Auxiliary Material for Assessment of relative Ti, Ta and Nb (TITAN) enrichments in ocean island basalts Bradley J. Peters*, James M.D. Day Scripps Isotope Geochemistry Laboratory, Geosciences Research Division, Scripps Institution of Oceanography, La Jolla, CA 92093-0208, United States of America *Corresponding author: Email: bpeters@ucsd.edu Phone: +1-858-534-8580 Submitted to: Geochemistry, Geophysics, Geosystems Introduction These supplementary materials provide additional information for methods discussed in the main text of the paper and cite data sources for the compilation highlighted in the main text. The authors consider this additional discussion prudent as it clarifies the treatment of a large quantity of published data. Four specific topics are addressed: 1. Data sources for OIB compilation (text, table S1) 2. Discussion of quantitative methods for determining compositions of parental magmas (text, figures S1-S5) 3. Iron oxidation state recalculation method from analytically determined total iron (text, figures S6-S11). 4. Discussion of partial melting-AFC model of trace element and radiogenic isotope compositions of OIB (text, tables S2-S4). Index of auxiliary files: TITANaux-text01 TITANaux-fs01 TITANaux-fs02 TITANaux-fs03 TITANaux-fs04 Contains all written supplementary information in titled sections with references to supplementary tables and figures Contains Supplementary Figure 1. Caption: Example of visual determination of reference MgO using Canary Island OIB. Black boxes represent areas in which the inflection point is expected to occur. Preferred reference MgO for this hotspot is 10 wt. %. Contains Supplementary Figure 2. Caption: Olivine compositions and corresponding whole-rock Mg# from hotspots. A number of olivines lie outside the narrow range of experimental equilibrium determined by Roeder & Emslie [1970]. For a more detailed explanation of the calculation of FeO from measured total Fe, please see Supplemental Information text. Contains Supplementary Figure 3. Caption: Comparison of parental TITAN anomalies determined from reference MgO determined by two methods: a visual estimate of reference MgO and olivine equilibria (see text for explanation of each method). Both methods result in similar calculated parental TITAN anomalies. Contains Supplementary Figure 4. Caption: Comparison of parental TITAN anomalies determined from visually-determined reference MgO contents and either direct or component calculation, as described in the text. Anomalies directly regressed against reference MgO values are systematically higher than those determined by regression of reference MgO against TITAN-involved elements and recalculation of parental TITAN anomalies (or the inverse of this). TITANaux-fs05 TITANaux-fs06 TITANaux-fs07 TITANaux-fs08 TITANaux-fs09 TITANaux-fs10 TITANaux-fs11 TITANaux-fs12 Contains Supplementary Figure 5. Caption: Parental TITAN anomalies (as in Figure 5) versus hotspot “scorecards” [Courtillot et al., 2003; Anderson, 2005] for a plume origin to OIB volcanism. All parental anomalies show poor to very poor correlations, indicating that the primitive component possibly proxied by a higher score is not present in the source of the positive TITAN anomalies. Symbols as in Figure 5. Contains Supplementary Figure 6. Caption: Relative abundance of FeO relative to total Fe in OIB and a subset of picritic (primitive) OIB. Contains Supplementary Figure 7. Caption: (a) Theoretical Fe redox state based on olivine accumulation in typical categories of geological materials. (b) Linear and quadratic approaches to recalculating Fe oxidation state and absolute deviation between the two methods. Contains Supplementary Figure 8. Caption: Propagated deviation on whole-rock Mg# for various values of total Fe content over a range of typical MgO contents for OIB using the two methods described in Supplementary Information text Percent deviation remains below 1% even for very high Fe contents. Contains Supplementary Figure 9. Caption: Comparison of methods of determining FeO (and, by this, Fe2+) from total Fe data when propagated to calculations of Mg# (Mg# = Mg / (Mg+Fe2+)). A 1:1 line (solid, bold) and other lines of constant ratio (dashed) are given with their respective slopes. Contains Supplementary Figure 