Abstract: In view of the serious mercury memory effect and the complicated elimination pretreatment process in determination of mercury element by inductively coupled plasma mass spectrometry (ICP-MS), this study was proposed. 0.500 0 g of sample was taken and placed in a polytetrafluoroethylene digestion tank, and 5 mL of nitric acid and 1 mL of 30% (mass fraction) hydrogen peroxide solution were added. The mixture was heated at 100 ℃ for 30 min in a constant temperature digestion instrument and cooled. The digestion tank was sealed and placed in a microwave digestion instrument. After digestion, the digestion tank was cooled and opened, and placed at 100 ℃ for 15 min in a constant temperature digestion instrument. The digestion solution was taken and placed in a 25 mL-volumetric flask, and the digestion tank was washed several times with water. The washing solutions were combined with the digestion solution, and the solution was made up to 25 mL with water. The 20 μg·L^{−1 186}Re internal standard solution, which was prepared with cleaning reagents 2 mg·L⁻¹ gold element standard solution or 2% (mass fraction, the same below) L-cysteine solution, was mixed online with the sample solution for determination by ICP-MS. As shown by the results, the cleaning reagents could eliminate the mercury memory effect of 0.5−5 μg·L⁻¹ mercury element standard solution, without affecting the determination of the next sample. Linear relationships between the ratios of the corresponding signal intensity to the internal standard signal intensity and mass concentrations of mercury element were found in the range of 0.5−5 μg·L⁻¹, with detection limit (3s) of 0.001 0 mg·kg⁻¹ by using 2 mg·L⁻¹ gold element standard solution to prepare an internal standard solution, and detection limit (3s) of 0.001 2 mg·kg⁻¹ by using 2% L-cysteine solution to prepare an internal standard solution. Test for recovery was made by the standard addition method, giving results in the range of 90.2%−104%, with RSDs (n=6) of the determined values less than 9.0%. This method was used for the analysis of 3 commercially available cosmetics, and the detection amounts, which obtained from the internal standard solution prepared by two cleaning reagents, were consistent and in the range of 0.008 0−0.13 mg·kg⁻¹.