Table of contents

Recent findings in precision platinum resistance and radiation thermometry at the Istituto di Metrologia "G. Colonnetti" are described. Two issues concern high-temperature platinum resistance thermometers. The first is the use of gas-controlled heat pipes for comparing these thermometers and the development of a novel approach to improve temperature stability and uniformity. The second deals with observations made at the freezing point of silver which suggest that humidity may lead to electric leakage. In radiation thermometry the main recent achievement has been the use of a photodiode-array thermometer to realize the ITS-90 above the silver point. Relative to single-detector thermometers, this new thermometer offers several advantages that are related to its spectral characteristics. Another issue of importance is the size-of-source effect (SSE). Practical rules for minimizing SSE errors in interlaboratory comparisons are given.

A brief summary is given of the developments in optical techniques for rapid temperature measurement during the last three decades in two national laboratories, the National Institute of Standards and Technology in the USA and the Istituto di Metrologia "G. Colonnetti" in Italy. Results of research conducted in originating and advancing the state-of-the-art in high-speed (millisecond- and microsecond-resolution) pyrometry for measurement of temperatures in the range 1000 K to 5000 K are discussed. The main emphasis is on several key issues related to high-speed pyrometry including: high-speed operation, calibration of pyrometers, determination of true temperature and high-temperature reference points. Anticipated future directions in research, developments and applications in high-speed pyrometry are briefly discussed.

We have measured the vapour pressure of caesium over the temperature range 370 °C to 660 °C using a pressure-controlled heat pipe. The equation log (p/Pa) = a + b/T + c log(T/K) + dT + eT² with the coefficients a = 34,573 234, b = -4979,5799 K, c= -9,323 424 7, d = 4,473 313 2 × 10^-3 K^-1 and e = -8,684 092 × 10^-7 K^-2 fits the data to within ±8 mK over the entire range. In addition, the non-uniqueness of the ITS-90 appears to be less than ±1,5 mK over the same temperature range based on the data from three thermometers.

The freezing temperature of a pure substance varies with pressure. In the case of a vertical solid/liquid interface as in a typical thermometric-quality freezing-point cell, where the hydrostatic pressure increases with the depth along the interface, successful tracking of the associated temperature gradient by an immersed platinum resistance thermometer is an accepted criterion for ensuring that the sensor is in adequate thermal equilibrium with the interface. In practice, however, the sensor is at a finite distance from the interface because the thermometer stem that contains the sensor is inside a thermometer well that is immersed in the solid or liquid phase. In consequence, it is the temperature gradient along the boundary between the well and the adjacent phase that is measured rather than that along the interface. The two gradients are not necessarily identical; we wish to determine whether the difference between them is sufficiently large that account should be taken of it in applying the criterion. In this paper we analyse numerically the heat flow in the adjacent phase so as to obtain an estimate of the difference between these gradients. It turns out that the temperature distribution is independent of the thermal conductivity of the substance. The difference between the two gradients is computed to be non-zero, but so small as to be irrelevant in practice.

Since Johnson's experimental observations of thermal noise in 1927, and Nyquist's explanation of the phenomenon shortly afterwards in 1928, thermal noise has attracted interest as a means of measuring temperature. The independence of the thermal noise from the material nature of the sensor makes it particularly attractive for metrological applications. However, the noise signals are extremely small and some ingenuity is required to make accurate measurements. This paper reviews the foundations of Johnson noise thermometry and the various techniques that have been employed to measure temperature via Johnson noise. Emphasis is placed on key developments in noise thermometers for metrological applications. The review includes the current activities of teams involved in noise thermometry research.

A newly developed Au/Pd thermocouple has shown very high reproducibility and good immersion characteristics. A stability test of the Au/Pd thermocouple has been conducted for 1000 h at 1000 °C in air to assess its suitability for practical application. Two thermocouples from two manufacturers were fabricated and their emf values were checked in a silver freezing-point cell before and after the soaking at 1000 °C. It was found that the high-temperature stability of the Au/Pd thermocouple is within ±30 mK (2σ) throughout the test time. This result implies that the Au/Pd thermocouple possesses characteristics which may make it suitable for use as a secondary high-temperature thermometer in industry.

Primary dielectric constant gas thermometry (DCGT) has been used to establish a quasi-continuous temperature scale in the range 4,2 K to 27,0 K with a measurement uncertainty which increases with rising temperature from 0,6 mK to 1,2 mK (confidence level 95%). (It should be noted that previous papers refer to the 68% level which yields about one half of the uncertainty values.) The large number of experimental data for ⁴He (about thirty isotherms, more than 200 triplets of pressure, temperature, and dielectric constant) allowed, furthermore, the determination of the temperature dependencies of the second and third virial coefficients, without any constraints. The virial coefficients thus obtained exhibit good agreement with data based primarily on conventional gas thermometry and potential calculations. The established temperature scale is compared with the isotherm and constant volume gas thermometry scale NPL-75. A flat sinusoidal form for the differences between the two scales may be seen, which is in accordance with other published data. However, the differences are well within the thermodynamic uncertainties estimated for the NPL-75 and the DCGT scale. This result supports the theoretical polarizability value with a relative combined uncertainty of 6 × 10^-5, i.e. the bounds for the polarizability value of ⁴He have been decreased by one order of magnitude compared with the available rigorous theoretical bounds.

The ITS-90 realized using high-temperature standard platinum resistance thermometers (HTSPRTs) at the National Physical Laboratory (NPL) was compared with the radiance temperature scale obtained using the NPL primary pyrometer. The scales were compared in the range 600 °C to 962 °C. The results show that the two scales generally agreed to within 2σ (about 0,1 °C) of the total combined uncertainty of the measurements, with the majority of the measurements agreeing to within 1σ. The ITS-90 value of the freezing point of silver, measured relative to that of aluminium, is confirmed within the uncertainty of the measurement.

