Non-linear mixing effects on mass-47 CO 2 clumped isotope thermometry: Patterns and implications William F. Defliese 1,2 * and Kyger C. Lohmann 1 1 Department of Earth and Environmental Sciences, University of Michigan, 1100 North University Ave, Ann Arbor, MI 48103, USA 2 Department of Earth, Planetary, and Space Sciences, University of California, Los Angeles, 595 Charles Young Drive East, Los Angeles, CA 90095, USA RATIONALE: Mass-47 CO 2 clumped isotope thermometry requires relatively large (~20 mg) samples of carbonate minerals due to detection limits and shot noise in gas source isotope ratio mass spectrometry (IRMS). However, it is unreasonable to assume that natural geologic materials are homogenous on the scale required for sampling. We show that sample heterogeneities can cause offsets from equilibrium Δ 47 values that are controlled solely by end member mixing and are independent of equilibrium temperatures. METHODS: A numerical model was built to simulate and quantify the effects of end member mixing on Δ 47 . The model was run in multiple possible configurations to produce a dataset of mixing effects. We verified that the model accurately simulated real phenomena by comparing two artificial laboratory mixtures measured using IRMS to model output. RESULTS: Mixing effects were found to be dependent on end member isotopic composition in δ 13 C and δ 18 O values, and independent of end member Δ 47 values. Both positive and negative offsets from equilibrium Δ 47 can occur, and the sign is dependent on the interaction between end member isotopic compositions. The overall magnitude of mixing offsets is controlled by the amount of variability within a sample; the larger the disparity between end member compositions, the larger the mixing offset. CONCLUSIONS: Samples varying by less than 2 ‰ in both δ 13 C and δ 18 O values have mixing offsets below current IRMS detection limits. We recommend the use of isotopic subsampling for δ 13 C and δ 18 O values to determine sample heterogeneity, and to evaluate any potential mixing effects in samples suspected of being heterogonous. Copyright © 2015 John Wiley & Sons, Ltd. Use of the mass-47 CO 2 clumped isotope thermometer has expanded greatly in the last few years, being applied to a diverse set of proxy materials. These materials include mollusks, [1–3] brachiopods, [4] corals, [5,6] speleothems, [7,8] foraminifera, [9] soil carbonates, [10–12] phosphate-associated carbonate, [13] and various diagenetic phases including phreatic and burial cements [14,15] and dolomite. [16,17] To achieve high-quality measurements, counting statistics requires large sample sizes, with typical replicates of 5–15 mg being analyzed multiple times for total sample sizes of up to 50 mg. [18] Alternatively, some researchers have applied techniques that use multiple (10+) aliquots of 150–200 μg of carbonate material, thus reducing the sample size to approximately 2 mg per replicate, although with a slight loss in sensitivity and throughput. [19,20] Most of the phases used in clumped isotope studies show isotopic heterogeneity on a small scale, such as the seasonal growth bands in many biogenic carbonates, and multiple stages of cementation in diagenetic carbonates, soil carbonates, and speleothems. Sampling procedures use either a drill or mortar and pestle to extract large quantities of sample powder, and can combine multiple growth bands or cement layers to reach the required amount. This leads to homogenization of an isotopically heterogeneous sample, each layer with potentially distinct values for δ 13 C, δ 18 O, and Δ 47 . Given that Δ 47 is calculated relative to the sample’s bulk δ 13 C and δ 18 O values, mixing phases with different compositions can have a dramatic effect on the calculated Δ 47 values, and thus on calculated paleotemperatures. Mixing relationships in multiply substituted isotopologues have been investigated by a number of authors, particularly focusing on the isotopologues of CO 2 [21–23] and N 2 O, [24] with only a few studies briefly considering CaCO 3 . [2,16] These studies have shown that mixing of two different populations of CO 2 (or N 2 O) causes deviations in measured isotope ratios from what would be expected from a linear mixing model. For example, Affek and Eiler [23] demonstrated that mixing of CO 2 from car exhaust and ambient atmosphere produced Δ 47 offsets of up to 0.042‰ compared with a linear mixing model, while Henkes et al. [2] suggested a maximum offset of 0.0014‰ in a hypothetical bivalve. In each of these examples, mixing effects would create an underestimate of the average equilibrium temperature, and they can be larger than typical analytical errors for Δ 47 measurement, such as the 0.017‰ standard deviation reported for interlaboratory comparison of NBS-19 by Dennis et al. [25] While it has been well documented that certain proxy materials, such as speleothems and corals, * Correspondence to: W. F. Defliese, Department of Earth and Environmental Sciences, University of Michigan, 1100 North University Ave, Ann Arbor, MI 48103, USA. E-mail: wdefliese@gmail.com Copyright © 2015 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2015, 29, 901–909 Research Article Received: 14 December 2014 Revised: 16 February 2015 Accepted: 18 February 2015 Published online in Wiley Online Library Rapid Commun. Mass Spectrom. 2015, 29, 901–909 (wileyonlinelibrary.com) DOI: 10.1002/rcm.7175 901