Abstract Marine shells incorporate oxygen isotope signatures during growth, creating valuable records of seawater temperature and marine oxygen isotopic compositions. Secondary ion mass spectrometry (SIMS) measures these compositions in situ at finer length-scales than traditional stable isotope analyses. However, determining oxygen isotope ratios in aragonite, the most common shell mineral, is hampered by a lack of ideal reference materials, limiting the accuracy of SIMS-based seawater temperature reconstructions. Here, we tested the capability of SIMS to produce seawater temperature reconstructions despite the matrix calibration challenges associated with aragonite. We cultured Anadara trapezia bivalves at four controlled seawater temperatures (13–28 °C) and used strontium labeling to mark the start of the temperature-controlled shell increment, allowing for more spatially precise SIMS analysis. An improved matrix calibration was developed to ensure more accurate bio-aragonite analyses that addressed matrix differences between the pure abiotic reference materials and the bio-aragonite samples with intricate mineral-organic architectures and distinct minor and trace element compositions. We regressed SIMS-IRMS biases of abiotic and biogenic aragonites that account for their systematic differences in major, minor, and trace elements, allowing for more accurate SIMS analyses of the temperature-controlled shell increment. The thorough matrix calibration allowed us to provide a SIMS-based seawater-corrected oxygen isotope thermometer of T (°C) = 23.05 ± 0.36 - 4.48 · (δ18Oaragonite [‰ VPDB] - δ18Oseawater [‰ VSMOW] ± 0.25) and 103lnαaragonite-water = (17.78 ± 0.88) · 103/T (K) - (29.44 ± 2.40) that agrees with existing aragonitic IRMS-based thermometer relationships and improves the applicability of SIMS-based paleo-environmental reconstructions of marine bio-aragonites.Plain Language Summary:In this study, we grew marine bivalves under tightly constrained aquaculture conditions at four different seawater temperatures and marked the start of the growth period in the shell structure using strontium labelling. The newly grown shell material between the strontium-labeled increment and the shell edge was analyzed for its oxygen isotopic composition. The compositions were measured in-situ, using a high resolution ion microprobe and a newly developed analytical post-processing strategy specifically designed for biomineral samples with mineral-organic architectures. The strategy involved two reference materials and the major, minor, and trace element content in the shell and the reference materials. The new approach resulted in an accurate and robust model for determining past seawater temperatures from fossil or historic shells based on their oxygen isotope composition at over an order of magnitude finer length scales than traditional oxygen isotope analyses. 1 Introduction:Marine calcifying organisms, including bivalve mollusks, grow their skeletal hard parts (i.e., shells) from calcium carbonate, the two most common polymorphs being calcite and aragonite. These occur together with organic phases in different hierarchical architectures (e.g., Boggild, 1930). During growth, shells incorporate minor and trace elements as well as isotopes from the marine environment that can be used for paleoclimate or paleoenvironmental reconstructions. Stable oxygen isotope ratios are commonly utilized as a sea surface temperature (SST) proxy from which other parameters, including seasonality and growth rates, are inferred. However, deciphering past environmental conditions encoded in the 18O/16O ratios (expressed as δ18O) of growing shells is complicated by a combination of different signals that are recorded in addition to ambient seawater oxygen isotopes (δ18Osw) (e.g., Emiliani, 1966; Shackleton, 1967) and the SSTs experienced during shell growth (e.g., Epstein et al., 1953). These additional signals include vital effects involved during shell biomineralization (Pérez-Huerta & Andrus, 2010; Urey, 1947; Weiner & Dove, 2003) and, in the case of fossil shells, post-depositional modification by surrounding pore water (Adams et al., 2023; Cisneros-Lazaro et al., 2022). Hence, δ18O paleothermometer calibrations have been developed using different approaches, from inorganic precipitation experiments (e.g., Kim & O'Neil, 1997; McCrea, 1950; O'Neil et al., 1969; Tarutani et al., 1969; Zhou & Zheng, 2003), through field-based approaches using live or recently live specimens (e.g., Aharon, 1991; Böhm et al., 2000; Carré et al., 2005; Chamberlayne et al., 2021; Grossman & Ku, 1986; Horibe, 1972; Rahimpour-Bonab et al., 1997) to highly-constrained species-specific aquaculture experiments (Al‐Qattan et al., 2023; Owen et al., 2008; Wanamaker et al., 2006).The desire for finer-scaled paleoenvironmental reconstructions of δ18O that are more contextualized with the biomineral architecture is driving in-situ Secondary Ion Mass Spectrometry (SIMS) instrumental developments (e.g., Green et al., 2022; He et al., 2021; Rollion-Bard et al., 2003; Vetter et al., 2013). Although these applications are becoming routine for some biomineral systems, quantitative applications and matrix calibrations of aragonite biominerals are still being developed (He et al., 2021; Long et al., 2020; Rollion‐Bard et al., 2007).Traditionally, instrumental and matrix isotope fractionation are corrected by bracketing the unknowns between microanalytical reference materials with matching matrices. Hence, aragonite applications rely on abundantly available, well characterized, and homogeneous microanalytical δ18O reference materials present as crystals or crystal fragments. However, there are currently no certified aragonite reference materials meeting these criteria. As a work-around, aragonite samples have been calibrated using in-house reference materials lacking international availability (e.g., Rollion-Bard et al., 2003; Rollion-Bard & Marin-Carbonne, 2011; Rollion‐Bard et al., 2007; Stern, 2024b) and newly developed potential reference materials (e.g., He et al., 2021). To our knowledge, no unpowdered, microanalytical aragonite reference material of biogenic origin exists that would allow for a more ideal matrix match with biogenic marine shells. The general lack of aragonite reference materials for in-situ analyses arises from the material being notoriously heterogeneous with respect to stable isotope signatures in both biogenic and geological environments. For instance, deep-sea corals reveal heterogeneous δ18O patterns despite the stable seawater temperature environment, due to biologically-mediated disequilibrium effects (e.g., Rollion-Bard et al., 2003; Saenger et al., 2017). The lack of in-situ aragonite reference materials leaves two options for improving SIMS matrix calibrations for biogenic aragonite: Develop new homogeneous reference materials, and refining δ18O calibration strategies for existing materials. This study focusses on the latter.This study aimed to refine δ18O calibrations of biogenic aragonite shells (in the following referred to as δ18OArg) by presenting a new procedure using the combined Na, Mg, Ca, and Sr abundances of abiotic aragonite reference materials to improve the accuracy of SIMS measurements of δ18O in biogenic aragonites. This approach allowed for a proof-of-concept study, establishing the first SIMS-based seawater thermometer relationship for the Sydney Cockle Anadara trapezia (Deshayes, 1839). We chose this intertidal, semi-endobenthic species for its wide range of seawater temperature tolerance, with populations adapted to living in seawater temperatures ranging from 9 °C in South Australia to 32 °C in Queensland (e.g., Kenny, 1974; Rochford, 1966). Although currently only distributed along the East coast of Australia and an additional, isolated occurrence in Western Australia, A. trapezia had a significantly wider distribution in the geological past, from New Zealand to Western Australia (Murray-Wallace et al., 2000; Pan et al., 2021; Ryan et al., 2021). Here, we cultured A. trapezia under controlled seawater conditions at 13, 18, 23, and 28 °C to quantify the influence of seawater temperature on shell δ18OArg, while carefully monitoring salinity, carbonate system parameters, seawater δ18O (in the following referred to as δ18OSW), and cation chemistry. The shell material grown under temperature-controlled conditions serves as in-house calibration material for the SIMS-based thermometer for A. trapezia. Strontium pulse-chase labelling (see Otter et al., 2019; Otter et al., 2023) provided precise control over SIMS spot locations, ensuring that only shell increments grown during temperature-controlled aquaculture were used for the seawater thermometer. The resulting thermometer, produced with 25 μm SIMS analyses, agrees well with previous biogenic aragonite relationships obtained by conventional stable isotope methods, validating SIMS for higher resolution SST reconstructions. 