Content uploaded by Jonathan Pearce
Author content
All content in this area was uploaded by Jonathan Pearce on Jul 07, 2016
Content may be subject to copyright.
A system for high-temperature homogeneity scanning of noble-metal
thermocouples
E. Webster1, R. Mason1, A. Greenen2, J. Pearce2
1Measurement Standards Laboratory of New Zealand
PO Box 31310 Lower Hutt 5040
New Zealand
2National Physical Laboratory
Hampton Road, Teddington, Middlesex, TW11 0LW
United Kingdom
Email: emile.webster@callaghaninnovation.govt.nz
Keywords: Drift, Homogeneity scanning, Inhomogeneity, Noble-metal, Thermocouples,
Uncertainty
Abstract
Noble-metal thermocouples are amongst the most widely used thermocouples for high-
temperature process measurement and as references. Although they are less susceptible to
inhomogeneity effects than the more-common base-metal thermocouples, inhomogeneity is
still the major source of uncertainty. Currently, most estimates of the uncertainty due to
inhomogeneity are based on thermocouple specifications or historical performance of similar
thermocouples. It is not common for the inhomogeneity to be measured directly, in part
because there is no accepted method for measuring the inhomogeneities, and in part because
there is no conclusive evidence linking the magnitude of inhomogeneities determined at the
scanning temperature to the effects of the same inhomogeneities at other temperatures. This
paper describes an inhomogeneity scanner able to be fitted to sodium heat-pipe furnaces to
operate between 600 C and 1000 C. Comparison of scans made at 100 C demonstrates the
scalability of some types of inhomogeneity in Type S and R thermocouples. It is concluded
that for Type R and S thermocouples, a robust uncertainty assessment can be obtained from a
scan made at a single temperature.
1. Introduction
Inhomogeneity causes a deviation in EMF as a function of position when the thermocouple
exposed to a constant temperature gradient. Homogeneity scanning is the process whereby a
thermocouple is moved between two stable isothermal zones, separated by a well-defined
temperature gradient. The width of the temperature gradient influences the scanner’s
resolving potential. Typically, the higher temperature zone should be low enough not to
induce changes in the thermocouple under test. By monitoring the thermocouple EMF during
the scanning process a measure of inhomogeneity is acquired. Although the process of
homogeneity scanning has been recognised for over a 100 years [1], it is only in the last 10
years that any quantitative analysis of the inhomogeneity data has been described [2-4].
Reference [4] provides the most rigorous approach to date, outlining the factors affecting
scanning and describing methods that enable comparisons between different scanners.
Although the methods described in that work allow an accurate assessment of the uncertainty
component due to inhomogeneity, the uncertainty may only be valid for temperature
measurements made near the scanning temperature. At higher scanning temperatures
inhomogeneities may be introduced or erased by the scanning process.
At the time of writing it is thought that the inhomogeneity signatures scale approximately
linearly with temperature [5,6]. If this scaling is true, then an uncertainty value derived from a
scan made at one temperature will be equally valid for all temperature measurements up to
some threshold where recrystallisation of the atomic structure occurs [7]. Inhomogeneities can
be divided into two categories; reversible and irreversible. Reversible inhomogeneities in
base-metal thermocouples include crystallographic ordering and strain defects (both point and
slip plane types). Irreversible inhomogeneities are predominantly caused by contamination
and oxidation. The contamination usually arises from diffusion of various alloying elements
and material from the immediate environment. For bare wires the oxidation of different
alloying elements over a range of temperatures is a major cause of irreversible
inhomogeneities. Ignoring contamination, the drift processes are somewhat simpler in noble-
metal thermocouples, with the two principal causes being oxidation and lattice structure
changes (vacancy movements and/or crystalline ordering [8]). Using a high-temperature
recrystallisation anneal, both oxidation (unlike base-metal types) and crystallographic effects
have been shown to be reversible in noble-metal thermocouples [8,9]. Experimentally,
recrystallisation temperatures have been found to be approximately 1100 °C in Pt/Rh alloys
[10] and 800 °C in platinum [11].
For noble-metal thermocouples homogeneity scans are often conducted in high-temperature
fixed-point furnaces; however, these scans do not constitute a true scan as discussed here [4].
A number of homogeneity scanner systems operating above 500 °C have also been described
in the literature; see, for example [12,13]. Yet another study employing a sodium heat-pipe
describes a system suitable for scanning at multiple temperatures [6]. Seldom, though, is the
temperature gradient given for the entrance of these furnace scanners. Hence, an assessment
of the temperature gradient applied to the thermocouples under test is not always possible and,
consequently, nor is information on the potential scanning resolution. If the resolution of the
scanner is not known, then nor is the correct magnitude of any inhomogeneities detected by
the scanner. The situation is further complicated by the erasure or generation of
inhomogeneity information during the scanning process.
In the study by Kim et al [6] they aimed to show a scaling relationship for the
inhomogeneities in base-metal thermocouples at various temperatures. Yet, due to a number
of complicating factors no strong evidence was demonstrated for this claim. As has now been
shown from two more recent studies [14,15], the drift mechanisms in base-metal
thermocouples start occurring at lower temperatures than previously thought; as low as
100 °C in many cases. The drift mechanisms included cold-work annealing, crystalline
ordering and diffusion [10,14,16-19]. None of these processes act in isolation at a single
temperature. To make matters worse there is also a high variability in the manufacturing
formulation of these alloys. Hence, no adequate scaling system seems likely for base-metal
thermocouples that would allow accurate forecasting of the effects of inhomogeneity.
