the fusion flux, cupel, and silver bead. The sums of the measured activities ranged from 97 to 103% of the standards. They differed from 100% because of the standard deviation associated with the level of 198Au activity in the silver bead. The crucibles were not examined for activity. Quartz was used to simulate the rock sample but this material contained negligible quantities of gold. The results indicate that a 97% or better recovery may be expected if 2 mg or more of silver are used during cupellation, a finding not significantly different from that reported by Fulton (15) and Coxon et al. ( 16) for larger amounts of gold. Since the recovery of gold is close to 100% and the standards, blanks, and samples are carried through the same procedure, no yield correction was made.Evaluation of the Gold Blank. Preliminary experiments indicated that the detection limit would be determined by the magnitude and reproducibility of the gold blank. The primary sources of gold contamination were the reagents used for the fire-assay flux and the clay crucibles used to fuse the sample. The contribution from both sources was about equal but varied considerably. The reagent blank was lowered by performing a preliminary fire-assay fusion on a modified flux yielding a flux composition close to that desired for the fire assay of the samples. The composition of a single charge of flux, before purification, is 100 grams of PbO, 25 grams of Na2C03, 5 grams of Na2B407, and 3 grams of flour. After purification, the composition is equivalent to 60 grams of PbO, 25 grams of Na2COs, 5 grams of Na2B407, and 0 gram of flour. The flux described here was satisfactory for most silicate rocks. If a sample requires a different flux composition, then adjustment should be made prior to purification.The clay crucibles, which are subjected to varying degrees of attack, were replaced by alumina crucibles. The attack on the alumina crucibles is considerably less and they can be re-used.The results of measurements on the gold blank from both the unpurified flux and two separately prepared portions of purified flux are shown in Table II. Sigma is the standard deviation of the average for the series of runs on each sample. If the limit of detection is defined as three times the standard deviation of the blank, then this technique should be able to detect 1.5 ng of gold or 0.1 ppb of gold in a 15-gram sample.Measurement of Gold in W-l. The procedure was tested on the U. S. Geological Survey's standard diabase, W-l, for which data have also been obtained by conventional neutron activation analysis (J, 17). This standard rock is supplied in bottles containing approximately 50 grams. Since 25-gram portions were used, the samples were not mixed nor quartered.The results, shown in Table III, indicate good agreement between these two quite different techniques.
ACKNOWLEDGMENTWe thank L. A. Harris and other members of the nuclear reactor staff, Naval Research Laboratory, Washington, D. C., for help in irradiating the lead buttons.