2 Materials and methods:2.1 Aquaculture experiments with living bivalves:We collected a total of 94 live A. trapezia specimens along a 50 km section of the New South Wales coastline (Figure S1) and distributed them evenly between four 220 L aquaria at the Australian National University (ANU). Each aquarium was stocked with three sand-filled polyethylene boxes, allowing the bivalves to burrow and for easier handling by moving the boxes instead of the bivalves to avoid handling stress (Otter et al., 2019). After the bivalves acclimated to 22 °C, seawater temperatures were adjusted stepwise at a rate of 1 °C·day-1 using combinations of heaters and chillers until the seawater temperatures reached 13, 18, 23, and 28 °C, respectively. Seawater temperature measurements were recorded at 15-minute intervals with a precision of 0.01 °C using TinyTag (Gemini, United Kingdom) data loggers. The chosen temperature range was inspired by the natural environment of the animal, which had a 20.2 °C annual mean SST in the previous year (see Table S1), as well as a maximum and minimum of 26.0 and 13.8 °C, respectively (IMOS, 2022).After reaching target seawater temperatures, we performed Sr pulse-chase labeling to mark the start of the experimental growth period in the growing shell. Pulse-chase labelling of marine calcifying organisms with trace elements or isotopes is an efficient strategy to chemically mark shell increments within growing calcified hard tissues (e.g., Gorzelak et al., 2011; Houlbreque et al., 2009; Otter et al., 2019; Otter et al., 2023; Winter et al., 2023). The Sr concentration in the seawater was raised to 18x the normal concentration (8 µg·g-1 at 35 psu salinity) for 72 h by dissolving 96.6 g of SrCl2·6H2O in 1L of pure water before adding it to the 220 L aquaria. Direct dissolution into the seawater was avoided as this can lead to the immobilization of excess Sr2+ via the precipitation of SrSO4, which has a low solubility and is resistant to re-dissolution. After 72 h, shells were returned to normal seawater and grown for 80 days. We minimized Sr exposure to the shortest duration and lowest concentration effective in previous studies (Otter et al., 2019; Otter et al., 2023) to avoid adverse effects on bivalves observed with higher concentrations and longer exposures (e.g., Cheng et al., 2024; Fujikura et al., 2003).Natural seawater sourced from the sample collection site was used throughout the experiment. The aquaria seawater was continuously cleaned and aerated by protein skimmers (Reef Octopus) and replaced weekly or when appropriate. Float valves maintained stable salinity levels by automatically adjusting dropping seawater levels by adding pure water to counteract evaporation. Seawater temperatures were recorded throughout the 80-day growth period (Table S2). Salinity measurements were logged (YSI, USA) every 3 days (Table S3). The bivalves were fed (Shellfish Diet 1800, Reed mariculture Inc, USA) once every three days for the first month of the experiment (including acclimatization). This rate was later increased to every other day and, finally, to daily for the last three weeks of the experiment. Protein skimmers were turned off for 4–6 h after each feeding. After the 80-day growth period, specimens were frozen, soft tissues removed, the shells rinsed in pure water and air-dried.2.2 Seawater characterization:Seawater samples for pH and alkalinity, aragonite saturation states, dissolved cation concentrations, and δ18OSW were collected every three days from all four aquaria (Tables S4 to S7). Aliquots for pH and alkalinity were filtered (0.45 μm) and analyzed immediately on a USB4000 Fiber Optic Spectrometer (Ocean Optics, Australia) at ANU. The pH values were obtained after pretreatment with 50 μL of meta-cresol purple and reported on the total pH scale (pHT). Alkalinities were obtained by adding Bromophenol Blue (BPB) dye to a 50 mL flask until the solution reached the endpoint, added to the test cell, and analyzed against undyed seawater. This approach has an analytical precision of ± 1.5 μmol·kg-1 for certified reference solutions (Nand & Ellwood, 2018).Aliquots for cation concentrations were acidified alongside Milli-Q blanks with 4 % HNO3 and stored at 4 °C. Seawater cation concentrations were analyzed using an Agilent Technologies 5110 Inductively Coupled Plasma-Optical Emission Spectrometer (ICP-OES) at ANU. Samples were diluted 50x with 2 % HNO3 and the following spectral lines were chosen: Na330.237, Na568.263, Na568.821, Na588.995, Na589.592, Mg202.582, Mg279.078, Mg280.270, Mg285.213, Mg383.829, K404.721, K693.876, K766.491, K769.897, Ca370.602, Ca422.673, Sr407.771, Sr421.552, and Sr460.733 (Table S6). Each sample was bracketed between an in-house seawater reference solution of a known concentration collected on an RV Investigator voyage across the Southern Ocean (IN2020_v08 at 57.977 S, 141.541 E at 10 m depth).For δ18Osw measurements, 2 mL glass vials were filled with seawater until overflowing to avoid entrapment of atmospheric air and stored at 4 °C. We used a Picarro L2140-i Cavity Ring-Down Spectrometer (CRDS) at ANU for stable oxygen isotope analyses in high-precision mode set to 7 injections per sample with dry air as the carrier gas. The first 3 injections were used to flush the system and were thus rejected in the final data. In addition, any injections where the median H2O concentration was outside the ± 1,500 µg·g-1 range around the run median (ca. 20,000 µg·g-1) were also rejected. Samples were calibrated against a set of commercially available waters used as in-house reference solutions: Fiji (-6.60 ± 0.02 ‰), Smart (-2.69 ± 0.08 ‰), and Kona (0.05 ± 0.03 ‰). These reference solutions were calibrated relative to international reference solutions IAEA SMOW2 and IAEA SLAP2 and reported in ‰ relative to the Vienna Standard Mean Ocean Water (‰ VSMOW). The precision of the δ18Osw measurements was monitored via repeat analysis of the in-house reference solutions (Table S7).2.5 Bivalve shell characterization:The left shell valve of each specimen was archived, while right valves were cut twice along the dorso-ventral axis of maximum growth (Figure 1a, orange line) using an IsoMet low-speed saw (Buehler, IL-USA). The resulting 3 mm thin shell slices were further shortened by cutting off the tip of the shell at about 7 mm from the ventral margin (Figure 1a, orange circle) using a Dremel power tool (Dremel, Australia). This produced cross-sections of the shell tips that included the most recently grown shell increments of the outer shell layer, which were then embedded together in epoxy resin mounts. Mounts were lapped and polished following the procedure in Otter et al. (2019) using colloidal silica (0.04 μm, 1 minute) as the final polishing step.A confocal micro-Raman spectrometer (Horiba Jobin Yvon LabRAM HR Evolution) at Macquarie University was used to verify the aragonitic composition of the shells. Spot measurements were acquired across all shell layers from a polished shell cross-section using a 633 nm laser at 10 mW. The acquisition time was 10 seconds per spectral window and 2 accumulations were averaged to eliminate spikes and reduce background noise. A confocal pinhole setting of 100, a grating with 1800 lines per mm, and a 50x objective lens ensured high-resolution and confocal measurement conditions. Data processing involved normalizing