Final report of APMP.T-S5. APMP Regional Comparison of Au/Pt Thermocouples from 0 C to 960 C

Transcription

1 Final report of APMP.T-S5 APMP Regional Comparison of Au/Pt Thermocouples from 0 C to 960 C Dr. F Jahan 1, Dr M J Ballico 1, Yong-Gyoo Kim 2, Hans Liedberg 3, Wang Li 4, Hideki Ogura 5 and C. M. Tsui 6 1 National Measurement Institute of Australia (NMIA), Australia 2 Korea Research Institute of Standards and Science (KRISS), Korea 3 National Measurement Institute of South Africa (NMISA), South Africa 4 National Measurement Centre, (NMC), Singapore 5 National Metrology Institute of Japan (NMIJ), AIST, Japan 6 The Standards and Calibration Laboratory (SCL), Hong Kong (corresponding author): Ferdouse.Jahan@nmi.gov.au Final report_ APMP.T-S5, Page 1 of 56

2 TABLE OF CONTENTS 1. INTRODUCTION SUMMARY OF COMPARISON MEASUREMENT SCHEDULE DESCRIPTION OF THE ARTEFACTS SUMMARY OF PARTICIPANTS MEASUREMENT RESULTS FURNACE TEMPERATURE PROFILES OF EACH PARTICIPANTS INHOMOGENEITY AND ITS CONTRIBUTION TO THE UNCERTAINITIES COMPARISON DATA ANALYSIS CONCLUSIONS REFERENCES APPENDIX A: Protocol for the APMP Regional Comparison of Au/Pt Thermocouple APPENDIX B: Calibration procedures supplied by participants APPENDIX C: Raw Calibration data from the participating Laboratories APPENDIX D: Instrument used by the participants APPENDIX E: Uncertainty Analysis given by the participants Final report_ APMP.T-S5, Page 2 of 56

3 1. INTRODUCTION The Au/Pt thermocouple is the most accurate and stable thermocouple, which can be calibrated from 0 C to 960 C with an uncertainty of 10 mk to 30 mk [1, 2]. A comparison of Au/Pt thermocouples APMP-T-S4 was piloted by KRISS, Korea, designed by the pilot lab, over the period , achieved results allowing the comparison of temperature scales at the 20 mk to 60 mk level over the temperature range of 0 C to 961 C [3]. Following this comparison, it was proposed to extend this comparison to a larger number of APMP laboratories, and try to achieve still better comparison uncertainties. In late 2007, laboratories from the APMP were invited to participate in a comparison of calibrations of Au/Pt thermocouples, using ITS-90 fixed points method. Seven NMIs of the APMP expressed interest. However, to ensure that the uncertainty of the reference value was as low as possible, participation was restricted to NMIs with a demonstrated capability in the area, and with uncertainties below 50 mk. Six NMIs met this criterion and were included in the comparison. The first protocol for this intercomparison was circulated in February 2008 by the pilot laboratory, NMIA, Australia and a final protocol (Appendix A) incorporating the changes suggested by the participants was circulated in May After construction of the artefacts and calibration by the pilot lab, the circulation among the participating labs started in September The measurement was completed in November The draft A report was circulated in November The Calibration procedure used by the individual laboratory is given in Appendix B. The raw calibration data from the participating laboratories are given in Appendix C. The instruments used by each participant are given in Appendix D and the uncertainty calculations as supplied by each individual laboratory, is given in Appendix E 2. SUMMARY OF COMPARISON As Au/Pt thermocouples are expensive to manufacture and the participation was limited to relatively few laboratories, a round-robin topology was selected for this comparison. Two Au/Pt thermocouples, designed [2, 4] and constructed by NMIA were used as the artefacts. Both thermocouples were calibrated by the pilot laboratory (NMIA) using the ice-point and the ITS-90 fixed points of Ga, Sn, Zn, Al and Ag. They were then circulated to the other 5 participating laboratories for calibration and returned to the pilot laboratory for a final calibration. Prior to calibration, each laboratory annealed the thermocouples to the 450 o C state, according to the procedure in the protocol, and performed the calibration from lower to higher temperatures to minimise any effects of reversible hysteresis. The calibration procedure used by the individual laboratory is given in Appendix B. The pilot laboratory measured the inhomogeneity of the thermocouples at the start and completion of the comparison using an oil-bath scanning system [6]. Each laboratory provided measurement data on the temperature profile of each of their fixed point furnaces to allow the pilot laboratory to determine the uncertainty contribution arising from the measured thermocouple inhomogeneity. Final report_ APMP.T-S5, Page 3 of 56

