X ray fluorescence spectrometry (XRF) is one of the important methods to test the homogeneity of geochemical reference materials, but there is still controversy about the application. The homogeneity test requires the prerequisite that the sample weight is the minimum. The result of homogeneity test using conventional powder tableting or melting samples by XRF is generally larger than the minimum sample weight, so the results obtained are not theoretically sufficient to support the homogeneity of the sample under the condition of minimum sample weight. In this study, the 0.1 g samples experienced melted preparation with a mixed flux of Li₂B₄O₇ LiBO₂ and LiF (mass ratio 45∶10∶5) and NH₄I release agent. The authors established an experiential method to analyzetotally 10 measuring components, i.e., SiO₂, Al₂O₃, TFe₂O₃, MgO, CaO, Na₂O, K₂O, Mn, Ti and P. The correlation coefficient of the calibration curve of each component was between 0.997 3 and 1.000 0. The experimental optimization results of sample preparation conditions showed that the ratio of sample to mixed flux was 1∶4, 2 drops of 0.2 g/mL and ammonia iodide were used as the mold release agent, melting at 1 050 ℃ for 10 min., leading to the best molding effect. The method parameters were studied. The relative standard deviation of each component was 0.2%~5.3 %, and the relative error was less than 6.2%. The method precision and method accuracy were high. Compared with the conventional sample weighing of 0.65 g, the experimental results of the two methods are consistent. This study provides the basis for the application of X ray fluorescence spectroscopy to the homogeneity test of geochemical reference materials.