each spectrum to its highest intensity and plotting in Origin Pro. We analyzed a line of spots (n = 31) across a polished shell cross-section perpendicular to the inner shell surface to probe the composition of all three shell layers: The outer crossed-lamellar shell layer (Figure 1b, red spectrum, n = 26) used for SIMS microanalyses, the inner shell layer (Figure 1b, blue spectrum, n=4), and 1 spot was situated on the thin myostracum (Figure 1b, green spectrum) that separates the outer and inner shell layers. In addition, Raman spectra were obtained for the microanalytical reference materials S0436 (Figure 1b, dark gray spectrum, n = 3) and VS001/1-A (Figure 1b, light gray spectrum, n = 1). All spectra are shown as averages (solid lines) with shaded envelopes representing the first standard deviation of the normalized spectra, except for VS001/1-A and the A. trapezia myostracum, for which only individual spectra were acquired (see He et al., 2021 for more Raman analyses of VS001/1-A).Shell growth was evaluated by pre-screening cross-sections of 42 randomly chosen shells across all cultures using BSE imaging during a first BSE session. We visualized the 72-hour Sr pulse-chase labeling event (Figure 1c) and measured its distance to the inner shell edge to quantify growth during the temperature-controlled experiment (Table S8). The Sr labeled shell increment is visible due to the increased Z-contrast caused by higher Sr mass fractions in the shell, whereas the unlabeled shell portions on either side of the label appear as darker greyscale intensities as these shell portions were grown in seawater with normal marine Sr concentrations. BSE imaging also revealed features of the crossed-lamellar shell architecture, including the alternating pattern of crystallographic orientation changes of the first-order lamellae (Agbaje et al., 2017; Böhm et al., 2016).We quantified major, minor, and trace element mass fractions for three specimens for each temperature group via Wavelength Dispersive (WDS) X-ray Spectroscopy using a JEOL JXA-8530FPlus Field Emission Gun-Electron Probe Micro Analyzer (FEG-EPMA) at ANU. Analyses were performed using an acceleration voltage of 15 keV, a beam current of 10 nA, and a spot size of 20 μm (Table S9) to match those of the ANU SIMS (SHRIMP-SI). The full EPMA calibration procedure is listed in Table S10. Based on the BSE images and shell growth measurements (Table S8), we selected the shells that had grown at least 25 µm, combined them into new mounts, and removed the carbon coating.2.6 Traditional stable isotope analyses by IRMS:We used a micromill to collect powdered shell samples that were analyzed by IRMS. The micromilled shell transects, sampled across shell increments grown in the wild, were collected from two specimens and were milled next to previously acquired SIMS and EPMA spots to support a co-located comparison between both bulk-conventional and in-situ d18O analyses. The transect on shell At-13-17R was milled using a 250 µm drill-bit and analyzed in a ThermoFisher Scientific DELTA mass spectrometer, while that of At-28-22R was milled using a 300 µm drill bit and was mostly analyzed on a MAT253 (Table S11). Both mass spectrometers were coupled to Kiel IV devices for calcium carbonate digestion. The shell samples were bracketed between three carbonate certified microanalytical reference materials NBS18, IAEA-603, and NBS19 (Table 1) to normalize the d18O to the Vienna Pee Dee Belemnite (VPDB) scale. For analysis in the MAT253 we used NBS18 and IAEA-603 for the normalization, and NBS19 was used for quality control (d18O = -2.20 ± 0.06 ‰ VPDB, n = 4). For analysis in the DELTA we used NBS18 and NBS19 for the normalization, and IAEA-603 was used for quality control (d18O =-2.40 ± 0.08 ‰ VPDB, n = 6). All d18O values for aragonite samples were further corrected for acid fractionation effects by subtracting 0.39 ‰ from the d18O value of each sample, after normalization to the VPDB scale. This value (0.39 ‰) was calculated by using the procedure described in Kim et al., 2015, and the following oxygen-isotope acid fractionation factors:1.00856 for calcite (Kim et al., 2007) and 1.00895 for aragonite, which was derived using the preferred temperature relationship for aragonite tabled in Kim et al. (2015) that builds on data presented in Kim et al. (2007). Both oxygen isotope acid fractionation factors are specific to the temperature during carbonate digestion, which was 75 °C for both mass spectrometer systems. The data were plotted using Origin Pro. 2.7 In-situ stable isotope analysis by ion microprobe:The polished epoxy mounts containing shell sections and the microanalytical calibration materials were cleaned in petroleum ether, RBS35 detergent solution, and Millipore water before vacuum drying for at least 12 hours at 60 °C and coated with a ca 10 nm layer of aluminum. This cleaning procedure removed all organic, inorganic, and particulate contaminants (e.g., residues from polishing and cleaning compounds), ensuring that the conductive coating adhered well and preventing loose particles from compromising the coating, secondary ion extraction, and vacuum system. Oxygen isotopic compositions were measured in-situ over five consecutive days using the Sensitive High Resolution Ion Microprobe-Stable Isotope (SHRIMP-SI) at ANU. Analyses were carried out in multi-collector mode using a 15 keV 133Cs+ primary beam. The primary beam was focused through a 200 μm aperture to achieve a final spot size of 25 μm. Excess charge at the sample surface was neutralized using a 1.5 keV focused electron beam. A mass resolution of 3,000 at 1 % peak height allowed for the interference-free detection of 16O- and 18O- with Faraday cups. The electrometers were operated in resistor mode (1011 and 1012 Ω, respectively). Each SIMS analysis took ~6 min (1 min for pre-sputter and baseline measurements, 3 min tuning, and 2 min data acquisition). For more details on SHRIMP-SI see Ávila et al. (2020). All data points were corrected for electrometer baselines and electron-induced secondary ion emission (EISIE: Ickert et al., 2008) and are expressed in per mil relative to VPDB.We used four reference materials for the in-situ stable isotope analyses, two calcites, S0161 and NBS19, as well as two aragonites, VS001/1-A and S0436 (Table 1). Calcite S0161 was used as primary calibration material, NBS19 was used for quality control, and the aragonites VS001/1-A and S0436 were used to calibrate residual matrix effects. We chose S0161 as the primary reference material as it is more homogenous in d18O than either of the aragonite reference materials, which are less suited for calibrating primary instrumental mass fractionation. The internal precision resulting from averaging 6 individual measurements for each SIMS spot, at a 95 % confidence level, was ± 0.15 and ± 0.16 ‰ for NBS19 and S0161, as well as ± 0.14 ‰ for both aragonite reference materials, respectively. The external repeatability that quantifies the analytical variance from repeated measurements on homogeneous calibration materials (Rollion-Bard et al., 2003; Rollion‐Bard et al., 2007) was ± 0.20 and ± 0.29 ‰ for NBS19 and S0161 and ± 0.37 ‰ for both aragonite reference materials, respectively.Table 1. List of microanalytical, carbonate reference materials used for stable isotope analyses. The materials used for SIMS analyses were NBS19, S0161 VS001/1-A, and S0436, while NBS18, NBS19, and IAEA-603 were used for calibrating the IRMS analyses. Reference material: Phase composition: Source: Assigned d18O [‰ VPDB]: Assigned d13C [‰ VPDB]: Reference: NBS18a Calcite Carbonatite -23.01 ± 0.10 -5.01 ± 0.04 Friedman et al. (1982) NBS19b Calcite Marble -2.20c +1.95” Friedman et al. (1982) IAEA-603 Calcite Carrara marble -2.37 ± 0.04 +2.46 ± 0.01 Assonov et al. (2020) VS001/1-A Aragonite Hydrothermal vein -12.41 ± 0.03 nA He et al. (2021) S0161 Calcite Calc-silicate xenolith -5.42 ± 0.03 +0.12 ± 0.04 Stern (2024a) S0436 Aragonite Speleothem -6.88 ± 0.03 +0.09 ± 0.07 Stern (2024b) aAlso known as NIST SRM 8543; bAlso known as NIST SRM 8544; cIsotope values are NIST information values defining the VPDB scale, hence, no