10. Caption: Olivine and whole-rock data for global OIB plotted with experimental equilibrium bounds of Roeder & Emslie (1970). (a) shows Mg# calculated using FeO determined by the method described in Section 3 of the Supplementary Information; (b) shows Mg# using a constant ratio of FeO/(FeO + Fe2O3) = 0.9. Contains Supplementary Figure 11. Caption: Comparison of Mg# calculated from measured redox states of Fe and re-calculations of redox state using the method described in Section 3 of the Supplementary Information and a constant ratio of FeO/(FeO + Fe2O3) = 0.9. A 1:1 line (solid, bold) and other lines of constant ratio (dashed) are given with their respective slopes. Mg# calculated using the method described in Section 3 of the Supplementary Information are a closer statistical fit with measured data than are Mg# calculated using a constant ratio of FeO to total Fe. Contains Supplementary Figure 12. Caption: Re-ordering of trace element spider diagrams based on relative incompatibilities of elements in OIB with various accumulations of olivine, clinopyroxene and oxide phenocryst phases for (A) typical ordering, (B) oxide accumulation >3 vol. %; (C) <10 vol.%; (D) between >3 vol. % and (E) >20 vol. % olivine and clinopyroxene. Accumulation of oxides, which strongly prefer HFSE (Table S5, S7), generally cause TITAN anomalies of the highest magnitude. In all cases, re-ordering the elements based on overall compatibility results in Ta and Nb being shifted to the far left, indicating that they are the most incompatible elements in these systems; this new position means that Ta and Nb anomalies can no longer be calculated. Titanium shifts variably, and in all cases Ti/Ti* becomes closer to 1. These re-ordering options are not meant to suggest re-use, but show how modal abundance of mineral phases in OIB may cause relative elemental incompatibilities that differ from those presented in a traditional spidergram; the most appropriate ordering of incompatible trace elements may be closely linked to modal mineral abundances. TITANaux-ts01 TITANaux-ts02 TITANaux-ts03 TITANaux-ts04 TITANaux-ts05 Contains Supplementary Table 1. Caption: Table of the nine hotspots primarily studied and corresponding indexed references. References are listed in Supplementary Information text. Contains Supplementary Table 2. Caption: Example of the number of samples filtered using criteria mentioned in Supplementary Information. Some samples are excluded by multiple factors. Contains Supplementary Table 3. Caption: Example of TITAN calculations using various normalization schemes. PM: primitive mantle (McDonough, 2000), CI: chondrite Ivuna (Wasson & Kallemeyn, 1988); BE: bulk Earth (McDonough, 2000). The largest difference between PMnormalized TITAN values (preferred) and another normalization reservoir are for CI-normalized Ta and Nb anomalies, which arise due to the low Sm and Tb contents of CI relative to PM. Maximum variation is less than 20%. Contains Supplementary Table 3. Caption: Partition coefficients for numerical modeling of partial melting and references (bracketed). References: [1] Adam & Green (2006); [2] McKenzie & O’Nions (1991); [3] Elkins, et al. (1991); [4] Hart & Brooks (1974); [5] Kelemen, et al. (1993); [6] Day (2013). Literature compositions of primitive mantle as comparisons to the starting point selected for the models (McDonough, 2000). None of the calculations results in a TITAN anomaly that is substantially different from 1.00; in the case of Nb/Nb*, partial melt models show that the increase in Nb/Nb* well exceeds the difference from 1.00 shown here (Figure 7). TITANaux-ts06 TITANaux-ts07 Contains Supplementary Table 6. Caption: Model parameters for OIB hotspot localities. For an explanation of variables, please see the text of Supplementary Information. Contains Supplementary Table 7. Caption: Partition coefficients for numerical modeling of AFC processes and references (bracketed). References: [1] McKenzie & O’Nions (1991); [2] Villemant, et al. (1981); [3] Hart & Brooks (1974); [4] Nikogosian & Sobolev (1997); [5] Adam & Green (2006); [6] Sobolev, et al. (1996); [7] Heber, et al. (2007); [8] Day (2013); [9] Nielsen (1992); [10] Ewart & Griffin (1992); [11] Nash & Crecraft (1985); [12] Okamoto (1979); [13] Lemarchand, et al. (1987).