Micromilling vs hand drilling in stable isotope analyses of incremental carbonates: The potential for δ13C contamination by embedding resin

Rationale Embedding resins are commonly used to facilitate high‐resolution sampling for stable isotope analysis but anomalous δ13C values have been observed in some cases. Here we compare the results of microsampling strategies for hand‐drilled versus resin‐embedded micromilled samples from the same marine shells to assess whether resin contamination is implicated in δ13C spikes. The comparison allows assessment of the relative benefits for spatial resolution, seasonal range for both δ18O and δ13C, and sample failure rates. Methods Hand‐drilled samples were obtained from two bivalve shells (Spisula sachalinensis), corresponding to micromilled samples on the same shells where high δ13C spikes were observed. All carbonate powders were analysed using a dual‐inlet Isoprime mass spectrometer and Multiprep device. Results from both sample sets were compared statistically. Results No anomalous high δ13C values and no failures due to insufficient gas were observed in the hand‐drilled samples in contrast to the embedded micromilled sequences. Spatial resolution was reduced (~2.5×) in the former compared with the latter, resulting in a small reduction in the total range observed in the micromilled δ13C and δ18O values. Reduced sampling resolution between the two datasets was only significant for δ18O. Conclusions For S. sachalinensis (as with other similar bivalves), rapid growth mitigates the reduced sampling resolution of hand drilling and does not significantly impact observed isotopic range and seasonal patterning. Occurrence of anomalous δ13C values were eliminated and failure rates due to insufficient sample size greatly reduced in the hand‐drilled dataset. We can find no other explanation for the occurrence of δ13C spikes than contamination by the embedding agent. We conclude that the logistical and interpretational benefits of careful hand drilling may be preferable to resin embedding for micromilling in marine shells, corals or speleothems where growth rate is rapid and the highest resolution is not required.

Rationale: Embedding resins are commonly used to facilitate high-resolution sampling for stable isotope analysis but anomalous δ 13 C values have been observed in some cases. Here we compare the results of microsampling strategies for handdrilled versus resin-embedded micromilled samples from the same marine shells to assess whether resin contamination is implicated in δ 13 C spikes. The comparison allows assessment of the relative benefits for spatial resolution, seasonal range for both δ 18 O and δ 13 C, and sample failure rates.
Methods: Hand-drilled samples were obtained from two bivalve shells (Spisula sachalinensis), corresponding to micromilled samples on the same shells where high δ 13 C spikes were observed. All carbonate powders were analysed using a dual-inlet Isoprime mass spectrometer and Multiprep device. Results from both sample sets were compared statistically.
Results: No anomalous high δ 13 C values and no failures due to insufficient gas were observed in the hand-drilled samples in contrast to the embedded micromilled sequences. Spatial resolution was reduced ($2.5Â) in the former compared with the latter, resulting in a small reduction in the total range observed in the micromilled δ 13 C and δ 18 O values. Reduced sampling resolution between the two datasets was only significant for δ 18 O.
Conclusions: For S. sachalinensis (as with other similar bivalves), rapid growth mitigates the reduced sampling resolution of hand drilling and does not significantly impact observed isotopic range and seasonal patterning. Occurrence of anomalous δ 13 C values were eliminated and failure rates due to insufficient sample size greatly reduced in the hand-drilled dataset. We can find no other explanation for the occurrence of δ 13 C spikes than contamination by the embedding agent. We conclude that the logistical and interpretational benefits of careful hand drilling may be preferable to resin embedding for micromilling in marine shells, corals or speleothems where growth rate is rapid and the highest resolution is not required.