In a similar study, Jahan and Ballico [5] conducted a number of high-temperature
homogeneity scans on Type R and S noble-metal thermocouples. The thermocouples had been
used for long periods in industrial applications. All the thermocouples where first given an
1100 °C quench anneal (QA) [8] to remove any reversible drift components, due to crystalline
ordering and oxidation effects. Therefore, the scans reflect an assessment of irreversible
inhomogeneity due to contamination. Although the results of the study suggest a reasonable
level of agreement for the inhomogeneity measured at the different scanning temperatures,
accurate comparisons were not possible as the resolution of each scan varied as a function of
scanning temperature. To build confidence in the scalability of inhomogeneity the resolution
of the scanner must be sufficiently similar at different scanning temperatures, and both
reversible and irreversible drift components must be assessed. This will enable a more
realistic estimate of the uncertainty due to inhomogeneity.
To confirm the preliminary work on Type S and R thermocouples by Jahan and Ballico [5] a
carefully designed high-temperature scanner was manufactured. Three necessary properties of
the scanner can be identified: the scanner must have an entrance gradient narrower than the
minimum in-use thermal conduction length to ensure the scanner is able to resolve details in
the thermocouple homogeneity to a higher resolution than any errors caused by in-use
temperature gradients; the scanner must also operate over a wide enough range of
temperatures to provide confidence in the scalability of inhomogeneity; lastly, the temperature
gradient of the scanner must be similar at all scanning temperatures to allow accurate
comparison of inhomogeneity data. To meet these requirements the system developed for this
study uses a sodium heat-pipe having a highly uniform temperature zone and is capable of
being operated at a range of temperatures. The entrance design uses thermal radiation
shielding coupled with a water chiller to provide a narrow temperature gradient.
This paper first gives a detailed description of the apparatus and an experimental method
suitable for measuring the inhomogeneity at a number of temperatures. Results from
measurements on Type R, S and Pt/Pd thermocouples are then provided with a discussion on
the drift mechanisms and analysis of the drift magnitudes at the different scanning
temperatures. Lastly, recommendations are given along with concluding remarks on the
validity of scaling inhomogeneities based on scans made at a single temperature.
2. Experimental
2.1 Scanning furnace
To provide a highly uniform temperature zone for the scanner a Carbolite CTF 12/100 single
zone furnace with Eurotherm 808 Controller was chosen. The furnace utilises a heat-pipe
inner sleeve with a length of 1 m, an outer diameter of 100 mm and an inner diameter of
60 mm. A quartz tube with an 8 mm external diameter and 6 mm internal diameter spans the
entire length of the furnace, providing contamination protection for the thermocouples from
the nickel/chromium heat-pipe sleeve. The furnace can safely operate between 600 °C and
1000 °C over long periods.
Pt foils and
silicon wafers
Press plate
Ceramic housing
Baffled water jacket
Aluminium flange
Sodium heat-pipe
furnace
Entrance apparatus
Quartz tube
2.2 Scanner entrance apparatus
The entrance design, shown Figure 1, provides thermal shielding to the thermocouple before
entering the furnace, thus yielding a narrow temperature gradient. The apparatus incorporates
a machinable ceramic housing with three 1 mm thick silicon wafer disks layered with
platinum foil. The foil/silicon disks are spaced approximately 3 mm apart with alumina wool.
This arrangement provides a useful radiation barrier, radiation being the dominant heat
transfer mechanism at temperatures over 400 °C. Behind the last silicon disk a stainless steel
press-plate compresses the alumina wool and silicon disks, locating these components. At the
rear of the ceramic housing a machined aluminium flange supports a cylindrical baffled water
jacket, which was connected to a Grant R1 water chiller with a Grant T100 thermostat set to
12 °C. The platinum foils, silicon wafers and water jacket all have an axial hole sufficient to
allow the quartz tube, described in Section 2.1, to pass through. The entrance apparatus is
130 mm long when assembled.
(A) (B)
Fig. 1 (A) Design of sodium heat-pipe entrance apparatus and (B) photograph of complete
assembly
2.3 Linear drive mechanism, software and DVM
To provide motion control for the thermocouples during scanning a Parker HD085 linear
actuator and Micromech Ltd controller were used. This system has sub-millimetre repeatable
movements over distances of 1.2 m. LabVIEW software, similar to that described in [14], was
used to provide automation, DVM communication and logging. The DVM was a calibrated
Keithley 2182A nano-voltmeter with a total uncertainty of 300 nV. Thermocouples were
terminated to solid copper extension leads which were in turn connected to the Keithley
2182A DVM. The thermocouple/copper junctions were maintained at 0 °C in a Fluke 9101
Zero-Point Dry-Well reference, stable to better than 20 mK.