uncertaintiesWe analyzed 578 spots by SIMS, of which 311 were located on the four microanalytical reference materials and 267 were situated on A. trapezia shell portions grown before and during aquaculture. We verified all SIMS spot locations on shells and reference materials by post-analysis BSE imaging, by which we identified 173 spots that were situated at compromised analysis locations. These were excluded from further consideration. For the reference materials, these included topographical features such as cracks and cleavage strikes. For shell portions grown during controlled aquaculture experiments (Table 2), we rejected any spots that touched the inner shell edge and thus the epoxy as well as those that overstepped or fell behind the Sr-labeled shell increment (Figure 1c, red circles) and only accepted spots falling between the Sr label and the inner shell edge as these were evidently grown during the temperature-controlled growth period (Figure 1c, blue circles). This selection procedure yielded 77 ideally situated shell SIMS analyses for shell portions grown during aquaculture and 17 spots on shell portions grown previously in the wild. Although the Sr pulse-chase labelling event was performed at target seawater temperatures, we chose to exclude any SIMS analyses that overstepped into the Sr labeled shell portion to avoid potential trace element-induced complications in bias correction as reported in previous studies (Allison & Finch, 2010; Rollion-Bard & Marin-Carbonne, 2011).2.8 Matrix bias correction for in-situ stable isotopes in biogenic aragonites:The δ18O ratios of aragonite samples obtained by ion microprobe required a two-step correction process addressing both instrumental and matrix-induced fractionation. The first step in quantifying ion microprobe δ18O ratios is to correct for instrumental mass fractionation by subtracting the difference between the oxygen isotope composition measured on the primary reference material by IRMS from all SIMS analyses (see equations provided in, e.g., Rollion-Bard & Marin-Carbonne, 2011). Secondly, matrix corrections are applied that further correct for the residual bias, i.e., the remaining matrix differences caused by compositional and structural aspects between the primary reference material and the unknowns. In this context, 'matrix' has a wider definition extending beyond mineralogy and polymorphism and also includes more subtle structural and chemical differences, e.g. within solid-solution series (Fayek et al., 2001; Rollion-Bard & Marin-Carbonne, 2011), and differences in elemental abundances observed between abiotic and biogenic aragonites (He et al., 2021). Collectively, these factors lead to matrix-specific behaviors in primary ion beam sputtering and secondary ion formation that need to be corrected to achieve accurate values.Following this approach, we first normalized the SIMS δ18O ratios to the primary calcite S0161 using the in-house software POXI-MC and refer to these values as δ18OSHRIMP in the following. Figure 2 shows the δ18O difference between SIMS measurements and the IRMS reference values (i.e., δ18OSHRIMP–IRMS bias) as a function of the combined abundances of Na, Mg, Ca, and Sr (Table 3) for VS001/1-A, S0436, and two wild shell samples, for which we have co-located SIMS, IRMS, and EPMA analyses, into a linear regression:Equation 1: y = 0.436 · Σ(Na, Mg, Ca, Sr) [% m/m] – 16.941 (n = 4, R2 = 0.91, p < 0.05)Both wild shell portions (Figure 2, gray diamonds) plot at lower combined element abundances than the two geological aragonite reference materials (Figure 2, black symbols). The linear regression expressed for the calcite-normalized aragonite reference materials and wild shell δ18O biases has a positive slope, illustrating an increases in δ18OSHRIMP bias with the combined element abundances. Next, we applied the regression to the δ18OSHRIMP analyses of the four temperature-controlled A. trapezia cultures using their combined abundances of Na, Mg, Ca, and Sr in Equation 1 to calculate their residual δ18O matrix bias: -0.38 (13 °C), -0.46 (18 °C), -0.33 (23 °C), and -0.29 (28 °C). We refer to these calculated biases as “biogenic δ18O matrix bias” in the following. Subtracting the biogenic δ18O matrix bias from the calcite-normalized δ18OSHRIMP values yielded the final matrix-corrected δ18OArg averages of the temperature-controlled shell portions (given in ‰ VPDB with ± first standard deviations): 2.83 ± 0.31 (13 °C), 1.67 ± 0.39 (18 °C), 0.78 ± 0.18 (23 °C), and 0.03 ± 0.24 (28 °C) (Table 4).It is critical to note here that the quasi-stoichiometric aragonite reference materials differ from the shells in three aspects: (1) Minor and trace elements (i.e., Na, Mg, and Sr) are significantly lower, (2) the Ca abundance is significantly higher, due to (3) the absence of organic phases (unlike the crossed-lamellar architecture of the shells known to contain up to 2 % m/m (Agbaje et al., 2017; Böhm et al., 2016)). Together, these three differences constitute the residual biogenic matrix effect that this approach effectively and comprehensively corrects for. This matrix calibration approach was inspired by the linear relationship between δ18O bias versus Ca abundance for abiotic and biogenic aragonites presented in He et al. (2021). Based on this observation and noting that the combined element abundances are dominated by Ca mass fractions, we also present a simplified linear regression showing the bias δ18OSHRIMP–IRMS as a function of only Ca (see also Figure S2):Equation 2: y = 0.418 · Ca [% m/m] – 16.079 (n = 4, R2 = 0.96, p < 0.02)to allow for a comparison with the calibration based on combined elemental abundances (Equation 1).However, follow-on experiments are needed to deepen our understanding of how robust the linearity of the combined elemental abundances and residual biogenic δ18O matrix bias is on a day-to-day basis by using different instrumental conditions and different SIMS instruments. 3 Results:3.1 Growing live Anadara trapezia shells under temperature-controlled aquaculture conditions:The 80-day aquaculture experiments with live A. trapezia shells yielded a closely monitored and controlled multi-variate dataset that includes all critical seawater chemistry parameters (Table 2). Target temperatures of 13, 18, 23, and 28 °C were met precisely, with averages measuring 13.0 ± 0.2, 17.9 ± 0.1, 23.0 ± 0.3, and 28.1 ± 0.5 °C, respectively (Tables 2 and S2). Average salinities were 35.8 ± 0.4, 35.6 ± 0.5, 35.3 ± 0.6, and 35.6 ± 0.9 ppt, respectively, for cultures reported in order of increasing temperature (Tables 2 and S3). Analytical precisions are provided as first standard deviations (1s) for all seawater parameters. The seawater pHT values averaged 8.02 ± 0.06, 8.03 ± 0.05, 8.02 ± 0.06, and 8.03 ± 0.04 and alkalinities averaged 2,150 ± 162, 2,100 ± 127, 2,037 ± 144, and 2,009 ± 186 µmol·kg-1, respectively, for cultures listed in order of increasing seawater temperature (see Tables 2 and S4). Both pHT and alkalinities reflect those of natural seawater and are critical parameters in characterizing marine seawater carbonate chemistry (Table 2). The in-house reference seawater solution was accurate within an RSD of 6 % for all cations measured by ICP-OES. The aragonite saturation state (Ωara), which is a measure of the thermodynamic behavior of aragonite to dissolve (Ω <1) or precipitate (Ω >1) was calculated for each culture using mean daily seawater temperatures, salinities, pHT, and alkalinity values together with the Ca2+ concentrations as input parameters and was 2.2 ± 0.4, 2.7 ± 0.4, 3.0 ± 0.4, and 3.6 ± 0.5, respectively (listed in order of increasing seawater temperature) and thus shows an expected correlation with seawater temperature (see Tables 2 and S5). The Na/Ca, Mg/Ca, K/Ca, and Sr/Ca do not show a significant difference between the temperature-specific cultures (Tables 2 and S6). Lastly, δ18OSW averaged 0.53 ± 0.15 (n = 27), 0.56 ± 0.16 (n = 27), 0.73 ± 0.21 (n = 27), and 0.87 ± 0.28 (n = 26) ‰ VSMOW, respectively, for cultures listed in order of increasing seawater temperature (Tables 2 and S7), with ± values representing first standard deviations. We observed a slight