4 3. MEASUREMENT SCHEDULE Several unavoidable delays led to some changes to the circulation schedule initially proposed. The actual schedule achieved was: June August 2008 Construction and calibration of the artefacts by the pilot lab. September October 2008 Calibration by NMISA South Africa. November December 2008 Calibration by SCL Hongkong. March April 2009 Calibration by NMC Singapore May June 2009 Calibration by KRISS Korea July August 2009 calibration by NMIJ Japan October November 2009 Final calibration by NMIA November 2010 Draft-A report circulated to the participating labs. 4. DESCRIPTION OF THE ARTEFACTS Two Au/Pt thermocouples (serial numbers: PtAu-0801 and PtAu-0802) were constructed, based on the NMIA design [2]. The thermocouple wires, Au (99.999% purity) and Pt wires (>99.995% purity) of 0.5 mm diameter, were purchased from Sigmund Cohn Corp. (USA) in The high purity (99.8%) twin bore alumina insulators (700 mm long and 4.75mm OD) were purchased from McDanel Advanced Ceramic Technologies (USA). Each bore in the insulator was 1.57mm in diameter to facilitate movement of the thermoelements. The insulators were prebaked at 1100 C for 6 hours prior to assembly. The Pt wires were electrically bare wire annealed at 1400 C for 3 hours and 1300 C for 7 hours. The Au wires were only furnace annealed at 1000 C for 7 hours and then at 450 C for more than 16 hours. The construction details are given in [2, 4]. After assembly into the insulator, they were given another 1 hour annealing at 1000 C and quenched and 16 hours anneal at 450 C. The ice point section was enclosed in a stainless steel tube of 220 mm long and of 6 mm in diameter. The total lengths of thermocouple wires from the tip to the bottom of the ice-point section was 1550 mm. Quartz sheaths 700 mm long and with 7 mm ID were provided to protect the thermocouples from contamination. 5. SUMMARY OF PARTICIPANTS MEASUREMENT RESULTS Each participating lab annealed the thermocouples to the 450 o C state using the procedures in the protocol (Appendix A) and measured the EMF in 3 freezing plateaus at each fixed point, using their own laboratory procedure (Appendix B). Table 1 summarises the average EMFs reported by each participant (see Appendix C), and Table 2 summarises the reported k=2 uncertainty (see Appendix E) estimated for each reported EMF value, excluding any contribution for thermocouple inhomogeneity. In order to better visualise these raw measurement results, a reference function (IEC 62460) [1]) was subtracted from each NMI s results, giving the values in table 3. These data are plotted in Figures 1 and 2. The pilot lab calibrated the thermocouples twice, initially and finally after circulation among the participants. The difference (in mk) between the two calibrations is shown in Figure 3. For TC(PtAu-0801) the difference is within 5 mk at all fixed points, whereas for TC(PtAu-0802) for high temperature fixed point the differences between initial and final calibration is within 10 mk to 15 mk, indicating good reproducibility of the thermocouples. Final report_ APMP.T-S5, Page 4 of 56

5 Table 1. Measured thermocouple EMF (V) at each fixed point, supplied by the participants. Table 2. The uncertainty values in V (k=2) provided by the participants (excluding the component for inhomogeneity). Final report_ APMP.T-S5, Page 5 of 56

6 Table 3. Calculated E-E REF data (in V) at each fixed point for each participant. Figure 1. The E-E ref in V of thermocouples PtAu-0801 at the fixed points given by all participants Final report_ APMP.T-S5, Page 6 of 56

7 Figure 2. The E-E ref in V of thermocouples PtAu-0802 at the fixed points given by all participants PtAu-0801 PtAu-0802 Initial - Final / C Temperature / C Figure 3. The difference in initial and final calibration in C of the two thermocouples, done by the pilot lab, NMIA. Final report_ APMP.T-S5, Page 7 of 56

8 6. FURNACE TEMPERATURE PROFILES OF EACH PARTICIPANTS Because the temperature profiles in the furnaces used by the participants are different, each participant is effectively calibrating a slightly different region of the thermocouple wire (as the voltage is generated in the temperature gradient region). Any inhomogeneity in the thermocouple wire will therefore appear as an apparent difference in the EMF measured by each participant. To properly estimate the magnitude of this additional uncertainty source in the measurement of the temperature differences between the laboratories, each NMI was asked to provide measurements of the temperature profile in each of their fixed point furnaces. The temperature profile of each of the fixed point furnaces given by the participating laboratories are replotted in Figure 4A to Figure 4E. The participants were also asked to estimate the effective immersion depth when calibrating a thermocouple in each fixed point which is summarizes in Table 4. This was defined in the protocol as the midpoint of the temperature profile. Additionally, the pilot laboratory has determined the points x 1 and x 2 from the supplied profile data (Figure 4), where the temperature has dropped by 10% and 80% respectively and given in Table 5, except at Ga fixed point, where 5% drops was considered. Table 4. The immersion length (in mm) in the various fixed point as estimated by the participants. Fixed point furnaces NMIA NMISA KRISS SCL NMC see note 1 Ag Al Zn Sn Ga NMIJ Icepoint Note 1 These lengths are the depth of the cell bottom to the furnace top. Table 5. Temperature gradient zone of each fixed point furnace used by the participating labs. These values (in mm) were calculated by the pilot from the data supplied by the participants. Final report_ APMP.T-S5, Page 8 of 56

9 Figure 4 A. Temperature profiles of the Ag fixed point furnaces used by the participants Figure 4 B. Temperature profiles of the Al fixed point furnaces used by the participants Final report_ APMP.T-S5, Page 9 of 56

10 Figure 4 C. Temperature profiles of the Zn fixed point furnaces used by the participants Figure 4 D. Temperature profiles of the Sn fixed point furnaces used by the participants Final report_ APMP.T-S5, Page 10 of 56

11 Figure 4 E Temperature profiles of the Ga fixed point furnaces used by the participants 7. INHOMOGENEITY AND ITS CONTRIBUTION TO THE UNCERTAINITIES The inhomogeneity of the thermocouples was measured by the pilot lab at 200 C [6] both before (Figure 5) and after (Figure 6) circulation to the participants. In the protocol it was optional for the participating labs to measure the inhomogeneity of the thermocouples, however, only two other laboratories, NMISA and NMC attempted to measure the thermocouple homogeneity. The measurements performed at NMISA were unsuccessful, as the signal was very noisy and their result was withdrawn. The inhomogeneity scan measured by the NMC is given in Figure 7. In the inhomogeneity graphs (Figure 5, 6 and 7), the position of the curves in the y-axis is not important. It is only the change in the measured EMF for different positions along a given wire that is important. Any arbitrary offset can be added to the data in these plots without changing the analysis or calculated values of inhomogeneity. Final report_ APMP.T-S5, Page 11 of 56

12 PtAu E/ V PtAu Immersion/mm Figure 5. The initial scan of Au/Pt thermocouples at 200 o C performed by NMIA. E is the change in EMF with immersion, [E(x) =E 200 C (x) constant] Manual scan after circulation and annealing PtAu-0801 E / V Immersion / mm PtAu-0802 Figure 6. The final scans of the thermocouples at 200 o C performed by NMIA after circulation. E is the change in EMF with immersion, [E(x) =E 200 C (x) constant]. Final report_ APMP.T-S5, Page 12 of 56