Most of the ITS-90 temperature fixed points are established by the melting or freezing of a specified substance. The question which may be asked is how closely this definition leads to a temperature value corresponding to the ideal thermodynamic state of the phase change temperature of a substance containing a small amount of impurities. In the light of experiments performed in the low temperature range it seems possible to use a calorimetric method to realize the temperature fixed points closer to the thermodynamic definition than by using the actual ITS-90 definition. This technique could lead to a more reliable (or more reproducible) temperature and could be a worthwhile improvement of the ITS-90.

This paper describes the realization of the International Temperature scale of 1990 from 1,2 K to 2473 K at the National Institute of Metrology (NIM), China, and gives the uncertainties in each range of the scale. Results of recent research in thermometry at the NIM are also reported. The non-uniqueness of SPRTs over the temperature range 13 K to 419 C and the nonlinearity of radiation thermometers have been investigated at the NIM. The results of international comparisons of the triple points of water and freezing points of silver between the NIM, the National Institute of Standards and Technology (NIST), USA, and the Nederlands Meetinstituut (NMi), the Netherlands, are presented.

The principles of total radiation thermometry as a method of primary thermometry are presented. Their practical application using a cryogenic radiometer is described with an account of the principal sources of uncertainty. It is shown that an uncertainty of about 0,3 mK may now be expected in the measurement of T in the room temperature range if advantage is taken of new commercial black surface preparations and new ideas in the design of black-body enclosures. Included also are comments on the use of the same equipment for a determination of the Stefan-Boltzmann constant with an estimated uncertainty approaching 1 part in 10⁵.

The ITS-90 is defined in the ranges 0,65 K to 3,2 K and 1,25 K to 5,0 K by the vapour-pressure/ temperature relations of ³He and ⁴He, respectively. An apparatus has been constructed at the National Institute of Standards and Technology (NIST) for realizing the ITS-90 below 84 K which includes cells for realizing the ITS-90 through vapour-pressure thermometry with ³He and ⁴He. We present here the results of our realizations with He vapour pressures. The expanded uncertainties (k = 2) of the ITS-90 temperature realizations are 0,12 mK or less over 97% of the ranges of the ITS-90 definitions. In the lower 3% of each range, the uncertainties of the realizations increase to 0,14 mK for ³He and 0,16 mK for ⁴He because of the growing size of the thermomolecular pressure correction. Comparisons are made with the previous wire scale of the NIST, which is traceable to the T_X1 and NPL-75 scales. Also presented are the results from direct comparisons of ³He and ⁴He realizations between 1,25 K and 3,2 K; these show the non-uniqueness of the ITS-90 in this region to be less than 0,3 mK.

The contribution of the National Measurement Laboratory (Australia) to the International Temperature Scale of 1990 (ITS-90) is stated briefly. Research relating to the measurement of thermodynamic temperature over the range 1 K to 1064 °C, improvements in the realization of fixed points and interpolation instruments are described and related to the content of the ITS-90.

The triple points of water, mercury and gallium are realized, mainly for the calibration of capsule-type thermometers, by a calorimetric method using platinum cells for water and mercury, and a small glass cell for gallium. Their molar impurities are estimated from the relation between the melting temperatures and the inverses of the fraction of melt. Almost all the samples show linear relations. For water, the values at 1/F = 0 of all the samples agree to within ±0,1 mK and, for mercury, the values of two samples are within ±0,05 mK. It is pointed out that the platinum cell is available for water and mercury, but is unsuitable for gallium.

The vapour-pressure versus temperature equations for liquid ⁴He and ³He that define the International Temperature Scale of 1990 (ITS-90) in the range 0,65 K to 5,0 K have been derived from earlier equations in the Echelle Provisoire de Temperature de 1976 (EPT-76), which span the wider range 0,2 K to 5,2 K. This paper investigates the extent to which, i.e. to within what uncertainty, the equations can be obtained from thermodynamic calculations using the most recent pVT and thermal data for the liquid and gas phases of ⁴He and ³He. Regarding the ³He vapour-pressure equation, the paper is an extension to higher temperatures of the paper by Fellmuth and Schuster which treated the important region below 1,5 K. The paper also deals with vapour pressures near the ⁴He λ-transition and near the critical points.

Since the International Temperature Scale of 1990 (ITS-90) was defined, there have been a number of significant developments in thermodynamic thermometry. New results have been published and further experiments are in progress or are planned. This special issue of Metrologia provides an appropriate occasion to review these developments. The new results show some differences between the ITS-90 and thermodynamic temperature, but they are generally within one standard uncertainty of the values assigned in the ITS-90. Only in the region of 150 K have larger differences been reported. Further results must be awaited before firm conclusions regarding a new consensus for thermodynamic values can be drawn.

The realizations of the fixed points of the International Temperature Scale of 1990 (ITS-90) maintained in Canada and Mexico were compared between -38 °C and 420 °C, using a portable fixed-point realization apparatus. In this range, the routinely realized defining fixed points of the ITS-90 in these countries were found to coincide to within 0,5 mK.

Double freezing curves of oxygen have been obtained in a cooling process with a rate of less than 40 mK/min. The plateaus of these double freezing curves show differences in temperature and width. The cause of the observed double freezing curve is presumably redistribution of impurities into relatively pure and impure regions by transportation during the slow cooling process. This separation is observed in the melting curve of oxygen using both continuous and pulse heating methods. We note the importance of impurity redistribution on the accuracy of the temperature measurement and the realization of the triple point of oxygen.