International Comparison Interlaboratory comparison of

Document Sample
International Comparison Interlaboratory comparison of Powered By Docstoc
					metrologia

International Comparison Interlaboratory comparison of realizations of the triple point of water
E. Renaot, M. Elgourdou and G. Bonnier

Abstract. An interlaboratory comparison of the different realizations of the triple point of water (TPW) in Europe has been organized by EUROMET (project No. 278). This project was based on the circulation of one TPW cell and an isothermal enclosure. The national metrology institutes of thirteen countries and the Bureau International des Poids et Mesures were involved in the work. Twenty-seven TPW cells were compared. Half of the results lie between –0.05 mK and +0.05 mK of the mean value, and all but two of the results are within 0.10 mK of the mean.

1. Introduction An interlaboratory comparison of the different realizations of the triple point of water in Europe has been organized under the auspices of EUROMET (project No. 278). The aim of this project, based on the circulation of one cell and an adapted isothermal enclosure, was to compare the customary realizations of the triple point of water in European national laboratories rather than to carry out further research on TPW behaviour, which has been well documented by several laboratories [1-5]. The Bureau National de M´ trologie-Institut Nae tional de M´ trologie (BNM-INM, France), as the pilot e
E. Renaot, M. Elgourdou and G. Bonnier: Bureau National de M´ trologie-Institut National de M´ trologie (BNM-INM), e e 292 rue Saint-Martin, F-75141 Paris Cedex 03, France.

laboratory, supplied the TPW cell and the isothermal enclosure. It established the schedule and followed the progress of the comparison, which was organized in four stages. Table 1 lists the participating laboratories and Table 2 the order in which the measurements were conducted. After measurements by a group of participants, the cell and the isothermal enclosure were returned to the BNM-INM for a stability test before shipping to the next group. The results of the stability tests are given in Section 5. The national metrology institutes of thirteen countries: the BNM-INM, NPL, NMi, PTB, DTI, IMGC, SP, CEM, IPQ, CMA, JV, OFMET and the UME (see Table 1 for acronyms), were involved in this work which lasted from January 1994 to June 1997. The BIPM took part in May 1997.

Table 1. Participating laboratories.
Laboratory Bureau National de M´ trologie-Institut National de M´ trologie e e (BNM-INM/CNAM), pilot laboratory National Physical Laboratory (NPL) Nederlands Meetinstituut (NMi) Physikalisch-Technische Bundesanstalt (PTB) Danish Technological Institute (DTI) Istituto di Metrologia “G. Colonnetti” (IMGC) Sveriges Provnings- och Forskningsinstitut (SP) Centro Espa˜ ol de Metrolog´a (CEM) n ı Instituto Portuguˆ s da Qualidade (IPQ) e Centre for Metrology and Accreditation (CMA) Justervesenet (JV) Office F´ d´ ral de M´ trologie (OFMET) e e e National Metrology Institute (UME) Bureau International des Poids et Mesures (BIPM) Metrologia, 2000, 37, 693-699 Country France United Kingdom Netherlands Germany Denmark Italy Sweden Spain Portugal Finland Norway Switzerland Turkey Participants E. Renaot (coordinator), M. Elgourdou, G. Bonnier M. V. Chattle, J. Gray M. de Groot F. Edler, U. Noatsch I. Wessel P. Marcarino, R. Dematteis J. Ivarsson V. Chimenti, F. Perezagua, D. del Campo E. Filipe, I. Lobo T. Weckstr¨ m o C. Rauta A. Steiner A. T. Ince, A. Kartal R. Pello, R. K¨ hler o

693

E. Renaot et al.

Table 2. Interlaboratory comparison programme.
First stage BNM-INM NPL NMi PTB DTI BNM-INM IMGC CEM IPQ SP CMA BNM-INM JV OFMET BNM-INM UME BIPM BNM-INM

Second stage

away. This water is then pumped out from the bottom of the enclosure. The BNM-INM studied the immersion-temperature effects of cell No. 679 in the enclosure using a Leeds and Northrup thermometer (No. 1807664). The experimental measurements agree with the theoretical values calculated using d /d given by the International Temperature Scale of 1990, see Figure 2. All the thermometers used in this comparison were glasssheathed SPRTs. The behaviour relative to the depth of immersion was expected to be the same for all SPRTs.

