The triple point of water serves to define the kelvin, the unit of thermodynamic temperature, in the International System of Units (SI). Furthermore, it is the most important fixed point of the International Temperature Scale of 1990 (ITS-90). Any uncertainty in the realization of the triple point of water contributes directly to the measurement uncertainty over the wide temperature range from 13.8033 K to 1234.93 K.The Consultative Committee for Thermometry (CCT) decided at its 21st meeting in 2001 to carry out a comparison of water triple point cells and charged the BIPM with its organization.Water triple point cells from 20 national metrology institutes were carried to the BIPM and were compared with highest accuracy with two reference cells. The small day-to-day changes of the reference cells were determined by a least-squares technique. Prior to the measurements at the BIPM, the transfer cells were compared with the corresponding national references and therefore also allow comparison of the national references of the water triple point.This report presents the results of this comparison and gives detailed information about the measurements made at the BIPM and in the participating laboratories. It was found that the transfer cells show a standard deviation of 50 µK; the difference between the extremes is 160 µK. The same spread is observed between the national references.The most important result of this work is that a correlation between the isotopic composition of the cell water and the triple point temperature was observed. To reduce the spread between different realizations, it is therefore proposed that the definition of the kelvin should refer to water of a specified isotopic composition.The CCT recommended to the International Committee of Weights and Measures (CIPM) to clarify the definition of the kelvin in the SI brochure by explicitly referring to water with the isotopic composition of Vienna Standard Mean Ocean Water (VSMOW). The CIPM accepted this recommendation and the next edition of the SI brochure will include this specification.Main text. To reach the main text of this paper, click on Final Report. Note that this text is that which appears in Appendix B of the BIPM key comparison database kcdb.bipm.org/.The final report has been peer-reviewed and approved for publication by the CCT, according to the provisions of the CIPM Mutual Recognition Arrangement (MRA).

Two cobalt-carbon (Co-C) eutectic point (1,324 • C) cells and one palladium-carbon (Pd-C) eutectic point (1,492 • C) cell were constructed for thermocouple calibration. The lengths of the Co-C and Pd-C cells were 297 mm, 140 mm, and 140 mm, respectively. The melting and freezing plateaux at the Co-C and Pd-C eutectic points were observed using Pt/Pd thermocouples. The repeatability of the plateau, the effect of the surrounding temperature, and the temperature profile in the cell were measured, and the heat flux effect along the thermometer well was evaluated. When the plateaux of Co-C (297 mm height), Co-C (140 mm height), and Pd-C cells, were measured three times, seven times, and six times, respectively, the standard deviations of the melting points were 0.1 µV, 0.1 µV, and 0.4 µV, respectively. According to the temperature profiles along the thermometer well during the melting plateaux, it was found that the Pt/Pd thermocouple should be inserted at least 9.5 cm, 5 cm, and 6 cm below the surface of the eutectic alloys in the Co-C (297 mm height), Co-C (140 mm height), and Pd-C cells with the furnace set-point 16 • C above the melting point.

The impurity effect on fixed-point temperature realization by thermal analysis has been assessed. For such an assessment, the following actions were conducted: (1) the fabrication of aluminium point cells using 6N or higher-grade aluminium samples from different sources (manufacturers), (2) temperature measurements during solidification and thermal analyses based on freezing curves obtained from the measurements, (3) direct cell comparison among cells of different nominal purities and (4) calculation of the departure of the freezing point from the ideally defined freezing point by applying the sum of individual estimates (SIE). Two aluminium point cells were prepared in action (1) using 6N-grade and one cell using 6N5-grade aluminium samples. To realize a fixed point using the cells, a fixed-point furnace was developed and evaluated. Temperature measurements in action (2) were conducted at different rates of solidification and in accordance with the one using the liquid–solid interface technique. Gradients of freezing curves were derived in the thermal analysis, and from their dependence on the rate of solidification, the impurity effect was evaluated. Indirect cell comparison was also derived using the difference in the gradients. It was found that the indirect cell comparison was in satisfactory agreement with the direct cell comparison, which was obtained from action (3). It was also found that the departure of the thermal analysis from the SIE obtained from action (4) was within the uncertainty. This fact may imply a possible application of thermal analysis for estimating the effect of impurities in the realization of the aluminium point, especially for 6N-grade aluminium fixed-point cell as used in the present study.

In this article, work concerning the thermal analysis of the tin fixed-point is reported. First, the development of a new fixed-point furnace is described. Improvements in the design of the furnace and in the control system enable measurement of the heater power during the phase change. The furnace is sufficiently thermally insulated to produce excellent uniformity and stability, leading to high quality freeze-initiation and minimal thermal influences on the freezing point. By employing the improved furnace and newly-fabricated tin fixed-point cells, the start and end of the melting plateau and the end of the freezing plateau were accurately determined, enabling reliable evaluation of the liquid fraction during the realization of the tin fixed-point compared to conventional methods. Two open-type tin fixed-point cells were fabricated using highpurity tin that was chemically analyzed for impurity content. Thermal analysis results of freezing-point depression are compared to those based on the chemical analysis.

CCT-WG1 has recommended the sum of individual estimates (SIE) method to correct for the influence of impurities on the realization of temperature fixed points when a detailed impurity analysis is available. The method to estimate the uncertainty of the SIE has also been reported. On the other hand, most cells are fabricated from commercial fixed-point metals that often have no detailed impurity analysis, so the SIE calculation is impossible in that case. Due to this circumstance, and with the focus on the silver fixed point, a new fixed-point cell was fabricated in such a way that a portion of the silver ingot used was extractable during the silver casting. This portion was then analyzed by glow discharge mass spectrometry (GDMS), and the result used to calculate the SIE correction and its uncertainty. Temperature measurements during melting and freezing were collected using new and existing silver fixed-point cells under various conditions. These measurements were used to derive the slope of the silver freezing curve, from which the effect of impurities was evaluated by thermal analysis. The difference between the SIE and the thermal analysis method was evaluated to check the inaccuracy of the thermal analysis from the SIE point of view.