13 E (uv) PtAu-0801 PtAu Immersion depth (mm) Figure 7. Inhomogeneity scan measured by NMC in an oil bath at 200 o C The thermocouple EMF measured by a participant in their fixed point cell is the mainly generated in the region of the thermocouple in the temperature gradient zone of furnace. As the furnaces and cells used by the various participants are different, each participant is effectively calibrating a slightly different region of the wire (as indicated in Table 5). When we then compare the EMFs measured by two participants for a given ITS-90 fixed point, some of the measured difference between the values will arise from inhomogeneity of the thermocouple. The participants were asked to provide uncertainty estimates excluding this factor, and the pilot was to assess the magnitude of the uncertainty and include it. Several different approaches may be taken to estimate this term: (a) Worst case overestimate: A conservative overestimate may be obtained by assuming that ALL the thermocouple signal is generated at a single point in the wire somewhere in the gradient region of the furnace. Then take the change in the thermocouple EMF in the inhomogeneity scan over the full region covering all the furnaces used by the participants. (b) Exact calculation: As the measured gradients in all the furnaces are available, so one could integrate the measured furnace profile together with the measured thermocouple inhomogeneity to come up with an actual expected EMF error for each furnace and participant. (c) Estimate based on linear approximation to the actual gradient: Another option is to assume that the thermocouple EMF is generated uniformly throughout the region from x1 to x2 and integrate this with the measured thermocouple inhomogeneity. This is approximately the same as assuming a linear furnace temperature gradient between x1 and x2; we will add an additional uncertainty for the difference between the measured temperature profiles and this approximate profile, by shifting the assumed gradient region by ±30 mm in either direction. Final report_ APMP.T-S5, Page 13 of 56

14 First let s consider the method (a): As the various furnace profiles (excl. Ga) cover a region from about 280 mm to 520 mm, the data in Figures 5 and 6 would indicate an inhomogeneity of 0.08 V in ( ) V, or % in the EMF. Assuming the error propagates proportionally to E, this corresponds to 0.64 V at 961 C, which is a significant term compared to the uncertainty of 0.2 V to 0.5 V estimated by some participants. However, it is clear from the furnace temperature profile of the furnaces (Figure 4) that there is very significant overlap of all the furnace profiles, so this is an overestimate. Method (b) is in principle possible, however in practical terms it is rather difficult to implement, as the furnace profile data is in some cases incomplete. Therefore this approach was not used. Method (c) is a practical compromise and is the method considered here. Mathematically, we determine difference in the measured thermocouple voltage E M for the ambient to T o part of the thermocouple, when the different gradients used by the participants are applied to the wire. dt E M S( T, x) dx ( 1 ) dx We need a mode for how the inhomogeneity, which is measured at about 200 o C, propagates to different temperatures. Here we assume is propagates proportionally to the EMF, ie. S(T,x) = S T (1+E 200 (x)/(e 200 -E amb )) Where S T is the Seebeck coefficient at T, and E 200 (x) is the measured thermocouple inhomogeneity curves (plotted as E in figures 5, 6 and 7). Note that a small arbitrary offset in the E200(x) curve is the first order equivalent to a small change in the assumed value of S(T) and will thus appear as a shift in the calculated value of EMF. However, as we will only be concerned with the differences in EMF between the various furnace profiles, this is unimportant. Consequentially, ( 2 ) E M x2 x1 x2 x1 S S T ( x) T ( x) (1 E dt dx dx 200 x2 x1 ( x) /( E S T ( x) E E amb ( x) /( E dt ) dx dx 200 E amb dt ) dx dx Changing the integrals from being over T instead of x, and noting that Equation ( 3) can be written as E M T0 ( E T0 S dt T Tamb E amb x2 x1 S T ( x) ) E 200 E E ( x) /( E T0 amb Tamb 200 E S( T ) E Now putting de S(T) dt, we see that amb 200 dt ) dx dx ( x( T )) dt ( 3 ) E( T ) S( T ) dt, ( 4 ) T 0 Final report_ APMP.T-S5, Page 14 of 56

15 E M T0 1 ET Eamb) E x T d E E ( ( ( ))) 0 ( amb Eamb E ( 5 ) Where we integrate the inhomogeneity or thermocouple error voltage E over the region of the thermocouple spanning T o to T amb, and the integral is in terms of voltage. Intuitively, we are adding up the voltage errors in each of the small differential batteries that make up the thermocouple signal from T o to T amb. Another way to consider this integral is as the average of the inhomogeneity error over the gradient region. If we assumed a temperature profile between x 1 and x 2 so that that E varies linearly from E amb to E To over the gradient region x 1 to x 2, i.e. E = E amb +(x-x 1 )*(E To -E amb )/(x 2 -x 1 ) Then d=( E To -E amb )/(x 2 -x 1 ) dx, and the expression for becomes x2 1 ET E 0 amb EM ( ET E ) E200( x) dx 0 amb E E x x ( E ( E T0 T0 E E amb amb E ) E E ) E T0 200 T E E E E amb amb amb amb amb x x 1 x1 average( E x1 to x2 1 x2 x1 E ( x) dx ( x)) ( 6 ) ( 7 ) The expected voltage error due to inhomogeneity is thus the average of the inhomogeneity over the temperature gradient region (i.e. average(e 200 )), scaled by the ratio of EMFs in the furnace and scan-bath (i.e. (E T -E amb )/(E 200 -E amb ) ). For each Fixed-Point and thermocouple, we calculate the expected EMF error for each participant s furnace gradient, and with the supplied furnace gradients shifted by 30 mm in each direction to allow for measurement errors in the supplied gradients. The inhomogeneity, calculated as the semi-range of these calculated maximum difference in the EMFs (divided by E 200 -E amb ), between a thermocouple placed in these various furnace gradients used by the participant, is summarised in Table 6. Table 6. The calculated effective inhomogeneity (expressed as % of the EMF) obtained from the measured inhomogeneity scan data and the temperature profiles of each fixed point furnaces used by the participants. Fixed Pt-Au-0801 Pt-Au-0802 Point initial final average initial final average Ga % % % % % % Sn % % % % % % Zn % % % % % % Al % % % % % % Ag % % % % % % Final report_ APMP.T-S5, Page 15 of 56