2.4 Interrogation of the sodium heat-pipe uniformity and scanning temperature gradient
To gain information on the uniformity of the temperature within the sodium heat-pipe and the
temperature gradient of the entrance apparatus an industrial platinum resistance thermometer
(IPRT) 2 m long (connected to a Fluke 525 bridge) was used. The IPRT was scanned from
ambient temperature through the furnace entrance apparatus and back to ambient at the far
end of the furnace at 10 mm·min-1. To verify no significant change had occurred in either the
IPRT or the temperature of the heat-pipe a reverse scan was also performed. The IPRT has a
small sensing element with minimal conduction losses. Therefore, the temperature gradient at
the entrance to the furnace measured with the IPRT was expected to be close to the true
temperature gradient, thus allowing estimation of the conduction losses when compared with
thermocouple scans. Prior to scanning through the furnace the IPRT sensing element was first
stabilised in a Jofra RTC-700 Dry-block calibrator by ramping the temperature between
400 °C and 600 °C twenty times over an approximately 48 hour period, after which the
IPRT’s repeatability was found to be within 50 mK at 600 °C. Multiple measurements were
then made at the water triple-point (TPW), zinc-point and aluminium-point to calibrate the
IPRT.
A thermally symmetric Type S thermocouple of similar construction to that described in [4]
was used to assess the conduction losses during scanning and gain information on the
temperature gradient experienced by a thermocouple under test. By locating the junction
90 mm from the tip the conduction region is symmetric about the measurement junction. The
thermocouple in all other respects is similar to a standard reference thermocouple: the
thermoelements are 0.5 mm in diameter and the sheath is made from 4 mm diameter twin-
bore high-purity alumina. Scans were made with the thermally symmetric test thermocouple
at 600 °C, 700 °C, 800 °C and 900 °C. Reference temperatures used for calculating the
Seebeck coefficient were derived from a reference Pt/Pd thermocouple held stationary for
10 min at 200 mm insertion into the sodium heat-pipe. Stabilisation of the entrance apparatus
and furnace was confirmed by holding the same Pt/Pd thermocouple at the point of greatest
temperature gradient in the entrance apparatus. Stabilisation time was typically 3 h.
2.5 Sample Thermocouples
To test the performance of the scanner system a number of Type R and S noble-metal
thermocouples were scanned at temperatures of 600 °C, 700 °C, 800 °C and 900 °C. These
temperatures were considered representative of the range where the scalability of
inhomogeneity could be assessed. Temperatures over 900 °C were avoided as recrystallisation
of platinum-rhodium thermoelements can start to occur [8]. A lower temperature scan of
100 °C was used on two of the thermocouples to gauge the level of modification caused by
the higher temperature scans. Scans at 600 °C were also made on a working standard Pt/Pd
thermocouple after a number of temperature measurements and after repeat anneals.
Type R and S thermocouples, two of each, were provided by NPL (Manufactured by CCPI,
UK) and are expected to be representative of long term high-temperature use. These
thermocouples were used for the European Metrology research programme ‘HiTEMS’ [20].
All four thermocouples had been used in the same furnaces and fixed points for long periods
well in excess of 1200 °C. Warping of the alumina protective sheaths required the 0.5 mm
thermoelements to be removed from their sheaths and carefully threaded into new 4 mm twin-
bore high-purity alumina sheaths, using a draw-through technique [21]. The warping of the
sheaths was not sufficient to impair measurements with these thermocouples, but would
otherwise prevent them from being scanned.
Representative samples, one of each of the R and S thermocouples, were first scanned in
MSL’s NZ dual heat-pipe (100 °C) scanner [14] prior to being scanned in the high-
temperature sodium heat-pipe scanner. Table 1 provides details on the thermocouples and the
scanning sequences. After the highest-temperature scan at 900 °C the Type R and S
thermocouples were given a second scan at 600 °C to verify changes caused by higher
scanning temperatures. Finally a 2 h 1100 °C QA was applied to these thermocouples and
another 600 °C scan was conducted. This last step enabled differentiation between reversible
and irreversible inhomogeneities.
Table 1. Description of thermocouples, their history and homogeneity scans
Thermocouple
Type
History
Low-T scan*
High-T scans†
114422 R1
R
HiTEMS
No
Yes
114422 R2
R
HiTEMS
Yes
Yes
114421 S1
S
HiTEMS
No
Yes
114421 S2
S
HiTEMS
Yes
Yes
Pt/Pd3
Pt/Pd
Eutectic/anneal
No
Yes
*Low temperature scan made in dual heat-pipe scanner at 100 °C
†High temperature scans made in sodium heat-pipe scanner at 600 °C, 700 °C, 800 °C and 900 °C
To provide further data on the homogeneity of palladium-based thermocouples some
representative scans were also made on a Pt/Pd working standard, used on a regular basis at
NPL. This Pt/Pd thermocouple is used to check the stability of the sodium heat-pipe furnace
and to monitor various fixed-point melts up to the Co-C high-temperature fixed-point (HTFP),
and was subjected to a number of 1100 °C quench anneals. There is some data in existing
literature on the homogeneity of Pt/Pd reference standards, with most scans being undertaken
in fixed-point furnaces, see, for example [22]. A number of purpose-designed homogeneity
scanners have also been used [13,23-26]. However, the accuracy of these scans is uncertain
for the reasons discussed in Section 1.
3. Results
A number of preliminary scans of the furnace/entrance apparatus were undertaken to optimise
the scanner system. The resolving potential of the scanner can be determined by its kernel and
is defined by dT/dx of the temperature gradient. Further information on scanning kernels can
be found in [14]. Dynamic scanning was used as opposed to scanning with pauses as this has
been shown to give results with the minimum of conduction error [4]. A scanning rate of
10 mm·min-1 was found to yield the most symmetric scanning kernel while keeping the
scanning time to a minimum.
