The argon-39 method requires neutron irradiation of K-bearing samples to convert 39K to 39Ar. The conversion factor J is determined by irradiating standards of known K-Ar age in the same container as the unknown samples. Spatial variation of neutron flux is monitored by loading several aliquots of the chosen standard at different positions in the sample container and determining J for each. Values of J for the unknown samples are then derived either by linear interpolation between standards or by plotting a J-curve. The difference in J between the ends of the container may vary from one or two percent to as much as thirty percent in some reactors. To minimise J-variations at right-angles to the axis the irradiation cannister, enclosing several separate sample containers, is usually turned through 180° half-way through the irradiation period.
After irradiation, the Ar-39 method depends exclusively upon argon isotope analysis, so that excellent internal precision (typically ~ 0.1%) is routinely obtainable for a suite of samples from a single irradiation. This fact was exploited by Roddick (1983) to make precise intercomparisons amongst a group of four standards (Hb3gr, MMHb-1, LP-6, FY12a) with K-Ar accuracies of 0.28% to 1.08% (1s). Roddick adjusted the Ar/K of the standards to fit on a single J-curve and assigned to the adjusted values uncertainties of 0.05% to 0.25%, without stating whether these figures represent internal or external precision. Several subsequent users of Roddick's data, when standardising their irradiations, appear to have assumed that the figures in fact represent external precision or accuracy, and may therefore be used when comparing 39Ar ages with Rb-Sr and U-Pb results.
To explore these statistical questions we have developed a least-squares curve-fitting program targetted particularly at multi-standard irradiations. For each standard of which four or more aliquots have been irradiated a polynomial
J(x) = J0 + J1x + J2x2 is fitted by an analytic algorithm and uncertainties in the Ji are determined. The curves are combined by making appropriately weighted averages of the Ji. Applying this program to Roddick's data we find an external precision for the adjusted Ar/K ratios of around 0.3%, very similar to the accuracy given for the best of the four K-Ar ages. If we use more recent data for the conventional ages of the standards, MMHb-1 must be left out because of probable inhomogeneity, and the remaining three curves agree less well, suggesting either systematic error(s) in the conventional ages or another source of error in the irradiation parameters.
The need for reliable standards for the 40Ar-39Ar method is not yet properly satisfied. Homogeneity on a small scale is not always acceptable (e.g. MMHb-1), and there remain worrying inter-laboratory differences amongst independent K-Ar measurements on standards. Even if we are satisfied on those counts, inconsistencies which may exceed the conventional uncertainties are not unknown in multi-standard irradiations. Possible explanations of such uncertainties include recoil loss of 39Ar, which may be grain-size dependent; across-axis variations in neutron flux (perhaps generated by a difference in average flux before and after turning the cannister); and compositional influences upon the validity of the isotopic corrections which are integral to the method. Statistical analysis of more recent irradiations is in progress with the aim of clarifying these issues.
Roddick, J.C., Geochim. Cosmochim. Acta 47, 887-898 (1983).