Supplementary Material to ’Robust Normal Mode Constraints on Inner Core Anisotropy From Model Space Search’ by C. Beghein and J. Trampert. Data

Data Two data sets were tested : the first one consisted of older normal mode splitting measurements (1, 2) (with 15 degree two and 7 degree four structure coefficients), and the second one consisted of the most recent splitting measurements that followed the great Bolivia and Kuril Islands earthquakes in 1994 (3–5) (with 22 degree two and 16 degree four data). The choice of data set one was based on the modes employed by Tromp (6) to derive his first model of inner core anisotropy. However, we decided to remove mode , due to a possible overlapping with mode , and , since it has been demonstrated (7) that different splitting functions could explain the spectrum of mode equally well. We also discarded the measurements made by Giardini, Li & Woodhouse (8) because their degree two splitting coefficients were systematically smaller than coefficients determined in other studies (1–5). In the end, data set one was composed of degree two measurements for modes , , , , , , , , , , , , , and ; degree four data were modes , , , , , and . Data set two was composed of degree two measurements for modes , , , , , , , , , , , , , , , , , , , , and , and degree four measurements for modes , , , , , , , , , , , , , , and . Original error bars for degree four data of mode were increased in data set two to account for discrepancies between the measurements. Modes , and were discarded as in data set one together with which is also difficult to measure.