16 Note that the inhomogeneity values are typically only 1/3 of the values (0.004%) that would be expected based on the overestimate method, except for the Gallium point, for which the measured fixed point gradients varied significantly between the participants. This sort of a reduction factor may be understood as arising from the fact that the variation in the furnace profiles between the participants is relatively small for the other fixed points, so the inhomogeneity error will be expected to partially cancel in a comparison measurement. As the average of the initial and final results for each thermocouple is to be reported in the overall uncertainty analysis, we use the average of the inhomogeneities from the initial and final inhomogeneity assessments as the semi-range, a, of a normal distribution ie. u i =a/2, which is added to the uncertainties supplied by each participant. 8. COMPARISON DATA ANALYSIS 8.1. Calculation of a Comparison Reference Value The comparison reference value has been calculated by three different methods using the following equations: i) Simple Mean: X simple = X i / n (8) u 2 simple = u i2 / n 2 (9) ii) Median: Computed using the MEDIAN function on Microsoft EXCEL. The uncertainty was calculated using equation given in reference [7]. X ref = median {x i } (10) 1.9 ( xref ) mediamxref xi n 1 iii) Weighted mean: X weighted = X i u i -2 u i -2 (11) u 2 weighted = 1 u i -2 (12) The calculated reference value by the above three methods and their associated uncertainties are given in Table 7 for two thermocouples of the comparison. All of them have reasonable uncertainties value and are consistent to each other. In the case of the weighted mean, the Birge Ratio [8] was calculated and given in Table 7, which is a measure of how well the estimated measurement uncertainties explain the measured dispersion of the actual data values. The Birge ratio is given by the following equation: Birge Ratio = (X i X weighted ) 2 u i -2 (n-1) (13) Final report_ APMP.T-S5, Page 16 of 56

17 which is approximately 1.3 for 6 laboratories. The table 7 shows that the birge ratio is less than or equal to 1 at all temperature points, except for thermocouple PtAu-0802 at aluminium point ( C). Thus it is justifiable to use the weighted mean as the reference value of the comparison. Figure 8 shows the calculated reference values and the values of E-E ref (in mk) calculated from the participants raw data (table 3) at each of the fixed point. The error bars are the uncertainties (k = 2) including average inhomogeneities from Table 6. Table 7. Calculated reference values and their associated uncertainties (k=2) in V and the Birge ratio. Fixed Points Simple Mean Median Weighted Mean Birge Ratio X mean U X median U X weighted U PtAu-0801 IP Ga Sn Zn Al Ag PtAu-0802 IP Ga Sn Zn Al Ag Final report_ APMP.T-S5, Page 17 of 56

18 Figure 8. The values of (E-E REF ) converted in to m calculated from the participants raw data. (closed circles PtAu-0801, open circles PtAu-0802). The error bars are the k=2 uncertainties supplied by the participants, including contribution for inhomogeneity. Final report_ APMP.T-S5, Page 18 of 56

19 8.2. Calculation of the Deviation from the Reference Value As the data from the two thermocouples are consistent with each other (as seen in Table 1), we consider the average of the two sets of data in the calculation of the degree of equivalence D i of the participating lab with respect to the comparison reference value. As the uncertainties related to the calibration of each of the two thermocouples by a participant are likely to be highly correlated, we take the simple means in our calculation of D i and its associated uncertainty u Di : D i = ½ ( X i, TC1 - X weighted,tc1 + X i, TC2 - X weighted,tc2 ) (14) u 2 Di = ½ ( u lab, TC1 2 + u weighted,tc1 2 + u lab, TC2 2 + u weighted,tc2 2 ) (15) * where the X i value for the pilot lab, NMIA, is taken as the average of the initial and final calibration values. The degree of equivalence, Di, at each of the fixed point is shown in Figure 9, where the average of two thermocouples results was used. It shows that at the Ag fixed point the maximum deviation from the reference value is 15 mk for all participants. At lower fixed points except IP point, all deviation is within 10 mk, with the exception of the NMIJ values. Figure 9. The Degree of Equivalence Di, for each participant is plotted as a function of temperature, where the average of two thermocouples results was used. The calculated values of Degree of Equivalence (DOE) and the expanded (k=2) uncertainty U Di = 2 u Di, for each participants are given in Table 8. As at the Ga point, there is a large variation in the temperature profile of the furnaces used by the participants (figure 4E), the Final report_ APMP.T-S5, Page 19 of 56

20 effective inhomogeneity is much larger than at other fixed point, which increased the uncertainty at the Ga point significantly. The same explanation also applies for IP values. Also as the Seebeck coefficient is much lower at these lower temperature, the converted temperature uncertainty increased. Table 8. The values of Degree of Equivalence (top left of each cell) and its k=2 uncertainty (bottom right of each cell), both expressed in mk. Final report_ APMP.T-S5, Page 20 of 56

21 Figure 10. The Degree of Equivalence D i, (average of two thermocouples results was considered) for each participating lab at each of the fixed points. The error bars represent the average uncertainty (k=2), U Di Final report_ APMP.T-S5, Page 21 of 56

