high-temperature - IEEE Xplore

4 downloads 0 Views 942KB Size Report
SICE Annual Conference in Sapporo, August 4-6,2004. Hokkaido Institute of Tecnology, Japan. HIGH-TEMPERATURE. FIXED-POINTS. AT THE NATIONAL.
SICE Annual Conference in Sapporo, August 4-6,2004 Hokkaido Institute of Tecnology, Japan

HIGH-TEMPERATURE FIXED-POINTS AT THE NATIONAL PHYSICAL LABORATORY D. Lowe, G.Machin National Physical Laboratory, Queens Road, Teddington, Middlesex TW11 OLW,UK

Abstract: High temperature fixed-points based on eutectic alloys of metals and carbon have been proposed as secondary reference points for the International Temperature Scale o f 1990. These allow checks of thermometry with a precision rivalling the uncertainties of many National Measurement Institutes. Aspects of the construction of such fixedpoints are considered with guidelines for making fixed-points that have best performance. Keywords: eutectic, fixed-points, high-temperature.

1. Introduction Thermodynamic temperature determination is difficult to carry out and has large uncertainties. For example, the uncertainty of the thermodynamic melting temperature of pure silver is 80 mK (k2). This compares to readily achievable uncertainties of 10 mK (k2) at the silver point for a good PRT. This illustrates the benefits of using a temperature scale based on defining fixed-points rather than the "true" temperature. Even if the fixed-point thermodynamic values are not well known a scale based upon them can give better repeatability and accessibility. In general reproducibility of a particular process is often far more important than the actual temperature. With the Intemational Temperature Scale (currently ITS90) this reproducibility is readily available to users.

ITS-90 allows best efficiency and consistency, and benefits free trade. The lowest uncertainties are found when the scale is based upon a scheme of fixed-points and interpolation functions. This is how JTS-90 is defined over most of the range from cryogenic temperatures to the freezing point of silver (962 "C). Above the freezing point of silver ITS-90 is based on a single fixed-point at the IOW end of the scale and relies on extrapolation using a wellcharacterised radiation thermometer. This has two important consequences: Firstly a lot o f time and effort goes into maintaining a scale above -1 100 "C as it relies on more of less continuous assessment of the pyrometer together with frequent recourse to standard reference sources. For this reason it is usual for National Measurement Lnstitutes (NMIs) to maintain the scale and disseminate it to industry through subsidiary calibration laboratories. Inevitably this intermediate step increases the calibration uncertainties. Secondly the uncertainties become large at high temperatures. This is because the available ITS-90

fixed-points are all below 1100 OC, at the lower end of a scale that extends from 962 "C to, typically, 3000 "C or more. As many of the uncertainties involved in scale realisation scale as T2, or as the temperature difference from the reference, small uncertainties at the reference temperature can quickly become large. It is well known that fixed-points at higher temperatures would be desirable 'I. The reason for not going above copper at 1084 "C is one of practicality; it has proved extremely difficult to make robust and repeatable fmed points at higher temperatures. For example, pure palladium in an alumina crucible has been tried ", but found to be too fragile to use. It seems that the combination of graphite crucible and metals insoluble to graphite is fortuitous as it allows pure metal phase transition temperatures to be easily realised. In general though crucible and fixed-point material react with unpredictable effects. In 1999 a solution was proposed by Y. Yamada of NMU (then NRLM) relying on eutectic alloys of metal and graphite '). These, it was expected, would have transition temperature independent of the exact carbon concentration, and so graphite could be used for the crucible without fear of contamination. In the last five years this idea has been a productive area of research, to the point where metalcarbon eutectic fixed-points from approximately 1300°C to 2500 "C have now been proposed as secondary reference points for F S 90. In the longer term they may lead to a redefinition of the F S . The range can be increased still further by using eutectic alloys of metal carbides and carbon 'I, in principle up to 3375 "C (TaC-C). The advantage of using eutectic alloys, rather than the pure materials associated with the ITS-90 fixedpoints, results from the solubility of graphite in metals. Graphite, the material of choice for crucible construction, is insoluble in all the metals used as defining fixed-points of ITS-90 (although standard