increase in δ18OSW, with increasing temperature, as an effect of increased evaporation rates, however, these were found to be within first standard deviations. Apart from the temperature-dependent ΩArg and a slight increase in average δ18OSW, seawater parameters do not show a correlation with seawater temperature. Measured seawater temperatures met target conditions with 5 °C intervals between cultures. Carbonate, cation, and oxygen isotope system parameters were controlled by regular water exchanges.Table 2. Overview of environmental and seawater chemistry parameters for the four temperature-controlled A. trapezia aquaculture experiments. The bivalves were grown for 80 days under tightly monitored and controlled conditions. Seawater temperatures were recorded every 15 minutes, while all other data were sampled at 3-day intervals. Salinity was measured via a hand-held logger, free pHT and total alkalinity were analyzed using a spectrophotometer and aragonite saturation states (Ωara) were calculated from seawater chemistry. δ18OSW were acquired on a Picarro CRD spectrometer and Element/Ca ratios by ICP-OES. All data are shown as averages with first standard deviations (1s). See Tables S2 to S8 in Medd et al., (2024) for full datasets. Target temperature [°C]: Measured temperature [°C]: Salinity [ppt]: pHT: Alkalinity [μmol·kg-1]: Na/Ca [mmol·mol-1]: Mg/Ca [mmol·mol-1]: K/Ca [mmol·mol-1]: Sr/Ca [mmo·mol-1]: ΩAraa: δ18OSW [‰, VSMOW]: 13 13.0 ± 0.2 35.8 ± 0.4 8.02 ± 0.06 2150 ± 162 46,838 ± 1,880 5,146 ± 13 1,016 ± 66 8.9 ± 0.3 2.2 ± 0.4 0.53 ± 0.15 18 17.9 ± 0.1 35.6 ± 0.5 8.03 ± 0.05 2100 ± 127 45,790 ± 1,656 5,140 ± 10 1,012 ± 58 9.1 ± 0.7 2.7 ± 0.4 0.56 ± 0.16 23 23.0 ± 0.3 35.3 ± 0.6 8.02 ± 0.06 2037 ± 144 46,781 ± 1,960 5,149 ± 15 1,000 ± 62 8.9 ± 0.3 3.0 ± 0.4 0.73 ± 0.21 28 28.1 ± 0.5 35.6 ± 0.9 8.03 ± 0.04 2009 ± 186 45,734 ± 1,870 5,058 ± 71 995 ± 50 8.9 ± 0.3 3.6 ± 0.5 0.87 ± 0.28 aAragonite saturation states calculated using salinity, Ca2+ concentrations, total alkalinity, and pHT as input into the ‘CO2Sys MS Excel Macro (Pierrot et al., 2011) using the carbonate species abundances based on the models of Millero et al. (2006) and the solubility product of Morse et al. (1980).3.2 Shell composition and growth:Anadara trapezia shell valves (Figure 1a) were found to consist of aragonite (Figure 1b), which was confirmed through peaks related to the Raman lattice mode region (153 and 206 cm-1), and those related to the intrinsic modes of the carbonate anion, namely the ν4 internal in-plane bending mode at 701 and 706 cm-1 as well as the ν1 internal symmetric stretching mode at 1085 cm-1 (e.g., Wehrmeister et al., 2011).The growth in temperature-controlled aquaculture was informed from BSE images. The 42 pre-screened shell cross-sections revealed different growth rates across the four temperature groups (Table S8): For the group cultured at 13 °C (n = 11), the average growth was 13.4 ± 0.5 µm, with only one specimen suitable for ion microprobe analyses. Shells cultured at 18 °C (n = 13) grew 28.7 ± 1.0 µm on average. The shells grown at 23 °C (n = 14) measured 27.2 ± 1.2 µm on average, with three specimens selected from both the 18 °C and 23 °C groups. Shells cultured at 28 °C (n = 4) grew 15.4 ± 0.4 µm on average, moreover, only one shell had survived the extended high temperature, which was selected for SIMS. The high mortality rate in the 28 °C culture indicates the species’ difficulty withstanding prolonged heatwaves, while the downscaled growth rate in the 13 °C group suggests the bivalves were approaching a growth hiatus. Only shells with sufficient growth were selected for in-situ stable oxygen analyses by ion microprobe.Elemental mass fractions of Na, Mg, Ca, P, Cl, and Sr were measured across a subset of 12 shells (3 individuals per temperature group), shell portions grown in the wild, and the microanalytical oxygen isotope reference materials (Table 3). The shell Ca abundance ranges from 37.1 to 37.8 % m/m, while the stable isotope reference materials have higher values ranging from 39.0 to 40.1 % m/m, which is a difference of up to 3 % m/m (Table 3). The shells also have detectable mass fractions of Na, Mg, and Sr, which combine to 0.4 to 0.8 % m/m that systematically decrease with seawater temperature. In contrast, these elements are mostly below detection limits in the SIMS calibration materials. Hence, the difference in Ca abundances between biogenic and abiotic aragonite cannot be explained by the presence of trace elements alone, further supporting the presence of up to 2 % m/m of total organic phases, in line with previous studies (Agbaje et al., 2017; Agbaje et al., 2019; Böhm et al., 2016).Table 3. Quantitative major and minor element mass fractions of A. trapezia shell portions and carbonate reference materials measured by WDS FEG-EPMA. Elemental averages and first standard deviations (1s) are shown in % m/m together with cation sums (with and without Ca), and Element/Ca ratios in mmol·mol-1. Measurements on shell portions grown in temperature-controlled aquaculture were carefully placed between the Sr labeled shell increment and the inner shell edge during BSE imaging, while the spots on shell portions grown in the wild followed micromill and SIMS transects. For the full dataset and details on the WDS calibration, see Tables S9 and S10, respectively. Elements below detectability in all measurements and are not shown (P <0.2, S <0.04, Cl <0.2). The uncertainties associated with cation sums are propagated from 1s of the combined elements. See Table S9 in Medd et al., (2024) for full dataset. Sample ID: n: Na [% m/m]: Mg [% m/m]: Cae [% m/m]: Sr [% m/m]: Σ(Na, Mg, Sr) [% m/m]: Σ(Na, Mg, Ca, Sr)f [% m/m]: Na/Ca [mmol·mol-1]: Mg/Ca [mmol·mol-1]: Sr/Ca [mmol·mol-1]: A. trapezia shell portions grown at target temperatures: 13 °Ca 14 0.53 ± 0.03 0.05 ± 0.01 37.2 ± 0.4 0.23 ± 0.08 0.8 ± 0.1 38.0 ± 0.4 24.9 ± 1.6 2.1 ± 0.6 2.8 ± 0.9 18 °Cb 15 0.49 ± 0.03 0.04 ± 0.01 37.1 ± 0.7 0.19 ± 0.04 0.7 ± 0.1 37.8 ± 0.7 22.9 ± 1.7 1.8 ± 0.6 2.3 ± 0.5 23 °Cc 14 0.40 ± 0.04 0.04 ± 0.02 37.5 ± 0.6 0.21 ± 0.08 0.7 ± 0.1 38.2 ± 0.6 18.5 ± 2.3 1.6 ± 0.7 2.7 ± 1.1 28 °Cd 15 0.35 ± 0.02 0.02 ± 0.01 37.8 ± 0.3 <0.08 0.4 ± 0.02 38.2 ± 0.3 16.2 ± 0.9 1.0 ± 0.5 - A. trapezia shell portions grown in the natural environment prior to temperature-controlled culturing period: At-13-17R 12 0.43 ± 0.05 0.03 ± 0.01 37.5 ± 0.3 <0.08 0.5 ± 0.01 37.9 ± 0.3 20.3 ± 2.2 1.3 ± 0.6 - At-28-22R 7 0.34 ± 0.05 <0.01 37.6 ± 0.3 <0.08 0.3 ± 0.05 38.0 ± 0.3 15.6 ± 2.3 - - Microanalytical carbonate reference materials: S0161 20 <0.01 0.05 ± 0.01 39.7 ± 0.3 <0.08 0.1 ± 0.01 39.8 ± 0.3 - 1.9 ± 0.4 0.6 ± 0.5 S0436 20 <0.01 <0.01 40.1 ± 0.3 <0.08 - 40.1 ± 0.3 - - - VS001/1-A 9 0.16 ± 0.03 <0.01 39.0 ± 0.2 0.61 ± 0.16 0.8 ± 0.2 39.8 ± 0.3 7.1 ± 1.4 - 7.2 ± 1.8 Detection limits: 0.01 0.01 0.01 0.08 - - - - - aMeasurements obtained from 3 shells: At-13-15R, At-13-17R, At-13-19R; bMeasurements obtained from 3 shells: At-18-07R, At-18-15R, At-18-18R; cMeasurements obtained from 3 shells: At-23-09R, At-23-11R, At-23-15R; dMeasurements obtained from 3 shells: At-28-18R, At-28-19R, At-28-22R; eCa abundances were applied to the simplified matrix calibration (Equation 2) to correct for δ18O matrix bias (see full dataset in Table S13); fCombined element abundances (Na, Mg, Ca, and Sr) were applied to the holistic matrix calibration (Equation 1) to correct for δ18O matrix bias (see full dataset in Table S12)3.3 In-situ stable oxygen isotope measurements of aragonite bivalve shells:Using our comprehensive matrix calibration (see Equation 1 and Figure 2), we obtained final matrix-corrected δ18OArg values for the temperature-controlled shell portions: 2.83 (13 °C), 1.67 (18 °C), 0.78 (23 °C), and 0.03 (28 °C) listed in Table 4. To demonstrate the improvement in matrix correction achieved here, we compared our biogenic δ18O bias calibration strategy with simpler calibrations and correlated IRMS data using the wild-grown shell portions (Figure 3). Both transects targeted the same growth increments, allowing us to correlate different SIMS calibration strategies with the conventional IRMS method: The closest agreement