017) in R (R Core Team, 2018).

140
The halogens Cl and F show distinct banding in the rotaliids, in particular in A.tepida ( Figure 1). Moreover, in the rotaliids, maxima of Cl/Ca and F/Ca are co-located with those of P/Ca, and correlate well in A. tepida, while the correspondence between P/Ca and F/Ca is moderate in A. lessonii, and spatially rather complex for Cl/Ca (see lateral profiles in Figure 1). Furthermore, the contrast between the high-intensity and low-intensity bands in F/Ca, Cl/Ca and P/Ca is higher in A. tepida than in A. lessonii.
In the miliolid foraminifers no banding of neither halogens nor P is visible, with the exception of a slight elevation in Cl and P 145 in areas of an image that were identified as a suture in SEM images. Lateral profiles in A. angulatus show a correlation of Cl with P, and no correlation of F with Cl or P. The lateral profile through the shell wall of S. marginalis shows similar patterns as the one of A. angulatus. The F/Ca ion count ratios in the miliolids A. angulatus and S. marginalis are in the same range and one order of magnitude higher than those in rotaliid A. tepida and A. lessonii (Figures 1 and 2). The Cl/Ca and P/Ca ion count ratios are in the same order of magnitude in all four species (Figures 1 and 2).

150
Ion count ratios of F/Ca and Cl/Ca correlate with each other in A. tepida, while there is no correlation in A. lessonii and the miliolids (Figure 2A and A3). Both F/Ca and Cl/Ca are correlated with P/Ca in the rotaliids, while only Cl/Ca is correlated with P/Ca in the miliolids ( Figure 2B,C). All correlations described here are significant to a level of p < 0.001; R 2 and p-values of the correlations are reported in Figure A3. These correlations are also seen when the elements F, Cl and P are normalized to O instead being normalized to Ca ( Figure A5).

Relation with cation incorporation and culture media properties
In all four species, the Cl content does not correlate to any of the elemental ratios measured by LA-ICP-MS (upper panels in Figure 3). However, the elevated F/Ca ratios in the miliolid foraminifera coincide with elevated Mg/Ca and Ba/Ca, which also are an order of magnitude higher in these species than in the rotaliid foraminifera (lower panels in Figure 3). Our data show no correlation of Cl/Ca or F/Ca with neither Na/Ca nor Sr/Ca (Figure 3).

160
Our data show no correlation of Cl/Ca or F/Ca with salinity or temperature ( Figure A4). Furthermore, crossplots of the NanoSIMS Cl/Ca and F/Ca ion count ratios show no correlation with carbonate system parameters for Cl/Ca (Figure 4). However, NanoSIMS F/Ca ion count ratios are higher in the miliolid foraminifera, which were cultured at higher DIC, corresponding to higher alkalinity and carbonate ion concentration as well ( Figure 4).