Third stage

Fourth stage

The number of laboratories participating in each stage was determined by their availability to carry out the tests and the constraints of the ATA carnet. 2. TPW cell and isothermal enclosure The circulating TPW cell, No. 679, was made by the National Physical Laboratory (NPL, UK). Prior to the EUROMET comparison this cell was compared with one cell (No. 673) belonging to the batch of TPW cells which constitutes the BNM-INM reference for this fixed point (Table 3). The mantle was realized three times on the dates indicated in the table and the measurements were performed five times for each mantle.
Table 3. Comparison of cell Nos. 679 and No. 673 prior to the EUROMET project.
Date 1993-08-12 1993-08-26 1993-09-28 Mean ( 673 +0.03 –0.02 +0.03 0.014
679 )/mK

The isothermal enclosure was designed and constructed at the BNM-INM (Figure 1). During the measurement with a standard platinum resistance thermometer (SPRT), length of glass sheath 48 cm, part of the head is inside the cover of the enclosure. This arrangement prevents visible and infrared radiation from penetrating the ice and reaching the thermometer sensor. The bottom of the SPRT is not in direct contact with the bottom of the thermometer well, but at a distance of approximately 5 mm. For other thermometers, this point was not explicitly described in the protocol but the participating laboratories were expected to adjust the position of the thermometer so that its base was approximately 5 mm above the bottom of the well. Several holes in the cell cover and the lower part of the support allow the water from melted ice to drain

Figure 1. Isothermal enclosure. 1: cover of the enclosure; 2: plastic cover of the cell full of crushed ice; 3: water extraction; 4: isothermal container; 5: cell holder; 6: crushed ice; 7: plastic container with foam rubber; 8: base containing drain holes; 9: water container.

3. Procedure The aim of this project was to allow each participating laboratory to compare the temperature of the triple point of water realized by their local facilities (cell, enclosure and procedure) with the temperature of the circulating TPW cell. The realizations using local cells and enclosures were performed according to local
Metrologia, 2000, 37, 693-699

694

Interlaboratory comparison of realizations of the triple point of water

Figure 2. Effect of immersion temperature. TPW : resistance of the TPW cell; : measurement 0 mA; ▫: calculation.

level. The uncertainty arising from this potential drift is included in component A2 (see Section 7). The number of independent measurements using different freezes was not prescribed by the comparison procedure. Each laboratory was allowed to fix the number of measurements according to its customary procedures. The uncertainty in the mean was expected to take into account the number of measurements made. The comparison was performed by measuring the difference in temperature between the circulating and local TPW cells. The difference in the observed resistances (corrected for the hydrostatic head effect and self-heating, and possible calibration of the measurement instrument) for the two cells was converted to a temperature difference using the d /d for the SPRT:

procedures, whereas the realization with the circulating cell was strictly defined by a precise procedure common to all the participating laboratories. The stability of the at circulating cell was tested by measuring the BNM-INM prior to and during the programme. All the participants were required to use the same procedure for preparing the circulating cell. First, the isothermal enclosure is filled with crushed ice. When the isothermal enclosure is cooled, the circulating cell is introduced into the plastic container (7 in Figure 1). After at least 3 h, alcohol is introduced into the thermometer well. In order to prepare the ice mantle, a metal rod precooled in liquid nitrogen is inserted into the thermometer well. This operation is repeated several times to obtain an adequate mantle thickness (4 mm to 8 mm) which is uniform over its whole length. The alcohol is then removed and the well is washed out with pure water or alcohol precooled to a temperature of 0 C. Finally, precooled pure water or alcohol is again poured into the well. The level of this liquid is adjusted to that of the free surface water in the cell when the thermometer is present. To avoid exothermic heat during measurements, special care must be taken to prevent mixing water and alcohol in the thermometer well. A second ice/water interface, immediately adjacent to the well surface, is formed by producing a layer of water (melted ice) by inserting a rod at room temperature into the well for about 30 s. The effectiveness of this layer should be verified prior to any measurement by checking the free rotation of the mantle. Each laboratory was allowed to fix the age of the ice mantle following its customary procedures, although it was specified that the cell must be left at rest for at least 20 h in order to release any mechanical stress in the ice mantle. A study by Ancsin and Murdock has shown this to be sufficient time to achieve temperature equilibrium [6]. The BNM-INM experience confirms this observation. Nevertheless, it has been observed by skilled laboratories that in some experiments it took more than 20 h to reach equilibrium at the 0.01 mK
Metrologia, 2000, 37, 693-699