3.1 Measurement of the scanner temperature gradient and resolution
The temperature gradient at 600 °C measured using the IPRT, is shown in Figure 2 and can be
seen to occupy a distance of approximately 50 mm. The conduction error caused by the
temperature survey procedure using the IPRT is significantly less than an equivalent alumina-
sheathed thermocouple and provides useful information on the true gradient of the scanner
system. Higher-temperature scans using the IPRT were not possible due to the 600 °C limit
for the sensing element. Plots for the temperature gradients as measured with the thermally
symmetric survey thermocouple are also shown in Figure 2 for the four scanning
temperatures. These plots reveal the significant conduction region either side of the junction
when compared to the IPRT measurement. Evidently, the scanner resolution is restricted by
the properties of the thermocouple, not the temperature gradient of the scanner.
Fig. 2 Gradient profiles of sodium heat-pipe and entrance apparatus, as measured with IPRT
at 600 °C and with thermally symmetric survey thermocouple at 600 °C, 700 °C, 800 °C and
900 °C.
The four temperature gradients measured using the symmetric survey thermocouple was
differentiated to identify their respective kernels, as seen in Figure 3. These kernels were then
least-squares fitted using a beta probability density function (pdf). The kernel was modelled
with four parameters, as described in [4]. The four parameters are made up of two curvature
components, alpha and beta, and two distance parameters, which define the kernel’s spatial
limit of influence. Modelling the kernels removes noise and acts as a smoothing function. The
standard deviation of the scanning kernel has been shown [4] to be a good measure of the
scanner’s resolving potential. The standard deviation for the kernels in this study are
20.25 mm, 19.28 mm, 17.04 mm and 17.14 mm, respectively, for 600 °C, 700 °C, 800 °C and
900 °C. Kernels with a lower standard deviation have better resolving potential.
Fig. 3 Kernels for a thermally symmetric survey thermocouple at 600 °C, 700 °C, 800 °C and
900 °C
3.2 Sodium heat-pipe uniformity and stability
To confirm temperature uniformity and stability of the sodium heat-pipe the IPRT was
scanned along its entire length. Figure 4 is a plot of the insertion and removal temperature
profile. A 3 °C linear slope is visible from the entrance to the exit of the furnace at the far end,
which is possibly due to light-piping of radiation in the quartz tube. The stability of the
furnace over the four hour scan period is within a few milli-Kelvin. The effect of the 3 °C
gradient on scanning results is negligible. The absolute error in scanning a 700 mm long
thermocouple could not exceed the non-uniformity divided by the scanning temperature (in
Kelvin), or ~1.5 K/873 K (~0.17%). The value of 1.5 K is used as the tip of a 700 mm
thermocouple would not extend beyond half the usable furnace length. At higher scanning
temperatures the non-uniformity is expected to increase, but the error should remain roughly
constant as the error is the ratio of the non-uniformity to the scanning temperature.
Fig. 4 Uniformity of sodium heat-pipe furnace when scanned in and out, using a calibrated
IPRT
3.3 Sample scans of used Type S, R and Pt/Pd thermocouples
Scanning data presented in Figures 5 to 9 is formatted as a percentage change in Seebeck
coefficient. This format is insensitive to scanning temperature and implicitly allows
comparison between scans of the same thermocouple made at different temperatures as well
as those of other thermocouple types. This presentation is more fully described in [14], but is
essentially the percentage error in Seebeck coefficient from reference functions
ref
refmeas
SSS
S
S
, (1)
where Smeas is the average Seebeck coefficient inferred from the measured voltage and the
temperatures measured either side of the scanner gradient, and Sref is the corresponding
Seebeck coefficient inferred from the thermocouple reference tables [27].
In Figure 5 plots are shown for 114421 S1 in the as-received state, scanned at 600 °C and
successive scans up to 900 °C. A repeat scan at 600 °C shows very little change in the
Seebeck coefficient due to the higher temperature scans. The initial scan at 600 °C has
removed a small peak at ~500 mm and once removed appears stable for later scans. There is
reasonable agreement in the magnitude of inhomogeneities at different scanning temperatures
except in the vicinity of a large depression located at 450 mm. The progressively earlier
detection of the thermocouple signatures at approximately 100 mm is caused by the
improving resolution of the scanner at higher temperatures. As the scanning temperature is
increased the resolution of the scanner also increases due to improved radiation heat transfer.
It might be thought this explanation should apply to the apparent increasing magnitude of the
depression at 450 mm. However, this is not true and is discussed in more detail in Section 4.4.
A final 2 h 1100 °C QA removes most of the inhomogeneity signatures and returns the
thermocouple to a near homogeneous state. Two persistent defects are a gentle gradient
between 200 mm and 400 mm and the residual dip at 450mm. These persistent defects render
this thermocouple too inhomogeneous to be used as a reference standard.
Fig. 5 Homogeneity of 114421 S1 when measured in sodium heat-pipe scanner at indicated
temperatures, before being given an 1100 °C quench anneal
In Figure 6 the thermocouple 114422 R1 has been exposed to similar temperatures and for
similar times to that of the previous thermocouple 114421 S1. The level of inhomogeneity is
not as great, but its location and the annealing processes occurring due to high temperature
scans are again visible between 450 mm and 550 mm. The consistency in the magnitude of
inhomogeneities at different scanning temperatures again suggests some linear scaling with
temperature. As with Figure 5 the improved resolution of the scanner at higher temperatures
leads to earlier detection of the homogeneity signature at approximately 100 mm. Unlike the
Type S thermocouple a 2 h 1100 °C QA is able to restore this thermocouple to an almost
homogeneous state.