22 9. CONCLUSIONS The comparison has successfully demonstrated the use of Au/Pt thermocouples to compare the calibration capabilities of laboratories at 10 mk to 16 mk level over the temperature range from 0 C to 961 o C. The comparison achieved a significant improvement in the uncertainty of the reference values (typically 10 mk 16 mk) compared to the previous APMP comparison using Au/Pt thermocouples [3], which achieved 10 mk-60 mk over this temperate range and also compared to the previous APMP comparison using type R thermocouples, which achieved 30 mk - 60 mk over 0 C to 1100 C [9]. The two Au/Pt thermocouples circulated as comparison artefacts gave consistent results, and the contribution to the uncertainty resulting from the artefact itself is a negligible component in the overall uncertainty of the degreeof-equivalence. The results from all laboratories were consistent with the reference value within the participants stated uncertainties. 10. REFERENCES 1. International Standard IEC 62460, Edition F. Jahan & M J Ballico, Stability studies of a new design Au/PT thermocouple without a strain relieving coil International Journal of Thermophysics, 2007, number 6, Y G Kim, Final Report on the Fixed point Comparison of Au/Pt Thermocouples, Draft B report APMP-T-S4, Metrologia, 2008, 45, Tech. Suppl., APMP.T-S4 Final Report, F. Jahan & M Ballico, Stability study of a simple design high precision Pt-Au thermocouples, Proc. of 6 th Biennial Conf. of MSA, (2005), D C Ripple and G W Burns, Standard Reference Material 1749: Au/Pt Thermocouple Thermometer, NIST Special Publication , F. Jahan and M. Ballico, A study of the Temperature Dependence of Inhomogeneity in Platinum Based Thermocouples, accepted, 8 th Symposium on Temperature: Its Measurement and Control in Science and Industry, Oct. (2002), Chicago, USA. 7. Andras E.- "EUROMET Project no Final report", National Office of Measures, Budapest, R. Kacker, R. Datla and A. Parr, Combined result associated uncertainty from interlaboratory evaluations based on the ISO Guide Metrologia, 39, , F. Jahan and M Ballico, Final report on the APMP Intercomparison of Type R Thermocouples Calibration from 0 to 1100 C, APMP-T-S1, Metrologia, 2007, vol 44, Tech. Suppl., Final report_ APMP.T-S5, Page 22 of 56

23 APPENDIX A: Protocol for the APMP Regional Comparison of Au/Pt Thermocouples APMP.T-S5 March 2008 F Jahan National Measurement Institute of Australia Final report_ APMP.T-S5, Page 23 of 56

24 1. INTRODUCTION The Au/Pt thermocouple is one of the most stable thermocouples, having a stability comparable to SPRTs (10 mk [1]), but more robust. The calibration procedure is also less time consuming than SPRT. Being made of pure element, the Au/Pt thermocouples are inherently more homogeneous, their thermoelectric stability is not limited by shifts in alloy composition caused by preferential oxidation, as in Type R or S thermocouples. Due to these excellent properties, this thermocouple has been considered as a possible alternative to the high temperature SPRT for dissemination of ITS-90 temperature in the industry. In 2005 a pilot comparison of Au/Pt thermocouples was run amongst three labs (KRISS Korea, NMISA (then CSIR) South Africa and NMIA Australia). The artifacts were supplied by the pilot lab, KRISS. A round robin topology using two Au/Pt thermocouples was used for the comparison. The thermocouples were calibrated by all participants using the ITS-90 fixed point method. Mechanical failure of both artifacts at several times during the comparison required them to be repaired, and delayed the comparison. The differences of the fixed point EMFs value between the three labs ranges from 20 mk to 60 mk [1]. A difference of 60 mk at Ga point, was also unexplained. The present comparison aims to improve on the results obtained in the pilot comparison by learning from the problems and limitations identified during the pilot study. Several studies on the Au/Pt thermocouples [2, 3, 4] have shown that calibration uncertainties and stabilities of less than 30 mk should be achievable. The key aspects of the present comparison are 2 Au/Pt thermocouples as designed in paper [2] by NMIA will be used in a round robin comparison. Participation will be restricted to labs with existing demonstrated experience with these artifacts (to reduce chance of breakage). Uncertainty due to inhomogeneity will be reduced by detailed gradient reporting. 2. DESCRIPTION OF THE THERMOCOUPLES Two Au/Pt thermocouples to be used supplied by pilot lab NMIA. Thermocouple is 160 cm long from tip to the bottom of reference junction. The Ice point section is 22 cm long and 6 mm in diameter. The quartz sheath of the thermocouple is 70 cm long with a diameter of 7.0 mm. 3. FACILITIES The lab should have annealing furnace with uniformity of at least ± 20 C. The length of uniform zone should be larger than the immersion length of the thermocouple in the fixed point enclosure. The lab should have Ga, Sn, Zn, Al and Ag freezing point cells with well diameter of minimum 7 mm. The lab should have a precise digital voltmeter with a resolution of 0.01 V. Final report_ APMP.T-S5, Page 24 of 56

