Evaluation on the Uncertainty Propagation With the Replacement of the Mercury Triple Point of ITS-90

The mercury triple point (Hg TP), one of the defining points of the international temperature scale of 1990 (ITS-90), has the distinguishing role for reducing Type 1 and 3 nonuniqueness with the subrange between the argon triple point (Ar TP) and the water triple point (TPW). However, mercury is highly toxic and will be banned from global commercial trade by the action of Minamata Convention on Mercury. The triple points of xenon (Xe TP), carbon dioxide (CO2 TP), and sulfur hexafluoride (SF6 TP) are the candidates for the replacement of the Hg TP. We presented in this study our analytical investigation of the propagation of uncertainties with the fixed points, the propagation of calibration errors for standard platinum resistance thermometers (SPRTs), and Type 3 nonuniqueness. We observed that, given the Hg TP as the reference, the interpolation formula using the single Xe TP for the replacement is capable for the reduction of 57% of the uncertainties of propagation and 48% of the errors of calibration for SPRTs. The formula using the single CO2 TP for the replacement is capable of the reduction of 31% of the uncertainties of propagation and 26% of the errors of calibration for SPRTs. Because of the lack of calibration data, we failed to get an entire insight of the effect on Type 1 and Type 3 nonuniqueness with the replacements of the typical interpolation formulas. Instead, we conducted a limit evaluation on Type 3 nonuniqueness only with three SPRTs calibrated at the CO2 TP. We observed no general effect on Type 3 nonuniqueness for the interpolation formula using the single CO2 TP.

Automated summary

This study investigates the replacements for the standard mercury triple point by analyzing the propagation of uncertainties, calibration errors of standard platinum resistance thermometers, and Type 3 nonuniqueness. It is found that using xenon or carbon dioxide as the replacement can significantly reduce the uncertainties and calibration errors. However, the effect on Type 1 and Type 3 nonuniqueness is not fully understood due to the lack of calibration data.