Post‐mortem oxygen isotope exchange within cultured diatom silica

Rationale Potential post‐mortem alteration to the oxygen isotope composition of biogenic silica is critical to the validity of palaeoclimate reconstructions based on oxygen isotope ratios (δ18O values) from sedimentary silica. We calculate the degree of oxygen isotope alteration within freshly cultured diatom biogenic silica in response to heating and storing in the laboratory. Methods The experiments used freshly cultured diatom silica. Silica samples were either stored in water or dried at temperatures between 20 °C and 80 °C. The mass of affected oxygen and the associated silica‐water isotope fractionation during alteration were calculated by conducting parallel experiments using endmember waters with δ18O values of −6.3 to −5.9 ‰ and −36.3 to −35.0 ‰. Dehydroxylation and subsequent oxygen liberation were achieved by stepwise fluorination with BrF5. The 18O/16O ratios were measured using a ThermoFinnigan MAT 253 isotope ratio mass spectrometer. Results Significant alterations in silica δ18O values were observed, most notably an increase in the δ18O values following drying at 40–80 °C. Storage in water for 7 days between 20 and 80 °C also led to significant alteration in δ18O values. Mass balance calculations suggest that the amount of affected oxygen is positively correlated with temperature. The estimated oxygen isotope fractionation during alteration is an inverse function of temperature, consistent with the extrapolation of models for high‐temperature silica‐water oxygen isotope fractionation. Conclusions Routinely used preparatory methods may impart significant alterations to the δ18O values of biogenic silica, particularly when dealing with modern cultured or field‐collected material. The significance of such processes within natural aquatic environments is uncertain; however, there is potential that similar processes also affect sedimentary diatoms, with implications for the interpretation of biogenic silica‐hosted δ18O palaeoclimate records.

Amorphous biogenic silica (Si(OSi≡) n (OH) 4-n , where n ≤4) contains a significant mass of molecular water and hydroxylated silica (e.g. silanol, Si-OH), in addition to tetrahedrally bonded Si-O-Si silica. [15][16][17][18][19] The proportional concentration of silanols is reported as Q n , where Q represents a silicon atom surrounded by four oxygen atoms, while the suffix n gives the number of surrounding oxygen atoms (out of four) that are bonded to another silicon atom. 17,18,20 Tetrahedrally bonded (Q 4 ) silica is not known to undergo isotope alteration at temperatures <550°C; however, hydroxyl-bound oxygen within hydrated silica (Q 1-3 , herein referred to as Q 3 ) readily exchanges with ambient water, such that a fraction of a given biogenic silica sample will undergo δ 18 O re-equilibration under ambient conditions. 9,14 Methods have been developed to counter the presence of exchangeable Q 3 silica during δ 18 O silica analysis, most commonly by stepwise fluorination 21,22 or by controlled isotope exchange, 9,[23][24][25] and those approaches have been demonstrated to produce reliable data between multiple laboratories. 26 However, a number of researchers have observed differences between the δ 18 O silica values of living and sedimentary diatoms, which suggest that the δ 18 O silica value may not always faithfully record the signal acquired during diatom growth.
For example, Schmidt et al, 10 Moschen et al 27  and those sampled from deep water traps or surface sediments. In some cases, these increases were also associated with the progressive loss of Q 3 silica. 10,27 Similarly, a comparison of δ 18 O silica values of living and surface sedimentary diatoms from Lochnagar, an upland lake in Scotland, revealed 4-6 ‰ δ 18 O silica offsets between living and recently sedimented diatoms. 28,29 In the latter case, the sedimentary diatoms had lower δ 18 O silica values than the living diatoms, with sedimentary values falling upon the line defined by a series of lake surface sediment δ 18 O silica values across Europe. 30 Partial dissolution of diatom frustules is one mechanism that may explain isotopic alteration during sedimentation; however, experimental studies suggest that the effects of dissolution upon diatom δ 18 O silica values are small to negligible, 31 contrasting with experimental evidence using phytoliths. 32 Conversely, the process of silica condensation has been frequently proposed to impart significant effects upon δ 18 O silica values. 10,13,14,27 Dodd et al 14