25 4. CALIBRATION PROCEDURE USED IN THE COMPARISON: Thermocouples will be calibrated by using ITS-90 fixed points, Ga, Sn, Zn, Al and Ag only using general procedure as referred to Supplementary information of the ITS-90. Thermocouples will be calibrated by all participants in the 450 C annealed state. If the lab has facility to measure inhomogeneity, then measure the inhomogeneity at a temperature lower than 250 C, so that no effects of heat treatment are introduced due to the measurement of inhomogeneity. Calibration of the thermocouples should be from lower to higher temperatures. The contribution of inhomogeneity to the comparison will depend upon the variation of applied temperature gradients fields amongst the participants. The pilot lab will use the measured inhomogeneity together with the reported gradients and the immersion length (Figure 1, Table 1) to determine the likely maximum error in comparison due to inhomogeneity. 5. PARTICIPANTS Name of Laboratory Contact Person Time Schedule NMIA Australia KRISS Korea NMISA South Africa SCL Hongkong NMC Singapore NMIJ Japan CMS Taiwan Ferdouse Jahan Ferdouse.Jahan@nmi.gov.au Kim Yong-Gyoo dragon@kriss.re.kr Hans Liedberg HLiedber@nmisa.org CM Tsui cmtsui@itc.gov.hk Wang Li wangli@nmc.a-star.edu.sg Hideki Ogura h.ogura@aist.go.jp Bor-Jiunn WenJiunner jiunner@itri.org.tw June-July 08 August-September 08 October-November 08 December January 09 February March 09 April May 09 June-July 09 NMIA Australia Ferdouse Jahan Ferdouse.Jahan@nmi.gov.au Aug Sep 09 * 4 weeks for calibration and 3 weeks for transportation and customs clearances 6. TIME SCHEDULE o Write and approve protocol by February 08 o Construct 2 thermocouples by Jun 08 o Calibration by the pilot lab by July 08 and then circulate o Each participants calibrate within 4/5 weeks of receiving date o Final Calibration by the pilot lab. Final report_ APMP.T-S5, Page 25 of 56

26 7. INSTRUCTIONS TO THE PARTICIPATING LABORATORY: Upon receiving the transfer thermocouples, the laboratory must inspect them for damage. If there is any damage, should inform the pilot lab, who will give instruction how to proceed. If there is no damage, anneal the thermocouples at 1000 C for 1 hour to anneal out any hysteresis or inhomogeneity introduced by the previous calibration (if any). Then anneal at 450 C for more than 16 hours to reduce the number of lattice vacancies quenched into the thermoelements, when the thermocouples are removed from higher temperatures. Optional: If the lab has facility to measure inhomogeneity, then measure the inhomogneity of the thermocouples at a temperature less than 250 C, after annealing and before starting calibration. Calibrate the thermocouple at Ice point (0 C), Ga, Sn, Zn, Al and Ag-point from lower to higher temperatures. During calibration CJ ends of the thermocouple should be immersed in to an ice point at least up to an immersion of 180 mm. A calibrated DVM should be used to measure thermocouple EMF. After completion of the calibration, the participant laboratory should transfer the data to the pilot lab and send the thermocouple to the next participating lab. For each of the thermocouples, 3 different realizations are needed at each fixed point. The participant should record at least 30 min of data for each realization. After completion of measurements, the lab should arrange to transfer the artifacts to the next participating laboratory. A) Sender to pay freight charges B) Receiver to pay customs clearances etc C) NMI will prepare a CARNET for applicable counties. Note: The participating laboratory should not dismantle the thermocouple and should handle the thermocouples with extreme care, as they are fragile. 8. REPORTING DATA TO PILOT: The participating laboratory must send to pilot lab the following information within 6 weeks of receiving the thermocouples: A general outline of the calibration procedure consisting of no more than one page and send this as an electronic file named procedure.doc. Details of instrumentation used in the fixed point calibration as an attached Excel spreadsheet named Instrument.xls as in Table 1, including the immersion length. The values of calibration results and the combined uncertainty as an Excel spreadsheet named Calibrationdata.xls. This should be in the format given in Table 2. If the inhomogeneity of the thermocouples is to be measured by the participant, supply the graphs of inhomogeneity. The measured temperature profiles of the enclosures used in the fixed point measurements should be send. Final report_ APMP.T-S5, Page 26 of 56

27 The uncertainty analysis according to the ISO Guide to the expression of Uncertainty in Measurement in terms of microvolt. Please send an electronic file named Uncertainty.xls. 9. Uncertainty Analysis The participating laboratory should send the calculated uncertainty of measurement to the pilot laboratory, as an Excel spreadsheet named Uncertainty.xls. To calculate the uncertainty of calibration, the participants should follow the guideline set out in the ISO Guide to the Expression of Uncertainty in Measurement. The various uncertainty components are given below as a guide: Uncertainty of the Fixed point temperature: an uncertainty value is assigned to the particular fixed point to cover the purity of the fixed point metal, (which can be estimated by melting range of the fixed point, Melt/freeze agreement, flatness of the freezing or melting curve) and also factors related to the realisation of the fixed point such as the choice of position on the freezing plateau. Measurement scatter at the fixed point: This is a type-a component. The reproducibility of EMF measurements in the fixed point. The standard deviation of 3 freezing point values should be used. Conduction errors: The thermocouple should be pulled out by 1 to 2 cm from full immersion and the change in EMF should be measured. This will reflect the minor variations in the equilibrium temperature of the phase transition due to crystal size and habit effects as well as local variations in impurity distribution. Uncertainty due to Inhomogeneity of thermocouple: Exclude this component The inhomogeneity in Seebeck coefficient along the thermocouple being calibrated would be measured by the pilot lab. The inhomogeneity component will be added by the pilot lab during analysis of the results from the reporting gradient zone/ immersion length of the thermocouples. CJ temperature: This component is estimated from the quality of the icepoint and also depends on the immersion and slight inhomogneity (if any) of this section of the thermocouples. DVM calibration and its use: This is calibration uncertainty of the DVM used and the drift of the DVM during use since calibration. Rounding/ Resolution error: The rounding error or the resolution of the reported EMF at each of the fixed point. Stray EMFs and Noise : Any spurious EMFs caused by AC pickup during measurement at the fixed points. 10. References: 10. Y G Kim, Final Report on the Fixed point Comparison of AU/Pt Thermocouples, Draft B report APMP-T-B F. Jahan and M J Ballico, Stability Studies of a New Design Au/Pt Thermocouple Without a Strain Relieving Coil Tempmeko 2007, Lake Louise, Canada. 12. Y G Kim, K S Gam and K H Kang, Thermoelectric properties of the AU/Pt Thermocouples, Rev. Scientific Instrum. 69, , D C Ripple and G W Burns, Standard Reference Material 1749: Au/Pt Thermocouple Thermometer, NIST Special Publication , 1997 Final report_ APMP.T-S5, Page 27 of 56