where the subscripts l and c indicate the local realization and the measurement of the circulating cell, respectively. 4. Description of local facilities Table 4 presents the features of the cells and the different local procedures used for this comparison. The electrical measurements were performed using a Guildline 9975 (four laboratories) or an ASL F18 (ten laboratories) bridge. The platinum resistance thermometers came from various manufacturers: Leeds and Northrup, Chino, Tinsley, Yellow Spring, Isotech, Hart Scientific. 5. Control of transfer cell during comparison In February 1993, prior to delivery to the BNM-INM, cell No. 679 was compared at the NPL with NPL cells 555 and 611. The results indicated that the mean of the temperatures realized by NPL cells 555 and 611 was 0.06 mK below that realized by cell No. 679, all three cells being stored and measured in the same NPL isothermal container. In August 1994 the NPL compared the same cells as part of this EUROMET project; in this case cell No. 679 was stored in the circulating isothermal enclosure. The results indicated that the mean of the temperatures realized by NPL cells 555 and 611 was 0.054 mK below the temperature realized in cell No. 679. The BNM-INM checked the stability of cell No. 679 during the comparison. It was compared with cell No. 673 at the beginning and end of the comparison and also between stages (Table 5). During this period cell No. 673 was compared with IMGC, KRISS and VNIIM cells as part of the BIPM interlaboratory comparison of the triple point of water [7]. were analysed assuming Variations in that the cells do not change in the same way. Taking into account the combined standard uncertainty in

695

E. Renaot et al.

Table 4. Features of cells, local procedures, results and associated uncertainties.
Laboratory Type BNM-INM NPL Cell No. 37 673 821 611 555 Delivery date 1977 1993 1995 1992 1991 Metal rod precooled in liquid nitrogen Carbon dioxide 0 0.014 –0.033 Isothermal enclosure full –0.050 of ice, cell in direct contact –0.057 with ice NMi bath –0.148 –0.111 Isothermal enclosure full –0.008 of ice 0.014 –0.020 0.020 Isothermal enclosure full of –0.021 ice + Plexiglas jacket and layer of air Isothermal enclosure full of 0.014 ice + Plexiglas jacket and layer of water Isothermal enclosure full of –0.037 ice + perforated plastic tube and layer of water Yellow Springs bath –0.05 Isothermal enclosure full –0.08 of ice Isothermal enclosure full of –0.003e ice and ISOTECH bath –0.264f ISOTECH bath –0.035 –0.33 –0.087 ISOTECH bath 0.035 Preparation Isothermal enclosure and bath mK
l c

/

mK

c

l

c

/

mK (coverage factor 2) 0.070

l

c

/

NPL

NPL

0.032

0.064

NMi PTB

VSL NPL PTB FGWa IMGC

IMGC

337 1993 338 1993 683 1993 692 1993 162/81 1981 308 ≤1989 1991 31b 1179c 1985

Circulation of liquid and gaseous nitrogen Metal rod precooled at –18  C

0.040 0.040

0.080 0.080

Metal rod precooled in liquid nitrogen Carbon dioxide + alcohol Carbon dioxide + alcohol Carbon dioxide Carbon dioxide Circulation of alcohol at –47  C Metal rod precooled in liquid nitrogen Carbon dioxide Metal rod precooled in liquid nitrogen

0.036

0.072

CEM

Jarret

0.035

0.070

DTI

Jarret

A-11 1987

0.041

0.082

IPQ CMA SP JV

NPL NPL

299 361

1986 1987

0.100 0.027 0.077 0.090

0.200 0.054 0.154 0.180

Spembly 100C 1974 NPL 815 823 753 1995 1995 1994

OFMET

NPL

Isothermal container full of ice + acrylic tube and layer of air

0.100

0.200

UME

BIPM

VSL IMGC UME UME UME ASMWd KRISS

293 24 4 6 13 131 1

1994 1980 1995 1995 1995 1980 1994

Cooling device with nitrogen gas Carbon dioxide + alcohol Hart Scientific bath

Carbon dioxide

Isothermal container full of ice + acrylic tube and layer of air

0.015 –0.114 0.097 0.080 0.081 0.026 0.072

0.047

0.094

0.040

0.080

a. Forschungsgemeinschaft f¨ r technisches Glas-Wertheim. u b. One of the batch of TPW cells that constitutes the IMGC reference (seven cells). c. One of the batch of TPW cells that constitutes the CEM reference (four cells). d. Now the PTB. e. Local cell 100C, age of mantle 30 h/circulating cell, age of mantle 53 h. f. Local cell 100C, age of mantle 170 h/circulating cell, age of mantle 29 h.

(0.035 mK), the temperature realized in cell No. 679 may be considered to have been stable during the entire comparison. 6. Results The aim of the comparison was to compare each local TPW realization with that realized in the circulating cell. Figure 3 shows the temperature differences derived from the resistance measurements given in Table 4. The cell differences are averages as more than one measurement was made in each laboratory. Depending on the laboratories, one, two or three measurement runs were carried out. A new mantle was

Table 5. Stability test on cell No. 679.
Date August 1993 First stage November 1994 Second stage September 1995 Third stage May 1996 Fourth stage May 1997 (
673 679

)/mK

0.014 –0.005 0.027 0.027 –0.030

prepared for each run, which lasted from a few days to two weeks.
Metrologia, 2000, 37, 693-699

696

Interlaboratory comparison of realizations of the triple point of water

Figure 3. Temperature differences

l

c

.