-802-

PR0002/04/0000-0802W O O 0 2004 SICE

texts suggest gold forms a eutectic alloy with carbon, wire bridge calibration of thermocouples indicate any change to the freezing temperature is much smaller than the phase diagram indicates). In all metals with melt temperatures above copper (1084 "C), even if no chemical reactions occur carbon dissolves into the fixed-point material changing the transition temperature. A number of research groups have constnrcted eutectic based fixed-points, and it appears that, as expected, the eutectic transition temperature is not affected by the exact graphite concentration of the alloy. Sa far as can be determined individual fixed-points remain stable, giving the same transition temperature after repeated melt and freeze cycling '). Therefore eutectic alloys of metal and graphite can be made as reference standards with graphite crucibles. In order to qualify as fixed-points these sources need to demonstrate reproducibility as well as repeatability. That is, while each individual cell may give the same temperature time after time, different cells from different makers may give different values for what is nominally the same material. It is therefore necessary to demonstrate reproducibility between sources, with different materials and different construction, if eutectic fixedpoints are to be used to define particular temperatures. The requirements for ITS defining fixed-points are demanding. Reproducibility better than 100 mK is suggested. To make reproducible fixed-points and comparative measurements with uncertainties less than 100 mK at 2500 "C to 3000 "C is proving to be extremely challenging and a lot of effort is going into this area. However the standards €or secondary fixedpoints are lower. Typical NMI uncertainties are much higher than 100 mK at high temperatures, and these uncertainties are likely to increase by a factor of 10 or more at the point of use. Therefore for industrial usage a fixed-point with an uncertainty even of one degree could be of benefit. Such references would even be of benefit to NMh as they allow quick and easy scale checks. Regardless of whether NMh need to improve their uncertainties, there is no doubt that industry could benefit from a scale that has less dependence on instrument characterisation and would be simpler to verify. For example, a requirement for 1 "C accuracy from radiation thermometers at a fairly modest temperature of 1300 "C presents little difficulty €or an NMI, but may already be problematic for a calibration laboratory. Relatively small improvements to NMI scales may not have much impact, NMI level uncertainties for calibration laboratories would. However these benefits can only become reality if fixed-points can be made robust enough to withstand repeated cycling to high temperature. This has caused

problems for some materials and if not solved could limit the usefulness of any future eutectic-based scale. The experience of fixed-point construction at NPL including this and other problems is demit& here. It is hoped that this paper will benefit others who may wish to make their own fixed-points.

2. Development at NPL The requirements for a fixed-point are repeatability of an identifiabIe feature (presumed to be the melt or freeze temperature), reproducibility between different sources, and practicality. It should be possible to realise a unique and specified temperature simply by following a written set of instructions.

It has been shown, theoretically at least, that three such fixed-points allow a scale to be created that needs minimal instrument characterisation, yet gives uncertainties comparable to that of the reference point uncertainty ')* 'I. This would allow anyone, anywhere, to realise a temperature scale that closely corresponds to the scales used elsewhere in the world without the need to transport any artefacts, Simply by having a relatively inexpensive (compared to the cost of maintaining the ITS-90 scale) facility of furnace, fixed-points and pyrometer, scale uncertainties of 12 * C can be realised up to 2500 'C and beyond. However for practical use the fixed-points should be robust enough for repeated use and also the feature in the melt or freeze that is to be assigned a temperature value should be unambiguously identifiable.

3. Practicality Issues NPL has been involved in making and assessing eutectic-based fixed-points since 1999. A number of practical issues have arisen, some of which were not apparent during development but only became clear at a relatively late stage as comparison studies at the highest level of accuracy were canied out.

.

The aspects considered are: The thermal expansion of graphite Fixed-point material composition The effect of furnace performance.

3.1 Thermal expansion Graphite is strong and easily machined. It is cheap to obtain even at high punty and has good emissivity. However it was found during development work that certain fixed points were prone to breakage. Previously made fixed-points at NPL 9, had shown no problems. Subsequent rhenium fixed-points, and then molybdenum-carbide fixed-points proved to be

-803-

extremely fragile. A typical breakage is shown in. Fig. 1. The fact that crucibles started to break was presumed to be due either to changes in the construction of the fixed-point crucible or due to use of a different grade of graphite. The fixed-point designs were checked for possible weaknesses resulting from small changes in design between the two sets, and from changes resulting from differences in machining. The strength and resistance to thermal shock of different grades of graphite were investigated.

D

~

l

b

X

l

~

~

E

O

3

J

w

o

Temperature K

Fig. 2. Values of linear expansion (percentage change in original length) for various fixed-point materials. Poco AXM is representative of graphite. Values from ' O ) * 'I)

Fig. 1. Broken molybdenum carbide-carbon eutectic fixedpoint.

The previously used graphite (Ringsdorff RW010) became unavailable after the Ringsdorff works was taken over by SGL. A change was made to using Poco SFG graphite. The way the fixed-points failed suggested thermal expansion was the cause, with failure in cases where the coefficient of thermal expansion (CTE) of the fixed-point material was less than that of the graphite crucible. In this case the crucible contracts more when cooled than the ingot. Depending on the mismatch in CTE and the temperature difference, enough strain can be induced to crack the crucible. Available data (Fig. 2) indicated that iridium was next most likely to break after rhenium. Due to the high cost of iridium, development of this fixed-point has been delayed untiI this issue has been resolved. Iridium-C, rhenium-C and molybdenum carbide-C have melting temperatures of 2290 "C, 2474 "C and 2584 "C respectively. A lack of practical fixed-points in the range 2000 "C to 2700 "C would restrict any possible future scale and would also limit options for sub-ranges.

