[EEL TKANSACTIONS ON llVSTRUMENlKl10N AND MEASUREMENT, VOL. 44, NO. 2. APRIL 19%

~

515

The ITS-90 Realization: A Survev L. Crovini and

Abstruct-The present state of the realization of the Interna- tional Temperature scale of 1990 is illustrated using the results of a recent inquiry conducted by the Consultative Committee for Thermometry of the CIPM among a sample group of metrology laboratories.

I. INTRODUCTION RECISION temperature measurements are necessary in P several scientific disciplines, chiefly in mechanical and

electrical metrology for the very accurate control and determi- nation of the temperature of the standards. Room temperature determinations is mostly required. However, the exploitation of superconductivity and of quantum Hall resistors requires the uniform and reliable realization of a temperature scale in the liquid helium range.

Since the introduction of the International Temperature Scale of 1990 (ITS-90) [ I], the national laboratories involved in temperature measurements have made great efforts to realize the new scale. In 1993, the Consultative Committee for Ther- mometry (CCT) of the CIPM carried out an inquiry among the participating national laboratories to assess the state of the ITS-90 realization [ 2 ] . The group includes 14 laboratories with responsibility for setting up the national measurement system in their respective countries. Although the ensemble of their answers cannot be taken as a representative sample of the situation world-wide, the analysis of these answers provides valuable information on the progress in the ITS-90 realization and the most relevant problems stemming from it. Fig. 1 presents the overall result.

11. THE STRUCTURE OF THE ITS-90

The scale entails several, partially-overlapping ranges. Dif- ferent interpolating instruments are used over them. These are as follows: 0.65 K to 3.2 K-'He vapor-pressure thermometer: 1.25 K to 5 K-4He vapor-pressure thermometer; 3 K to 24.6 K-both "e and 'He interpolating, constant-volume gas ther- mometers (CVGT's); 13.8 K to 1235 K-platinum resistance thermometer (PRT); above 1 235 K-monochromatic-radiation themlometer. In its tum, the realization of the ITS-90 by means of the PRT is carried out over ' 1 1 subranges, all of them including the triple point of water (O.Ol°C). They too overlap. There are, in total, 18 defining fixed points, which constitute a redundant set of calibration points. By this method, interpolating instruments of the above types and of the required quality can be calibrated in a flexible way in

Manuscript received July I , 1994; r e v i d October 15, 1994. The authors are with CNR, Istituto di Metrologia G. Colonnetti, Strada

IEEE Log Number 9409309 delle Cacce 73, 1-10135 Torino, Italy.

J

P. P. M. Stem

.! - ' 1

2 2 NRLM

m N NPL

: NET

! NIM

.4

R

KRISS

INM i ; f IMGC U CSlRO b

r r r

)i / ainpadiua!

Fig. 1. ITS-90 realization (in black).

Range of temperature covered in the various laboratories with an

order to obtain temperature values as precise and reproducible as required by the most demanding users, while at the same time approximating the corresponding thermodynamic values as closely as possible.

111. TEMPERATURES BELOW 24.6 K

In this temperature range, and more evidently below 14 K, the realization of the ITS-90 is slower. The effort required in order to realize, for instance, a CVGT operating between 3 K and 24.6 K is substantial. The thermodynamic Constant Volume Gas Thermometer behaves in the temperature range between 3 K and 24.6 K and over a wide spectrum of specified parameters 131 in a known and regular way, such that this range can be covered with a quadratic interpolation function requiring three calibration points and with an uncertainty not exceeding 0.5 mK. After the introduction of the ITS-90 only one laboratory (KRISS) appears to have an operating CVGT (4He) to date, although with an uncertainty of 3 mK. Three oIhei laboratories have projects in an advanced stage of development, i.e., IMGC ('He), NIST (4He), and PTB

0018-9456/95$04.00 0 1995 IEEE

576 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT. VOL. 44, NO. 2, APRIL 199.5

(4He). NRLM has at present a .5He CVGT in development for thermodynamic temperature measurement. Its evolution to become a primary interpolating instrument may be envisaged.

To date, the international traceability in temperature mea- surements below 24.6 K greatly depends on the existence of approximate realizations of the ITS-90, currently referred to as secondary scales [4]. In most cases, rhodium-iron resistance thermometers are used. Their calibration was carried out with reference to the EPT-76, a provisional scale closely matching the ITS-90 below 4.2 K and presenting a known difference from it up to 24.6 K [3]. The accuracy of such secondary scales critically depends on the thermometer stability. A similar approach, albeit not as accurate, is provided using germanium resistance thermometers.