between δ18OIRMS (Figure 3, red) was observed for our biogenic matrix calibration based on combined element abundances, δ18OArg (Figure 3, blue), followed by the calcite-normalized δ18OSHRIMP data without further matrix correction (Figure 3, white), while δ18OSHRIMP matrix corrected using only one aragonite reference material (VS001/1-A light gray and S0436 dark gray in Figure 3) was most offset from the δ18OIRMS. The simpler SIMS calibration approach using only one aragonite reference material are ca. 0.5 (using VS001/1-A) to 0.9 ‰ (using S0436) lower than the δ18OIRMS and δ18OArg data. These numbers are consistent with previous studies showing SIMS δ18O values to be 0.5 to 0.8 ‰ lower than δ18OIRMS in other bio-aragonites including otoliths and gastropod shells (e.g., Aubert et al., 2012; Helser et al., 2018; Long et al., 2020). These previous studies attributed the lower δ18O values obtained by SIMS to the simultaneous measurement of aragonite and organic phases in the biomineral architecture. The agreement between δ18OIRMS and δ18OArg confirms that our biogenic matrix calibration based on combined element abundances indeed accounts for the structural and compositional differences between abiotic and biogenic aragonites. Further, using a single aragonite reference material in a simple, one-dimensional calibration results in greater bias, because abiotic reference materials have a positive bias, while bio-aragonites have a negative bias. These opposing biases amplify the matrix discrepancy, reducing accuracy and highlighting the need for element-dependent matrix bias corrections to address this issue. Table 4. Ion microprobe δ18O ratios of A. trapezia shell portions and carbonate reference materials shown together with corresponding biases and analytical precision values. Biogenic aragonite matrix biases were obtained using Equation 1 and the combined element abundances from EPMA, eliminating structural and compositional differences between abiotic reference materials and biogenic shell samples. All δ18O values given in ‰ VPDB. See Table 1 for details on the microanalytical, carbonate reference materials used and see Table S12 in Medd et al., (2024) for the full dataset. Specimen ID: n: Calcite-normalized δ18OSHRIMP weighted meane: δ18OSHRIMP- δ18OIRMS bias: Biogenic aragonite matrix bias: Final bias corrected δ18OArg weighted mean: ± 1s: ± Internal precision (95 % conf.): ± Precision estimate: A. trapezia shell portions grown at stable temperatures: 13 °Ca 16 2.45 - -0.38 2.83 0.31 0.10 0.02 18 °Cb 35 1.20 - -0.46 1.67 0.39 0.11 0.02 23 °Cc 12 0.45 - -0.33 0.78 0.18 0.11 0.03 28 °Cd 14 -0.26 - -0.29 0.03 0.24 0.11 0.03 A. trapezia shell portions grown in the natural environment prior to temperature-controlled culturing period: At-13-17R 9 -0.27 -0.27 -0.38 0.10 0.36 0.09 0.03 At-28-22R 8 -1.46 -0.50 -0.42 -1.04 0.52 0.12 0.04 Microanalytical calibration materials: S0161 111 -5.42 0.00 - - 0.29 0.16 0.01 NBS19 75 -2.34 -0.10 - - 0.20 0.15 0.01 S0436 95 -6.19 0.69 - -6.88 0.37 0.14 0.01 VS001/1-A 30 -12.18 0.23 - -12.41 0.37 0.14 0.02 aMeasurements obtained from 1 specimen: At-13-15R; bMeasurements obtained from 3 specimens: At-18-07R, At-18-13R, At-18-14R; cMeasurements obtained from 3 specimens: At-23-09R, At-23-11R, At-23-15R; dMeasurements obtained from 1 specimen: At-28-22R; eCalcite-normalized (δ18OSHRIMP) values represent final values for calcites, while aragonites required further matrix corrections (see calculations used in Table S12). Note: The alternative, simplified SIMS dataset evaluated using Ca abundances applied to Equation 2 is provided in Table S13 for comparison3.4 A novel SIMS-based SST versus δ18O thermometer relationship:The shell material grown at four tightly controlled seawater temperatures (Table 2) served as crucial in-house calibration materials for the first SIMS-based seawater thermometer for A. trapezia. The relationship between seawater temperature and aragonite-seawater δ18O fractionation is provided in Equation 3, while the relationship for the fractionation factor 103 lnαArg-SW is provided in Equation 4 (Table S14); both relationships are based on the SIMS δ18O values using the matrix calibration with combined major, minor, and trace element abundances (outlined in 2.8):Equation 3: T(°C) = 23.05 ± 0.36 - 4.48 · (δ18OArg [‰ VPDB] - δ18OSW [‰ VSMOW] ± 0.25 (n = 77, R2 = 0.994, p <0.004)Equation 4: 103lnαArg-SW = 17.78 ± 0.88) · 103/T [K] - (29.44 ± 2.40 (n = 77, R2 = 0.995, p <0.003)The regressions are based on the means of the four temperature groups to minimize sampling bias, as the 77 SIMS analyses were unevenly distributed, with more measurements concentrated in the middle temperatures where multiple individuals were available (Table 4). A p-value of 0.05 was set as the criterion for identifying significant effects. The seawater thermometer regressions (Equations 3 and 4) are visualized in Figure 4 (solid black line). The underpinning data are shown as both averages (black squares, error bars propagated from SIMS δ18OArg and δ18Osw first standard deviations) and individual data points (black circles) together with the confidence interval (gray shaded envelope). We used the 95 % confidence level as a reasonable cut-off for identifying statistical outliers, which returned one value in the 18 °C group (shown as red plus). As a single outlier seems negligible and prescreening of each ion microprobe spot was extensive (i.e., based on both optical and electron imaging), we retained this spot in the regression. The degree of remaining scatter observed within the 95 % confidence interval is suggested to result from natural variability within the realm of isotopic equilibrium fractionation. Both thermometer relationships (Figure 4a and b) were calculated from the four temperature group averages to counteract bias from differing numbers of data points and shell specimens.Using the simplified SIMS matrix calibration based only on shell Ca abundances and Equation 2, additional seawater thermometer relationships provided in Equations 5 and 6 show the seawater temperature and aragonite-seawater δ18O fractionation, as well as the fractionation factor 103 lnαArg-SW, respectively (Figure S3, Table S15):Equation 5: T(°C) = 23.31 ± 0.34 - 4.31 · (δ18OArg [‰ VPDB] - δ18OSW [‰ VSMOW] ± 0.22 (n = 77, R2 = 0.995, p < 0.003)Equation 6: 103lnαArg-SW = 18.67 ± 0.71) · 103/T [K] - (32.39 ± 2.40 (n = 77, R2 = 0.997, p < 0.002)4 Discussion:4.1 Strontium pulse-chase labeling improves confidence in SIMS δ18O analyses:Strontium pulse-chase labelling was found to be an effective strategy to mark the start of the bivalve growth experiment as a precondition for identifying shells with enough growth for SIMS analyses and obtaining well-targeted SIMS analyses (Figure 1c). Combining optical microscopy during the SIMS session with subsequent BSE imaging allowed more precise analyses within narrow growth increments, compared to previous strategies, such as literature-based growth rates (e.g., Böhm et al., 2000; Carré et al., 2005; Chamberlayne et al., 2021) or shell height measurements (e.g., Wanamaker et al., 2006), which would not have been precise enough here. As the pulse-chase Sr label is visible in high-resolution BSE images, it also surpasses the imaging resolution of fluorescent markers (e.g., Al‐Qattan et al., 2023) and is therefore even better suited for marking narrow growth increments at high resolution as well as for species with inconclusive natural growth banding (e.g., Long et al., 2020).4.2 Quantitative matrix corrected δ18O for bio-aragonites with organic-inorganic architectures:The difference in Ca mass fractions between biogenic and abiotic aragonites is linked to their distinct formation processes. The abiotic aragonites used here (VS001/1-A and S0436) formed as vein precipitates (He et al., 2021) and speleothem (Stern, 2024b), with both processes linked to inorganic supersaturation-driven crystallization (e.g., Brown et al., 1962; Riechelmann et al., 2014) although some speleothems reveal more complex formation processes (Frisia et al., 2022). On the other hand, the biogenic aragonite shells are mineral-organic nano-composite materials with