It was also found that while the problem of breakage occurred with some workers in the field 12), in other cases no breakages occurred 13). There is wide variation in the thermal expansion of different grades of graphite 'I), Different manufacturers figures claim values more than 4 times that of others. The Poco graphite used is near the upper end of the range, while the previously used Ringsdorff is much lower. Different grades of graphite have been tried with rhenium and molybdenum-carbide fixed-point$. Ringsdorff RW010 and Schunk FP379 have both been used with rhenium without further breakage problems. A problem with the furnace coaling water meant switching off the furnace power supply while at 2200 "C with the rhenium fixed-point (made from RWO10) in place. The resulting rapid cooling to room temperature did not break the crucible, and it has since been extensively used for scale comparisons 14) and reproducibility studies without problems. Graphite from the Ringsdorff plant believed to be very similar to the RWOlO grade is now available from SGL. It was concluded that the graphite used for the crucible should have lower thermal expansion than the fixedpoint material. Molybdenum carbide-carbon fixedpoints made with Ringsdorff and Schunk graphite both broke and it is suspected that this may not be suitable as a fixed-point material. This may also be due to the complicated nature of the phase diagram, with four structural phase changes occumng in the solid state as the fixed-point cools to room temperature. The thermal expansion of most of these phases is unknown.

3.2 Fixed point material composition During development at NPL experience was gained and the plateau shape of the eutectics improved. It was found that when making the fixed points using material close to the published eutectic composition the amount needed was usually less than that calculated to fill the crucible. Even taking account of

-804-

the filling method leaving a space at the top of the fixed-point the ingot density was less than expected. The eutectic alloy as a Iiquid appears to be much more viscous than is generally found for pure metals, making it difficult to form a uniform, solid ingot. It seems to be that using material in powder form, and at the compositions used, results in voids in the ingot. In some cases multiple peaks in the freeze suggested the ingot was in several discrete lumps (Fig. 3).

24W 2479

2178

24n 2476 2475 2474

2173

Fig. 4. Comparing different realisations of rhenium-carbon eutectic melt. ....I ..... ./ ......

\

~

4

3.3 Furnace performance The importance of furnace conditions has been shown

in a number of compatisons. BPM, NMU and NPL 1

17351



n Tlm

Fig. 3. Multiple peaks in platinum freeze indicate poor ingot formation.

The effect of this became clear during comparison measurements between BIPM, NMIJ and NPL Is)with the NPL cells showing more of a slope in the melt curve. Investigation after this comparison showed that the NPL rhenium cell blackbody cavity was not fully surrounded by the metal ingot. The outer crucible (Poco graphite) cracked during the comparison. This was machined off and showed that the cell had not been filled properly. The top of the blackbody tube was exposed over a significant area. The ingot and blackbody tube were installed in a new outer crucible made of Schunk FP379 graphite, and pure rhenium was added. Subsequently fixed-points are being made with substantially lower carbon concentration ”), and it was found that this could improve the melt and freeze plateaux. Fig. 4 shows the improvement in rhenium melt curve due to the better ingot.

fixed-points were realised in two different furnaces (NMIJ Nagano and NPL Thermogage) with different results Is). Fig. 5 shows the qualitative difference found with one of the rhenium cells in the different furnaces. Although the melt curve looks very different, the actual melt value determined from the point of inflexion is unchanged within the uncertainties. 1

Fig. 5. BIPM rhenium fixed-point measured in Nagano and Thermogagefurnaces on consecutive days.

Following the problems described above with the cooling supply of the Thennogage parts of the electrodes were replaced and there was a marked improvement. It is thought that degradation of some of the components had resulted in increased temperature gradients in the furnace. This effect appears very similar to incomplete filling of the crucibles. More significant differences in melt temperature were found comparing rhenium fixed-points constructed by

-805-

NPL and VNIIOFI 13). In this case measurements made with a filter radiometer and using a Vega UHTBB 3500 gave good agreement on the freeze, but the NF’L cell melted 0.6 ‘C higher. In the Thermogage furnace however the melts agreed while the VNIIOFI freeze was 0.6 “C higher. Because of the difficulty in obtaining good melt and freeze curves with the VNIIOFI cell in the Thermogage it is believed that the larger physical s i x (60 mlong compared to NPL 40 mm) is at least partly to blame for this.