Fig. 6 Homogeneity of 114422 R1 when measured in sodium heat-pipe scanner at indicated
temperatures, before being given an 1100 °C quench anneal
In Figure 7 extra homogeneity detail is provided by a scan made prior to the high-temperature
scans, using the MSL dual heat-pipe scanner operated at 100 °C. The profile is similar to that
of Figure 5 with a large depression at 450 mm and a more noticeable peak at 500 mm.
Annealing is occurring beyond 500 mm for scans made at higher temperatures, where the
Seebeck coefficient can be seen to increase by ~0.2%. The MSL scanner has at least 2-3 times
the spatial resolution of the high-temperature scanner, as revealed by the significantly earlier
signal detection at approximately 50 mm. The difference in amplitude of the depression at
450 mm suggests that the drift mechanism for this region does not scale linearly with
temperature, yet the drift features either side do.
Fig. 7 Homogeneity of 114421 S2 when measured in MSL dual heat-pipe scanner and sodium
heat-pipe scanner at indicated temperatures, before been given an 1100 °C quench anneal
The second Type R thermocouple, Figure 8, when scanned at 100 °C appears to have been
exposed to two high-temperature gradients, one at ~450 mm and another at ~550 mm. The
region between 200 mm and 350 mm is similar to that of 114422 R1 with an initial
homogeneous section followed by a small peak at 400 mm. This is then followed by a large
depression at 450 mm and a partial recovery between 500 mm and 550 mm. As with Figure 7
the drift between 500 mm and 550 mm is significantly altered by the higher temperature
scans.
Fig. 8 Homogeneity of 114422 R2 when measured in MSL dual heat-pipe scanner and sodium
heat-pipe scanner at indicated temperatures, before been given an 1100 °C quench anneal
The scans of a working standard Pt/Pd thermocouple are shown in Figure 9. The initial scan,
run1, shows the inhomogeneity along the length of the thermocouple to be about 0.05%. After
a number of hours measuring a Co-C HTFP a slight drop in performance can be seen, likely
due to palladium oxide formation in the furnace gradient region (approximately 200 mm to
400 mm). The thermocouple was then placed in an 1100 °C annealing furnace for three 16 h
periods, after which a gradual recovery was observed, most noticeably in the 100 mm to
150 mm region. A slight dip is evident at approximately 200 mm with a persistent and gradual
reduction in Seebeck coefficient along the remainder of the thermocouple.
Fig. 9 Homogeneity of Pt/Pd3 in initial condition, post Co-C HTFP use, and three consecutive
applications of 16 h 1100 °C anneals.
4. Discussion
4.1 Scanning resolution required to derive a meaningful uncertainty value
The conduction region seen with the symmetric junction thermocouple in Figure 2 is expected
to represent the smallest sustainable temperature gradient (for a typical noble-metal
thermocouple in a 4 mm twin-bore alumina sheath). The IPRT was able to show the entrance
apparatus has a sharper temperature gradient when the survey thermocouple is not present. A
consequence of the large conduction region generated by the thermocouple suggests that
homogeneity scanners used for noble-metal thermocouples do not need a scanning kernel with
a standard deviation of less than ~20 mm. In any real world application in which a wider
temperature gradient exists, the conduction region is anticipated to be even larger. In such
instances localised inhomogeneities will become smeared by the filtering effect of the
temperature gradient. However, inhomogeneities that occur over the length of the
thermocouple, like those seen for the Pt/Pd thermocouple in Figure 9 may become
increasingly significant when placed in wider temperature gradients.
The standard deviations of the four scanning kernels of this study, Figure 3, are similar to the
value derived for the oil based scanner used by NMIA [28], but significantly wider than the
dual heat-pipe scanner used at MSL [14]. From the furnace scans of this study and oil scans
made at NMIA it can be concluded a practical standard deviation for the kernel width should
be between 20 mm and 25 mm if real world estimates of the uncertainty caused by
inhomogeneity are required. Higher-resolution scans may need to be filtered using a kernel
with a standard deviation of 20 mm to provide comparable uncertainty estimates.
4.2 Sources of Inhomogeneity in Type R and S thermocouples
The migration of rhodium in the form of gaseous oxide at high-temperatures within an
alumina insulator has been suspected for many years [29]. However, due to a lack of
consistent data this has not been conclusively demonstrated. In a recent study [30] platinum-
rhodium thermocouples with an alloy leg containing 17% rhodium were found to be more
stable than alloys with a lower percentage of rhodium after various annealing and aging
cycles. Further evidence for this conclusion can be considered when comparing the drift for
the first 300 mm section in the Type R and S thermocouples, seen in Figures 6 and 8 and
Figures 5 and 7, respectively. The significant depression at 450 mm is consistent with
localised rhodium-oxide formation in the temperature gradient region of the furnace and has
been demonstrated in a previous study [8]. However, other influencing factors have been
shown to affect oxide formation; these include the air gap surrounding the wire, impurities,
grain-size and grain-orientation [31]. After being given an 1100 °C QA both Type R
thermocouples were returned to a more homogeneous state than the Type S thermocouples.