| INTRODUCTION
Carbon and oxygen stable isotope (δ 13 C/δ 18 O) analysis of marine bivalves can be used to provide (palaeo)environmental information on marine conditions, including water temperature, 1-4 salinity 5 and productivity, 6,7 as well as biogeographic information on the species analysed. [8][9][10] Depending on the size and shape of the material/ species being sampled, and the resolution and precision requirements of the study, a variety of sampling techniques can be used to produce powder samples for mass spectrometry analysis. These have been discussed by other authors, primarily referring to hand drilling (sometimes known as microdrilling, but referred to as hand drilling throughout this paper), micromilling, SIMS (Secondary Ion Mass Spectrometry) and laser ablation techniques in the sampling of marine and freshwater carbonates such as mollusc shells, corals and otoliths, 9,11 as well as in the related field of speleothem research. [12][13][14][15] Although SIMS and laser ablation are increasingly used, hand drilling and micromilling remain the most commonly used sampling methods, 11,16 due to their balance of accessibility, precision and resolution, and because they produce the powdered samples necessary for high-precision MC-ICPMS (Multicollector Inductively Coupled Plasma Mass Spectrometry) and ICP-AES (Inductively Coupled Plasma Atomic Emission Spectroscopy). [17][18][19] Micromilling provides the opportunity to precisely mill from very specific regions of the sample and at high resolution, but is more time consuming than hand drilling and the equipment is often temperamental. The collection of micromilled powder samples is most commonly accomplished manually using a combination of scalpels and razor blades, 16 so the process requires a flat and relatively wide sampling surface. This means that in order to be micromilled, most materials need sectioning in order to provide such a surface. In the case of fragile or small/thin material (including, e.g., shells and otoliths) this also necessitates embedding specimens in a stabilising medium to make them less susceptible to breakage, and to increase the size of the flat surface available when collecting samples. The embedding and sectioning process can be laborious, further increasing the time and resources required for this approach.
By contrast, hand drilling is expedient and inexpensive.
Depending on the size of the drill bit, it usually produces larger sample sizes, although with lower resolution due to increased time averaging within and between samples. Unlike micromilling, hand drilling can be performed on flat or curved surfaces. This allows sampling to be done around the curve of a shell and reduces the need for extensive preparation of the sampling surface.
This study was prompted by the results of δ 13 C and δ 18 O analysis on archaeological samples of the marine bivalve Spisula sachalinensisa large and long-lived bivalve found in the northwest Pacific. 20 Results were obtained from a resin-embedded specimen which had been sampled using a computer-operated New Wave Research MicroMill. δ 18 O values show a seasonal annual temperature curve as expected, but the δ 13 C record of two of the seven shells showed regions where the values were significantly more positive than expected (up to $5‰). These positive spikes were especially apparent when compared with the rest of the sequences from these individual shells, as well as results from other shells analysed, where δ 13 C is relatively stable at $ +1 ± 0.5‰.
The species lives above the thermocline at depths <15 m, 23,[28][29][30] where there is not a high degree of seasonal change in surrounding seawater DIC (dissolved inorganic carbon). We therefore expect relative stability in shell δ 13 C values. Moreover, δ 13 C values as high as $5‰ would be considered implausibly high in all cases for marine bivalves. In McConnaughey and Gillikin's 6 overview of the physiological and environmental controls affecting shell δ 13 C values, there is no mention of any condition or species where marine shell δ 13 C would near 5‰, nor any mechanism that could account for such large and sudden spikes in δ 13 C. We are unaware of any literature that reports shell δ 13 C values this high and considers it a true reflection of the growing shell's isotopic composition. The large δ 13 C excursions seen in these two specific shells (labelled II_1 and VIII_3) discussed in this paper were therefore considered highly anomalous, leading us to investigate the potential role of sample preparation and sampling strategy in affecting δ 13 C results.
As the micromilled shells were embedded in polyester resin prior to sampling, one possible consideration is resin contamination. and a 0.17‰ depletion in δ 13 C, but in the context of their analysis this was not considered significant as these differences were smaller than their inter-sample variation. This study has since been cited as evidence that polyester resin 'does not contribute to the CO 2 evolved by acid digestion that is used for the isotopic measurements'. 33 However, the study of Mortensen et al 32  specific mass spectrometry of lipids in archaeological sediments, particular affecting δ 13 C 18:0 . Resin-associated analytical issues have also been previously noted by the BGS in high-resolution speleothem work, but these were never fully investigated. This existing evidence highlights the issue of resin contamination in carbonate palaeothermometry, but has not yet dissuaded the common use of sample pretreatment, embedding, and gluing in subsequent studies (e.g. 36,37 ). Moreover, previous studies do not explore whether changes to sampling strategy to avoid such contaminants results in an unacceptable loss of precision and/or resolution.
Here we tested corresponding hand-drilled, non-embedded samples from the same valves to assess whether the high δ 13 C values occurred in both datasets. If so, then we must re-examine our initial position and examine how these high δ 13 C results could be otherwise explained. Alternatively, if these δ 13 C peaks are not present in the hand-drilled samples then we can consider how sampling strategy and/or polyester resin contamination are potential issues for future carbonate stable isotope analyses. This would have relevance not only to the shells used in this particular study, but also to any researcher contemplating the use of embedding resin in a strategy for highresolution sampling of carbonates for stable isotopes.