28 . Table 1: Description of the Measuring equipments used in the comparison. Laboratory Name: Devices r Manufacture Model Serial number Immersion / mm Remarks DVM used Scanner(if used) Ice-Point used Ag Al Zn Sn Ga Ice point Cell Enclosure Cell Enclosure Cell Enclosure Cell Enclosure Cell Enclosure Cell Enclosure Note: Immersion - should be the length at which the mid-point of the temperature gradient zone of the enclosure located on the thermocouples (see Figure 1). Final report_ APMP.T-S5, Page 28 of 56

29 Table 2: Measurement Data of the Thermocouples Fixed Point Freezing Point EMF/ V Combined Uncertainty u c, K= 2 File Name TC 1 TC 2 TC 1 TC 2 Ice point Ga point Sn point Zn point Al point Ag point Final report_ APMP.T-S5, Page 29 of 56

30 Temperature profile of a furnace Furnace Top Immersion length Figure 1. Schematic diagram showing the temperature gradient and immersion of thermocouple. The immersion length is the length from the tip of the thermocouple to the mid point of B and C. TO THE MIDPOINT, NOT TO THE TOP OF THE FURNACE Final report_ APMP.T-S5, Page 30 of 56

31 APPENDIX B: Calibration procedures supplied by participants. Participating Laboratory - NMISA, South Africa Measurements at NMISA (2-10 October 2008): Thermocouples annealed for one hour at 1000 C, removed to room temperature, then annealed for 16 hours at 450 C. Thermocouples scanned for thermoelectric inhomogeneity in stirred oil bath: temperature instability and electrical noise limited the resolution of the scan to ± V / V (0801) or ± 10-3 V / V (0802) and the usable immersion range to approximately 160 mm to 400 mm. Homogeneity test: 200 C Deviation from ref function ( C) Distance from thermocouple tip (mm) AuPt-0801: insertion Insertion AuPt-0801: withdrawal Withdrawal Homogeneity test: 200 C Deviation from ref functionn ( C) Distance from thermocouple tip (mm) AuPt-0802: insertion AuPt-0802: withdrawal Thermocouples measured for three half-hour periods in an ice point and at Ga, Sn, Zn, Al and Ag points (on three separate freeze or melt plateaux). Temperature profiles in cells measured using another Au/Pt thermocouple. Final report_ APMP.T-S5, Page 31 of 56

32 Participating Laboratory - SCL, Hongkong Calibration Procedure for Au/Pt Thermocouple Laboratory: Standards and Calibration Laboratory 1 For measurements at aluminium and silver points, the fixed point cells were placed inside a sodium heat pipe furnace. At zinc and tin points, the cells were placed inside a 3-zone furnace. The cells were heated overnight to a temperature slightly above the freezing point to melt all the metal. To start the freezing, the temperature of the furnace was lowered. For silver and aluminium cells, a cool quartz rod was inserted into the thermometer well of the cell for 1 to 2 minutes to create an inner freeze. For zinc cell, the SPRT was removed and then re-inserted into the well after 30 seconds to create an inner freeze. For tin cell, the whole cell assembly was temporarily removed from furnace to create an inner freeze. An SPRT was used to monitor the temperature of the cell to ensure the freezing plateau was reached. After half an hour, the SPRT was removed and the Au/Pt thermocouple under test was inserted. The open-circuit emf of the test thermocouple was measured directly by a digital voltmeter without using scanner. 2 For measurements at gallium point, the cell was cooled overnight by placing it inside the control unit with power off. To melt the cell, power was applied to the heater of the control unit. A monitor SPRT was placed inside the thermometer well of the cell. When the melting temperature was reached, 9 ml of water at about 40 C was added to the well to create an inner melt. The SPRT was used to monitor the cell temperature to ensure the melting plateau was reached. After about an hour, the SPRT was removed and the Au/Pt thermocouple under test was inserted. The open-circuit emf of the test thermocouple was measured directly by a digital voltmeter without using scanner. 3. At 0.01 C, the emf of the Au/Pt thermocouples was measured using a triple-point-ofwater cell. The emf at ice point (0 C) was calculated using the following formula emf(0 C) = emf(0.01 C) 0.01C x 6.04 V/C 4. In all the measurements, the cold junctions of the test thermocouple were maintained at 0 C by immersing in an ice cell. 5. The calibration sequence was from lower to higher temperatures. Final report_ APMP.T-S5, Page 32 of 56

33 Participating Laboratory - KRISS, Korea Au/Pt FP calibration procedures by KRISS by Yong-Gyoo Kim, Kee Sool Gam and Young Hee Lee 1. TCs were annealed in horizontal furnaces as indicated by protocol. Please see an attached excel file to look into the temperature gradient. 2. After annealing, they were measured at ice -> Ga -> Sn -> Zn -> Al -> Ag points in sequence 3 times. 3. From Sn to Ag, after confirming supercool and recovery, quartz chilled-rod was inserted for 1 min one time. Monitoring thermometer was a type S thermocouple. 4. At first realization, 0801 was inserted, and then In next, sequence was reversed. At last measurement, 0801 was inserted firstly. At third realization, thermocouples were pulled out by 1 cm and 2 cm. 5. DVM 2182A configurations were: analog/digital filter ON, number of reading 30, REL function ON, Moving average function ON 6. After each measurement, DVM null tests were performed to measure the short-term stability of DVM. 7. Inhomogeneity test was not performed at KRISS in order to prohibit any contamination. 8. Noise tests were performed by turn-off the furnace abruptly during realization. There was not a significant emf change. 9. All processes were done through the protocol. 10. All tcs were handled in as-received form. Quartz tubes were not withdrawn at all. Final report_ APMP.T-S5, Page 33 of 56

