Modeling and Measurement of 39Ar Recoil Loss From Biotite as a Function of Grain Dimensions

The call for age measurements with less than 1 per mil error puts a demand upon geochronologists to be aware of and quantify a number of problems which were previously negligible. One such factor is ³⁹Ar recoil loss during sample irradiation, a phenomenon which is widely assumed to affect only unusually small crystals having exceptionally high surface/volume ratios. This phenomenon has important implications for thermochronologic studies seeking to exploit a range of closure temperatures arising from variable diffusion radii. Our study focuses on biotite, in which spatial isotope distributions cannot be reliably recovered by stepwise heating and which therefore lack recoil-diagnostic age spectrum behavior. Previous work by Renne et al. [Application of a deuteron-deuteron (D-D) neutron generator to ⁴⁰Ar/³⁹Ar geochronology, Applied Radiation and Isotopes, in press] used the SRIM code to calculate a ∼20% ³⁹Ar recoil loss from the outermost 0.25 μ m of an infinite slab of phyllosillicate. This result is applied to measured grains of the biotite standard GA1550, a hypabyssal granite from the Mount Dromedary Complex, Australia. We measure the thickness and surface area of 166 grains and approximate the shape of each grain as a cylinder. Grain thickness ranges from 3 to 210 μ m, with an average grain radius of 350 μ m. We predict the amount of ³⁹Ar recoil loss from each grain, finding an expected age error >0.1 % for grains thinner than 150 μ m, a >1% error for grain less than 10 μ m thick, and up to a 3% error for grains less than 3 μ m thick. These modeling results will be tested by analysis of the measured grains after irradiation in the Oregon State University TRIGA reactor. It is important to either account for ³⁹Ar loss in thin biotite grains, or use sufficiently thick ones so that recoil loss is negligible. Our results indicate that only biotite grains thicker than 150 μ m should be used for neutron fluence monitoring in order to avoid bias greater than the limit of analytical resolution ( ∼0.1%). We will focus our discussion on: (1) evaluating the accuracy of the recoil model and (2) application of the results to thermochronology.