The larger rhodium-oxide depression seen for the Type S thermocouples, when compared to
the Type R, is in agreement with other recent findings [30], for which it was shown
platinum/rhodium alloys containing lower proportions of rhodium are more sensitive to drift
due to oxide formation.
The first 400 mm of the Type R and S thermocouples have been exposed to long periods to
temperatures over 1200 °C. At these temperatures a complex vapour/decomposition process is
known to take place for platinum- and rhodium-oxides. Chaston [32] gives a transition
temperature of 1400 °C in oxygen, at one atmosphere, for complete decomposition of RhO2
gas into solid rhodium. More recent experimental work suggests this temperature is somewhat
lower in air [33]. A study by Li et al [34] revealed a platinum depleted region exists on the
surface of platinum-rhodium alloys heated in oxygen over 600 °C. The sub-surface, only a
few nano-meters thick, is shown to be made up of rhodium and rhodium-oxide. They also
demonstrate that as oxygen diffuses into the surface layer, forming rhodium-oxide, the
platinum, with substantially lower oxide formation energy and with higher partial pressure, is
most likely driven from the surface, transitioning to a PtO2 gaseous state. Whether the
gaseous oxide then decomposes on the thermoelement or on the surround surfaces is not
known. Within the bore of an alumina insulator the oxygen partial pressure will be heavily
influenced by the gaseous formation of both platinum- and rhodium-oxides and is very likely
oxygen depleted [30]. It is suspected these gaseous oxides cling closely to the surface of the
thermoelement and their molecular diffusion is low due to their high mass, but could certainly
migrate over long time periods at high temperatures [35,36]. The gradual slope observed in
the first 400 mm for the Type S thermocouples, seen in Figures 5 and 7, may be evidence of
oxide migration; likewise the small bumps seen in Figures 6 and 8 at 400 mm for the Type R
thermocouples.
The Type R and S thermocouples in this study were additionally protected by a closed end
high-purity alumina sheath during their high-temperature aging. Despite this extra protection,
at temperatures in excess of 1300 °C chemical diffusion of contaminants is a real possibility
given the long immersion times [37]. An incomplete recovery of these thermocouples, when
given a 2 h 1100 QA, indicates some permanent change has occurred in the thermocouples.
Whether this change is due to diffusion of material from external sources or movement of
platinum/rhodium oxide vapours within the sheath remains uncertain.
The small peak seen at 500 mm in Figures 5 to 8, which is more clearly visible in the 100 °C
scan in Figure 7, is consistent with crystalline ordering [8]. The erasure of this peak during the
600 °C scan is in agreement with temperatures reported for atomic mobility [10]. The
magnitude of the Seebeck coefficient to the far right in Figure 7, when scanned at 100 °C,
represents their initial anneal condition, which would appear to be either 1100 °C QA or
1100 °C furnace cooled. This is further confirmed after multiple scans, as the Seebeck
coefficient can be seen to slowly increase, indicating a vacancy annealing process.
4.3 Inhomogeneity in Pt/Pd reference standard
There was some evidence for scalability of inhomogeneity with temperature for the Pt/Pd
thermocouple used in this study, when scanned at temperatures between 600 °C and 900 °C
(data not presented here). However, because the inhomogeneities were small and only one
sample was available, further investigation is required to confirm this. The recovery in
Seebeck coefficient after annealing in the first 300 mm, seen in Figure 9, suggests
contamination of this region. The contamination may have occurred over many uses when
measuring a number of high-temperature eutectics. The cause of the increase in Seebeck
coefficient between 150 mm and 400 mm after multiple anneals is associated with some
recovery process, as it appears to slow after 48 h. It has been suggested by a number of
authors that up to 200 hours may be required to fully stabilise the palladium in Pt/Pd
thermocouples [38-40]. After the first anneal a significant grey/black deposit was observed on
the palladium wire exiting the annealing furnace in the region of the temperature gradient.
Clearly, substantial oxide was present in parts of the wire, now annealed, and had formed over
a number of past uses. The propensity for the oxides to migrate to the temperature gradient
region provides compelling evidence that the volatile noble-metal oxides can be mobile and
has been reported on by at least one other investigator [41].
4.4 Factors affecting linear scalability at different scanning temperatures
A rhodium-oxide depression signature at 450 mm, seen in Figures 5 to 8, formed during long
aging periods, can be used to assess the linear scalability of inhomogeneity due to rhodium-
oxide. Comparisons of high-resolution scans made at 100 °C with those made at the 600 °C to
900 °C show the depression has the same general shape. By plotting the characteristic peak-
to-peak inhomogeneity in the 400 mm to 500 mm region, as seen in Fig. 10, the magnitude of
the depression can be evaluated as a function of scanning temperature. The temperature
dependence in these plots has been fitted using 2nd order polynomials for 114422 R2 and
114422 S2.
To assess the scanner’s changing resolution at the different scanning temperatures a
convolution filter was applied to all Type R and S homogeneity scans. A beta pdf with a
standard deviation of 21 mm and alpha and beta values of 6 was chosen for the convolution
kernel. The kernel was selected to be slightly broader than the widest kernel of the scanner,
which was measured to have a standard deviation of 20.25 mm at 600 °C. The convolution
was also applied at 600 °C and revealed there was little influence on the magnitude of the
rhodium-oxide depression caused by the variation in scanner resolution. Therefore, it was
concluded inhomogeneity caused by rhodium-oxide does not scale linearly with temperature,
and the oxide must not be present when homogeneity scanning or calibrating Type R and S
thermocouples.