.

4. Conclusions Correct choice of graphite is an important consideration. Robust fixed-points can be made so long as the graphite thermal expansion is lower than that of the fixed-point material. Some of the factors considered here are perhaps insignificant at the uncertainty levels likely if M-C fixed-points are used as secondary references. Comparisons involving the NPL Thennogage before improvement, and an NMU Nagano furnace gave qualitatively very different results, but differences of just 100 - 200 mK in terms of melt temperature. This also includes the effect of poor ingot formation. For measurements at uncertainties of 1 O C the issue of furnace performance may not be important. Where the furnace may have a large effect is if the fixedpoint is too large. Since there is no melt plateau, the melt temperature is taken to be the point of inflexion in the melt curve which is independent of the ingot mass 16), there is no benefit to a cell larger than the aperture and emissivity require. Therefore fixed-point cells should be designed to be compact. Poor filling can result in badly defined melt curves. Making fixed-points with below eutectic carbon composition improved this. As with furnace gradients the effect in many cases may not be significant, but no extra effort is involved in making a good ingot than a poor one.

References [ 11 CcT196 Recommendation T2, see also T. J. Quinn, “News from the B I P M , Metroloeia, Vo1.34,

pp.187-194, (1997). [2] Machin G., McEvoy H. C., Boyes S.I., “A realisation of the Pd Point at NPL”, Pmceedings of Tempmeko 96, ed. P Marcanno pp301-303, (1997). [3] Yamada Y., Sakate H., Sakuma F., Ono A., “A possibility of practical high temperature fixed points above the copper point”, Proceedings Tempmeko 99, ed. I. F. Dubbeldam and M. J. de Croot, pp 535-540, (1999). [4] N. Sasajima., Y. Yamada, F. Sakuma, “Investigation of fixed points exceeding 2500 “C using metal carbide-carbon eutectics”, Proc. TMCSI vol 7, ed D. Ripple, pp 279-284, (2003). [5] Binary Alloy Phase Diagrams, ed. Massalski, ASM (1990). [6] D. Lowe, G. Machin, “Development of metal-carbon

eutectic based high-temperature fixed-points for reproducibility studies”, Tempmeko 04, to be published. I71 Bloembergen P., Yamada Y., Yamamoto N., Hartmann J., “realising the high-temperature part of if future lTS with the aid of eutectic metal-carbon fixed-points”,Proc. TMCSI vol7, ed D. Ripple, pp 291-296, (2003). [8] Saunders P., White D. R., “Physical basis of interpolation equations for radiation thermometry”, Metrologia 40, pp 195-203, (2003). [9] D. b w e , G. Machin, “Development of metak-carbon eutectic blackbody cavities to 2500 “C at NPL”, Proc Tempmeko 2001, pp 519-524, (2003). [IO] Touloukian Y.S., Kirby R. K., Taylor R. E., Lee T. Y. R., ‘“nophysical properties of matter Vol 12 Thermal expansion Metallic elements and alloys”, IFVPlenum (1975). 11 11 Touloukian YS.,

R., ‘“nophysical

Kirby R. K., Taylor R. E., Lee T. Y. properties of matter Vol 13 Thermal

expansion Nonmetallic solids”, IWlenum (1977). 1121 Y. Yamada, personal correspondence [13] E. R. Woolliams, 3.B. Khlevnoy, S. R. Montgomery, D.Lowe, N. J. Harrison, N. P. Fox, S.A.Ogarev, V. B. Khromchenko, V. I. Sapritsky, Tempmeko 04, to be published [14] G. Machin, C. E. Gibson, D. Lowe, D. W. Allen, H. W. Yoon, Tempmeko 04, to be published I151 G. Machin, Y. Yamada, D. Lowe, N. Sasajima, K. Anhalt, I. Hartmann, R. Goebel, H. McEvoy, Tempmeko

NPL is continuing research and maintains close links with N M U in this area. The experience being gained

04,to be published [I61 Sasajima N., Yamada Y., Zailani B. M., Fan K., Ono

will lead to further improved fixed-points that will be assessed as a possible means of defining an interpolation based temperature scale.

eutectic fixed-point blackbodies” Proc Tempmeko 200 1, pp

ACKNOWLEDGEMENT This was part supported by the European Commission “GROWTH” Programme Research Project ‘Novel high temperature metal-carbon eutectic fixed-points for

A., “Meking and freezing behaviour o f metal-carbon

501-506, (2002).

Contact - David Lowe dave.lowe8nol.co.uk +44 20 8943 63 12 (tel) +44 20 8943 6755 (fax)

Radiation Thermometry, Radiometry and Thermocouples” (HIMERT), contract number: G6RD-CT-2000-006 10

-806-