| METHODS
The samples used in this experiment came from two archaeological S. sachalinensis valves (labelled VIII_3 and II_1) both collected from the site of Hamanaka 2, Rebun Island, Japan. In the initial stable isotope analysis undertaken on these shells, the valves had been sectioned as Milling was carried out up to 300 μm depth in multiple milling passes of 50 μm depth/pass, and using a 0.3 mm diameter diamond-coated dental drill bit. No chemical or physical pretreatments were applied to the samples before analysis. The non-embedded halves of the same valves were sampled using a Buehler hand drill, from the area corresponding to the original sampling locations (Figure 3). For both the micromilled and hand-drilled samples, the cross-lamellar layer of the shell was targeted, as is most common with shell palaeothermometry studies. 38 In this large species, the cross-lamellar layer is relatively thick (ca. 4 mm width after the second year of growth; Figure 3), so it was possible to avoid mixing material from the surrounding outer and inner layers by eye during hand drilling. In other smaller species of shell, it may be more difficult to specifically target one microstructural layer during hand drilling. Samples were spot drilled using a 1 mm diameter drill bit, and aluminium foil was used to collect the resulting aragonite powder before transferring it into microcentrifuge tubes. During hand drilling, the shell was supported using the non-dominant hand, which rested firmly on the lab bench and provided a 'cushion' between the shell itself and the bench to prevent direct contact between the two which could cause damage to the specimen. The sectioned surface of the shell was held angled down towards the aluminium foil to direct the shell powder onto the foil and reduce sample loss. Each sample was checked for purity of the aragonite using Fourier-transform infrared spectroscopy, 20 as per Loftus et al. 39 Once in the microcentrifuge tubes, the powders obtained using both the hand-drilling and micromilling methods were analysed identically, using a dual-inlet Isoprime mass spectrometer interfaced with a Multiprep autosampler (hereafter referred to as the Isoprime plus Multiprep) at the BGS. We aimed to produce 50 to 100 μg of carbonate per sample for isotope analysis; however, micromilled samples were often below 50 μg, smaller than those produced by hand drilling which were typically >100 μg (with 50-100 μg subsamples used for analysis). Samples are loaded into glass vials and sealed with septa, evacuated and anhydrous phosphoric acid delivered to the carbonate at 90 C. Evolved CO 2 is collected for 15 min, H 2 O is removed, and pure, dry CO 2 introduced into the mass spectrometer for measurement. Isotope ratios of carbon and oxygen ( 13 C/ 12 C, 18  Therefore, this was not the best approach to determine the effects of resin traces on the shell carbonates; it would have been preferable to analyse the resin sample on the Multiprep, but this ran the risk of a costly contamination of the instrument and could not be justified.

| RESULTS AND DISCUSSION
Of the 36 hand-drilled samples added for this paper, all produced sufficient gas for isotopic measurement. Of the pre-existing micromilled samples from these two shells, 21 of the 91 samples failed to produce enough gas to measure isotopic composition. This represents a failure rate of 23%, likely due to challenges with handling very small amounts of powder during the period between micromilling and mass spectrometry. The results of all (hand-drilled and micromilled) samples are summarised in Table 1 resolution of approximately one sample per 1.7 mm distance along the shell. Drill bits smaller than 1 mm diameter could be sourced to increase the resolution of hand drilling, but the spatial precision achievable by hand drilling will still be limited by the steadiness and coordination of the drill operator, and the fragility of increasingly small drill bits. Furthermore, this study does not represent the maximum spatial precision achievable with micromilling, as we sampled using discrete consecutive milling trenches (as seen in  Despite the lower spatial resolution and compression of δ 18 O values in the hand-drilled samples, the sequences appear more coherent and are easier to interpret in terms of seasonal shifts compared with the micromilled results ( Figure 2). This is at least partly because the increased resolution achieved in the micromilled results ( Figure 2A) is more complex and appears 'noisy', rendering seasonal Where there is concern about the possibility of resin Sloane and Jack Lacey for their assistance in the sample preparation and mass spectrometry processes. They are also grateful to two anonymous reviewers for their helpful comments and encouragement.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/rcm.9318.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.