| EXPERIMENTAL
In order to test the isotope stability of biogenic silica under heating and drying conditions, a series of experiments was conducted using freshly cultured diatom silica. These experiments provide the basis for oxygen isotope mass balance calculations of the fractional mass of affected oxygen and the associated isotopic fractionation at temperatures relevant to natural sedimentation and laboratory pretreatment. has low silica mass/biovolume which contrasts with the high silica/ biovolume of Cyclotella meneghiniana. 34 Given the differences in silica/biovolume, it is reasonable to assume that frustules of these two diatom taxa also differ in their surface area/mass ratio, which in turn has been demonstrated to influence silica dissolution rate. 35,36 We therefore expected the denser frustule of Cyclotella meneghiniana to be less susceptible to oxygen isotopic exchange or dissolution. Each batch was cultured at 15°C in an LMS 400 controlled temperature cabinet (LMS Ltd, Sevenoaks, UK) illuminated with a photon flux of 200 μmol/m 2 /s (approx. 43.6 W/m 2 ) using in-built fluorescent tubing on an 18:6 h day:night cycle. For growth medium, the CCAP recipe 'Diatom Medium' (DM) was used, 37 buffered to pH 7 using tris(hydroxymelthyl)aminomethane (TRIS). Each sample represented an amalgamation of 80 L of culture, spread over eight 10-L polycarbonate carboys and two cabinets. Each carboy was aerated by bubbling sterile filtered air, which ensured perpetual cell suspension within the culture.

| Experimental setup
Periodic subsamples were collected for cell density determination using a Coulter counter (Beckman Coulter Inc., Brea, CA, USA) and a Neubauer haemocytometer. Cells were harvested by centrifugation at the onset of the stationary growth phase in order to maximise harvested biomass. Following centrifugation, the organic diatom pellet was frozen at −80°C and freeze-dried for storage.
Removal of organic matter was undertaken using a sequential digestion using H 2 O 2 , HNO 3 + HCl (aqua regia; 1:3 ratio) and conc.
HCl at 60°C in a heated water bath, with repeated rinsing in 10 MΩ deionised water between phases. Due to the amount and resistance of the organic-rich samples, the digestion procedure took 1 week to reach completion. The final mass of silica per sample was of the order of 0.5 g.
In order to assess the oxygen isotopic exchange between fresh diatoms and water, two waters of different δ 18  week, after which time the silica was concentrated by centrifugation, and the supernatant water was decanted before the sample was transferred to a 2-mL centrifuge tube. The silica samples were then frozen (4°C) and freeze-dried.
The oxygen isotope composition of water used in each aliquot was monitored in order to account for any changes during the course of the experiment. A subsample of the initial water source was set aside ('start water', kept constant for each aliquot) and 5-mL samples of supernatant water were collected at the end of each experiment ('end water'). The water samples were stored refrigerated in air-tight HDPE bottles with minimal headspace and analysed within 1 month of collection. In some circumstances during storage at 80°C, the pressure and temperature within the experimental vial led to cracking of the lid and some evaporation of water. In these cases, the δ 18 O water value of the end water was higher than that of the start water by a maximum of~5 ‰ and the value used in subsequent calculations is the mean of the two valuesi.e. assuming a linear progression in both evaporative enrichment of 18 O in water and the degree of silica-water isotopic exchange.

| Isotope analysis
Stable isotope analyses were conducted at the NERC Isotope Geosciences Laboratory (Nottingham, UK). The oxygen isotopic composition of waters was analysed using the equilibration method 38 with a Sira 10 dual-inlet isotope ratio mass spectrometer with an For determination of the oxygen isotope ratio of biogenic silica, the hydrous outer layer of the frustules was removed by stepwise fluorination whereby a stoichiometric deficiency of the reagent, bromine pentafluoride (BrF 5 ), was used to partially react the samples before full reaction at 500°C with an excess of BrF 5 . 22 The oxygen liberated was converted into CO 2 by passing over hot graphite, following the method of Clayton and Mayeda. 39 Oxygen yields were monitored for comparison with the calculated theoretical yield. A random selection of samples was analysed in duplicate and gave a mean reproducibility of 0.3 % (1σ), which is comparable with the 0.3% reproducibility of the standard laboratory diatomite control (BFC mod ) both within individual batches and over the longer term. 22 The 18 O/ 16 O ratios were measured on a MAT 253 isotope ratio mass spectrometer (ThermoFinnigan, Bremen, Germany), and normalised through laboratory standards and NBS28. 22 The data are reported in the usual δ form, as per mil (‰) deviations from VSMOW. values, these mass balances are described as:

| Isotope exchange mass balance equations
where δT (x) is the δ 18 O value of the total sample and δW (x) is the δ 18 O value of the storage water. The subscript (x) denotes the storage water, either DI or BAS-Lo, and Δ BW(x) is an approximation of the fractionation factor between water and the altered silica (B): Solving Equations 1a and 1b simultaneously gives the following for A: enabling an estimate of the mass of Q 4 -bound oxygen affected during storage in water x. Equation 3 means that it is possible to calculate A without knowing δA, assuming that δA was the same in both experiments.
Solving for A also permits an estimate of δB and Δ BW for a given value of δA: δA is known in cases where the δ 18 O value of a sample was measured prior to the experiment; however, due to limited sample availability this was not always possible. Where δA was not measured, it was estimated by applying a constant offset equivalent to that observed between the measured δA and δT at 4°C. Note that estimates of Δ BW are extremely sensitive to the value of δA used, which limits accuracy but does not affect the overall trend.
These mass balance calculations are equivalent to those used by Labeyrie and Juillet 9our value of B approximates their value of xbut with an important difference. The use of stepwise fluorination means that only Q 4 silica was analysed in this study, whereas Labeyrie and Juillet 9 analysed the whole sample comprising Q 4 and Q 3 silica, following controlled isotope exchange. Therefore, Labeyrie and Juillet's x represents the total amount of secondary isotope exchange within the whole sample, with an assumption that all exchange takes place between water and Q 3 silica. By contrast, our B is the fraction of Q 4 -bound oxygen altered during treatment.

| Effect of temperature on isotope fractionation during secondary alteration of oxygen isotopes in diatom silica
The oxygen isotope fractionation (Δ BW ) between altered Q 4 silica (B) and water was estimated using Equation 4. When expressed as a fractionation, much of the scatter in Figure 2A is diminished, suggesting that a large proportion of the between-sample differences can be explained as a function of variations in δW and in the amount of oxygen affected ( Figure 4A). In this case, an added source of uncertainty is the oxygen isotopic composition of the initial Q 4 silica (δA). Here, δA is assumed to be equivalent to the oxygen isotopic composition of the total sample at the beginning of the experiment, whereby prefluorination removes all non-Q 4 silica. 22  Application of this correction to all samples defines a temperaturefractionation relationship that is consistent with that defined by the data from Stephanodiscus experiment #3 (Figure 4), except for a single sample, Stephanodiscus #2, 25°C, which is a clear outlier ( Figure 4A).
Irrespective of this outlier, there is a negative correlation between Δ BW and temperature (Figure 4), a relationship which appears to conform with the low-temperature extrapolation of published hightemperature silica-water oxygen isotope fractionation models ( Figures 4A and 4B). These published models were established through a range of experiments, including secondary alteration of diatom silica 9 and precipitation of amorphous silica 40  precipitation; condensation of Q 3 to Q 4 silica). In this circumstance, we discount methodological artefacts with the following justification.
First, the stepwise fluorination method is designed to react away loosely bonded hydration water and silanol prior to the liberation of Q 4 -bound oxygen at 500°C. 22 In principle, it is possible that some Q 3 silica survives this process, which would lead to oxygen isotope contamination by formerly exchangeable silica; however, this is deemed unlikely given the high reactivity of silanol with BrF 5 .
Furthermore, inter-laboratory comparisons with other methods which do not involve pre-fluorination indicate that our method is robust, including during analysis of silanol-rich silica standards. 26 Retention of Q 3 silica post-prefluorination should also lead to a consistent  Stephanodiscus hantzschii

30.5
Stephanodiscus hantzschii #3  (Figures 1 and 2). Secondly, contamination is discounted on the basis that the diatom culture material was never in contact with mineral silica and a prolonged oxidation stage was carried out in excess of usual requirements for removal of organic constituents. 45 Even in the case of some organic matter being present, this would be constant between samples for a particular experiment, and would therefore not account for the differences between treatments. The consistency between experiments, including between different diatom taxa, also attests to the likelihood of differences in organic contamination having a negligible effect on our results.
With respect to naturally occurring processes, dissolution is a common feature of diatom silica diagenesis 46        water fractionation factor measured in our experiments also appear to conform with the above-mentioned studies, though we add the caveat that our experiments were not optimised to precisely constrain the silica-water oxygen isotope fractionation factor due to the large volumes of diatoms cultured and the intensive oxidation of organic matter which was necessary prior to analysis. In contrast, post-mortem isotope fractionation appears to exhibit a temperature dependency of −0.36 ‰/°C (in the range 0-40°C), analogous to equilibrium silicawater oxygen isotope fractionation at high temperatures 9,14,41,42,[51][52][53] ( Figure 4). Sedimentary diatom silica values usually fall between these endmembers, suggesting either partial or complete oxygen isotope reequilibration within sediments 13,14,44 (Figure 6). Such re-equilibration will lead to higher δ 18 O silica values within sediments than in living diatoms, and a smoothing of seasonal patterns towards multiannual average conditions; however, further research is required in order to fully describe the rate and extent of secondary silica condensation and isotopic re-equilibration in natural and experimental settings.

| CONCLUSIONS
The