For the Type R and S thermocouples the regions not affected by rhodium-oxide do appear to
scale linearly with temperature. However, an annealing process at the scanning temperature of
600 °C has erased crystalline ordering features in the 500 mm to 600 mm region. Evidence for
this erasure can be seen in Figures 7 and 8 when comparing the 100 °C and 600 °C scans.
Consequently, a recrystallisation anneal is also required before homogeneity scanning if the
scanning results are to be used for a calibration. An 1100 °C QA is able to remove both
oxides and ordering features, leaving the thermocouple in a suitable state for low-temperature
homogeneity scanning. From these results some conclusions can be reached: rhodium-oxide is
persistent below 900 °C, but does not scale linearly, whereas ordering does scale linearly but
is erased at temperatures in the vicinity of 600 °C. Therefore, meaningful homogeneity
scanning of Type R and S thermocouples can only be undertaken for irreversible
inhomogeneities. These inhomogeneities do appear to scale linearly with temperature and is in
agreement with results reported elsewhere [5].
Fig. 10 Peak-to-peak inhomogeneity between 400 mm and 450 mm for thermocouples in
figures 5 to 8 at scanning temperatures of 100 °C, 600 °C, 700 °C, 800 °C and 900 °C
4.5 Validity of scans made at different temperatures
If the thermocouple is left in the 1100 °C QA state the scanning temperature must be less than
200 °C to avoid vacancy annealing and crystalline ordering effects [8]. However, if the
thermocouple is subsequently placed in the 450 °C vacancy anneal (VA) state [42], then
scanning may be undertaken at any temperature up to 450 °C. In a previous study [8] the drift
effects in Type S thermocouples were shown to be effectively suppressed at temperatures less
than 500 °C when given the recommended annealing treatments of an 1100 °C QA followed
by a 450 °C VA. Therefore, Type R and S thermocouples scanned according to these
guidelines will produce an inhomogeneity profile sufficient to calculate an uncertainty
component that is valid when measuring temperatures up to 900 °C.
For thermocouples in any other state the scanning may lead to modification of the
thermocouple homogeneity, even when data from the scan suggests otherwise. This study
demonstrated times as short as 30 min at 600 °C are sufficient to significantly alter the
crystalline structure of Type R and S thermocouples. Consequently, for research on the drift
mechanisms in noble-metal thermocouples a high-resolution scanner operated between
100 °C and 200 °C is needed. This temperature range is low enough to avoid changes in the
thermocouple under test.
The success of the scanner system described in this study for identifying inhomogeneities in
noble-metal thermocouples has also been seen in scans made on Type B and 20/40 ‘Land-
Jewell’ thermocouples, and is the subject of a companion paper describing platinum/rhodium
alloyed noble-metal thermocouples. Further investigation is required on the scanner’s ability
to detect changes in Pt/Pd and Au/Pt thermocouples and whether scalability is also true for
these types. Bentley’s work [24,43] on noble-metal thermocouples has the potential to be
substantially added to through the use of the variable temperature sodium heat-pipe scanner
described in this study.
Conclusion
A specialised entrance design for a sodium heat-pipe furnace is described with a narrow
temperature gradient region, suitable for homogeneity scanning noble-metal thermocouples
over the temperature range of 600 °C to 900 °C. The entrance design incorporates
components that minimise both radiation and conduction heat transfer as the thermocouple is
progressively inserted into the high-temperature sodium heat-pipe. This arrangement is shown
to have a spatial resolution of approximately 20 mm, limited only by conduction losses in the
thermoelements. Because of the significant conduction region present in typical noble-metal
thermocouples it has been shown that homogeneity scanners do not need spatial resolution
better than 20 mm to determine the uncertainty due to inhomogeneity.
A number of example scans are included which demonstrate that many of the drift
mechanisms within Type S and R thermocouples scale linearly with temperature, as has been
partially shown in an earlier study [5]. The reversible drift processes, which are sensitive to
scanning temperature, can be removed by an appropriate anneal prior to calibration. Hence,
only the irreversible inhomogeneities need to contribute to the uncertainty budget. Annealing
is especially important as there was evidence for the erasure of some ordering/oxidation
signatures that occurred as part of the scanning process. In scans made on a Pt/Pd
thermocouple the effects of drift caused by eutectic measurements and various annealing
times were also revealed.
It is concluded the apparatus performs with sufficient accuracy and resolution to allow good
quantitative estimates of inhomogeneity within the temperature range of 600 °C to 900 °C.
For thermocouples given an appropriate preconditioning anneal, a robust quantitative
uncertainty value can be derived. This value is specific to the reference thermocouple under
test and does not rely on a type B assessment. However, at temperatures over 1000 °C, atomic
recrystallisation processes prevent the extrapolation of inhomogeneity data and a secondary
uncertainty component will likely be required.
Acknowledgements
The author wishes to acknowledge the financial assistance and resources provided by NPL in
completing this work.
References
1. W. P. White, Phys. Rev. (Series I) 23, 449-474 (1906).
2. J. V. Pearce, Meas. Sci. Technol. 18, 3489-3495 (2007).
3. M. Ballico and F. Jahan, Temperature, its measurement and control in science and industry, vol. 8, part
1, ed. by C. W. Meyer, (AIP, 2013), pp. 544-548.