There is a growing demand for an internationally-agreed temperature scale below 0.65 K. that would extend to the niillikelvin region. This demand stems from the need to ensure the uniqueness of low-temperature physics results. One proposal is to resort to the use of the 3He melting curve [ 5 ] .

IV. TEMPERATURES FROM 24.6 K TO 1235 K

This temperature range, covered by the Platinum Resistance Thermometer, is the most realized in the laboratories. Its 11 subranges offer great flexibility of realization according to the different applications. In contrast to the previous scale, the difficult realization of the two vapor-liquid equilibrium points of equilibrium hydrogen (ca 17 K and 20.3 K, respectively) can be avoided when measurements below 24.6 K are not envisaged. However, five laboratories out of 14 maintain the PRT calibration facility down to 13.8 K. Furthermore, since the exposure of a PRT to a high temperature affects its stability, the best accuracy in the measurement of medium temperatures, 01 of room temperature, is achieved using the appropriate calibration rubrange among the many available, an option not provided by the previous scale. Thus, since the laboratories started with the ITS-90 realization with the aim to respond to the more pressing needs, it is not surprising that almost all have reached the argon triple point (83.8058 K) and quite a few the neon triple point (24.5561 K). For the same reasons, the subrange -383°C to 29.7"C and the 0-29.7"C, 0-156.6"C, 0-23 I .9"C, and 0-419.5"C subranges have almost universally been realized.

The ITS-90 has introduced resistance thermometry above 903 K and, consequently, the measurement capability of several laboratories has been upgraded for what concerns both the resistance measuring equipment and the fixed point apparatus. This effort resulted in the ITS-90 realization up to both the freezing point of aluminium (660.323"C) and the freezing point of silver (961.78"C) in 1 1 cases out of the 14 examined.

There are not many commercial sources of high quality PRT's. The types in use at both low and medium tem- perature have remained substantially unchanged for many years. Though their stability is generally satisfactory, their shape and size is not always best fitted to their use. For instance, high-stability, miniaturized capsule types would be beneficial for accurate room temperature measurements in

C U a w

Fabrication Fig. 2. Type and range of use source of high-temperature, long-stem PRT's.

length metrology. In the case of high temperature PRT's the situation is more serious. Their stability and somewhat limited life concern both the national laboratories and the users. The contamination of the platinum coil with metals diffusing through the silica sheath was another concern that prompted specific investigations. some of them currently under way. Fig. 2 summarises the situation in the laboratories for the calibration ranges of high temperature PRT's, their resistance values, and their fabrication or commercial sources.

V. ROOM-TEMPERATURE MEASUREMENTS There are three subranges more specifically suited to room-

temperature measurements: (a) -383°C to 29.7"C; (b) 0°C to 29.7"C; and (c) 0°C to 156.6"C. All three subranges entail the use of defining fixed points of very high reproducibility: all the laboratories that have realized them estimate for their reproducibility a standard uncertainty not exceeding 0.2 mK, with the exception of one case for the triple point of mer- cury where a substantially larger uncertainty was estimated. Subrange (a) provides the best reproducibility both below and above the triple point of water and ensures an almost perfect smoothness through this temperature range; ten laboratories have realized it. The adoption of either the subrange (a) or (b) greatly improves the accuracy of the temperature measure- ments necessary to reproduce material standards in electrical, dimensional and mass metrology and to carry out very high accuracy measurements, such as the determination of silicon lattice spacing by means of X-ray interferometry; practically all laboratories have realized it. Range (b) can also be realized by any user possessing the two required fixed-points, i.e., the cells for the water triple point and the melting point of gallium. Eleven laboratories have realized this subrange. Lastly, subrange (c) enables capsule-type PRT's to be used in a broader room-temperature range still preserving the integrity of the capsule and minimizing the effects of vacancy anneal and oxidation of platinum. Eleven laboratories have realized it.

It is often suggested that the 0-29.7"C subrange be ex- clusively adopted for room temperature measurement in me-

CROWN1 AND STEUR: THE ITS-90 REALIZATION: A SURVEY 571

chanical and electrical metrology. The standard uncertainty (one standard deviation) of a temperature calibration, ut, as estimated for a measurement close to 20"C, is expressed by the following approximate equation:

where W20 and WG, express the PRT resistance ratio at 20°C and at the gallium melting point, respectively; ut(Ga) and ut(O.OloC) are the standard uncertainty of the gallium melting point and the water triple point calibrations, respectively. With a state-of-the-art realization of the triple point of water and of the melting point of gallium, the results from the inquiry [2] show that:

ut(O.OloC) = 0.03 mK ut(Ga) = 0.03 mK

whence ut (c? 2OoC = 0.04 mK. To fully exploit such an accuracy the resistance-ratio measurements have to be carried out to better than A further uncertainty of about the same value may come from the ITS-90 nonuniqueness in this subrange, when only a realization with the triple point of water and the melting point of gallium is considered, forcing a linear interpolation, thus raising the total standard uncertainty to 0.05 mK.