Limited environmental control on Cl and F incorporation into foraminiferal shells
The NanoSIMS data presented clearly show that biomineralization pathways govern the incorporation and distribution of Cl and F within foraminiferal shells: the rotaliid species show distinct banding in Cl, F and P, while the F-rich miliolid species do not. Biologic control is known to affect incorporation of most elements into foraminiferal shells, while at the same time 170 relationships with environmental conditions have proven robust tools for paleo reconstructions (Eggins et al., 2004;Kunioka et al., 2006;Paris et al., 2014;Spero et al., 2015;Fehrenbacher et al., 2017;Geerken et al., 2019). In our data set comprised of a very limited amount of specimens, we see no overall trend in Cl/Ca ratios in foraminiferal calcite with chemical properties of the culture media. The high intra-shell variability in rotaliid foraminifera and the spatial correlation with P on the location of organic linings suggest that Cl is associated with organic linings in rotaliid foraminiferal shells. Furthermore, Cl/Ca is highly 175 variable within a single section through a foraminiferal wall in all the species measured and the range of Cl/Ca ratios is similar in all investigated specimens.
The overall absence of trends with environmental conditions as well as the high intra-specimen variability lowers the confidence in potentially using Cl/Ca for paleo reconstructions. However, as this study does not cover a range of physicochemical parameters for a single species, but rather presents a collection of different species that were also grown in different condi-180 tions, we cannot exclude that any of the presented physicochemical parameters may exert an influence on the Cl content of foraminifera on a species-specific level. A definite conclusion regarding proxy applicability would require culturing studies including 20-30 specimens per species per environmental condition.
Moreover, there is no discernable trend of Cl/Ca ratios in foraminiferal calcite with any of the measured trace elements ( Figure 3). Cl/Ca ratios are in the same range for species with low-Mg calcite (rotaliid) as they are for those with high-Mg 185 calcite (miliolid), suggesting that chlorine incorporation is systematically different from the incorporation of these cations.
In inorganically precipitated calcite, chlorine contents are an order of magnitude lower than sodium contents (Kitano et al.,  1975), suggesting that Cl is incorporated neither as fluid inclusions (Wit et al., 2013) nor as solid solutions of NaCl into calcite.
As chlorine content seems not to reflect any environmental parameter, and Cl/Ca correlates well with P/Ca in all the species investigated here, we suspect that chlorine incorporation into foraminiferal calcite is closely related to organic molecules 190 involved in calcification or to a solid-solution between calcite and chlorapatite (Ca 5 (PO 4 ) 3 (OH,Cl)).
The F/Ca ratios in the miliolid species are about an order of magnitude higher than those in the rotaliid species. The elevated F/Ca ratios in miliolids coincide with overall higher CO 2 -3 -ion concentration in the culture media of the miliolid species. This might indicate a relation between foraminiferal F/Ca ratios and carbonate ions, but the relationship is inverse to what one would expect for inorganic ion exchange (Ichikuni, 1979) and what has been observed in corals (Tanaka et al., 2013). However, 195 the high intra-shell variability in F/Ca ratios of single specimens and the co-variation of environmental conditions with mineralization pathway complicates attributing F/Ca ratios to environmental parameters. Species-specific culturing studies could provide more insight into whether F/Ca ratios of benthic foraminifera on a species-specific level correlate with environmental conditions, as then the effect of different biomineralization pathways would not hamper interpretation as is the case in our data set.
200 Opdyke et al. (1993) suggested that the presence of photosynthetic symbionts in foraminifera impacts their F/Ca ratio.
Symbionts influence the intracellular carbonate chemistry by photosynthesis, which could link to fluoride incorporation via the intracellular CO 2 -3 ion activity. A. tepida is the only symbiont-barren species we investigated, and indeed its F/Ca ratios are lower than in the miliolid symbiont-bearing species. However, the symbiont-bearing rotaliid A. lessonii exhibits the lowest mean F/Ca ratios, which are in the same order of magnitude as in A. tepida, but less variable throughout the shell wall. We 205 therefore conclude that F/Ca ratio is unlikely directly related to the presence of symbionts in foraminifera.
Notably, F/Ca ratios are higher in specimens with a higher Mg and Ba content (Figure 3). A correlation of trace element content with Mg content within and between species has been found for several elements, including Sr, Zn, and Na (Evans et al., 2015;van Dijk et al., 2017;Geerken et al., 2018), and also F (Opdyke et al., 1993). The miliolid species have generally a much higher Mg content than the rotaliid species and their biomineralization mechanisms are thought to be substantially 210 different ( Figure 5). The fact that higher F content corresponds with higher Mg content may point towards a strong biological control on F incorporation. Fluorine may be incorporated in solid solutions. Fluorite (CaF 2 ) solid solution has been suggested as the incorporation mechanism for fluorine in calcite (Carpenter, 1969), but also fluorapatite (Ca 5 (PO 4 ) 3 (OH,F)) solid solution would be a possible option.