4. E. S. Webster and D. R. White, Metrologia 52, 130-144 (2015).
5. F. Jahan and M. Ballico, Temperature, its measurement and control in science and industry, vol. 7, part
1, ed. by D. C. Ripple, (AIP, 2002), pp. 523-528.
6. Y. G. Kim, C. H. Song, K. S. Gam and I. Yang, Meas. Sci. Technol. 20, 1-5 (2009).
7. E. H. McLaren and E. G. Murdock, The Properties of Pt/PtRh Thermocouples For Thermometry In The
Range 0 - 1100 °C Part 3, NRCC 17407 edn. (National Research Council Canada, 1983).
8. E. S. Webster, Int J Thermophys (2015). (Under review - Effect of annealing procedure in determining
drift as a function of temperature between 170°C and 900°C in Type S thermocouples)
9. F. Jahan and M. Ballico, Int J Thermophys 31, 1544–1553 (2010).
10. R. E. Bentley, Theory And Practice Of Thermoelectric Thermometry, 1st edn. (Springer, 1998).
11. I. Jursic and S. Rudtsch, Int J Thermophys 35, 1055-1066 (2014).
12. R. E. Bentley, Aust. J. Instrum. Control 4, 4-9 (1989).
13. M. Gotoh, Temperature, its measurement and control in science and industry, vol. 7, part 1, ed. by D.
C. Ripple, (AIP, 2002), pp. 481-484.
14. E. S. Webster, D. R. White and H. Edgar, Int J Thermophys 36, 444-466 (2014).
15. E. S. Webster, Int J Thermophys 35, 574-595 (2014).
16. T. G. Kollie, J. L. Horton, K. R. Carr, M. B. Herskovitz and C. A. Mossman, Rev. Sci. Instrum. 46,
1447-1461 (1975).
17. A. W. Fenton, Temperature, its measurement and control in science and industry, vol. 4, part 3, ed. by
H. H. Plumb, (Instrument Society of America, 1972), pp. 1973-1990.
18. N. A. Burley, R. M. Hess, C. F. Howie and J. A. Coleman, Temperature, its measurement and control
in science and industry, vol. 5, part 2, ed. by J. F. Schooley, (Instrument Society of America, 1982), pp. 1159-
1166.
19. D. D. Pollock, Thermocouples theory and properties, (CRC press Boca Raton, 1991).
20. G. Machin, K. Anhalt, F. Edler, J. Pearce, M. Sadli, R. Strnad and E. Vuelban, 16th International
Congress of Metrology, (EDP Sciences, 2013), .
21. R. E. Bentley, Measurement 23, 35-46 (1998).
22. K. D. Hill, Metrologia 31, 51-58 (2002).
23. Y. G. Kim, K. S. Gam and J. H. Lee, Meas. Sci. Technol. 8, 317-321 (1997).
24. R. E. Bentley, Meas. Sci. Technol. 12, 1250-1260 (2001).
25. O. Ongrai, J. Pearce, G. Machin and S. Sweeney, International Journal of Thermophysics 31, 1506-
1516 (2010).
26. J. Tamba, K. Yamazawa, S. Masuyama, H. Ogura and M. Izuchi, Int J Thermophys 32, 2436-2451
(2011).
27. G. W. Burns, M. G. Scroger, G. F. Strouse, M. C. Croarkin and W. F. Guthrie, Temperature-
electromotive Force Reference Functions And Tables For The Letter-designated Thermocouple Types Based On
The ITS-90, (NIST, 1993).
28. R. E. Bentley, Meas. Sci. Technol. 11, 538-546 (2000).
29. P. Kinzie, Thermocouple Temperature Measurements, 1st edn. (John Wiley & Sons, New York, 1973).
30. F. Edler and P. Ederer, Temperature, its measurement and control in science and industry, vol. 8, part
1, ed. by C. W. Meyer, (AIP, 2013), pp. 532-537.
31. M. Rubel, Materials Science and Engineering 9-12 (1987).
32. J. C. Chaston, Platinum Met. Rev 9, 51-56 (1965).
33. J. C. Chaston, Platinum Met. Rev 19, 135-140 (1975).
34. T. Li, E. A. Marquis, P. A. J. Bagot, S. C. Tsang and G. D. W. Smith, Catalysis Today 175, 552-557
(2011).
35. J. C. Chaston, Platinum Met. Rev 8, 50-54 (1964).
36. J. C. Chaston, Platinum Met. Rev 10, 91-93 (1966).
37. R. E. Bentley, Int J Thermophys 6, 83-99 (1985).
38. F. Edler and A. C. Baratto, Metrologia 42, 201-207 (2005).
39. J. V. Pearce, H. Ogura, M. Izuchi and G. Machin, Metrologia 46, 743-749 (2009).
40. Y. Kim, I. Yang and K. Gam, Instrumentation Science and Technology 36, 257-266 (2008).
41. E. H. McLaren and E. G. Murdock, Temperature, its measurement and control in science and industry,
vol. 4, part 3, ed. by H. H. Plumb, (Instrument Society of America, 1972), pp. 1543-1560.
42. E. H. McLaren and E. G. Murdock, Temperature, its measurement and control in science and industry,
vol. 5, part 2, ed. by J. F. Schooley, (Instrument Society of America, 1982), pp. 953-975.
43. R. E. Bentley, Meas. Sci. Technol. 12, 627-634 (2001).