VI. TEMPERATURES ABOVE 1235 K Where the ITS-90 is defined in terms of the Planck

radiation-law a reference black-body source is needed. The freezing points of silver, gold, and copper can variously be used for that purpose. At the time that the introduction of the ITS-90 was discussed, some concern was raised as to the accuracy and durability of a blackbody source at the freezing point of silver. Molten silver may dissolve oxygen, which depresses its freezing point. In such an application there is no way to keep silver in a sealed enclosure that is accurate enough. This concem was, in fact, dismissed on the grounds of the results of subsequent experiments in several laboratories. Nowadays, the realization of a black-body source at the silver freezing point is preferred in about 60% of the cases. But also the other two altematives are used: the freezing point of gold, when it was already available, and the much cheaper freezing point of copper. The three fixed points can be realized with comparable accuracy so that the most convenient among them can be chosen. Such a flexibility, however, results in increasing nonuniqueness: for instance, when using a blackbody source at the freezing point of gold, the radiation-thermometer branch of the ITS-90 may not meet the PRT branch at 961.78OC, the freezing point of silver, although a difference larger than 20 mK is unlikely.

VII. REPRODUCIBJLITY OF THE DEFINING FIXED-POINTS The participating laboratories have provided their own es-

timate of the reproducibility of the realized fixed points. Data have been provided for all the 17 defining fixed points of the ITS-90. On average, the values of the standard uncertainty due to reproducibility ranges from 0.2 mK below 84 K to about 25 mK for the freezing point of silver used as a blackbody

50 40 30 U) 10 0

Oxygen Triple Point og-.15 ,120

1-2 2 4 4-8 8-16 PPm ppm PPm PPm

Mercury Triple Point .l-.2

m .?!

Argon Triple Point .1-.2 :k 20

0 5 1 1-2 PPm PPm PPm

Water Triple Point .06-.1

50 , ."M mK - . l- .2

I -- 1 GalliumMeltingPoint I 1 TinFreezingPoint 1

I .l-.2 .2-.4 .4-.8 1 1 .2-.4 .4-.8 3-1.6 > 1.6 PPm PPm PPm PPm PPm PPm PPm I

Zinc Freezing Point Aluminium Freezing

1 < .2 .2-.4 .4-.8 3-1.6 71.6 1 1 " <.5 5 1 1-2 2 4 1 PPm PPm PPm ppm PPm PPm PPm Ppm PPm

Fig. 3. Frequencies, in percent, of the estimated reproducibility of some defining fixed points of the ITS-90 as declared by the interviewed laboratories. The reproducibility is expressed in parts per million of the Kelvin temperature and in mK..

source, with a minimum of 0.03 mK at the water triple point. However, the values supplied for every particular fixed point show a wide range, much larger than expected. Fig. 3 comparatively presents the declared values of the fixed points that have more frequently been realized. Values ranging over one decade or more are found for some of them, a situation also occurring for other fixed points not shown in the figures. Differences in capabilities and, also, different interpretations of the reproducibility concept may explain these results.

The term reproducibility and the ensuing uncertainty may not be a definition as precise as needed for fixed-point ther- mometry. Quoting from [6], reproducibility means closeness of the agreement between the results of measurements of the same measure und carried out under changed conditions of measurement. But, which changed conditions of measurement have to be considered with a thermometric fixed point? Is it enough to reproduce, for instance, a freezing plateau with a single selected cell or are all factors influencing the repro- ducibility of a given fixed point, as realized according to the current procedure of the particular laboratory, to be accounted for as well? In the second case, the selection of high-purity samples of the thermometric substance and the particular cell-

57s IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 44, NO. 2. APRIL 1995

preparation technique have to be considered. In the first case, the reproducibility would be estimated from the properties of a single fixed-point cell whereas in the second one it would be derived from a group of nominally equal cells. The CCT will have to clarify this matter in order to better evaluate the results of future interlaboratory comparisons.

