Accuracy of Be-9-data and its influence on Be-10 cosmogenic nuclide data

A Be-9-solution has been chemically prepared from phenakite (Be2SiO4) mineral grains as commercial Be-9-solutions are too high in long-lived Be-10. The solution is intended to be used as a carrier for radiochemical separation of Be-10 to be measured by accelerator mass spectrometry (AMS). Thus, accurate data of the Be-9-concentration of this solution is essential to guarantee for high-accuracy Be-10 data in the future. After devastating preliminary results (similar to 8 % standard deviation), eight laboratories finally produced twelve individual results by four different analytical methods. A certain lab and method bias might be identified by sophisticated statistical evaluation. Some laboratories also (grossly) underestimate their uncertainties. Thus, the simple weighted mean of this round-robin exercise needed to be corrected by introducing additional allowances (Paule-Mandel-approach). The final result has been calculated to (2,246 +/- A 11) mu g Be-9/(g solution) with a reasonably low weighted standard deviation of 0.49 %. The maximum deviation of a single lab value from the weighted mean is 2.4 % when removing one Grubbs outlier (11 % off from the mean) from the data set. As Be-10-data, which is usually calculated from measured Be-10/Be-9 by AMS and stable Be-9, cannot be more accurate than the determined Be-9-concentration, it seems highly advisable to establish or improve quality assurance by having self-made carrier-solutions analysed at more than a single lab and regularly taking part in round-robin exercises