Abstract
The triple point temperatures of pure mercury, water and gallium are estimated by a thermal analytical method using mercury samples with nominal purities of 7N and 9N, gallium with 7N and water with 5N. The differences between the triple point temperatures realized by the pure samples and by the conventional triple point equipment are less than 0.1mK, which is estimated from the relation of melting temperature vs. the inverse of the fraction of melt. Concerning the triple point of water, the main source of the uncertainty is the ambiguity of the isotopic composition, rather than the error from impurities. The uncertainty of the unit of temperature, kelvin, is in the same situation. The definition of the unit by the gas constat has now a high potential to replace the present definition of the kelvin.