Cl and F incorporation in foraminifera is primarily controlled by biomineralisation pathway 215
In the two species of rotaliid foraminifera that were investigated here, Cl and F show strong banding. The Cl and F bands are co-located with P in the foraminiferal shell walls. Since phosphorus is present in organic molecules like phospholipids in membranes, P/Ca can be used to image organic linings in between the lamella in hyaline foraminiferal shell walls, as demonstrated in Geerken et al. (2019). In A. tepida, the correlation of Cl and F with P is tight, and we conclude that both elements are primarily associated with the organic linings. In A. lessonii, the peaks in Cl/Ca and F/Ca lateral profiles are also 220 co-located with peaks in P/Ca. However, in the specimens we analyzed, there seems to be substantial additional Cl and F also in some calcite lamella, as can be seen in the lateral profiles. Moreover, the contrast between high-intensity bands and lowintensity bands is less prominent in A. lessonii than in A. tepida. We suggest that also in A. lessonii, association with organic linings is the primary mode of incorporation of both elements in the foraminiferal shells. Using NanoSIMS and the very same species and specimens, Geerken et al. (2019) reported co-occurrence of organic linings and banding of Mg, Na, Sr, K, S, P and 225 N. Moreover, they showed that elemental incorporation in A. lessonii was overall higher than in A. tepida, consistent with our observations for the halogens (Figures 1, 2).
In the miliolids, the distributions of Cl and F are distinctly different from those of the rotaliids: since miliolids do not calicify by adding subsequent lamella of calcite ( Figure 5), no patterns of alternating high and low concentration banding are visible. Moreover, Cl and F are spatially not correlated throughout the shell walls of these miliolids, indicating that Cl and 230 F may have different modes of incorporation. The correlation of Cl and P within the shell wall supports the hypothesis that Cl is incorporated in the calcite as a solid solution of chlorapatite, a mineral containing both Cl and P, or associated with organic molecules as for the rotaliids, but then distributed in a less organized way (no banding) within the calcite. The type of organics being present in foraminiferal calcite is determined by the precipitation pathway: rotaliids initiate calcification around a primary organic sheet (POS) and cover their shell with organic linings, while miliolid shells comprise of randomly oriented 235 calcite needles held together in an organic matrix. As these organics may differ in their P content, it is possible to measure comparable P contents in both rotaliid and miliolid calcite, even though their absolute organic matter content within the calcite is expected to be different. Similar Cl content in both rotaliids and miliods may thus be due to differences in the composition of the organics or may hint to an incorporation via a different pathway. Alternatively, apatite may form via the adsorption of phosphate to calcite and amorphous calcium phosphates at low Mg concentrations in solution (Martens and Harriss, 1970;240 Millero et al., 2001). The incorporation of Cl via chlorapatite could also explain the spatial correlation of Cl and P in the species we analysed. If F would be incorporated similarly as Cl, we would expect the F and Cl distribution to be comparable. As this is not the case, we conclude that F is incorporated primarily in a different way than Cl, e.g. as a solid solution of fluorite, as suggested before (Rosenthal and Boyle, 1993).
The observed difference between F-rich miliolids lacking organic linings and clear banding of trace elements on the one 245 hand, and the rotaliids with F, Cl and P rich bands on the other hand, is consistent with known differences in calcification mechanisms ( Figure 5) that have developed during the evolution of foraminifera (Debenay et al., 1996;Bentov and Erez, 2006).
Hyaline (including the rotaliid) foraminifera precipitate calcite onto organic templates within an extracellular but confined space, and add a new lamella to the entire shell each time they produce a new chamber (Hemleben et al., 1986;de Nooijer et al., 2014). For some intermediate Mg-calcite producing foraminifera (like A. lobifera or also A. lessonii), transport of precipitate calcium crystals in intracellular vesicles prior to arranging them in the shape of the new chamber wall (Angell, 1980;Hemleben et al., 1986;Debenay et al., 1998Debenay et al., , 2000Bentov and Erez, 2006). Furthermore, there are species within the suborder of the miliolids that show features of both biomineralization pathways, such as Archaias angulatus, which appears to precipitate calcite at the site of the new chamber wall, opposed to other miliolid species (Wetmore, 1999). and F support this.

Conclusions
Here we investigated for the first time the spatial distribution of the halogens Cl and F in foraminiferal shell walls. In the rotaliid benthic species Ammonia tepida and Amphistegina lessonii, Cl and F are distributed in bands within the chamber walls, which co-locate with P banding. In the miliolid benthic species Sorites marginalis and Archaias angulatus Cl and F were not 270 found to occur in bands. However, the rather homogeneously-distributed Cl was found to correlate with P content, while F did not correlate with either P or Cl. Based on these findings we suggest that Cl and F are predominately associated with organic linings in the rotaliid species. We further propose that Cl may be incorporated in miliolid species as a solid solution of chlorapatite or be associated with organics. Our data in the miliolid species suggests that F is incorporated in a different way than Cl, as F does not correlate with P.       The error bars depict one standard error of the mean NanoSIMS ion count ratios where more than one image was analysed per specimen.    Figure A5. Illustration to show that the patterns in figure 2 do not change much when normalizing to 16