fundamentally different crystallization pathways formed by living organisms (Lowenstam & Weiner, 1989). In fact, mollusk shells have been extensively shown to grow via non-classical pathways that involve transient phases including amorphous calcium carbonate (ACC), which later transforms to the final crystalline product, (i.e., aragonite in the case of A. trapezia, see Figure 1b) (e.g., Addadi et al., 2003; Jacob et al., 2011; Wolf et al., 2016). In bivalve shells, mineralization is orchestrated by an organic matrix with mineral-organic interfaces present at all length scales (Wolf et al., 2016). The outer shell layer of A. trapezia analyzed here has a crossed-lamellar shell architecture (Figure 1c), which is the most common mollusk shell architecture (Boggild, 1930), prevalent since the Middle Cambrian (Runnegar, 1985). It has a plywood-like organization with lamellae of alternating crystallographic orientations (Figure 1c) separated by thin organic sheaths (Agbaje et al., 2017; Agbaje et al., 2019; Böhm et al., 2016). This shell architecture contains about 2 % m/m of total organic content (Agbaje et al., 2017; Agbaje et al., 2019; Böhm et al., 2016). This total organic content is in accordance with the remaining difference of about 2 % m/m between the biogenic aragonite of our A. trapezia shells and the geological aragonites when all major and trace elements are considered (Table 3). Hence, the two-dimensional strategy using combined element abundances to correct residual SIMS matrix bias in biogenic aragonites applied to Equation 1 (Figure 2) provides an improvement in accuracy compared to simpler, one-dimensional calibration approaches using only a single aragonite reference material (Figure 3).4.3 The influence of minor and trace elements on the δ18O thermometer relationship: Figure 5a and b present the seawater temperature and carbonate-fluid δ18O fractionation relationship and the relation for the fractionation factor (α), respectively, together with previously published relationships. Equations 3 (in Figure 5a) and 4 (in Figure 5b) were matrix corrected using combined element abundances of Na, Mg, Ca, and Sr (solid black line), while Equations 5 (in Figure 5a) and 6 (in Figure 5b) were matrix corrected using Ca mass fractions (dashed black line). Both matrix bias calibrations are compared with published IRMS-based carbonate-fluid thermometers (colored lines in Figure 5a and b). In Figure 5a, both SIMS-based thermometers plot congruently at higher SSTs, however, a deviation occurs below 16 °C and at 10 °C the influence of minor and trace elements on δ18O becomes more pronounced, with a 0.2 ‰ difference observed between the two SIMS-based thermometers (i.e., Equations 3 and 5). This observation confirms that Na, Mg, and Sr are systematically influenced by SST in the aragonitic shells of A. trapezia, though their effect is smaller compared to the stronger correlation of δ18O with SST. If trace elements were independent of seawater temperature, Equations 3 and 5 would run parallel. As the combined temperature-sensitivity of Na, Mg, and Sr is incorporated into the SIMS-based thermometer (Equation 3), this approach can be understood as a paired proxy strategy for bio-aragonite sclerochronological archives (Trofimova et al., 2020).4.4 SIMS-based SST versus δ18O thermometer relationship comparison:Our SIMS-based relationships (solid and dashed black lines in Figure 5a and b) plot close to other aragonite biomineral thermometers (Figure 5a and b), with the closest relationship observed to the mollusk-based thermometer by Grossman and Ku (1986) and the aragonitic Arthritica helmsi bivalve shells presented by Chamberlayne et al. (2021). Further away (i.e. with a lower intercept for the isotope-temperature relationship in Figure 5a and a higher isotope fraction factor in Figure 5b), plot inorganically precipitated aragonite (Wang et al., 2013), aragonitic bivalve larvae shells of Placopecten magallanes (Owen et al., 2008), and the thermometer relationship of Böhm et al. (2000) that consists of aragonitic coralline sponges as well as mollusks and foraminifera data of Grossman and Ku (1986). The aragonitic shells of Mesodesma donacium bivalves (Carré et al., 2005) have a different intercept, causing them to plot with calcite thermometers at higher SST and with aragonite thermometers at lower SST. The prismatic calcite shell layer of Mytilus edulis bivalves (Wanamaker et al., 2006) plot further away (Figure 5a and b). The largest deviation from the thermometers developed in this study is observed in inorganically precipitated calcite (Kim & O'Neil, 1997). The general difference between calcite and aragonite thermometers is perhaps expected due to their distinct isotope fractionation factors, stemming from variations in internal vibrational frequencies and carbonate anion bond strengths (e.g., O'Neil et al., 1969; Tarutani et al., 1969). The insignificant differences between the SIMS-based thermometer and the IRMS-based relationships from Grossman and Ku (1986) and Chamberlayne et al. (2021) demonstrate, as proof-of-concept, that SIMS can produce thermometer calibrations for paleoenvironmental reconstructions similar to those from traditional methods for aragonite matrices, provided matrix bias is carefully corrected.4.5 Validation of the SIMS-based δ18O thermometer relationship and recommendations for application:To test the performance of the SIMS-based thermometer, we used an independent literature dataset (Chamberlayne et al., 2021; Chamberlayne, 2021), which includes measured SST, δ18OSW and shell δ18OArg obtained by IRMS. The short-lived, estuarine micro-mollusk species, A. helmsi, is only distantly related to A. trapezia, sharing ancestry at the order level. Figure 5c compares measured with predicted SST using different thermometer relationships: The SIMS-based thermometer calibrated from combined major to trace element abundances (Equation 3, Figure 5c, blue squares) is shown together with predictions from Chamberlayne et al. (2021) (Figure 5c, green triangles) and the mollusk-based version of Grossman and Ku (1986) (Figure 5c, red triangles). We combined δ18OArg with the δ18OSW from the previous month relative to the shell collection date, following the approach used in Chamberlayne et al. (2021), and compiled the monthly averages (Table S16). The predicted and measured SSTs are positively correlated, following the 1:1 line. All thermometers intersect this line, with the SIMS-based thermometer at slightly higher SSTs. The thermometers by Grossman and Ku (1986) and Chamberlayne et al. (2021) align at lower SSTs but diverge above 20 °C. Our SIMS-based thermometer regression converges with Grossman and Ku (1986) and runs parallel to Chamberlayne et al. (2021). At higher SST, predictions from the SIMS-based thermometer, using a simplified Ca-dependent calibration (Equation 5, Figure 5c, light blue circles), align with those based on combined element abundances. However, they diverge below 20 °C, indicating that incorporating minor and trace elements improved SIMS matrix calibration and thermometer accuracy at lower SST. The predictions for the three lowest measured SSTs deviate by up to 5 °C across all thermometer relationships, with estimates by Grossman and Ku (1986) and Chamberlayne et al. (2021) deviating towards lower, and the SIMS-based thermometers towards higher estimates. These deviations may arise from measuring whole, crushed shells necessary for IRMS due to the small sample sizes ranging between 1.3–2.6 mm that may introduce bias towards the warmer growth season. Four seawater-shell datapoints with salinities <12 psu were rejected, as the predicted SSTs from all thermometers deviated significantly from the corresponding measured SSTs, by up to 13 °C for Grossman and Ku (1986) and 11 °C for this study, indicating that these data fall too far outside the calibrated thermometer ranges.Figure 5d further compares predicted and measured SSTs using box and whisker plots, with whiskers showing the full temperature range and the box highlighting the second and third quartile. The SIMS-based thermometer, calibrated using combined element abundances, predicts a