34 Participating Laboratory - NMC, Singapore Calibration procedure for APMP Au/Pt thermocouple comparison 1. The two Au/Pt thermocouples (S/N 0801 and 0802) were calibrated by using fixed point method. Ice, Ga, Sn, Zn, Al and Ag fixed points were used for the calibration. The emf of the Au/Pt thermocouples was measured by using a digital nanovoltmeter together with a scanner (see attached file Instrument.doc for details of the equipment). The calibration was done from low to high temperatures. 2. The two thermocouples were annealed at about 1000 C for about 1 hour at a Sodium heatpipe furnace upon arrival. During the annealing, each thermocouple was inserted into a quartz tube to protect the thermocouple from any possible contamination. 3. The two thermocouples were then inserted into a Potassium heatpipe furnace which was kept at about 450 C (again with quartz tubes). The thermocouples were annealed for about 17 hours. 4. The inhomogeneity of the thermocouples was tested at 200C in liquid bath with immersion up to 400mm after the annealing. The measurement result is attached in file Inhomogeneity.xls. 5. After the inhomogeneity test, the thermocouples were calibrated at the various fixed points. For ice point, the measurement was done after annealing, after inhomogeneity measurement and after Ga point for each thermocouple. The average of the three measurements was taken. For Ga, Sn, Zn, Al and Ag points, 3 to 5 plateaus were realized. The average value of plateaus was taken for each fixed point. 6. At each plateau, the TC was inserted into the cell bottom fist. After the measurement data were collected, the thermocouple was lifted up step by step for 10mm each step until 20 to 90 mm from the cell bottom (depending on the plateau durations) to check any immersion error. Same measurement was repeated at each step. The TC was finally replaced at the cell bottom and the measurement was repeated. The two measurement results at the cell bottom were averaged to get the emf value at the cell bottom. 7. The emf measurement data were taken through an automatic data acquisition system. During each measurement, the TC s emf was monitored and the data were taken only when it was stabilized. For each measurement at the cell bottom, data in at least 30 minute period (more than 200 readings) were collected and averaged. For immersion test, data in a period of about 10 minutes were collected due to the limitation of the plateau duration. 8. The measured emf of the Au/Pt thermocouple and estimated measurement uncertainty at each fixed point was given in file Comparisondata.xls. 9. Measurement uncertainties were estimated according to the comparison protocol (see file Uncertainty.xls). Compiled by: Wang Li, National Metrology Centre, A-Star, Singapore Date: 07 July 2009 Final report_ APMP.T-S5, Page 34 of 56

35 Participating Laboratory - NMIJ, Japan General outline of the calibration procedure 1) The thermocouples were annealed at 1000 C for 1 hour then at 450 C for 16 hours. The annealing furnace used in this work has a uniform zone of over 60 cm lengths with uniformity of ± 15 C at 1000 C and ± 5.5 C at 450 C. Each thermocouple was inserted 65 cm into the furnace to ensure that 55cm from the tip of the thermocouple was held within this temperature uniform zone. 2) After annealing, the thermocouples were calibrated at the ice point (0 C), the triple point of Ga, and the freezing points of Sn, Zn, Al and Ag from lower to higher temperatures. The depth of the full immersion position of the Sn, Zn, Al and Ag points was about 55 cm from the top of the thermometer well. Temperature profiles of these furnaces are presented in an Excel sheet named Temperature profile in the Instrument.xls file. 3) Instead of using the quartz sheaths provided, protective quartz sheaths for the thermocouples were newly prepared for exclusive use by NMIJ and used at the Sn, Zn, Al and Ag points. 4) During calibration CJ ends of the thermocouples were immersed 190 mm into a mixture of shaved ice and distilled water in a Dewar flask. 5) A calibrated DVM was used to measure thermocouple EMF. 6) Three different plateaus were realized on different days at each fixed point. During a plateau, which continued over 8 hours, two thermocouples were immersed into the cell in turn. 7) EMF data were recorded for 55 min at the full immersion position and then for 15 min at the 15mm upper position for each thermocouple during each plateau. The EMF value at each fixed point was the average of data for the last 30 min at the full immersion position of 3 plateaus. The values reported in Table 2 are after correction of the DVM calibration. Final report_ APMP.T-S5, Page 35 of 56

36 Participating Laboratory - NMIA, Australia General outline of the calibration procedure: The thermocouples were annealed in horizontal annealing furnace of 1000 mm long at 1000 C for 1 hour and quenched. The temperature uniformity of the furnace was ±10 C for the central 700 mm length. Thermocouples were then annealed at 450 C for more than 16 hours. After annealing the thermocouples were scanned in an oil bath to measure the inhomogeneity at a temperature of 200 C. Then they were calibrated at ITS-90 fixed points of ice point, Ga, Sn, Zn, Al and Ag points. During calibration the cold junction ends of the thermocouples were immersed up to 180 mm to a crushed icepoint in a 30 cm long dewar. The EMF was measured by a calibrated Agilent Nanovoltmeter, model 34420A and the data was logged by using automatic data acquisition program. For higher temperature fixed point Al and Ag, the temperature of the fixed point was monitored by another Au/Pt thermocouple or a Type R thermocouple. After nucleation, when freezing plateau was stabilised, then the test thermocouple was inserted to the full immersion after pre annealing for a 10 minutes at the required temperature. As the freezing plateau was lasted for more than 3 hours, two thermocouples were measured for 40 to 50 minutes. Three different freezing plateaus were realised for each of the fixed point. For lower temperature fixed point (Ga, Sn and Zn) monitor thermocouple was not used. The EMF value given was the average of 25 to 30 minutes of measured EMF. Conduction error was tested in each fixed point by pulling out thermocouples from 1 to 4 cm from full immersion. Final report_ APMP.T-S5, Page 36 of 56