The following sub-sections give specific remarks by two of the participating laboratories, whose cells were excluded from the final analysis of the results. 6.1 SP For measurements in the first run the mantle of the local cell (No. 100C) was only 1 day old and later measurements have shown that this cell exhibits a maximal temperature when the mantle is about 2 days old. The temperature then decreases by 0.25 mK and levels off after about 5 days. The second run was taken after one week. The age of the mantle of EUROMET cell 679 was about 2 days and 1 day, respectively, for the different runs. 6.2 JV Two remarks concern the results obtained with cell No. 823: is significant. (a) The dispersion of With two thermometers the minimum and the maximum differences are respectively –0.30 mK and –0.50 mK. (b) During this comparison the ice mantle was formed by two different methods for cell Nos. 823 and 815: (i) for No. 823 the ice mantle was prepared with crushed solid carbon dioxide; (ii) for No. 815 the ice mantle was prepared with a metal rod precooled and in liquid nitrogen. The values of may suggest that the temperatures realized by cell Nos. 823 and 815 were not the same. On another occasion, however, cell Nos. 823 and 815 were compared directly. The ice mantles were formed by the same method (metal rod precooled in liquid nitrogen). In this case the difference observed 0.067 mK with a combined was standard uncertainty of 0.09 mK.
Metrologia, 2000, 37, 693-699

7. Uncertainties In this interlaboratory comparison we applied the general rules for expressing uncertainty in physical measurement as established in the ISO Guide [8]. The comparison consisted of measuring the difference in temperature between the circulating cell and a local one, therefore the laboratories had to evaluate the uncertainty components related exclusively to this difference. They had to take into account expected correlations: for example the electrical measurements on the circulating and local cells are made with the same bridge and at practically the same ratio. The standard uncertainty in the difference is smaller than the standard uncertainty in the value of the SPRT resistance. Some comments on the uncertainty components are given below. 7.1 Type A evaluation Component A1 : This corresponds to the repeatability of measurement results, under the following conditions: • • • • • • same measurement procedure; same observer; same thermometer left in place; same bridge; same isothermal enclosure; repetition over a short period (i.e. without changing the ice mantle).

Component A2 : Through different measurements performed in a given laboratory on TPW cells, the standard deviation of the measurements can be obtained. This standard deviation can be considered to be an estimate of the reproducibility of the measurements due to changes in the influencing quantities.

697

E. Renaot et al.

The observed temperature differences between circulating and local cells depend on various factors including the following: • • • • crystal size [1, 9]; age of mantles; differences between mantles; method of handling the cells before preparation of the mantle (if a cell is stored for a considerable time in the same position before preparation of the mantle, the impurities may not be uniformly distributed in the water and handling can then cause a change in the local impurity concentration); undesirable thermal reaction in thermal contact fluids (mixing alcohol and water); stability of SPRT (repeated measurements with reinsertion of SPRT); different types of SPRT (geometric characteristics of the thermometer can influence the observed difference in temperature between circulating and local TPW cells); influence of spurious heat flux linked with the temperature of the surroundings.

measurements are corrected for the self-heating effect. As the thermal resistances have approximately the same magnitude in circulating and local cells, the difference between the self-heating corrections is very small. In addition, the uncertainties in self-heating corrections in circulating and local cells are strongly correlated. For each laboratory the combined standard uncertainty is calculated from the different components presented in Table 6. 8. Analysis of results Figure 4 presents the results as a histogram. This simplified presentation shows clearly that only two of the results differ significantly from the general behaviour and that approximately half of the results range between –0.05 mK and +0.05 mK. Table 7 lists data. some statistical properties of the

• • •

•

7.2 Type B evaluation Component B1 : The source of this standard uncertainty is electrical measurement (as indicated above). Component B2 : This is the uncertainty of the correction related to hydrostatic pressure. In this comparison we considered the contribution of some uncertainties to be negligible, either because their values are very small, or because these uncertainties are strongly positively correlated. This applies in particular to the uncertainty in self-heating corrections. All the
Table 6. Uncertainty budgets.
Component
A1 A2 B1 B2

Figure 4. Histogram of results for

l

c

.