median SST of 20.4 °C, closely aligning within 0.1 °C of the measured median SST of 20.5 °C. In comparison, predictions from the Grossman and Ku (1986) thermometer show a median SST of 19.1 °C deviating by 1.4 °C from the measured values, which aligns well with their sensitivity estimate of 1.6 °C (Grossman & Ku, 1986). The mean average SST for the SIMS-based thermometer is 20.5 °C deviating by 0.9 from the mean measured SST of 19.6 °C. Meanwhile, the Grossman and Ku (1986) thermometer shows a deviation of 0.4 °C, with a mean SST of 19.2 °C. This independent application example demonstrates that despite limited knowledge of A. helmsi's growth rates and lack of quantitative element abundances, the SIMS-based thermometer shows comparable sensitivity compared to traditional methods with potential for further improvement when growth rates and shell composition were better constrained. As A. helmsi is a short-lived micro-mollusk, in-situ δ18O analyses using SIMS could offer micron-scale SST chronologies at resolutions down to monthly or weekly increments, at a comparable sensitivity of whole shell IRMS analyses. Drawing on the experience gained here, we recommend that future users interested in quantitative SIMS δ18O-based SST reconstructions with bio-aragonites test the quantitative matrix calibration strategy outlined here, particularly for samples with fine-scaled growth increments, small shell sizes or those too precious for more destructive analyses with traditional methods. Further, SST reconstructions of bio-aragonites with similar mineral-organic properties are likely to benefit from the increased accuracy of the SIMS-based thermometer. The calibration using shells grown in stable aquaculture conditions, combined with precise analysis locations via Sr-labelling, provided a significantly more robust relationship compared to field-based methods used in previous studies. For example, the thermometers developed by Grossman and Ku (1986) involved whole-shell analyses from samples growing in heterogeneous natural environments. While pioneering at the time, this field-based sampling approach introduced greater uncertainties due to the variability in growth conditions, the averaging of shell growth increments, and seawater temperatures that were not aligned with the majority of the shell growth period. In contrast, the approach used here overcomes these challenges and enhances confidence through tightly controlled experimental conditions and BSE imaging to validate each SIMS analysis location against the Sr label before calibrating the thermometer. The uncertainties associated with current aragonite benchmark thermometers compared to new aquaculture-based calibration approaches merit further investigation to deepen our understanding of the robustness of paleothermometer relationships.Although the A. trapezia thermometer relationship should broadly apply to bio-aragonites with similar ranges of minor and trace elements as well as comparable total organic contents (1–2 % m/m), separate calibrations may be required for shells with significantly different properties and those living in seawater conditions falling too far outside the calibrated range, such as low salinities. Indeed, some paleothermometer calibrations have been achieved by adjusting salinities to force certain δ18OSW levels across different SST groups (e.g., Wanamaker et al., 2006). In contrast, we considered salinity to be less critical, only confirming that it fell within A. trapezia's natural habitat range and remained constant during the experiment (Tables 2 and S3). Given the significant regional and temporal variability in δ18O-salinity relationships within estuaries, deliberate control over salinity and its control over δ18OSW was deemed non-unique. Instead, we chose to ensure the broader applicability of our thermometer relationship by using unaltered, natural seawater during the shell growth period and ensuring that stable Ca and Mg levels were maintained (Tables 2 and S6). However, this explains the observed inaccuracy in reconstructing the low salinity conditions in the application example for the SIMS-based thermometer as well as the ones from Grossman and Ku (1986) and Chamberlayne et al. (2021).Lastly, the downscaled growth rates obtained in aquaculture conditions that limited the number of shells suitable for SIMS analysis were believed to be unconcerning as they fall within the range of natural growth rates. In the natural environment, shells exhibit cyclic growth patterns, at annual as well as tidal and/or daily resolutions following stochastic growth processes (Gim et al., 2021) indicating that achieving certain growth rates in aquaculture settings may not be of particular concern for thermometer calibrations. However, future studies aiming to grow similar temperature-controlled growth increments in their shells may benefit from a longer culturing period to increase their yield of acceptable SIMS analyses.Altogether, we demonstrated that high-resolution ion microprobes are indeed capable of producing paleothermometer relationships of similar, if not improved, sensitivity and confidence compared to traditional approaches, as long as rigorous calibration procedures and sampling strategies are applied with care.5 ConclusionsThe main objective of this study was to test whether in-situ ion microprobe analyses can produce a δ18O versus SST relationship for aragonite marine shells, despite difficulties in making accurate matrix bias corrections. This proof-of-concept study demonstrated that using suitable aragonite oxygen isotope reference materials with known element abundances allowed for the development of a linear calibration that eliminated the residual δ18O bias. This two-dimensional calibration approach yielded significantly more accurate δ18O values compared to the simple one-dimensional calibration procedure with a single reference material, where there is no flexible way to account for the structural and compositional differences within natural bio-aragonites. This allowed us to produce the first in-situ δ18O versus SST relationship from a suite of well-characterized Anadara trapezia bivalve shells grown under tightly controlled aquaculture conditions (Table 2). We marked the start of the growth period in the aragonitic shells using Sr pulse-chase labelling visualized using BSE imaging that ensured the selection of shells with enough growth for SIMS analyses and precise control over the exact analyses locations. The similarity between our in-situ thermometer and previously published ones produced by the traditional stable isotope analyses approach demonstrates the utility of ion microprobe analyses. In addition, our in-situ approach offers the advantage of measuring δ18O with a spot size of down to 25 μm, which is more than one order of magnitude finer than the traditional approach using micromilling and IRMS. Lastly, our application example validates SIMS as a powerful tool for accurate, micron-scale, high-resolution, and sensitive SST reconstructions for aragonite samples with organic-inorganic shell architectures, especially those not amenable to bulk analyses due to finely scaled growth banding, small sample sizes (i.e., otoliths), or those too precious for more destructive conventional analyses. Acknowledgments We thank Monika Misztela for her assistance with ICP-OES experiments. SHRIMP analytical data were obtained using instrumentation funded by ARC LIEF (LE22010083), NCRIS/AuScope and the Australian National University. We acknowledge the ANU Centre for Advanced Microscopy for access to its instrumentation and support from its staff members. The SST remote sensing data for the bivalve sampling locations was sourced from IMOS/NCRIS. We acknowledge the use of the CSIRO Marine National Facility (https://ror.org/01mae9353) and grant of sea time on RV Investigator in undertaking this research. We are grateful to Janaina Avila (University of Queensland) and Oscar Branson (University of Cambridge) for helpful discussions. This work was funded via an ANU Research School of Earth Science Award and an MSA Malacological Research Grant both awarded to LMO. This study is part of OMM.’s B.Sc.