VIII. APPROXIMATE REALIZATIONS OF THE ITS-90 AND TRANSFER STANDARDS

are keen to supply their users, chiefly their accredited cal- ibration laboratories, with one or more alternatives to the high-temperature PRT. Gold-platinum and platinum-palladium thermocouples can offer accuracy of transfer of 0.1 K or better, moderate cost, lower sensitivity to temperature gradi- ents and better stability in the high temperature environment than, for instance, platinum-rhodium thermocouples. For these last, however, improved, ITS-90-based reference tables have recently been produced [7].

The national laboratories appear to resort to an approximate realization of the ITS-90 essentially in four circumstances: I x . LABORATORIES CONSIDERED IN THIS SURVEY I ) below 24.6 K in the absence of a CVGT and/or of an helium vapor-pressure scale; 2) in the PRT range, adopting special subranges without one of the fixed points required for the corresponding official subranges or converting IPTS-68 calibrations; 3) extending the use of the PRT up to the freezing point of copper (1084.62”C); and 4) extending the use of the radiation thermometer below 96 1°C. Although 2) and, to some extent in l), can be considered as temporary measures, the other secondary scales respond to permanent practical needs. For instance, a radiation thermometer interpolating between four calibration points is used to provide traceability to a class of radiation thermometers with an accuracy not otherwise attainable.

For the dissemination of the ITS-90 several, high-quality transfer standards are needed. Only the PRT is at the same time an ITS-90 interpolating instrument and a transfer standard.

CSIRO, Division of Applied Physics (Australia). KRISS, Korean Research Institute for Standard and Science (Rep. of Korea). IMGC, Istituto di Metrologia “G. Colonnetti,” CNR (Italy). INM, Institut National de Metrologie, CNAM (France). NIM, National Institute of Metrology, Heat Division (People’s Rep. of China). NIST, National Institute of Standards and Technology (USA). NPL, National Physical Laboratory (United Kingdom). NRC, Institute for National Measurement Standards (Canada). NRLM, National Research Laboratory for Metrology (Japan). PTB, Physikalisch-Technische Bunde- sanstalt (Germany). SMU, Slovak Metrology Ustav, Tempera- ture and Heat Department (Slovak Rep.). VNIIFTRI, Cryomet Division (Russia). VNIIM, The D. I. Mendeleyev Institute for Metrology (Russia). VSL, van Swinden Laboratorium, NMI (The Netherlands).

Therefore, special transfer standards are used both below 24.6 K and above 1235 K. As already seen, Rh-Fe resistance thermometers are the most accurate transfer standards in use

REFERENCES

[ I ] H. Preston-Thomas, “The international temperature scale of 1990,” below 24.6 K. Their resistance must be measured with an uncertainty lower than O.5.1Oe6 in order to ensure a transfer

Metrologiu, vol. 27, pp. 3-10, Jan. 1990. [2] L. Crovini and P. P. M. Steur, “ITS-90 realisation in 14 national

standards laboratories. Results of a CCT inauirv.” Doc. CCT/93-1(Rl ~ /.

to within 0.1 mK, a goal attainable with the modem electrical measuring instrumentation. Quite different is the of

ribbon lamps, transfer standard pyrometers, blackbody sources

than the accuracy of the ITS-90 realizations in the national laboratories. Without the development of better transfer stan- dards it is not possible to compare different realizations of the ITS-90 above 1235 K exoloitine their best accuracv.

in 18th Meeting CCT, BIPM, S h e s , Franc:, 1593.

France, 1990, pp. 6 2 8 and pp. 130-142.

scale of 1990,” BIPM. S2vres. France, 1990, pp. 42-87.

meltingpressure scale.” Doc. CCT/93-8, 18th Meeting CCT, BIPM, SBvres, France, 1993.

[61 BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML, "International vocabu- lary of basic and general terms in metrology.” ISO, Geneva, Switzerland, 1993, Sec. 3.7.

171 G. W. Bums. M. G. Scroger. G. F. Strouse. M. C. Croarkin. and

[3] BIPM, “The supplementary information of the ITS-90,” BIPM, S h e s ,

temperatures above 235 K, as the reproducibility of tungsten- [4] -, “Techniques for approximating the intemational temperature

[SI B. Fellmuth, w. Buck, and G. Schuster, “Remarks on a future ’He with thermocouple control and others is poorer

Y - _ ., B~~~~~~ 6 6 0 0 ~ and 9620c, and also up to 10840c, trans-

fer standard thermocouples have been developed and used. W. F. Guthrie, “Temperature-electromotive force reference functions and tables for the letter-designated thermocouples types based on the ITS-90,’’ NIST Monograph 175, US Dept. of Commerce, 1993, pp,

Though not necessary in principle, the national laboratories 9-10.