BNMINM 0.015 0.025 0.020 0.005 0.035
c

NPL 0.020 0.020 0.015 0.001 0.032

NMi 0.030 0.025 0.010 0.040

PTB 0.010 0.030 0.025 0.006 0.04

IMGC 0.015 0.030 0.012 0.006 0.036

CEM 0.031 0.011 0.012 0.004 0.035

DTI 0.041 0.0002 0.005 0.041

IPQ 0.03 0.064 0.066 0.001 0.10

CMA 0.009 0.020 0.015 0.004 0.027

SP 0.008 0.075 0.014 0.001 0.077

JV 0.020 0.085 0.015 0.006 0.09

OFMET 0.05 0.08 0.03 0.01 0.10

UME 0.015 0.040 0.013 0.004 0.047

BIPM 0.002 0.017 0.019 0.030 0.040

Combined standard uncertainty

Table 7. Statistical properties of the

l

c

data.
Standard deviation/mK 0.097 0.064 Standard deviation of mean/mK 0.018 0.013 Mean of combined standard uncertainties/mK 0.053 0.051

Arithmetic mean/mK All results Without cells 100C and 823 –0.037 –0.017

698

Metrologia, 2000, 37, 693-699

Interlaboratory comparison of realizations of the triple point of water

8.1 Investigation of possible bias introduced by method of realizing ice mantle For this comparison, the methods used for the preparation of the ice mantle can be classified into two main groups: (a) methods using liquid nitrogen (fast freezing of the water); (b) other methods. Cell Nos. 100C and 823 showed unstable behaviour (see Section 6); consequently their results have been excluded from the analysis. For each group the arithmetic mean of and the standard deviation of the results was calculated (Table 8).
Table 8. Influence of preparation of ice mantle.
Method Liquid nitrogen Other Arithmetic mean of l c /mK –0.052 0.007 Standard deviation of l c /mK 0.059 0.057

Figure 5.

l

c

according to delivery date of cell.

enclosure. Twelve European laboratories, the Turkish national laboratory and the BIPM were involved in the comparison of twenty-seven TPW cells. During the comparison, the stability of the circulating TPW cell was periodically checked by comparisons with another BNM-INM cell. Half of the results lie between –0.05 mK and +0.05 mK of the mean value, and all but two of the results are within 0.10 mK of the mean. References
1. McAllan J. V., In Temperature: Its Measurement and Control in Science and Industry, Vol. 5 (Edited by J. F. Schooley), New York, American Institute of Physics, 1982, 285-290. 2. Berry R. J., Can. J. Phys., 1959, 37, 1230. 3. Furukawa G. T., Mangum B. W., Strouse G. F., Metrologia, 1997, 34, 215-233. 4. Furukawa G. T., Bigge W. R., In Temperature: Its Measurement and Control in Science and Industry, Vol. 5 (Edited by J. F. Schooley), New York, American Institute of Physics, 1982, 291-297. 5. Bonhoure J., Pello R., In Temperature: Its Measurement and Control in Science and Industry, Vol. 6 (Edited by J. F. Schooley), New York, American Institute of Physics, 1992, 299-303. 6. Supplementary Information for the International Temperature Scale of 1990, S` vres, Bureau International des Poids e et Mesures, 1990, 107. 7. Pello R., Goebel R., K¨ hler R., Report on the international o comparison of water triple-point cells, BIPM Rapport BIPM/96-8, 1996. 8. Guide to the Expression of Uncertainty in Measurement, Geneva, International Organization for Standardization, 1993. 9. Mendez-Lango E., Proc. TEMPMEKO’96, Turin, Levrotto & Bella, 1997, 57-62.

Because the manufacturers of the cells are not the same in the two groups, it is very difficult to state objectively that the difference between the two arithmetic means depends only on the methods used for the preparation of the ice mantle. This difference is also well within the combined uncertainty of the two values. Studies from several laboratories show that the method of preparation of the mantle does not affect the equilibrium value of the cell although it might do so during the first few days after preparation [2, 3]. 8.2 Investigation of possible bias introduced by ageing of cell Figure 5 presents the different results versus the age of the cell (date of delivery to the user), cell Nos. 100C and 823 being excluded. Statistical analysis indicates that the correlation coefficient between the different results versus time is very close to zero (0.005), therefore no relation can be found between the results and the age of the cell. It must nevertheless be pointed out that it is common practice in any laboratory to discard a cell presenting unusual behaviour. The cells that become difficult to use are therefore withdrawn from the reference batch. 9. Conclusions EUROMET project No. 278 was based on the circulation of one TPW cell and an isothermal

Received on 13 October 1998 and in revised form on 29 June 2000.

Metrologia, 2000, 37, 693-699

699